science 2009-01-30

CONTENTS Volume 323 Issue 5914

EDITORIAL

561 U.S.-China S&T at 30Norman P. Neureiter and Tom C. Wang

NEWS OF THE WEEK

568 Celebration and Concern Over U.S. Trial of Embryonic Stem Cells

569 Early Start for Human Art? Ochre May Revise Timeline

570 European Science Not As Intense As Hoped

570 Iraq Museum May Reopen Amid Controversy

571 A Stimulus for Science

572 Report Calls for Fresh Look at What Happens Outside School

572 Fingerprints Enhance the Sense of Touch>> Science Express Report by J. Scheibert et al.

573 How Sorghum Withstands Heat and Drought

NEWS FOCUS

574 Seeds of Discontent>> Science Podcast

576 DOD Funds New Views on Conflict With ItsFirst Minerva Grants

578 Beset by Delays, U.S. Astronomers Ponder a Better To Do' ListPriorities Nearer to Home in Need of Better

Cost Estimates

580 HPV Casts a Wider Shadow

LETTERS

582 Systems Politics and Political Systems H. Ashrafian

Scientists Not Immune to PartisanshipD. H. Guston et al.

Law and Science Not Mutually Exclusive D. D. deRosier

Credit for Coauthors N. T. Hagen

Response C. H. Sekercioglu

page 574

page 584

www.sciencemag.org SCIENCE VOL 323 30 JANUARY 2009 553

COVER

Desert locusts, Schistocerca gregaria, are normally solitary and

avoid each other (green individual, lower right). Forced proximity

between them leads to the release of the neurotransmitter

serotonin, which changes their behavior to mutual attraction

and initiates a cascade of events leading to swarm formation,

including a change in appearance (black and yellow group).

See page 627.

Image: Tom Fayle, Swidbert Ott, and Stephen Rogers/University

of Cambridge

DEPARTMENTS

559 This Week in Science

562 Editors' Choice564 Science Staff565 Random Samples567 Newsmakers596 AAAS News & Notes655 New Products656 Science Careers

583 CORRECTIONS AND CLARIFICATIONS

583 TECHNICAL COMMENT ABSTRACTS

BOOKS ET AL.

584 Science on the AirM. C. LaFollette, reviewed by A. J. Wolfe

585 HumanM. S. Gazzaniga, reviewed by R. Adolphs

EDUCATION FORUM

586 Learning and Scientific ReasoningL. Bao et al.

PERSPECTIVES

588 The Force Is with UsM. A. Schwartz

>> Reports pp. 638 and 642

589 Transforming GrapheneA. Savchenko

>> Report p. 610

590 An Abnormal Normal StateG. S. Boebinger

>> Research Article p. 603

592 Pores in PlaceF. D. Sack and J.-G. Chen

>> Report p. 649

593 Unfolding the Secrets of CalmodulinR. B. Best and G. Hummer

>> Report p. 633

594 The Key to Pandora' s BoxP. A. Stevenson

>> Report p. 627

REVIEW

602 Sudden Death of Entanglement T. Yu and J. H. Eberly

BREVIA

602 Facile Synthesis of AsP3

B. M. Cossairt et al.

The question of the stability of solid AsP3, a

simple inorganic molecule, has been settled

by its synthesis.

CONTENTS continued >>

Published by AAAS

CONTENTS

RESEARCH ARTICLE

603 Anomalous Criticality in the Electrical Resistivity of La2± xSrxCuO4

R. A. Cooper et al.

High magnetic fields can strip away the

superconducting regime of a cuprate

superconductor, revealing the presence

of an enigmatic quantum critical point.

>> Perspective p. 590

REPORTS

607 Revealing the Maximum Strength in Nanotwinned CopperL. Lu et al.

Studies of nanocrystalline copper reveal

changes in deformation mechanisms with grain

size and the role played by twin boundaries.

610 Control of Graphene' s Properties byReversible Hydrogenation: Evidence for GraphaneD. C. Elias et al.

Graphene can be transformed from a conductor

to an insulator by exposure to hydrogen atoms

and reversed by a thermal treatment.


614 Dynamical Quorum Sensing and Synchronization in Large Populations of Chemical OscillatorsA. F. Taylor et al.

Communication between chemical oscillators

in solution can mimic that of large populations

of single-celled organisms.

617 Single Nanocrystals of Platinum Preparedby Partial Dissolution of Au-Pt Nanoalloys M. Schrinner et al.

Gold-platinum nanoparticles, held in polymer

networks on latex beads, are converted into

platinum nanocrystals.

620 Cascadia Tremor Located Near Plate Interface Constrained by S Minus P WaveTimes M. La Rocca et al.

A series of microearthquakes near Puget

Sound originate near or on the subduction

zone fault from a recurrent source.

623 Divergent Evolution of Duplicate GenesLeads to Genetic Incompatibilities WithinA. thaliana

D. Bikard et al.

The divergent evolution of a duplicated gene

results in genetic incompatibilities between

strains of the plant Arabidopsis.

627 Serotonin Mediates Behavioral Gregarization Underlying Swarm Formation in Desert Locusts M. L. Anstey et al.

Serotonin induces the phenotypic switch

from solitary to gregarious behavior in

desert locusts.

>> Perspective p. 594; Science Podcast

630 Survival from Hypoxia in C. elegans

by Inactivation of Aminoacyl-tRNASynthetases L. L. Anderson et al.

Reduced activity of aminoacyl± transfer RNA

synthetases allows for survival from hypoxic

insult in the nematode C. elegans.

633 Ligand-Dependent Equilibrium Fluctuationsof Single Calmodulin MoleculesJ. P. Junker et al.

Single-molecule force spectroscopy reveals the

equilibrium dynamics of calmodulin folding

and how it is modulated by peptide ligands.


638 Stretching Single Talin Rod Molecules Activates Vinculin Binding A. del Rio et al.

Force-induced stretching of proteins can

expose previous cryptic binding sites and

promote binding to their ligands.


642 Mechanically Activated Integrin SwitchControls α5β1 Function J. C. Friedland et al.

Myosin contraction and extracellular matrix

stiffness drive a tension-induced cell-surface

integrin switch that regulates cell signaling.


644 A Human Telomerase Holoenzyme ProteinRequired for Cajal Body Localization andTelomere Synthesis A. S. Venteicher et al.

Telomerase Cajal body protein 1 (TCAB1)

is the fourth discovered subunit of the

chromosome end-capping enzyme telomerase.

649 PAN1: A Receptor-Like Protein That Promotes Polarization of an AsymmetricCell Division in Maize H. N. Cartwright et al.

Asymmetric cell division in plants is regulated

by a receptor-like kinase, implicating a

signaling cascade in cell polarization.


651 Calcineurin/NFAT Signaling Is Required for Neuregulin-Regulated Schwann CellDifferentiation S.-C. Kao et al.

The cell signaling components

calcineurin/NFATc and Sox10 control

Schwann cell myelination.

CONTENTS continued >>


page 589 & 610

page 590 & 603

page 614

Published by AAAS


CONTENTS

SCIENCEXPRESS

www.sciencexpress.org

A Role for RNAi in the Selective Correction of DNA Methylation Defects F. K. Teixeira et al.

An RNA interference± dependent DNA methylation rescue system helps to preserve a subset of DNAmethylation marks in Arabidopsis.

10.1126/science.1165313

The Role of Fingerprints in the Coding of TactileInformation Probed with a Biomimetic SensorJ. Scheibert et al.

Fingertip ridges improve the tactile perception of fine features.>> News story p. 57210.1126/science.1166467

Zircon Dating of Oceanic Crustal AccretionC. J. Lissenberg et al.

Zircon dates from the slow-spreading mid-AtlanticRidge show that magmatic intrusions formed newoceanic crust regularly and evenly, thereby providingcooling times.10.1126/science.1167330

Inducing a Magnetic Monopole with TopologicalSurface StatesX.-L. Qi et al.

A magnetic monopole is theoretically predicted to beinduced at the surface of a topological insulator.10.1126/science.1167747

TECHNICALCOMMENTS

Comment on ª Arsenic (III) Fuels Anoxygenic Photosynthesis in Hot Spring Biofilms from Mono Lake, CaliforniaºB. Schoepp-Cothenet et al.

full text at www.sciencemag.org/cgi/content/full/323/5914/583c

Response to Comment on ª Arsenic(III) FuelsAnoxygenic Photosynthesis in Hot Spring Biofilmsfrom Mono Lake, Californiaº R. S. Oremland et al.

full text at www.sciencemag.org/cgi/content/full/323/5914/583d

SCIENCENOW

www.sciencenow.org

Highlights From Our Daily News Coverage

A Millennia-Long Greenhouse DisasterThe physics of global warming would make for

essentially irreversible global damage.

No Feces for This Species Dung beetle gives up excrement for the life of a

hunter.

Death March of the Penguins? Shrinking sea ice may decimate penguin population

that starred in 2005 movie.

SCIENCESIGNALING

www.sciencesignaling.org

The Signal Transduction Knowledge Environment

EDITORIAL GUIDE: Freedom of MaterialsM. B. Yaffe

Intellectual property concerns bring into conflict datasharing and accessibility of materials.

RESEARCH ARTICLE: Direct Response to Notch ActivationÐ Signaling Crosstalk and Incoherent LogicA. Krejcí et al.

Identification of the direct target genes of Notchreveals complex input into multiple signaling pathways that goes beyond coordination of transcriptional networks.

RESEARCH ARTICLE: Differential Requirement ofmTOR in Postmitotic Tissues and TumorigenesisC. Nardella et al.

Conditional inactivation of mTor has little effect in theadult mouse prostate, but it suppresses tumor initiation and progression.

RESEARCH ARTICLE: Characterization of the Intrinsic and TSC2-GAP± Regulated GTPase Activity of Rheb by Real-Time NMR C. B. Marshall et al.

An NMR-based assay enables real-time analysis of theGTPase activity of Rheb and of the effect of itsGTPase-activating protein TSC2.

RESEARCH ARTICLE: EGFR Signals to mTORThrough PKC and Independently of Akt in Glioma Q.-W. Fan et al.

Akt is dispensable for signaling between EGFR andmTOR in glioma cells, whereas PKC is critical.

PERSPECTIVE: Grab Your Partner with Both HandsÐ Cytoskeletal Remodeling by Arp2/3 Signaling S. H. Soderling

Dimerization provides an additional layer of regulation to the activity of a family of actin-nucleating factors.

PODCASTJ. F. Foley and A. M. VanHook

Leptin secreted by adipose tissue decreases insulinproduction through both direct and indirect mechanisms.

NETWATCH: iBioSeminars

View seminars from leading cell biology researchers;in Web Broadcasts.

NETWATCH: Foldit

A computer game teaches students about the energet-ics and dynamics of protein folding; in Educator Sites.

SCIENCEONLINE

SCIENCE (ISSN 0036-8075) is published weekly on Friday, except the last week

in December, by the American Association for the Advancement of Science, 1200

New York Avenue, NW, Washington, DC 20005. Periodicals Mail postage (publica-tion No. 484460) paid at Washington, DC, and additional mailing offices. Copyright ©2009 by the American Association for the Advancement of Science. The title SCIENCE isa registered trademark of the AAAS. Domestic individual membership and subscription(51 issues): $146 ($74 allocated to subscription). Domestic institutional subscription(51 issues): $835; Foreign postage extra: Mexico, Caribbean (surface mail) $55; othercountries (air assist delivery) $85. First class, airmail, student, and emeritus rates onrequest. Canadian rates with GST available upon request, GST #1254 88122. Publica-tions Mail Agreement Number 1069624. Printed in the U.S.A.

Change of address: Allow 4 weeks, giving old and new addresses and 8-digit accountnumber. Postmaster: Send change of address to AAAS, P.O. Box 96178, Washington, DC20090± 6178. Single-copy sales: $10.00 current issue, $15.00 back issue prepaidincludes surface postage; bulk rates on request. Authorization to photocopy materialfor internal or personal use under circumstances not falling within the fair use provisionsof the Copyright Act is granted by AAAS to libraries and other users registered with theCopyright Clearance Center (CCC) Transactional Reporting Service, provided that $20.00per article is paid directly to CCC, 222 Rosewood Drive, Danvers, MA 01923. The identifi-cation code for Science is 0036-8075. Science is indexed in the Reader' s Guide to Period-ical Literature and in several specialized indexes.

SCIENCESIGNALING

Specifying muscle.

SCIENCECAREERS

www.sciencecareers.org/career_magazine

Free Career Resources for Scientists

Women M.D.-Ph.D.' s: Life in the TrenchesK. Hede

Physician-scientists urge a focus on solutions, notproblems, to advance women in academic medicine.

Perspective: Ensuring Retention of Women inPhysician-Scientist TrainingJ. M. Pauff and M. C. Richards

Why do more women than men drop out of M.D.-Ph.D.programs?

Feet on the Ground, Head in the SkyE. Pain

Spain' s Juan A. Añel has quickly established himself in the field of atmospheric physics.

SCIENCEPODCAST

www.sciencemag.org/multimedia/podcast

Free Weekly Show

Download the 30 January Science Podcast to hearabout the influence of serotonin on swarming behavior, predicting protests in China, your letters to Science, and more.

ORIGINSBLOG

blogs.sciencemag.org/origins

A History of Beginnings

SCIENCEINSIDER

blogs.sciencemag.org/scienceinsiderScience Policy News and Analysis

CR

ED

IT: (S

CIE

NC

E S

IGN

ALIN

G) SA

RA

H J

. B

RA

Y/U

NIV

ER

SIT

Y O

F C

AM

BR

IDG

E

Published by AAAS


CR

ED

ITS: (L

EF

T T

OP

AN

D B

OT

TO

M) A

AA

S; (R

IGH

T) A

SSO

CIA

TE

D P

RE

SS

EDITORIAL

U.S.-China S&T at 30IN 1972, THE SHANGHAI COMMUNIQUE OF PRESIDENT NIXON AND CHINESE PREMIER ZHOU ENLAIended 23 years of isolation between the United States and China. Tucked into a single sentencewas a brief reference to cooperation in science and technology (S&T). Visits by scientists andscholars then gradually increased, guided on the U.S. side by the nongovernmental NationalAcademy of Sciences. Six years later, and presaging formal diplomatic relations in 1979, camethe breakthrough science diplomacy mission of the President' s Science Advisor Frank Press,accompanied by representatives of nearly every technical federal agency. That trip laid thegroundwork for the formal Agreement on Cooperation in S&T, signed exactly 30 years ago thisweek by President Carter and the Chinese leader Deng Xiaoping.

The new U.S. president firmly believes in the power of science to address domestic andglobal challenges, making this 30th anniversary year an excellent time to assess the U.S.-ChinaS&T relationship and ensure that it is on the right track for the future.These ties have come a long way. Over a million Chinese students havestudied at U.S. universities, some two-thirds of them in S&T. Many havestayed, and today about 8% of science and engineering Ph.D.s in theUnited States were born in China.* In recent years, almost 40% ofChinese science and engineering articles in international journals havehad U.S. coauthors, and almost 8% of U.S. international papers have hadChinese coauthors. In 2004, $622 million of China' s total R&D wasperformed by U.S.-owned companies and affiliates in China. With theexception of one areaÐ space explorationÐ research cooperationbetween universities as well as government labs is broad and diverse.Furthermore, China has declared that its path to the future will be drivenlargely by S&T, and it is making the needed commitments to education,facilities, and research.²

However, the overall relationship with China is not without tensions. The massive U.S. tradedeficit is equated with the loss of U.S. jobs, the alleged manipulation of currency exchange rates,and concerns over U.S. competitiveness. China' s economic boom and soaring demand forresources (now easing) could be a source of future conflict. Taiwan remains an area of potentialconfrontation. And China' s increasing investment in its military has raised concerns, being usedin the United States to justify more investment in new weapons and defense systems. China isalso in a special category for U.S. visas and export controls, due to a controversial 1999 U.S. con-gressional report charging China with espionage in nuclear and missile technologies. Despitebeing America' s principal creditor, the Chinese have concerns about U.S. long-term intentionstoward Asia and their country. There is a level of mistrust on both sides.

Science provides a common language that can help bridge cultures and serve to lessen mistrustand increase transparency. The Obama administration should work to raise S&T cooperation withChina to a new level of partnership. The existing Joint S&T Commission, chaired by the Presi-dent' s Science Advisor and China' s S&T minister, should meet annually (instead of every 2years), and the 2009 meeting should lay out a 10-year plan of cooperative research focused onglobal challenges faced by both countries, including climate change, energy, food, health, andsecurity. On the U.S. side, new funds will be needed to complement the agencies' domestic pro-grams in these areas. The United States must also ensure that all qualified Chinese (and otherforeign) scientists can obtain visas on a timely basis, and that our export controls protect nationalsecurity but do not prevent U.S. corporations from fully participating in the global civil economy.³The United States and China must be true partners in seeking technical solutions that will supporta global population of some nine billion people by 2050. Such cooperation can also mitigateinevitable tensions in the overall relationshipÐ both splendid goals for the next 30 years.

± Norman P. Neureiter and Tom C. Wang

10.1126/science.1170938

*National Science Board Science and Enginering Indicators 2008, www.nsf.gov/statistics/seind08/.

² Wen Jiabao, Science 322, 649 (2008).

³ See Beyond ª Fortress Americaº (National Academies Press, Washington, DC, 2009); available at www.nap.edu/catalog.php?record_id=12567.

Tom C. Wang is director

for international cooper-

ation at the AAAS,

Washington, DC.

Norman P. Neureiter is

director of the Center

for Science, Technology,

and Security Policy at

the American Associa-

tion for the Advance-

ment of Science (AAAS),

Washington, DC.

Published by AAAS

...

......

http://www.sciencemag.org

30 JANUARY 2009 VOL 323 SCIENCE www.sciencemag.org564

John I. Brauman, Chair, Stanford Univ.Richard Losick, Harvard Univ.Robert May, Univ. of OxfordMarcia McNutt, Monterey Bay Aquarium Research Inst.Linda Partridge, Univ. College London Vera C. Rubin, Carnegie Institution Christopher R. Somerville, Univ. of California, Berkeley

Joanna Aizenberg, Harvard Univ.David Altshuler, Broad InstituteArturo Alvarez-Buylla, Univ. of California, San FranciscoRichard Amasino, Univ. of Wisconsin, MadisonAngelika Amon, MITMeinrat O. Andreae, Max Planck Inst., MainzKristi S. Anseth, Univ. of ColoradoJohn A. Bargh, Yale Univ.Cornelia I. Bargmann, Rockefeller Univ.Ben Barres, Stanford Medical SchoolMarisa Bartolomei, Univ. of Penn. School of Med.Facundo Batista, London Research Inst.Ray H. Baughman, Univ. of Texas, Dallas Stephen J. Benkovic, Penn State Univ. Ton Bisseling, Wageningen Univ. Mina Bissell, Lawrence Berkeley National LabPeer Bork, EMBLRobert W. Boyd, Univ. of Rochester Paul M. Brakefield, Leiden Univ. Dennis Bray, Univ. of Cambridge Stephen Buratowski, Harvard Medical SchoolJoseph A. Burns, Cornell Univ. William P. Butz, Population Reference BureauMats Carlsson, Univ. of Oslo Peter Carmeliet, Univ. of Leuven, VIBMildred Cho, Stanford Univ. David Clapham, Children' s Hospital, Boston David Clary, Oxford University J. M. Claverie, CNRS, Marseille Jonathan D. Cohen, Princeton Univ.

Andrew Cossins, Univ. of Liverpool Robert H. Crabtree, Yale Univ.Wolfgang Cramer, Potsdam Inst. for Climate Impact ResearchF. Fleming Crim, Univ. of Wisconsin William Cumberland, Univ. of California, Los Angeles Jeff L. Dangl, Univ. of North CarolinaStanislas Dehaene, Collège de FranceEdward DeLong, MITEmmanouil T. Dermitzakis, Wellcome Trust Sanger Inst.Robert Desimone, MITDennis Discher, Univ. of Pennsylvania Scott C. Doney, Woods Hole Oceanographic Inst.W. Ford Doolittle, Dalhousie Univ.Jennifer A. Doudna, Univ. of California, BerkeleyJulian Downward, Cancer Research UK Denis Duboule, Univ. of Geneva/EPFL Lausanne Christopher Dye, WHO Gerhard Ertl, Fritz-Haber-Institut, Berlin Mark Estelle, Indiana Univ.Barry Everitt, Univ. of Cambridge Paul G. Falkowski, Rutgers Univ. Ernst Fehr, Univ. of ZurichTom Fenchel, Univ. of Copenhagen Alain Fischer, INSERM Scott E. Fraser, Cal TechChris D. Frith, Univ. College London Wulfram Gerstner, EPFL LausanneCharles Godfray, Univ. of OxfordDiane Griffin, Johns Hopkins Bloomberg School of

Public HealthChristian Haass, Ludwig Maximilians Univ.Niels Hansen, Technical Univ. of DenmarkDennis L. Hartmann, Univ. of Washington Chris Hawkesworth, Univ. of BristolMartin Heimann, Max Planck Inst., Jena James A. Hendler, Rensselaer Polytechnic Inst.Ray Hilborn, Univ. of WashingtonOve Hoegh-Guldberg, Univ. of QueenslandBrigid L. M. Hogan, Duke Univ. Medical CenterRonald R. Hoy, Cornell Univ.Olli Ikkala, Helsinki Univ. of TechnologyMeyer B. Jackson, Univ. of Wisconsin Med. School Stephen Jackson, Univ. of Cambridge Steven Jacobsen, Univ. of California, Los Angeles

Peter Jonas, Universität FreiburgBarbara B. Kahn, Harvard Medical SchoolDaniel Kahne, Harvard Univ.Gerard Karsenty, Columbia Univ. College of P&SBernhard Keimer, Max Planck Inst., Stuttgart Elizabeth A. Kellog, Univ. of Missouri, St. LouisAlan B. Krueger, Princeton Univ. Lee Kump, Penn State Univ.Mitchell A. Lazar, Univ. of PennsylvaniaVirginia Lee, Univ. of PennsylvaniaOlle Lindvall, Univ. Hospital, LundMarcia C. Linn, Univ. of California, BerkeleyJohn Lis, Cornell Univ.Richard Losick, Harvard Univ.Ke Lu, Chinese Acad. of SciencesAndrew P. MacKenzie, Univ. of St Andrews Raul Madariaga, École Normale Supérieure, Paris Anne Magurran, Univ. of St AndrewsCharles Marshall, Harvard Univ.Virginia Miller, Washington Univ.Yasushi Miyashita, Univ. of TokyoRichard Morris, Univ. of EdinburghEdvard Moser, Norwegian Univ. of Science and TechnologyNaoto Nagaosa, Univ. of Tokyo James Nelson, Stanford Univ. School of Med. Timothy W. Nilsen, Case Western Reserve Univ. Roeland Nolte, Univ. of Nijmegen Helga Nowotny, European Research Advisory BoardEric N. Olson, Univ. of Texas, SWStuart H. Orkin, Dana-Farber Cancer Inst.Erin O' Shea, Harvard Univ. Elinor Ostrom, Indiana Univ.Jonathan T. Overpeck, Univ. of ArizonaJohn Pendry, Imperial CollegeSimon Phillpot, Univ. of Florida Philippe Poulin, CNRS Mary Power, Univ. of California, Berkeley Molly Przeworski, Univ. of ChicagoColin Renfrew, Univ. of CambridgeTrevor Robbins, Univ. of Cambridge Barbara A. Romanowicz, Univ. of California, BerkeleyEdward M. Rubin, Lawrence Berkeley National LabShimon Sakaguchi, Kyoto Univ.Jürgen Sandkühler, Medical Univ. of Vienna

David W. Schindler, Univ. of AlbertaGeorg Schulz, Albert-Ludwigs-Universität Paul Schulze-Lefert, Max Planck Inst., Cologne Christine Seidman, Harvard Medical SchoolTerrence J. Sejnowski, The Salk Institute David Sibley, Washington Univ.Joseph Silk, Univ. of Oxford Montgomery Slatkin, Univ. of California, BerkeleyDavor Solter, Inst. of Medical Biology, SingaporeJoan Steitz, Yale Univ.Elsbeth Stern, ETH ZürichJerome Strauss, Virginia Commonwealth Univ.Jurg Tschopp, Univ. of Lausanne Derek van der Kooy, Univ. of Toronto Bert Vogelstein, Johns Hopkins Univ.Ulrich H. von Andrian, Harvard Medical SchoolBruce D. Walker, Harvard Medical SchoolChristopher A. Walsh, Harvard Medical School Graham Warren, Yale Univ. School of Med. Colin Watts, Univ. of DundeeDetlef Weigel, Max Planck Inst., TübingenJonathan Weissman, Univ. of California, San FranciscoEllen D. Williams, Univ. of Maryland Ian A. Wilson, The Scripps Res. Inst. Jerry Workman, Stowers Inst. for Medical ResearchXiaoling Sunney Xie, Harvard Univ.John R. Yates III, The Scripps Res. Inst. Jan Zaanen, Leiden Univ.Huda Zoghbi, Baylor College of Medicine Maria Zuber, MIT

John Aldrich, Duke Univ.David Bloom, Harvard Univ.Angela Creager, Princeton Univ.Richard Shweder, Univ. of ChicagoEd Wasserman, DuPontLewis Wolpert, Univ. College London

EDITOR-IN-CHIEF Bruce AlbertsEXECUTIVE EDITOR Monica M. Bradford

DEPUTY EDITORS NEWS EDITOR

R. Brooks Hanson, Barbara R. Jasny, Colin NormanKatrina L. Kelner

EDITORIAL SUPERVISORY SENIOR EDITOR Phillip D. Szuromi; SENIOR EDITOR/

PERSPECTIVES Lisa D. Chong; SENIOR EDITORS Gilbert J. Chin, Pamela J.Hines, Paula A. Kiberstis (Boston), Marc S. Lavine (Toronto), Beverly A.Purnell, L. Bryan Ray, Guy Riddihough, H. Jesse Smith, Valda Vinson;ASSOCIATE EDITORS Kristen L. Mueller, Jake S. Yeston, Laura M. Zahn;ONLINE EDITOR Stewart Wills; ASSOCIATE ONLINE EDITORS Robert Frederick,Tara S. Marathe; WEB CONTENT DEVELOPER Martyn Green; BOOK REVIEW

EDITOR Sherman J. Suter; ASSOCIATE LETTERS EDITOR Jennifer Sills; EDITORIAL MANAGER Cara Tate; SENIOR COPY EDITORS Jeffrey E. Cook, CynthiaHowe, Harry Jach, Barbara P. Ordway, Trista Wagoner; COPY EDITORS

Chris Filiatreau, Lauren Kmec; EDITORIAL COORDINATORS Carolyn Kyle,Beverly Shields; PUBLICATIONS ASSISTANTS Ramatoulaye Diop, Joi S.Granger, Jeffrey Hearn, Lisa Johnson, Scott Miller, Jerry Richardson,Jennifer A. Seibert, Brian White, Anita Wynn; EDITORIAL ASSISTANTS

Carlos L. Durham, Emily Guise, Patricia M. Moore; EXECUTIVE ASSISTANT

Sylvia S. Kihara; ADMINISTRATIVE SUPPORT Maryrose MadridNEWS DEPUTY NEWS EDITORS Robert Coontz, Eliot Marshall, JeffreyMervis, Leslie Roberts; CONTRIBUTING EDITORS Elizabeth Culotta, PollyShulman; NEWS WRITERS Yudhijit Bhattacharjee, Adrian Cho, JenniferCouzin, David Grimm, Constance Holden, Jocelyn Kaiser, Richard A.Kerr, Eli Kintisch, Andrew Lawler (New England), Greg Miller,Elizabeth Pennisi, Robert F. Service (Pacific NW), Erik Stokstad; INTERN

Jackie D. Grom; CONTRIBUTING CORRESPONDENTS Jon Cohen (San Diego,CA), Daniel Ferber, Ann Gibbons, Robert Koenig, Mitch Leslie,Charles C. Mann, Virginia Morell, Evelyn Strauss, Gary Taubes; COPY

EDITORS Linda B. Felaco, Melvin Gatling, Melissa Raimondi;ADMINISTRATIVE SUPPORT Scherraine Mack, Fannie Groom; BUREAUS NewEngland: 207-549-7755, San Diego, CA: 760-942-3252, FAX 760-942-4979, Pacific Northwest: 503-963-1940PRODUCTION DIRECTOR James Landry; SENIOR MANAGER Wendy K. Shank;ASSISTANT MANAGER Rebecca Doshi; SENIOR SPECIALISTS Steve Forrester,Chris Redwood; SPECIALIST Anthony Rosen; PREFLIGHT DIRECTOR David M.Tompkins; MANAGER Marcus Spiegler

ART DIRECTOR Yael Kats; ASSOCIATE ART DIRECTOR Laura Creveling;ILLUSTRATORS Chris Bickel, Katharine Sutliff; SENIOR ART ASSOCIATES HollyBishop, Preston Huey, Nayomi Kevitiyagala; ART ASSOCIATE JessicaNewfield; PHOTO EDITOR Leslie Blizard

SCIENCE INTERNATIONAL

EUROPE ([email protected]) EDITORIAL: INTERNATIONAL MANAGING

EDITOR Andrew M. Sugden; SENIOR EDITOR/PERSPECTIVES Julia Fahrenkamp-Uppenbrink; SENIOR EDITORS Caroline Ash, Stella M. Hurtley, Ian S.Osborne, Peter Stern; ASSOCIATE EDITOR Maria Cruz; EDITORIAL SUPPORT

Deborah Dennison, Rachel Roberts, Alice Whaley; ADMINISTRATIVE

SUPPORT John Cannell, Janet Clements; NEWS: EUROPE NEWS EDITOR JohnTravis; DEPUTY NEWS EDITOR Daniel Clery; CONTRIBUTING CORRESPONDENTS

Michael Balter (Paris), John Bohannon (Vienna), Martin Enserink(Amsterdam and Paris), Gretchen Vogel (Berlin); INTERN Sara Coelho

ASIA Japan Office: Asca Corporation, Eiko Ishioka, Fusako Tamura, 1-8-13, Hirano-cho, Chuo-ku, Osaka-shi, Osaka, 541-0046 Japan; +81(0) 6 6202 6272, FAX +81 (0) 6 6202 6271; [email protected]; ASIA

NEWS EDITOR Richard Stone (Beijing: [email protected]); CONTRIBUTING

CORRESPONDENTS Dennis Normile (Japan: +81 (0) 3 3391 0630, FAX +81 (0) 3 5936 3531; [email protected]); Hao Xin (China: +86 (0)10 6307 4439 or 6307 3676, FAX +86 (0) 10 6307 4358; [email protected]); Pallava Bagla (South Asia: +91 (0) 11 22712896; [email protected])

www.sciencemag.org EXECUTIVE PUBLISHER Alan I. Leshner

PUBLISHER Beth Rosner

FULFILLMENT SYSTEMS AND OPERATIONS ([email protected]); DIRECTOR

Waylon Butler; SENIOR SYSTEMS ANALYST Jonny Blaker; CUSTOMER SERVICE

SUPERVISOR Pat Butler; SPECIALISTS Latoya Casteel, LaVonda Crawford,Vicki Linton, April Marshall; DATA ENTRY SUPERVISOR Cynthia Johnson; SPECIALISTS Eintou Bowden, Tarrika Hill, William Jones

BUSINESS OPERATIONS AND ADMINISTRATION DIRECTOR Deborah Rivera-Wienhold; ASSISTANT DIRECTOR, BUSINESS OPERATIONS Randy Yi; MANAGER,

BUSINESS ANALYSIS Michael LoBue; MANAGER, BUSINESS OPERATIONS JessicaTierney; FINANCIAL ANALYSTS Priti Pamnani, Celeste Troxler; RIGHTS AND

PERMISSIONS: ADMINISTRATOR Emilie David; ASSOCIATE Elizabeth Sandler;MARKETING DIRECTOR Ian King; MARKETING MANAGER Allison Pritchard; MARKETING ASSOCIATES Aimee Aponte, Alison Chandler, Mary EllenCrowley, Julianne Wielga, Wendy Wise; INTERNATIONAL MARKETING

MANAGER Wendy Sturley; MARKETING EXECUTIVE Jennifer Reeves; MARKETING/MEMBER SERVICES EXECUTIVE Linda Rusk; DIRECTOR, SITE LICENSING

Tom Ryan; DIRECTOR, CORPORATE RELATIONS Eileen Bernadette Moran; PUBLISHER RELATIONS, eRESOURCES SPECIALIST Kiki Forsythe; SENIOR PUBLISHER

RELATIONS SPECIALIST Catherine Holland; PUBLISHER RELATIONS, EAST COAST

Phillip Smith; PUBLISHER RELATIONS, WEST COAST Philip Tsolakidis; FULFILLMENT SUPERVISOR Iquo Edim; FULFILLMENT COORDINATOR LauraClemens; ELECTRONIC MEDIA: MANAGER Lizabeth Harman; PROJECT MANAGER

Trista Snyder; ASSISTANT MANAGER Lisa Stanford; SENIOR PRODUCTION

SPECIALISTS Christopher Coleman, Walter Jones; PRODUCTION SPECIALISTS

Nichele Johnston, Kimberly Oster

ADVERTISING DIRECTOR, WORLDWIDE AD SALES Bill Moran

PRODUCT ([email protected]); MIDWEST/WEST COAST/W. CANADA

Rick Bongiovanni: 330-405-7080, FAX 330-405-7081; EAST COAST/

E. CANADA Laurie Faraday: 508-747-9395, FAX 617-507-8189;UK/EUROPE/ASIA Roger Goncalves: TEL/FAX +41 43 243 1358; JAPAN

Mashy Yoshikawa: +81 (0) 3 3235 5961, FAX +81 (0) 3 3235 5852;SENIOR TRAFFIC ASSOCIATE Deiandra Simms

COMMERCIAL EDITOR Sean Sanders: 202-326-6430

PROJECT DIRECTOR, OUTREACH Brianna Blaser

CLASSIFIED ([email protected]); INSIDE SALES MANAGER:

MIDWEST/CANADA Daryl Anderson: 202-326-6543; INSIDE SALES

REPRESENTATIVE Karen Foote: 202-326-6740; KEYACCOUNT MANAGER JoribahAble; NORTHEAST Alexis Fleming: 202-326-6578; SOUTHEAST Tina Burks:202-326-6577; WEST Nicholas Hintibidze: 202-326-6533; SALES

COORDINATORS Rohan Edmonson, Shirley Young; INTERNATIONAL: SALES

MANAGER Tracy Holmes: +44 (0) 1223 326525, FAX +44 (0) 1223326532; SALES Susanne Kharraz, Dan Pennington, Alex Palmer; SALES

ASSISTANT Louise Moore; JAPAN Mashy Yoshikawa: +81 (0) 3 3235 5961,FAX +81 (0) 3 3235 5852; ADVERTISING PRODUCTION OPERATIONS MANAGER

Deborah Tompkins; SENIOR PRODUCTION SPECIALIST/GRAPHIC DESIGNER AmyHardcastle; SENIOR PRODUCTION SPECIALIST Robert Buck; SENIOR TRAFFIC

ASSOCIATE Christine Hall; PUBLICATIONS ASSISTANT Mary Lagnaoui

AAAS BOARD OF DIRECTORS RETIRING PRESIDENT, CHAIR David Baltimore;PRESIDENT James J. McCarthy; PRESIDENT-ELECT Peter C. Agre; TREASURER

David E. Shaw; CHIEF EXECUTIVE OFFICER Alan I. Leshner; BOARD Lynn W.Enquist, Susan M. Fitzpatrick, Alice Gast, Linda P. B. Katehi, NancyKnowlton, Cherry A. Murray, Thomas D. Pollard, Thomas A. Woolsey

INFORMATION FOR AUTHORS

See pages 634 and 635 of the 1 February 2008 issue or accesswww.sciencemag.org/about/authors

1200 New York Avenue, NWWashington, DC 20005

Editorial: 202-326-6550, FAX 202-289-7562

News: 202-326-6581, FAX 202-371-9227

Bateman House, 82-88 Hills RoadCambridge, UK CB2 1LQ

+44 (0) 1223 326500, FAX +44 (0) 1223 326501

SUBSCRIPTION SERVICES For change of address, missing issues, neworders and renewals, and payment questions: 866-434-AAAS (2227)or 202-326-6417, FAX 202-842-1065. Mailing addresses: AAAS, P.O.Box 96178, Washington, DC 20090-6178 or AAAS Member Services,1200 New York Avenue, NW, Washington, DC 20005

INSTITUTIONAL SITE LICENSES please call 202-326-6755 for anyquestions or informationREPRINTS: Author Inquiries 800-635-7181

Commercial Inquiries 803-359-4578

PERMISSIONS 202-326-7074, FAX 202-682-0816

MEMBER BENEFITS AAAS/Barnes&Noble.com bookstore www.aaas.org/bn;AAAS Online Store www.apisource.com/aaas/ code MKB6; AAASTravels: Betchart Expeditions 800-252-4910; Apple Store www.apple/eppstore/aaas; Bank of America MasterCard 1-800-833-6262priority code FAA3YU; Cold Spring Harbor Laboratory PressPublications www.cshlpress.com/affiliates/aaas.htm; GEICO Auto Insurance www.geico.com/landingpage/go51.htm?logo=17624; Hertz 800-654-2200 CDP#343457; Office Depot https://bsd.officedepot.com/portalLogin.do; Seabury & Smith Life Insurance 800-424-9883; Subaru VIP Program 202-326-6417; VIP Moving Serviceswww.vipmayflower.com/domestic/index.html; Other Benefits: AAAS Member Services 202-326-6417 or www.aaasmember.org.

[email protected] (for general editorial queries)

[email protected] (for queries about letters)

[email protected] (for returning manuscript reviews)

[email protected] (for book review queries)

Published by the American Association for the Advancement of Science(AAAS), Science serves its readers as a forum for the presentation anddiscussion of important issues related to the advancement of science,including the presentation of minority or conflicting points of view,rather than by publishing only material on which a consensus has beenreached. Accordingly, all articles published in Science Ð includingeditorials, news and comment, and book reviewsÐ are signed and reflectthe individual views of the authors and not official points of view adoptedby AAAS or the institutions with which the authors are affiliated.

AAAS was founded in 1848 and incorporated in 1874. Its mission is toadvance science, engineering, and innovation throughout the world for the benefit of all people. The goals of the association are to: enhancecommunication among scientists, engineers, and the public; promote anddefend the integrity of science and its use; strengthen support for the science and technology enterprise; provide a voice for science on societalissues; promote the responsible use of science in public policy; strengthenand diversify the science and technology workforce; foster education in science and technology for everyone; increase public engagement with science and technology; and advance international cooperation in science.

SENIOR EDITORIAL BOARD

BOARD OF REVIEWING EDITORS

BOOK REVIEW BOARD

Published by AAAS

568

NEWS>>

THIS WEEK Why we havefingerprints

Sorghumdecoded

572 573

Almost exactly 10 years after two groups iso-lated human embryonic stem cells, ignitingtremendous hope for new cures, the cells areabout to be injected into humans for the firsttime. Last week, the U.S. Food and DrugAdministration (FDA) gave Geron in MenloPark, California, permission to conduct asafety test in a handful of patients with arecent spinal cord injury.

For Geron and the scientists who workwith it, FDA' s decision was the culminationof a huge effortÐ including studies of nearly2000 rodents with spinal-cord injuries and a22,500-page application. ª I actually have aglass of champagne in my hand right now,ºsays a key player, Hans Keirstead, a neuro-scientist at the University of California,Irvine. Several years ago, he approachedGeron with the idea to commercialize hisfinding that stem cells could be used to mit-igate spinal cord injury in rodents. He hasbeen working with the company ever since. ª Idon' t expect this treatment to allow patientsto jump out of wheelchairs and play soccer,ºbut ª a meaningful and incrementaladvanceº in mobility is a real pos-sibility, he says.

But many stem cell researchers,particularly those in academia, whohave struggled since 2001 with theBush Administration' s strict limitson the development and use of newstem cell lines, are concerned thatthis trial may not be the best firstcandidate. Safety is one worry: Forexample, a big fear is that the cells could form atype of tumor called a teratoma. Some alsoquestion the trial' s scientific rationale.

Evan Snyder, a neuroscientist who directsthe stem cell research center at the nonprofitBurnham Institute for Medical Research inSan Diego, California, warns that a shakystart could set the field back enormously.ª There' s a lot of debate among spinal cordresearchers that the preclinical data itselfdoesn' t justify the clinical trial,º says Snyder,who is working on using neural stem cells fordrug delivery. Among the concerns he cited:

The rodents Geron studied had more moder-ate injuries than patients expected in the trial,suggesting that the results might not trans-late, and the therapy has not been tried inlarger animals. John Gearhart of the Univer-sity of Pennsylvania, who led one of theteams that isolated the cells in 1998, addsthat ª we' re still ¼ a long way from reallyunderstanding a good deal about these cellsand how to use them safely.º

Geron will be testing oligodendrocyteprogenitor cells, precursors to some nervoussystem cells the company developed from

one of the original human embryonic stemcell linesÐ created with Geron funding inJames Thomson' s lab at the University ofWisconsin, Madison. Eight to 10 patientswill receive the cells a week or two after aserious spinal cord injury. The goal is not tocreate new nerve fibers but to support thosestill intact by making the nerve insulatormyelin. To prevent rejection, patients willtake immune-suppressing drugs for about60 days. Although the primary goal is toassess safety, Geron will be looking forhints that the cells had an effectÐ for exam-

ple, improving bladder and bowel function,sensation, or mobility.

Geron CEO Thomas Okarma says heisn' t concerned about one of the risks thatpeople mention: ending up with the wrongtype of cell. In rodents, he says the injectedcells only formed glial cells, as expectedÐexactly the result he and FDA wanted.Geron has also performed extensive rodentstudies that assured the company and FDAthat the experimental cells did not causetumors in the animals. Keirstead andOkarma assert that, despite the criticisms,they' ve done everything they can before tak-ing the next step. ª There' s nothing we can dobut go to humans now,º says Keirstead. Ani-mal testing has its limitations, he addsÐincluding the fact that there are no large ani-mal models of spinal cord injury. (FDAdeclined to comment in detail on its decision

to let the trial begin.) Okarma suggests that aca-

demic researchers may be con-cerned because they' re notfully aware of what the com-pany has accomplished. Geronhas published or presented lit-tle of its oligodendrocytework; so far only FDA offi-cials have been privy to mostof it. ª There is so little expert-ise in the academic worldabout cell therapy that thesepeople are rightly nervous,ºOkarma says. ª We are so farahead of them.º Geron is alsoexamining whether its oligo-dendrocytes might help Alz-heimer' s disease, stroke, ormultiple sclerosis sufferers.

Other companies, mean-while, are developing products

derived from embryonic stem cells, and it' sexpected that upcoming trials will advancemore easily. Keirstead, for example, is work-ing with a second California company that iscoaxing the cells to form motor neurons andplans to test them in infants with spinalmuscular atrophy.

Gearhart says that for years ª we werealways told, ` Cure a patient and then all ofthis [controversy] will go away,' º andembryonic stem cells will quickly gainacceptance. Now, he says: ª Here comes thefirst test out of the box.º ± JENNIFER COUZIN

Celebration and Concern Over U.S. Trial of Embryonic Stem Cells

BIOTECHNOLOGY

CR

ED

ITS: D

AN

IEL A

. A

ND

ER

SO

N, U

C IR

VIN

E C

OM

MU

NIC

AT

ION

S; (IN

SE

T) G

ER

ON

30 JANUARY 2009 VOL 323 SCIENCE www.sciencemag.org

All smiles. Neuroscientist Hans Keirstead initiatedthe work that led to Geron' s new therapy for spinalcord injury, using cells derived from an embryonicstem cell line (inset).

Published by AAAS

...

......



FOCUS Forecastingprotests in China

574

A military forayinto social science

576

CR

ED

IT: C

HR

IST

OP

HE

R H

EN

SH

ILW

OO

D A

ND

FR

AN

CE

SC

O D

'ER

RIC

O

In 2002, a discovery at Blombos Cave inSouth Africa began to change howresearchers view the evolution of modernhuman behavior. Archaeologists reportedfinding two pieces of red ochre engravedwith crosshatched patterns, dated to 77,000years ago. Many experts interpreted theetchings as evidence of symbolic expressionand possibly even art, 40,000 years earlierthan many resea rchers had though t(Science, 11 January 2002, p. 247). Nowthe Blombos team reports on anadditional 13 engraved ochrepieces, many dated to100,000 years ago. Theresearchers suggest thatsome of the engravingsmay represent an artisticor symbolic tradition. Ifso, the timeline for the earliest known sym-bolic behavior must once again be redrawn.

ª [I] almost fell off my chairº on seeingthe latest ochre etchings, says archaeologistPaul Mellars of the University of Cambridgein the United Kingdom. At least some ª areunquestionably deliberate designs; ¼ theyhave to be some kind of symbols,º he says.Archaeologist Paul Pettitt of the Universityof Sheffield, U.K., a skeptic about the origi-nal discovery, says, ª The new materialremoves any doubt whatsoever.º

Others remain cautious, however, sug-gesting that the etched lines may have beenproduced incidentally when working ochrefor utilitarian reasons.

Pett i t t , Mellars, and other exper tsattended a meeting earlier this month inCape Town, South Africa, where the Blom-bos paper was presented; it is also in pressat the Journal of Human Evolution (JHE).After the meeting, researchers toured thesite, which has become crucial for under-standing early human behavior. Archaeol-ogist Christopher Henshilwood of the Uni-versity of the Witwatersrand in Johannes-burg, South Africa, lead author and thecave' s discoverer, has reported numeroussigns of apparent symbolic behavior fromBlombos, including incised ochre, shellbeads, and sophisticated tools, all presum-ably crafted by Homo sapiens (Science, 16

April 2004, p. 369). Blombos is one of sev-eral sites in Africa and the Near East thathave challenged the notion that full-fledged symbolism, such as cave paint-ings, did not appear before about 40,000years ago in Europe. ª There is now noquestion that explicitly symbolic behaviorwas taking place by 100,000 years ago orearlier,º says Mellars.

He and others say that the Blombos dates

seem accurate. Henshilwood' s team has usedat least four dating methods, including opti-cally stimulated luminescence (OSL) datingof quartz grains from the cave' s sedimentsand thermoluminescence dating of stonetools. The most recent round of OSL datingput the earliest archaeological levels atBlombosÐ where eight of the 13 new ochrepieces were foundÐ at about 99,000 yearsago. ª The stratigraphy is impeccable, withremarkably well-layered and discrete lensesof material,º says Pettitt, a dating expert.

To analyze the latest finds, Henshilwoodteamed up with Francesco d' Errico of theUniversity of Bordeaux in France and inde-pendent ochre expert Ian Watts, who isbased in Athens. The trick with ancientochre is to f igure out what early humanswere using it for. Many previous studieshave concluded that ochre was often groundto make a powder, which could have been

used to paint bodiesÐ a form of social iden-tification usually considered symbolicÐ orfor more utilitarian purposes. For example,Lynnette Wadley of Witwatersrand hasargued from modern-day experiments thatground ochre could have been used as akind of glue to haft stone tools into woodenor bone handles.

So Henshilwood and colleagues focusedtheir attention on 13 pieces engraved in

ways that seemed inconsistentwith grinding alone. Some

pieces have lines arranged inapparent fan-shaped orcrosshatched designs; oth-ers are etched in wavy pat-terns. Microscopic exami-

nation showed that these engrav-ings had been made with a pointed stone

tool and a finely controlled hand.Wadley agrees that ª some of the pieces

seem engraved for reasons other than ochrepowder removal,º although she is not yetconvinced that those reasons were sym-bolic. Archaeologist Richard Klein of Stan-ford University in Palo Alto, California,says that ultimately the question of whetherthe engravings were symbolic ª is not some-thing that science can resolve.º The teampoints out, however, that some of the oldestpieces have a crosshatched pattern similar tothat of the two original ochre pieces dated to77,000 years ago. And other researchershave very recently discovered similar cross-hatched patterns on a few African stoneand bone objects thought to be as old as thenew finds, or nearly so. This refutes sugges-tions that the marks are merely doodles,Henshilwood says, and suggests a 25,000-year tradition of symbolic representation.

If so, modern humans were probablyengaging in symbolic behavior even beforethe 100,000-year mark at Blombos and pos-sibly since the origin of our species, some-time between 160,000 and 200,000 yearsago, says Mellars.

Still, even if the engravings are sym-bolic, Klein says, the question remains:ª What did they symbolize?º Researchersagree that we may never know.

± MICHAEL BALTER

Early Start for Human Art? Ochre May Revise TimelineHUMAN EVOLUTION

Traces of an artist' s hand? Engravings on this

ancient ochre may have had symbolic meanings.

Published by AAAS

...

......



SO

UR

CE

: S

CIE

NC

E, T

EC

HN

OLO

GY

AN

D C

OM

PE

TIT

IVE

NE

SS K

EY

FIG

UR

ES R

EP

OR

T 2

008/2

00

9

NEWS OF THE WEEK

Mediterranean Sea

United

Kingdom

Sweden

Spain

Slovenia

Slovakia

Romania

Portugal

Poland

Neth.

Malta

Lux.

Lithuania

Latvia

Italy

Ireland

Hungary

Greece

Germany

F r a n c e

Finland

Estonia

Denmark

Czech Rep.

Bulgaria

Bel.

Austria

Percentage Change in R&D Intensity (2000± 2006)

Estonia +146

Latvia +97

Cyprus +90

Lithuania +56

Spain +40

Hungary +36

Czech Republic +36

Austria +33

Romania +33

Ireland +24

Slovenia +16

Portugal +11

Denmark +9

Italy +5

Finland +4

Germany +3

Malta +2

Greece ± 2

France ± 3

U.K. ± 4

Belgium ± 8

Netherlands ± 9

Bulgaria ± 11

Sweden ± 13

Luxembourg ± 14

Poland ± 17

Slovakia ± 35

+146%0± 35%

European research got a mixed report card in an analysisreleased last week by the European Union. The report saysthat the 27 E.U. nations have done well at increasing theirresearch work force: Numbers grew twice as fast as in theUnited States between 2000 and 2006, reaching 640,000researchers, while also attracting more foreign researchersto come and work there. Europe also attracted recordamounts of private R&D funding from U.S. companiesduring that period. But at the same time its ª R&D inten-sityº (research spending as a percentage of gross domesticproduct) pretty much stuck at about 1.84%Ð a long wayfrom the E.U.' s self-imposed goal of reaching an R&Dintensity of 3% by 2010.

Across the E.U., there is much variation in R&D intensity,with 17 countries, particularly new members in easternEurope, making marked improvements. But 10 countries,including science powers such as France and the UnitedKingdom, declined. Meanwhile, Japan increased its inten-sity from 3.04% to 3.39%, Korea from 2.39% to 3.23%,and China from 0.90% to 1.42%. (U.S. research intensityfell, from 2.74% to 2.61%.) ± DANIEL CLERY

European Science Not As Intense As HopedRESEARCH FUNDING

A dispute over whether it is safe to reopenIraq' s renowned archaeology museum inBaghdad has cost the head of the country' sarchaeology board her job. The battle over theBaghdad museum, closed since before the U.S.invasion in 2003, is one of several stickingpoints in the ongoing debate over how to man-age the country' s cultural heritage.

Iraq' s new minister for tourism and antiq-uities, Qahtan al-Juburi, visited the IraqMuseum on 3 January and demanded that themuseum be opened to the public by mid-February, according to several Iraqi andAmerican sources. The acting head of theState Board of Antiquities and Heritage,Amira Edan, argued against the minister' sproposal for security reasons, says DonnyGeorge, former head of SBAH now teachingin the United States, who spoke with Edan

about the incident. Edan had previouslyoffered to resign because she lacked the con-fidence of the ministry, says another sourcewho requested anonymity, but that offer wasignored. Following Iraqi media reports of heropposition to reopening the museum,however, al-Juburi accepted her resignationon 11 January. Neither the ministry nor Edanresponded to requests for interviews.

The ministry is controlled by a Shiite partyeager to see U.S. troops depart Iraq, and severalU.S. and Iraqi archaeologists say that reopen-ing the museum would be a potent politicalsymbol. ª That would be a message to theworld that everything is f ine and that theAmericans can leave,º says George, whobelieves that unlocking the museum doors ª is aterrible thing to do.º Another researcher famil-iar with the situation, however, says that open-

ing some of the galleries poses no major threatbecause ª there is a good security systeminstalled.º But the source adds that the SBAHchief should have a say in the decision and thather dismissal ª is disturbing.º

Al-Juburi also refused to allow a team ofIraqi archaeologists, including Edan and herreplacement, Qais Hussein Rashid, to visitWashington, D.C., this month to discuss howto spend a $700,000 grant from the U.S. StateDepartment to help develop a master plan forthe ancient Mesopotamian capital of Baby-lon, once the world' s largest and richest city.Babylon has suffered from years of neglect,shoddy reconstruction, and damage duringrecent occupation by U.S. and Polish troops.Provincial authorities are eager to open thefragile site to tourism, but archaeologistswant to preserve it.

Provincial authorities have also assertedtheir claims to ancient objects discovered byfarmers and construction workers, although bylaw such objects must be sent to the national

Iraq Museum May Reopen Amid Controversy

ARCHAEOLOGY

Cyprus

Published by AAAS

...

......



NEWS OF THE WEEKSO

UR

CE

: (T

OP

) U

.S. H

OU

SE

OF

RE

PR

ESE

NTA

TIV

ES, U

.S. SE

NA

TE

; (B

OT

TO

M) A

LI Y

USSE

F/A

FP

/GE

TT

Y IM

AG

ES

AGENCY HOUSE OF REPRESENTATIVES SENATE

NIH Total: $3.5 billion Total: $3.5 billion

$1.5 billion for extramural research, $1.5 billion for extramural facilities, $2.7 billion split between director’s office and the institutes, $500 million and $500 million for on-campus buildings. for on-campus buildings, and $300 million for extramural instrumentation.

NSF Total: $3 billion Total: $1.4 billion

$2 billion for research, $100 million for education, $400 million for new facilities, $1 billion for research, $50 million for education, $150 million $300 million for instrumentation, and $200 million for academic renovation. for new facilities, and $200 million for instrumentation.

DOE Total: $41 billion Total: $40 billion

$2 billion for Office of Science and $400 million for ARPA-E. $430 million for Office of Science, $2.6 billion for energy efficiency and renewable-energy research.

NASA Total: $600 million Total: $1.5 billion $400 million for science, $150 million for aeronautics, and $50 million $500 million for science, $250 million for aeronautics, $500 million for hurricane repairs. for human exploration, and $250 million for hurricane repairs.

NOAA Total: $1 billion Total: $1.2 billion

$600 million for climate sensors and modeling, $400 million for habitat restoration. $795 million for facilities/equipment, $427 million for restoration/maintenance.

NIST Total: $520 million Total: $575 million

$300 million for extramural buildings, $100 million for intramural research, $357 million for intramural facilities, $218 million for competitive grants. and $70 million for Technology Innovation Program.

Biodefense Total: $900 million $870 million for pandemic flu vaccine. $430 million for BARDA and $420 million for pandemic flu vaccine.

USGS Total: $200 million Total: $135 million

Upgrade laboratories; enhance National Map project and various monitoring networks.

As the U.S. economy slides deeper into a recession, universities arefollowing other sectors in freezing salaries, canceling job searches,and trimming expenses. At the same time, however, academicresearchers are on the verge of receiving a major influx of federalfunding as part of a 2-year, $825 billion economic stimulus packagemoving rapidly through Congress.

The bills, drawn up in consultation with the new Obama Admin-

istration, include some $360 billion in new spending, along with$275 billion in tax breaks and a large expansion of mandatory pro-grams. The research and science infrastructure components tuckedinto the first category amount to roughly $15 billion spread acrossseveral federal research agencies. As Science went to press, the Houseof Representatives was preparing to vote on legislation introduced on15 January by Democratic leaders, while the Senate was just begin-ning deliberations on its version. Democratic leaders have promisedto have the bill ready for the president' s signature by mid-February.

The table below shows the major science components, byagency, of each bill. ± JEFFREY MERVIS

U.S. BUDGET

museum for cataloging and analysis. ª Buteach [province] hopes to become independent,so they won' t send in the antiquities,º says anarchaeologist close to Edan, who has pleadedwith the provinces to cooperate.

Another source of tension is the fate ofancient Jewish manuscripts captured duringthe invasion. Widah Na' srat, a member of theIraq Interior Ministry' s Criminal Investiga-tions Department, told the London-basedpublication Al-Hayyat on 18 January that hesuspects U.S. contractors of smuggling someof the manuscripts to Israel. He did not elabo-rate but said he would visit Washington soonto investigate the matter. Jeffrey Spurr, aHarvard University researcher, says the manu-scripts were housed in Saddam Hussein' ssecret service headquarters and damaged bywater during the fighting, then frozen andflown to Texas for conservation in the summerof 2003 with the permission of SBAH. Theyare now in a Maryland facility, he says, buthave not been cataloged. ± ANDREW LAWLER

Safety first. Amira Edan led visiting dignitaries and Iraqi officials, shown here in the Assyrian gallery, on a

November tour of the closed and heavily guarded Baghdad museum.

A Stimulus for Science

Published by AAAS

...

......



NEWS OF THE WEEK

A 5-year-old who devourseverything she can f ind ondinosaurs is not only learningabout the natural world, she' salso experiencing the joy ofbecoming an expert. Thatfeeling, says Dennis Bartels,executive director of thefamed Exploratorium in SanFrancisco, California, is noless important to becomingscientifically literate than isher ability, as she movesthrough school, to define pho-tosynthesis or explain New-ton' s laws of motion. Stan-dardized tests may be good atassessing the latter type ofknowledge, Bartels says, butthey don' t capture what gets students excitedabout science.

A new report from the U.S. NationalAcademies comes to grips with a knottyproblem for science educators, namely, track-ing the science that people learn when they' renot in school. Part of the solution, accordingto Learning Science in Informal Settings:

People, Places, and Pursuits, is to stop taking

snapshots of a person' s understanding of atopic. Instead of quizzing them on a museumexhibit they' ve attended or a television pro-gram they' ve just watched, researchers needmetrics that recognize that learning occurs incountless locations and spans a lifetime.

ª People don' t realize that they are learn-ing science through everyday experiences,ºsays David Ucko, head of informal science

education at the U.S. Nat-ional Science Foundation inArlington, Virginia, whichspends $65 million a year onsuch activities and requestedthe academies' report. That' strue not just for children,Ucko adds, but also for theadults who teach them.

The idea that people learnabout science in many waysmay seem obvious, admitspanel chair Bruce Lewenstein,a professor of science commu-nications at Cornell Univer-sity. But too few researcherstake that into account whenstudying the impact of infor-mal science activities. The

result, says Lewenstein, is an ª uneven bodyof researchº on the subject.

ª Although there is strong evidence for theimpact of educational television on sciencelearning,º the report concludes, ª there is sub-stantially less evidence regarding the impactof other mediaÐ newspapers, magazines,gaming, radio Ð on science learning.ºLewenstein says that ª what' s been missing

Report Calls for Fresh Look at What Happens Outside SchoolINFORMAL EDUCATION

Fingerprints can help nab a criminal, but that' snot why they evolved. In fact, scientists aren' tentirely sure of the purpose of the tiny ridgeson our fingertips. Some have argued that theyimprove our grip on slippery objects, much asthe treads on tires help grip the road. Othershave suggested that they improve our sense oftouch. The two hypotheses aren' t mutuallyexclusive, but online in Science this week(www.sciencemag.org/cgi/content/abstract/1166467), a team of physicists presentscircumstantial evidence for the latter theory.

After a series of experiments with a sensordesigned to mimic a small patch of skin on ahuman fingertip, Alexis Prevost, GeorgesDebrégeas, and their colleagues at the ÉcoleNormale Supérieure in Paris conclude thatfingerprints likely enhance the perception oftexture by increasing vibrations in the skin asfingers rub across a textured surface. In par-ticular, fingerprints amplify vibrations in thefrequency range that best stimulates Pacinian

corpuscles, mechanoreceptors in the skinimportant for texture perception.

ª Texture information plays a huge role inour ability to identify objects by touch,º saysSliman Bensmaïa, a neuroscientist at JohnsHopkins University in Baltimore, Maryland.The new paper demonstrates how finger-prints could enhance this ability, Bensmaïasays. ª Their evidence ispretty compelling.º

The artificial fingertipused in the experimentsconsists of a half-millimeter-wide sensor covered witha dome of an artif icial,rubberlike material withmechanical propertiessimilar to those of humanskin. The researchers cre-ated two versions of thisª skinº : a smooth versionand one with parallel

ridges whose size and spacing approximatedthose of human fingerprints. Then they com-pared the vibrations picked up by the sensorwhen they slid a glass slide etched with finelines across the two types of skin. (To thehuman touch, the glass has a slight roughness,Debrégeas says.)

The ridges made the vibrations picked upby the underlying sensor up to 100 timesstronger, the researchers found. Moreover,when the glass slid over the skin at a speedcomparable to the typical motion of a person

scanning his fingers over asurface, the resulting vibra-tions tended to be in the sen-sitivity sweet spot for Pacin-ian corpuscles. ª It' s a reallyinteresting finding becauseit demonstrates the extent towhich the physical andmechanical properties of asensor can perform a com-putation,º says Mitra Hart-mann, a biomedical engi-neer at Northwestern Uni-versity in Evanston, Illinois.

Key evidence. Investigations with a

biomimetic sensor suggest that finger-

prints aid texture perception. CR

ED

ITS (T

OP

TO

BO

TT

OM

): S

ER

GIO

DIO

NIS

IO/G

ET

TY

IM

AG

ES; C

LO

UD

S H

ILL IM

AG

ING

LT

D./

CO

RB

IS

Fingerprints Enhance the Sense of Touch

NEUROSCIENCE

Bare-bones education? Museums are only one of many informal settings in which

people learn about science.

Published by AAAS

...

......



NEWS OF THE WEEK

are studies that measure the acquisition ofknowledge across time and outside of the tra-ditional educational media. Such studiesaren' t easy, but they can be done.º

In addition to conducting new types ofresearch, the report says scientists shouldadd ª excitementº and ª self-identificationwith scienceº to the definition of what itmeans to learn about science. A 2007 acade-mies' report, titled Taking Science to School,used traditional metrics such as whether stu-dents understand and apply the scientificmethod to explore the natural world, thinkabout science as a way of knowing, andengage in hands-on activities. But that' s toonarrow a definition, agrees developmentalpsychologist Kevin Crowley of the Univer-sity of Pittsburgh in Pennsylvania, whoworked on the 2007 report.

ª Schools have so many requirements onhow to do STEM [science, technology, engi-neering, and math] education that aren' tcompatible with informal science educa-tion,º he notes. ª Not only are kids compelledto be in school, but the adults in their liveswho are most excited about science are oftennot their teachers. We need to do a better jobof taking into account those out-of-schoolexperiences.º

Amen, says Bartels. ª As my colleagues liketo say, Nobody ever flunked a nature walk.' º

± JEFFREY MERVIS

Sorghum is no fair-weather cereal. Able tothrive in hot, semidry places, it feeds morethan 500 million people in 98 countries, pri-marily in the developing world. Sorghum isalso an important U.S. biofuel. Now, theanalysis of its genome sequence hasrevealed clues about how this crop toughsout subpar growing conditions. Drought tol-erance may arise from extra copies of keygenes, and sorghum' s eff icient photo-synthetic pathway was gradually cobbledtogether from existing photosynthetic genesand duplicated genes that shifted their func-tion over millions of years, says AndrewPaterson, a plant geneticist at the Universityof Georgia in Athens.

ª Having the genome sequence ofsorghum is a significant landmark,º saysWilliam Dar, director-general of the Inter-national Crops Research Institute for theSemi-Arid Tropics in Andhra Pradesh, India.ª We anticipate using a variety of approachesfor harnessing this genome sequence in ourapplied crop-improvement programs.º

At 730 million bases, the Sorghum

bicolor genome is 75% larger than that ofrice, the first sequenced cereal, but only aquarter that of maize, whose genome hasbeen sequenced but not yet analyzed. Manyplants, including rice and wheat, depend onC3 photosynthesis, so named because car-bon dioxide is initially converted to a three-carbon compound. But sorghum, whichoriginated in Africa, uses the C4 pathway,initially making four-carbon compounds.This pathway involves a special enzyme thattakes up carbon dioxide faster than doesC3 photosynthesis, enabling the plant tocurtail water loss through carbon dioxide±absorbing pores. As a result, C4 plants dobetter than C3 plants in hot climates.

When Paterson and his 44 colleaguesstudied the 34,496 putative genes unearthedin the sorghum genome, they discoveredthat the C4 pathway includes genes derivedfrom several belonging to the C3 pathway,as well as recently duplicated genes andgenes that date back 70 million years to awhole-genome duplication. ª This all sug-gests that it took quite a long time toevolve,º says Paterson.

With respect to withstanding drought,sorghum has four more copies than rice ofa regulatory gene that activates a key genefamily in a wide variety of plants during

droughts, Paterson and his colleaguesreport in the 29 January issue of Nature.Sorghum also has a surplus of genes forproteins called expansins, which may helpsorghum bounce back from water short-ages. In addition, it has 328 cytochromeP450 genes, which help plants respond tostress, whereas rice has 228 such genes.These specializations ª bear explorationºand may be useful for improving othercrops, says Neal Gutterson, CEO of MendelBiotechnology Inc. in Hayward, California.

Given the toll global warming is taking

on agriculture, ª understanding how majorcereal crops can be made to adapt to condi-tions of high temperature, high light inten-sity, and limited water supply, which can beelucidated from the sorghum genomesequence, will make a great impact in agri-culture,º says Takuji Sasaki, a rice geneticistat the National Institute of AgrobiologicalSciences in Tsukuba, Japan.

With respect to biofuels, sorghum' ssequence may help researchers improvemore complex biofuel crops such as Mis-

canthus and sugarcane. ª The smallgenome should help [in] identifying candi-date flowering time genes that may directlylead to manipulating biomass production,ºsays Thomas Brutnell, a plant geneticist atCornell University.

± ELIZABETH PENNISI

PLANT GENETICS

CR

ED

IT: C

OU

RT

ESY

OF

AN

DR

EW

PA

TE

RSO

N

Selecting and amplifying the signals importantfor texture perception could in principle beaccomplished within the nervous system, Hart-mann says. But in this case, it seems to be thedesign of the hardware rather than the program-ming of the neural software that does the trick.

The new results leave open why humanfingerprints are arranged in elliptical swirls.Bensmaïa notes that the amplification effectwas strongest when the textured glass slidperpendicular to the ridges, so it' s possiblethat the loops ensure that no matter how thefingers move, some ridges are always opti-mally oriented. It' s puzzling, however, thatmacaque monkeys have ridges parallel to thelong axis of their fingers, Bensmaïa says.The loops could represent an evolutionaryupgrade in humans, he suggests. Or perhapsthe monkeys use their f ingers differentlywhen exploring a surface, says Hartmann.

The work may one day lead to improvedprosthetic hands, Bensmaïa adds. ª It would bepretty straightforward to take their device andput it in a prosthetic hand, and I think that couldenhance tactile feedback quite a bit.º

± GREG MILLER

New cereal sequence. Like rice (foreground),sorghum now has a deciphered genome that ' srevealing its genetic history.

How Sorghum Withstands Heat and Drought

Published by AAAS

...

......


BEIJINGÐ On 22 June 2008, the body of ayoung woman was fished out of a river inWeng' an County in southwestern China. Thedeceased was quickly identified as 17-year-oldLi Shufen. Rumors began to swirl about howthe student died. Relatives went toWeng' an' s police station to inquireabout the investigation but wererebuffed; an uncle was beaten. On28 June, the police released theirfindings: Li had committed sui-cide. A few hundred angry peo-ple, convinced that the police were wrong,descended on the station. Word spread and thecrowd swelled to 30,000Ð nearly half the pop-ulation of Weng' an' s biggest town, Yongyang.That evening, a mob torched three govern-ment buildings and burned cars before policearrived with tear gas.

The seeds of unrest were sown long beforeLi' s death. In 2003, Weng' an County, inGuizhou Province, approved construction of adam but failed to adequately compensatethousands of people who were ordered to relo-cate, says Wang Erping of the Institute of Psy-chology of the Chinese Academy of Sciences(CAS). Later, Weng' an officials mandatedsetting aside nearly 20% of the county' s arableland for peppers, overriding farmers' objec-tions. And in 2006, another young womandied mysteriously; that case remains open.ª People were already unhappy with the localgovernment,º says Wang, who says Li' s deathwas the ª sparkº that ignited pent-up anger.

Until a year or so ago, China' s central gov-ernment routinely censored reports of massdisturbances. But the Guizhou incident, likeother recent protests, received extensive cov-erage in the Chinese press. The change is duelargely to the rising influence of China' s socialscientists. ª We want to tell the public what

really happened,º says Shan Guangnai, a sen-ior fellow at the Institute of Sociology of theChinese Academy of Social Sciences (CASS)here, which in 2007 advised the governmentto publicize mass incidents. ª It' s critical to

stop rumors from spreading.º That glasnost will likely be put

to the test in the coming months.Social scientists predict that theglobal financial crisis will sharplyescalate unrest in China. ª The firsthalf of 2009 will be a hard time,º

says Shan. Already, sagging demand for Chi-nese goods has forced some 670,000 compa-nies to close, resulting in millions of joblosses. In an annual rite, tens of millions ofChineseÐ including now-jobless migrantworkersÐ returned to their home villages forthe Lunar New Year holiday that started on 26 January. Many are unlikely to be satisfiedwith jobs in the countryside after havingtasted city life, says Shan.

Despite President Hu Jintao' s intention tocreate a ª harmonious society,º protests havegrown more frequent with China' s rapiddevelopment. Top officials lay the blame forsocial unrest primarily on labor disputes andenvironmental woes. Although numbers arehard to come by, in 2005, for instance, therewere 51,000 conflicts over ecological degra-dation, such as contaminated water, dust, andlandslides (Science, 1 August 2008, p. 611).

Hindsight may be 20/20. But can sciencepredict mass incidents before they happen?Wang thinks so. After compiling statistics andconducting interviews across China, he andhis team have developed a methodology that,they say, enables them to forecast the likeli-hood that a county will experience a distur-bance in the coming months. CAS briefed thecentral government on the findings last Octo-ber, and the State Council' s Emergency Man-agement off ice has since asked Wang toadvise it on how to monitor social attitudes inareas traumatized by last May' s devastatingearthquake in Sichuan Province. The projectª has potential to generate findings of consid-erable policy value,º says Li Lianjiang, anexpert on public administration at the ChineseUniversity of Hong Kong.

Others are skeptical that such forecasts areworthwhile. ª Who can observe or measure` the potential' to riot?º asks Ching Kwan Lee,a sociologist at the University of California,Los Angeles. ª But there is no point in arguingone way or the other. If they think they canpredict, let them and see if they get it right.º

If forecasting pans out, it could be a boon toleaders here. ª It can help the government dealwith mass disturbances before they occur,ºsays Wan Chuan of the Beijing Academy ofSocial Sciences. Knowing which tinderboxesare hottest might enable the central govern-

Seeds of DiscontentSocial scientists blame poor local governance for China' s rising unrest;

the global financial crisis, they warn, could make things far worse

Train station scrum. Tens of millions of

Chinese migrated home this month for

Lunar New Year celebrations. After the glow

fades, unrest may surge.


CR

ED

ITS (T

OP

TO

BO

TT

OM

): A

ND

Y W

ON

G/A

P; A

LVIN

CH

AN

/RE

UT

ER

S

NEWSFOCUS

OnlinePodcast interviewwith author

Richard Stone.

sciencemag.org

Published by AAAS

...

......


ment to focus resources and ameliorate condi-tions in the most dysfunctional districts. Suchforecasts might also be used to suppress dis-sent, some experts warn. But Wang and Shandon' t see a dark side: They say that the centralgovernment consistently strives to rectify thepoor local governance and policy failureslying at the heart of unrest. Toward that end,Wang says, ª we need a national social moni-toring system for social stability.º

A wake-up callFor decades in China, research on social prob-lems was taboo. The central government, asarchitect of wrenching societal transformationssuch as the Cultural Revolution, was responsi-ble for many societal illsÐ and didn' t wantunauthorized opinions being aired. Reformpolicies, initiated 30 years ago bythe late top leader Deng Xiaoping,opened the door to social research.In 1998, Wang' s team at the Insti-tute of Psychology focused on pub-lic administration. ª It was not aseasy as we imagined it would be,ºWang says. They scraped for fund-ing for several years. Then theWanzhou incident occurred.

On 18 October 2004, an explo-sive riot ensued when police triedto break up a street fight in theWanzhou district of Chongqing, amajor city on the Yangtze River incentral China. Unlike most massincidents in China, which takeplace in towns or rural areas, this one roiledChongqing, a city province, threatening todestabilize a region inhabited by more than30 million people. ª Wanzhou was a wake-upcall,º says Shan.

But CASS and CAS social monitoringresearch met strong resistance. When Wang' s18-strong team fanned out across China to col-lect statistics on crimes, protests, and citizens'complaints, local officials often claimed thedata were too sensitive to share. ª Many localgovernments try to hide their administrativefaults,º Wang says.

The farther from Beijing the scientistsroam, the less cooperation they tend toreceive. Last September, for example, Wangand a doctoral student made a fact-findingtrip to Menglian County in Yunnan Province,just across the border from Myanmar. Assoon as they arrived, officials advised them tohigh-tail it back to Beijing. Local authorities,they explained, were preoccupied with com-bating drug traffickers, and their research

ª would not be convenient.º Shan adds thatgathering data often requires subterfuge.When he and his two dozen CASS colleagueswho study social unrest anticipate localantipathy, he says, ª we pretend to be touristsor relatives of local residents.º

Despite such obstacles, Chinese social sci-entists have amassed a wealth of data. Wang' steam, for instance, has compiled dossiers on132 counties in the past 4 years. Commonthreads have emerged. The root causes of thevast majority of incidents, Wang says, arepoor local governance and miscarriages ofjustice. ª Grievances accumulate over time,ºhe says. Whatever the source of resentment,Chinese communities follow similar arcs.ª The f irst reaction is to endure, to avoidfights,º Wang says. When confronted withunfair treatment, Chinese usually attempt toprivately negotiate a solution, he says. Whenthat fails, they may seek redress from the gov-

ernment, the courts, or through the media. Ifgrievances remain, resentment buildsÐ andcan prime a community for an outburst.

A new wildcardShan insists that the protests and mass inci-dents now plaguing China are fundamentallydifferent from those other countries experi-ence. According to CASS research, he says,incidents in China are directed strictly at thegovernment. Crowds may turn violent, butviolence is focused. ª People don' t rob banks,they don' t loot, they don' t burn private cars,ºShan says. ª Our disturbances are not anabstract phenomenonº but usually arise fromlocal grievances. For that reason, he argues,incidents like the one in Weng' an are exceed-ingly unlikely to spread to other regions.

Trickier to anticipate are incidents steepedin religious conflict or separatism. Last March,for example, rioters rampaged through Lhasa,the capital of Tibet, killing at least a dozen peo-ple. Wang' s CAS group plans to launch sur-

veys of ethnic and religious conflicts later thisyear. But Tibet and Xinjiang Province in west-ern ChinaÐ the site of deadly terrorist attackslast summer attributed to Uighur separatistsÐare too sensitive for such research now, Wangsays. His group will cut its teeth on ethnicenclaves in Guangxi, Ningxia, or Qinghai.

The global financial crisis is also terranova. ª The crisis may stimulate a more wide-spread disaffection,º Shan says. Even beforethe crisis, peasants and laborersÐ the majorityof China' s 1.3 billion peopleÐ were strugglingto pay for education and medical care. ª Thefinancial crisis is pounding these people,º saysShan. The result, he says, is simmering resent-ment: not because blue-collar workers areedging into poverty, but because they feel asense of injusticeÐ real or imaginedÐ that thefinancial crisis is hitting them harder than it isChina' s well-heeled white-collar workers. Thegap between societal benefits and people' s

expectations is widening, warnsShan. That' s a recipe for mass dis-content. ª We' re very afraid of thisphenomenon,º he says.

The financial crisis offers aprime window for Wang' s CASteam to hone its forecasting. Teammembers search for correlationsbetween how people feel and howthey act. Based on data gathered in2007, they forecast the likelihoodof mass incidents in 35 countieslast year; they were correct for 33.CASS researchers say they arereluctant to forecast unrestÐ butlike their CAS colleagues, theyhave a good idea where blood is

boiling. ª We can' t say exactly where and whenmass disturbances will occur, but we can tellyou that they may happen because there areconflicts of interest or arguments over land,ºsays Yu Jianrong of CASS' s Rural Develop-ment Institute. The problem, Shan says, is thatit' s impossible to predict when a sparkÐ like LiShufen' s death in Weng' anÐ will set off amass incident.

Last July, several days after the incident,Shan' s team visited Weng' an. The provincehad already replaced the county' s top fourofficials. The CASS researchers didn' t have todisguise their objectiveÐ the new leadersreadily met with them. ª They learned theirlesson,º Shan says. One of their first acts wasto open a grievance bureau that by August hadlogged more than 2000 complaints. Theyfired the police officer who had assaulted Li' suncle, and they reopened the investigationinto her death.

It was ruled a suicide.± RICHARD STONE

Case study. Paramilitary police restore order afterlast June' s violent incident in Weng' an County.

Beyond the pale. Research on incidentsin Tibet, like last March' s violent protestagainst Chinese rule, is off limits.


CR

ED

IT: A

ND

RE

AS S

TE

INB

ICH

LE

R/A

PSOCIAL SCIENCE AND POLITICS | NEWSFOCUS

Published by AAAS

...

......



CR

ED

ITS: S

ITE

IN

ST

ITU

TE

/VID

EO

GR

AB

/HA

ND

OF

F/E

PA

/CO

RB

IS; P

HO

TO

S.C

OM

; (IN

SE

T) M

IT

NEWSFOCUS | SOCIAL SCIENCE AND POLITICS

Mark Woodward is an unlikely soldier inthe global war on terrorism. A professor of religious studies at Arizona State University (ASU), Tempe, and a lifelongacademic, he says ª a lot of my researchinvolves sitting in coffee shops and talkingto people.º Woodward has spent much of thepast 30 years trying to understand how localcommunities throughout Southeast Asiapreserve their own religious and culturalidentities as radical and violent Islamicmovements gain strength around the world.

Currently a visiting professor at GadjahMada University in Yogyakarta, Indonesia,Woodward recalls a recentvisit to a mosque nearlydestroyed by an earth-quake. A Saudi Arabianfoundation that was financ-ing its reconstruction alsowanted to provide a teacherwho would disseminateWahabi-style Islam. Vil-lage elders politely butfirmly declined the instruc-tional assistance, Woodward says. ª This isWahabi colonialism,º said one local leader.ª We don' t need Arabs to teach us Islam.º Thatreaction is why Woodward believes that ª theforces of localityº will prevail in a clash of ide-ologies. ª I think that attempts to estab-lish hegemonic Islam are going to fail,through very creative uses of tradi-tional rituals and language,º he says.

U.S. military leaders want to findout if he' s right. Last month, the U.S.Department of Defense (DOD), whichis waging wars in Iraq and Afghanistanand spending billions at home tocounter the threat from Islamicextremists, chose Woodward to lead ateam that proposed a study of ª thediffusion and influence of counter-radical Muslim discourseº in South-east Asia, West Africa, and Europe.The project is one of seven led bysocial scientists that were selected ear-lier this month to receive a total of $45 million from a controversial DODprogram called the Minerva ResearchInitiative. The Pentagon plans to issue a

second solicitation this spring of roughly thesame size, and officials have hinted stronglythat there will be subsequent rounds.

Minerva is a banquet for a field accus-tomed to living on scraps. But some socialscientists see it as a threat to academic free-dom. They cite the military' s history of ques-tionable research practices and worse goingback to the Vietnam War, running through theinterrogation of prisoners at GuantánamoBay, Cuba, and continuing with the much-maligned Human Terrain Teams now inAfghanistan. They say DOD' s choice of top-ics reflects a narrow, military perspective on

the world. In addition to ª thestrategic impact of religious andcultural changes in the Islamicworldº that Woodward' s projectaddresses, DOD solicited propos-als relating to ª terrorist organiza-tions and ideologies,º the relation-ship between Chinese technologi-cal and military growth, andBa' athist Party materials seized atthe start of the Iraq war. Critics

also worry that the lure of so much moneywill cause researchers to shift their attentionfrom more important issues.

ª The problem is the process,º saysBrown University professor Catherine

Lutz, one of many anthropologists who havebeen scornful of the program. ª DOD should-n' t be involved because it' s not likely to fundthe best work. My fear is also that DOD willchoose researchers who agree with themabout the problems that the world is facing.º

DOD officials say they' ve bent over back-ward to address those concerns. When Secre-tary of Defense Robert Gates unveiled Min-erva last spring in a talk to the Association ofAmerican Universities, he acknowledged theoften ª hostileº relationship between the mili-tary and social scientists and pledged thatMinerva would abide by a policy of ª completeopenness and rigid adherence to academicfreedom and integrity.º

Toward that end, DOD held a well-attended community workshop in August.And there' s a Web site run by the Social Sci-ence Research Council (SSRC), a venerableNew York City± based nonprofit researchorganization, that has published 18 essays on the controversy (www.ssrc.org/essays/minerva/). Webmaster Thomas Asher, ananthropologist by training, says he hopes thedialogue will improve future solicitations;ironically, SSRC' s own bylaws preclude itfrom accepting military funding.

William Rees, deputy undersecretary ofdefense for labs and basic research, whooversees the Minerva initiative, emphasizesthat the research is unclassif ied and thatresults will be posted on the project' s Website (Minerva.dtic.mil). He says his goal is toattract the best researchers, to expand thepool of scientists addressing these ques-tions, and to foster collaborations among

researchers from many fields.An examination of the f irstcohort of winners suggeststhat he' s come close to hittingall three targets.

DOD received 211 initialqueries from researchers seek-ing funding in one of five cate-gories, four times the numbercommunity leaders had told

him to expect, says Rees. ª Rees wasworried about getting the topresearchers to participate,º saysHoward Silver, executive director ofthe Washington, D.C.± based Consor-tium of Social Science Associations.ª My sense is that he got the A team.º

Many of the Minerva granteesalready have ties to the defenseestablishment. One such grantee isDavid Matsumoto, a professor ofpsychology at San Francisco StateUniversity in California. His teamwill study the role of emotion in stok-

DOD Funds New Views on Conflict

With Its First Minerva GrantsThe Pentagon makes a $45 million bet that social scientists can help it understand

the worldÐ and protect the United States

SOCIAL SCIENCES

Imagemakers. A 2007 video of Osama bin Laden, released by a groupthat monitors terror messages, is part of the cyberworld of diplomacythat Nazli Choucri (inset) is studying.

ª In the cyberworld,

anybody can play.

We need ... the

conceptual tools

to couple the virtual

and the real worlds.ºÐ NAZLI CHOUCRI

Published by AAAS

...

......



SOCIAL SCIENCE AND POLITICS | NEWSFOCUS

ing or quelling ideologically driven move-ments. A longtime collaborator with psychol-ogist Paul Ekman in his work on microexpres-sions, Masumoto has helped train airportscreeners for the Transportation SecurityAdministration and has worked with severalDOD agencies over the years on what he callsª behavior-detection techniques.º

On the other end of the career continuum isJacob Shapiro, an assistant professor of publicaffairs at Princeton University. A former U.S.naval officer who was on active duty from1998 to 2002, he completed his postdoc only 1year ago and is lead researcher on a project tounderstand the economics of counterinsur-gency movements around the world. ª Ientered the academic community because Ifelt there were not enough veterans in theacademy, and that is not a good thing,º he says.ª The military represents all of society and soshould the academy.º

Rees hopes the Minerva program will bringtogether scientists who haven' t had a chance towork on a problem of mutual interest and allowsmall interdisciplinary groups to expand theiractivities. That' s what Nazli Choucri, a politi-cal scientist at the Massachusetts Institute ofTechnology (MIT) in Cambridge, is hoping toaccomplish with her project to examine cyberinternational relationships. The team includesforeign policy and national intelligence heavy-weights such as Harvard University' s AshtonCarter and Joseph Nye, as well as Internet andartificial intelligence gurus such as MIT' sDavid Clark.

ª Our current theories are inadequate, and what we know now is anecdotal,º saysChoucri. ª In the cyberworld, anybody can play.We need a fuller vocabulary to understandcyberspace as an environment, as well as theconceptual tools to couple the virtual and thereal worlds.º Choucri is in line to receive thelargest single Minerva grant, which Pentagonofficials expect to be approximately $10.4 mil-lion over 5 years. (Grantees are still negotiatingwith DOD on funding levels.)

That type of funding is on a scale mostsocial scientists have only dreamed about.ª We' re talking about a huge order of magnitude biggerº than a typical grant, saysWoodward, who requested $5.8 million.Woodward is working with Muhammad SaniUmar of Northwestern University inEvanston, Illinois, an expert on Islam in west-ern Africa, and David Jacobson, a professor ofglobal studies at ASU, who' ll examine Islamiccommunities in France and Germany. Theproject will combine ethnographic fieldworkat each location with global survey data onpublic attitudes toward Muslims. It will alsofeature a Web component to track the flow of

ideas across the various Islamic communitiesand analyze their influence on daily life.

For labor economist Eli Berman of theUniversity of California, San Diego (UCSD),the Minerva grant is a game changer. He isworking with Shapiro tounderstand what it takes forcommunities to counteractgrass-roots movementssuch as Hamas or the TamilTigers. ª Instead of just asummer salary and a gradu-ate student, I' ll be able to dosurveys and experimentsaround the world, partnerwith additional organiza-tions, and bring on postdocs as well as severalgraduate students,º he says. ª We' ll be able toaccomplish things in a matter of years ratherthan decades.º

Berman is also a research director for theUC-wide Institute on Global Conflict andCooperation, based at UCSD, that receiveda grant to explore how China' s growingtechnological prowess is fueling the mod-ernization of its military forces. The drivingforce behind the project is the institute' s TaiMing Cheung, a former journalist who hasseen the literature on the topic explode overthe past 2 decades in step with China' sbooming economy.

ª Tai Ming has been working on this foryears as the lonely monk scholar, and this grant will allow us to engage many otherresearchers,º says Susan Shirk, who directs theinstitute and is the named principal investiga-tor on the grant. ª Most of the social scientistsworking in China are looking at rural develop-ment, or urbanization, issues that are a lot eas-ier to study and less sensitive. It' s hard to find

academic jobs from which you can look at[Chinese] national security issues.º

None of the grantees who spoke toScience expressed concern about limitationson their research or on how it could be pre-

sented. ª We' re not in thebusiness of providing DODwith information that is tac-tical or operational,º saysWoodward. ª This is basicsocial science research. It' snot telling the governmentwhat it wants to hear.º

In fact, one grantee whohas written about ª why wegot it so wrongº on the sta-

tus of Iraqi biological weapons before the U.S.invasion hopes her project will help policy-makers understand that uncertainty isinevitable and that perfect knowledge isimpossible. Patricia Lewis, a nuclear physicistwho directs nonproliferation research at theMonterey Institute of International Studies inCalifornia, will lead a team analyzing materi-als captured in 2003 that many social scientistssay do not even belong in U.S. hands. ª I' minterested in how we interpret information andhow we too often see things in the light of whatwe already believe,º says Lewis. ª Biases areeverywhere, and the worst sorts are those thataren' t disclosed.º

That culture of openness apparently meanssomething different to political scientistJames Lindsay of the University of Texas,Austin, who refused to discuss his Minervaproject, which is titled ª Climate Change, StateStability, and Political Risk in Africa.º He toldScience, ª I don' t owe you an explanation, andI have nothing to say about the program.º

± JEFFREY MERVIS

Bombs or bombast? Patricia Lewis (inset) and her team will explore the context ofmaterials seized during the U.S. invasion of Iraq in April 2003, like these purporteddescriptions of chemical and biological weaponry.

ª I' m interested in how we

interpret information and

how we too often see

things in the light

of what we already

believe.ºÐ PATRICIA LEWIS

CR

ED

ITS: A

NJA

NIE

DR

ING

HA

US/A

P P

HO

TO

; (IN

SE

T) JA

ME

S M

AR

TIN

CE

NT

ER

FO

R N

ON

PR

OLIF

ER

AT

ION

ST

UD

IES

Published by AAAS

...

......


NEWSFOCUS

Wendy Freedman yearns to take the sharpestlook ever at the wonders of the cosmos. Anastronomer at the Carnegie Observatories inPasadena, California, Freedman is one of 50researchers developing the Giant MagellanTelescope, a $700 million behemoth thatwould combine seven 8.4-meter-wide mir-rors into one enormous ª segmented mirrorºeffectively 24.5 meters across. It wouldbring celestial marvels into the sharpestfocus yet achieved.

In fact, Freedman had expected to be wellon the way to completing such a telescope bynow. In 2001, astronomers in the UnitedStates ranked a giant segmented-mirror tele-scope as their first choice for ground-basedfacilities to be built in the following 10 years.Usually, that endorsement suffices to push aproject along. ª Certainly it was the expecta-tion that the number-one ranked projectwould move forward,º Freedman says. Butplans for a giant telescopeÐ and half the pro-posals in the 2001 surveyÐ have yet to cometo fruition.

That presents astronomers with anunusual problem. Early each decade sincethe 1960s, they have developed a list of prior-ity projects Ð an exercise known as aª decadal surveyº Ð to tell funding agenciesand the U.S. Congress which projects theircommunity wants most. Conducted by theNational Research Council, those surveyshave earned astronomers a reputation forself-discipline. ª Congress listens to themand says, ` Okay, these people have got theiract together,' º says John Mather, an astro-physicist at NASA' s Goddard Space FlightCenter in Greenbelt, Maryland.

But now, as a new committee begins the18-month process of mulling over proposalsfor inclusion in the next decadal survey, 15 ofthe 20 projects listed in the especially ambi-tious 2001 survey remain unfinished. Onlytwo of the major initiatives (ground-basedprojects with capital costs greater than $50

million and space missions with capital costsabove $500 million) have received approvalfor construction (see table), and the commit-tee will have to reconsider the unapprovedproposals even as it evaluates myriad newideas. In weighing more than 170 proposals,the committee will surely have to make someunpopular decisions.

ª The pressure is on as never before tomake tough choices,º says Roger Blandford,a theoretical astrophysicist at Stanford Uni-versity in Palo Alto, California, who chairsthe new survey committee. ª There are goingto be a lot of proposals for very expensiveprojects, and the very good is going to betrumped by the excellent.º

Astronomers will have to come up withmore than a wish list, however. Many of the

cost estimates in the 2001 survey turned out tobe far too low, so this time officials at the U.S.National Science Foundation (NSF), NASA,and the Department of Energy want numbersthat will stick. They also want guidance onkeeping projects on trackÐ or reshuffling thelist should some fall behind, says Craig Foltz,acting director of NSF' s astronomical sci-ences division. ª There' s a logjam,º Foltz says,ª and we need help breaking it.º

Bigger, pricier, slowerNot every large project in past surveys hasbeen completed within 10 years. The top pri-ority from the 1991 survey, an infrared satel-lite observatory now called the Spitzer SpaceTelescope, rocketed into orbit in August 2003.The 1991 survey also called for an array ofmillimeter-wavelength radio dishes that isonly now being built in Chile. However,astronomers and government officials worrythat uncompleted projects are becoming therule rather than the exception.

Part of the problem is that projects havegrown far bigger and more expensive, saysMartha Haynes, an astronomer at CornellUniversity and a member of the survey com-mittee. The 1991 survey listed one ground-

578

Beset by Delays, U.S. AstronomersPonder a Better To Do' ListWith only five of 20 projects from the last decadal survey completed and only five

more started, scientists search for ways to make their list of priorities more effective

ASTRONOMY

NEWSFOCUS

COST ESTIMATE

PROJECT 2001 Current

James Webb Space Telescope $1 billion $4.5 billion

Giant segmented-mirror telescope $700 million $700 million

Orbiting x-ray observatory $800 million New int' l project

Upgrade of Very Large Array of radio dishes $140 million $82 million*

Large Synoptic Survey Telescope $170 million $400 million

Spacecraft to find Earth-like planets $1.7 billion Project in limbo

Far-infrared space telescope $600 million >$1.5 billion

*Phase one only

CR

ED

ITS:JU

PIT

ER

IM

AG

ES; N

ASA

Unfinished business. Only two of the seven majorinitiatives listed in the 2001 survey have beenapproved for construction (checks). For the last twoprojects, only research and development money wasrecommended in 2001.

30 JANUARY 2009 VOL 323 SCIENCE www.sciencemag.org

Published by AAAS

...

......


based project with a price above $100 million.NSF is currently developing three that cost$250 million or moreÐ more than theagency' s annual astronomy budget. ª Thesemegaprojects just take a very long time,ºHaynes says.

ª Woefully lowº cost estimates in the2001 survey compounded the problem, saysLennard Fisk, a space scientist at the Uni-versity of Michigan, Ann Arbor. In that sur-vey, the construction cost of a successor tothe Hubble Space Telescope was estimatedat $1 billion. Now that satellite, NASA' sJames Webb Space Telescope, is approvedfor construction and expected to launch in2013 or later with a total mission cost of$4.5 billion. The Advanced TechnologySolar Telescope, to be built on Haleakala inHawaii, was estimated to cost $60 million;the current price tag is $250 million.

Reasons for the underestimates vary. Forexample, Mather, who leads development ofthe Webb telescope, says the $1 billion fig-ure for it in the 2001 report was not a rigor-ous estimate but a goal set by NASA offi-cials as part of then± NASA AdministratorDaniel Goldin' s push to do things ª faster,better, cheaper.º ª I told [the 2001 commit-tee] that this is what Dan Goldin wanted usto try to achieve,º he says. ª I didn' t promisethem that we could do it.º

The poor cost estimates played havoc withagency plans. At NASA, which funds space-based astronomy, the Webb telescope gobbledup money intended for other missions. AtNSF, spiraling costs left the agency strugglingto develop several big projects at the sametime. Such problems might have been avoidedif the 2001 survey had provided guidance onadjusting priorities as projects developed, Fisksays. ª What happens if the cost [of a project]increases by a factor of 2 or 3?º he says. ª Thesurvey didn' t address that issue.º

This decade' s theme: CredibilityGiven the problems getting through the 2001ª to doº list, both scientists and fundingagency officials are looking for somethingdifferent this time. Above all, the committeewill try to nail down a realistic cost estimatefor each proposal. ª If your budget and sched-ule and scope are not believable, you willlose,º Haynes warns.

In fact, the committee will hire cost con-tractors who will independently estimate thecosts of the various proposals, says MarciaRieke, an astronomer at the University of Ari-zona, Tucson, who is a member of the com-mittee. ª I don' t want to advertise that ournumbers will be good to 10%, but I' m hopingwe can do better than a factor of 3 or 4,º she

says. The numbers should be ª free of any spinfrom an agency or proponent group,º she says.

Officials would also like a protocol forreordering projects should one or anotherburst its budget or should funding to agenciesfail to rise as expected. Rieke says the com-mittee will try to identify technological mile-stones that each project must reach to stay ontrack and will suggest ways of shuffling prior-ities should a project fall behind.

Some scientists say there is a limit to howmuch guidance the ad hoc committee can pro-vide on reprioritizing. ª I think that searchingfor a mechanism that will work like clockworkwhen the committee is gone is a mistake,ºMather says. Rather, he says, agency officialsª need to be empowered to make the decisionsthey need to make.º

Foreign observers have some reservationsabout the current survey. Committee mem-bers ª are in a terrible position because if theydon' t recommend the things that were put for-ward last time, they' ll most likely get can-

celed,º says Simon White, a theorist at theMax Planck Institute for Astrophysics inGarching, Germany.

Martin Rees, an astronomer at the Univer-sity of Cambridge in the U.K., wonderswhether the decadal survey is losing its punch.Head-to-head comparisons aren' t much use,he notes, for big projects funded by differentagenciesÐ such as the Webb telescope and thegiant segmented-mirror telescope. And nowa-days, when most big projects are internationalefforts, a national ranking has less impact thanit used to.

Still, U.S. astronomers and officials saythe decadal survey serves a crucial purposeby identifying the proposals with the greatestscientific potential. ª It' s the science ques-tions that will drive the f ield in the nextdecade,º says Jon Morse, director of NASA' sastrophysics division. It' s less clear that amention in the decadal survey will still pro-pel a project to completion.

± ADRIAN CHO


NEWSFOCUSC

RE

DIT

:N

ASA

/JP

L-C

ALT

EC

H

Priorities Nearer to Home in Need of Better Cost Estimates

Planetary scientists are generally pleased with the results so far of their first-ever prioritization ofnew solar system missions, covering the years 2003 to 2013. But like their colleagues in astronomy(see main text), they have seen early cost estimates for two missions blow up, sending shock wavesinto their next decadal survey due in 2011. ª The survey process was the way to go,º says WesleyHuntress, director of the Carnegie Institution for Science' s Geophysical Laboratory in Washington,D.C., but those tackling the next planetary decadal surveyneed ª a less naïve idea of what missions will cost.º

Some cost estimates served well enough. Small missions(costing less than $325 million) should launch every 18 months,the study conducted by the National Research Council (NRC)concluded. Launches occurred in 2004, 2005, and 2007, andmore are scheduled for 2009 and 2011. Amedium-cost mission(less than $650 million) is on its way to Pluto, and another is inthe works for Jupiter. Among small Mars missions (a separateprioritization), the Phoenix lander has finished its mission inthe martian arctic, while a mission to study the martian upperatmosphere has been selected.

But priorities for large missions (costing more than $650 million) did not fare as well. A mis-sion to Jupiter' s moon Europa turned out to be wildly more costly than $650 million, so it droppedout of the decade. And returning samples from MarsÐ always seen as a huge-ticket itemÐ is noteven getting a serious look. Most surprising, though, was the ballooning of costs for the Mars Sci-ence Laboratory (MSL). Pegged in the survey at less than $650 million, its cost has swelled to morethan $2 billion as technical problems triggered a delay of 2 years. MSL ª sucked big money out of[the planetary program], but it is what we want to do,º says John Mustard of Brown University,chair of NASA' s Mars Exploration Program Analysis Group.

Getting what scientists want without jeopardizing future missions will require better cost esti-mates early on, says Huntress, who chaired an NRC midterm review of the decadal survey. ª NASAshould spend enough money before the decadal survey so you have a better idea what class a mis-sion will be,º he says. The situation is improving, he adds, and Congress is pitching in. It has man-dated that decadal surveys ª include independent estimates of the life cycle costs and technicalreadiness of missions ¼ whenever possible.º Scientists will no doubt find out whether their eyesare still bigger than NASA' s stomach.

± RICHARD A. KERR

MSL rover. Bigger,

costlier.

Published by AAAS

...

......



CR

ED

ITS (T

OP

TO

BO

TT

OM

): C

OU

RT

ESY

OF

MA

UR

A G

ILLIS

ON

/OSU

AN

D W

ILLIA

M W

EST

RA

/JO

HN

S H

OP

KIN

S U

NIV

ER

SIT

Y; SO

L S

ILV

ER

MA

N/U

CSF

NEWSFOCUS

When he began practicing in 1994, headand neck cancer specialist WendellYarbrough knew just what type of patientswould walk through his clinic doors. Theywere typically in their 50s or 60s, and inmost cases he attributed their throat andtonsil cancer to years of lighting up tobaccoand chugging down alcohol. Yarbrough' spatient roster at the Vanderbilt-Ingram Can-cer Center in Nashville, Tennessee, nowtells a different story: It' s not unusual to seepatients in their late 20s or 30s, he says, andfew of the younger patients smoke.

Nationally, while the total number oforopharyngeal cancersÐ affecting the areafrom the base of the tongue through the ton-silsÐ is in decline, the rate among Americansyounger than age 50 is creeping upward.According to surveillance data from theNational Cancer Institute (NCI), rates oforopharyngeal cancer in the 20 to 49 agegroup more than doubled from 1975 to 2005.

A search for the cause has led to a famil-iar suspect: the human papilloma virusesÐspecifically HPV 16 and to a lesser extentHPV 18, the same strains that cause 70% ofcervical cancers in the United States. Sev-eral papers in recent years have documentedactive HPV DNA in tumor samples fromoral cancer patients. In November, an epi-demiological analysis from the Centers forDisease Control and Prevention (CDC) in

Atlanta, Georgia, linked up to 60% oforopharyngeal cancer cases from 1998 and2003 to HPV. The link is ª as solid as it is forcervical cancer,º says pathologist LubomirTurek of the University of Iowa in Iowa City.

How significant HPV infection is as arisk factorÐ and what behaviors put some-one at risk for cancerÐ remain unclear. Butthe realization that HPV is behind a subsetof head and neck cancers is promptingresearchers and clinicians to reevaluate howthey diagnose and treat oral cancer, whichstrikes about 34,000 Americans each year.And it' s adding another facet to the pictureof the health threat of HPV as public healthofficials weigh expanding vaccination pro-grams. This year, Merck is seeking to get its

anti-HPV vaccine Gardasil, which wasapproved for women in 2006, approved foruse in men.

Viral connection

One of the first to publish suspicions thatHPV might cause oral cancer is dentist-turned-researcher Stina Syrjänen of theUniversity of Turku in Finland. She and hercolleagues started to examine oral cancertumors for HPV after virologist Harald zurHausen began isolating HPV DNA in cervi-cal cancer lesions in the early 1980s. TheSyrjänen group ' s initial surveys andimmunochemistry staining of 40 lesionspublished in 1983 turned up nearly a quarterthat were shaped like HPV lesions in cervi-cal cancer. But there wasn' t much interest inthis at the time, Syrjänen says.

Papers from other European labs trickledin documenting the presence of HPV anti-bodies in some oral cancer tumors. Onepublished in 1995 by R. D. Steenbergen andcolleagues at the Free University Hospitalin Amsterdam showed that HPV 16 wasintegrated and its onco proteins wereexpressed in an oral cancer tumor cell line.

Those f indings had a big impact onMaura Gillison, then a postdoc at JohnsHopkins University in Baltimore, Mary-land. ª I was pretty convinced from this onecaseº that the link between the virus andthis oral cancer was real, she says. Gillison,now a chair of cancer research at Ohio StateUniversity, Columbus, began looking formore evidence that the virus actually wastriggering oral cancers. Using PCR,sequencing, Southern blot assays, and insitu hybridization, Gillison and colleaguesat Johns Hopkins examined 253 head andneck tumors and identified HPV in 62 ofthem, as they reported in the 3 May 2000Journal of the National Cancer Institute.

Recent research on HPV and cervicalcancer suggests how the process works. Twospecif ic HPV genes, E6 and E7, turn offproteins in cervical cells that suppresstumor growth. In their 2000 study, Gillisonand colleagues examined these tumor sup-pressor proteins in head and neck cancersamples. Previous work revealed that thetumor suppressor proteins of HPV-negativeoral cancers often show mutations, likelycaused by long-term exposure to carcino-gens and alcohol. But the researchers foundfew such mutations in the HPV-positivesamples, so they argued that the onco-

High-risk group? The U.S. rate of oropharyngeal

cancer among people aged 20 to 49 rose sharply

between 1975 and 2005.

HPV Casts a Wider ShadowRecent studies link certain oral cancers to the virus that causes cervical cancer; some

researchers want to vaccinate both men and women against it

CANCER

Telltale linkage. Molecular studies have identified

strain 16 of the human papilloma virus in oral can-

cers, such as this tonsillar tumor.

Published by AAAS

...

......



CR

ED

IT: C

OU

RT

ESY

OF

OH

IO S

TA

TE

UN

IVE

RSIT

Y M

ED

ICA

L C

EN

TE

RNEWSFOCUS

proteins were disarming the cell' s tumorsuppressors, just as they do in cervical cancer. Over the next several years, labs inthe United States and in Europe confirmedthe findings. At the 2007 American Societyfor Clinical Oncology (ASCO) meeting inChicago, Illinois, Yale University researcherAmanda Psyrri reported that suppressing E6and E7 RNA in HPV-positive oral cancercells reactivated the cells' natural tumor sup-pression genes.

The molecular and pathological evi-dence that HPV, particularly HPV 16, cancause oral cancer is convincing, researcherssay. ª HPV-related head and neck cancerrepresents a new entity that is now well-def ined,º Dana-Farber Cancer Instituteoncologist Robert I. Haddad wrote to theOral Cancer Foundation after the 2007ASCO meeting.

Gillison and colleagues at Johns Hop-kins and NCI have been building up an epi-demiological picture of HPV-positive oralcancer. A May 2007 paper in The New Eng-

land Journal of Medicine (NEJM) exam-ined 100 cases of oropharyngeal cancer andfound that HPV infection could be a riskfactor independent of smoking or drinking history. In February 2008, the researchersreported in the Journal of Clinical Oncol-

ogy that U.S. National Institutes of Healthrecords show that rates of HPV-related oralcancers signif icantly increased between1973 and 2004 while rates of non-HPV oralcancers declined.

When the researchers interviewedpatients about their health behaviors, theyreported in the 19 March 2008 Journal of

the National Cancer Institute, two differentprof iles emerged. Risk factors for HPV-negative tumors included several decades ofpack-a-day smoking, years of heavy drink-ing, and losing teethÐ but not having moresex partners. In contrast, having more oralsex partners significantly increased the riskof developing HPV-positive cancer. Unex-pectedly, a history of heavy marijuanausage also seemed to increase risk for HPV-negative tumors in these patients.

Gillison says she has concluded thatHPV-positive oral cancer isn' t just a subsetof oral cancer but a completely independentdisease: ª If the risk factors are that differ-ent, what other evidence do you need?º

What you need, Turek and Syrjänen say,are more confirmatory studies. They agreethat Gillison' s team is on the right track indoing the larger, case-control studies of thepast 2 years. But questions remain. As Turekand Syrjänen have noted, Gillison' s groupused the anatomic site of some tumors

(around the tonsils, where HPV-positivetumors overwhelmingly occur) as a proxyfor evidence of HPV' s presence. Gillisonsays her team is now doing a molecularanalysis of these tumor samples and expectsresults this year.

Large epidemiological studies couldhelp clinch the risk-factor profile. StephenSchwartz, an epidemiologist at the Univer-sity of Washington, Seattle, who studiesHPV-related cancers, says he' s skeptical of aproposed link between HPV-positive oralcancer and marijuana usage, because onlyone study has demonstrated it so far. But theevidence that disease risk increases withmore sexual partners seems stronger, hav-

ing been suggested in several studies byindependent researchers, he says.

One question all the researchers agreeremains unresolved is the natural history ofHPV-positive oral cancer. Researchers knowit can take 20 to 30 years after HPV infectionfor cervical cancer to develop in an otherwisehealthy woman. They don' t know how long ittakes oral cancer to develop or how much ofthe virus must be present in oral cells to trig-ger cancer, Turek says. A single study of 292HPV-positive oral cancer patients publishedin 2001 in NEJM suggests it might take onlya decade, but researchers agree that studiesare needed to conf irm that f inding andexplain why oral cancer might develop so

much faster than cervical cancer. Gillison argues that researchers don' t

need to wait for al l questions to beresolved before reassessing how they diag-nose and treat HPV-positive oral cancerpatients. In November, Gillison pressedher case to more than 80 head and neckcancer experts at a closed meeting at NCIin Bethesda, Maryland. She and otherresearchers, including Turek, believe thestandard treatment protocol for oral cancerÐ chemotherapy and radiation, fol-lowed by possible surgeryÐ may be tooaggressive for patients with HPV-positivetumors. Researchers and clinicians saythey have noticed that these patients tendto respond much better to treatment,specif ically radiation: About 85% ofpatients with HPV-positive tumors are stillalive within 5 years of their cancer diagno-sis, compared with about 45% of thosewith non-HPV tumors. So radiation alonemight be enough to treat some of thesepatients, Gillison argues.

Public health officials, meanwhile, areasking what can be done to prevent the dis-ease. CDC has approved four vaccinesagainst HPV for use in young women toprevent cervical cancer; now the question iswhether to expand the recommendations toinclude young men. Preliminary resultsfrom a Merck-funded study found that Gar-dasil effectively prevented HPV infection inthe genitals of men, says study leader JoelPalefsky of the University of California,San Francisco. Full results are expectedsome time this spring, he says. NCI epi-demiologists are also conducting clinicaltrials to assess whether the vaccine can pre-vent oral HPV infection. It may be yearsbefore those results come in, says studyleader Aimee Kreimer.

Well before that, CDC' s Advisory Com-mittee on Immunization Practices (ACIP)hopes to give its own recommendations onwhether men should get vaccinated for HPVand if so, at what age, says CDC medical epi-demiologist Lauri Markowitz. She saysACIP is currently reviewing data from theMerck trials, along with studies on HPV andoral cancer. Schwartz says he thinks the vac-cine will be approved for use in men soon,before researchers fill in the lingering gapsin knowledge about HPV and oral cancer orhow effectively vaccination programs mightprevent disease. That means taking a bit onfaith, he says: ª I don' t have any real reason tothink it won' t work, but we have to acknowl-edge we' re just taking more of a leap here.º

± RACHEL ZELKOWITZ

Rachel Zelkowitz is a writer in Washington, D.C.

"I was pretty convinced fromº

a 1995 study that the link

between the virus and this

oral cancer was real.Ð MAURA GILLISON, OHIO STATE UNIVERSITY

Published by AAAS

...

......



CR

ED

IT: JO

E S

UT

LIF

F

LETTERS I BOOKS I POLICY FORUM I EDUCATION FORUM I PERSPECTIVES

584

Cellular force mechanics

588

Broadcasting science

LETTERS

Systems Politics and Political Systems

IN HIS EDITORIAL ª A SCIENTIFIC APPROACH TO POLICYº (5 DECEMBER 2008, P. 1435), B. ALBERTSwrote that when dealing with complex political issues, policy-makers would benefit fromapplying the experiment-based, problem-solving approach of scientists and engineers.

The suggestion that the complexity of intercellular and subcellular interactions is equivalentin a sense to political interchange holds true at many levels. Indeed, many parallels have been

made between politics and systemic biol-ogy. For example, autoimmunity resultsfrom a misguided national defense (im-mune) force attacking its own subjects(cells), and regulatory steps of a govern-mental construction policy mirror the com-munication, trafficking, and machinerycontrol involved in cell polarization.

These comparisons can work both ways,and therefore it is incumbent on scientists tolearn from policy-makers as well. Leader-ship, management, public engagement, andeconomic awareness are all characteristicsthat can enhance the phenotype of the nextgeneration of scientists.

HUTAN ASHRAFIAN

Department of Biosurgery and Surgical Technology, Imperial College London, London W2 1NY, UK. E-mail:[email protected]

edited by Jennifer Sills

Scientists Not Immune

to Partisanship

THE EDITORIAL BY B. ALBERTS (ª A SCIENTIFICapproach to policy,º 5 December 2008, p.1435) persists in promoting a long-refutednotion that scientists constitute an apoliticalelite that will, by bypassing partisanship andideology, improve policy-making. Althoughthe three ª scientific habits of mindº that hedescribesÐ optimism, long-term thinking,and pragmatismÐ are admirable virtues, theyare not exclusive to members of the scientificcommunity, nor universal among scientists,nor guarantees of wise action. Who is todecide the ends to which such virtues areapplied? A can-do attitude and a pragmaticapproach to realizing a long-term goal of

dubious value gets society nowhere (orworse). Indeed, some scientists (in their opti-mism about getting rich) applied these veryqualities to the development of financial mod-els that have contributed to the current eco-nomic meltdown that Alberts blames entirelyon the unrealistic assumptions of financialmanagers (1). David Halberstam (2) docu-mented the disastrous application of similarfaith among such ª nonideological policy-makersº in the conduct of the Vietnam War.The obvious point is that the application ofour scientific knowledge is inseparable fromthe subjective contexts in which it is applied.Science and technology bring not only won-derful benefits, but also challenges and risks,from threats to personal and national security,to skewed distribution of wealth and socialcapital, to environmental and cultural degra-

dation. While we may be pleased that theprospects of a new administration in Wash-ington will correct the recent era of hyperpar-tisanship, we note that in the 1990s it was con-servatives who were calling for more ª soundscienceº in policy matters. Science will con-tinue to be used by players at all positionswithin the political spectrum as a cover forpolitical agendas until we abandon the doc-trine that Alberts insists on perpetuating. Theclaim that science and values can be kept apartin the policy world confuses the means of sci-ence with the ends of democracyÐ a confu-sion that is dangerous for the health of both.

DAVID H. GUSTON,* DANIEL SAREWITZ,

CLARK MILLER

Consortium for Science, Policy, and Outcomes, ArizonaState University, Tempe, AZ 85287, USA.

*To whom correspondence should be addressed. E-mail:[email protected]

References 1. J. Nocera, ª Risk Management,º The New York Times

Magazine (4 January 2008), pp. 4± 33.2. D. Halberstam, The Best and the Brightest (Random

House, New York, 1969).

Law and Science Not

Mutually Exclusive

AS SOMEONE WITH DEGREES IN BOTH PHYSICSand law, I agreed with the Editorial ª A callto serveº by W. A. Wulf and A. K. Jones (14November 2008, p. 1025) until I reached thefifth paragraph, in which the authors seemedto suggest that lawyers think differently fromscientists and engineers. Do the authors thinkthat those lawyers who also have science andengineering degrees were rendered incapableof thinking as scientists and engineers oncethey received their law degrees?

The authors also suggest that lawyers ª dis-proportionately populate government posi-tions.º What do the authors believe the correctproportion to be? Many people get law de-grees with the intent of public service. I sus-pect that an understanding of law makes one amore effective public servant, just as knowl-edge of science helps somebody who serveson the National Science Foundation or on theNational Science Board. Are the authors

COMMENTARY

Published by AAAS

...

......


willing to suggest that those positions are dis-proportionately taken up by people with back-grounds in science and engineering?

The authors suggest that scientists andengineers are ª trainedº to follow whereverthe evidence leads. I disagree that majoringin science or engineering is the only way tolearn how to consider a problem from allangles. The authors have set up a falsedichotomyÐ being trained to form an argu-ment that best supports a given conclusiondoes not preclude the ability to examine asituation dispassionately and reach the cor-rect conclusion. Certainly, there arelawyers who are unqualified to make policydecisions on scientific issues, both becausethey lack basic scientific understandingand because they have other interestsbesides promoting the best available poli-cies, but the authors have made unwar-ranted (and insulting) generalizationsabout lawyers.

DONALD D. DEROSIER

Carmichael, CA 95608, USA. E-mail: [email protected]

Credit for Coauthors

IN THE LETTER ª QUANTIFYING COAUTHORcontributionsº (17 October 2008, p. 371), C.H. Sekercioglu' s proposal that the kth rankedcoauthor be considered to contribute 1/k asmuch as the first author is not novel. It wasoriginally made in 1981 in a letter to Science

by Susan E. Hodge and David A. Greenbergtitled ª Publication creditº (1).

Hodge and Greenberg were responding toa plea from Derek De Solla Price for dividingauthorship credit equally among all coauthors

(2). They detailed the advantages of theirscheme and proposed the formula: Points =[(1/i)/(1 + (1/2) + ¼ + (1/N)] ´ 100for the ith author out of N. The denominator inthe formula is the well-known harmonicseries; and the points are standardized to 100per paper. Apart from the standardization, thisis exactly what Sekercioglu proposes.

Sekercioglu' s Letter opens by citing anever-cited 40-year-old contribution to Science

that argued that more than three authors onone paper is not justifiable. It is interestingto note that both Price and Hodge andGreenberg proposed division of authorshipcredit ª ¼ to discourage putting many authorson a single paperº (1).

Authorship trends have not progressed asPrice and Hodge and Greenberg had hoped.Sekercioglu' s letter highlights the need for acritical reappraisal of the way modern bib-liometry allocates publication credit (3).

NILS T. HAGEN

Department of Biosciences and Aquaculture, Bodù UniversityCollege, N-8049 Bodù , Norway. E-mail: [email protected]

References1. S. E. Hodge, D. A. Greenberg, Science 213, 950 (1981).2. D. D. S. Price, Science 212, 986 (1981).3. N. T. Hagen, PLos ONE 3, e4021 (2008).

ResponseI THANK HAGEN FOR POINTING OUT A CRITI-cal reference (1) that I inadvertently omitted.A key difference in my formulation is thatauthor rank can be independent from authororder. This is essential in many cases wherethe last author is the senior author or wheresome authors may have contributed equally tothe publication. Hodge and Greenberg' s pro-posal does not allow for this. They state thatª the first author always receives twice asmany points as the secondº and they havedeliberately not allowed for the last authorbeing the senior author, as they ª do not wishto encourage this pernicious habit.º

We need to encourage the conversation onquantifying coauthor credit so that proposalsare rigorously debated, improved, and decidedupon by the scientific community. Otherwise,we may still be having this discussion threedecades from now.

CAGAN H. SEKERCIOGLU

Center for Conservation Biology, Department of Biology,Stanford University, Stanford, CA 94305, USA. E-mail:[email protected]

Reference1. S. E. Hodge, D. A. Greenberg, Science 213, 950 (1981).


Plant pores

592

Explaining locust behavior

594

Letters to the EditorLetters (~300 words) discuss material published in Science in the previous 3 months or issues ofgeneral interest. They can be submitted throughthe Web (www.submit2science.org) or by regularmail (1200 New York Ave., NW, Washington, DC20005, USA). Letters are not acknowledged uponreceipt, nor are authors generally consulted beforepublication. Whether published in full or in part,letters are subject to editing for clarity and space.

CORRECTIONS AND CLARIFICATIONS

Reports: ª A null mutation in human APOC3 confers a favorable plasma lipid profile and apparent cardioprotectionº by T. I.Pollin et al. (12 December 2008, p. 1702). On page 1703, the exon was mischaracterized. The second sentence in thesecond full paragraph of the first column should read as follows: Sequencing of the coding region of APOC3 revealed a C → Tsubstitution at the terminal nucleotide of exon 2, the 55th nucleotide from the ATG start codon; this substitution resulted ina premature stop codon for an arginine residue at amino acid position 19 (R19X).

TECHNICAL COMMENT ABSTRACTS

COMMENT ON ª Arsenic (III) Fuels Anoxygenic Photosynthesis in Hot SpringBiofilms from Mono Lake, Californiaº

B. Schoepp-Cothenet, S. Duval, J. M. Santini, W. Nitschke

Kulp et al. (Reports, 15 August 2008, p. 967) described a bacterium able to photosynthetically oxidize arsenite[As(III)] via arsenate [As(V)] reductase functioning in reverse. Based on their phylogenetic analysis of As(V) reduc-tase, they proposed that this enzyme was responsible for the anaerobic oxidation of As(III) in the Archean. We chal-lenge this proposition based on paleogeochemical, bioenergetic, and phylogenetic arguments.

Full text at www.sciencemag.org/cgi/content/full/323/5914/583c

RESPONSE TO COMMENT ON ª Arsenic (III) Fuels Anoxygenic Photosynthesis in HotSpring Biofilms from Mono Lake, Californiaº

R. S. Oremland, J. F. Stolz, M. Madigan, J. T. Hollibaugh, T. R. Kulp, S. E. Hoeft, J. Fisher, L. G.

Miller, C. W. Culbertson, M. Asao

Schoepp-Cothenet et al. bring a welcome conceptual debate to the question of which came first in the course ofplanetary biological evolution, arsenite [As(III)] oxidation or dissimilatory arsenate [As(V)] reduction. However, wedisagree with their reasoning and stand by our original conclusion.

Full text at www.sciencemag.org/cgi/content/full/323/5914/583d

Published by AAAS

...

......


Comment on “Arsenic (III) FuelsAnoxygenic Photosynthesis in Hot SpringBiofilms from Mono Lake, California”B. Schoepp-Cothenet,1* S. Duval,1 J. M. Santini,2 W. Nitschke1

Kulp et al. (Reports, 15 August 2008, p. 967) described a bacterium able to photosyntheticallyoxidize arsenite [As(III)] via arsenate [As(V)] reductase functioning in reverse. Based on theirphylogenetic analysis of As(V) reductase, they proposed that this enzyme was responsible for theanaerobic oxidation of As(III) in the Archean. We challenge this proposition based onpaleogeochemical, bioenergetic, and phylogenetic arguments.

The report by Kulp et al. (1) demonstratesthe ability of anoxygenic photosyntheticbacteria to use reduced arsenic compounds

as electron-donating substrates. As pointed outby the authors, this result provides a possible so-lution to a paleogeochemical/phylogenetic conun-drum. Phylogenetic analyses of relevant proteinshave been interpreted to indicate ancient (i.e.,pre-Archaea/Bacteria divergence) roots for thebioenergetic mechanism of AsIII oxidation (2).By contrast, the ultimate electron-accepting sub-strate observed so far in most extant arsenite ox-idizers, that is, O2, is unlikely to have been presenton the early Earth. We have recently suggested

nitrogen oxides as possible alternatives to O2

playing the role of strong oxidants in the earlyArchean (3). The existence of organisms feedingelectrons from the oxidation of As(III) into ananoxygenic photosynthetic chain constitutes an-other solution to this nagging contradiction.

The observation of photosynthetic As(III) ox-idation per se substantially advances our under-standing of a possible bioenergetic role for As(III)in the Archean era. Kulp et al. (1), however, fur-ther conclude that the existence of a biologicalpathway generating the oxidized element, arse-nate (AsV), “requires that the antiquity of pro-karyotic arsenate respiration be reevaluated….” Itwas indeed assumed previously that dissimilatoryarsenate reduction could only have evolved afterthe increase in oxidation level of the environmentbrought about by oxygenic photosynthesis (4).

We challenge this conclusion by Kulp et al.based on the following paleogeochemical, bio-

energetic, and phylogenetic arguments. The abun-dance of reduced compounds, and especially ofhigh concentrations of Fe(II), in the Archeanbiosphere provides a powerful chemical reduc-tant for the photosynthetically produced As(V),keeping the total bioavailable amount of As(V)at very low levels. This also implies that the As(III)/As(V) ratio has remained high despite the photo-synthetic production of As(V). The resultingoperating potential of the As(III)/As(V) couple istherefore not likely to substantially exceed 0 mV(versus standard hydrogen electrode), renderingit bioenergetically useful only as an acceptor forrather low potential substrates. Finally, ourphylogenetic analysis of arsenate reductase (5)suggests a late origin for this enzyme followedby subsequent dissemination by horizontal genetransfer. In this context, we note that theoutgroup position of the branch in figure 3D in(1) labeled Pyrobaculum aerophilum andconsidered to represent the ArrA protein of thisArchaeon, would indeed support the authors’conclusion on an early origin of dissimilatoryarsenate reductase. However, several argumentsindicate this enzyme to be a polysulphide/thiosul-phate reductase rather than an arsenate reductase(5). In summary, we conclude that the presentlyavailable evidence argues against a role ofdissimilatory arsenate reductase in arsenic cy-cling on the early Earth.

References1. T. R. Kulp et al., Science 321, 967 (2008).2. E. Lebrun et al., Mol. Biol. Evol. 20, 686 (2003).3. A.-L. Ducluzeau et al., Trends Biochem. Sci. 34, 9 (2009).4. R. S. Oremland, J. F. Stolz, Science 300, 939 (2003).5. S. Duval et al., BMC Evol. Biol. 8, 206 (2008).

21 August 2008; accepted 31 December 200810.1126/science.1164967

TECHNICALCOMMENT

1Bioénergétique et Ingénierie des Protéines, IFR88, CNRS, 13402Marseille, France. 2Department of Structural and Molecular Bi-ology, University College London, London WC1E 6BT, UK.


www.sciencemag.org SCIENCE VOL 323 30 JANUARY 2009 583c

...

......


Response toComment on “Arsenic(III) FuelsAnoxygenic Photosynthesis in Hot SpringBiofilms from Mono Lake, California”R. S. Oremland,1* J. F. Stolz,2 M. Madigan,3 J. T. Hollibaugh,4 T. R. Kulp,1 S. E. Hoeft,1J. Fisher,4† L. G. Miller,1 C. W. Culbertson,5 M. Asao3

Schoepp-Cothenet et al. bring a welcome conceptual debate to the question of whichcame first in the course of planetary biological evolution, arsenite [As(III)] oxidation ordissimilatory arsenate [As(V)] reduction. However, we disagree with their reasoning andstand by our original conclusion.

Schoepp-Cothenet et al. (1) challenge ourconclusions regarding dissimilatory arse-nic reduction on primordial Earth (2). The

authors argue their points based on suppositionsconcerning the redox state of the Archean, andtheir entrained bioenergetic/phylogenetic reason-ing. With regard to their point about redox state,although Fe(II) is believed to have been abun-

dant in the anoxic Archean, this chemical spe-cies does not reduce As(V) to As(III). Free sulfidecan reduce As(V), but this requires a very low pHfor any reasonable kinetic resolution (3). In ad-dition, the production of As(V) by As(III)-linkedanoxygenic phototrophy would have openedlocalized niches for As(V) respiration withoutaffecting the overall redox balance of the ArcheanEarth. This is similar to what has been proposedfor photosynthetically linked Fe(II) oxidationand banded iron formations (4). Indeed, the tightcoupling of As(V) reduction/As(III) oxidationhas been seen in periphyton samples (5) and salt-saturated anoxic sediments from Searles Lake(6). We have also observed this phenomenon in“red” hot spring biofilms from Mono Lake aspart of an ongoing study. Finally, although in silicophylogenetic analyses of genomes can providegrist for evolutionary schemes, their validity is

dependent on correct annotation. Unless proofof function is known for a given homolog, bio-chemical or genetic confirmation is necessary(7). For example, the dimethyl sulfoxide reduc-tase family of molybdenum enzymes is composedof oxidases, reductases, and dehydrogenases, andenzymes sharing high sequence identity can havevery different functions (8). Moreover, arsenaterespiration has been demonstrated in two Pyro-baculum species; however, the respiratory arse-nate reductase in these organisms has yet to beidentified. Investigations of the arrA ampliconsobtained from the extreme environment of SearlesLake, California, suggest unexpectedly high se-quence diversity, with most lying well outside theboundaries established with cultured anaerobesfrom the domain Bacteria (9). Such observationsindicate that there are many more “novel” prokary-otes out there (including Archaea) with the ca-pacity to respire As(V), awaiting isolation andcharacterization and, ultimately, genomic analysis.

References1. B. Schoepp-Cothenet, S. Duval, J. M. Santini, W. Nitschke,

Science 323, 583 (2009); www.sciencemag.org/cgi/content/full/323/5914/583d.

2. T. R. Kulp et al., Science 321, 967 (2008).3. J. A. Cherry et al., J. Hydrol. (Amst.) 43, 373 (1979).4. A. Ehrenreich, F. Widdell, Appl. Environ. Microbiol. 60,

4517 (1994).5. T. R. Kulp et al., Appl. Environ. Microbiol. 70, 6428

(2004).6. R. S. Oremland et al., Science 308, 1305 (2005).7. R. S. Oremland et al., Nat. Rev. Microbiol. 3, 572 (2005).8. J. F. Stolz et al., Annu. Rev. Microbiol. 60, 107 (2006).9. T. R. Kulp et al., Appl. Environ. Microbiol. 72, 6514

(2006).

25 September 2008; accepted 31 December 200810.1126/science.1166435

TECHNICALCOMMENT

1U.S. Geological Survey (USGS), Menlo Park, CA 94025, USA.2Department of Biological Sciences, Duquesne University,Pittsburgh, PA 15282, USA. 3Department of Microbiology,Southern Illinois University, Carbondale, IL 62901–6508, USA.4Department of Marine Sciences, University of Georgia,Athens, GA 30602–3636, USA. 5USGS Water Sciences Center,Augusta, ME 04330, USA.

*To whom correspondence should be addressed. E-mail:[email protected]†Present address: Division of Earth and Ecosystem Science,Desert Research Institute, Las Vegas, NV 89119, USA.

www.sciencemag.org SCIENCE VOL 323 30 JANUARY 2009 583d

...

......



BOOKS ET AL.

Writing a history of print journalismis a fairly straightforward process:microform and now digital collec-

tions have made even obscure, short-livedpublications widely available. But radio?That' s a different story. There is no such thingas a comprehensive Reader' s Guide to

Periodic Literature for radio, and what listingsare available tell little about what exactly hap-pened on the shows. For science journalism,the problem is compounded by the fact thatscience programming rarely made it onto thenetworks, appearing instead as filler onregional stations. And yet, in the fascinatingScience on the Air, Marcel LaFollette man-ages to suss out most of the major players inscience popularization in the first threedecades of broadcast journalism. It is aremarkable achievement.

Ironically, LaFollette' s task was madesomewhat easier by the rarity of scientific pro-gramming in the golden age of radio. Thebook focuses on the story of how Americanscientific institutions failed to grasp the powerof broadcast media to shape public attitudesabout and understanding of science. Their fewforays into the medium were marked by a pro-found unwillingness to take the concerns ofaudiences, advertisers, and network execu-tives seriously. The few examples of commer-cially successful science programming thatLaFollette, an independent historian, unearthswere mostly developed by corporate interests.

It didn' t have to be this way. The storyopens in 1923 with a phone call from a stationmanager at WRC, a Washington, DC± basedaffiliate of RCA, to the secretary of theSmithsonian Institution, Charles D. Walcott.Like many of his peers, this station managerwas open to suggestions on how to distinguishhis station' s offerings from the typical fare ofsermons, jazz bands, and sports. The call waspassed along to Austin Clark, an ambitiouscurator and invertebrate biologist, who wassoon scheduling lectures by Smithsonian sci-entists on such topics as ª Creatures That Flyand How They Do Itº and ª Animal Terrors ofPast Ages.º

For Clark and his colleagues, radio was anextension of the Smithsonian' s public lecture

programming. Many of thescientists who appeared onthe show did not own radiosets, had never listened toradio programming, anddid not understand that theairwaves were mostly dom-inated by musical perform-ances. Clark further insistedthat his guests don tuxedosat the recording booth inkeeping with the dignity of science. This levelof stodginess was acceptable in 1923, whenmuch of the radio dial was occupied byearnest, amateurish programming, but fell outof favor during the wave of consolidation thatsoon affected the industry. Whereas stationmanagers defined a successful program by thenumber of listeners (and potential advertisingrate), officials at the Smithsonian judged theshow by the caliber of the guests and Clark' sability to get the lectures published as pam-phlets or magazine articles. WRC canceledRadio Talks in 1927.

LaFollette tells a nearly identical story forprograms developed by the Franklin Institute,the American Museum of Natural History,and a few entrepreneurial scientists, includingnaturalist Thornton Burgess and Harvardastronomer Harlow Shapley. The American

Psychological Association and the AmericanMedical Association were slightly more suc-cessful in developing programs that could sus-tain listener interest in the 1930s, largely

because they were forced torespond to the charismatic huck-sters who peddled quack remediesover the airwaves. All of theseorganizations emphasized educa-tion over entertainment. The lonevoice advocating a more popularapproach belonged to WatsonDavis, first as managing editor andlater as director of ScienceService, Incorporated, a news syn-dication service that produced along-running audio news digestcalled Science Service Talks. That

Davis' s contributionsÐ theme music, intro-ductions by nonscientists, interviews insteadof straight lecturesÐ were seen as radicalgives some indication of how tedious earlyscience radio programming must have been.

Innovation in science radio was instead leftto the corporate giants who had the necessaryresources both to produce quality program-ming and to purchase air time. General Electric,Westinghouse, and, especially, DuPont createdlavish radio productions that incorporateddramatization, fictionalization, and quiz showsinto scientific programming. DuPont producednearly 800 episodes of its Cavalcade of

America for radio, followed by an additional200 for television before finally taking it off theair in 1957. The critical acclaim and popularsuccess of this program finally convinced thestaid American scientific establishment to

Faulty TransmissionAudra J. Wolfe

SCIENCE COMMUNICATION

The reviewer is at the Chemical Heritage Foundation, 315Chestnut Street, Philadelphia, PA 19106, USA. E-mail:[email protected]

Science on the Air

Popularizers and

Personalities on Radio

and Early Television

by Marcel Chotkowski

LaFollette

University of Chicago

Press, Chicago, 2008.

324 pp. $$27.50, £14.50.

ISBN 9780226467597.

Advocate for popularization. Beginning in the 1920s, Science Service journalist Watson Davis (right) useda series of radio programs to deliver news about science and scientists.

CR

ED

IT: SM

ITH

SO

NIA

N IN

ST

ITU

TIO

N A

RC

HIV

ES, SC

IEN

CE

SE

RV

ICE

RE

CO

RD

S, N

EG

AT

IVE

SIA

20

08

-18

17

Published by AAAS

...

......


experiment with broadcasting from planetari-ums and scientific expeditions, but it was toolittle, too late. By then, the airwaves were beingcontrolled by the national networks, and thenetworks' gatekeepers had decided that sciencecouldn' t pay its own way.

The pattern would be repeated in the earlyyears of television, where complex productionrequirements, expensive sets, and demandingvisuals left little room for amateurish produc-tions. As on the radio, the most creative usesof the medium for science popularizationcame from media behemoths like ABC andDisney rather than scientific institutions.

LaFollette minces no words in referring tothe ª imaginative failuresº of the Americanscientific establishment in making use ofmass media technologies. Commenting on aneffort to increase the amount of educationalprogramming on 1930s-era radio, for exam-ple, she argues that the scientific organiza-tions' ª inability ¼ to cooperate with eachother, their intellectual snobbery and undis-guised disdain for the very medium they weresupposedly trying to utilize, and their unwill-ingness to invest significant resources in pro-duction of quality programs all hobbled theireffectiveness.º LaFollette' s extensive use ofarchival collections, including meeting min-utes and personal correspondence, offersample support for her damning conclusions.Her criticism is bracing but fair. Now, sheargues, podcasts, online video-sharing sites,and blogs are giving scientists another oppor-tunity to communicate directly with the pub-lic. The question is: Will they take it?

10.1126/science.1168311

NEUROSCIENCE

Who Are We?Ralph Adolphs

It is blatantly apparent that humans arequite different from all other animals, inways both good and bad. But articulating

what that difference consists of and uncover-ing the biology behind it have formed a largeand difficult project tackled by biologists,anthropologists, psychologists, and philoso-phers. Michael Gazzaniga' s Human: The

Science Behind What Makes Us Unique pro-vides a masterful overview of what we know,who the key players are, and what the futuremight hold. The book is at once dense with

facts (there are 748 endnotes) and an easyread; it is both entertaining and informative.

Gazzaniga, a professor of psychology atthe University of California, Santa Barbara, isthe director of the SAGE Center for the Studyof the Mind at UCSB. There he has beenabsorbing the views of scholars in residencewho visit for weeks to months, a rich source ofinformation reflected in the book' s contents.Human is organized into four parts: The firstdescribes some of the genetic, cellular, andbehavioral ways in which humans may beunique and sets the stage for the subsequentsections, which emphasize cognition and thebrain. The second and third parts treat topicsranging from morality to empathy to art andconclude with two chapters discussing con-sciousness. The fourth part provides a view tothe future, exploring such possibilities asbrain-machine interfaces and artificial intelli-gence. Roughly speaking, the book becomesmore speculative, and tackles tougher topics,the further on one reads.

Although the majority of topics thatGazzaniga discusses in the book are con-tentious, many of them highlyso, his treatment of them isscholarly and balanced. Nogrand conclusions are drawn,and he does not offer specifictheories of his own so much assurvey those of others. Thebook is also highly accessible,quite a feat given its scope anddensity. Brief interludes pro-vide the science backgroundfor those who need it. For in-stance, the last chapter includes, sandwichedbetween discussions of fyborgs (functionalcyborgs) and cyborgs, a wonderful 5-pageintroduction to cellular neurophysiology tohelp readers understand what follows. Eachchapter also ends with a brief conclusion,which summarizes the main points. Theauthor is clearly someone who has writtentextbooks and knows how to teach.

Perhaps not surprisingly, Gazzaniga favorsa modular view of the mind informed by evo-lutionary psychology, a view he previouslyadvocated in The Social Brain (1) and one inline with people he frequently cites (StevenPinker, Leda Cosmides, John Tooby; the lattertwo colleagues of his at UCSB). Modulesmake fun reading and serve to quantize theexposition. But how do they work together?The penultimate chapter tackles the issue ofhow various modules coordinate their activityand selectively contribute to the contents ofour integrated conscious experience. It beginswith a brief reminiscence of how the authoronce lost consciousness as an intoxicated col-

lege student (he was at the time a member ofthe original Animal House at Dartmouth) andquickly moves on to survey some of the mostpopular theories and review his own. In one ofhis early books [(2), co-authored with his stu-dent Joseph LeDoux], he introduced readersto ª the Interpreterº : a left-hemisphere mecha-nism that analyzes our actions, integrates theoutputs from many modules, and generates anarrative that constitutes our stream of con-scious experience. Yet Gazzaniga clearlynotes that our ability to consciously experi-ence the world is shared with many other ani-mals, even though the nature of human con-sciousness may be unique.

So what are the aspects of cognition thatmake humans unique? Gazzaniga considers along list: control over our thoughts, emotions,and actions; planning into the future; self-reflection and self-consciousness; language;aspects of imitation and social learning;episodic memory; imagination; creativity;cooperation and altruism; theory of mind;and many more. Trying to find a single themethat ties all these together or subsumes them

is daunting. Nor is it clearwhether these are differencesin degree or in kind. Gazzanigadoes not attempt an answerhere, although he hints at onein the book' s short Afterword:ª Just like other animals, we areconstrained by our biology. ¼But the ability to wish or imag-ine that we can be better is not-able. No other species aspiresto be more than it is. Perhaps

we can be.º This ability to step outside of our-selvesÐ to view the world and us in it fromarbitrarily abstract perspectivesÐ does indeedseem uniquely human, and it cuts across manyof the specific abilities in the above list. Itmakes us responsible for our actions and inac-tions in a way that other animals are not. It isour burden and yet offers hope for our speciesand our planet.

Human delivers what the best popular sci-ence writing should. Gazzaniga tackles themost difficult questions, provides an expertsurvey of the field, and, most important,instills a sense of wonder and enjoyment aboutthe subject matter. Lay readers and young sci-entists alike should benefit, and perhaps ourspecies will too.

References

1. M. S. Gazzaniga, The Social Brain: Discovering the

Networks of the Mind (Basic, New York, 1985).

2. M. S. Gazzaniga, J. E. LeDoux, The Integrated Mind

(Plenum, New York, 1978).

10.1126/science.1169620


BOOKSETAL.

The reviewer is in the Division of Humanities and SocialSciences 228-77, California Institute of Technology, 1200East California Avenue, Pasadena, CA 91125, USA. E-mail:[email protected]

Human

The Science Behind

What Makes Us Unique

by Michael S. Gazzaniga

Ecco (HarperCollins),

New York, 2008. 461 pp.

$27.50, C$29.50, £14.99.

ISBN 9780060892883.

Published by AAAS

...

......



EDUCATIONFORUM

The development of general scien-tif ic abilities is critical to enablestudents of science, technology,

engineering, and mathematics (STEM) tosuccessfully handle open-ended real-worldtasks in future careers (1 ± 6). Teachinggoals in STEM education include fosteringcontent knowledge and developing generalscientific abilities. One such ability, scien-tif ic reasoning (7 ± 9), is related to cogni-tive abilities such as critical thinking andreasoning (10 ± 14). Scientif ic-reasoningskills can be developed through trainingand can be transferred (7, 13). Training inscientific reasoning may also have a long-term impact on student academic achieve-ment (7). The STEM education communityconsiders that transferable general abilitiesare at least as important for students tolearn as is the STEM content knowledge(1 ± 4). Parents consider science and mathe-matics to be important in developing rea-soning skills (15).

We therefore asked whether learningSTEM content knowledge does in fact havean impact on the development of scientific-reasoning ability. The scientific-reasoningability studied in this paper focuses ondomain-general reasoning skills such as theabilities to systematically explore a prob-lem, to formulate and test hypotheses, tomanipulate and isolate variables, and toobserve and evaluate the consequences.

Research Design

Students in China and the United States gothrough very different curricula in scienceand mathematics during their kindergartenthrough 12th grade (K± 12) school years.This provides systemically controlled long-term variation on STEM content learning,which we used to study whether or not such

learning has any impact on the developmentof scientific-reasoning ability. Scientificreasoning is not explicitly taught in schoolsin either country.

In China, K± 12 education is dominatedby the nationwide college admission examgiven at the end of grade 12. To comply withthe requirements of this exam, all Chinese

schools adhere to a national standard withinall courses. In physics, for example, everystudent goes through the same physicscourses, which start in grade 8 and continueevery semester through grade 12, providing5 years of continuous training on introduc-tory physics topics (16). The courses arealgebra-based with emphasis on develop-ment of conceptual understanding and skillsneeded to solve problems.

In contrast, K± 12 physics education inthe United States is more varied. Althoughstudents study physics-related topics withinother general science courses, only one ofthree high school students enrolls in a two-semester physics course (17). As a result,the amount of instructional time and theamount of emphasis on conceptual physics

understanding and problem-solving skillsare very different in the two countries.Similar curriculum differences between theUnited States and China are reflected inother STEM areas such as chemistry, biol-ogy, and mathematics (16).

Chinese students go through rigorousproblem-solving instruction in all STEM

subject areas throughout most of theirK± 12 school years and become skillful atsolving content-based problems. It remainsunclear, however, whether this training istransferable beyond the specific contentareas and problem types taught.

We used quantitative assessment instru-ments (described below) to compare U.S.and Chinese students' conceptual under-standing in physics and general scientific-reasoning ability. Physics content was cho-sen because the subject is conceptually andlogically sophisticated and is commonlyemphasized in science education (15).Assessment data were collected from bothChinese and U.S. freshmen college studentsbefore college-level physics instruction. Inthis way the data reflect students' knowledge

Comparisons of Chinese and U.S. students

show that content knowledge and reasoning

skills diverge.

Learning and Scientific ReasoningLei Bao,1* Tianfan Cai,2 Kathy Koenig,3 Kai Fang,4 Jing Han,1 Jing Wang,1 Qing Liu,1

Lin Ding,1 Lili Cui,5 Ying Luo,6 Yufeng Wang,2 Lieming Li,7 Nianle Wu7

PHYSICS

1Department of Physics, The Ohio State University,Columbus, OH 43210, USA. 2Department of Physics,Beijing Jiaotong University, Beijing 100044, China.3Department of Physics, Wright State University, Dayton,OH 45435, USA. 4Department of Physics, TongjiUniversity, Shanghai 200092, China. 5Department ofPhysics, University of Maryland, Baltimore County,Baltimore, MD 21250, USA. 6Department of Physics,Beijing Normal University, Beijing 100875, China.7Department of Physics, Tsinghua University, Beijing100084, China.

*Author for correspondence. E-mail: [email protected]

0%

5%

10%

15%

0 5 10 15 20

Popula

tion p

erc

enta

ge

Score

Scientific ReasoningUSA

China

0%

5%

10%

15%

20%

0 5 10 15 20 25 30

Popula

tion p

erc

enta

ge

Score

MechanicsUSA

China

0%

5%

10%

15%

20%

0 5 10 15 20 25 30

Popula

tion p

erc

enta

ge

Score

Electricity & Magnetism

USA

China

TEST SCORES (%)

TestChina

(n)

FCI 1.98

BEMA 3.53

LCTSR 0.03

49.3 ± 19.3

(2681)

26.6 ± 10.0

(650)

74.2 ± 18.0

(1061)

85.9 ± 13.9

(523)

65.6 ± 12.8

(331)

74.7 ± 15.8

(370)

USA

(n)

Effect

size

Content knowledge and reasoning skills diverge. Comparisons of U.S. and Chinese freshmen college

students show differences on tests of physics content knowledge but not on tests of scientific reasoning.

Published by AAAS

...

......


and skill development from their formal andinformal K± 12 education experiences.

Data Collection and Analysis

From the early 1980s, researchers and edu-cators in psychology and cognitive science(11 ± 14) have developed many quantitativeinstruments that assess reasoning ability.Some are included as components in stan-dard assessments such as the GraduateRecord Examination, whereas others arestand-alone tests such as Lawson' s Class-

room Test of Scientific Reasoning (LCTSR)(8, 9). We used the LCTSR because of itspopularity among STEM educators andresearchers. Common categories of reason-ing ability assessments include propor-tional reasoning, deductive and inductivereasoning, control of variables, probabilityreasoning, correlation reasoning, and hypo-thesis evaluation, all of which are crucialskills needed for a successful career in STEM.

Research-based standardized tests thatassess student STEM content knowledge arealso widespread. For example, in physics,education research has produced manyinstruments. We used the Force Concept

Inventory (FCI) (18, 19) and the Brief Elec-

tricity and Magnetism Assessment (BEMA)(20). These tools are regularly administeredby physics education researchers and educa-tors to evaluate student learning of specificphysics concepts.

Using FCI (mechanics), BEMA (elec-tricity and magnetism), and LCTSR (sci-entific reasoning), we collected data (seefigure, page 586) from students (N = 5760)in four U.S. and three Chinese universities.All the universities were chosen to be ofmedium ranking (15). The students testedwere freshmen science and engineeringmajors enrolled in calculus-based intro-ductory physics courses. The tests wereadministered before any college-levelinstruction was provided on the relatedcontent topics. The students in China usedChinese versions of the tests, which werefirst piloted with a small group of under-graduate and graduate students (n = 22) toremove language issues.

The FCI results show that the U.S. stu-dents have a broad distribution in themedium score range (from 25 to 75%). Thisappears to be consistent with the educationalsystem in the United States, which producesstudents with a blend of diverse experiencesin physics learning. In contrast, the Chinesestudents had all completed an almost identi-cal extensive physics curriculum spanningfive complete years from grade 8 throughgrade 12. This type of education background

produced a narrow distribution that peaksnear the 90% score.

For the BEMA test, the U.S. studentshave a narrow distribution centered a bitabove the chance level (chance 20%). TheChinese students also scored lower thantheir performance on the FCI, with thescore distribution centered around 70%.The lower BEMA score of students in bothcountries is likely due to the fact that someof the topics on the BEMA test (for exam-ple, Gauss' s law) are not included in stan-dard high school curricula.

The FCI and BEMA results suggest thatnumerous and rigorous physics courses inthe middle and high school years directlyaffect student learning of physics contentknowledge and raise students to a fairly highperformance level on these physics tests.

The results of the LCTSR test show acompletely different pattern. The distribu-tions of the Chinese and U.S. students arenearly identical. Analyses (15) suggest thatthe similarities are real and not an artifactof a possible ceiling effect. The resultssuggest that the large differences in K± 12STEM education between the UnitedStates and China do not cause much varia-tion in students' scientific-reasoning abili-ties. The results from this study are consis-tent with existing research, which suggeststhat current education and assessment inthe STEM disciplines often emphasizefactual recall over deep understanding ofscience reasoning (2, 21 ± 23).

What can researchers and educators doto help students develop scientific-reason-ing ability? Relations between instructionalmethods and the development of scientificreasoning have been widely studied andhave shown that inquiry-based scienceinstruction promotes scientific-reasoningabilities (24 ± 29). The current style of con-tent-rich STEM education, even when car-ried out at a rigorous level, has little impacton the development of students' scientific-reasoning abilities. It seems that it is notwhat we teach, but rather how we teach,that makes a difference in student learningof higher- order abilities in science reason-ing. Because students ideally need todevelop both content knowledge and trans-ferable reasoning skills, researchers andeducators must invest more in the develop-ment of a balanced method of education,such as incorporating more inquiry-basedlearning that targets both goals.

Our results also suggest a differentinterpretation of assessment results. Asmuch as we are concerned about the weakperformance of American students in

TIMSS and PISA (30, 31), it is valuable toinspect the assessment outcome from mul-tiple perspectives. With measurements onnot only content knowledge but also otherfactors, one can obtain a more holisticevaluation of students, who are indeedcomplex individuals.

References and Notes1. R. Iyengar et al., Science 319, 1189 (2008).

2. A. Y. Zheng et al., Science 319, 414 (2008).

3. B. S. Bloom, Ed., Taxonomy of Educational Objectives:The Classification of Educational Goals, Handbook I:Cognitive Domain (David McKay, New York 1956).

4. National Research Council (NRC), National ScienceEducation Standards (National Academies Press,

Washington, DC, 1996).

5. NRC, Learning and Understanding: Improving AdvancedStudy of Mathematics and Science in U.S. High Schools(National Academies Press, Washington, DC, 2002).

6. H. Singer, M. L. Hilton, H. A. Schweingruber, Eds.,

America' s Lab Report (National Academies Press,

Washington, DC, 2005).

7. P. Adey, M Shayer, Really Raising Standards: CognitiveIntervention and Academic Achievement (Routledge,

London, 1994).

8. A. E. Lawson, J. Res. Sci. Teach. 15, 11 (1978).

9. Test used in this study was Classroom Test of ScientificReasoning, rev. ed. (2000).

10. P. A. Facione, Using the California Critical Thinking SkillsTest in Research, Evaluation, and Assessment (California

Academic Press. Millbrae, CA, 1991).

11. H. A. Simon, C. A. Kaplan in Foundations of CognitiveSciences, M. I. Posner, Ed. (MIT Press, Cambridge, MA,

1989), pp. 1± 47.

12 R. E. Nisbett, G. T. Fong, D. R. Lehman, P. W. Cheng,

Science 238, 625 (1987).

13. Z. Chen, D. Klahr, Child Dev. 70, 1098 (1999).

14. D. Kuhn, D. Dean, J. Cognit. Dev. 5, 261 (2004).

15. See Supporting Online Material for more details.

16. This is based on the Chinese national standards on K± 12

education (www.pep.com.cn/cbfx/cpml/).

17. J. Hehn, M. Neuschatz, Phys. Today 59, 37 (2006).

18. D. Hestenes, M. Wells, G. Swackhamer, Phys. Teach. 30,

141 (1992).

19. The test used in this study is the 1995 version.

20. L. Ding, R.Chabay, B. Sherwood, R. Beichner, Phys. Rev.ST Phys. Educ. Res. 2, 010105 (2006).

21. M. C. Linn et al., Science 313, 1049 (2006).

22. A. Schoenfeld, Educ. Psychol. 23, 145 (1988).

23. A. Elby, Am. J. Phys. 67, S52 (1999).

24. C. Zimmerman, Dev. Rev. 27, 172 (2007).

25. P. Adey, M. Shayer, J. Res. Sci. Teach. 27, 267 (1990).

26. A. E. Lawson, Science Teaching and the Development ofThinking (Wadsworth, Belmont, CA, 1995).

27. R. Benford, A. E. Lawson, Relationships Between EffectiveInquiry Use and the Development of Scientific ReasoningSkills in College Biology Labs (Arizona State University,

Tempe, AZ, 2001); Educational Resources Information

Center (ERIC) accession no. ED456157.

28. E. A. Marek, A. M. L. Cavallo, The Learning Cycle andElementary School Science (Heinemann, Portsmouth,

NH, 1997).

29. B. L. Gerber, A. M. Cavallo, E. A. Marek, Int. J. Sci. Educ.23, 5359 (2001).

30. Trends in International Mathematics and Science Study

(TIMSS), http://nces.ed.gov/timss/.

31. Programme for International Student Assessment (PISA),

www.pisa.oecd.org/.

32. We wish to thank all the teachers who helped with this

research.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/323/5914/586/DC1

10.1126/science.1167740


EDUCATIONFORUM

Published by AAAS

...

......



CR

ED

IT: K

. SU

TLIF

F/S

CIE

NC

E

PERSPECTIVES

The ability of cells and tissues torespond to mechanical force is centralto many aspects of biology. Areas re-

levant to human physiology and diseaseinclude development and maintenance ofbone, blood vessels, and muscles; regulationof blood pressure; motility of cells; andregulation of cell proliferation (1).Mechanosensitive adhesions mediatedby membrane proteins called integrinsare believed to underlie a number ofthese behaviors (2). Forces applied eitherby the cells' own cytoskeleton or fromexternal sources induce strengthening ofthese adhesions and trigger a varietyof intracellular signals through theseeffects. Two papers in this issue, one byFriedland et al. on page 642 (3) and theother by del Rio et al. on page 638 (4),increase our understanding of the molec-ular basis for these phenomena.

Current models for mechanotrans-duction generally involve alterations inprotein conformation by forces (1, 5).Stretch-activated membrane channels,present in cells from all organisms thathave been examined, undergo conforma-tional transitions in response to changesin membrane tension. For cytoskeletaland extracellular matrix proteins thatform linear polymers, tension is thoughtto cause partial or complete unfolding ofdomains, exposing new binding sites orother motifs. The extracellular matrixprotein fibronectin is the best-studiedexample. Fibronectin requires tension toassemble into fibrils (6, 7), and fluores-cence resonance energy transfer meas-urements showed that fibronectin in fib-rils is in an extended state (8). Stretchingfibronectin exposes a cryptic binding sitein the first type III domain, leading toassociations between fibronectin pro-teins that promote fibril formation (6).

Integrins are one link in a connectionthat runs from extracellular matrix pro-teins such as fibronectin across theplasma membrane to the actin cytoskele-ton. Integrins form this connection in part

by binding the cytoskeletal protein talin, whichbinds actin directly, as well as by recruiting thecytoskeletal protein vinculin, which also bindsactin (9). Because every component in thischain experiences tension, each one is in princi-ple a potential mechanotransducer.

The best-defined response of integrin-dependent cell adhesions to tension is rein-forcement of the adhesions to resist the appliedforce (2). These responses occur over severaltime scales. Over tens of seconds, vinculin isrecruited to small adhesions, most likely with-

out recruitment of additional integrins(10). Over a few minutes, adhesions growlarger and lengthen in the direction ofthe applied force (11), which appears toinvolve recruitment of additional inte-grins. Indeed, application of force bystretching elastic substrata triggers con-version of unoccupied, low-affinity inte-grins to a high-affinity state, whichinduces new binding to the extracellularmatrix (12). These reinforcement mech-anisms may also mediate changes indownstream signaling. Indeed, the activ-ity of focal adhesion kinase (FAK), atyrosine kinase that transduces multiplesignals from integrins, requires actin-and myosin-dependent tension (13).

Friedland et al. provide evidence thatthe major fibronectin-binding integrin,α

5β

1, undergoes a force-dependent con-

formational transition. This conclusionis based on the ability of integrin α

5β

1to

be chemically cross-linked to fibro-nectin. Force is required for conversionof the bond from a non± cross-linkable toa cross-linkable state, suggesting achange in proximity of key residues.Integrin α

5β

1binds to two sites in

fibronectin: an Arg-Gly-Asp (RGD)sequence in the 10th type III repeat anda secondary (so-called) synergy site inthe 9th repeat (7) (see the figure).Friedland et al. also show that conver-sion to the cross-linkable state requiresthe synergy site and that these eventscorrelate with phosphorylation (thus,activation) of FAK. The picture thatemerges from these results is that initiallow-tension binding of integrin α

5β

1to

fibronectin involves association ofthe integrin with the RGD sequence,which under force converts to a higher-strength, more readily cross-linked bondthat involves the synergy site. Only thissecond conformation can activate FAKand transmit downstream signals.

Talin is the focus of the study by delRio et al. It binds directly to the β

1inte-

Through changes in protein conformation

and interactions, cells sense and respond

to forces at their point of attachment to

extracellular matrix.The Force Is with UsMartin A. Schwartz

CELL BIOLOGY

TalinFAK

Vinculin

Extracellularmatrix

Actin

Integrin

Fibronectin

RGDSynergy

Low tensionbinding of integrin

High tensionbinding of integrin

Proteinconformational

changes

Cytoplasm

Membrane

Altered proteininteractions

PO4

* *

Force mechanics. At points of cell attachment to extracellular

matrix, when tension is low, integrin α5β

1binds to the RGD

sequence in fibronectin. The cytoplasmic domain of integrin α5β

1

is associated with talin, but most vinculin binding sites on talin are

inaccessible. Increased tension triggers conformational changes

in the integrin, talin, and fibronectin. These alterations reveal new

vinculin binding sites in talin, enhance the affinity of α5β

1for the

synergy site in fibronectin, and reveal new self-association sites in

fibronectin. These changes lead to increased vinculin binding to

talin, tighter binding of the integrin to fibronectin, increased

fibronectin matrix assembly, and activation of FAK. These events

may also promote increased clustering of integrins.

Departments of Microbiology, Cell Biology, andBiomedical Engineering, Cardiovascular ResearchCenter and Mellon Urological Cancer ResearchInstitute, University of Virginia, Charlottesville, VA22908, USA. E-mail: [email protected]

Published by AAAS

...

......


grin cytoplasmic domain and forms one of thelinks to actin. Talin binding is also crucial inthe conversion of low-affinity to high-affinityintegrins (14) and in the recruitment of vin-culin to sites of adhesion (2). Interestingly, thetalin rod domain contains multiple vinculinbinding sites, but in the intact molecule, mostof these sites are buried among bundles of αhelices. Indeed, these structural data led to theprediction that unfolding of the talin roddomain under force might expose these cryp-tic sites to recruit vinculin (15). Because vin-culin can also connect to actin, it would rein-force the link between the integrin and thecytoskeleton. Using single-molecule tech-niques, del Rio et al. show that application offorce at the piconewton level results in unfold-ing of the talin rod domain and binding of vin-culin, thus confirming this key prediction.

The advances made by Friedland et al. anddel Rio et al. will facilitate answering a num-

ber of long-standing questions about integrin-mediated mechanotransduction. What is theprecise nature of the force-activated confor-mation of the integrin and how does thisconformation promote FAK activation? It' salso not yet clear how recruitment of vinculinto talin mediates downstream signaling, orwhether there are other intracellular compo-nents recruited to newly exposed bindingsites in talin that mediate these functions.Finally, does altered integrin binding tofibronectin directly affect fibronectinmatrix assembly, or does this occur throughthe previously described unfolding offibronectin domains under force and subse-quent self-association (6)? Growing evi-dence suggests that the entire adhesionapparatus functions as a force-transducingand -sensing machine. These two studiesbring us several steps closer to understand-ing its detailed dynamics.

References1. A. W. Orr, B. P. Helmke, B. R. Blackman, M. A. Schwartz,

Dev. Cell 10, 11 (2006).2. A. D. Bershadsky, N. Q. Balaban, B. Geiger, Annu. Rev.

Cell Dev. Biol. 19, 677 (2003).3. J. C. Friedland, M. H. Lee, D. Boettiger, Science 323, 642

(2009).4. A. del Rio et al., Science 323, 638 (2009).5. V. Vogel, Annu. Rev. Biophys. Biomol. Struct. 35, 459

(2006).6. C. Zhong et al., J. Cell Biol. 141, 539 (1998).7. I. Weirzbicka-Patynowski, J. E. Schwarzbauer, J. Cell Sci.

116, 3269 (2003).8. G. Baneyx, L. Baugh, V. Vogel, Proc. Natl. Acad. Sci.

U.S.A. 99, 5139 (2002).9. K. Burridge, M. Chrzanowska-Wodnicka, Annu. Rev. Cell

Dev. Biol. 12, 463 (1996).10. C. G. Galbraith, K. M. Yamada, M. P. Sheetz, J. Cell Biol.

159, 695 (2002).11. D. Riveline et al., J. Cell Biol. 153, 1175 (2001).12. A. Katsumi, T. Naoe, T. Matsushita, K. Kaibuchi, M. A.

Schwartz, J. Biol. Chem. 280, 16546 (2005).13. R. W. Tilghman, J. T. Parsons, Semin. Cancer Biol. 18, 45

(2008).14. S. Tadokoro et al., Science 302, 103 (2003).15. D. R. Critchley, Biochem. Soc. Trans. 33, 1308 (2005).

10.1126/science.1169414


CR

ED

IT: P. H

UE

Y/S

CIE

NC

EPERSPECTIVES

Hydrogenation is a process in which anadded hydrogen atom bonds toanother atom by sharing its only elec-

tron with this atom. Organic chemists usehydrogenation to convert unsaturated into sat-urated fats and to transfer alkenes into alkanes(1). On page 610 of this issue, Elias et al. (2)show that a similar process can also be realizedin condensed matter physics. Incorporatinghydrogen into grapheneÐ a single layer of car-bon atoms arranged in a hexagonal crystal lat-tice and the thinnest material conducting elec-tricityÐ transforms it into graphane, a newmaterial whose physical properties are verydifferent from those of graphene.

A graphene layer can be separated from acrystal of graphite by a peeling process whilepreserving its crystal structure and excellentelectrical conduction (3). Furthermore, itselectrical conduction can be tunedÐ similar tothat of a field-effect transistorÐ by putting it ona substrate and varying the density of mobileparticles (electrons and holes) by applying avoltage between the layer and substrate. Theunique physical and electrical properties ofgraphene allow not only the experimental test-ing of fundamental principles of quantumphysics but also the fabrication of new devices

for future carbon-based nanoelectronics (4).Unlike in conventional semiconductors,

the conducting electrons and holes ingraphene are massless and behave like pho-tons. In addition, there is no gap between thetwo energy bands in which the electrons andholes are located. Therefore, the electricalconductance of graphene is nonzero, evenwhen the voltage between the layer and sub-

strate is tuned so that there are no free chargedparticles in the graphene layer (4). Somedevice applications, however, require a gapbetween the two bands. One way to createsuch a gap is to cut a narrow ribbon from thegraphene sheet with a width of less than 100nanometers, thereby confining the electronsand holes to a ª quantum boxº and splitting theenergies of the two bands (5, 6). Such cutting

Synthesis of a new material by the

hydrogenation of graphene offers the

opportunity for wider device applications.Transforming GrapheneAlex Savchenko

MATERIALS SCIENCE

School of Physics, University of Exeter, Exeter EX4 4QL, UK.E-mail: [email protected]

Hydrogen atoms

Carbon atoms

Graphene layer

A

B

Graphene hydrogenation. (A) A graphene layer, where delocalized electrons are free to move between car-bon atoms, is exposed to a beam of hydrogen atoms. (B) In nonconductive graphane, the hydrogen atomsbond their electrons with electrons of carbon atoms and pull the atoms out of the plane.

Published by AAAS

...

......



PERSPECTIVES

can modify further the properties of graphene,because the dangling electron bonds at the rib-bon edges are chemically active and can cap-ture elements from the environment (7).

The surface of graphene is thought to berelatively inert. The electrons that are not partof the bonds between carbon atoms in theplane are shared between all atoms. It is thesemobile electrons that are responsible for elec-trical conduction. In a recent theoretical study,Sofo et al. predicted that hydrogenation of thegraphene surface should create a stable mate-rial, graphane, with a large energy band (8). Inhydrogenated graphene, alternating carbonatoms are pulled out of the plane in oppositedirections by the attached hydrogen atoms(see the figure, panel B). This atomic arrange-ment resembles that in diamond, a noncon-ductive form of carbon with a large energygap. In hydrogenated graphene, though, thesize of the gap can be controlled by varyingthe amount of hydrogen on its surface (9).

Elias et al. present the first experimentaldemonstration that graphane can be synthe-sized, showing that the surface of graphene ischemically active and that the energy gap canbe produced without cutting graphene into aribbon. The authors also show that the chemi-cal modification of graphene is reversible; theoriginal properties of graphene can be largelyrestored by annealing the hydrogenated sam-ples at high temperatures.

Adding atomic hydrogen to graphene is nota simple task: Hydrogenation first requiresbreaking apart the diatomic molecules ofhydrogen gas; this is why hydrogenation inchemistry is usually performed with a hot cata-lyst. Instead, Elias et al. exposed the graphenesamples to a hydrogen plasma discharge, inwhich the hydrogen gas is dissociated intohydrogen ions (see the figure, panel A). Thesample has to be placed some distance from theplasma discharge to avoid mechanical damageof graphene by energetic ions.

The emergence of the energy gap manifestsitself in the electrical conductance, whichbecomes strongly temperature dependent. Itshows an exponential dependence correspon-ding to so-called electron ª hopping.º Suchdependence has been seen in thin electron lay-ers confined within semiconductors with anenergy gap, such as silicon (10) and galliumarsenide (11). Although in pristine graphene,electrons are free to move across the crystal, thepresence of the gap after hydrogenation forcesthem to ª hopº from one site to another, and thisconduction mechanism is much more efficientat higher temperatures.

Another direct confirmation of the modifi-cation due to hydrogenation is provided by thediffraction image of the new crystal lattice in

transmission electron microscopy, where anoticeable decrease of the separation betweenthe carbon atoms caused by the pulling effect ofhydrogen atoms is seen. Raman spectroscopy,which is sensitive to the vibrations of the atomsin the crystal lattice (phonons), also shows theevolution of the lattice during hydrogenation.

The demonstration of reversible hydro-genation of graphene is an important step inexpanding the device applications for thisexciting material. There will be a search foroptimal ways of graphene hydrogenation,including chemical methods: Recently therehas been a study of graphene hydrogenationusing dissociation of a hydrogen-rich materialdeposited on graphene (12). Future researchshould aim at using graphene for hydrogenstorage in hydrogen-fuel technologies,because graphane has a very high hydrogendensity (8). This extremely thin material withan energy gap is also likely to find use in nano-electronics. But research will probably not be

limited to adding only hydrogen to graphene.Elias et al. show that the graphene surface canbe used as a base for creating new materials,and it will be interesting to study the effects ofincorporating other elements into its structure.

References1. R. O. C. Norman, D. J. Waddington, Modern Organic

Chemistry (Collins Educational, London, 1995).2. D. C. Elias et al. Science 323, 610 (2009).3. K. S. Novoselov et al., Science 306, 666 (2004). 4. K. S. Novoselov, A. K. Geim, Nat. Mater. 6, 183 (2007).5. M. Y. Han, B. Özyilmaz, Y. Zhang, P. Kim, Phys. Rev. Lett.

98, 206805 (2007).6. X. Li, X. Wang, L. Zhang, S. Lee, H. Dai, Science 319,

1229 (2008).7. T. Wassmann, A. P. Seitsonen, A. M. Saitta, M. Lazzeri, F.

Mauri, Phys. Rev. Lett. 101, 096402 (2008).8. J. O. Sofo, A. S. Chaudhari, G. D. Barber, Phys. Rev. B 75,

153401 (2007).9. D. W. Boukhvalov et al., Phys. Rev. B 77, 035427 (2008).

10. M. Pepper, S. Pollitt, C. J. Adkins, R. E. Oakeley, Phys.

Lett. A 47, 71 (1974).11. E. I. Laiko et al., Sov. Phys. JETP 66, 1258 (1987).12. S. Ryu et al., Nano Lett. 8, 4597 (2008).

10.1126/science.1169246

High-temperature cuprate supercon-ductors were discovered in 1986, yetmany properties of these materials

remain unexplained. When confronted with anew material with complicated behavior,physicists often map out its ª phase diagramº ;they plot the different phases assumed by thematerial (solid, liquid, gas, magnet, metal,insulator, and many more complicated ver-sions of these) as a function of thermody-namic parameters (including temperature,pressure, magnetic field, and number ofcharge carriers). In the phase diagram of thecuprates, the superconducting phase is notthe only one that is lacking a clear theoreticalbasis. Several ª normalº insulating and metal-lic phases form at higher temperatures or inhigh magnetic fields that are also puzzling.The unusual way in which the electrical resis-tivity changes with temperature in these ª nor-malº phases could provide clues to the mech-anism of electron pairing in the supercon-ducting phase.

On page 603 of this issue, Cooper et al. (1)use high magnetic fields to suppress the super-conducting state in a cuprate, La

2-xSr

xCuO

4,

which extends the study of the resistivity of itsmetallic phase to lower temperatures. TheCooper et al. data fail to support the standardworking model phase diagram for the cuprates:that a quantum phase transition between theinsulating and metallic phases occurs at a sin-gle point (2, 3). Instead, the results add to the listof strange cuprate behaviors.

The only well-understood phase of thecuprates, an antiferromagnet, has no transitionto the superconducting phase (see the figure,panel A). As strontium doping adds holes tothe system, the static magnetism of the antifer-romagnet gives way to magnetic fluctuations.There is a broad consensus that these magneticfluctuations provide the pairing mechanismrequired for superconductivity in the cuprates.The development of a detailed theory of high-temperature superconductivity might benefitfrom understanding how magnetic fluctua-tions manifest themselves in the nearby nor-mal phases. In particular, electrical resistivityprovides insights into how electrons scatter inthe material, and different scattering mecha-

A metallic phase of a high-temperature superconductor reveals unexpected properties in the

zero-temperature limit.

An Abnormal Normal StateGregory S. Boebinger

PHYSICS

National High Magnetic Field Laboratory, Florida StateUniversity, Tallahassee, FL 87545, USA. E-mail: [email protected]

Published by AAAS

...

......


nisms give rise to character-istic temperature dependen-ces of the resistivity.

In the underdoped region(fewer holes than optimalfor superconductivity), aª messy insulating phaseºforms that is not yet under-stood. Unlike a normal insu-lator, this phase does nothave a true energy gap thatprevents electrons from re-aching the conduction band.In this phase, a dramaticsuppression of the density ofstates at low energies robselectrons of the states theyuse to move through thematerial, making them proneto localization and corre-lated behavior that arisefrom the Coulomb andmagnetic interactions be-tween electrons.

The electron transport istwo-dimensional (2D); theelectron momentum is notisotropic but is largely restricted to the copper-oxygen planes. When magnetic fields suppresssuperconductivity, the resistivity, ρ, of thepseudogap state at low temperatures increaseslogarithmically with decreasing temperature(T ). This characteristic feature of the pseudo-gap state has been observed in several of thecuprates, but it is not yet understood in terms ofan electron-scattering mechanism. (4).

In the overdoped region, the three-dimen-sional (3D) metallic phase also shows anunexpected temperature dependence. Insteadof ρ changing as T2, as expected for electronsscattering off other electrons, it follows a curi-ous T 3∕2 behavior (5).

Between the messy insulator and 3Dmetal, in the region of optimum hole dopingfor superconductivity, lies the ª 2D strangemetalº (see the figure, panel A). In early stud-ies, samples exhibiting the highest supercon-ducting transition temperature displayed anormal-state resistivity that varied linearly inT (6). This observation has been borne out bymore recent systematic studies (7). The linearρ± T relation extends over such a large temper-ature range with such striking precision (8, 9)that understanding its origin might lead to anunderstanding of the mechanism of high-tem-perature superconductivity.

The temperature dependence of the resistiv-ity of the normal states is unusual but not totallyunprecedented. Similarities have long beennoted between the high-temperature supercon-ductors and the heavy fermion systems, in

which magnetic fluctuations are also believedto cause superconductivity and in which therole of electron conduction is better under-stood. The phase diagrams are similar (10), andthe 3D resistivity in heavy fermion systemsarising from magnetic fluctuations varies asT 3/2 (11), just as in the 3D metal regime of thehigh-temperature superconductors.

The heavy fermions exhibit a zero-temper-ature quantum phase transition that is nor-mally obscured by the superconducting phasebut that can be revealed by applying a strongmagnetic field to suppress the superconduc-tivity. Much of the interest in examining thelow-temperature limiting behavior of the nor-mal state in the high-temperature supercon-ductors is to see whether they exhibit a similarzero-temperature quantum phase transitionnear optimum doping.

In this model, there should be a singlepoint in the low-temperature normal stateÐ aparticular degree of hole dopingÐ that marksthe location of a quantum phase transition.This point would be found by extrapolatingthe 2D strange metal regime to zero tempera-ture (see the figure, panel A). The lineardependence of ρ with T might then be causedby critical fluctuations that extend to hightemperatures. Many physical quantities, suchas specific heat and the Knight shift, suggestthat there might be such a phase transition (3).Anomalies in the Hall coefficient observednear optimum doping might be direct evi-dence of such a transition (12, 13) because the

Hall coefficient is sensitive tochanges in the electronic den-sity of states that would beexpected to change at the quan-tum phase transition betweenthe 2D pseudogap state andthe 3D metallic state.

Through a systematic studyof resistivity in overdopedsamples, Cooper et al. nowshed a different light on theabnormal normal state of thehigh-temperature supercon-ductors. Instead of a singlepoint at which the resistiv-ity shows linear-T behavior,Cooper et al. argue that theresistivity should not be inter-preted as a single power-lawdependence on T, but rather asa mixture of linear-T behaviorand T 2 behavior throughoutthe right-hand side of the super-conducting dome, once super-conductivity is suppressed byhigh magnetic fields (see thefigure, panel B). Resistivity

consistent with a sum of T and T2 terms downto low temperatures had been noted previ-ously at a single point in the 3D metal regime(14). However, Cooper et al. report the firstsystematic study of many samples throughoutthis entire regime.

A linear-T component existing over abroad range of doping is a challenge to theprevailing thinking that the convergingboundaries of the 2D strange metal wouldextrapolate in the zero-temperature limit to aquantum phase transition at a single specificdoping. To paraphrase Alice in Wonderland,the abnormal normal state gets curiouserand curiouser.

References1. R. A. Cooper et al., Science 323, 603 (2009); published

online 11 December 2008 (10.1126/science.1165015).

2. G. S. Boebinger et al., Phys. Rev. Lett. 77, 5417 (1996).

3. J. L. Tallon, J. W. Loram, Physica C 349, 53 (2001).

4. Y. Ando, G. S. Boebinger, A. Passner, T. Kimura, K. Kishio,

Phys. Rev. Lett. 75, 4662 (1995).

5. H. Takagi et al., Phys. Rev. Lett. 69, 2975 (1992).

6. H. L. Stormer et al., Phys. Rev. B 38, 2472 (1988).

7. Y. Ando et al., Phys. Rev. Lett. 93, 267001 (2004).

8. S. Martin, A. T. Fiory, R. M. Fleming, L. F. Schneemeyer, J.

V. Waszczak, Phys. Rev. Lett. 60, 2194 (1988).

9. S. Martin, A. T. Fiory, R. M. Fleming, L. F. Schneemeyer, J.

V. Waszczak, Phys. Rev. B 41, 846 (1990).

10. N. D. Mathur et al., Nature 394, 39 (1998).

11. P. Gegenwart et al., Physica C 408-410, 157 (2004).

12. F. F. Balakirev et al., Nature 424, 912 (2003).

13. F. F. Balakirev et al., Phys. Rev. Lett. 102, 017004 (2009).

14. A. P. Mackenzie, S. R. Julian, D. C. Sinclair, C. T. Lin, Phys.

Rev. B 53, 5848 (1996).

10.1126/science.1168952


PERSPECTIVES

A B

UNDERDOPED OVERDOPED

1992(5 years after the discovery of the SC phase)

Today(with magnetic field suppressing the SC phase)

0.0 0.2 0.30.1

AFI SC

3D metal 2D strange metal

Messy

insulating

phase

T

T + T 2

T 2 ?

UNDERDOPED OVERDOPED

0.0 0.2 0.30.1

AFI SC

3D metal 2D strange metal

T

T 3/2

Pseudogap

regime

Hole doping in the CuO2 planes

Tem

per

atu

re

The evolving cuprate phase diagram. (A) Schematic phase diagram of the high-temperaturesuperconductors from 1992. The solid lines delineating the antiferromagnetic insulator (AFI)and superconducting (SC) phases are the only true phase transitions, the extent of the otherregimes remaining ill-defined. The doping level for optimum superconductivity is the same asfor the strange metal. The temperature dependence depicted refers to resistivity, as describedin the text. (B) The phase diagram presented by Cooper et al. when sufficiently intense mag-netic fields suppress the superconducting phase. The boundary of the suppressed supercon-ducting phase is shown with a dashed line. A much broader strange metal regime forms at lowtemperatures than had been expected (the region labeled T + T2)

Published by AAAS

...

......



CR

ED

IT: K

. SU

TLIF

F/S

CIE

NC

E

PERSPECTIVES

The plant shoot epidermis is studdedwith stomata, pores (each surroundedby a pair of guard cells) that regulate

gas and water vapor exchange between plantsand the atmosphere. Stomata develop throughunequal cell divisions, but the mechanismsthat establish this polarity and regulate cellfate are mostly unknown (1). On page 649 ofthis issue, Cartwright et al. identify a receptorthat positions cell divisions with respect toadjacent cells (2). And a recent study byLampard et al. shows that stomatal number isregulated by the phosphorylation of a specifictranscription factor (3).

Because a plant cell is surrounded by awall of carbohydrates, the position of the twodaughter cells that arise from cell divisiondepends on the position of the new cell wallthat forms to separate the cytoplasm, two newnuclei, and other cellular constituents. Thiswall placement can involve communicationbetween adjacent cells. In maize, for example,subsidiary mother cells produce cells thatflank stomata at maturity. This relationship isestablished when young flanking cells arepositioned next to the stomatal precursor cell(guard mother cell) (4). This suggests that theguard mother cell broadcasts signals to thesubsidiary mother cells to orient unequal divi-sion (see the figure). Other polarities developwithin neighbor cells as well, includingnuclear migration and the formation ofcytoskeletal (actin) patches close to the guardmother cell. But the specific regulators thatestablish polarity across plant cells are stillpoorly understood.

The PANGLOSS1 (PAN1) gene was identi-fied by screening for mutant maize plants thatdisrupt stomatal development (4). PAN1 isrequired to orient nuclear migration, actindistribution, and asymmetric cell division(2, 4). It encodes a transmembrane, receptor-like kinase that is preferentially localized in thesubsidiary mother cells next to the guardmother cell. The polar distribution of PAN1develops before nuclear and actin polarizationand cell division. Thus, PAN1 likely perceivescues from the guard mother cell that orientsubsidiary mother cell polarity and division.Other receptors, such as ERECTA (ER) and

TOO MANY MOUTHS (TMM) in the plantArabidopsis thaliana, also orient asymmetriccell division in stomatal development (5, 6).Like PAN1, TMM likely orients the divisionaxis in response to intercellular cues (6). Butunlike PAN1, TMM is dispensable for asym-metric division and is distributed throughoutthe cell. Thus, PAN1 seems to be a novel plantexample of a cell-intrinsic receptor proteinpolarized by extrinsic cues (7). How PAN1 actsis unclear, given that it lacks apparent kinaseactivity as well as critical amino acids present infunctional kinases. But it might function in anetwork, as it is required for the phosphoryla-tion of another membrane protein.

In Arabidopsis, protein phosphorylation isessential for stomatal development via a mito-gen-activated protein kinase (MAPK) cas-cade, a phosphorylation signal amplificationpathway that here includes YODA andMPK3/6 kinases (8, 9). This signaling path-way restricts stomatal initiation, because loss-of-function mutations in components of thepathway induce excess stomata. However,

neither the target of the stomatal pathwayMAPK nor that of any developmentallyrelated MAPK cascade in plants was knownbefore the work of Lampard et al. (3). Two setsof transcription factorsÐ proteins that regu-late the expression of target genesÐ have beenshown to control cell fate in the stomatal path-way (10 ± 13). One such factor is SPEECH-LESS (SPCH), because spch mutations blockstomatal initiation (10, 11).

Lampard et al. (3) recently showed thatSPCH is a substrate of MPK3/6 and that itsphosphorylation limits the number of divi-sions initiated by SPCH. SPCH is thereforerequired for stomatal formation, whereasthe MAPK cascade limits stomatal output.Together they constitute a core module thatintegrates positive and negative signals toregulate stomatal frequency (see the fig-ure). It is possible, but not shown, that anER receptor kinase complex restricts divi-sions by triggering the MAPK cascade.Thus, the first developmental MAPK sub-strate to be identified in plants is one that

Signals that control plant cell division and fate

also control epidermal pore development and

gas exchange.Pores in PlaceFred D. Sack and Jin-Gui Chen

PLANT SCIENCE

Asymmetriccell division

Arabidopsis stomatal development

Maize stomatal development

Subsidiarymother cell

Subsidiarycell

Guardmother cell

Guardcells

PAN1localization


(+SPCH)(-MAPK, ER, and TMM)


Stoma

Guard cells

Guardmother cell

sends signals

Division control. In maize, unequal division of an epidermal cell (pink) produces a guard mother cell (blue),which sends signals to adjacent subsidiary mother cells that polarize the distribution of the protein PAN1(red). PAN1 is required for orienting division of the subsidiary mother cells so that the smaller daughter cells(purple) flank the stomata. In Arabidopsis, stomatal initiation starts when an epidermal cell (pink) dividesunequally. Initiation is promoted by SPCH and restricted by a MAPK signaling pathway. Final stomatal num-ber depends on the extent of phosphorylation of SPCH by MAPK.

Department of Botany, University of British Columbia,Vancouver, BC V6T 1Z4, Canada. E-mail: [email protected]; [email protected]

Published by AAAS

...

......


produces an adaptive number of stomataand helps put them in their place. Thisstomatal module might be widespread,given that SPCH-like genes are present ingrasses and other plants.

Collectively, the studies by Cartwright et

al. and Lampard et al. (3) reveal mechanismsthat control plant cell fate specification anddivision polarization. Questions outstandinginclude the nature of intercellular cues thatpolarize PAN1, how PAN1 coordinates polarcell division and cell fate allocation, and howa MAPK cascade that controls many differ-ent plant processes specifically regulatesstomatal development.

Life on Earth depends in no small part onstomata that control plant evaporative coolingand CO

2uptake for photosynthesis, thereby

influencing global water and carbon cycling.Pangloss pontificated in Voltaire' s Candide

that ª this is the best of all possible worlds.º Itis fitting that his eponymous protein and theloquacious SPEECHLESS make this planetlush and green.

References1. D. C. Bergmann, F. D. Sack, Annu. Rev. Plant Biol. 58, 163

(2007).

2. H. N. Cartwright, J. A. Humphries, L. G. Smith, Science 323,

649 (2009).

3. G. R. Lampard, C. A. MacAlister, D. C. Bergmann, Science

322, 1113 (2008).

4. K. Gallagher, L. G. Smith, Curr. Biol. 10, 1229 (2000).

5. E. D. Shpak, J. M. McAbee, L. J. Pillitteri, K. U. Torii, Science

309, 290 (2005).

6. J. A. Nadeau, F. D. Sack, Science 296, 1697 (2002).

7. C. A. ten Hove, R. Heidstra, Curr. Opin. Plant Biol. 11, 34

(2008).

8. D. C. Bergmann, W. Lukowitz, C. R. Somerville, Science

304, 1494 (2004).

9. H. Wang, N. Ngwenyama, Y. Liu, J. C. Walker, S. Zhang,

Plant Cell 19, 63 (2007).

10. C. A. MacAlister, K. Ohashi-Ito, D. C. Bergmann, Nature

445, 537 (2007).

11. K. Ohashi-Ito, D. C. Bergmann, Plant Cell 18, 2493

(2006).

12. L. J. Pillitteri, D. B. Sloan, N. L. Bogenschutz, K. U. Torii,

Nature 445, 501 (2007).

13. M. M. Kanaoka et al., Plant Cell 20, 1775 (2008).

10.1126/science.1169553


PERSPECTIVES

Single-molecule experiments are pro-viding an increasingly detailed pictureof the function of biomolecules and

their assemblies. By watching one moleculeat a time, one can observe phenomena thatwould be lost in macroscopic measurementsaveraged over large numbers of molecules.Force spectroscopy is particularly well suitedto probe changes in biomolecular conforma-tion (1± 4). Like other single-molecule meth-ods (5), it monitors the distance between twosites in a macromolecule, but in addition, avariable pulling force is coupled to this dis-tance. Because stability is roughly exponen-tially sensitive to applied force, conforma-tional equilibria can be probed over a widefree-energy range.

On page 633 of this issue, Junker et al.(1) report the use of atomic forcemicroscopy (AFM) to explore the force-dependent kinetics of calmodulin foldingand ligand binding. Calmodulin, a calciumion (Ca2+)± sensing protein ubiquitous ineukaryotes, contains two Ca2+-binding do-mains connected by a flexible linker. UponCa2+ binding, the domains open and partiallyexpose their hydrophobic interior. Thisconformational change enables bindingto various targets involved in biologicalprocesses ranging from apoptosis to musclecontraction (6).

Junker et al. elucidate calmodulin' s con-formational dynamics at unprecedented reso-lution. In addition to the now familiar ª saw-toothº patterns in the force-extension curves,in which a sharp drop of the force indicatesunfolding, the authors directly observe re-folding of the protein: They record a remark-able succession of unfolding and foldingevents near the rupture point, where previ-ously only a single unfolding event wouldhave been recorded. The observation of

refolding was made possible by operatingcloser to equilibrium, taking advantage ofimproved instrument stability. Together withoptical tweezers (2, 3) and force-clamp AFM(4), this technique is part of a new generationof force-spectroscopy experiments forobserving continuous dynamical trajectoriesclose to equilibrium.

The ability to watch folding or bindingdynamics in real time opens the way for directobservations of molecular reaction mecha-

nisms as a sequence of structuralevents along the microscopicpathways (5). In a classical analy-sis of bulk experiments that probechanges in population of meta-stable species, a reaction mech-anism would be inferred fromthe combined effects of perturba-tions (such as mutations) on thekinetics and thermodynamics (7).For example, macroscopic kineticmeasurements have shown thatribonuclease H folds via a stableintermediate (8). However, theseexperiments did not reveal whetherthe intermediate was on thepathway between the unfoldedand folded states or was an off-pathway trap. A single-moleculeoptical-tweezer experiment de-monstrated an on-pathway inter-mediate by directly observingtransitions between the threestates near equilibrium (2).

Junker et al. now provide newinsights into the mechanisms of

Researchers have now observed in real time

how a single molecule of calmodulin refolds

and how it binds to a ligand.Unfolding the Secrets of CalmodulinRobert B. Best,1 Gerhard Hummer2

BIOCHEMISTRY

1Department of Chemistry, University of Cambridge,Lensfield Road, Cambridge CB2 1EW, UK. 2Laboratory ofChemical Physics, National Institute of Diabetes andDigestive and Kidney Diseases (NIDDK), National Institutesof Health, Bethesda, MD 20892, USA. E-mail: [email protected]

PREEQUILIBRIUM SCENARIO

INDUCED-FIT SCENARIO

Time

Stable

intermediate

Unstructured

transition stateMole

cula

r exte

nsio

n

Mechanisms of coupled folding and binding. These schematictrajectories portray the extension of a protein (green) with a cova-lently attached ligand (purple) as a function of time. In a preequi-librium scenario, folding precedes binding, resulting in an inter-mediate; Junker et al. observed this scenario for binding of theMLCK peptide to calmodulin. In contrast, in an induced-fit sce-nario, binding and folding would occur together via an unstruc-tured transition state; Junker et al. deduced this scenario for thebinding of Ca2+ at high concentrations.

Published by AAAS

...

......



PERSPECTIVES

coupled folding and binding, a central issuein the function of natively unstructured pro-teins (9± 11). Calmodulin has long served asa textbook example of the preequilibriumscenario in ligand binding (see the figure). Inthis scenario, unbound proteins are predomi-nantly in the ligand-free (ª apoº ) structure.However, ligands bind not to these apo struc-tures, but to a preexisting small population ofª holo-likeº proteins that are structurally sim-ilar to the ligand-bound (ª holoº ) state. In thecase of calmodulin, the folded apo and holo-like states are in rapid exchange in theabsence of Ca2+ (12, 13), and peptide ligandsand Ca2+ are thought to bind to the alreadyfolded holo-like structure. By using force tocounteract the stabilizing effects of boundCa2+, Junker et al. could resolve folding ofthe calmodulin carboxyl-terminal domain.Their results imply that the preequilibriumscenario does not hold under all conditions.

The authors find that Ca2+ stabilizes theprotein by increasing the folding rate withoutaffecting the unfolding rate. Thus, under theconditions of high Ca2+ concentration in thisstudy, calmodulin folds by a different path-way, in which folding and Ca2+ binding occurtogether, bypassing the folded Ca2+-free apoand holo-like intermediates. This binding-induced folding (11) follows an induced-fitscenario, in which the ligand itself triggersthe conformational change (see the figure).At low Ca2+ concentration, the preequilib-rium mechanism is expected to prevail. Thesetwo alternative folding mechanisms have

also been seen in molecular simulations (14).By contrast to Ca2+ binding, Junker et al.

find that peptide binding to calmodulin has adifferent effect on the folding energy land-scape. The two peptide ligands that they con-sider bind only to the folded protein and sta-bilize it by decreasing the unfolding rate,conforming to a preequilibrium mechanism.To demonstrate this mechanism directly, theauthors linked one of the peptides, thecalmodulin-binding domain of myosin light-chain kinase, to calmodulin in such a waythat binding of the peptide induces a changein length of the protein-ligand construct.They observed an obligatory intermediatecorresponding to the folded protein in theabsence of the ligand, thus confirming themechanism inferred from their kineticmeasurements: Only in the preequilibriumcase would it be possible to observe a dis-tinct binding-competent intermediate beforeligand binding (see the figure).

Future single-molecule studies may beable to probe the effect of force on the fold-ing mechanism itself. If the pulling directionis not correlated with the motions along theintrinsic unfolding (or unbinding) route, themechanism should switch from the intrinsicmechanism at low force to a different mech-anism at high force (15). This switch mayhave a role in the mechanical strength ofproteins and could make a protein moreresistant to unfolding by tensile force.Previous work inferred such a switch bycomparing the extrapolation of unfolding

kinetics at high force with ensemble meas-urements at zero force (16). The ability tostudy folding over a wider range of forcesopens the possibility of observing thisswitch directly.

References and Notes

1. J. P. Junker, F. Ziegler, M. Rief, Science 323, 633

(2009).

2. C. Cecconi, E. A. Shank, C. Bustamante, S. Marqusee,

Science 309, 2057 (2005).

3. W. J. Greenleaf, K. L. Frieda, D. A. N. Foster, M. T.

Woodside, S. M. Block, Science 319, 630 (2008).

4. J. M. Fernandez, H. Li, Science 303, 1674 (2004).

5. B. Schuler, W. A. Eaton, Curr. Opin. Struct. Biol. 18, 16

(2008).

6. M. Ikura, G. M. Clore, A. M. Gronenborn, G. Zhu, C. B.

Klee, A. Bax, Science 256, 632 (1992)

7. A. R. Fersht, A. Matouschek, L. Serrano, J. Mol. Biol. 224,

771 (1992).

8. T. M. Raschke, J. Kho, S. Marqusee, Nat. Struct. Biol. 6,

825 (1999).

9. Y. Levy, S. S. Cho, J. N. Onuchic, P. G. Wolynes, J. Mol.

Biol. 346, 1121 (2005).

10. K. Sugase, H. J. Dyson, P. E. Wright, Nature 447, 1021

(2007).

11. A. G. Turjanski, J. S. Gutkind, R. B. Best, G. Hummer,

PLoS Comp. Biol. 4, e1000060 (2008).

12. N. Tjandra, H. Kuboniwa, H. Ren, A. Bax, Eur. J. Biochem.

230, 1014 (1995).

13. A. Malmendal, J. Evenäs, S. Forsén, M. Akke, J. Mol. Biol.

293, 883 (1999).

14. Y.-G. Chen, G. Hummer, J. Am. Chem. Soc. 129, 2414

(2007).

15. R. B. Best, E. Paci, G. Hummer, O. K. Dudko, J. Phys.

Chem. B 112, 5968 (2008).

16. P. M. Williams et al., Nature 422, 446 (2003).

17. R.B.B. is supported by a Royal Society University

Research Fellowship. G.H. acknowledges support by the

intramural research program of the NIDDK, NIH. We

thank A. Bax, W. A. Eaton, and A. Szabo for their com-

ments on the manuscript.

10.1126/science.1169555

On page 627 of this issue, Anstey et al.

(1) describe how social encountersbetween desert locusts promote re-

lease of the neurochemical serotonin, whichturns out to be both sufficient and necessaryfor switching their behavior from mutualavoidance to aggregationÐ a prerequisite forbuilding swarms of migrating plague locusts.This finding advances our understanding ofhow social interactions implement behav-ioral adaptations. Further research on amine-containing neurons and their interactions

with other messenger system± orchestratingbehavior in insects could lead to new pestcontrol strategies.

The notorious desert locust Schistocerca

gregaria lives mostly an inconspicuous solitaryexistence, cryptic in color (green) and behavior,avoiding others and flying alone at nighttime.This changes dramatically when, under favor-able climatic conditions and flourishing vege-tation, their numbers explode, triggering astriking transformation to the crowded, gregar-ious phase (2, 3). Bright yellow in color, thelocusts now attract each other, forming bandsof marching hoppers and eventually swarmscomprising billions of voracious plague locuststhat are tuned for long-distance migration andmass devastation of crops. This change is

highly complex and encompasses many mor-phological, physiological, and behavioral fea-tures controlled by numerous chemical mes-sengers (2, 3) and involving more than 500genes (4). Until now, however, no single agenthas been identified as primarily responsible fortransforming the meek solitary phase into theswarming gregarious phase.

The breakthrough by Anstey et al. capsa series of investigations coordinated mostlyby Stephen Simpson in collaboration withMalcolm Burrows (5± 8). From observationaldata of hundreds of locusts, the authors con-structed a mathematical model for assessingan individual' s degree of gregariousness basedon walking speed and time spent near otherlocusts, motionless or grooming. The results

When desert locusts meet up, their nervous

systems release serotonin, which causes them

to become mutually attracted, a prerequisite

for swarming.The Key to Pandora' s BoxP. A. Stevenson

ECOLOGY

Institute for Biology II, Faculty for Biosciences, Psychologyand Pharmacology, Leipzig University, Talstrasse 33,04103 Leipzig, Germany. E-mail: [email protected]

Published by AAAS

...

......


revealed that solitary locustsexhibit full gregarious be-havior within only 2 hoursof enforced crowding. Gre-garious behavior was alsoinduced by the mere sightand smell of other locusts,and most effectively by sim-ply stroking touch-sensitivehairs on a leg or stimulatingthe leg nerve electrically.Regardless of how it wasinduced, the degree of gre-gariousness was greater inlocusts with more serotoninin their nervous system.Furthermore, treatment withthis amine, its precursor, orreceptor agonists inducedgregarious behavior in soli-tary locusts, whereas sero-tonin receptor antagonistsand synthesis inhibitionprohibited it.

Serotonin is thus suf-ficient and necessary forinducing gregarious behav-ior (see the figure). Hence,serotonin is the key thatturns the behavior of soli-tary locusts from mutualrepulsion to attraction, aprerequisite for swarmingof gregarious locusts andopening Pandora' s Box.

Will this insight lead tomore effective pest controlmeasures? In the shortterm, probably not. To beeffective, antiserotonin-likechemicals would need tobe applied when the ani-mals are solitary locustsand scarce targets in vastexpanses of desert (aboutthree locusts per 100 m2).Current serotonergic drugsare not designed for pass-ing through the insect cuti-cle and sheath encasing thenervous system, nor arethey insect-selective, hencetheir use is ecologicallyunjustifiable.

Nonetheless, the newinsights into the mechanismunderlying locust phasechange harbors consider-able potential. But first weneed to know more. Thefirst insect serotonin recep-

tor was cloned in fruit flies 17 years ago (9),but still little is known about the function andpharmacology of the different receptor sub-types in other insects. Serotonin-containingneurons were identified in locusts 25 yearsago (10), and although they comprise onlyfive cell pairs in each thoracic compartment ofthe nervous system (see the figure, panel C),their function and connectivity are unknown.Can these neurons drive the complete processof gregarization? How does serotonin interactwith other messengers controlling phase poly-morphism? Can gregarious locusts be chemi-cally reconverted to the solitary phase, as canoccur naturally after prolonged isolation?

Clues will also probably emerge frominvestigating other behavioral roles of aminesin insects. Serotonin-induced aggregation oflocusts is analogous to suppressing escapebehavior by serotonin in other insects (11) andarthropods, where its effect depends on socialstatus (12). This and many other actions ofserotonin are opposed by octopamine, theinvertebrate' s adrenaline, which regulates thedecisions to fight, retreat, or court in cricketsand fruit flies (11, 13, 14) and to rest or foragein cockroaches and honey bees (15, 16) bybiasing the operation of neural circuits (17).Complete understanding of locust phasechange, and with it viable solutions for pestcontrol, will thus depend on continued basicresearch on all aspects of insect biology,where the aim is to discover how animalsadapt to ensure survival in the face of chang-ing environmental conditions.

References1. M. L. Anstey, S. M. Rogers, S. R. Ott, M. Burrows, S. J.

Simpson, Science 323, 627 (2009).

2. M. P. Pener, Y. Yerushalmi, J. Insect Physiol. 44, 365

(1998).

3. S. W. Applebaum, Y. Heifetz, Annu. Rev. Entomol. 44,

317 (1999).

4. L. Kang et al., Proc. Natl. Acad. Sci. U.S.A. 101, 17611

(2004).

5. P. Roessingh, S. J. Simpson, S. James, Proc. R. Soc.

London Ser. B 252, 43 (1993).

6. S. J. Simpson, E. Despland, B. F. Hagele, T. Dodgson,

Proc. Natl. Acad. Sci. U.S.A. 98, 3895 (2001).

7. S. M. Rogers et al., J. Exp. Biol. 206, 3991 (2003).

8. S. M. Rogers et al., J. Exp. Biol. 207, 3603 (2004).

9. R. Hen, Trends Pharmacol. Sci. 13, 160 (1992).

10. N. M. Tyrer, J. D. Turner, J. S. Altman, J. Comp. Neurol.

227, 313 (1984).

11. P. A. Stevenson, V. Dyakonova, J. Rillich, K. Schildberger,

J. Neurosci. 25, 1431 (2005).

12. D. H. Edwards, S.-R. Yeh, B. E. Musolf, B. L. Antonsen,

F. B. Krasne, Brain Behav. Evol. 60, 360 (2002).

13. S. C. Hoyer et al., Curr. Biol. 18, 159 (2008).

14. S. J. Certel, M. G. Savella, D. C. F. Schlegel, E. A. Kravitz,

Proc. Natl. Acad. Sci. U.S.A. 104, 4706 (2007).

15. D. Wicher, Endocr. Metab. Immune Disord. Drug Targets

7, 304 (2007).

16. D. J. Schulz, A. B. Barron, G. E. Robinson, Brain Behav.

Evol. 60, 350 (2002).

17. F. Libersat, H.-J. Pflüger, Bioscience 54, 17 (2004).

10.1126/science.1169280


PERSPECTIVES

SWARMING

MIGRATORY

PLAGUE

F

Favorable environmentalfactors

BANDS OF FULLY GREGARIOUS LOCUSTSE

Neuromodulators,pheromones, hormones

GREGARIOUS BEHAVIORD

Aggregation

Modulation of neuralcircuits by serotonin

100 µm

SEROTONIN RELEASE IN THORACIC GANGLIAC

Mechanical, olfactory,and visual stimulation

Serotonin-containingneurons

SOLITARY PHASE

HIGH POPULATION DENSITYB

Mutual avoidance

Frequent social contact

Favorable environmentalfactors

A

Serotonin' s role in locust swarming. (A) Solitary locusts are scarce andavoid each other. (B) Under favorable climatic conditions, populationdensity and hence social contact increases. (C) As shown by Anstey et al.,the resulting mechanical, olfactory, and visual stimuli induce serotoninrelease, possibly from neurons in the thoracic nervous system (arrow). (D)Serotonin promotes gregarious behavior and aggregation (1), presum-ably by selectively modulating specific neuronal circuits (7, 8). (E)Crowding recruits additional chemical messengers, leading to acquisitionof all gregarious traits, including color change from green to yellow(2, 3). (F) Environmental factors (such as drought or food shortage) pro-mote swarming and migration of adult gregarious locusts (2, 3).C

RE

DIT

S: LO

CU

ST

CA

RT

OO

NS IN

PA

NE

LS A

, D

, A

ND

E, FR

OM

(1); P

HO

TO

MIC

RO

GR

AP

H IN

PA

NE

L C

, P.

A. ST

EV

EN

SO

N/U

NIV

ER

SIT

Y O

F L

EIP

ZIG

; P

HO

TO

GR

AP

H IN

PA

NE

L F

, E

. B

UH

L/U

NIV

ER

SIT

Y O

F L

EIP

ZIG

Published by AAAS

...

......



In the middle years of his career, long beforehe had won the Nobel Prize for solving a central mystery of cell biology, Peter Agre f igured that it would be natural to make acareer shift at about the time he turned 50. Hewould spend more time outside the lab, and hewould devote more energy to public service.

Since following through on that plan, Agre has helped found Scientists and Engineers for America, which promotesresearchers' involvement in politics and publicpolicy. He briefly Ð and very publicly Ðconsidered running for the U.S. Senate. He served as an adviser to a presidential candidate. He even did a late-night TV turn on The Colbert Report.

Agre is planning to expand his role as anambassador for science when he becomespresident of AAAS in February. Because climate change and the financial crisis openthe door to an era of innovation and trans-formation, he sees an opportune moment forresearchers to reach out to schoolchildrenand their teachers, local and national policy-makers, and S&T colleagues overseas.

ª Scientists have been so worried about getting funded that they probably have not invested as much as they should in termsof public awareness,º Agre said in a recentinterview. ª But it seems to me that everytenured faculty in America owes something,and my idea would be tithing 10% of your timefor the public good¼ .I think being part of thepublic debate is very importantÐ and that' swhere we' re overdue.º

It' s impossible to understand Agre' s workwithout understanding his youth in Min-nesota. His father, a chemist, brought his children to the lab on weekends and set upsimple experiments for them. Agre, with hisbrother Jim, became an Eagle Scout. He didget a D in chemistry during his senior year ofhigh schoolÐ an anomaly he attributes to thedistractions and rebellion of late adolescence.

Now 60, his medical experience andresearch give him enormous credibility in thepolicy world, and that' s enhanced by his easy-going, self-effacing manner. In 1970, he grad-uated from Augsburg College in Minneapoliswith a degree in chemistry. Four years later, hereceived his M.D. from Johns Hopkins; it was

there, he says, that he became committed to biomedical research. After clinical training at Case Western University Hospitals inCleveland and postgraduate medical trainingat the University of North Carolina at ChapelHill, he returned to Hopkins.

In 2003, Agre won the Nobel Prize in Chemistry for his team' s 1991 discovery of aquaporinsÐ channels that allow water molecules to pass through cell membranes, a process essential to life. (He shared the prize with Roderick MacKinnon of Rocke-feller University.)

Heartland values still echo in his outlook.Public service is crucial, he says. Science mustrespect people' s religious beliefs, even whenthey question scientific findings. He has deepadmiration for teachersÐ and he insists theydeserve better salaries.

Science teachers ª are the ones who makescience interesting,º he said. ª I remember inmy Nobel banquet speech reciting a line that I realized really resonated with the audience:` Early in the life of every scientist, a child' sf irst interest in science was sparked by a teacher.' º

Agre was a science adviser to the campaignof Barack Obama, and the new president' s

respect for science ª gives me great confidence,ºhe said. ª Until this economic recession or fearof depression is remedied, there may not beincreased budgets¼ But if the government is listening, great things can happen.º

Agre will become AAAS president on 17 February, at the end of the AAAS AnnualMeeting in Chicago. He succeeds Harvard climate and ocean researcher James J.McCarthy, who will become chair of the AAAS Board of Directors. Already, Agrehas had an extensive interview with the New York Times.

Though he is a strong public advocate for science, he' s not ready to quit the lab. After astint as vice chancellor for science and technology at Duke University Medical Center, he returned to Hopkins last year tohead its Malaria Research Institute.

ª Science is about new adventures,º Agresaid. ª I still have a few years ahead of me, andI don' t want to just rest on the past.º

S C I E N C E P O L I C Y

Obama Taps Past AAASLeaders for Key S&T Posts

U.S. President Barack Obama has chosen 2006AAAS President John P. Holdren to serve as his top science adviser and has selected threeother distinguished researchers with ties toAAAS for key positions in his administration.

Holdren, a climate and energy scholar atHarvard and director of the Woods HoleResearch Center, has been appointed assistantto the president for science and technology anddirector of the White House Office of Scienceand Technology Policy.

Oregon State University marine ecologistJane Lubchenco has been appointed adminis-trator of the National Oceanic and Atmo-spheric Administration. Lubchenco served asAAAS president in 1997 and received the2005 AAAS Award for Public Understandingof Science and Technology.

Eric Lander and Harold Varmus, bothAAAS Fellows, will join Holdren on the Presi-dent' s Council of Advisors on Science andTechnology. Lander, who received the 2004AAAS Public Understanding award, is a principal leader of the Human Genome Projectand founding director of the Broad Institute.Varmus is president and CEO of MemorialSloan-Kettering Cancer Center and a formerdirector of the National Institutes of Health.

Ð Becky Ham

AAAS NEWS&NOTES EDITED BY EDWARD W. LEMPINEN

P R O F I L E

Agre: A New Era of Science Outreach at Home and Abroad

Peter Agre

Published by AAAS

...

......


A A A S

Call for Nomination of

2009 AAAS Fellows

AAAS Fellows who are current members of the association are invited to nominate members for election as Fellows. A Fellow is defined as a member ª whose efforts on behalfof the advancement of science or its applications

are scientifically or socially distinguished.º A nomination must be sponsored by threeAAAS Fellows, two of whom must have no affiliation with the nominee's institution.

Nominations undergo review by the steeringgroups of the association© s sections (the chair,chair-elect, retiring chair, secretary, and fourmembers-at-large of each section). Each steeringgroup reviews only those nominations desig-nated for its section. Names of Fellow nomineeswho are approved by the steering groups are

presented to the AAAS Council for election.Nominations with complete documentation

must be received by 11 May 2009. Nominationsreceived after that date will be held for the following year. The nomination form and a list of current AAAS Fellows can be found atwww.aaas.org/aboutaaas/fellows. To request ahard copy of the nomination form, please contactthe AAAS Executive Office, 1200 New YorkAvenue N.W., Washington, DC 20005, USA; at 202-326-6635; or at [email protected].


Results of the 2008 Election of AAAS OfficersFollowing are the results of the 2008 election. Terms begin on 17 February 2009.

General OfficesPresident-Elect: Alice S. HuangBoard of Directors: Julia M. Phillips, David S. SabatiniCommittee on Nominations: Steven Chu, SusanHackwood, Sallie Keller-McNulty, Jack Dixon

Section on Agriculture, Food, and RenewableResourcesChair-Elect: Brian A. LarkinsMember-at-Large: Barbara ValentElectorate Nominating Committee: Ann M. Hirsch,Mark E. SorrellsCouncil Delegate: Daniel J. Cosgrove

Section on AnthropologyChair-Elect: Clark Spencer-LarsenMember-at-Large: Carol V. WardElectorate Nominating Committee: Anne C. Stone,Dolores R. Piperno

Section on AstronomyChair-Elect: Alan P. BossMember-at-Large: Donald CampbellElectorate Nominating Committee: Michael Werner,Giovanni G. Fazio

Section on Atmospheric and HydrosphericSciencesChair-Elect: Alan RobockMember-at-Large: Kevin E. TrenberthElectorate Nominating Committee: Margaret(Peggy) LeMone, Jim Coakley

Section on Biological SciencesChair-Elect: Trudy MackayMember-at-Large: Margaret Werner-WashburneElectorate Nominating Committee: Joan W. Bennett,Eric Ursell Selker

Section on ChemistryChair-Elect: Charles P. CaseyMember-at-Large: Ronald W. WoodardElectorate Nominating Committee: Brian M. Stoltz,Jonathan A. Ellman

Section on Dentistry and Oral Health SciencesChair-Elect: Margarita Zeichner-DavidMember-at-Large: Ira B. LamsterElectorate Nominating Committee: Laurie K.McCauley, Frank A. Scannapieco

Section on EducationChair-Elect: Joseph KrajcikMember-at-Large: Jay B. LabovElectorate Nominating Committee: Suzanne

O' Connell, Mary Monroe Atwater

Section on EngineeringChair-Elect: Duncan T. MooreMember-at-Large: Christine M. MaziarElectorate Nominating Committee: KristenFichthorn, Pradeep K. KhoslaCouncil Delegate: Gail H. Marcus, James L. Merz

Section on General Interest in Science andEngineeringChair-Elect: Kathryn D. SullivanMember-at-Large: Sharon M. FriedmanElectorate Nominating Committee: Gloria J. Takahashi, Robert J. Griffin

Section on Geology and GeographyChair-Elect: Malcolm HughesMember-at-Large: Jean Lynch-StieglitzElectorate Nominating Committee: Timothy GordonFisher, Elizabeth A. Canuel

Section on History and Philosophy of ScienceChair-Elect: Richard CreathMember-at-Large: Heather E. DouglasElectorate Nominating Committee: Nancy J.Nersessian, Alain TouwaideCouncil Delegate: Virginia Trimble

Section on Industrial Science and TechnologyChair-Elect: Jennie C. Hunter-CeveraMember-at-Large: Harry S. HertzElectorate Nominating Committee: Qinghuang Lin,Robert BoilyCouncil Delegate: Steven W. Popper

Section on Information, Computing, andCommunicationChair-Elect: Bart SelmanMember-at-Large: Julia GelfandElectorate Nominating Committee: Christine L.Borgman, Bonnie C. Carol

Section on Linguistics and Language ScienceChair-Elect: David W. LightfootMember-at-Large: Suzanne FlynnElectorate Nominating Committee: Elizabeth C.Traugott, Douglas H. Whalen

Section on MathematicsChair-Elect: Kenneth C. MillettMember-at-Large: Tony F. ChanElectorate Nominating Committee: Douglas N.Arnold, Robert M. Fossum

Section on Medical SciencesChair-Elect: Judy LiebermanMember-at-Large: Robert DomsElectorate Nominating Committee: Beverly Davidson, Wendy C. BrownCouncil Delegate: Jennifer M. Puck, James H.Hughes, Etty (Tika) Benveniste, Reed E. Pyeritz, Terence S. Dermody

Section on NeuroscienceChair-Elect: Michael T. ShipleyMember-at-Large: Gail MandelElectorate Nominating Committee: Michael S.Wolfe, Erik D. Herzog

Section on Pharmaceutical SciencesChair-Elect: Gary M. PollackMember-at-Large: William T. BeckElectorate Nominating Committee: Craig K. Svensson, Per Artursson

Section on PhysicsChair-Elect: Charles W. ClarkMember-at-Large: Alexander L. FetterElectorate Nominating Committee: Elizabeth H.Simmons, Ali Yazdani

Section on PsychologyChair-Elect: Stephen J. SuomiMember-at-Large: Jenny SaffranElectorate Nominating Committee: Denise Park,David ShapiroCouncil Delegate: John Gabrieli

Section on Social, Economic, and PoliticalSciencesChair-Elect: Eugene A. RosaMember-at-Large: Wendy BaldwinElectorate Nominating Committee: Robert F. Rich,Anil B. DeolalikarCouncil Delegate: Nicholas Christakis

Section on Societal Impacts of Science andEngineeringChair-Elect: Bruce V. LewensteinMember-at-Large: Melanie LeitnerElectorate Nominating Committee: Robert M.Simon, Michael L. Telson

Section on StatisticsChair-Elect: Joel B. GreenhouseMember-at-Large: Kenneth W. WachterElectorate Nominating Committee: Nancy Reid,Mary A. Foulkes

Published by AAAS

...

......


Sudden Death of EntanglementTing Yu1* and J. H. Eberly2*

A new development in the dynamical behavior of elementary quantum systems is the surprisingdiscovery that correlation between two quantum units of information called qubits can be degradedby environmental noise in a way not seen previously in studies of dissipation. This new route fordissipation attacks quantum entanglement, the essential resource for quantum information as wellas the central feature in the Einstein-Podolsky-Rosen so-called paradox and in discussions ofthe fate of Schrödinger’s cat. The effect has been labeled ESD, which stands for early-stagedisentanglement or, more frequently, entanglement sudden death. We review recentprogress in studies focused on this phenomenon.

Quantum entanglement is a special type ofcorrelation that can be shared only amongquantum systems. It has been the focusof foundational discussions of quantum

mechanics since the time of Schrödinger (whogave it its name) and the famous EPR paper ofEinstein, Podolsky, and Rosen (1, 2). The degreeof correlation available with entanglement is pre-dicted to be stronger as well as qualitatively dif-ferent compared with that of any other knowntype of correlation. Entanglement may also behighly nonlocal—e.g., shared among pairs ofatoms, photons, electrons, etc., even though theymay be remotely located and not interacting witheach other. These features have recently promotedthe study of entanglement as a resource that webelieve will eventually find use in new approachesto both computation and communication, for ex-ample, by improving previous limits on speedand security, in some cases dramatically (3, 4).

Quantum and classical correlations alike al-ways decay as a result of noisy backgrounds anddecorrelating agents that reside in ambient envi-ronments (5), so the degradation of entanglementshared by two or more parties is unavoidable(6–9). The background agents with which we areconcerned have extremely short (effectively zero)internal correlation times themselves, and theiraction leads to the familiar law mandating thatafter each successive half-life of decay, there isstill half of the prior quantity remaining, so that adiminishing fraction always remains.

However, a theoretical treatment of two-atomspontaneous emission (10) shows that quantumentanglement does not always obey the half-lifelaw. Earlier studies of two-party entanglement indifferent model forms also pointed to this fact(11–15). The term now used, entanglement sud-den death (ESD, also called early-stage disentan-glement), refers to the fact that even a veryweaklydissipative environment can degrade the specif-ically quantum portion of the correlation to zero

in a finite time (Fig. 1), rather than by successivehalves. We will use the term “decoherence” torefer to the loss of quantum correlation, i.e., lossof entanglement.

This finite-time dissipation is a new form ofdecay (16), predicted to attack only quantum en-tanglement, and not previously encountered in thedissipation of other physical correlations. It hasbeen found in numerous theoretical examinationsto occur in a wide variety of entanglements in-volving pairs of atomic, photonic, and spin qubits,continuous Gaussian states, and subsets of mul-tiple qubits and spin chains (17). ESD has alreadybeen detected in the laboratory in two differentcontexts (18, 19), confirming its experimental re-ality and supporting its universal relevance (20).However, there is still no deep understanding ofsudden death dynamics, and so far there is nogeneric preventive measure.

How Does Entanglement Decay?An example of an ESD event is provided bythe weakly dissipative process of spontaneousemission, if the dissipation is “shared” by twoatoms (Fig. 1). To describe this we need a suit-able notation.

The pair of states for each atom,sometimes labeled (+) and (−) or (1)and (0), are quantum analogs of“bits” of classical information, andhence such atoms (or any quantumsystems with just two states) arecalledquantumbits or “qubits.”Unlikeclassical bits, the states of the atomshave the quantum ability to exist inboth states at the same time. This isthe kind of superposition used bySchrödinger when he introduced hisfamous cat, neither dead nor alivebut both, in which case the state ofhis cat is conveniently coded by thebracket (+ ⇔ −), to indicate equalsimultaneous presence of the oppo-site + and − conditions.

This bracket notation can be ex-tended to show entanglement. Sup-posewehave twoopposing conditionsfor two cats, one large and one small,

and either waking (W) or sleeping (S). Entangle-ment of idealized cats could be denoted with abracket such as [(Ws) ⇔ (Sw)], where we havechosen large and small letters to distinguish a bigcat from a little cat. The bracket would signal viathe term (Ws) that the big cat is awake and thelittle cat is sleeping, but the other term (Sw)signals that the opposite is also true, that the bigcat is sleeping and the little cat is awake.

One can see the essence of entanglement here:If we learn that the big cat is awake, the (Sw) termmust be discarded as incompatible with what welearned previously, and so the two-cat state re-duces to (Ws). We immediately conclude that thelittle cat is sleeping. Thus, knowledge of the stateof one of the cats conveys information about theother (21). The brackets are symbols of infor-mation about the cats’ states, and do not belong toone cat or the other. The brackets belong to thereader, who can make predictions based on theinformation the brackets convey. The same is trueof all quantum mechanical wave functions.

Entanglement can be more complicated, evenfor idealized cats. In such cases, a two-party jointstate must be represented not by a bracket asabove, but by a matrix, called a density matrixand denoted r in quantum mechanics [see (22)and Eq. S3]. When exposed to environmentalnoise, the density matrix r will change in time,becoming degraded, and the accompanying changein entanglement can be tracked with a quantummechanical variable called concurrence (23), whichis written for qubits such as the atoms A and B inFig. 1 as

CðrÞ ¼ max½0,QðtÞ ð1Þ

where Q(t) is an auxilliary variable defined interms of entanglement of formation, as given ex-plicitly in Eq. S4. C = 0 means no entanglementand is achieved whenever Q(t) ≤ 0, while for

REVIEW

1Department of Physics and Engineering Physics, Stevens In-stitute of Technology, Hoboken, NJ 07030–5991, USA. 2Roch-ester Theory Center and Department of Physics and Astronomy,University of Rochester, Rochester, NY 14627–0171, USA.

*E-mail: [email protected] (T.Y.), [email protected](J.H.E.)

Cavity a Cavity b

Atom A Atom B

Deg

ree

of e

ntan

glem

ent

Asymptotic decay

Sudden death

Fig. 1. Curves show ESD as one of two routes for relaxation ofthe entanglement, via concurrence C(r), of qubits A and B thatare located in separate overdamped cavities.


...

......


maximally entangled states C = 1, and C islimited to the range 1 ≥ C ≥ 0.

In the case of spontaneous emission there isno environment at all, except for the vacuum. Thevacuum can still have a noisy degrading effectthrough its quantum fluctuations, which cannotbe avoided, so both atoms in Fig. 1must eventuallylose their excitation and come to their ground stateswith the rate G. Then their state is simply (––), acompletely disentangled situation because learningthat the state of one atom is (−) does not changeour information about the other, also (−). Thus,disentanglement is the eventual fate of the pair.

The question is, how quickly do they meettheir fate? For the initial density matrix shown inEq. S5, the answer is supplied by the surfacegraphed in Fig. 2, which shows possible path-ways for entanglement dissipation as a functionof time. The k axis shows that the time evolutionof entanglement depends on the value of the pa-rameter that encodes the initial probability for thetwo atoms to be in the doubly excited state (++).The two extreme concurrence curves for k = 1and k = 0 are the ones already shown in Fig. 1.

The sudden death behavior shown in the righthighlighted curve of Fig. 2 is a new feature forphysical dissipation (16, 20) and is induced byclassical as well as quantum noises (24). It iscounterintuitive, based on all previous single-atomexperience, because spontaneous emission is aprocess that obeys the half-life rule rigorously forindividual atoms. But it turns out that the two-qubit correlation does not follow the one-qubitpattern. That is, the sudden death does not comefrom a shorter half-life; the entangled joint cor-relation does not even have the half-life property.

As reported in (18), the first experimentalconfirmation of ESDwasmadewith an all-opticalapproach focusing on photonic polarization. It wasachieved by the tomographic reconstruction ofr(t), and from it the Q(t) variable, and thus the

concurrence C(r). In the exper-iment, both amplitude and phasenoises that can degrade entangle-ment were realized by combin-ing beam splitters and mirrors.

Can Sudden Death Be Avoidedor Delayed?The issue of how to avoid ESD-type decorrelation in a realisticphysical system is incompletelyresolved at this time. A numberof methods are known to pro-vide protection against previous-ly known types of decorrelation(3). Some methods have clas-sical analogs in information the-ory. One engages appropriatelydesigned redundancyand isknownas quantum error correction. An-other relies on using a symmetrythat can isolate entanglementfrom noise, effectively providinga decoherence-free subspace to

manage qubit evolution.Error correction is most useful when the dis-

turbing noise is below some threshold (25). Inpractice error correction can be complicated, be-cause a noisy channel is a dynamical process andits physical features are often not fully under-stood or predictable. An example is atmosphericturbulence during open-air communication. An-other issue is the cost associated with providingredundancy. Additionally, it has been reported thatsome quantum error correction algorithms couldactually promote rather than mitigate ESD (26).Symmetries that avoid decoherence by providingisolation from noise during evolution (27) havealso been examined as a way to postpone or avoidESD, but knowledge of the noise to be combattedappears unlikely to be available because qubitsremote from each other would rarely share eithersymmetry properties or noise descriptions.

Other methods considered use dynamic ma-nipulation such as mode modulation (28) or thequantum Zeno effect (29), which can be regardedas extensions of the so-called bang-bang method(30). Another proposal is to use feedback control(31) to prevent ESD and other decoherenceeffects in the presence of hostile noise. None ofthese methods is perfect, but they can be moreeffective if designed for specific noise avoidance.

Does the Number of Noises Matter?Nonlocal entanglement raises the issue of ESDtriggered by different noise processes and can referto more than one noise source acting together oncolocated entangled qubits, or to independent noisesources acting separately on remotely locatedmem-bers of a qubit pair. The rate of dissipation in thepresence of several noise sources is normally thesum of the individual dissipation rates. More ex-plicitly, if decay rates G1 and G2 come from theaction of two distinct weak noises, then when thetwo noises are applied together to a physical sys-

tem, the resulting relaxation rate is simply givenby the sum of the separate rates: G1 + G2.

However, such a long-standing result doesnot hold for entanglement decay. This was dis-covered (16) by examining entanglement evo-lution of a set of X-form mixed-state matrices(Eq. S3) with d = 0, where d is the probability thatboth qubits are in their ground states. Straight-forward calculations for the entire class are shownby the diagrams in Fig. 3, illustrating two qubitsexposed together to amplitude noise as well asphase noise. The top two time-dependent curvesshow that the application of either noise separatelyallows long-running entanglement decay of thehalf-life type (no ESD). By contrast, the bottomcurve reaches zero in a finite time. That is, the com-bined effect of the two noises causes ESD. Thecaption explains the colored squares. This illustratesthe “supervulnerability” of paired-qubit entangle-ment when attacked by different independentnoises. This result is universal in the sense that itcontinues to hold (16) even if the two noisesattack one of the qubits but not the other, and alsoif each of the two qubits is remotely attacked byjust one of the noises.

Is There an “Anti-ESD” or Rebirth Effect?Special circumstances are needed to see “anti-ESD,” the creation or rebirth of entanglement

Fig. 2. Atom-atom entanglement is plotted as a function of time fork values in the range 0 to 1. [Adapted from (20)] For all values of kless than 1/3, the half-life rule is obeyed, but for k between 1/3 and 1,it is not. For those values, the curves show ESD, i.e., becoming zeroin a finite time and remaining zero thereafter. The two curvesmarked with arrows are similar to the curves in Fig. 1. Time isrepresented by p = 1 – exp (−Gt).

0.00.0

0.1

0.2

0.3

0.4

0.5

0.6

0.2 0.4 0.6 0.8 1.0

Co

ncu

rren

ce

Time

Amplitude

Phase

Mixed

EXP

EXP

ESD

Amplitude Phase Mixed

ESDa

|z|

Fig. 3. The top two time-dependent curves showexponential (smooth half-life type) decay of concur-rence for a qubit pair exposed to phase noise andto amplitude noise, respectively. The bottom curveshows nonsmooth decay, i.e., ESD occurs for thequbit pair when both noises are acting together.The color-coded squares apply to any two-party Xmatrix (Eq. S3) having d = 0. They present thepredicted results for the entire accessible physicaldomain, which means throughout 1 ≥ a ≥ 0 and1 ≥ |z| ≥ 0. The blue zones labeled EXP designatedomains where smooth exponential (half-life type)evolution occurs, and the orange zones show whereESD occurs. The left andmiddle squares apply whenamplitude and phase noises are applied separately,and have smooth decay regions, while the orangesquare at right shows that sudden death is universalin the entire region for any X matrix (Eq. S3).


REVIEW

...

......


from disentangled states. By imposing the rightinteractions almost anything can be made tohappen, but we are concerned with evolution ofjoint information in a pure sense and focus ontwo-party entanglement that evolves withoutmutual interaction or communication.

The same two-atom situation shown in Fig. 1can be made relevant to anti-ESD. In solving forthe surface of solutions plotted in Fig. 2, thecavities were taken as fully overdamped, so thatany photon emitted by either atomwas irreversiblyabsorbed by the walls, but they could also betreated as undamped mirror-like cavities, such asused in the Jaynes-Cummings (JC)model for light-atom interactions (32). This situation produces aperiodic sequence of perfect rebirths of atom ABentanglement (33, 34). An early mathematicalmodel of two-qubit evolution (11) can be inter-preted as treating an underdamped cavity and alsoshows rebirths.

The panels in the top row of Fig. 4 showrebirth scenarios. They occur for states that areinitially of the cat type, such that both atoms areexcited and both are un-excited at the same time.The cat-type bracket for them is

Fa ¼ ½ðþþÞcos a ⇔ ð−−Þsin a ð2Þ

where different values of sine and cosine producethe different curves in the figure (35).

Starting from the photon vacuum state in eachcavity, the JC-type evolutionwill permit only zeroor one photon to reside in each cavity at any latertime, so each of the two modes is a two-statesystem (a qubit), and counting the two atoms thereare now four qubits on hand. This provides sixconcurrences that can be computed: CAB, Cab,CAb, CaB, CAa, and CBb, where the capital lettersidentify atoms and the small letters identify thephotons in cavity modes a or b. Concurrence isdefined only for pairs of qubits, not quartets ofthem, so the label CAB implies that the a and bdegrees of freedom are not available to observa-tion and have been ignored (technically, have beentraced out), and for photon concurrence Cab theatomicA andB properties are traced out, and so on.

This idealized model provides a convenientframework to analyze entanglement in a simpli-fied but still multi-qubit framework. As shown inthe top half of Fig. 4, ESD takes place for atomconcurrence (in the left panel, almost allCAB curvesreach zero and remain zero for finite times). How-ever, in the right panel, the photon concurrenceCab behaves in a manner opposite to that of CAB,showing anti-ESD. That is, initially Cab is zero,but it immediately begins to grow. The photonsjointly experience entanglement “sudden birth,”but this is followed byESD a half cycle later. All ofthis occurs via pure “informatics,” i.e., without en-ergy exchange or other interaction between the sites.

The reason for the rebirths is obvious—thephotons emitted cannot really get lost among thefew joint states available. If a larger number ofcavity modes would be provided, a longer timewould be needed for a rebirth to be complete, and

as a limiting case, the cavities producing the curvesin Fig. 2 have an infinite number of modes, so thelost quantum correlation cannot be reborn in anyfinite time. If there are enough states available inone mode, as is the case for coherent-state modepreparation, then ESD and true long-time reviv-als are also predicted (36).

What Are the Future Prospects?Quantum memory banks. Clearly, ESD can belargely ignored, to a first approximation, whendesirable quantum operations can bemanipulated

at sufficiently high speed. The key goal of mem-ory is opposite to that of speed, i.e., to preservequantum state features semi-indefinitely. Quan-tum memory networks will be sensitive to theconsequences if ESD occurs. ESD will probablyhave to be taken into account if practical versionsof quantum memories are built to operate inmixed-state configurations.

Disentanglement control. Over a given noisychannel, it appears that some entangled statesmay bemore robust against the influence of noisethan others. To control decoherence optimally, itwill be useful to learn how to identify the robuststates separately from the fragile ones. Control

issues also include the use of external fields tomanipulate qubit states (37) as in gate operations,to create transient decoherence-free subspaces, asmentioned already. A qualitatively different routeto combat decoherence specifically of the ESDtype is illustrated in Fig. 2. Two evolution tracksare highlighted to show that for qubits preparedwith the same value of initial entanglement, theirconcurrences may evolve very differently. In thatillustration, the right track is subject to ESD,whilethe left one is not. Decoherence per se is notavoided, because the non-ESD track shows steady

dissipation, but it always remains finite. Otherexamples of this are known, and in some cases apurely local operation (i.e., a manipulation of onlyone of the two entangled qubits) can be under-taken to change the state matrix r without chang-ing its degree of entanglement, but in a way thatswitches the evolution trajectory from ESD tonon-ESD (38), effectively putting it on a half-lifetrack as in the figure. Similar studies (39) haveexamined the effect of local operations at inter-mediate stages of evolution. Use of this methodrequires detailed knowledge of the state matrix r,which may not be practical, particularly at timeslate in the evolution.

0 0.5 1 1.5 20

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 0.5 1 1.5 20

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 0.5 1 1.5 20

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 0.5 1 1.5 20

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

gt/π

C(ρ

AB)

gt/π

C(ρ

ab)

gt/π

C(ρ

AB)

gt/π

C(ρ

AB)

α=π/4α=π/6α=π/12

α=π/4α=π/6α= pi/12

Fig. 4. Entanglement birth, death, and rebirth. [Adapted from (35)] In the bottom pair of panels the riseand fall of AB entanglement is exactly compensated by the fall and rise of ab entanglement. This is not thecase in the top pair of panels, but a more subtle form of compensation still occurs, as reported in (35). Itinvolves not concurrence but the auxiliary variable Q(t) defined in Eqs. 1 and S4.


REVIEW

...

......


Entanglement invariants. Entanglement flowin small reservoirs has led to the recent dis-covery of entanglement “invariants” (35) by in-spection of the curves in the bottom row ofFig. 4, which repeats the top row except that aslightly different cat state is used for the atomsinitially:

Ya ¼ ½ðþ−Þcos a ⇔ ð−þÞsin a ð3ÞIn the bottom left panel of Fig. 4, it appears

that each AB atomic concurrence curve is com-pensated at all times by the corresponding abphotonic concurrence curve in the right panel,one going up as the other falls.

In fact, exact compensation can be confirmedanalytically (35), but one notes that the same be-havior does not appear in the top-row curves inFig. 4, where there is no perfect compensation ofab for AB. For example, it is easy to see that theAB and ab red curves can be zero at the sametime. A natural question is, where does the miss-ing information go in that case? Because thetwo-site JC model is unitary, preservation of allfour-qubit information is guaranteed, so it shouldbe located “somewhere.” Careful examinationshows that concurrence is not conserved, butratherQ(t) is conserved, spread among all six dif-ferent types in that case (35).

This identification of four-particle memoryflow channels is unusual and clearly deservesfuture examination (40, 41). One can say aboutthese invariants that they emerge only from akind of analytic continuation of the bipartite con-currence function C(t) to un-physically negativevalues, which is permitted via Q(t). The entan-glement flow issue (42) is also related to, andappears to expand considerably, the theory asso-ciatedwith entanglement swapping,which is underactive exploration, and has been realized withparticle pairs from independent sources (43).

Non-Markovian noises. Dissipative entangle-ment evolution is critically dependent on the typesof the noises acting on the system. Markov envi-ronments are those for which a noise signal hasno self-correlation over any time interval, andunder Markov conditions noise typically resultsin a quantum irreversible process. Non-Markoviannoise arising from a structured environment orfrom strong coupling appears to be more fun-damental (44, 45). Recent studies have suggestedthat correlated noises may cause new difficultiesin using quantum error correction codes (46) anddynamic decoupling (47). Although some progresshas been made, extending the current research onESD into physically relevant non-Markovian sit-uations remains a challenge. High-Q cavity QEDand quantum dot systems are two possible exper-imental venues.

Qutrits and beyond.Many-qubit entanglementand entanglement of quantum systems that are notqubits (i.e., those having more than two states) islargely an open question, and one that is embar-rassing in a sense, because the question has beenopen since quantum mechanics was invented inthe 1920s. There is still no known finite algo-

rithm for answering the simple-seeming questionof whether a given mixed state is entangled ornot, if it refers to more than two systems, and it isanswerable for two mixed-state systems only inthe case of pairs of qubits (as we have been dis-cussing) and the case of one qubit and one qutrit(a three-state system such as spin-1). Investiga-tion into ESD of qubit-qutrit systems has begun(48, 49), but generalizations to many-qubit sys-tems are daunting tasks due to both technical andconceptual difficulties (7–9).

Topological approach. N-party entanglementdynamics will presumably become simpler to pre-dict if a computable entanglement measure for amixed state of more than two qubits can be dis-covered. However, an alternative approach is toavoid dynamics through topological analysis (50).One now knows (51, 52) that ESD is necessary(i.e., must occur) in arbitrary N-party systems ofnoninteracting qubits if they are exposed to ther-mal noise at any finite T > 0 temperature. Thesteady state of any noninteractingN-qubit systemhas a neighborhood in which every state is sepa-rable. In this case, any prearranged subsystementanglement will inevitably be destroyed in afinite time. This is a universal result showing howentanglement evolves in the absence of externalnoise control.

Clearly, the holy grail for research on entan-glement dynamics is to find an efficient real-timetechnique for tracking and controlling the entan-glement evolution of a generic many-qubit system.Another important open question is to determinea generic method for direct experimental registra-tion of entanglement, for which there is no currentanswer. We believe that many surprising resultsare awaiting discovery.

References and Notes1. E. Schrödinger, Naturwissenschaften 23, 807 (1935).2. A. Einstein, B. Podolsky, N. Rosen, Phys. Rev. 47, 777

(1935).3. M. A. Nielsen, I. L. Chuang, Quantum Computation

and Quantum Information (Cambridge Univ. Press,Cambridge, 2000).

4. C. H. Bennett, D. P. DiVincenzo, Nature 404, 247 (2000).5. W. H. Zurek, Rev. Mod. Phys. 75, 715 (2003).6. T. Yu, J. H. Eberly, Phys. Rev. B 66, 193306 (2002).7. C. Simon, J. Kempe, Phys. Rev. A 65, 052327 (2002).8. W. Dür, H. J. Briegel, Phys. Rev. Lett. 92, 180403 (2004).9. For an earlier review of entanglement dynamics, see (53).10. T. Yu, J. H. Eberly, Phys. Rev. Lett. 93, 140404 (2004).11. K. Zyczkowski, P. Horodecki, M. Horodecki, R. Horodecki,

Phys. Rev. A 65, 012101 (2001).12. A. K. Rajagopal, R. W. Rendell, Phys. Rev. A 63, 022116

(2001).13. L. Diosi, in Irreversible Quantum Dynamics, F. Benatti,

R. Floreanini, Eds. (Springer, New York, 2003),pp. 157–163.

14. S. Daffer, K. Wodkiewicz, J. K. McIver, Phys. Rev. A 67,062312 (2003).

15. P. J. Dodd, J. J. Halliwell, Phys. Rev. A 69, 052105 (2004).16. T. Yu, J. H. Eberly, Phys. Rev. Lett. 97, 140403 (2006).17. For details, see section 1 of the Supporting Online

Material.18. M. P. Almeida et al., Science 316, 579 (2007).19. J. Laurat, K. S. Choi, H. Deng, C. W. Chou, H. J. Kimble,

Phys. Rev. Lett. 99, 180504 (2007).20. J. H. Eberly, T. Yu, Science 316, 555 (2007).21. Our example illustrates the distinction between entangled

and nonentangled states, but quantum entanglement has

the ability to encode oppositeness that goes further than theawake-asleep dichotomy of the cats or the heads-tailsdichotomy of coins. One extension is the anti-alignmentof the hands of a clock, not uniquely found at 6:00,but realized also at different angles of the hands, atapproximately 7:05, 8:10, etc. Orthogonal photonpolarizations, which may be oriented to any direction,extend the concept to continuous “opposites,” exploitedby Clauser and Freedman in their famous experimentshowing quantum violation of a Bell Inequality (54).

22. Matrix encoding of the entanglement is more complicatedthan the bracket notation used for the cats, but it isimportant because we want to allow for more possibilities,which can be visualized by thinking that any of the fourjoint state labels ++, +−, −+, or −− can be correlated withitself or any other, making 16 basic correlation labels, andthese are conventionally arranged in a 4 × 4 matrix. Thisis the density matrix r referred to in the text. See theSupporting Online Material for an example, and a sketchof the theory of time evolution of the density matrix ina noisy environment.

23. W. K. Wootters, Phys. Rev. Lett. 80, 2245 (1998).24. T. Yu, J. H. Eberly, Opt. Commun. 264, 393 (2006).25. J. Preskill, Proc. R. Soc. Lond. A 454, 385 (1998).26. I. Sainz, G. Björk, Phys. Rev. A 77, 052307 (2008).27. D. A. Lidar, K. B. Whaley, in Irreversible Quantum

Dynamics, F. Benatti, R. Floreanini, Eds. (Springer,Berlin, 2003), pp. 83–120.

28. G. Gordon, G. Kurizki, Phys. Rev. Lett. 97, 110503 (2006).29. S. Maniscalco, F. Francica, R. L. Zaffino, N. Lo Gullo,

F. Plastina, Phys. Rev. Lett. 100, 090503 (2008).30. For example, see L. Viola, S. Lloyd, Phys. Rev. A 58, 2733

(1998).31. A. R. R. Carvalho, A. J. S. Reid, J. J. Hope, Phys. Rev. A

78, 012334 (2008).32. E. T. Jaynes, F. W. Cummings, Proc. IEEE 51, 89 (1963).33. M. Yönac, T. Yu, J. H. Eberly, J. Phys. B 39, S621 (2006).34. Z. Ficek, R. Tanas, Phys. Rev. A 74, 024304 (2006).35. M. Yönac, T. Yu, J. H. Eberly, J. Phys. B 40, S45 (2007).36. M. Yönac, J. H. Eberly, Opt. Lett. 33, 270 (2008).37. H. Nha, H. J. Carmichael, Phys. Rev. Lett. 93, 120408

(2004).38. T. Yu, J. H. Eberly, Quant. Inf. Comput. 7, 459 (2007).39. A. R. P. Rau, M. Ali, G. Alber, Eur. J. Phys. 82, 40002

(2008).40. I. Sainz, G. Björk, Phys. Rev. A 76, 042313 (2007).41. D. Cavalcanti et al., Phys. Rev. A 74, 042328 (2006).42. T. S. Cubitt, F. Verstraete, J. I. Cirac, Phys. Rev. A 71,

052308 (2005).43. M. Halder et al., Nat. Phys. 3, 692 (2007).44. M. Ban, J. Phys. A 39, 1927 (2006).45. B. Bellomo, R. Lo Franco, G. Compagno, Phys. Rev. Lett.

99, 160502 (2007).46. E. Novais, E. R. Mucciolo, H. U. Baranger, Phys. Rev. A

78, 012314 (2008).47. K. Shiokawa, B. L. Hu, Quant. Inf. Proc. 6, 55 (2007).48. K. Ann, G. Jaeger, Phys. Rev. B 75, 115307 (2007).49. A. Checinska, K. Wodkiewicz, Phys. Rev. A 76, 052306

(2007).50. B. V. Fine, F. Mintert, A. Buchleitner, Phys. Rev. B 71,

153105 (2005).51. T. Yu, J. H. Eberly, http://arxiv.org/abs/0707.3215 (2007).52. A. Al-Qasimi, D. F. V. James, Phys. Rev. A 77, 012117

(2008).53. F. Mintert, A. R. R. Carvalho, M. Kus, A. Buchleitner,

Phys. Rep. 415, 207 (2005).54. S. J. Freedman, J. F. Clauser, Phys. Rev. Lett. 28, 938

(1972).55. T.Y. and J.H.E. acknowledge grant support from the

U.S. Army Research Office (48422-PH), the NSF(PHY06-01804 and PHY07-58016), and U.S.Department of Energy, Office of Basic EnergySciences (DE-FG02-05ER15713).

Supporting Online Materialwww.sciencemag.org/cgi/content/full/323/5914/598/DC1SOM TextFig. S1References

10.1126/science.1167343


REVIEW

...

......


Facile Synthesis of AsP3Brandi M. Cossairt,1 Mariam-Céline Diawara,2 Christopher C. Cummins1*

White phosphorus consists of tetrahedralP4 molecules and is the key indus-trial intermediate for most phospho-

rus compounds of commercial importance (1).In contrast, in condensed phase the correspond-ing arsenic molecule, As4 (yellow arsenic), is boththermally and photochemically unstable, revert-ing readily to the stable gray allotrope with aninfinite sheet structure (2). To date, essentiallyall that is known of the binary molecules AsnP4–n(n = 1 to 3) has come from gas-phase studieswherein hot (660°C) vapors of phosphorus andarsenic mixtures under equilibrium reaction con-ditions were subjected to Raman spectroscopicanalysis (3). To determine the properties of AsP3as a pure and isolated substance, we set out tosynthesize it by using a method hinging ontransition-metal chemistry.

Two salts of the [P3Nb(ODipp)3]− anion,

of interest as a P33− synthon, were synthesized

by P4 activation with dichloride precursorCl2Nb(ODipp)3(THF) in the presence of reduc-ing agents (Fig. 1A), where Dipp indicates 2,6-C6H3[CH(CH3)2]2 and THF, tetrahydrofuran (4).The anion [P3Nb(ODipp)3]

− was characterized byx-ray crystallography as its sodium salt and foundto contain chemically equivalent P atoms and astructure in which a Nb(ODipp)3 unit replacesone vertex of the P4 tetrahedron (5). Solution nu-clear magnetic resonance (NMR) spectroscopy(31P, 13C, and 1H) studies of [P3Nb(ODipp)3]

− salts

are consistent with the structural assignment, withthe 31P NMR data [singlets at a chemical shift, d,of −206 and −170 parts per million (ppm) for thesodium and cobaltocenium salts, respectively] serv-ing as spectroscopic signatures of this system (5).

AsP3 itself was obtained from the reaction inTHF solvent of either [P3Nb(ODipp)3]

− salt withAsCl3; dichloride Cl2Nb(ODipp)3(THF) is regen-erated in the process (Fig. 1A). Purification ofAsP3 after workup is achieved either by crystal-lization from ether at −35°C or by sublimation. Ina typical synthesis, ~100 mg of bright white solidAsP3 was obtained, representing a 75% yield (5).AsP3 is identified with a single high-field

31P NMRresonance at −484 ppm, in good agreement withthe prediction given by our theoretical calculations[supporting online material (SOM) text]. Further-more, gas chromatography–mass spectrometry(GC-MS) analysis of a toluene solution of the whitepowder revealed a single band of retention time of9.1 min, corresponding to the molecular weightof AsP3 at 168 mass/charge (m/z) (Fig. 1B). High-resolution mass spectrometry data obtained byusing the electron impact technique featured aparent ion signal at 167.842 m/z (5).

We acquired Raman spectra of microcrystal-line AsP3 by using a solid state laser at 785 nm(Fig. 1D) (5). The four expected bands are prom-inent and compare well with theory (SOM text)and with those bands observed by Ozin for themolecule as one component of a hot gas (3).

AsP3 melts without decomposition at 71° to73°C and is thermally quite stable, withstandingtemperatures of 120°C for more than 1 week ina toluene solution. No precautions against irra-diation were necessary when handling isolatedsamples of AsP3 in ambient light; however, muchlike P4, AsP3 must be handled anaerobically be-cause it is pyrophoric.

For structural characterization, AsP3 wasderivatized by ligation to molybdenum in the(AsP3)Mo(CO)3(PiPr3)2 coordination complex,obtained as orange crystals in 63% yield (6).Single-crystal x-ray diffraction (Fig. 1C) revealedthe AsP3 tetrahedron bound to the metal centerby a single phosphorus vertex at a distance of2.487(1) Å. The three P–P bonds in the tetra-hedron average to 2.177 Å, whereas the As–Pbonds are longer, at an average of 2.305 Å (5).The observation of specific AsP3 coordination tomolybdenum at a phosphorus vertex is an initialexample of selectivity in a reaction of AsP3.

The synthetic strategy developed here forP3

3− transfer with generation of an EP3 tetrahe-dron is not limited to E = As. When using SbCl3in place of AsCl3 in the synthesis, we generatedthe exotic SbP3 molecule and identified it with abroadened resonance in the 31P NMR spectrumat d of −461.8 ppm (5).

One possible application for AsP3 might beas a stoichiometrically exact 1:3 source of ar-senic and phosphorus atoms for the synthesis ofadvanced materials (7).

References and Notes1. J. Emsley, The 13th Element: The Sordid Tale of Murder,

Fire, and Phosphorus (Wiley, New York, 2000).2. N. N. Greenwood, A. Earnshaw, Chemistry of the Elements

(Butterworth-Heinemann, Oxford, ed. 2, 1997).3. G. A. Ozin, J. Chem. Soc. A 1970, 2307 (1970).4. J. R. Clark et al., Inorg. Chem. 36, 3623 (1997).5. Materials and methods are detailed in supporting

material available on Science Online.6. T. Groer, G. Baum, M. Scheer, Organometallics 17, 5916

(1998).7. O. Manasreh, Semiconductor Heterojunctions and

Nanostructures (McGraw-Hill, New York, 2005).8. We thank the NSF (grant CHE-719157), Thermphos

International, and the Massachusetts Institute of Technology forfunding and N. A. Piro and H. A. Spinney for helpfuldiscussions. Crystallographic parameters for (Na)[P3Nb(ODipp)3],Cl2Nb(ODipp)3(THF), and (AsP3)Mo(CO)3(PiPr3)2 are availablefree of charge from the Cambridge Crystallographic DataCentre (CCDC 704466 to 704468).

Supporting Online Materialwww.sciencemag.org/cgi/content/full/323/5914/602/DC1Materials and MethodsSOM TextFigs. S1 to S5References

14 October 2008; accepted 12 November 200810.1126/science.1168260

BREVIA

1Department of Chemistry, Massachusetts Institute of Tech-nology, 77 Massachusetts Avenue, Room 6-435, Cambridge,MA 02139, USA. 2École Normale Supérieure de Lyon, Uni-versité Claude Bernard Lyon 1, 46, Allée d‘Italie, 69364 LyonCedex 07, France.


ODipp

ODippDippO

Nb

PP

P

Nb

Cl

ClODipp

DippO

DippO

THF

P P

P

P

P As

P

P

0.75

- M+

i or ii iii Mo1

P1

P2

P3

As1P31P32

0 500 1000 1500 2000 2500 3000 3500

0

5000

10000

15000

20000

25000

30000

35000

40000

Inte

nsity

(a.

u.)

200 300 400 500 600

Raman shift (cm-1)

Raman shift (cm-1)

313 cm-1

345 cm-1

428 cm-1

557 cm-1

A

B

C

D

60 80 100 120 140 1600

50000

100000

150000

200000

250000

300000

Abu

ndan

ce

m/z

GCMS of AsP3 in toluene solventretention time: 9.1 minutes

168AsP3

+

137AsP2

+

106AsP+

93P3

+

75As+

62P2

+

Fig. 1. (A) Synthetic scheme: (i) 0.5% Na/Hg, THF. (ii) 1. SmI2, THF; 2. 2 CoCp2 (M is Na or CoCp2).(iii) AsCl3, THF. (B) GC-MS spectrum of AsP3. (C) Thermal ellipsoid plot of (AsP3)Mo(CO)3(PiPr3)2,with 50% thermal ellipsoids and hydrogen atoms omitted for clarity. (D) Raman spectrum of AsP3.


...

......


Anomalous Criticality in theElectrical Resistivity of La2–xSrxCuO4R. A. Cooper,1 Y. Wang,1 B. Vignolle,2 O. J. Lipscombe,1 S. M. Hayden,1 Y. Tanabe,3 T. Adachi,3Y. Koike,3 M. Nohara,4* H. Takagi,4 Cyril Proust,2 N. E. Hussey1†

The presence or absence of a quantum critical point and its location in the phase diagram of high-temperature superconductors have been subjects of intense scrutiny. Clear evidence for quantumcriticality, particularly in the transport properties, has proved elusive because the important low-temperature region is masked by the onset of superconductivity. We present measurements of thelow-temperature in-plane resistivity of several highly doped La2–xSrxCuO4 single crystals in whichthe superconductivity had been stripped away by using high magnetic fields. In contrast to otherquantum critical systems, the resistivity varies linearly with temperature over a wide doping rangewith a gradient that scales monotonically with the superconducting transition temperature. It ismaximal at a critical doping level (pc) ~ 0.19 at which superconductivity is most robust. Moreover,its value at pc corresponds to the onset of quasi-particle incoherence along specific momentumdirections, implying that the interaction that first promotes high-temperature superconductivitymay ultimately destroy the very quasi-particle states involved in the superconducting pairing.

An important theme in strongly correlatedelectron systems is quantum criticalityand the associated quantum phase tran-

sitions that occur at zero temperature upon tuninga nonthermal control parameter, g (e.g., pressure,magnetic field H or composition), through acritical value, gc. One feature of such a system isthe influence that critical fluctuations have onthe physical properties over a wide region inthe (T, g) phase diagram above the quantumcritical point (QCP), inside which the systemshows marked deviations from conventionalLandau Fermi-liquid behavior. A number of can-didate non–Fermi-liquid systems have emerged,particularly in the heavy fermion family (1), al-though there are others, for example, certaintransition metal oxides (2), that display similarcharacteristics.

The physics of copper-oxide high-temperaturesuperconductors may also be governed by prox-imity to a QCP. The generic temperature-doping(T, p) phase diagram resembles that seen in theheavy fermions, with an apparent funnel-shapedregion that either pierces or skirts the supercon-ducting dome (3). Above this region, cupratesdisplay an in-plane resistivity, rab, that varieslinearly with temperature over a wide tempera-ture (4) yet narrow doping (5) range. This T-linear

resistivity has been widely interpreted, in tan-dem with other anomalous transport properties(6), as a manifestation of scale-invariant physicsborne out of proximity to the QCP. This view-point has remained untested, largely because ofthe high upper critical field Hc2 values in high-Tc cuprates that restrict access to the importantlimiting low-temperature region below Tc( p).We used a combination of persistent and pulsedhigh magnetic fields to expose the normal stateof La2–xSrxCuO4 (LSCO) over a wide dopingand temperature range and studied the evolutionof rab(T) with carrier density, from the slightlyunderdoped (p = 0.15) to the heavily overdoped( p = 0.33) region of the phase diagram. Our anal-ysis reveals the presence of a singular dopingconcentration in LSCO at which the electronicresponse changes, although in a manner distinctfrom that observed in other candidate quantumcritical systems.

In-plane resistivity of La2–xSrxCuO4. A seriesof high-field rab(T, H) measurements were car-ried out on overdoped LSCO single crystals withdoping levels of p = 0.18, 0.21, and 0.23 (labeledhereafter LSCO18, LSCO21, and LSCO23, re-spectively) with the field aligned perpendicularto the CuO2 planes in order to suppress the su-perconductivity. Figure 1A shows the rab(T, H)data obtained on LSCO23. In order to track thetemperature dependence of the zero-field resistivityr(T, 0) below Tc, we used a simple, transparenttechnique to extrapolate the high-field rab(T, H)data to the zero-field axis (Fig. 1B). The re-sultant r(T, 0) values, plotted in Fig. 1C togetherwith the zero-field rab(T) curve below 70 K,are found to exhibit a T-linear dependence downto 1.5 K. For comparison, we also plotted the ab-solute values of r(T, 48) at a fixed high field of48 T obtained directly from the vertical dashedline in Fig. 1A. The temperature dependence of

the latter (analysis-free) values is identical to thatof r(T, 0) and is consistent with earlier 60-T datataken on LSCO22 (7), showing that the anal-ysis itself has not introduced any additional,artificial temperature dependence in r(T, 0). Sim-ilar pulsed-field measurements and analysis werecarried out for the two other doping levels assummarized in fig. S1.

Figure 2 shows the resultant r(T, 0) valuesplus zero-field rab(T) data for seven differentconcentrations ranging from optimal doping(p = 0.17) to the heavily overdoped, nonsuper-conducting region (p = 0.33). The gradual cross-over in the temperature dependence of rab(T),from quasi-linear for LSCO17 to approximatelyquadratic for LSCO33, is evident in the raw dataand is consistent with previous studies carriedout above Tc (5, 8, 9). At low temperatures, how-ever, rab(T) develops predominantly T-linearbehavior for the entire doping range 0.18 ≤ p ≤0.29 [for p = 0.17, data exists only above Tc(H =0)]. Although evidence for a low-T T-linear re-sistivity has emerged for single doping concen-trations in both electron- (10) and hole-doped(11, 12) cuprates, our measurements show thatthe low-T linearity in fact persists over a broadrange of doping.

Single-component analysis. In heavy fermi-on systems, Dr(T), the T-dependent part of r(T),is often described by a single term anT

n whoseexponent n(T, H) evolves from the Fermi-liquidvalue n = 2 to some anomalous value less than 2over a narrow temperature and magnetic fieldwindow (13–15). The anomalous exponent inDr(T) persists to low temperatures only at thecritical field, Hc. In Fig. 3, we plotted a com-parative n(T, p) = d(lnDr)/d(ln T) for LSCO byusing the resistivity curves shown in Fig. 2.

For T > 50 K, the resultant phase diagram re-sembles that seen in prototypical quantum criticalsystems, with a narrow region in which rab(T) isapproximately (although not strictly) T-linear sep-arated from a region where rab(T) varies approx-imately as T2. As the temperature is lowered,however, the situation becomes markedly dif-ferent. Rather than collapsing to a single (critical)point, the T-linear region in LSCO fans out anddominates the low-T response. Intriguingly, thisT-linear regime (or more precisely, the regionwhere n < 1.1) is coincident with both the Tcparabola (long-dashed white line) and the super-conducting fluctuation regime (short-dashed whiteline) and has thus been obscured until now bythe veil of superconductivity.

Dual-component analysis. Previously, Drab(T)in overdoped, hole-doped cuprates has been ex-pressed either as above, that is, as anT

n (1 ≤n ≤ 2) (16), or as the sum of two components,a1T + a2T

2 (11, 17, 18). In fig. S2, we describein detail why the latter is in fact the more ap-propriate expression for LSCO. In Fig. 4, A andB, we show the doping dependences of a1 anda2, respectively, for two different fitting proto-cols. The solid squares are coefficients obtained

RESEARCHARTICLE

1H. H. Wills Physics Laboratory, University of Bristol, TyndallAvenue, Bristol, BS8 1TL, UK. 2Laboratoire National des ChampsMagnétiques Pulsés (LNCMP), UMR CNRS-UPS-INSA 5147,Toulouse 31400, France. 3Department of Applied Physics, Grad-uate School of Engineering, Tohoku University, 6-6-05 Aoba,Aramaki, Aoba-ku, Sendai 980-8579, Japan. 4Department ofAdvanced Materials Science, Graduate School of Frontier Sci-ence, University of Tokyo, Kashiwa-no-ha 5-1-5, Kashiwa-shi,Chiba 277-8651, Japan.

*Present address: Department of Physics, Okayama University,Tsushima-naka, Okayama 700-8530, Japan.†To whom correspondence should be addressed. E-mail:[email protected]


...

......


from least-square fits of the rab(T) curves forT ≤ 200 K to the expression rab(T) = a0 + a1T +a2T

2, whereas the solid circles are obtained fromfits over the same temperature range to a parallel-resistor formalism 1/rab(T) = 1/(a0 + a1T + a2T

2) +1/rmax, where rmax (=900 T 100 mohm•cm) rep-resents a maximum (saturation) resistivity value(19, 20). The inclusion of rmax helps to accountsmoothly for the escalation of rab(T) to highertemperatures and to make the values of a1 and a2insensitive to the temperature range of fitting.

[For a discussion of the physical meaning of rmaxand its relevance to the data, please refer to thesupporting online material (SOM).] For allsamples, the parallel-resistor fits (blue dashedcurves in Fig. 2) are nearly indistinguishable fromthe rab(T) curves over a fitting range that extends(with the exception of LSCO17) over 2 decades intemperature and a factor of 2 in doping. The opensymbols in Fig. 4 relate to coefficients obtainedfrom corresponding fits to earlier zero-field rab(T)data (5) for comparison.

For both sets of data and analysis, a1 is seento grow rapidly with decreasing p, attaining amaximum value of around 1 mohm cm/K at pc =0.185 T 0.005. In the parallel-resistor fit, a2 re-mains finite at all doping concentrations and es-sentially p-independent down to the same dopinglevel, at which point it rises sharply. Without rmaxin the fitting procedure, a2 becomes negligiblearound p = 0.20. It is important to note that noneof the main findings of this study—the ubiquityof the T-linear term, the nondivergence of a2, the

Fig. 1. Electrical resistivity ofoverdoped La2–xSrxCuO4 atlow temperatures. (A) In-planeresistivity rab of LSCO23 as afunction of temperature and mag-netic field. The vertical dashedline indicates the temperaturevariation of rab at 48 T. (B)Mode of extraction of the zero-field resistivity r(H = 0) forLSCO23 at each temperature.The primary data were fitted tothe form r(H) = r(0) + AH2 [asargued in the supporting onlinematerial (SOM) text] betweensome lower bound Hcut-off andthe maximum field strengthused, and the deduced valuesof r(0) are shown plotted as afunction of these Hcut-off values,the sample held at the varioustemperatures indicated. For Hcut-offless than some indicated criticalfield, defined as Hc2, the extractedr(0) values rise monotonicallywith increasing cut-off field.When Hcut-off becomes of orderHc2, r(0) reaches a plateauvalue. Lastly, as the field rangefor fitting becomes too narrow, the extrapolated value of r(0) begins to oscillatewildly. For each temperature, r(0) is identified by its value in the plateauregion. (C) Zero-field rab(T) of LSCO23 (solid green line) plotted with the

extrapolated values of r(0) (green solid squares) and of r(m0H = 48 T) (redopen squares) for the various temperatures indicated. The dashed line is aguide to the eye.

30K

15K6K

10

20

30

40

50

60

0 10 20 30 40 50 60

ρ(0)

( µΩ

cm)

µ0H

cut-off(Tesla)

50K

B

0

20

40

60

80

0 10 20 30 40 50 60 70

C

T (K)

ρ ab(µ

Ωcm

)

H = 48 TH = 0 T

0

10

20

30

40

50

60

0 10 20 30 40 50 60

A

ρ ab( µ

Ωcm

)

µ0H (Tesla)

40K35K30K27K24K21K18K15K12K

10K8K6K

4.2K3.4K2.5K1.5K

Fig. 2. Fitting of the in-planeresistivity of overdoped LSCO.Solid red lines are measuredzero-field rab(T) for (A) LSCO17,(B) LSCO18, (C) LSCO21, (D)LSCO23, (E) LSCO26, (F)LSCO29, and (G) LSCO33;whereas the red diamondsare corresponding extrapo-lated r(H = 0) values. Theblue dashed lines representfits to the data below 200 Kusing the expression 1/ =1/(a0 + a1T + a2T

2) + 1/rmaxwhere rmax = 900 mohm•cm.

0 50 100 150

LSCO26

T (K)

0

50

100

150

200

250

0 50 100 150

LSCO17

ρ ab( µ

Ωcm

)

T (K)0 50 100 150

LSCO18

T (K)0 50 100 150

LSCO21

T (K)0 50 100 150

LSCO23

T (K)0 50 100 150

LSCO29

T (K)0 50 100 150 200

LSCO33

T (K)

A B C D E F G


RESEARCH ARTICLE

...

......


position of the kink, nor the value of a1 at p =pc—are dependent on the type of dual-componentanalysis used.

The coexistence of two distinct componentsto rab(T) is suggestive of some form of two-fluidmodel, although the fact that the two compo-

nents add in the resistivity (i.e., in series) ratherthan in the conductivity (in parallel) implies thatthe two subsystems would have to coexist on amicroscopic scale. Indeed, it is well documentedthat cuprates possess substantial microscopic in-homogeneity (21), and overdoped LSCO in par-ticular is prone to microscopic phase separationinto superconducting and non-superconductingregions with a physical extent on the order ofthe superconducting coherence length (22, 23).One might then be tempted to attribute the T-linear component to the superconducting region(because its value scales roughly with Tc) and thequadratic term to that portion of the sample withan effective carrier density p > 0.27.

An alternative approach, however, is to as-sign the two additive coefficients to two distinct,independent quasi-particle scattering processesthat coexist on the cuprate Fermi surface. Such apicture was recently proposed on the basis of angle-dependent magnetoresistance (ADMR) mea-surements on overdoped Tl2Ba2CuO6+d (Tl2201)that revealed that the scattering rate was com-posed of two distinct terms, a T 2 scattering termthat is isotropic within the basal plane and a T-linear scattering rate that is strongly anisotropic,exhibiting a maximum near the Brillouin zoneboundary, where the pseudogap is maximal andvanishing along the zone diagonal (18). In a sub-sequent doping-dependent study, it was intimatedthat this anisotropic T-linear term correlates withTc, whereas the quadratic term remains constantas a function of p (24). Both of these trends inTl2201 now appear to be confirmed in LSCO butover a much wider range of doping.

Anomalous criticality in LSCO. Critical fluc-tuations are a definitive signature of a QCP. Innearly ferromagnetic (FM) or antiferromagnetic(AFM) metals, for example, critical spin fluctu-ations give rise to physical properties with anom-alous temperature dependences whose criticalexponents depend on the nature of the spin fluc-tuations (FM or AFM) and on the dimensionality.For two-dimensional (2D) AFM spin fluctuations,a T-linear resistivity is expected that extends tovery low temperatures at or near the QCP (25).Away from the critical point, the resistivity showsa crossover to a Fermi-liquid–like T2 resistivity asT decreases. As the critical point is approached,the crossover temperature TF decreases, and thecoefficient of the T2 resistivity diverges. Such be-havior is typified by the heavy-fermion com-pound YbRh2Si2, where Dr is found to be strictlyT-linear over 2 decades in temperature at a criticalmagnetic field, Hc (13). Either side of Hc, Dr isproportional to T2 with a coefficient a2 (TF) thatdiverges (falls linearly) as H approaches Hc. Sim-ilar behavior is exhibited in the quasi-2D heavy-fermion compound CeCoIn5, where the QCPcoincides with the upper critical field Hc2 (14),and in the bilayer ruthenate Sr3Ru2O7 (2), wherethe QCP coincides with a metamagnetic transition.

There are three aspects of the resistivity be-havior in LSCO that conflict with this con-ventional quantum critical picture: (i) the two

Fig. 3. Schematic of theevolution of the exponentn with temperature anddoping in LSCO. The phasediagram is obtaineddirect-ly by interpolating plots ofd[(lnrab–a0)]/d(lnT) versusT for all the resistivity curvesshown in Fig. 2. In markedcontrast to other quantumcritical systems, the T-linearregime is found to growwider with decreasing tem-perature. This unusual ex-pansion of the T-linearregion at low temperaturescoincides with the super-conducting dome (longdashed white line) and the region where superconducting fluctuations become significant (short dashedwhite line). The red dashed line represents T* = Eg/2 (36). As stated in the text and argued in more detail inthe SOM, this form of Dr(T) is shown here only for comparison with the heavy-fermion compounds, but it isnot the most appropriate means of describing Dr(T) in LSCO or indeed other cuprates (18).

n

0

0.002

0.004

0.006

0.15 0.2 0.25 0.3 0.35

α 2(µ

Ωcm

/K2 )

Hole doping p

0

0.5

1

1.5

α 1(µ

Ωcm

/K)

A

B

Fig. 4. Doping evolution of the temperature-dependent coefficients of rab(T). (A) Doping depen-dence of a1, the coefficient of the T-linear resistivity component. (B) Doping dependence of a2, thecoefficient of the T2 resistivity component. In both panels, solid squares are coefficients obtainedfrom least-square fits of the rab(T) curves for T ≤ 200 K to the expression rab(T) = a0 + a1T + a2T

2,whereas the solid circles are obtained from fits over the same temperature range to a parallel-resistor formalism 1/rab(T) = 1/(a0 + a1T + a2T

2) + 1/rmax with rmax = 900 T 100 mohm cm. Theopen symbols are obtained from corresponding fits made to the rab(T) data of Ando et al. (5)between 70 K and 200 K. The dashed lines are guides to the eye. The error bars are a convolutionof standard deviations in the values of a1 and a2 (1s) for different temperature ranges of fittingplus systematic uncertainty in the absolute magnitude of rmax.


RESEARCH ARTICLE

...

......


T-dependent coefficients appear to coexist for allT and p, rather than simply merge into one an-other; (ii) the coefficient of the T2 term does notdiverge—it either remains constant or diminishesas one approaches pc from the overdoped side;and (iii) the T-linear scattering term in LSCOpersists at low T over an extended doping range.This is evident not only in the evolution of a1 (Fig.4A) but also in the single-exponent plot (Fig. 3).This extended regime of T-linear resistivity is apossible manifestation of a second tuning pa-rameter other than doping, that is, a third axis inthe phase diagram (e.g., magnetic field or dis-order) along which the QCP may be located. InYbRh2Si2, for example, alloying with Ge broadensthe field range over which the T-linear resistivityextends to low T (13). Alternatively, the breadthof the critical region in LSCO suggests that thestrangemetal physics in cuprates is associated notwith a QCP but with a novel, extended quantumphase, as seen, for example, in the itinerant fer-romagnet MnSi beyond a critical pressure (26).

Discussion. The marked kinks in a1 and a2at pc = 0.185 T 0.005 are the most revealingfindings of this study. The value of pc coincideswith that inferred from a wealth of other experi-mental probes (3) and is associated with the open-ing of the pseudogap, an anisotropic (nodal) gapin the normal state excitation spectrum. Belowpc, a new energy (E*) or temperature (T*) scaleemerges, which grows sharply with decreasingdoping and appears to extrapolate to a value aboutequal to the AFM exchange energy, J. There aretwo competing scenarios describing the effect ofthe pseudogap on the Fermi surface in cuprates:one involving Fermi surface reconstruction intohole and electron pockets (27) and the other in-volving Fermi surface degradation into a seriesof disconnected Fermi patches or arcs, centeredabout the zone diagonals, with an arc length thatis progressively reduced upon underdoping (28).Although the actual doping dependence of thepseudogap has been hotly debated (3), the con-sistent and systematic analysis reported here ap-pears to confirm that there is indeed a fundamentalchange in the transport properties of cuprates atthis specific, well-defined doping level sandwichedbetween optimal doping and the upper edge ofthe superconducting dome (16, 29).

A particular feature of Fig. 4 is the anticor-relation of a1 and a2 for p < pc; whereas a2 risesor recovers, a1 falls away from its doping tra-jectory inferred from p > pc (thin dashed line inFig. 4A). Recent experiments on less-disorderedcuprates with p ≤ 0.1 have shown that this trendcontinues well into the underdoped regime withrab(T), becoming purely T2 at low to interme-diate temperatures (30). The distinct momentumdependences of the two scattering processes, sug-gested by the ADMR experiments on Tl2201(18), provides a means to interpret qualitativelythe anticorrelation of a1 and a2 within a scenarioof progressive destruction of the underlyingFermi surface (31). As p falls below pc, sectionsof Fermi surface near the Brillouin zone bound-

ary begin to disappear. Although this may not nec-essarily have a detrimental effect on the isotropicterm a2, for a1 the reduction in the total numberof coherent states is more than offset by the re-moval of the strong scattering sinks near (p, 0),leading to an overall decrease, or at best saturation,in a1 with further reduction in carrier density.

Recent angle-resolved photoemission spec-troscopy (ARPES) studies of underdoped cup-rates have hinted at a gradual erosion of the Fermiarc length with decreasing temperatures belowT* (32). For 0.15 ≤ p ≤ 0.19, T* is about equalto E*/2 and is of the order of 100 K (33). Be-cause our fitting range extends up to 200 K, thechanges observed in a1(p) and a2(p) for p < pcimply that actually quasi-particle states are lostat temperatures T > T*. Moreover, given that ourfitting parameters are relatively insensitive to thetemperature range, we conclude that the anti-nodal states must play little or no role in theconductivity at any given temperature. One hinttoward understanding this lies in the absolutevalue of a1 (=1 mohm•cm K–1) at p = pc. For a2D metal, drab/dT = (2pℏd/e2vFkF)d(1/t)/dT,where d is the interlayer spacing (=0.64 nm inLSCO), vF is the Fermi velocity, and kF the Fermiwave vector. Taking typical (p-independent) val-ues of vF (=8.0 × 104 ms−1) and kF (=7 nm−1) forthe antinodal states in overdoped LSCO (34), wefind that at p = pc the momentum-averaged scat-tering rate ℏ/t ~ pkBT. Given that the anisotropicscattering rate varies as cos22ϕ within the plane(ϕ being the angle between the k vector and theCu-O-Cu bond direction) (18), we find that ℏ/t ~2pkBT for states near (p, 0). This level of scatteringintensity is consistent with the so-called Planckiandissipation limit (35) beyond which Bloch-wavepropagation becomes inhibited, that is, the quasi-particle states themselves become incoherent.

The picture that emerges then is the follow-ing. As the carrier number falls and the systemapproaches the Mott insulating state at p = 0, theeffective interaction responsible for the (aniso-tropic) T-linear scattering term becomes progres-sively stronger. Because the strength of this termis closely correlated with Tc, the same interactionalso drives up both the superconducting transi-tion temperature and the condensation energy,the latter reaching a maximum at p = pc (33, 36).At this point, however, scattering intensity isso strong that those states near (p, 0) begin tode-cohere, and, given the proportionality withtemperature, incoherence occurs at all finite tem-peratures, not just below T*. With further re-duction in doping, the strength of the interactioncontinues to rise, causing yet more states to losecoherence and leading to an overall reduction orsaturation in a1(p) and a corresponding suppres-sion in the condensation energy (36) and thesuperfluid density (37).

One is then left with the intriguing possibil-ity that the pseudogap itself forms in response tothis intense scattering, the electronic ground statelowering its energy through gapping out the in-coherent, highly energetic antinodal states and,

in so doing, preventing scattering of the remnantquasi-particle states into those same regions. Thetemperature-invariant violation of the Planckiandissipation limit then provides a possible ex-planation as to why the pseudogap itself is a non–states-conserving gap (3, 33): In contrast to aconventional (Bardeen-Cooper-Schrieffer) super-conducting gap where the full density of statesis recovered above the (two-particle) gap energy,pseudogapped states near (p, 0) are effectivelyremoved at all energies up to the bandwidth (38)or the Coulomb energy (39). Along the nodaldirections, where this anisotropic scattering termis absent, however (18), the nodal quasi-particlestates remain intact (protected) at all T and pacross the entire superconducting dome.

Concluding remarks. It has been a long-standingmystery why the pseudogap in cuprates van-ishes at a singular doping level, pc, beyond op-timal doping. By tracking the evolution of thein-plane resistivity from the overdoped side, wehave uncovered a change in the electrical trans-port at p = pc that appears to coincide with theonset of incoherence (at the antinodal regions),thus offering a possible explanation for its pre-cise location in the cuprate phase diagram andthe subsequent reduction in the strength of thesuperconductivity with lowering carrier concen-tration. And although Planckian dissipation isquantum criticality of sorts, it does not appear tobe consistent with any simple one-parameter scal-ing hypothesis (40) in just the same way as thedoping evolution of the transport coefficients isdifficult to reconcile with conventional quantumcriticality.

References and Notes1. P. Gegenwart, Q. Si, F. Steglich, Nat. Phys. 4, 186

(2008).2. S. A. Grigera et al., Science 294, 329 (2001).3. J. L. Tallon, J. W. Loram, Physica 349C, 53 (2001).4. M. Gurvitch, A. T. Fiory, Phys. Rev. Lett. 59, 1337 (1987).5. Y. Ando, S. Komiya, K. Segawa, S. Ono, Y. Kurita, Phys.

Rev. Lett. 93, 267001 (2004).6. D. van der Marel et al., Nature 425, 271 (2003).7. G. S. Boebinger et al., Phys. Rev. Lett. 77, 5417 (1996).8. H. Takagi et al., Phys. Rev. Lett. 69, 2975 (1992).9. S. Nakamae et al., Phys. Rev. B 68, 100502(R) (2003).

10. Y. Dagan, M. M. Qazilbash, C. P. Hill, V. N. Kulkarni,R. L. Greene, Phys. Rev. Lett. 92, 167001 (2004).

11. A. P. Mackenzie, S. R. Julian, D. C. Sinclair, C. T. Lin,Phys. Rev. B 53, 5848 (1996).

12. R. Daou et al., Nat. Phys. 5, 31 (2009).13. J. Custers et al., Nature 424, 524 (2003).14. J. Paglione et al., Phys. Rev. Lett. 91, 246405 (2003).15. S. Nakatsuji et al., Nat. Phys. 4, 603 (2008).16. S. H. Naqib, J. R. Cooper, J. L. Tallon, C. Panagopoulos,

Physica 387C, 365 (2003).17. C. Proust et al., Phys. Rev. Lett. 89, 147003 (2002).18. M. Abdel-Jawad et al., Nat. Phys. 2, 821 (2006).19. H. Wiesmann et al., Phys. Rev. Lett. 38, 782 (1977).20. N. E. Hussey, Eur. Phys. J. B 31, 495 (2003).21. S. H. Pan et al., Nature 413, 282 (2001).22. Y. Tanabe, T. Adachi, T. Noji, Y. Koike, J. Phys. Soc. Jpn.

74, 2893 (2005).23. Y. Tanabe et al., J. Phys. Soc. Jpn. 76, 113706 (2007).24. M. Abdel-Jawad et al., Phys. Rev. Lett. 99, 107002

(2007).25. T. Moriya, K. Ueda, Adv. Phys. 49, 555 (2000).26. N. Doiron-Leyraud et al., Nature 425, 595 (2003).27. D. LeBoeuf et al., Nature 450, 533 (2007).28. M. R. Norman et al., Nature 392, 157 (1998).


RESEARCH ARTICLE

...

......


29. F. F. Balakirev et al., Phys. Rev. Lett. 102, 017004 (2009).30. F. Rullier-Albenque et al., Phys. Rev. Lett. 99, 027003

(2007).31. T. Senthil, Phys. Rev. B 78, 035103 (2008).32. A. Kanigel et al., Nat. Phys. 2, 447 (2006).33. J. W. Loram, K. A. Mirza, J. R. Cooper, J. L. Tallon, J. Phys.

Chem. Solids 59, 2091 (1998).34. T. Yoshida et al., J. Phys. Condens. Matter 19, 125209

(2007).35. J. Zaanen, Nature 430, 512 (2004).36. J. W. Loram, J. Luo, J. R. Cooper, W. Y. Liang, J. L. Tallon,

J. Phys. Chem. Solids 62, 59 (2001).

37. C. Panagopoulos et al., Phys. Rev. B 67, 220502 (2003).38. H. J. A. Molegraaf, C. Presura, D. van der Marel, P. H. Kes,

M. Li, Science 295, 2239 (2002).39. S. Chakraborty, D. Galanakis, P. Phillips, http://arxiv.org/

abs/0807.2854 (2008).40. P. Phillips, C. Chamon, Phys. Rev. Lett. 95, 107002

(2005).41. We acknowledge technical and scientific assistance from

S. L. Kearns, J. Levallois, and N. Mangkorntang andcollaborative support from H. H. Wen. This work wassupported by Engineering and Physical Sciences ResearchCouncil (UK), the Royal Society, Laboratoire National des

Champs Magnétiques Pulsés, the French AgenceNationale de la Recherche IceNET, and EuroMagNET.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/1165015/DC1Materials and MethodsFigs. S1 and S2References

22 August 2008; accepted 21 November 2008Published online 11 December 2008;10.1126/science.1165015Include this information when citing this paper.

REPORTS

Revealing the Maximum Strengthin Nanotwinned CopperL. Lu,1* X. Chen,1 X. Huang,2 K. Lu1

The strength of polycrystalline materials increases with decreasing grain size. Below a critical size, smallergrains might lead to softening, as suggested by atomistic simulations. The strongest size should arise at atransition in deformation mechanism from lattice dislocation activities to grain boundary–relatedprocesses. We investigated the maximum strength of nanotwinned copper samples with different twinthicknesses. We found that the strength increases with decreasing twin thickness, reaching a maximum at15 nanometers, followed by a softening at smaller values that is accompanied by enhanced strain hardeningand tensile ductility. The strongest twin thickness originates from a transition in the yielding mechanismfrom the slip transfer across twin boundaries to the activity of preexisting easy dislocation sources.

The strength of polycrystalline materialsincreases with decreasing grain size, asdescribed by the well-known Hall-Petch

relation (1, 2). The strengthening originates fromthe fact that grain boundaries block the latticedislocation motion, thereby making plastic defor-mationmore difficult at smaller grain sizes. How-ever, below a certain critical size, the dominatingdeformation mechanism may change from latticedislocation activities to othermechanisms such asgrain boundary–related processes, and softeningbehavior (rather than strengthening) is expected(3, 4). Such a softening phenomenon has beendemonstrated by atomistic simulations, and a crit-ical grain size of maximum strength has beenpredicted (5–7). In pure metals, an impediment todetermining the grain size that yields the higheststrength is the practical difficulty of obtaining sta-ble nanostructures with extremely small structuraldomains (on the order of several nanometers). Thedriving force for growth of nanosized grains inpure metals, originating from the high excess en-ergy of numerous grain boundaries, becomes solarge that grain growthmay take place easily evenat ambient temperature or below.

Coherent twin boundaries (TBs), which aredefined in a face-centered cubic structure as the(111) mirror planes at which the normal stackingsequence of (111) planes is reversed, are known tobe as effective as conventional grain boundariesin strengthening materials. Strengthening has beenobtained in Cu when high densities of nanometer-thick twins are introduced into submicrometer-sized grains (8–10). In addition, coherent TBs aremuch more stable against migration (a fundamen-tal process of coarsening) than conventional grainboundaries, as the excess energy of coherent TBs isone order of magnitude lower than that of grainboundaries. Hence, nanotwinned structures areenergetically more stable than nanograined coun-terparts with the same chemical composition. Thestable nanotwinned structure may provide samplesfor exploring the softening behavior with very smalldomain sizes. Here, we prepared nanotwinned pureCu (nt-Cu) samples with average twin thicknessranging from a few nanometers to about 100 nm.

High-purity (99.995%) Cu foil samples com-posed of nanoscale twin lamellae embedded insubmicrometer-sized grains were synthesized bymeans of pulsed electrodeposition. By increasingthe deposition rate to 10 nm/s, we succeeded inrefining the mean twin thickness (i.e., the meanspacing between adjacent TBs, hereafter referredto as l) from a range of 15 to 100 nm down to arange of 4 to 10 nm (see supporting online ma-terial). The as-deposited Cu foils have an in-planedimension of 20 mm by 10 mm and a thicknessof 30 mm with a uniform microstructure. Shown

in Fig. 1, A to C, are transmission electron mi-croscopy (TEM) plane-view images of three as-deposited samples with l values of 96 nm, 15 nm,and 4 nm, respectively. The TEM images indicatethat some grains are irregular in shape, but low-magnification scanning electronic microscopyimages, both cross section and plane view, showthat the grains are roughly equiaxed in three di-mensions. Grain size measurements showed asimilar distribution and a similar average diam-eter of about 400 to 600 nm for all nt-Cu samples.Twins were formed in all grains (see the electrondiffraction pattern in Fig. 1D), and observationsof twins in a large number of individual grainsrevealed no obvious change in the twin densityfrom grain to grain. Note that in all samples, theedge-on twins that formed in different grains arealigned randomly around the foil normal (growth)direction (8, 11), in agreement with a strong [110]texture determined by x-ray diffraction (XRD). Foreach sample, twin thicknessesweremeasured froma large number of grains, whichwere detected fromnumerousTEMandhigh-resolutionTEM(HRTEM)images, to generate a distribution. Figure 1E illus-trates the 492 measurements for the sample withthe finest twins; the majority yielded spacings be-tween twins smaller than 10 nm, with a mean of4 nm. For simplicity, each nt-Cu sample is iden-tified by its mean twin thickness; for example, thesample with l = 4 nm is referred to as nt-4.

Figure 2 shows the uniaxial tensile true stress–true strain curves for nt-Cu samples of various lvalues. Also included are two stress-strain curvesobtained from a coarse-grained Cu (cg-Cu) andan ultrafine-grained Cu (ufg-Cu) that has a sim-ilar grain size to that of nt-Cu samples but is freeof twins within grains. Two distinct features areobserved with respect to the l dependence of themechanical behavior of nt-Cu. The first is the oc-currence of the l giving the highest strength. Allstress-strain curves of nt-Cu samples in Fig. 2, Aand B, are above that of the ufg-Cu, indicating astrengthening by introducing twins into the sub-micrometer grains. However, such a strengthen-ing does not show a linear relationship with l. Forl > 15 nm (Fig. 2A), the stress-strain curves shiftupward with decreasing l, similar to the strength-ening behavior reported previously in the nt-Cu(9, 11) and nanocrystalline Cu (nc-Cu) (12–15)samples (Fig. 3A). However, with further de-

1Shenyang National Laboratory for Materials Science, Instituteof Metal Research, Chinese Academy of Sciences, Shenyang110016, P.R. China. 2Center for Fundamental Research: MetalStructures in Four Dimensions, Materials Research Department,Risø National Laboratory for Sustainable Energy, Technical Uni-versity of Denmark, DK-4000 Roskilde, Denmark.



...

......


creases of l down to extreme dimensions (i.e.,less than 10 nm), the stress-strain curves shift down-ward (Fig. 2B). As plotted in Fig. 3A, the mea-sured yield strength sy (at 0.2% offset) shows amaximum value of 900 MPa at l ≈ 15 nm.

The second feature is a substantial increase intensile ductility and strain hardening when l < 15nm. As seen in Fig. 2, the tensile elongation ofthe nt-Cu samples increases monotonically withdecreasing l. When l < 15 nm, the uniform ten-sile elongation exceeds that of the ufg-Cu sample,reaching a maximum value of 30% at the finesttwin thickness. Strain-hardening coefficient (n)values were determined for each sample by fittingthe uniform plastic deformation region to s =K1 +K2e

n, where K1 represents the initial yield stressand K2 is the strengthening coefficient (i.e., thestrength increment due to strain hardening at straine = 1) (16, 17). The n values determined for all thent-Cu samples increase monotonically with de-creasing l (Fig. 3B), similar to the trend of uniformelongation versus l. When l < 15 nm, n exceedsthe value for cg-Cu (0.35) (16, 17) and finallyreaches amaximumof 0.66 at l = 4 nm. The twinrefinement–induced increase in n is opposite tothe general observation in ultrafine-grained andnanocrystalline materials, where n continuouslydecreases with decreasing grain size (Fig. 3B).

The strength of the nt-Cu samples has beenconsidered to be controlled predominantly by thenanoscale twins via the mechanism of slip trans-fer across the TBs (10, 18), and it increases withdecreasing l in a Hall-Petch–type relationship(9) similar to that of grain boundary strength-ening in nanocrystalline metals (12). Our re-sults show that such a relationship breaks downwhen l < 15 nm, although other structural pa-rameters such as grain size and texture are un-changed. The grain sizes of the nt-Cu samplesare in the submicrometer regime, which is toolarge for grain boundary sliding to occur at roomtemperature, as expected for nanocrystalline ma-terials with grain sizes below 20 nm (3). There-fore, the observed softening cannot be explainedby the initiation of grain boundary–mediatedmechanisms such as grain boundary sliding andgrain rotation, as proposed by molecular dynam-ics (MD) simulations for nanocrystalline mate-rials (3).

To explore the origin of the twin thickness giv-ing the highest strength, we carried out detailedstructural characterization of the as-deposited sam-ples. HRTEM observations showed that in eachsample TBs are coherent S3 interfaces associatedwith the presence of Shockley partial dislocations(as steps), as indicated in Fig. 1D. These partialdislocations have their Burgers vector parallel tothe twin plane and are an intrinsic structural fea-ture of twin growth during electrodeposition. Thedistribution of the preexisting partial dislocationsis inhomogeneous, but their density per unit areaof TBs is found to be rather constant among sam-ples with different twin densities. This suggeststhat the deposition parameters and the twin re-finement have a negligible effect on the nature of

TBs. Therefore, as a consequence of decreasingl, the density of such TB-associated partialdislocations per unit volume increases.

We also noticed that grain boundaries in thent-Cu samples with l ≤ 15 nm are characterizedby straight segments (facets) that are often

0 5 10 15 20 25 30 350 5 10 150

200

400

600

800

1000

1200

ufg

True strain (%)

Tru

e st

ress

(M

Pa)

nt-15

nt-10

ufg

True strain (%)

nt-8

nt-4

cg

nt-96

nt-35

nt-15A B

Fig. 2. Uniaxial tensile true stress–true strain curves for nt-Cu samples tested at a strain rate of 6 × 10−3 s−1.(A) Curves for samples with mean twin thickness varying from 15 to 96 nm; (B) curves for samples withmean twin thickness varying from 4 to 15 nm. For comparison, curves for a twin-free ufg-Cu with a meangrain size of 500 nm and for a cg-Cu with a mean grain size of 10 mm are included.

300 nm

C 10 nm

0 10 20 30 40 50 600

5

10

15

20

25

30

Fre

quen

cy (

%)Twin thickness (nm)

E

200nm

200 nm

A

B

bb

D

Fig. 1. TEM images of as-deposited Cu samples with variousmean twin thicknesses. (A) l = 96 nm. (B) l =15 nm. (C) l = 4 nm. (D) The same sample as (C) but at higher resolution, with a corresponding electrondiffraction pattern (upper right inset) and a HRTEM image of the outlined area showing the presence ofShockley partials at the TB (lower right inset). (E) Distribution of the lamellar twin thicknesses determinedfrom TEM and HRTEM images for l = 4 nm.


REPORTS

...

......


associated with dislocation arrays (19), whereasin samples with coarser twins, grain boundariesare smoothly curved, similar to conventionalgrain boundaries. The microstrain measured byXRD was a negligible 0.01% for samples with l ≥15 nm, but increased gradually from 0.038% to0.057% when l decreased from 10 to 4 nm,which also indicates a gradual increase in the de-fect density.

Recent experimental studies and MD simula-tions (3, 20, 21) have shown that an increase inthe density of preexisting dislocations in nano-scale materials will cause softening. In the nt-Cusamples studied, both the dislocation arrays asso-ciated with the grain boundaries and the stepsassociated with the preexisting partial dislocationsalong TBs could be potential dislocation sources,which are expected to affect the initiation of plas-tic deformation (22) and to provide the disloca-tions required for the dislocation-TB interactionsthat cause work hardening. The preexisting par-

tial dislocations can act as readily mobile dislo-cations, and their motion may contribute to theplastic yielding when an external stress is appliedto the sample. The plastic strains induced by themotion of preexisting partial dislocations can beestimated as e = r0bsd/M (where r0 is the initialdislocation density, bs is the Burgers vector ofShockley partial dislocation, d is the grain size,andM is the Taylor factor). Calculations showedthat for the samples with l > 15 nm, the preexist-ing dislocations induce a negligibly small plasticstrain (<0.05%). However, for the nt-4 specimen,a remarkable amount of plastic strain, as high as0.1 to 0.2%, can be induced just by the motionsof high-density preexisting dislocations at TBs(roughly 1014 m−2), which could control the mac-roscopic yielding of the sample. The above anal-ysis suggests that for extremely small values of l,a transition in the yielding mechanism can resultin an unusual softening phenomenon in which thepreexisting easy dislocation sources at TBs and

grain boundaries dominate the plastic deforma-tion instead of the slip transfer across TBs.

Shockley partial dislocations are always in-volved in growth of twins during crystal growth,thermal annealing, or plastic deformation. Shock-ley partials might be left at TBs when the twingrowth is interrupted. Therefore, the presence ofShockley partials at some TBs is a natural phe-nomenon. Although these preexisting dislocationsmay have a small effect on the mechanical be-havior of the samples with thick twins, the effectwill be much more pronounced in the sampleswith nanoscale twins and/or with high preexist-ing TB dislocation densities such as those seen indeformation twins (23).

To understand the extraordinary strain hard-ening, we analyzed the deformation structures ofthe tensile-deformed samples. In samples withcoarse twins, tangles and networks of perfect dis-locations were observed within the lattice betweenthe TBs (Fig. 4A), and the dislocation density wasestimated to be on the order of 1014 to 1015 m−2.In contrast, high densities of stacking faults andShockley partials associated with the TBs werefound to characterize the deformed structure ofthe nt-4 sample (Fig. 4, B and C), indicating theinteractions between dislocations and TBs. RecentMD simulations (18, 24, 25) showed that whenan extended dislocation (two Shockley partialsconnected by a stacking fault ribbon) is forced byan external stress into a coherent TB, it recom-bines or constricts into a perfect dislocation con-figuration at the coherent TB and then slips throughthe boundary by splitting into three Shockley par-tials. Two of them glide in the slip plane of theadjacent twin lamella, constituting a new extendeddislocation, whereas the third one, a twinning par-tial, glides along the TB and forms a step. It isexpected that with increasing strain, such an in-teraction process will generate a high density ofpartial dislocations (steps) along TBs and stack-ing faults that align with the slip planes in thetwin lamellae, which may (or may not) connectto the TBs. Such a configuration of defects wasobserved, as shown in Fig. 4C. The density ofpartial dislocations in the deformed nt-4 sample

Fig. 3. Variation of (A)yield strength and (B)strain hardening coef-ficient n as a functionof mean twin thicknessfor the nt-Cu samples.For comparison, the yieldstrength andn values fornc-Cu [ (12), (13), (14), and (15)],ufg-Cu [ (9)], andcg-Cusamples reported in theliterature are included.Amaximum in the yieldstress is seen for thent-Cu with l = 15 nm,but this has not beenobserved for the nc-Cu, even when the grain size is as small as 10 nm.

0.0

0.2

0.4

0.6

0.8

nt-Cu nc- & ufg-Cu

n

or d (nm)

B

cg-Cu: n=0.35

0

200

400

600

800

1000

σ y (M

Pa)

λ or d (nm)

nt-Cu nc-Cu

A

0 2 0 4 0 6 0 8 0 1 0 0 1 2 0 1 1 0 1 0 0 1 0 0 0 1 0 0 0 0

λ

2 nm

TT

T

SF

SF

SF

SF

SF

SF

TB

TB

TB

SFB

2 nm

SF SF

SF SF

SF

SF

SF SF

SF

200 nm

A C

Fig. 4. (A) A typical bright TEM image of the deformed nt-96 sample showing the tangling of latticedislocations. (B) An HRTEM image of the nt-4 sample tensile-deformed to a plastic strain of 30%, showinga high density of stacking faults (SF) at the TB. (C) The arrangement of Shockley partials and stackingfaults at TBs within the lamellae in the nt-4 sample. Triangles, Shockley partial dislocations associatedwith stacking faults; ⊥, partials with their Burgers vector parallel to the TB plane.


REPORTS

...

......


was estimated to be 5 × 1016 m−2 on the basis ofthe spacing between the neighboring partials andl. This is two orders of magnitude higher thanthat of the preexisting dislocations and the latticedislocations stored in the coarse twins. Such afinding suggests that decreasing the twin thick-ness facilitates the dislocation-TB interactions andaffords more room for storage of dislocations,which sustain more pronounced strain hardeningin the nt-Cu (26, 27).

These observations suggest that the strain-hardening behavior of nt-Cu samples is governedby two competing processes: dislocation-dislocationinteraction hardening in coarse twins, anddislocation-TB interaction hardening in fine twins. With arefining of l, the contribution from the latter mech-anism increases and eventually dominates the strainhardening, as revealed by the continuous increaseof n values (Fig. 3B). However, the former hard-eningmechanism usually leads to an inverse trend,diminishing with size refinement (17).

Twins are not uncommon in nature, and theyappear in various metals and alloys with differentcrystallographic structures. Extremely thin twinlamellae structures can possibly be achieved underproper conditions during crystal growth, plasticdeformation, phase transformations, or thermalannealing of deformed structures. Our finding ofthe twin thickness giving maximum strength il-

lustrates that the scale-dependent nature of plasticdeformation of nanometer-scale materials is notnecessarily related to grain boundary–mediatedprocesses. This finding also provides insight intothe development of advanced nanostructuredmaterials.

References and Notes1. E. O. Hall, Proc. Phys. Soc. London Ser. B 64, 747 (1951).2. N. J. Petch, J. Iron Steel Inst. 174, 25 (1953).3. J. Schiøtz, K. W. Jacobsen, Science 301, 1357 (2003).4. S. Yip, Nature 391, 532 (1998).5. M. A. Meyers, A. Mishra, D. J. Benson, Prog. Mater. Sci.

51, 427 (2006).6. P. G. Sanders, J. A. Eastman, J. R. Weertman, Acta Mater.

45, 4019 (1997).7. C. C. Koch, K. M. Youssef, R. O. Scattergood, K. L. Murty,

Adv. Eng. Mater. 7, 787 (2005).8. L. Lu et al., Acta Mater. 53, 2169 (2005).9. Y. F. Shen, L. Lu, Q. H. Lu, Z. H. Jin, K. Lu, Scr. Mater. 52,

989 (2005).10. X. Zhang et al., Acta Mater. 52, 995 (2004).11. L. Lu, Y. Shen, X. Chen, L. Qian, K. Lu, Science 304,

422 (2004); published online 18 March 2004(10.1126/science.1092905).

12. J. Chen, L. Lu, K. Lu, Scr. Mater. 54, 1913 (2006).13. S. Cheng et al., Acta Mater. 53, 1521 (2005).14. Y. Champion et al., Science 300, 310 (2003).15. Y. M. Wang et al., Scr. Mater. 48, 1851 (2003).16. A. Misra, X. Zhang, D. Hammon, R. G. Hoagland,

Acta Mater. 53, 221 (2005).17. M. A. Meyers, K. K. Chawla, in Mechanical Behavior of

Materials, M. Horton, Ed. (Prentice Hall, Upper SaddleRiver, NJ, 1999), pp. 112–135.

18. Z. H. Jin et al., Scr. Mater. 54, 1163 (2006).19. X. H. Chen, L. Lu, K. Lu, J. Appl. Phys. 102, 083708

(2007).20. X. Huang, N. Hansen, N. Tsuji, Science 312, 249

(2006).21. Z. W. Shan, R. K. Mishra, S. A. Syed Asif, O. L. Warren,

A. M. Minor, Nat. Mater. 7, 115 (2008).22. K. Konopka, J. Mizera, J. W. Wyrzykowski, J. Mater.

Process. Technol. 99, 255 (2000).23. Y. S. Li, N. R. Tao, K. Lu, Acta Mater. 56, 230 (2008).24. S. I. Rao, P. M. Hazzledine, Philos. Mag. A 80, 2011

(2000).25. Z. H. Jin et al., Acta Mater. 56, 1126 (2008).26. M. Dao, L. Lu, Y. Shen, S. Suresh, Acta Mater. 54, 5421

(2006).27. T. Zhu, J. Li, A. Samanta, H. G. Kim, S. Suresh, Proc. Natl.

Acad. Sci. U.S.A. 104, 3031 (2007).28. Supported by National Natural Science Foundation of

China grants 50431010, 50621091, 50725103, and50890171, Ministry of Science and Technology of Chinagrant 2005CB623604, and the Danish National ResearchFoundation through the Center for FundamentalResearch: Metal Structures in Four Dimensions (X.H.).We thank N. Hansen, Z. Jin, W. Pantleon, and B. Ralphfor stimulating discussions, X. Si and H. Ma for samplepreparation, S. Zheng for TEM observations, and Y. Shenfor conducting some of the tensile tests.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/323/5914/607/DC1Materials and MethodsTable S1References

24 October 2008; accepted 30 December 200810.1126/science.1167641

Control of Graphene’s Propertiesby Reversible Hydrogenation:Evidence for GraphaneD. C. Elias,1* R. R. Nair,1* T. M. G. Mohiuddin,1 S. V. Morozov,2 P. Blake,3 M. P. Halsall,1A. C. Ferrari,4 D. W. Boukhvalov,5 M. I. Katsnelson,5 A. K. Geim,1,3 K. S. Novoselov1†

Although graphite is known as one of the most chemically inert materials, we have found thatgraphene, a single atomic plane of graphite, can react with atomic hydrogen, which transforms thishighly conductive zero-overlap semimetal into an insulator. Transmission electron microscopyreveals that the obtained graphene derivative (graphane) is crystalline and retains the hexagonallattice, but its period becomes markedly shorter than that of graphene. The reaction with hydrogenis reversible, so that the original metallic state, the lattice spacing, and even the quantum Halleffect can be restored by annealing. Our work illustrates the concept of graphene as a robustatomic-scale scaffold on the basis of which new two-dimensional crystals with designed electronicand other properties can be created by attaching other atoms and molecules.

Graphene, a flat monolayer of carbon atomstightly packed into a honeycomb lattice,continues to attract immense interest, most-

ly because of its unusual electronic propertiesand effects that arise from its truly atomic thick-ness (1). Chemical modification of graphene hasbeen less explored, even though research on car-bon nanotubes suggests that graphene can be al-tered chemically without breaking its resilient C-Cbonds. For example, graphene oxide is graphenedensely covered with hydroxyl and other groups(2–6). Unfortunately, graphene oxide is stronglydisordered, poorly conductive, and difficult to

reduce to the original state (6). However, one canimagine atoms or molecules being attached tothe atomic scaffold in a strictly periodic manner,which should result in a different electronic struc-ture and, essentially, a different crystalline mate-rial. Particularly elegant is the idea of attachingatomic hydrogen to each site of the graphenelattice to create graphane (7), which changes thehybridization of carbon atoms from sp2 into sp3,thus removing the conducting p-bands and open-ing an energy gap (7, 8).

Previously, absorption of hydrogen on gra-phitic surfaces was investigated mostly in con-

junction with hydrogen storage, with the researchfocused on physisorbed molecular hydrogen(9–11). More recently, atomic hydrogen chem-isorbed on carbon nanotubes has been studiedtheoretically (12) as well as by a variety of exper-imental techniques including infrared (13), ultra-violet (14, 15), and x-ray (16) spectroscopy andscanning tunnelingmicroscopy (17).We report thereversible hydrogenation of single-layer grapheneand observed dramatic changes in its transportproperties and in its electronic and atomic struc-ture, as evidenced by Raman spectroscopy andtransmission electron microscopy (TEM).

Graphene crystals were prepared by use ofmicromechanical cleavage (18) of graphite ontop of an oxidized Si substrate (300 nm SiO2) andthen identified by their optical contrast (1, 18)and distinctive Raman signatures (19). Three typesof samples were used: large (>20 mm) crystalsfor Raman studies, the standard Hall bar de-vices 1 mm in width (18), and free-standing mem-branes (20, 21) for TEM. For details of samplefabrication, we refer to earlier work (18, 20, 21).

1School of Physics and Astronomy, University of Manchester,M13 9PL, Manchester, UK. 2Institute for Microelectronics Tech-nology, 142432 Chernogolovka, Russia. 3Manchester Centrefor Mesoscience and Nanotechnology, University of Manches-ter, M13 9PL, Manchester, UK. 4Department of Engineering,Cambridge University, 9 JJ Thomson Avenue, Cambridge CB3OFA, UK. 5Institute for Molecules and Materials, RadboudUniversity Nijmegen, 6525 ED Nijmegen, Netherlands.

*These authors contributed equally to this work.†To whom correspondence should be addressed. E-mail:[email protected]


REPORTS

...

......


We first annealed all samples at 300°C in anargon atmosphere for 4 hours in order to removeany possible contamination (for example, resistresidues). After their initial characterization, thesamples were exposed to a cold hydrogen plas-ma. We used a low-pressure (0.1 mbar) hydrogen-argon mixture (10% H2) with dc plasma ignitedbetween two aluminum electrodes. The sampleswere kept 30 cm away from the discharge zonein order to minimize any possible damage byenergetic ions. We found that it typically re-quired 2 hours of plasma treatment to reach thesaturation in measured characteristics. As a ref-erence, we used graphene samples exposed toa pure Ar plasma under the same conditions,which showed little changes in their transportand Raman properties (22).

Typical changes induced by the hydrogena-tion in electronic properties of graphene are il-lustrated in Fig. 1. Before plasma exposure, ourdevices exhibited the standard ambipolar fieldeffect with the neutrality point (NP) near zerogate voltage (18). For the device shown in Fig. 1,mobility m of charge carriers was ≈14,000 cm2

V–1 s–1. This device exhibits a weak temper-ature dependence of its resistivity at all gatevoltages (not visible on the scale of Fig. 1A).We observed metallic dependence close to theNP below 50 K (23) and the half-integer quan-tum Hall effect (QHE) at cryogenic temperatures(Fig. 1B), both of which are hallmarks of single-layer graphene [(1) and references therein].

This behavior completely changed after thedevices were treated with atomic hydrogen (Fig. 1,C and D). The devices exhibited an insulatingbehavior such that the resistivity r grew by twoorders of magnitude with decreasing tempera-ture T from 300 to 4 K (Fig. 1C). Carrier mo-bility decreased at liquid-helium temperaturesdown to values of ~10 cm2 V–1 s–1 for typical car-rier concentrations n of the order of 1012 cm−2.The quantum Hall plateaus, so marked in theoriginal devices, completely disappeared, withonly weak signatures of Shubnikov–de-Haas os-cillations remaining in magnetic field B of 14 T(Fig. 1D). In addition, we observed a shift of NPto gate voltages Vg ≈ +50 V, which showed thatgraphene became doped with holes in concen-tration of ≈3 × 1012 cm−2 (probably due to ad-sorbed water). At carrier concentrations of lessthan 3 × 1012 cm−2, the observed temperaturedependences r(T) can be well fitted by thefunction exp[(T0/T)

1/3] (T0 is the parameter thatdepends on Vg) (Fig. 2), which is a signature ofvariable-range hopping in two dimensions (24).T0 exhibits a maximum at NP of ~250 K andstrongly decreases away from NP (Fig. 2B). Atn > 4 × 1012 cm−2 (for both electrons and holes),changes in r with T became small (similar to thosein pristine graphene), which indicates a transi-tion from the insulating to the metallic regime.

The hydrogenated devices were stable atroom T for many days and showed the samecharacteristics during repeated measurements.However, we could restore the original metallic

state by annealing (we used 450°C in Ar atmo-sphere for 24 hours; higher annealing T dam-aged graphene). After the annealing, the devicesreturned practically to the same state as beforehydrogenation: r as a function of Vg reachedagain a maximum value of ≈h/4e2, where h isPlanck’s constant and e is the electron charge,and became only weakly T-dependent (Figs. 1Eand 2). Also, m recovered to ~3500 cm2 V–1 s–1,and the QHE reappeared (Fig. 1F). Still, the

recovery was not complete: Graphene remainedp-doped, the QHE did not restore at fillingfactors n larger than T2 (compare Figure 1, Band F), and zero–B field conductivity s (=1/r)became a sublinear function of n, which indi-cates an increased number of short-range scat-terers (23). We attribute the remnant features tovacancies induced by plasma damage or residualoxygen during annealing. To this end, after an-nealing, the distance (as a function of Vg) between

Fig. 1. Control of the electronic properties of graphene by hydrogenation. The electric field effect forone of our devices at zero B at various temperatures T (left column) and in B = 14 T at 4 K (right). (Aand B) The sample before its exposure to atomic hydrogen; curves in (A) for three temperatures (40, 80,and 160 K) practically coincide. (C and D) After atomic hydrogen treatment. In (C), temperatureincreases from the top; T = 4, 10, 20, 40, 80, and 160 K. (E and F) The same sample after annealing.(E) T = 40, 80, and 160 K, from top to bottom. (Inset) Optical micrograph of a typical Hall bar device.The scale is given by its width of 1 mm.

Fig. 2. Metal-insulator transition inhydrogenated graphene. (A) Tem-perature dependence of graphene’sresistivity at NP for the sample shownin Fig. 1. Red circles, blue squares,and green triangles are for pristine,hydrogenated, and annealed graphene,respectively. The solid line is a fit bythe variable-range hopping depen-dence exp[(T0/T)

1/3]. (B) Characteristicexponents T0 found from this fittingat different carrier concentrations.


REPORTS

...

......


the peaks in rxx at n = 0 and n = T4 becamenotably greater (~40%) than that between all theother peaks for both annealed and original de-vices. The greater distance indicates the pres-ence of mid-gap states (25) [such as vacancies(26)] that were induced during the processing,which was in agreement with the observed sub-linear behavior of the conductivity. The extracharge required to fill these states (25) yields theirdensity as of about 1 × 1012 cm−2 (with an aver-age spacing of ≈10 nm).

The changes induced by hydrogenation havebeen corroborated by Raman spectroscopy. Themain features in the Raman spectra of carbon-based materials are the G and D peaks that lie ataround 1580 and 1350 cm−1, respectively. The Gpeak corresponds to optical E2g phonons at theBrillouin zone center, whereas the D peak iscaused by breathing-likemodes (corresponding totransverse optical phonons near the K point) andrequires a defect for its activation via an interval-ley double-resonance Raman process (19, 27–29).

Both the G and D peaks arise from vibrations ofsp2-hybridized carbon atoms. The D peak inten-sity provides a convenient measure for the amountof disorder in graphene (27–29). Its overtone, the2D peak, appears around 2680 cm−1 and its shapeidentifies monolayer graphene (19). The 2D peakis present even in the absence of any defects be-cause it is the sum of two phonons with oppositemomentum. In Fig. 3, there is also a peak at~1620 cm−1, called D′, which occurs via an in-travalley double-resonance process in the pres-ence of defects.

Figure 3A shows the evolution of Ramanspectra for graphene crystals that are hydrogenatedand annealed simultaneously with the device inFig. 1 (the use of different samples for Ramanstudies was essential to avoid an obscuring con-tribution to the D and D′ peaks caused by theedges of the Hall bars, which were smaller thanour laser spot size of about 1 mm). Hydrogenationresulted in the appearance of sharp D and D′peaks, slight broadening and a decrease of theheight of the 2D peak relative to the G peak, andthe onset of a combination mode (D + D′) around2950 cm−1, which, unlike the 2D and 2D′ bands,requires a defect for its activation because it is acombination of two phonons with different mo-mentum. The D peak in hydrogenated graphene isobserved at 1342 cm−1 and is very sharp, as com-pared with that in disordered or nanostructuredcarbon-based materials (29). We attribute theactivation of this sharpDpeak in our hydrogenatedsamples to breaking of the translational symmetryof C-C sp2 bonds after the formation of C-H sp3

bonds. Although the majority of carbon bonds inhydrogenated graphene are expected to acquire sp3

hybridization, we do not expect to see any Ramansignature of C-C sp3 bonds because their crosssection at visible light excitation is negligible ascompared with that of the resonant C-C sp2 bonds,and therefore even a small residual sp2 phaseshould generally dominate our spectra, as happensin other diamondlike compounds (22, 29).

After annealing, the Raman spectrum recov-ered to almost its original shape, and all of the

Fig. 4. Structural studies of graphane via TEM [we used a Tecnai F30 (FEI,Eindhoven, the Netherlands)]. (A) Changes in the electron diffraction after~4 hours exposure of graphene membranes to atomic hydrogen. Scale bar,5 nm−1. The blue hexagon is a guide to the eye and marks positions of thediffraction spots in graphane. The equivalent diffraction spots in grapheneunder the same conditions are shown by the red hexagon. (B) Distribution of

the lattice spacing d found in hydrogenated membranes. The green dashedline marks the average value, whereas the red solid line shows d alwaysobserved for graphene (both before hydrogenation and after annealing). (C andD) Schematic representation of the crystal structure of graphene and theo-retically predicted graphane. Carbon atoms are shown as blue spheres, andhydrogen atoms are shown as red spheres.

Fig. 3. Changes in Raman spectra of graphene caused by hydrogenation. The spectra are normalized tohave a similar intensity of the G peak. (A) Graphene on SiO2. (B) Free-standing graphene. Red, blue,and green curves (top to bottom) correspond to pristine, hydrogenated, and annealed samples, re-spectively. Graphene was hydrogenated for ~2 hours, and the spectra were measured with a Renishawspectrometer at wavelength 514 nm and low power to avoid damage to the graphene during mea-surements. (Left inset) Comparison between the evolution of D and D′ peaks for single- and double-sided exposure to atomic hydrogen. Shown is a partially hydrogenated state achieved after 1 hour ofsimultaneous exposure of graphene on SiO2 (blue curve) and of a membrane (black curve). (Right inset)TEM image of one of our membranes that partially covers the aperture 50 mm in diameter.


REPORTS

...

......


defect-related peaks (D, D′, and D+D′) werestrongly suppressed. However, two broad low-intensity bands appeared, overlapping a sharperG and residual D peaks. These bands are indi-cative of some residual structural disorder (29).The 2D peak remained relatively small with re-spect to the G peak when compared with thesame ratio in the pristine sample, and both be-came shifted to higher energies, indicating thatthe annealed graphene is p-doped (30). The ob-served changes in Raman spectra are in broadagreement with our transport measurements.

For graphene on a substrate, only one side isaccessible to atomic hydrogen, and the plasmaexposure is not expected to result in graphane(which assumes hydrogen atoms attached onboth sides). For more effective hydrogenation,we employed free-standing graphene membranes(Fig. 3B, inset) (20, 21). The experiments de-scribed below refer to membranes that had somefree edges to facilitate the relaxation of straininduced by hydrogenation [membranes with allthe sides fixed to a metal scaffold are discussedin (22)]. Raman spectra for hydrogenated andsubsequently annealed membranes (Fig. 3B)were rather similar to those described above forgraphene on SiO2, but with some notable dif-ferences. If hydrogenated simultaneously and be-fore reaching the saturation, the D peak for amembrane was by a factor of two greater thanthat for graphene on a substrate (Fig. 3A, inset),which indicates the formation of twice as manyC-H bonds in the membrane. This result agreeswith the general expectation that atomic hydrogenattaches to both sides of membranes. Moreover,the D peak could become up to three times greaterthan the G peak after prolonged exposures ofmembranes to atomic hydrogen (Fig. 3B).

Further information about hydrogenated mem-branes was obtained with TEM. For graphene,the electron-diffraction (ED) patterns observed ondozens of the studied membranes were alwaysthe same, exhibiting the hexagonal symmetry withthe lattice constant d = 2.46 T 0.02 Å. Prolongedexposure to atomic hydrogen preserved the hex-agonal symmetry and hence crystallinity, but ledto drastic changes in the lattice constant d, whichcould decrease by as much as 5% (Fig. 4A). Gen-erally, the compression was not uniform, anddifferent parts of membranes exhibited locallydifferent in-plane periodicities (Fig. 4B; diame-ters of the selected area for the ED and studiedmembranes were 0.3 mm and 30 to 50 mm, re-spectively). Such nonuniformity is generally notunexpected because the crystals were fixed tothe scaffold (Fig. 3) that restricted their isotropicshrinkage. We found that the more extended freeedges a membrane had, the more uniformly itbecame hydrogenated (22). In the extreme case ofall the edges being fixed to the scaffold, evendomains with a stretched lattice could be ob-served (22). Annealing led to complete recoveryof the original periodicity observed in TEM.

The in-plane compression of graphene’s lat-tice can only be the result of chemical modi-

fication as opposed to physical forces, becauseany compression that is not stabilized on anatomic scale should cause the membranes tobuckle. Furthermore, strains of the order of a fewpercent would result in massive variations of theRaman peaks, which was not the case. The mostobvious candidate for the modified crystal latticeis graphane (7, 8). In this until-now-theoreticalmaterial, hydrogen attaches to graphene’s sub-lattices A and B from the two opposite sides,and carbon atoms in A and B move out of theplane (“buckle”), as shown in Fig. 4D. The in-plane periodicity probed by TEM would thensubstantially shrink if the length a of the C-Cbond were to remain the same as in graphene(1.42 Å). However, the change in hybridizationfrom sp2 to sp3 generally results in longer C-Cbonds, which is the effect opposing to the latticeshrinkage by atomic-scale buckling. Recent cal-culations (8) predicted a in graphane to be ≈1.53 Å(near that of diamond) and the in-plane peri-odicity d to be ≈1% smaller than in graphene.Although the maximum in the observed distri-bution of d occurs at ≈2.42 Å (that is, near thetheoretical value for graphane) (Fig. 4B), theobservation of more compressed areas (such asin Fig. 4A) suggests that the equilibrium d (with-out strain imposed by the scaffold) should besmaller. The latter implies either shorter or stron-ger buckled C-C bonds, or both, are present. Al-ternatively, the experimentally produced graphanemay have a more complex hydrogen bondingthan the one suggested by theory.

Finally, let us return to the graphene hydro-genated on a substrate (Figs. 1 and 3). Single-sided hydrogenation of ideal graphene wouldcreate a material that is thermodynamically un-stable (7, 8), and therefore our experiments seemto be in conflict with the theory [for the case ofgraphene on a substrate, we can exclude the pos-sibility of double-sided hydrogenation becausethe diffusion of hydrogen along the graphene-SiO2 interface is negligible (31)]. However, re-alistic graphene samples are not microscopicallyflat but always rippled (20, 21), which shouldfacilitate their single-sided hydrogenation. Indeed,attached hydrogen is expected to change the hy-bridization of carbon from sp2 to (practically)sp3 with angles of ~110° acquired between all ofthe bonds (7). These constraints necessitate themovement of carbon atoms out of the plane inthe direction of the attached hydrogen, at thecost of an increase in elastic energy. However,for a convex surface, the lattice is already de-formed in the direction that favors sp3 bonding,which lowers the total energy. As shown in (22),single-sided hydrogenation becomes energetical-ly favorable for a typical size of ripples observedexperimentally (20). Because of the random na-ture of ripples, single-sided graphane is expectedto be a disordered material, similar to grapheneoxide, rather than a new graphene-based crystal.The formation of a disordered material also ex-plains the observation of variable-range hoppingin our transport experiments. The importance of

ripples for hydrogenation of graphene on a sub-strate is further evidenced in experiments involv-ing bilayer samples, which show a substantiallylower level of hydrogenation than monolayersunder the same conditions (22). We attribute thisobservation to the fact that bilayer graphene isless rippled (20).

The distinct crystal structure of hydrogenatedgraphene and pronounced changes in its electronicand phonon properties reveal two new graphenederivatives, one crystalline and the other disordered.The results show that conversion of graphene intoother giant molecules with a regular structure ispossible.

References and Notes1. A. K. Geim, K. S. Novoselov, Nat. Mater. 6, 183 (2007).2. S. Stankovich et al., J. Mater. Chem. 16, 155 (2006).3. S. Stankovich et al., Nature 442, 282 (2006).4. X. Wang, L. Zhi, K. Mullen, Nano Lett. 8, 323 (2008).5. S. Gilje, S. Han, M. Wang, K. L. Wang, R. B. Kaner, Nano

Lett. 7, 3394 (2007).6. C. Gomez-Navarro et al., Nano Lett. 7, 3499 (2007).7. J. O. Sofo, A. S. Chaudhari, G. D. Barber, Phys. Rev. B 75,

153401 (2007).8. D. W. Boukhvalov, M. I. Katsnelson, A. I. Lichtenstein,

Phys. Rev. B 77, 035427 (2008).9. A. C. Dillon et al., Nature 386, 377 (1997).

10. F. L. Darkrim, P. Malbrunot, G. P. Tartaglia, Int. J.Hydrogen Energy 27, 193 (2002).

11. A. Züttel et al., Int. J. Hydrogen Energy 27, 203 (2002).12. T. Yildirim, O. Gulseren, S. Ciraci, Phys. Rev. B 64,

075404 (2001).13. B. N. Khare, M. Meyyappan, A. M. Cassell, C. V. Nguyen,

J. Han, Nano Lett. 2, 73 (2002).14. P. Ruffieux et al., Phys. Rev. B 66, 245416 (2002).15. D. Neumann et al., Appl. Phys. A 55, 489 (1992).16. A. Nikitin et al., Phys. Rev. Lett. 95, 225507 (2005).17. L. Hornekær et al., Phys. Rev. Lett. 96, 156104 (2006).18. K. S. Novoselov et al., Science 306, 666 (2004).19. A. C. Ferrari et al., Phys. Rev. Lett. 97, 187401 (2006).20. J. C. Meyer et al., Nature 446, 60 (2007).21. T. J. Booth et al., Nano Lett. 8, 2442 (2008).22. Supporting online material is available on Science

Online.23. S. V. Morozov et al., Phys. Rev. Lett. 100, 016602

(2008).24. N. F. Mott, Philos. Mag. 19, 835 (1969).25. T. O. Wehling et al., Nano Lett. 8, 173 (2008).26. V. M. Pereira, F. Guinea, J. M. B. Lopes dos Santos,

N. M. R. Peres, A. H. Castro Neto, Phys. Rev. Lett. 96,036801 (2006).

27. A. C. Ferrari, Solid State Commun. 143, 47 (2007).28. F. Tuinstra, J. L. Koenig, J. Chem. Phys. 53, 1126 (1970).29. A. C. Ferrari, J. Robertson, Phys. Rev. B 61, 14095 (2000).30. A. Das et al., Nat. Nanotechnol. 3, 210 (2008).31. J. S. Bunch et al., Nano Lett. 8, 2458 (2008).32. This work was supported by Engineering and Physical

Sciences Research Council (UK), the Royal Society, theEuropean Research Council (programs “Ideas” and “Newand Emerging Science and Technology,” project“Structural Information of Biological Molecules at AtomicResolution”), Office of Naval Research, and Air ForceResearch Office of Scientific Research. D.C.E. acknowledgesfinancial support from the National Council for Scientificand Technological Development (Brazil). The authors aregrateful to Nacional de Grafite for providing high-qualitycrystals of natural graphite.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/323/5914/610/DC1SOM TextFigs. S1 to S7References



REPORTS

...

......


Dynamical Quorum Sensing andSynchronization in Large Populationsof Chemical OscillatorsAnnette F. Taylor,1 Mark R. Tinsley,2 Fang Wang,2 Zhaoyang Huang,2 Kenneth Showalter2*

Populations of certain unicellular organisms, such as suspensions of yeast in nutrient solutions,undergo transitions to coordinated activity with increasing cell density. The collective behavior isbelieved to arise through communication by chemical signaling via the extracellular solution.We studied large, heterogeneous populations of discrete chemical oscillators (~100,000) withwell-defined kinetics to characterize two different types of density-dependent transitions tosynchronized oscillatory behavior. For different chemical exchange rates between the oscillatorsand the surrounding solution, increasing oscillator density led to (i) the gradual synchronization ofoscillatory activity, or (ii) the sudden “switching on” of synchronized oscillatory activity. We analyzethe roles of oscillator density and exchange rate of signaling species in these transitions with amathematical model of the interacting chemical oscillators.

From the periodic firing of neurons to theflashing of fireflies, the synchronization ofrhythmic activity plays a vital role in the

functioning of biological systems (1–3). Themechanisms by which single cells or wholeorganisms coordinate their activity continue toinspire research over a range of disciplines (4–6).Synchronization may occur by global coupling,where each oscillator is connected to every otheroscillator through a common (mean) field. Withthis mechanism, mathematically formalized byKuramoto (7), oscillators are regulated by theaverage activity of the population (the meanfield) and a collective rhythm emerges above acritical coupling strength. A number of oscillato-ry systems are thought to synchronize by thismechanism (8, 9), such as stirred suspensions ofthe cellular slime mold Dictyostelium discoidium(10), but it has been experimentally characterizedonly recently in a system of coupled electro-chemical oscillators (11).

A distinctly different type of transition tosynchronized oscillatory behavior has been ob-served in suspensions of yeast cells (12). Thedensity-dependent transition, discovered byAldridge and Pye (13) more than 30 yearsago, recently has been studied in a stirred-flowreactor configuration (14). Relaxation experi-ments reveal that slightly below the critical celldensity, the system is made up of a collection ofquiescent cells rather than unsynchronized os-cillatory cells, whereas slightly above the criticaldensity, the cells oscillate in nearly completesynchrony (12). This type of transition is muchlike quorum-sensing transitions in bacteria pop-ulations, where each member of a populationundergoes a sudden change in behavior with asupercritical increase in the concentration of a

signaling molecule (autoinducer) in the extracel-lular solution (15). Many examples of quorum-sensing transitions have been found, such as theappearance of bioluminescence in populations ofVibrio fischeri (16–18) and biofilm formation inPseudomonas aeruginosa (19).

We studied a population of chemical oscil-lators to characterize the transition to synchroni-zation as a function of population density andtransport rate of signaling species to the sur-rounding solution. We used porous catalytic par-ticles (~100 mm in radius) that were suspendedin a fixed volume of catalyst-free Belousov-Zhabotinsky (BZ) reaction mixture (20, 21). Thecatalyst for the reaction, Fe(phen)3

2+ (ferroin), isimmobilized on the cation exchange particles(22, 23). Reagents in the solution react with theferroin to produce an activator, HBrO2, whichcatalyzes its own production, and an inhibitor,

Br−, which inhibits autocatalysis. A catalyst-loaded particle changes from red to blue as ferroinis oxidized and HBrO2 is produced. The oxidizedmetal catalyst reacts with solution reagents toregenerate the reduced form of the catalyst and Br−. The cycle repeats when the inhibitor level fallssufficiently. Each catalyst-loaded particle has itsown oscillatory period, on the order of 1 min(45 T 13 s), which depends on the catalyst load-ing and the particle size. The period distribution(fig. S1) is obtained by monitoring the colorchange associated with the oxidation of theferroin catalyst on each particle in unstirred solu-tions (21, 24).

A sketch of the experimental setup is shownin Fig. 1. Both the activator HBrO2 and inhibitorBr− are exchanged between the catalyst particlesand the surrounding solution, with the exchangerate depending on the stirring rate (21). The sur-rounding solution is monitored with a Pt elec-trode, where the potential increases with anincreasing concentration of HBrO2 relative toBr−, and the oxidation state of the particles ismonitored by high-speed video (21). The firsttime series in Fig. 1 illustrates the change inamplitude of the potential as the oscillator densityis increased. The second time series shows thecorresponding average intensity obtained fromthe mean of the individual particle intensities ineach image (scaled by the average intensity offully oxidized particles). When all particles aresimultaneously oxidized, the scaled average in-tensity increases from 0 (all particles red) to 1 (allparticles blue), and the value is less than 1when afraction of the population is oxidized. Thecoherence of the population can be seen inimages of the stirred particles (Fig. 1).

We observed two distinct types of transitionsto synchronous activity, depending on the stirringrate. One type occurs at low stirring rates, as

1School of Chemistry, University of Leeds, Leeds LS2 9JT, UK.2C. Eugene Bennett Department of Chemistry, West VirginiaUniversity, Morgantown, WV 26506, USA.


Fig. 1. Experimental setup (21). Catalytic microparticles are globally coupled by exchange of species withthe surrounding catalyst-free BZ reaction solution. Electrochemical time series illustrates the change inoscillatory amplitude and period with increasing particle density (red line) for a stirring rate of 600 rpm. Atypical series of images obtained during one oscillation is shown, from which the (normalized) averageintensity of the particles is calculated as a function of time. The associated time series illustrates thechange in oscillatory amplitude and period with increasing particle density (red line) for a stirring rate of600 rpm. A density of 0.02 g cm−3 corresponds to ~1.3 × 104 particles cm−3.


REPORTS

...

......


shown in Fig. 2A for 300 rpm. At low particledensities, the global electrochemical signal isnoisy with no regular oscillations. There are alsono oscillations in the average intensity of theparticle images; however, an approximatelyconstant fraction of oxidized catalyst particles isobserved (~20%), representing the fraction of theoscillatory cycle in the oxidized state. As thedensity is increased, small-amplitude oscillationsemerge, and there is a gradual growth in theamplitude of the global electrochemical signal.Corresponding oscillations are also observed inthe average image intensity. With a density of0.02 g cm−3, ~50% of the particles are simulta-neously oxidized in an oscillation, and this ap-proaches 100% for densities greater than 0.04 gcm−3, as shown by the maximum average inten-sity as a function of particle density. Also shownis the period of the oscillations decreasingslightly with increasing particle density.

The other type of transition to synchronousactivity occurs at high stirring rates. At lowparticle density, there is no oscillatory signal inthe electrochemical potential and there are nooxidized catalyst particles in the associated im-ages, as shown in Fig. 2B for 600 rpm. As thedensity is increased beyond a threshold value,large-amplitude oscillations in the global signalsuddenly appear, and analysis of the averageimage intensity shows that ~80% of the particlesare simultaneously oxidized each oscillation. Asthe density is increased further, the maximumfraction of oxidized particles during an oscil-lation approaches a constant value. The period ofoscillation, which is greater than in the low stir-ring rate case, decreases with increasing density.

Insights into these transitions to synchro-nized oscillatory behavior can be gained byexamining a model of the oscillatory particlesystem (20, 21). A sketch of the exchange be-tween the particles and the surrounding solutionis shown in Fig. 3. The model is based on thethree-variable ZBKE scheme (25) for theferroin-catalyzed BZ reaction (26), with varia-bles X for the autocatalyst, Y for the inhibitor,and Z for the oxidized form of the metal ioncatalyst. The concentration of the autocatalystfor the ith particle (i = 1, … N) is given by

dXi

dt¼ −kexðXi − XsÞ þ f ðXi, Yi, ZiÞ ð1Þ

and in the surrounding solution

dXs

dt¼ V

Vs∑iNkexðXi − XsÞ þ gðXs, YsÞ ð2Þ

where f (Xi, Yi, Zi) represents the chemicalreaction on the particle, and g(Xs, Ys) representsthe reaction in the surrounding solution (21).The exchange rate of the autocatalyst betweenthe particle and the surrounding solution isgiven by the term –kex(Xi – Xs), where the valueof the exchange rate constant kex increases withincreasing stirring rate (21). The sum of the con-tributions from each particle gives the concen-tration in the surrounding solution, with thedilution factor V=Vs, where Vs is the totalvolume of solution and V is the average vol-ume of a particle. The particle dynamics is pe-riodic for kex = 0 and depends on the numberdensity and the exchange rate for nonzero kex.

There is no mechanism for oscillation in thesurrounding solution in the absence of particles;however, in the presence of oscillatory particles,oscillations in Xs and Ys arise from the exchangebetween the particles and the solution.

The coherence of the population of oscil-lators is determined for a sample of N = 1000using the order parameter K introduced byShinomoto and Kuramoto (27):

K ¼*

N−1∑N

jexpðiqjÞ −

N−1∑

N

jexpðiqjÞ

+

ð3Þ

where qj is the phase of the jth oscillator andangle brackets indicate the time average. Thiscoherence measure is 0 when all particles os-cillate out of phase with each other (or when theparticles are not oscillating) and 1 when allparticles oscillate in perfect synchrony. For kex =0, the particles have a broad distribution ofnatural periods (38 T 6 s).

The surface plot in Fig. 3A shows the am-plitude of the activator species Xs in the sur-rounding solution as a function of the numberdensity of particles n, where n = N/Vs, and theexchange rate constant kex. The growth in theamplitude of Xs with increasing n depends onthe value of kex. For kex = 0.3, there is a dis-tribution in the natural period for low n (Fig. 3B).There is a corresponding gradual increase in thecoherence parameter K with increasing particledensity as the individual oscillators graduallyalign their frequencies and phases. For kex = 3.0,the catalyst particles are not oscillatory for lowvalues of n (Fig. 3C); however, when n is in-creased beyond a threshold value, the particlessuddenly begin to oscillate in perfect synchrony.The coherence parameter switches from 0 to 1.The gradual transition from unsynchronized tosynchronized oscillations at low values of kex,and the sharp switching transition from steady-state behavior to synchronized oscillations athigh values of kex can be seen in the surface plotof the global signal Xs (Fig. 3A).

The gradual synchronization of particles forlow kex can be understood within the frameworkof the Kuramoto model, with Xs playing the roleof the mean-field coupling. The time variationof Xi, Xs, and the exchange rate –kex(Xi – Xs) fora particle are shown in Fig. 4A for one cycle,with kex = 0.3 and n = 4000. The value of Xsincreases when several of the oscillators firetogether (Fig. 4A, top). This increase in turninfluences the individual oscillators through theexchange rate, which becomes positive for aparticle when the concentration is higher in thesolution (Fig. 4A, middle), leading to an in-crease in Xi. When Xi crosses a threshold value,the particle fires in synchrony with the others(Fig. 4A, bottom). For low n, particles with fre-quencies that differ from that of the global os-cillation are little affected by the weak signal inXs; however, as n is increased, the magnitude of

Fig. 2. Dynamical tran-sitions with increasingparticle density at lowstirring rate (A) and highstirring rate (B). From topto bottom: Amplitude ofoscillation in the sur-rounding solution fromthe electrochemical sig-nal; maximum of theaverage particle intensity,corresponding to the de-gree of coherence of theoscillators; and period ofoscillation.


REPORTS

...

......


the Xs signal increases, and more of the oscilla-tors join the rhythm. In this way, the particles aresynchronized by an internally generated signal.

The exchange rate for all particles averagedover one cycle is negative and therefore consti-tutes an overall loss rate in Xi, shown as a func-

tion of kex in Fig. 4B. As kex increases for fixedn, the average concentration of activator in thesurrounding solution Xs increases (blue line), butthe loss rate also increases (red line), and hencethe average concentration of Xi on a particledecreases (black line). When it decreases belowa threshold value, oscillations are no longersupported and a sudden transition to the steadystate is observed. The Xi loss rate increasesslightly as kex is further increased.

The model predicts that both a dynamicalquorum-sensing transition and a desynchroni-zation transition will be observed upon decreas-ing the stirring rate (Fig. 4C, upper panel). Theoscillations suddenly appear and then decreasein amplitude with decreasing kex because ofthe loss of synchronization until a noisy non-oscillatory signal corresponding to out-of-phaseoscillators is observed. The transition to the de-synchronized state is best observed at low par-ticle densities, as can be seen in Fig. 3A for n =1.8 × 104 cm−3. The transition from steady-stateto oscillatory behavior followed by a gradualdesynchronization is also observed experimen-tally (Fig. 4C, lower panel).

For high kex and low n, the catalyst particlesare quiescent, as oscillations are not supportedbecause of the high loss rate of Xi. As the num-ber density is increased, the concentration bothin the surrounding solution Xs and on the indi-vidual particles Xi increases, with a correspond-ing slight decrease in the average loss rate of Xion the particles (Fig. 4D). At a critical density n,the value of Xi on the individual particlesreaches the threshold for the transition fromsteady-state behavior to oscillatory behavior. Al-though there is a considerable spread in oscil-lator frequencies, the transition is sharp, with allparticles oscillating synchronously, indicatingcooperative behavior in the transition (21). Thetime series in Fig. 1 show experimental measure-ments of the dynamical quorum-sensing transi-tion from steady-state to oscillatory behavior withincreasing particle density, with a Pt electrodegiving the global signal and video images givingthe average particle intensity.

We have shown that there are two distincttypes of transitions to synchronized oscillatorybehavior with increasing oscillator density, wherecoupling occurs by exchange of signaling speciesthrough the surrounding solution. For lowexchange rates, the oscillators gradually synchro-nize their rhythms via a weak coupling throughthe low solution concentration of activator. Forhigh exchange rates, the oscillators are quiescentbelow a critical density, as the activator concen-tration on the particles is not sufficient to supportoscillatory behavior. Higher exchange rates giverise to lower activator concentrations on individ-ual particles, because only activator-consumingprocesses occur in the surrounding solution. Asthe density is increased, the activator concentra-tion in the surrounding solution and on eachparticle increases until a threshold concentrationis reached, and synchronous oscillations sudden-

Fig. 3. Behavior of BZparticle population model(20, 21) with oscillatorscoupled by exchange ofspecies with the sur-rounding solution (seetable S1 for parametervalues). (A) The ampli-tude of the oscillationsin autocatalyst in the sur-rounding solution as afunction of number den-sity of oscillators and ex-change rate constant. (B)The order parameter K asa function of n for low ex-change rates. Lower plotsshow the distribution ofoscillator periods at indi-cated densities (i to iii).(C) The order parameter Kat high exchange ratesand corresponding oscil-lator periods.

Fig. 4. Influence of exchange rate –kex(Xi – Xs) on model dynamics. (A) Variation of autocatalyst insolution (blue line), on a particle (black line), and the exchange rate (red line) during one oscillation, withn= 1.8 × 104 cm−3 and kex = 0.3 s−1. (B) Transition from oscillations to steady state with increasing kex (n=1.8 × 104 cm−3). Time-averaged autocatalyst on the particles (black line), in solution (blue line), and lossrate of autocatalyst from the particles (red line = ⟨kexðXi − XsÞ⟩). (C) Appearance and desynchronization ofoscillations in autocatalyst in solution (blue line) and electrode potential (black line) with decreasingkex (n = 1.8 × 104 cm−3) and stirring rate (density = 0.0162 g cm−3) shown as red lines. (D) Transitionfrom steady state to oscillations with increasing n (kex = 3.0 s−1). Color coding is same as in (A).


REPORTS

...

......


ly appear at the critical density. The sudden tran-sition has features that are suggestive of an“oscillator death” transition (28); however, thecoupling between particles occurs through thesurrounding solution in which there is an addi-tional decay of the activator species.

The exchange rate of the activator HBrO2 andthe inhibitor Br− determines whether the transitionto synchronized oscillatory behavior occurs bya synchronization mechanism or a dynamicalquorum-sensing mechanism. It is possible thatsuspensions of cellular organisms, such as yeastorD. discoidium, may also undergo both types oftransitions with different efficacies of molecularsignaling. Our system of chemical oscillators sus-pended in a stirred solution is an idealized configu-ration of oscillators interacting by global coupling.

References and Notes1. A. T. Winfree, The Geometry of Biological Time (Springer,

New York, ed. 2, 2001).2. A. Goldbeter, Biochemical Oscillations and Cellular

Rhythms: The Molecular Bases of Periodic andChaotic Behaviour (Cambridge Univ. Press, Cambridge,1996).

3. L. Glass, M. C. Mackey, From Clocks to Chaos: The Rhythmsof Life (Princeton Univ. Press, Princeton, NJ, 1988).

4. S. Strogatz, Sync: The Emerging Science of SpontaneousOrder (Hyperion, New York, 2003).

5. A. S. Pikovsky, M. Rosenblum, J. Kurths, Synchronization:A Universal Concept in Nonlinear Sciences (CambridgeUniv. Press, Cambridge, 2001).

6. S. C. Manrubia, A. S. Mikhailov, D. H. Zanette, Emergenceof Dynamical Order: Synchronization Phenomena inComplex Systems (World Scientific, Singapore, 2004).

7. Y. Kuramoto, Chemical Oscillations, Waves andTurbulence (Springer, Berlin, 1984).

8. A. T. Winfree, Science 298, 2336 (2002).9. J. Garcia-Ojalvo, M. B. Elowitz, S. H. Strogatz,

Proc. Natl. Acad. Sci. U.S.A. 101, 10955 (2004).10. G. Gerisch, B. Hess, Proc. Natl. Acad. Sci. U.S.A. 71,

2118 (1974).11. I. Z. Kiss, Y. Zhai, J. L. Hudson, Science 296, 1676 (2002).12. S. De Monte, F. d’Ovidio, S. Danø, P. G. Sørensen,

Proc. Natl. Acad. Sci. U.S.A. 104, 18377 (2007).13. J. Aldridge, E. K. Pye, Nature 259, 670 (1976).14. S. Danø, P. G. Sørensen, F. Hynne, Nature 402, 320

(1999).15. A. Camilli, B. L. Bassler, Science 311, 1113 (2006).16. E. P. Greenberg, ASM News 63, 371 (1997).17. A. Eberhard et al., Biochemistry 20, 2444 (1981).18. J. Engebrecht, K. Nealson, M. Silverman, Cell 32, 773

(1983).19. G. M. Patriquin et al., J. Bacteriol. 190, 662 (2008).

20. R. Toth, A. F. Taylor, M. R. Tinsley, J. Phys. Chem. B 110,10170 (2006).

21. See supporting material on Science Online.22. J. Maselko, K. Showalter, Nature 339, 609 (1989).23. J. Maselko, J. S. Reckley, K. Showalter, J. Phys. Chem. 93,

2774 (1989).24. K. Yoshikawa, R. Aihara, K. Agladze, J. Phys. Chem. A

102, 7649 (1998).25. A. M. Zhabotinsky, F. Buchholtz, A. B. Kiyatkin,

I. R. Epstein, J. Phys. Chem. 97, 7578 (1993).26. A. Zaikin, A. Zhabotinsky, Nature 225, 535 (1970).27. S. Shinomoto, Y. Kuramoto, Prog. Theor. Phys. 75, 1105

(1986).28. G. B. Ermentrout, Physica D 41, 219 (1990).29. Supported by UK Engineering and Physical Sciences

Research Council grant GR/T11036/01 (A.F.T.) and by NSFgrant CHE-0809058. We dedicate this paper to the memoryof Anatol M. Zhabotinsky, a pioneer of modern researchin oscillations and pattern formation in chemical systems.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/323/5914/614/DC1Materials and MethodsSOM TextFigs. S1 and S2Table S1References


Single Nanocrystals of PlatinumPrepared by Partial Dissolution ofAu-Pt NanoalloysMarc Schrinner,1 Matthias Ballauff,1* Yeshayahu Talmon,2 Yaron Kauffmann,3Jürgen Thun,4 Michael Möller,4 Josef Breu4

Small metal nanoparticles that are also highly crystalline have the potential for showing enhancedcatalytic activity. We describe the preparation of single nanocrystals of platinum that are 2 to 3nanometers in diameter. These particles were generated and immobilized on sphericalpolyelectrolyte brushes consisting of a polystyrene core (diameter of ~100 nanometers) onto whichlong chains of a cationic polyelectrolyte were affixed. In a first step, a nanoalloy of gold andplatinum (a solid solution) was generated within the layer of cationic polyelectrolyte chains. In asecond step, the gold was slowly and selectively dissolved by cyanide ions in the presence ofoxygen. Cryogenic transmission electron microscopy, wide-angle x-ray scattering, and high-resolution transmission electron microscopy showed that the resulting platinum nanoparticles arefaceted single crystals that remain embedded in the polyelectrolyte-chain layer. The compositesystems of the core particles and the platinum single nanocrystals exhibit an excellent colloidalstability, as well as high catalytic activity in hydrogenation reactions in the aqueous phase.

Metallic nanoparticles (NPs) of controlledsize and shape have been of great inter-est recently for a number of possible

applications in electronic or optical materials, aswell as in catalysis (1–8). Particle morphologycan play a central role in catalysis. For example,faceted Pt crystals can exhibit higher catalytic

activity than spherical particles (9, 10), and theactivity of the exposed facets may vary consider-ably (11, 12). The reactivity and selectivity ofNPs can be tuned by controlling their morphol-ogy; for example, amorphous Pt NPs haveexhibited a much reduced catalytic activity (13).

However, faceted nanocrystals (NCs) with awell-developed shape and a narrow size dis-tribution that have been reported generally werein the size range of 100 nm and more. Forinstances, Sun and Xia obtained well-defined Auand Ag crystals with sizes on the order of 100 nm(4). Nanoprisms of Ag with dimensions around100 nm were prepared by Jin et al. by photo-chemical conversion of Ag spheres (3). Aniso-

tropic Ag NPs of similar size were synthesizedby Liz-Marzan and co-workers through carefulchoice of a suitable surfactant (14), and Pt NPswith high-index facets were obtained recentlyby Tian and co-workers (6). Again, the typicalsizes ranged between 50 and 200 nm. The onlyreported route to faceted single crystals in the sizerange of a few nanometers was the synthesis ofwell-defined clusters, for example, Au55 cluster,and a subsequent heat treatment (15).

Platinum NPs have been of particular interestbecause of their role in many catalytic reactionsand because the growth of specific surfaces canbe controlled by amphiphilic polymers or suit-able surfactants (9, 16–19). However, downsiz-ing the Pt NPs to a few nanometers is oftenaccompanied by a broadening of their size dis-tribution and partial loss of the control of theparticle shape (5). The reduction of themetal ionsand the generation of the NPs seem to proceedvery rapidly and leads to disordered structuresand distorted crystal shapes.

We present a simple synthesis of faceted,well-defined Pt single NCs with a typical size of2 to 3 nm. As shown in Fig. 1, the Pt NCs areobtained by partial dissolution of nanoalloys ofPt and Au. In case of larger nanoalloy particles,this procedure leads to spongelike or hollowstructures (20, 21). We found that the dissolutionof the Au component of the nanoalloy leads toreorganization of the Pt atoms and the formationof well-defined, faceted NCs. We started from ananoalloy of Au and Pt generated on the surfaceof a spherical polyelectrolyte brush (SPB), asdescribed recently (22). These particles consist ofa solid polystyrene (PS) core with colloidal di-mensions (diameter ~ 100 nm) onto which longcharged polymer chains are grafted. Sphericalpolyelectrolyte brushes are well suited for the

1Physikalische Chemie I, University of Bayreuth, 95440Bayreuth, Germany. 2Department of Chemical Engineering,Technion-Israel Institute of Technology, Haifa 32000, Israel.3Department of Materials Engineering, Technion-Israel Insti-tute of Technology, Haifa 3200, Israel. 4Anorganische ChemieI, University of Bayreuth, 95440 Bayreuth, Germany.



REPORTS

...

......


generation and immobilization of metallic NPs inthe aqueous phase (23, 24) because anionic com-plex ions of metals, like Au, Ag, Pd, or Pt, can beimmobilized in the surface layer of cationic SPB.The confinement of these ions was attributed tothe strong electrostatic interaction of the counter-ion with the highly charged macro-ion. The com-plex metal ions like PtCl6

2– are thus localizedsolely in the layer of polyelectrolyte chains of theSPB, so that virtually no metal salt is releasedinto the aqueous phase. Subsequent reductionwith NaBH4 under mild conditions led to theformation of metallic NPs, firmly embedded in adense layer of polyelectrolyte chains.

In the same way, particles of nanoalloys (25)can be synthesized and immobilized on thesurface layer of SPB particles. We demonstratedthis approach by reducing a mixture of theAuCl4

– and PtCl62– ions in the surface layer of

a cationic SPB (22). High-resolution transmis-sion electron microscopy (HR-TEM) has shownthat the nanoalloy particles do not exhibit a core-shell structure [compare with the discussion ofthis point in (25)] but form homogeneous solidsolutions (22). The composition of the metals canbe varied continuously without disturbing thestructure of the particles that consist of a randommixture of Au and Pt (22). Moreover, the finalcomposition is within narrow limits, determinedby the ratio of the metal ions introduced into thebrush layer earlier (22). The composite particlesthat consist of the carriers and the metallic NPsexhibit excellent colloidal stability in aqueoussolution, which arises from the strong interaction

of the alloy NPs with the polyelectrolyte chainsaffixed to their surface. As depicted in Fig. 1, thecationic polymer chains are interwoven with theNPs and form a dense mesh on the surface ofthe core particles. This model is supported by care-ful measurements of the hydrodynamic radius, RH,by dynamic light scattering [see the discussion ofthis point in (26)]. The thickness, L, of the surfacelayer is given by L= RH – R, where R denotes theradius of the cores. As will be shown below, thecolloidal stability of the composites allows us totransform the NPs while keeping them on thesurface of the core particles.

We prepared composite particles containingAu-Pt NPs with the composition Au45Pt55, asshown by elemental analysis. Partial dissolutionof these alloy NPs then leads to pure Pt NCs. TheAu atoms were slowly dissolved by a treatment

of the alloy NPs with cyanide ions in presence ofoxygen (Fig. 1). In a typical experiment, 4 mlof a 1.9 × 10−5 M NaCN solution was addeddropwise within 25 min to 13 ml of a AuPt-SPBsuspension [0.04 weight % (wt. %)] at roomtemperature under air with vigorous stirring. Thehigh dilution of the cyanide ions is crucial foravoiding coagulation or complete dissolution ofthe NPs. Air was bubbled slowly through thesolution to achieve complete removal of the Auatoms. After 3 hours, the dispersion turned blue,indicating the formation of pure Pt NCs (26).

The de-alloying of the Au-Pt-nanoalloy pro-ceeded surprisingly smoothly. Micrographs ofthe composite particles before and after the leach-ing process (Fig. 2) were obtained by cryogenictransmission electron microscopy (cryo-TEM)that allow us to analyze the NPs in their native

100 nm100 nm

Au

CN-/O2

A B

Fig. 2. (A) Cryo-TEM micrographs of the Au-Pt-nanoalloy particles (composition: Au45Pt55)generated on the surface of the spherical polyelectrolyte brushes. (B) Composite particles aftercomplete removal of the gold atoms from the Au-Pt-nanoalloy by a mixture of CN– ions and O2.

Fig. 1. Synthesis schemeof Pt NCs by de-alloying of aAu-Pt nanoalloy. The carrierparticles are SPBs that con-sist of a solid PS core (RH =50 nm) onto which cationicpolyelectrolyte chains of 2-aminoethylmethacrylate (2-AEMH) are attached. In afirst step, the chloride coun-terions were exchangedagainst AuCl4

– ions; in asecond step the rest of Cl–

ions were exchanged againstPtCl4

2– ions. BimetallicAu45Pt55 nanoalloy particleswere generated by reductionof the mixture of these ionsby NaBH4 (22). The compo-sition of the resulting nano-alloy can be adjusted by theratio of the metal ions in thebrush layer. In the final step,cyanide ions and oxygenwere used to leach out theAu atoms from the nano-alloy under very mild con-ditions. This procedure leadsto faceted Pt NCs with a fewnanometers in diameterembedded in the surface layer of polyelectrolyte chains.


REPORTS

...

......


state (27). Thus, a small drop of the suspension ofthe composite particles is shock-frozen and ana-lyzed by TEM (see supporting online material fordetails). The Pt NCs can be seen in Fig. 2B as blackdots distributed over the SPB.Both the alloyNPs aswell as the Pt NCs are distributed more or less uni-formly over the carrier particles. The alloyNPs havea rather narrow size distribution that is preservedduring the leaching process (fig. S3 histogram).

The formation of the cyanide complex isselective for Au (28). Energy-dispersive x-rayspectroscopy (EDX) (figs. S1 and S2) shows thatthe NPs resulting from de-alloying consist onlyof Pt. The Pt NPs are still embedded on thesurface of the SPB-carrier particles. The colloidalstability of the composite particles was not lostduring the reaction with the cyanide ions, nor isany coagulation of the NPs on the surface of thecarrier particles observed during this process. Wereiterate that rapid addition or high concentra-tions of the cyanide ions lead to coagulation andcoarsening of the metal NPs on the surface.

The structure of the Pt NPs was analyzed bycombining high-angle annular dark-field scan-ning TEM (HAADF-STEM) and HR-TEM withelectron diffraction (ED) andwide-angle x-ray scat-tering (WAXS). A low-magnification HAADF-STEM micrograph of the PS spheres on thesupporting holey carbon film is shown in Fig.3A, and Fig. 3B (from the same area at highermagnification) shows the uniform distribution ofthe Pt NPs on the PS spheres. In order to avoidany disturbance of this analysis by the core par-ticles, only NPs sitting on the periphery of thecarrier spheres were analyzed by HR-TEM (Fig.3, C to F). The HR-TEM shows that the Pt NPscontain no grain boundaries and are single crys-tals. In several cases, the facets can be indexedbecause the NCs are aligned by chance; in Fig. 3,E and F, the electron diffraction shows directlythe hexagonal symmetry of the cubic crystal.

The lattice spacing derived from the (200)reflection as seen in ED (d200 = 0.20 nm) is ingood agreement with the bulk value [d200 = 0.196nm; spacing 0.392 nm (29, 30)]. The overall sizeof the Pt NCs lies within the range estimatedfrom the volume contraction caused by Au leach-ing. Our results are consistent with the Au-Ptnanoparticles converted individually into Pt NCs,without coagulation or exchange between differ-ent alloy NPs having taken place. Some of theNPs moved under the intense electron beam androlled on the surface of the PS spheres, but theirshape remained stable (nomelting was observed).

The high crystallinity observed by HR-TEMmight have been induced by the electron beamthrough local heating and subsequent crystalliza-tion. To rule out this possibility, we examinedpowders of the dried composite particles beforeand after leaching by WAXS (fig. S4). The aver-age crystallite sizes as derived from the Rietveldanalysis were 7.01 T 0.003 [7.01(3)] nm for theAu45Pt55-NPs and 3.80(4) nm for the resulting PtNCs. These values agree quite well with crystalsizes as observed by HR-TEM showing that the

Fig. 3. (A and B) HAADF-STEM micrographs of the Pt NPs (bright spots) embedded and uniformlydispersed on a surface layer of the spherical polyelectrolyte. (C) HR-TEMmicrograph of NPs on the surfaceof two adjacent carrier particles. (D) HR-TEMmicrograph of several NCs. (E and F) HR-TEMmicrographs oftwo different Pt single NCs of sizes 4.6 and 2.8 nm, respectively, showing well-defined facets. Allmicrographs were acquired at 300 keV. (Insets) The Fourier transforms of the images.

Fig. 4. Catalytic activity of the PtNCs. k1 (Eq. 1), obtained for thecatalytic reduction of p-nitrophenolto p-aminophenol is plotted againstthe specific surface, S, of the PtNCs in the solution. S is the totalsurface of all particles per unitvolume.

0.008

0.004

0.000

0.00 0.01 0.02

S (m2/ l)

k ap

p (

s-1 )


REPORTS

...

......


particles are single-domain NCs. Also, the latticeparameter of 0.401 (T 0.001) nm obtained for thealloy NPs agrees very well with the value ex-pected from Vegard’s law, that is, a linear inter-polation of the lattice parameters of Au and Pt[see the discussion of this point in (22)], if a solidsolution had formed. After the Au leaching step,the lattice parameter decreased to 0.391 (T 0.001)nm, in agreement with the known lattice param-eter of pure Pt [0.392 nm (29)]. Thus, full dealloy-ing of the Au-Pt-nanoalloy could be achieved.Wenote that a partial dealloying by an electrochem-ical method has been recently reported by Kohand Strasser (31).

As a final point, we discuss the catalytic ac-tivity of the Pt NCs affixed to the surface of theSPBs. Previous work (23, 32, 33) showed thatthe catalytic reduction of p-nitrophenol to p-aminophenol can be used for the analysis of thecatalytic activity of Pt NPs. We assumed thatreduction rates were independent of the concen-tration of sodium borohydride because it was inexcess compared to p-nitrophenol. Moreover, theapparent rate constant, kapp, was found to beproportional to the surface S of the Pt NPs presentin the system (32, 33):

−dctdt

¼ kappct ¼ k1Sct ð1Þ

where ct is the concentration of p-nitrophenol attime t and k1 is the rate constant normalized to S,the surface area normalized to the unit volume ofthe system. k1 is plotted against the specific sur-face S of the Pt NCs in the systems in Fig. 4 (32).The Pt NCs exhibit a high catalytic activity and

turnover numbers as high as 1580 T 50, whichare among the highest turnover numbers mea-sured so far for this reaction (23, 32, 33). Hence,the composite particles consisting of the SPBsand the Pt NCs present a system with high col-loidal stability that may be used for catalysis in anaqueous environment.

References and Notes1. C. Burda, X. Chen, R. Narayanan, M. A. El-Sayed, Chem.

Rev. 105, 1025 (2005).2. S. K. Ghosh, T. Pal, Chem. Rev. 107, 4797 (2007).3. R. Jin et al., Science 294, 1901 (2001).4. Y. Sun, Y. Xia, Science 298, 2176 (2002).5. R. Narayanan, M. A. El-Sayed, J. Am. Chem. Soc. 126,

7194 (2004).6. N. Tian, Z.-Y. Zhou, S.-G. Sun, Y. Ding, Z. L. Wang,

Science 316, 732 (2007).7. S. C. Warren et al., Science 320, 1748 (2008).8. E. Sulman et al., Top. Catal. 39, 187 (2006).9. T. S. Ahmadi, Z. L. Wang, T. C. Green, A. Henglein,

M. A. El-Sayed, Science 272, 1924 (1996).10. R. Narayanan, M. A. El-Sayed, J. Am. Chem. Soc. 126,

7194 (2004).11. R. Narayanan, M. A. El-Sayed, J. Phys. Chem. B 107,

12416 (2003).12. R. Narayanan, M. A. El-Sayed, Nano Lett. 4, 1343 (2004).13. Y. Sun, L. Zhuang, J. Lu, X. Hong, P. Liu, J. Am. Chem.

Soc. 129, 15465 (2007).14. I. Pastoriza-Santos, L. Liz-Marzan, Nano Lett. 2, 903 (2002).15. M. Turner et al., Nature 454, 981 (2008).16. H. Lee et al., Angew. Chem. Int. Ed. 45, 7824 (2006).17. S. Kinge, H. Bönnemann, Appl. Organomet. Chem. 20,

784 (2006).18. J. Ren, R. D. Tilley, Small 3, 1508 (2007).19. E. Ramirez, L. Eradés, K. Philippot, P. Lecante,

B. Chaudret, Adv. Funct. Mater. 17, 2219 (2007).20. S. Guo, Y. Fang, S. Dong, E. Wang, J. Phys. Chem. C 111,

17104 (2007).21. J. Snyder, P. Asanithi, A. B. Dalton, J. Erlebacher,

Adv. Mater. 20, 4883 (2008).

22. M. Schrinner et al., Adv. Mater. 20, 1928 (2008).23. M. Ballauff, Prog. Polym. Sci. 32, 1135 (2007).24. Y. Lu et al., J. Phys. Chem. C 111, 7676 (2007).25. R. Ferrando, J. Jellinek, R. L. Johnston, Chem. Rev. 108,

845 (2008).26. M. Schrinner et al., Macromol. Chem. Phys. 208, 1542

(2007).27. Z. Li, E. Kesselman, Y. Talmon, M. A. Hillmyer,

T. P. Lodge, Science 306, 98 (2004).28. J. E. Huheey, Inorganic Chemistry: Principles of

Structure and Reactivity (Pearson/Addison-Wesley,Boston, ed. 5, 2004).

29. D. Brooksbank, K. W. Andrews, J. Iron Steel Inst. London206, 595 (1968).

30. T. Swanson, Natl. Bur. Stand. (U.S.) Circ. 31, 539 (1953).31. S. Koh, P. Strasser, J. Am. Chem. Soc. 129, 12624 (2007).32. Y. Mei, Y. Lu, F. Polzer, M. Ballauff, F. Polzer, Chem.

Mater. 19, 1062 (2007).33. Y. Mei et al., Langmuir 21, 12229 (2005).34. Financial support by the Deutsche Forschungsgemeinschaft,

SFB 481, Bayreuth; by the BASF SE; and by the Fonds derChemischen Industrie is gratefully acknowledged.M.S. thanks the Minerva Foundation, the German AcademicExchange Service (DAAD), and the Technion Russell BerrieNanotechnology Institute (RBNI) for the support duringa stay at the Technion in Haifa. Y.T. thanks the vonHumboldt Foundation for the support, through a vonHumboldt–Meitner Prize, of the collaboration with theUniversity of Bayreuth. Cryo-TEM was performed at theHannah and George Krumholz Laboratory for ElectronMicroscopy of Soft Matter, part of the Technion Project onComplex Liquids, Microstructure, and Macromolecules. TheHR-TEM work was done at the Electron Microscopy Centerat the Department of Materials Engineering, the Technion.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/323/5914/617/DC1Materials and MethodsFigs. S1 to S4Table S1References


Cascadia Tremor Located NearPlate Interface Constrainedby S Minus P Wave TimesMario La Rocca,1* Kenneth C. Creager,2 Danilo Galluzzo,1 Steve Malone,2John E. Vidale,2 Justin R. Sweet,2 Aaron G. Wech2

Nonvolcanic tremor is difficult to locate because it does not produce impulsive phases identifiableacross a seismic network. An alternative approach to identifying specific phases is to measurethe lag between the S and P waves. We cross-correlate vertical and horizontal seismograms toreveal signals common to both, but with the horizontal delayed with respect to the vertical. Thislagged correlation represents the time interval between vertical compressional waves andhorizontal shear waves. Measurements of this interval, combined with location techniques, resolvethe depth of tremor sources within T2 kilometers. For recent Cascadia tremor, the sourceslocate near or on the subducting slab interface. Strong correlations and steady S-P time differencesimply that tremor consists of radiation from repeating sources.

Deep nonvolcanic tremor (NVT) has beenobserved in several tectonically activeregions. In Cascadia and Japan, a rela-

tion with the subduction dynamics is inferredby the contemporaneous occurrence of slow slipdetected geodetically (1–3). In these regions,the occurrence of major episodic tremor and

slip (ETS) events is surprisingly periodic (2).Minor NVT has been triggered by the stressesimposed by low-frequency seismic waves ra-diated by major distant earthquakes (4–6), andNVT is modulated by ocean tidal loading (7),indicating that small stress perturbations triggerit. NVT is characterized by low amplitudes, a

lack of energy at high frequency, emergent onsets,an absence of clear impulsive phases, and dura-tions from minutes to days. The lack of impul-sive phases identifiable across seismic networksin Cascadia hinders accurate source location byconventional techniques.

NVT recorded in Shikoku, Japan, containsidentifiable low-frequency earthquakes (LFEs)(8, 9), suggesting that tremor is primarily madeup of many back-to-back LFEs. On individualseismograms, S waves are sometimes identi-fied. Stacking seismograms of repeating eventsrecorded at a given station after aligning themon the S waves produces a clear P wave. The Sminus P times allow an accurate depth to bedetermined. Such Japanese tremor locations arefound to be near the subduction plate interface(8, 10). Individual LFEs have not yet been re-ported in tremor recorded in Cascadia.

Tremor episodes that continue for severaldays are accompanied by geodetically observedslow slip in both Cascadia and Shikoku, leading

1Istituto Nazionale di Geofisica e Vulcanologia–OsservatorioVesuviano, Via Diocleziano 328, 80124 Napoli, Italy. 2Depart-ment of Earth and Space Science, University of Washington,Box 351310, Seattle, WA 98195, USA.



REPORTS

...

......


to the widely accepted interpretation for Japanthat tremor and slow slip are manifestations ofthe same process. In contrast, attempts to locatetremor in Cascadia have resulted in a broad, 40-km-thick, distribution in depth, leading to theinterpretation that much of the tremor is relatedto fluids throughout a large volume above theplate interface (11, 12). Three different locationtechniques for tremor in Cascadia produce depthestimates over a range of 40 km. One techniqueuses the arrival times of seismic signal envelopes(12–14) and conventional location techniquesbased on an S-wave velocity model. A secondtechnique, the Source Scanning Algorithm (15),stacks seismograms such that theoretical S-wavetime delays correspond to a grid of possible sourcelocations. Formal uncertainties are as small as 5 kmbut still show a wide depth range (11, 16). A thirdtechnique uses small-aperture arrays to computeslowness vectors for seismic wavefronts andback projects those for a common source (17).In all of these cases, common signal shapes areprocessed under the assumption that they aredirect S phases.

Cascadia tremor depth estimates likely havea greater uncertainty than the reported formalerrors of 10 km in epicenter and more in depth.We tested this assumption by using two of thetechniques to locate earthquakes in the same re-gion as the tremor. The differences between themore accurately determined catalog earthquakedepths and those we computed are much greaterthan the formal errors, sometimes tens of kilome-ters different (14, 17). In the case of back pro-jection of slowness vectors, unmodeled lateralvelocity variations may bend the wave frontssufficiently to explain these large errors in esti-mated tremor depths. Errors in depth determinedfrom arrival times of seismogram envelopes arecaused largely by unmodeled variations in S-wavecoda duration. Here, we apply a new location pro-cedure that greatly improves the depth resolu-tion of the tremor sources.

Our data come from low-noise stations ofthe Pacific Northwest Seismograph Networkand from three dense arrays composed of sixthree-component short-period seismic stationsthat were deployed specifically to record theJuly 2004 ETS event in Cascadia (18) (Fig. 1).Tremor periods with depths from 20 to 60 kmhave been reported (12, 16, 17). Polarizationanalysis of array data indicates that most of thesignal has S-wave polarization (13) with appar-ent velocity higher than 5 km/s, as expected forshear waves from a deep source (17). P-wavepolarization is observed in some of the tremorsignals (18).

To improve the signal-to-noise ratio, we firststacked the seismograms over all the stations foreach component of each array. Such stacks em-phasize signals arriving with a steep incidenceangle, i.e., from sources directly below an array.Then we computed the cross-correlations betweenthe vertical and each horizontal stacked traceby using 300-s windows bandpass-filtered from

Fig. 2. Envelopes of the cross-correlations between vertical and east components for 5-min windowsrecorded at Sequim array (top) and Sooke array (bottom).

Fig. 1. Map showing locations of the three seismic arrays (large triangles), other stations used for location(small triangles), depth (km) to the plate interface [assumed to be 7 km above the slab Moho (20)], andlocation of cross section on Fig. 4 (black line) and tremor epicenters (dots), color-coded by distance above(red) and below (blue) the plate interface. Tremor epicenters are determined from cross-correlations ofseismogram envelopes, and depths come from additional ts-p constraints. Theoretical ray incident angles atthe arrays are restricted to be less than 15°.


REPORTS

...

......


2 to 8 Hz. On many traces, the cross-correlogramenvelopes show a distinct and persistent peakat positive lag times of 3.5 to 6.0 s, and no suchpeaks at negative lag times (Fig. 2). Positive lagscorrespond to the vertical component leadingthe horizontal component. These are particularlyevident at the Sequim and Sooke arrays, wherethey were observed in many 5-min windowsevery day from 9 to 18 July. At the Lopez array,

similar lags were visible only sporadically on 9and 10 July, when tremor was nearest the array.

The distinct peaks at positive lags are con-sistent with the arrival of first vertical P wavesand then horizontal S waves. The sequentialcorrelations are characterized by tens of minuteswith almost constant S minus P times (ts-p). Inseveral cases, when these peaks are most evi-dent, visual inspection of the stacked seismo-

grams revealed many subtle individual P- andS-wave arrival pairs with the same ts-p (Fig. 3).This pattern implies that at least some Cascadiatremor is made up of individual LFEs from asingle source region occurring rapidly one afteranother in a sporadic sequence such as observedin Japan (9).

Figure 2 shows lag times of 5.2 s at Sequimand of 3.7 s at Sooke. At Sequim, there were nointervals during the entire tremor period with ts-pless than 4.5 s. The low horizontal slowness mea-sured at this array during times correspondingto ts-p near 4.5 s indicate that the sources weredirectly beneath the array. In contrast, higherslowness and back-azimuths to the north sys-tematically correspond to ts-p = 5.0 to 5.4 s. Atthe Sooke array, all S-P times were between3.7 and 4.4 s. The smallest value of ts-p = 3.7 s(Fig. 2) is consistent with sources located al-most directly beneath the array. At the Lopezarray, there were only a few periods of tremorwith clear cross-correlation peaks, and those hadts-p = 5.5 s.

Combining the ts-p-determined distanceswith slowness determined by array analysis al-lows an accurate estimate of the source depth.Based on a standard layered Earth model (19),tremor below the Sequim array (ts-p = 4.5 s) waslocated at a depth of 39 T 2 km, whereas it was31 T 2 km below the Sooke array (ts-p = 3.7 s).To better compare the range of depths deter-mined with and without the ts-p constraints, weapplied an envelope cross-correlation tremorlocation method (14) to the same periods oftime for which good S-P cross-correlation timesare available. We located these tremor burstsboth with and without the constraints of dis-tance based on array measurements of ts-p andshow the resulting epicenters in Fig. 1 and cross-section in Fig. 4. We kept all observations forwhich the theoretical ray angles at the arrayswere less than 15° from the vertical. Depths areshown relative to the plate interface depth asdetermined from slab-Moho reflections (20).The depth range for locations determined withjust envelope cross-correlations is large (smallcircles) as compared with those of the sametremor located with the additional ts-p constraints(large circles). At Lopez and Sequim, the refineddepths are close to the plate interface. On aver-age, the tremor beneath Sooke appears to befarther above the slab interface than that beneathSequim and Lopez. This is likely an artifact ofunmodeled high wave speeds associated withthe Crescent Basalts beneath Sooke (21). Takentogether, out of 128 observations, the mean sourcedistance above the plate interface is 3 km andthe standard deviation is T5 km.

The clear correlations and stable ts-p in thesignals suggest that, as was the case in Shikoku,where individual low-frequency events are foundin the tremor and precisely located (8), at leastsome of the deep tremor in Cascadia consists ofhighly repetitive, continuous sequences of indi-vidual events, which radiate both P- and S-wave

Fig. 4. Cross-section of tremors shown in Fig. 1 along the southwest-northeast profile. Small dotsshow hypocenters determined solely from envelope cross correlations. Large dots are hypocentersof the identical set of tremor periods, but with their depths shifted to satisfy the additional ts-pconstraints from the Sooke (black), Sequim (blue), and Lopez (red) arrays.

Fig. 3. Three-component seismograms in which several P waves, marked by arrows on the verticalcomponent, are followed by S waves 5.1 T 0.2 s later, as marked by arrows on the horizontalcomponents. Seismograms are the stacked signals of Sequim array bandpass filtered between 2and 10 Hz. The reference time is 15:00 GMT of 9 July 2004.


REPORTS

...

......


energy from a volume small enough to givealmost constant ts-p over periods of tens ofminutes or more. We cannot locate all tremorwith this method for two reasons. First, thestable determination of ts-p depends on tremorover an extended period coming from only oneplace, which is not always the case. Second, ourtechnique is most sensitive to tremor beneatheach array where the P and S waves are clearlyseparated into vertical and horizontal compo-nents. There are many periods of time, particu-larly at the Lopez array, in which tremor waslocated some distance from an array. In thesecases, there was likely too much S-wave energyon the vertical component and P energy on thehorizontal to get a strong cross-correlation.Although we cannot determine the depth of allobserved tremor, the tremor under each arrayoccurred near the subduction zone interface, andwe see no evidence for shallow tremor, in con-trast to other studies (11, 12, 16, 17). Our meth-od is ideally suited to find shallow tremor if itexisted beneath an array.

If our results are representative of most NVTin Cascadia, then the contemporaneous occur-

rence of Global Positioning System–detected slowslip (22) indicates that deep tremor and slow slipare two manifestations of the same source pro-cess in Cascadia, as has been inferred for Japan(8, 10, 23). This hypothesis is further supportedby the polarization characteristics of tremor sig-nals, which are in good agreement with the ex-pected polarization of signals produced by faultslip along the subduction megathrust (13).

References and Notes1. K. Obara, Science 296, 1679 (2002).2. G. Rogers, H. Dragert, Science 300, 1942 (2003).3. K. Obara, H. Hirose, Tectonophysics 417, 33

(2006).4. J. L. Rubinstein et al., Nature 448, 579 (2007).5. J. Gomberg et al., Science 319, 173 (2008).6. M. Miyazawa, J. Mori, Geophys. Res. Lett. 32, L10307

(2005).7. J. L. Rubinstein, M. La Rocca, J. E. Vidale, K. C. Creager,

A. G. Wech, Science 319, 186 (2008).8. D. R. Shelly, G. C. Beroza, S. Ide, S. Nakamula, Nature

442, 188 (2006).9. D. R. Shelly, G. C. Beroza, S. Ide, Nature 446, 305

(2007).10. Shelly D. R., G. C. Beroza, S. Ide, Geochem. Geophys.

Geosys. 8 (2007).11. H. Kao et al., Nature 436, 841 (2005).

12. W. McCausland, S. Malone, D. Johnson, Geophys. Res. Lett.32, L24311 (2005).

13. Wech A. G., K. C. Creager, Geophys. Res. Lett. 34,L22306 (2007).

14. A. G. Wech, K. C. Creager, Geophys. Res. Lett. 35,L20302 (2008).

15. H. Kao, S.-J. Shan, Geophys. J. Int. 157, 589 (2004).16. H. Kao et al., J. Geophys. Res. 111, B03309 (2006).17. M. La Rocca et al., Bull. Seismol. Soc. Am. 98, 620

(2008).18. M. La Rocca et al., Geophys. Res. Lett. 32, L21319

(2005).19. R. S. Crosson, J. Geophys. Res. 81, 3047 (1976).20. L. A. Preston, K. C. Creager, R. S. Crosson, T. M. Brocher,

A. M. Trehu, Science 302, 1197 (2003).21. K. Ramachandran, R. D. Hyndman, T. M. Brocher,

J. Geophys. Res. 111, B12301 (2006).22. T. I. Melbourne, W. M. Szeliga, M. M. Miller,

V. M. Santillan, Geophys. Res. Lett. 32, L04301 (2005).23. Ide S., D. R. Shelly, G. C. Beroza, Geophys. Res. Lett. 34,

L03308 (2007).24. Istituto Nazionale di Geofisica e Vulcanologia and

IRIS/PASSCAL provided instruments for the arrays; otherseismograms were provided by the Pacific NorthwestSeismic Network. Discussions with T. Pratt, J. Gomberg,H. Houston, A. Ghosh, E. Del Pezzo and comments from twoanonymous reviewers improved this manuscript. Fundingwas provided by INGV, NSF and the U.S. Geological Survey.


Divergent Evolution of DuplicateGenes Leads to GeneticIncompatibilities Within A. thalianaDavid Bikard,1 Dhaval Patel,2 Claire Le Metté,1 Veronica Giorgi,1 Christine Camilleri,1Malcolm J. Bennett,2 Olivier Loudet1*

Genetic incompatibilities resulting from interactions between two loci represent a potentialsource of postzygotic barriers and may be an important factor in evolution when they impair theoutcome of interspecific crosses. We show that, in crosses between strains of the plant Arabidopsisthaliana, loci interact epistatically, controlling a recessive embryo lethality. This interaction isexplained by divergent evolution occurring among paralogs of an essential duplicate gene, forwhich the functional copy is not located at the same locus in different accessions. These paralogsdemonstrate genetic heterogeneity in their respective evolutionary trajectories, which results inwidespread incompatibility among strains. Our data suggest that these passive mechanisms,gene duplication and extinction, could represent an important source of genetic incompatibilitiesacross all taxa.

When crossing individuals from differ-ent species is feasible, offspring oftenhave reduced viability or fertility (1, 2).

The Bateson-Dobzhansky-Muller model explainssuch incompatibilities on the basis of the syner-gistic interaction of genes that have functionallydiverged among the respective parents (3–5).Elucidating the molecular basis of such geneticincompatibilities is of great importance to the

science of evolution as well as to plant breeding.Whether these incompatibilities mostly appearconcurrently with speciation (arising, for exam-ple, in geographically isolated populations) or afterspeciation has occurred as a consequence of theirdivergence remains a considerable question (4); itis now known that such incompatibilities can seg-regate within species first (5, 6). A limited numberof genes interacting to cause hybrid incompati-bility have been identified at the molecular level,such as the Lhr/Hmr system responsible forlethality of male F1s from a cross between twoDrosophila species (7). Recently, an interactionbetween zeel-1 and peel-1 loci was discoveredto causewidespread genetic incompatibility amongCaenorhabditis elegans strains (8). Also at the

intraspecific scale, a typical dominant case of in-compatibility in Arabidopsis thaliana has beenidentified that may establish a link between hy-brid necrosis and the plant immune system (9).

While generating homozygous progeny fromcrosses between A. thaliana wild strains, it is fre-quently witnessed that physically unlinked locido not always segregate independently and that,often, one homozygous allelic combination at twoindependent loci is rare or totally absent in thedescendants of a specific cross (10, 11). This phe-nomenon explains part of the segregation distor-tion inherited in suchmaterial and is viewed as therecessive version of Bateson-Dobzhansky-Muller–type incompatibilities, although other models thanfunctional divergence could apply (12, 13). Suchepistasis-based recessive incompatibilities couldresult in reduced fitness in the progeny and limitthe extent of rearrangements among parental ge-nomes. Furthermore, if several incompatibilitieswere to segregate within a cross they should leadto conflicts between the genomes of divergedstrains, which could result in isolating barriersand, ultimately, speciation (14, 15).

A cross between the Arabidopsis referenceaccession Columbia-0 (Col) and the Cape VerdeIsland accession Cvi-0 (Cvi) was generated, and367 F6 recombinant inbred lines (RILs) weregenotyped, revealing that two pairs of unlinkedloci did not segregate independently from eachother (10). For both of these pairs, a specific com-bination of Col and Cvi alleles was not found inthe RIL set, resulting in a transchromosomal link-age disequilibrium pattern (pseudo-LD) and ex-hibiting segregation distortion (fig. S1). Byfocusing on one of these two-locus interactions(labeled LD1 in fig. S1), we realized that a homo-zygous combination of the Col allele at the LD1.1

1Genetics and Plant Breeding, INRA, SGAP UR254, F-78026Versailles, France. 2Centre for Plant Integrative Biology,Division of Plant Sciences, School of Biosciences, Universityof Nottingham, Sutton Bonington Campus, LoughboroughLE12 5RD, UK.



REPORTS

...

......


locus (bottom of chromosome 1) and the Cviallele at the LD1.5 locus (top of chromosome 5)caused arrested embryo development, resultingin seed abortion (Fig. 1 arrows). A large F2 pop-ulation generated from the same cross recapitu-lated this recessive incompatibility (Fig. 1) andshowed complete penetrance. We also noted that,in one intermediate heterozygous combination,the embryo developed normally but the primaryroot of the resulting seedling was shortened to~one-third of its regular size, adding a quantita-tive phenotype we named “weak root” to thecomplexity of this epistasis. Taking advantageof heterogeneous inbred families [HIFs; (16)] de-rived from F6 RILs segregating for each locuswhile being fixed for the appropriate (incompat-ible) genotype at the other locus, we fine-mappedthe LD1 epistatic interaction (17), reducing thecandidate intervals of LD1.1 and LD1.5 to 65 and15 kb, containing 11 and 4 annotated genes,respectively.

Gene annotation within these intervals revealeda candidate gene pair; the histidinol-phosphateamino-transferase gene codes for a protein (HPA)that catalyzes an important step in the biosyn-thetic pathway leading to histidine (His), anessential amino acid incorporated into proteinsand hence required in many aspects of plantgrowth and development (18). Two paralogouscopies of this gene are found in the Col ge-nome; one (HPA1/HISN6A/At5g10330) lies with-in the LD1.5 candidate interval, and the other(HPA2/HISN6B/At1g71920) lies within the LD1.1candidate interval. These paralogs appear to have

arisen from a recent single gene duplication eventresulting in a dispersed duplicate pair, a mode ofgene duplication that typically affects fewer genesthan tandem or segmental duplication events inA. thaliana (19). InCol,At1g71920 andAt5g10330coding sequences differ by two synonymous sin-gle nucleotide polymorphisms (SNPs). Intraspe-cific sequence analyses and comparison to A. lyratashow that the ancestral locus is represented byAt1g71920 (19) and that At5g10330 arose froma 3.3-kb duplication centered on At1g71920. Inthe Col background, transferred DNA (T-DNA)–insertion mutants in At5g10330 are embryo-lethalat the homozygous state [emb2196; (20)], and wehave confirmed this result in other mutant lines,including SALK_089516. Recently, a weak pointmutation allele of At5g10330 (hpa1) was shownto affect the maintenance of the root meristemand primary root elongation (21), an effect typ-ically associated with the lack of free His in theplant. Both arrested embryo development androot growth impairment have been comple-mented by supplying emb2196 and hpa1 plantswith exogenous histidine (18, 21).

The SNPs in the coding sequence were usedto distinguish between the two paralogs’mRNA,confirming that there is no detectable transcrip-tional activity from At1g71920 in Col, whereasAt5g10330 is expressed in both shoot and root[fig. S2; (19)]. Sequence and reverse transcrip-tion polymerase chain reaction (RT-PCR) analy-ses in Cvi showed that the gene at LD1.1(At1g71920) was expressed, whereas there wereno traces of an HPA coding sequence at LD1.5 in

this background (fig. S2). Instead, according tosequence results, a 6.4-kb region appears to bedeleted in Cvi (compared with the Col sequence)that encompasses the entire Col duplication stretch-ing to 3 kb beyond the Col duplicated region onone side. Furthermore, the borders of this addi-tional 3-kb deletion contained traces of transpos-able elements: ATGP9-LTR andVANDAL-18NA.The region homologous to LD1.5 inA. lyrata lacksa HPA gene but is also slightly different from Cvi,with deletions and insertions of hundreds of basepairs at the locus. Then, the possibility that theLD1.5 locus might have undergone multiple re-arrangements makes it difficult to determinewhether the HPA gene was deleted in Cvi orwhether it was ever present in this background.

The situation at each paralog is summarizedin Fig. 1, with Col and Cvi retaining alternatefunctional copies of the essential HPA gene. Thecombination of the two silenced copies in a prog-eny homozygous for the Col allele at LD1.1 andthe Cvi allele at LD1.5 leads to arrested seeddevelopment, presumably because the embryois unable to synthesize His. Similarly, the limitedprimary root growth in weak-root plants could beexplained by a reducedHis quantity in these plantsbecause they have a single functional HPA copyoriginating from the Col LD1.5 locus (Fig. 1).Given that the Col LD1.5 allele is less expressedthan the Cvi LD1.1 allele (fig. S2), this alleliccombination may be specifically limiting for theroot, an organ particularly sensitive to a shortagein His (21). All other genotypes should result ingreater HPA activity, enough to sustain normal

Fig. 1. Geneticmechanism underlyingLD1 interaction and incompatibility. Allcombinations of alleles at LD1.5 and itsinteractor LD1.1 result in phenotypical-ly normal plants, except for the combi-nation Col at LD1.1/Heterozygous (Het)at LD1.5, which shows a reduced pri-mary root length (weak-root pheno-type), and another combination (Colat LD1.1/Cvi at LD1.5), which showsembryo lethality at an early stage inthe silique (seeds under developmentcontaining lethal embryos are indicatedby arrows). Fine-mapping identified aduplicate gene for which Col and Cvishow reciprocal gene loss explainingthe interaction and incompatibility.


REPORTS

...

......


embryo and root development.Moreover, limitedroot growth in weak-root individuals can be ex-plained by an alteration of the cell production ratebecause cell elongation remains normal in theseplants (fig. S3), which is consistent with the lackof His and its associated defect in root meristemmaintenance (21).

One way to prove the link between LD1and HPA genes is to show that LD1 can be

complemented by adding exogenous His. Weak-root plants were gradually and quantitatively res-cued to the phenotype of the control by growingthe plants on increasing concentrations of His invitro (Fig. 2). Furthermore, watering heterozygousplants from the bolting stage with sufficientamounts of His restored the viability of embryoswith incompatible allelic combinations (17), pro-viding complementation of the embryo lethality

phenotype as well. From these experiments, weconcluded that the LD1 incompatibility resultsfrom a shortage of His in certain genotypes.

To prove the causative role of HPA geneson LD1, we performed an allelic (quantitative)complementation test by combining different al-leles at LD1.5 and At5g10330 in an identical F1background. Crossing segregating HIF lines todifferent mutants (hpa1 and emb2196) revealedhow the different alleles at the two duplicate genesinteract to qualitatively control embryo develop-ment and quantitatively limit primary root growth(Fig. 3). From crosses with the hpa1mutant (Fig.3A), we observed that a Cvi allele at LD1.5 isunable to complement the EMS mutant allele atAt5g10330 (whereas a Col allele does). This geno-type (Cvi/hpa1) leads to embryo lethality, a pheno-type even stronger than homozygous hpa1mutants(these embryos survive), indicating that the caus-ative Cvi allele at LD1.5 is more deleterious thanthe EMS mutation (which fits well with the genebeing completely deleted in Cvi). The significantinteraction between the LD1.5 alleles and HPAgenotypes argues that HPA is involved in LD1epistasis. Crosses to the emb2196 mutant led tothe same conclusion when Col and Cvi alleleswere compared versus the stronger emb2196 al-lele (Fig. 3B). However, a heterozygous seed-ling for emb2196 mutation (Col/emb2196 atAt5g10330) had a significantly (P < 0.01) longerprimary root than a heterozygous seedling with aCvi/Col genotype at LD1.5 (Fig. 3B). This, to-gether with the fact that another T-DNA line inthe same gene (SALK_089516) has a shorter rootthan emb2196 in the heterozygous state (similarto weak-root plants), indicated that the emb2196T-DNA insertion is probably not a complete nullallele (in contrast to SALK_089516). From theseexperiments, we concluded that epistasis at LD1is most likely explained by allelic variation atHPA genes as depicted in Fig. 1 and that the in-compatibility observed between Col and Cvi rep-resents an example of intraspecific divergence ofa duplicate gene pair.

Similar patterns of segregation distortioninvolving regions at the bottom of chromo-some 1 and top of chromosome 5 were detectedin a RIL set derived from the cross betweenCvi and Landsberg erecta (Ler), indicatingthat there may be a similar interaction betweenthese loci (22). Nearly isogenic lines derivedfrom this population (23) confirmed epistasisand incompatibility by showing a specific pat-tern of segregation and the weak-root pheno-type. Both HPA genes were expressed in Ler;however, the chromosome 1 paralog containsa premature stop codon (Fig. 4). This encour-aged us to analyze the extent of functionalnatural variation at those loci and further char-acterize intraspecific evolution of these genesin A. thaliana.

We analyzed 30 accessions derived fromdistinct natural populations representing mostknown variation in Arabidopsis (24) for theexpression of each gene copy, potential deletions,

Fig. 2. The weak-root phenotype is quantitatively complemented by exogenous His. Typicalphenotypes of descendants from a plant segregating at the LD1.5 locus when grown on mediasupplemented with different histidine concentrations. Plants were identified on the basis of theirgenotype at the LD1.5 locus. The complementation of the expected weak-root phenotype wascomplete when supplied with 10−2 mM His.

Fig. 3. Allelic complementation of LD1 interaction. Quantitative complementation tests were performedby combining different alleles at LD1 loci in F1 backgrounds. The relative complementation of either (A)an EMS mutant allele (hpa1) or (B) a T-DNA insertion mutant allele (emb2196) at At5g10330 by a Col orCvi allele at LD1.5 was measured through both the length of the primary root (shown in mm) and embryolethality. Individuals with a phenotype depicted on the x axis (null) underwent seed abortion. All plants areCol at LD1.1. Arrows along the y axis represent the typical root length of control genotypes [WT indicateswild-type Col plants; hpa1 (Hom), homozygous hpa1 mutant plants; and emb2196 (Het), heterozygousindividuals for the emb2196mutation]. Each data point represents the mean T standard error of about 40plants. The LD1.5 allele × HPA1 genotype interaction term as tested by analysis of variance is significantin crosses to hpa1 (P < 0.0001) and crosses to emb2196 (P < 0.05).


REPORTS

...

......


and deleterious mutations. We also looked forLD1.1/LD1.5 genetic incompatibilities (accom-panied by segregation of the weak-root pheno-type) in 28 of the possible crosses to Col or Cvi(either in RIL sets or F2 populations) to confirmthe relevance of variation detected at the nucleo-tide and/or expression level. Divergent evolutionof the HPA gene pair was dramatic and wide-spread (Fig. 4 and table S1) because most (22/30)of the accessions tested have silenced one or theother copy in at least six different ways (earlystop, no expression, a combination of both, anddifferent deletions). Cvi- and Col-incompatiblegroups represent 14 and 8 accessions, respective-ly, and, on the basis of these data, we estimatethat at least one-fourth of all possible crossesamong these 30 strains would show HPA in-compatibility. We observed no particular correla-tion with geographical population structure (25).Most of the four structure groups defined pre-viously (26) included strains belonging to thetwo incompatibility groups as well as compatibleaccessions.

We confirmed that the rapid and commonevolution of duplicate genes provides an impor-tant source of Bateson-Dobzhansky-Muller–likeepistatic interactions after paralogs have been re-ciprocally silenced or lost in diverged strains, asproposed (27, 28). Depending on the functionand essential nature of the gene, this may resultin hybrid and/or F2 defective fitness in the de-scendants of certain intraspecific crosses andcould contribute to reproductive isolation (29).Passive gene loss is recognized as the most prob-able fate of duplicate genes and is especiallylikely at early stages after small-scale duplica-tion events (12, 30). However, direct evidenceof gene loss as a neutral mechanism generatingpostzygotic isolating barriers within existing spe-

cies with no prior fitness consequences in theparental strains (because only the location ofthe functional copy changes) has been lacking(4, 31, 32). Staal et al. (33) described how atransposition of a resistance gene found in theLer strain of Arabidopsis was responsible for var-iation in disease susceptibility in crosses to Col.Similarly, transposition of an essential gene hasrecently been associated with the sterility of a hy-brid between two Drosophila species (29), anddescriptive work on three related yeast speciesindicated that divergent resolution events afterwhole-genome duplication may have contributedto their speciation (34). Our study extends theseobservations in demonstrating the link betweengene duplication and genetic incompatibility.

References and Notes1. K. Bomblies, D. Weigel, Nat. Rev. Genet. 8, 382 (2007).2. J. Mallet, Trends Ecol. Evol. 21, 386 (2006).3. T. Dobzhansky, Genetics and the Origin of Species

(Columbia Univ. Press, New York, 1937).4. J. A. Coyne, H. A. Orr, Speciation, M. A. Sunderland, Ed.

(Sinauer, Sunderland, MA, 2004).5. K. Bomblies, D. Weigel, Curr. Opin. Genet. Dev. 17, 500

(2007).6. A. Kopp, A. Frank, Genetica 125, 55 (2005).7. N. J. Brideau et al., Science 314, 1292 (2006).8. H. S. Seidel, M. V. Rockman, L. Kruglyak, Science 319,

589 (2008); published online 10 January 2008(10.1126/science.1151107).

9. K. Bomblies et al., PLoS Biol. 5, e236 (2007).10. M. Simon et al., Genetics 178, 2253 (2008).11. O. Törjék et al., Theor. Appl. Genet. 113, 1551 (2006).12. M. Lynch, J. S. Conery, Science 290, 1151 (2000).13. H. A. Orr, J. P. Masly, N. Phadnis, J. Hered. 98, 103

(2007).14. H. A. Orr, M. Turelli, Evolution 55, 1085 (2001).15. H. A. Orr, J. P. Masly, D. C. Presgraves, Curr. Opin. Genet.

Dev. 14, 675 (2004).16. O. Loudet, V. Gaudon, A. Trubuil, F. Daniel-Vedele,

Theor. Appl. Genet. 110, 742 (2005).17. Materials and methods are available as supporting

material on Science Online.

18. R. Muralla, C. Sweeney, A. Stepansky, T. Leustek,D. Meinke, Plant Physiol. 144, 890 (2007).

19. R. C. Moore, M. D. Purugganan, Proc. Natl. Acad. Sci.U.S.A. 100, 15682 (2003).

20. I. Tzafrir et al., Plant Physiol. 135, 1206 (2004).21. X. Mo et al., Plant Physiol. 141, 1425 (2006).22. C. Alonso-Blanco et al., Plant J. 14, 259 (1998).23. J. J. Keurentjes et al., Genetics 175, 891 (2007).24. H. I. McKhann et al., Plant J. 38, 193 (2004).25. M. Nordborg et al., PLoS Biol. 3, e196 (2005).26. M. F. Ostrowski et al., Mol. Ecol. 15, 1507 (2006).27. M. Lynch, A. G. Force, Am. Nat. 156, 590 (2000).28. C. R. Werth, M. D. Windham, Am. Nat. 137, 515

(1991).29. J. P. Masly, C. D. Jones, M. A. F. Noor, J. Locke, H. A. Orr,

Science 313, 1448 (2006).30. S. Maere et al., Proc. Natl. Acad. Sci. U.S.A. 102, 5454

(2005).31. M. Lynch, J. S. Conery, J. Struct. Funct. Genomics 3, 35

(2003).32. M. Semon, K. H. Wolfe, Curr. Opin. Genet. Dev. 17, 505

(2007).33. J. Staal, M. Kaliff, S. Bohman, C. Dixelius, Plant J. 46,

218 (2006).34. D. R. Scannell, K. P. Byrne, J. L. Gordon, S. Wong,

K. H. Wolfe, Nature 440, 341 (2006).35. We thank K. Bomblies and D. Weigel for providing F2

seeds for specific test crosses (Bay-0 × Cvi and Tsu-0 ×Cvi), X. Mo for hpa1 seeds, and R. Mercier and F. Rouxfor discussion and critical reading of the manuscript.Funded partly by an INRA–Département de Génétique etd’Amélioration des Plantes Innovative Project grant toC.C. and O.L. Funding by the University of Nottingham toD.P. and the Biotechnology and Biological SciencesResearch Council/ Engineering and Physical SciencesResearch Council Centres for Integrative SystemsBiology program to M.J.B. Sequences have beendeposited in GenBank under accessions FJ436446and FJ436447.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/323/5914/623/DC1Materials and MethodsFigs. S1 to S3Table S1References

15 September 2008; accepted 25 November 200810.1126/science.1165917

Fig. 4. Divergent evolution of du-plicate genes among A. thalianaaccessions. Groups of accessions arepresented according to At5g10330and/or At1g71920 genotypes andtranscript accumulation pheno-types. Accessions (underlined) fromeach group were crossed to Coland/or Cvi and tested for LD1incompatibility and compatibilityto confirm the loss of function ofone or the other duplicate gene.Dash-underlined accessions showconditional incompatibility (tableS1). The incompatible groups arecircled: green-circled genotypes areincompatible with Cvi; purple-circledgenotypes are incompatible withCol. Genotypes not circled are fullycompatible with both Col and Cvi.


REPORTS

...

......


Serotonin Mediates BehavioralGregarization Underlying SwarmFormation in Desert LocustsMichael L. Anstey,1*StephenM.Rogers,1,2*†Swidbert R.Ott,2MalcolmBurrows,2 Stephen J. Simpson1,3

Desert locusts, Schistocerca gregaria, show extreme phenotypic plasticity, transforming betweena little-seen solitarious phase and the notorious swarming gregarious phase depending onpopulation density. An essential tipping point in the process of swarm formation is the initialswitch from strong mutual aversion in solitarious locusts to coherent group formation and greateractivity in gregarious locusts. We show here that serotonin, an evolutionarily conserved mediatorof neuronal plasticity, is responsible for this behavioral transformation, being both necessaryif behavioral gregarization is to occur and sufficient to induce it. Our data demonstrate aneurochemical mechanism linking interactions between individuals to large-scale changesin population structure and the onset of mass migration.

Phenotypic plasticity, the differential expres-sion of alternative phenotypes from a sin-gle genotype depending upon environmental

conditions, is of considerable evolutionary im-portance, but controlling mechanisms remainelusive (1). Changes in population density can bea substantial source of environmental variabil-ity and can trigger phenotypic changes that equipanimals for increased competition for resourcesas well as for dispersal or migration (1–3). Desertlocusts undergo an extreme and economicallydevastating form of this kind of phenotypic plas-ticity, changing reversibly between two extremeforms or phases (4–6). These differ extensively inmorphology and physiology, but behavior is thekey to both establishing and maintaining eachphase (7). Swarming begins with a rapidly in-duced switch frommutual repulsion in solitariouslocusts to attraction and aggregation after just afew hours of forced crowding (7, 8). Althoughthe sensory stimuli triggering behavioral gregar-ization have recently been identified (9–11), itwas unknown how they mediated their effect.

Solitarious locusts acquire full gregarious be-havioral characteristics within the first 2 hours offorced crowding [Fig. 1, A to C, fig. S1, and sup-porting online material (SOM) text]. This periodcoincides with a substantial but transient (<24hours) increase in the amount of serotonin [5-hydroxytryptamine (5-HT)] specifically in oneregion of the central nervous system (CNS), thethoracic ganglia, but not the brain (12) (Fig. 1D).To determine whether this increase caused gre-garization, we first analyzed the relationship be-tween the degree of behavioral gregarization andthe amountof serotonin in individual locusts crowdedfor different times (13). Behavior was character-

ized by using a binary logistic regression model(8), which produced a single probabilistic metricof gregariousness Pgreg that encompassed four dif-ferent variables (table S1). A Pgreg of 0 meant ananimal behaved solitariously, whereas a Pgreg of1 indicated fully gregarious behavior. Amountsof serotonin in the thoracic ganglia were mea-sured by using high-performance liquid chroma-tography (HPLC).

We crowded solitarious locusts for 0, 1, or2 hours to generate the entire gamut of behavior,from solitarious to gregarious (Fig. 2A). The

amount of serotonin was significantly positivelycorrelated with the extent of gregarious behavioracross this entire range (analysis of covariance,5-HT loge transformed; F1,35 = 21.817, r2 =0.429, P < 0.001). Locusts that behaved the mostgregariously (Pgreg > 0.8) had approximately threetimes more serotonin (12.78 T 1.85 pmol; mean TSD, n = 10 locusts) than more solitariouslybehaving (Pgreg < 0.2) locusts (4.18 T 0.27 pmol;n = 7 locusts). Furthermore, the amount of sero-tonin only corresponded with the degree of gre-garization but not the duration of crowding, perse (F3,35 = 1.218, P = 0.318).

Behavioral gregarization can be acquired viatwo distinct sensory pathways: a thoracic path-way driven by mechanosensory stimulation ofthe hind legs as locusts jostle each other and acephalic pathway in which the combined sightand smell of other locusts is the necessary stim-ulus (9–11). Locusts stimulated via either sensorypathway displayed similar levels of gregariousbehavior after 2 hours (Fig. 1C, fig. S1, and SOMtext).We tested whether gregarization induced bythese separate pathways showed the same rela-tionship between serotonin and behavior that wesaw in crowded solitarious locusts. The thoracicpathway was activated by either stroking a hindfemur (10) or electrically stimulating metatho-racic nerve 5, which simulated mechanosensorystimulation (11). In both instances, the amount ofserotonin in the thoracic ganglia significantlyincreased and was correlated with the extent of

1Department of Zoology, University of Oxford, South Parks Road,Oxford, OX1 3PS, UK. 2Department of Zoology, University ofCambridge, Downing Street, Cambridge, CB2 3EJ, UK. 3Schoolof Biological Science, University of Sydney, Sydney, NSW 2006,Australia.


Fig. 1. (A) Final larvalinstar solitarious and gre-garious locusts. Scale bar,1 cm. (B) Trajectories (over500 s) of a solitarious(upper) and gregarious(lower) locust in the be-havioral arena. A group of20 long-term gregarious-phase locusts was placedbehind a clear partitionon the left. (C) Solitariouslocusts undergo rapid be-havioral gregarizationwithappropriate stimulation;median Pgreg of locuststreated for 0 to 4 hoursby either forced crowd-ing with gregarious lo-custs (circles), stroking ahind femur (squares), elec-trically stimulating theprincipal hind-leg nerve(diamonds), or exposureto the sight and smell ofother locusts (triangles).See SOM text for analysis. (D) The CNS of alocust consists of the brain, which receives ol-factory and visual information, and a chain ofsegmental ganglia linked by paired connectives.The three thoracic ganglia receive direct mech-anosensory and proprioceptive inputs from thelegs.


REPORTS

...

......


gregarization [linear regressions; in stroked locusts,F1,24 = 15.027, P = 0.001, r2 = 0.39 (Fig. 2B); inelectrically stimulated locusts, F1,13 = 7.457, P =0.017; r2 = 0.37 (Fig. 2C)]. Likewise, gregariza-tion through exposure to just the sight and smellof other locusts led to an increase in the amountof serotonin in the thoracic ganglia that correlatedwith the degree of behavioral gregarization [F1,11 =23.065, P = 0.001, r2 = 0.7 (Fig. 2D)]. Theseresults indicate that gregarizing stimuli from bothsensory pathways converge in the thoracic ganglia,but it is unknownwhether they have a cumulativeeffect on serotonin production.

This strong correlation between phase stateand serotonin levels in the thoracic ganglia ledus to ask whether blocking the action of serotonincould prevent behavioral gregarization. Amixtureof two 5-HTreceptor antagonists, ketanserin (1mM)(14, 15) and methiothepin (1 mM) (16, 17), or asaline control were injected directly into the meso-and metathoracic ganglia of solitarious locusts.The locusts then received either mechanosensoryor olfactory and visual gregarizing stimuli for1 hour. The locusts injected with the antagonistsfailed to gregarize in response to either stimulusregime (Fig. 3A), in contrast to the saline-injectedcontrols [analysis of variance (ANOVA) of nor-mal rank transformed data, F1,49 = 17.169, P <0.0005; there was no interaction between stimu-lus regime and degree of gregarization, F1,49 =0.001,P= 0.987].MedianPgreg in the antagonist-injected locusts was 0.27 for the thoracic and0.07 for the cephalic pathways (Fig. 3A). Bycontrast, the median Pgreg values of the saline-

injected controls were 0.91 and 0.74, respective-ly. Next, we inhibited serotonin synthesis by usinga-methyltryptophan (AMTP), a competitive an-tagonist of tryptophan hydroxylase (18, 19). Lo-custs were given repeated systemic injections ofeither 40 ml of 0.1 mM AMTP in locust saline orjust saline (controls) over 5 days (13). TheAMTP-injected locusts (Fig. 3B) showed little behavioralgregarization after having their hind-femorastroked for 2 hours (median Pgreg = 0.13), instrong contrast to the controls (median Pgreg =0.91; Mann-WhitneyU = 30.500, n = 23 locusts,P = 0.032).

We next determined whether serotonin orserotonin receptor agonists were sufficient toinduce behavioral gregarization in the absenceof stimuli associatedwith other locusts. Serotonin(1mM) in saline or a saline control were topicallyapplied to the exposed thoracic ganglia over 2hours. Serotonin-treated locusts (Fig. 4A) weresignificantlymore gregarious in behavior (medianPgreg = 0.6) than control animals, which remainedhighly solitarious (median Pgreg = 0.07; Mann-Whitney U = 41, n = 30 locusts, P = 0.004). In asecond experiment, animals injected in the thoracicganglia with a mixture of two serotonin receptor ag-onists, 1 mM a-methylserotonin (20, 21) and 1 mM5-carboxamidotryptamine (16), showed a signifi-cant shift toward gregarious behavior (medianPgreg =0.4) as comparedwith saline-injected controlsafter 1 hour [medianPgreg = 0.13;Mann-WhitneyU = 63, n = 30 locusts, P = 0.042 (Fig. 4B)].

Gregarizing stimuli cause serotonin to in-crease in the thoracic CNS, and exogenous sero-

tonin increases the likelihood of locusts behavinggregariously. We next asked whether enhancedendogenous serotonin synthesis could amplifythe effect of stimuli presented for a brief period.Locusts were given single 40-ml injections in thethoracic haemocoel of either the serotonin pre-cursor 5-hydroxytryptophan (5-HTP) (10 mM)(22, 23) or saline controls, and their behavior wasassayed after either 30 min of further solitude or30 min of crowding. Treatment regime had asignificant effect on behavior [ANOVA of nor-mal rank transformed Pgreg data; F3,59 = 5.6, P =0.002 (Fig. 4C)]. Control locusts that just re-ceived saline and were kept in isolation remainedhighly solitarious in behavior (median Pgreg =0.17). ThemedianPgreg of crowded saline-injectedlocusts was 0.46, suggesting some change in

Fig. 2. The amount ofserotonin in the thoracicCNS is correlated with thedegree of behavioral gre-garization. Relationshipsbetween the amount ofserotonin in the thoracicCNS (loge scale) and thedegree of behavioral gre-garization (Pgreg) after soli-tarious locustshavebeen (A)crowded for 1 hour (trian-gles), 2 hours (squares),or unstimulated controls(circles); (B) stroked on theleft hind femur for 2 hours;(C) given patterned elec-trical stimulation to meta-thoracic nerve 5, simulatingthe effect of mechanosen-sory stimulation for 2hours;or (D) presented with thesight and smell of ~1000locusts for 2 hours.

Fig. 3. Serotonin is necessary to induce behavior-al gregarization. Behavior of locusts injected withsubstances that block either the action or the syn-thesis of serotonin and then exposed to sensorystimuli that normally induce gregarization (left column)is shown. Saline-injected controls are shown in theright column. (A) Locusts injected with serotonin-receptor antagonists ketanserin and methiothepin(1 mM) and given either 1 hour of femoral mech-anosensory stimulation or 1 hour of olfactory andvisual stimulation from other locusts. (B) Locustsinjected with 0.1 mM AMTP, an inhibitor of sero-tonin synthesis coupled with 2 hours of mechano-sensory stimulation. Arrows indicate median Pgregvalues.


REPORTS

...

......


behavior; however, the distribution was not sig-nificantly different from isolated saline-injectedlocusts (Dunnet’s post-hoc test, P = 0.383),which indicates that 30 min was too brief a pe-riod of crowding to induce full behavioralgregarization. Locusts that received an injectionof 5-HTP but were not crowded also showed asimilar but also nonsignificant increase in Pgreg(median 0.38; P = 0.448), which indicates thatthis treatment was also insufficient to induce fullbehavioral gregarization. However, locusts thathad received both the serotonin precursor andhad also been exposed to a brief period of crowd-ing became highly gregarious (median Pgreg =0.81, P = 0.001 as compared with undruggedand uncrowded controls). Thus, 5-HTP can po-tentiate the effect of gregarizing stimuli appliedfor a brief period. Unlike the direct application ofserotonin, this experiment suggests that endoge-nous serotonin synthesis driven by sensory stimuliis mechanistically responsible for inducing behav-ioral gregarization. There are few serotonergicneurons in the locust CNS (24), which suggeststhat the individual neurons driving behavioralgregarization can be identified.

Serotonin and other monoamines have beenimplicated in changing behavior after social inter-actions in a number of contexts, including intra-specific aggression, status, and courtship in manyspecies, including crickets (25), crustaceans (26, 27),and rats (28). All of these interactions, includingbehavioral gregarization in locusts, require theinterpretation of complex signals from conspecif-ics leading to long-lasting changes in the wayindividuals interact during future encounters. Be-havioral gregarization therefore resembles mem-ory formation, with specific sensory experiences

altering future behavior; in the case of locusts,this entails a suite of changes that creates an in-tegrated behavioral phenotype adapted to a changedbiotic environment. Locusts that have been rearedgregariously for many generations have lower titersof serotonin than long-term solitarious animals,(12) which strongly suggests that gregarious be-havior is not maintained by a long-term seroto-nergicmodulation of neuronal circuits. Furthermore,solitarious behavior is acquired more slowly onthe isolation of long-term gregarious phase lo-custs (8) than gregarious behavior is acquired bysolitarious locusts, implying that gregarious be-havior becomes more ingrained during prolongedcrowding. This ingraining process may entailserotonin-mediated gene transcription and/ortranslation-dependent mechanisms similar tothose associated with other serotonin-mediatedneuronal plasticity (29, 30). Although serotoninclearly mediates the critical change in behaviorthat drives the early process of gregarization, wedo not know whether it directly initiates the fulland complex suite of changes associated with thefull gregarious phenotype. Serotonin, by provid-ing a rapidly acquired and stable behavioral sub-strate for group living, may enable slower butfully independent mechanisms of phenotypicchange to activate through a process of envi-ronmental feedback (4, 5, 7, 31).

Could serotonin antagonists be effective lo-cust control agents? Given the ubiquity of sero-tonin signaling in the animal kingdom, any agentwould have to be specific for the serotoninreceptor mediating phase change, which is yet tobe characterized. To be effective, it would have tobe targeted at regions of incipient swarm forma-tion to prevent locusts coalescing further into groups.

The whole multilayered process of phasechange depends upon a simple behavioral deci-sion whether to avoid other locusts or band to-gether. Without this initial behavioral choice, nofurther physiological and morphological changecan occur (12, 31), and there is no possibility offurther escalation in group size. Our data dem-onstrate a mechanism by which sensory signalsgauging population density alter neuronal circuitsunderlying this fundamental decision. The con-sequences ramify upwards into population struc-ture and ultimately provide the essential conditionsfor mass migration and swarming. Phase changeis the defining character of locust biology and isthe reason why theymake such devastating pests.Serotonin-mediated behavioral plasticity is a piv-otal mechanism in this transformation.

References and Notes1. M. J. West-Eberhard, Developmental Plasticity and

Evolution (Oxford Univ. Press, Oxford, 2003).2. J. Buhl et al., Science 312, 1402 (2006).3. S. Bazazi et al., Curr. Biol. 18, 735 (2008).4. B. Uvarov, Grasshopper and Locusts, vol. 1 (Cambridge

Univ. Press, Cambridge, 1966).5. M. P. Pener, Y. Yerushalmi, J. Insect Physiol. 44, 365

(1998).6. M. Enserink, Science 306, 1880 (2004).7. S. J. Simpson, A. R. McCaffery, B. F. Hagele, Biol. Rev.

Camb. Philos. Soc. 74, 461 (1999).8. P. Roessingh, S. J. Simpson, S. James, Proc. R. Soc.

London B Biol. Sci. 252, 43 (1993).9. P. Roessingh, A. Bouaichi, S. J. Simpson, J. Insect Physiol.

44, 883 (1998).10. S. J. Simpson, E. Despland, B. F. Hagele, T. Dodgson,

Proc. Natl. Acad. Sci. U.S.A. 98, 3895 (2001).11. S. M. Rogers et al., J. Exp. Biol. 206, 3991 (2003).12. S. M. Rogers et al., J. Exp. Biol. 207, 3603 (2004).13. Materials and methods are available as supporting

material on Science Online.14. D. Parker, J. Neurophysiol. 73, 923 (1995).15. L. Gatellier, T. Nagao, R. Kanzaki, J. Exp. Biol. 207, 2487

(2004).16. A. J. Tierney, Comp. Biochem. Physiol. 128A, 791

(2001).17. N. Spitzer, D. H. Edwards, D. J. Baro, J. Exp. Biol. 211, 92

(2008).18. B. D. Sloley, S. Orikasa, J. Neurochem. 51, 535 (1988).19. P. A. Stevenson, H. A. Hofmann, K. Schoch, K. Schildberger,

J. Neurobiol. 43, 107 (2000).20. G. Molaei, A. B. Lange, J. Insect Physiol. 49, 1073

(2003).21. J. F. Colas, J. M. Launay, O. Kellermann, P. Rosay,

L. Maroteaux, Proc. Natl. Acad. Sci. U.S.A. 92, 5441(1995).

22. M. Ureshi, M. Dainobu, M. Sakai, J. Comp. Physiol. A188, 767 (2002).

23. N. N. Osborne, V. Neuhoff, Brain Res. 74, 366 (1974).24. N. M. Tyrer, J. D. Turner, J. S. Altman, J. Comp. Neurol.

227, 313 (1984).25. H. A. Hofmann, P. A. Stevenson, Nature 403, 613 (2000).26. R. Huber, K. Smith, A. Delago, K. Isaksson, E. A. Kravitz,

Proc. Natl. Acad. Sci. U.S.A. 94, 5939 (1997).27. E. A. Kravitz, R. Huber, Curr. Opin. Neurobiol. 13, 736

(2003).28. K. A. Miczek et al., J. Neurosci. 27, 11803 (2007).29. C. Pittenger, E. R. Kandel, Philos. Trans. R. Soc. London B

358, 757 (2003).30. C. A. Hoeffer, S. Sanyal, M. Ramaswami, J. Neurosci. 23,

6362 (2003).31. A. I. Tawfik et al., Proc. Natl. Acad. Sci. U.S.A. 96, 7083

(1999).32. We thank J. Niven, G. Sutton, and G. Sword for reading

and commenting upon the manuscript; C. de la Riva atthe Babraham Institute (UK) for technical support withthe HPLC and K. Kendrick for permission to use the

Fig. 4. Serotonin is sufficient to induce gregarious behavior. Behavior of locusts after they havebeen treated with (A) serotonin topically applied to the thoracic ganglia (left column) and pairedsaline controls (right column); (B) a mixture of serotonin agonists (1 mM a-methylserotonin and1 mM 5-carboxamidotryptamine) and paired saline-injected controls; and (C) the serotonin precursor5-HTP, either with or without 30 min of crowding with other locusts.


REPORTS

...

......


facility; T. Dodgson and E. Miller for locust rearing andtechnical support; and M. Ungless for advice and supportto M.A. This work is supported by grants from theBiotechnology and Biological Sciences Research Council,UK (to M.B, S.S., and S.O.), and a University ResearchFellowship from the Royal Society, UK, to S.O. M.A. wassupported by the Natural Sciences and EngineeringResearch Council of Canada, an Overseas Research

Student Award, UK, and the Hope Studentship inEntomology from Jesus College and the University ofOxford, UK. S.S. is supported by an Australian ResearchCouncil Federation Fellowship.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/323/5914/627/DC1Materials and Methods

SOM TextFig. S1Table S1References

15 September 2008; accepted 14 November 200810.1126/science.1165939

Survival from Hypoxia inC. elegans by Inactivation ofAminoacyl-tRNA SynthetasesLori L. Anderson,1 Xianrong Mao,1 Barbara A. Scott,1 C. Michael Crowder1,2*

Hypoxia is important in a wide range of biological processes, such as animal hibernation and cellsurvival, and is particularly relevant in many diseases. The sensitivity of cells and organisms tohypoxic injury varies widely, but the molecular basis for this variation is incompletely understood.Using forward genetic screens in Caenorhabditis elegans, we isolated a hypoxia-resistantreduction-of-function mutant of rrt-1 that encodes an arginyl–transfer RNA (tRNA) synthetase, anenzyme essential for protein translation. Knockdown of rrt-1, and of most other genes encodingaminoacyl-tRNA synthetases, rescued animals from hypoxia-induced death, and the level ofhypoxia resistance was inversely correlated with translation rate. The unfolded protein responsewas induced by hypoxia and was required for the hypoxia resistance of the reduction-of-functionmutant of rrt-1. Thus, translational suppression produces hypoxia resistance, in part byreducing unfolded protein toxicity.

Oxygen requirements of cells and orga-nisms have wide-ranging implications inbehavior and disease. Forward genetic

screens offer the possibility of discovering genesnot previously known to control hypoxic sensi-tivity. Such genes are likely to play an importantrole in emergent organismal traits such as habitatrange and ability to hibernate. Additionally, thesegenes may lead to the development of noveltherapies for conditions where cellular hypoxicsensitivity is a pathological determinant, such asstroke, myocardial infarction, and cancer. Wild-type C. elegans when placed in a severe hypoxicenvironment (oxygen concentration <0.3 volumepercent) become immobile but fully recover whenreturned to normoxia within 4 hours (1). After4 hours, permanent behavioral deficits and cel-lular death ensue, and after a 22-hour hypoxicincubation, >99% of wild-type animals are dead.To identify genes that control hypoxic sensitivity,we screened for ethylmethane sulfonate (EMS)–derived mutants that survived a 22-hour hypoxicincubation. In a screen of 3884 F1 mutant wormgenomes, we recovered 14 mutants that had ahypoxia-resistant phenotype (table S1). These mu-tants fell into 13 complementation groups. We se-lected gc47, one of the strongest hypoxia-resistantmutants, for further characterization and mapping.

After outcrossing to the wild-type strain N2,the hypoxia resistance of gc47 was quantified.Immediately after removal from a 20-hour hy-poxic incubation, both N2 and gc47 were para-lyzed, but the gc47 worms recovered the abilityto move completely over the next 1 to 2 hours.After a 24-hour recovery, essentially all of the

gc47 animals were alive, whereas almost all wild-type worms failed to survive (Fig. 1, A and B);gc47 prolonged the hypoxic incubation timerequired for complete killing by a factor of >3(Fig. 1C). The hypoxia-resistant phenotype wasfully recessive and segregated as a single locus ina Mendelian fashion (Fig. 1D) (2). gc47 wasmapped to a 106-kb interval on the left arm ofchromosome III (Fig. 2A) (2). Double-strandedRNA interference (RNAi) of 29 of the 32 genesin the interval identified only one gene, rrt-1,whose knockdown produced high-level hypoxiaresistance (Fig. 2B). Simultaneously, five fosmidsthat together spanned the entire interval wereindividually injected to attempt transformationrescue of gc47. Only one fosmid restored normalhypoxia sensitivity to gc47 (Fig. 2C); the res-cuing fosmid contained the rrt-1 gene implicatedby RNAi. Sequencing rrt-1 in gc47 found asingle G→ A transition, resulting in a change ofamino acid residue 271 from an aspartate to anasparagine in gc47 (D271N, Fig. 2A). Thus, gc47is an allele of rrt-1 and behaves like a reduction-of-function allele.

rrt-1 encodes an arginyl-tRNA synthetase,one of the aminoacyl-tRNA synthetases (AARSs).AARSs catalyze the adenosine triphosphate(ATP)–dependent acylation of their cognatetRNA(s) with a specific amino acid (3). AARSs

1Department of Anesthesiology, Washington UniversitySchool of Medicine, St. Louis, MO 63110, USA. 2Depart-ment of Developmental Biology, Washington UniversitySchool of Medicine, St. Louis, MO 63110, USA.


Fig. 1. gc47 is a potentregulator of hypoxic celldeath in C. elegans. (Aand B) Time-lapse im-ages (magnification 4×)of gc47 (A) and N2 (B)adult worms after a 24-hour recovery from a20-hour hypoxic insult.(C) Percent dead animalsfor N2 (squares) and gc47(circles) after a 24-hourrecovery as a function oflength of hypoxic insult.(D) Percent death amonghomozygous gc47 (openbar, 13 trials), heterozy-gous gc47/+ (hatched bar,9 trials), andN2 (solid bar,16 trials) animals; theresults show that gc47 isrecessive. Data aremeansTSEMwith >30 animals pertrial; *P<0.01 (two-tailedt test).


REPORTS

...

......


fall into two distinct structural classes. RRT-1 is aclass I enzyme, characterized by HIGH andMRSKdomains (4, 5). In higher eukaryotes, the RRT-1ortholog has been isolated in a cytoplasmic com-plex with six other AARSs and three accessorysubunits (6, 7). Some AARSs are specific formitochondrial DNA, and besides their role intranslation, a subset has been implicated in non–translation-related functions (8, 9). To test whethercontrol of hypoxic sensitivity is a general functionof AARSs or is restricted to a particular subset,we used feeding RNAi constructs against 23 ofthe 33 predicted AARSs in the C. elegans ge-nome. All but one of the 23 RNAi constructsconferred significant hypoxia resistance (table S2).Thus, most, if not all, AARSs control hypoxicsensitivity. Knockdown of someAARSs producedstronger hypoxia-resistant phenotypes than others,which was not explained by the class of AARSsor whether the tRNA substrate was cytoplasmicor mitochondrial. Although the degree of RNAiknockdown was variable (table S2), RNAi ef-ficacy did not explain all of the phenotypic var-iance. To examine whether the variable levels ofhypoxia resistance could be explained by differ-ences in translational suppression, we measured[35S]methionine incorporation in animals treatedwith select AARSRNAi constructs aswell as in rrt-1(gc47) (fig. S1). The level of hypoxia resistancehad a strong inverse correlation with the relativetranslation rate. However, the absolute amount oftranslational suppression was relatively modesteven in the strongest hypoxia-resistant animals,where the [35S]methionine incorporation was abouthalf that of vector controls. Thus, hypoxic sensi-tivity appears to be exquisitely sensitive to evensmall changes in translation rate. Consistent withtranslational suppression as the proximate mech-anism of hypoxia resistance, treatment withcycloheximide also conferred hypoxia resistance(fig. S2).

Reduced translation has been shown to length-en life span in C. elegans and other organisms(10–12) and a known, long-livedmutant was alsofound to be highly hypoxia-resistant (1). Thus,we hypothesized that rrt-1(gc47) would be long-lived. Indeed, rrt-1(gc47) had a small but sig-nificantly increased life span (fig. S3). Althoughthis further links hypoxia resistance and long lifespan, the difference in the strength of the twophenotypes suggests that their mechanisms down-stream of RRT-1 are distinct or that hypoxic sen-sitivity is much more responsive than life span toalterations in the translation machinery. Other phe-notypes seen in gc47 were a modest decrease infecundity, a small but significant level of em-bryonic lethality, a reduction in speed of move-ment to about half that of wild-type animals, anda slight developmental delay (table S3). Thus, asexpected for a gene with such an essential func-tion, rrt-1(gc47) has a pleiotropic phenotype butthe strongest observed phenotype was hypoxiaresistance.

RRT-1 presumably functions in all cells tomediate translation. However, because of their

Fig. 2. gc47 is an alleleof rrt-1. (A) Genetic map-ping of gc47. Three-factormapping with the in-dicated visible markers(black text) and single-nucleotide polymorphisms(gray text) placed gc47 ina 106-kb interval on chro-mosome III between CE3-141 and snpLA3. Fosmidsinjected for transformationrescue are shown. Aminoacid alignment is withrrt-1 orthologs; the gc47mutation is boxed. Aminoacid abbreviations: A, Ala;C, Cys; D, Asp; E, Glu; G,Gly; H, His; I, Ile; L, Leu;M, Met; N, Asn; R, Arg; S,Ser; T, Thr; V, Val. (B)RNAi of genes within the106,000–base pair map-ping interval. Death wasinduced by hypoxia inanimals treated with 29of 32 predicted geneswithin the mapping inter-val. rrt-1 RNAi conferredhighly significant hypoxiaresistance relative to emp-ty vector (*P < 0.01; two-tailed t test, n = 30animals per data point).(C) Transformation rescue of gc47. Hypoxia-induced animal death was scored for N2, gc47 [full genotype:rrt-1(gc47) dpy-17(e164)], and gc47 transformed with the transformation marker pPHGFP alone or inaddition to the rescuing fosmid WRM0615bG07. Data are means T SEM of at least two trials, >20 animalsper trial; *P < 0.01 (two-tailed t test).

Fig. 3. rrt-1 acutely controls hypoxic sensitivity, both during and after the insult. (A) Schematic of theexperimental protocol. Wild-type animals were exposed to rrt-1 or L4440 empty vector RNAi at thedevelopmental stages indicated, and were then treated with a 20-hour (B) or 16-hour (C) hypoxicincubation (HYP, yellow boxes); Emb, embryo. (B) Hypoxia resistance by rrt-1(RNAi) is not dependent ondevelopmental stage [n = 40 animals per condition; *P < 0.05 versus L4440 (Fisher’s exact test, two-sided)]. (C) Inhibition of rrt-1 is effective after hypoxic insult. Percent survived = [(number of animals aliveat day of interest)/(number of animals alive initially after 24-hour recovery)] × 100; *P< 0.05 versus L4440(Fisher’s exact test, two-sided); n > 300 initially alive worms per RNAi over three independent trials.


REPORTS

...

......


high metabolic activity, germ cells might be par-ticularly vulnerable to hypoxic injury and therebydetermine the hypoxic sensitivity of the wholeorganism. To examine this issue, we made use ofa mutation in the rrf-1 gene, which encodes anRNA-directed RNA polymerase required for thesomatic but not germline actions of RNAi (13).The hypoxia resistance of rrt-1(RNAi)was great-ly reduced in an rrf-1 mutant versus wild-typebackground (fig. S4), indicating that RRT-1 actsin both somatic and germline cells to mediatehypoxic sensitivity. Myocytes and neurons aretwo major somatic cell types where hypoxiaproduces characteristic pathological changes(1, 14). Hypoxia-induced myocyte nuclear frag-mentation and neuron axonal degeneration wereboth abated in rrt-1(gc47) (fig. S5).

To determine when RRT-1 functions to regu-late hypoxic sensitivity, we applied RNAi towild-type animals early in development, early inadulthood before the hypoxic insult, or in adult-hood after the hypoxic insult (Fig. 3A). RNAieither during early development or during earlyadulthood protected equally well from subsequenthypoxic death (Fig. 3B). Thus, the hypoxia-resistant phenotype of rrt-1 reduction of functionis not dependent on developmental stage and canbe induced after development is complete.Exposure to rrt-1(RNAi) only after the hypoxicinsult also increased survival from delayed hy-poxic death (Fig. 3C).

The reduction-of-function mutant rrt-1(rf) re-duces global translation rate and consequentlyshould reduce oxygen consumption. Indeed, the

rate of paralysis by hypoxia, which should cor-relate with oxygen consumption, is decreased inrrt-1(gc47) (fig. S6). Reduced oxygen consump-tion by translational arrest is a logical and estab-lished mechanism for reducing cellular injuryduring hypoxia but not after (15–17). Thus, themechanism of protection by rrt-1 knockdown, atleast that functioning after the hypoxic insult,appears to be more complex than a global reduc-tion in oxygen consumption by translational arrest.

Hypoxia produces intracellular misfolded pro-teins and thereby induces the unfolded proteinresponse (UPR) (18). One effect of UPR induc-tion is phosphorylation of the translation initia-tion factor eIF2-a, thereby suppressing translation(19); this has been proposed as an adaptive mech-anism to reduce the load of newly synthesized andunfolded proteins, particularly in the context ofcancer cell biology. To determine whether thehypoxia-resistant phenotype of rrt-1(rf) may bedue to a reduction in unfolded proteins, we usedstrains carrying a transgene, Phsp-4::GFP, con-sisting of a fusion between the hsp-4 promoterand GFP (green fluorescent protein). Phsp-4::GFPexpression has been shown to be a reliable in-dicator of the level of unfolded proteins and ofactivation of the UPR (20–22). Hypoxia induced asignificant increase in expression of Phsp-4::GFPthat peaked 4 hours after recovery from hypoxiaand was dependent on the length of hypoxic in-cubation (Fig. 4, A to C). The glycosylation in-hibitor tunicamycin, which increases the level ofunfolded proteins, also induced Phsp-4::GFP ex-pression. Induction of Phsp-4::GFP expression byeither hypoxia or tunicamycin was blocked vir-tually completely by a loss-of-function mutationin ire-1 [ire-1(lf)], which encodes an endoplasmicreticulum transluminal kinase essential for the UPR(22, 23) (Fig. 4, A and B). Consistent with a re-duction in unfolded protein load, rrt-1(RNAi)completely blocked hypoxic induction of Phsp-4::GFP (Fig. 4, A and B); however, it did notdiminish induction by tunicamycin. Thus, thelevel of translational suppression by rrt-1(RNAi)does not preclude synthesis of the GFP markerunder strong inducing conditions. Further sup-porting the hypothesis that rrt-1(rf) reduces theload of unfolded proteins, rrt-1(gc47)was highlyresistant to tunicamycin-induced developmentalarrest (Fig. 4D). Finally, we found that ire-1(lf)and its downstream target xbp-1(lf) (22) signifi-cantly suppressed the hypoxia resistance producedby rrt-1(RNAi), but neither ire-1(lf) nor xbp-1(lf)induced hypersensitivity in rrt-1(+) animals (Fig.4E). Rather, the ire-1(lf) and xbp-1(lf) animalswere weakly resistant. These data indicate thatinhibition of translation by rrt-1(lf) and the UPRinteract synergistically to reduce hypoxic sensi-tivity, but that in the absence of translational sup-pression by rrt-1(lf), an intact UPR can promotedeath after a severe hypoxic insult.

Translational repression is a well-establishedmechanism of survival for hibernating animals ina prolonged hypoxic environment (24). Transla-tional mechanisms are important in the tumori-

Fig. 4. The unfolded protein response is induced by hypoxia and is required for high-level hypoxiaresistance of rrt-1(RNAi). (A) Phsp-4::GFP expression in age-matched young adult C. elegans afterincubation for 6 hours in M9 buffer in a normoxic or hypoxic environment or with tunicamycin (25 mg/ml).zcIs4[Phsp-4::GFP] animals were raised on empty vector or rrt-1(RNAi) bacteria; ire-1(zc14);zcIs4 animalswere raised on empty vector. Scale bar, 200 mm. (B) Quantification of Phsp-4::GFP expression. *P < 0.01,unpaired two-sided t test; au, arbitrary units. (C) Time course of Phsp-4::GFP induction after hypoxic (H) ornormoxic (N) incubations of 2, 4, or 8 hours. (D) Sensitivity to developmental arrest by tunicamycin in N2and rrt-1(gc47) animals. Freshly laid eggs were allowed to develop on agar plates containing theindicated concentrations of tunicamycin; percent of animals reaching adulthood after 7 days ofdevelopment was scored. (E) Hypoxic sensitivity of wild-type or UPR pathway mutant animals exposed torrt-1(RNAi) (30-hour hypoxic incubation) or empty vector control (20-hour hypoxic incubation). *P < 0.05,paired t test.


REPORTS

...

......


genicity of cancer cells (25) and are increasinglyimplicated in the sensitivity of normal cell typesto hypoxic and ischemic injury (15, 26). In thesediverse scenarios, translational repression resultsin several secondary changes in the biology ofthe cell, including decreases in ATP consumptionand protein aggregates and an alteration of theproteome. Our data show that a modest suppres-sion of translation that allows relatively normalgrowth and physiology can produce a profoundhypoxia resistance that requires the UPR for itsfull phenotypic expression. One effect of UPRactivation is translational suppression, which oc-curs by a mechanism distinct from limiting amino-acylated tRNA levels. A logical model is that areduction in translation rate by a decrement inAARS activity reduces the unfolded protein loadto a level that is manageable by the UPR andmay synergize with the translational suppressionproduced by the UPR itself. However, withouttranslational inhibition, the activity of the UPRmay be maladaptive in the context of hypoxic in-jury. This interaction between translational activityand the UPR may be exploited to regulatehypoxic cell death.

References and Notes1. B. A. Scott, M. S. Avidan, C. M. Crowder, Science 296,

2388 (2002); published online 13 June 2002(10.1126/science.1072302).

2. See supporting material on Science Online.3. M. Ibba, D. Soll, Annu. Rev. Biochem. 69, 617 (2000).4. C. Landes et al., Biochimie 77, 194 (1995).5. B. Delagoutte, D. Moras, J. Cavarelli, EMBO J. 19, 5599

(2000).6. S. V. Kyriacou, M. P. Deutscher, Mol. Cell 29, 419

(2008).7. J. M. Han, J. Y. Kim, S. Kim, Biochem. Biophys. Res.

Commun. 303, 985 (2003).8. P. Sampath et al., Cell 119, 195 (2004).9. S. G. Park, K. L. Ewalt, S. Kim, Trends Biochem. Sci. 30,

569 (2005).10. P. Syntichaki, K. Troulinaki, N. Tavernarakis, Nature 445,

922 (2007).11. K. Z. Pan et al., Aging Cell 6, 111 (2007).12. M. Hansen et al., Aging Cell 6, 95 (2007).13. T. Sijen et al., Cell 107, 465 (2001).14. N. Dasgupta, A. M. Patel, B. A. Scott, C. M. Crowder,

Curr. Biol. 17, 1954 (2007).15. W. Paschen, C. G. Proud, G. Mies, Curr. Pharm. Des. 13,

1887 (2007).16. B. G. Wouters et al., Semin. Cell Dev. Biol. 16, 487

(2005).17. D. J. DeGracia, R. Kumar, C. R. Owen, G. S. Krause,

B. C. White, J. Cereb. Blood Flow Metab. 22, 127(2002).

18. C. Koumenis et al., Methods Enzymol. 435, 275 (2007).

19. C. Koumenis et al., Mol. Cell. Biol. 22, 7405 (2002).20. T. Yoneda et al., J. Cell Sci. 117, 4055 (2004).21. F. Urano et al., J. Cell Biol. 158, 639 (2002).22. M. Calfon et al., Nature 415, 92 (2002).23. X. Shen et al., Cell 107, 893 (2001).24. K. B. Storey, J. M. Storey, Biol. Rev. Camb. Philos. Soc.

79, 207 (2004).25. T. van den Beucken, M. Koritzinsky, B. G. Wouters,

Cancer Biol. Ther. 5, 749 (2006).26. D. J. DeGracia, B. R. Hu, J. Cereb. Blood Flow Metab. 27,

875 (2007).27. We thank M. Mabon for early efforts on screen design,

L. Metz for technical help, and M. Hansen for advice andstrains. Supported by the National Institute ofNeurological Disorders and Stroke, a Neuroscience ofBrain Disorders Award from the McKnight EndowmentFund for Neuroscience, and an American HeartAssociation Established Investigator Award. Some of thestrains used in this work were provided by theCaenorhabditis Genetics Center, which is funded by theNIH National Center for Research Resources.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/323/5914/630/DC1Materials and MethodsFigs. S1 to S6Tables S1 to S4References


Ligand-Dependent EquilibriumFluctuations of SingleCalmodulin MoleculesJan Philipp Junker,1 Fabian Ziegler,1 Matthias Rief1,2*

Single-molecule force spectroscopy allows superb mechanical control of protein conformation.We used a custom-built low-drift atomic force microscope to observe mechanically inducedconformational equilibrium fluctuations of single molecules of the eukaryotic calcium-dependentsignal transducer calmodulin (CaM). From this data, the ligand dependence of the full energylandscape can be reconstructed. We find that calcium ions affect the folding kinetics of theindividual CaM domains, whereas target peptides stabilize the already folded structure.Single-molecule data of full length CaM reveal that a wasp venom peptide binds noncooperativelyto CaM with 2:1 stoichiometry, whereas a target enzyme peptide binds cooperatively with1:1 stoichiometry. If mechanical load is applied directly to the target peptide, real-timebinding/unbinding transitions can be observed.

Single-molecule mechanical methods havemade it possible to study and control biomo-lecular conformations with unprecedented

precision (1–6). Whereas classical methods suchas changes in temperature or chemical environ-ment act globally and rather unspecifically on theenergy landscape, forces can be applied locally toprecisely manipulate selected structural elementsof a protein and thus explore protein energylandscapes in a controlled manner (5). Due tolimited resolution, single-molecule mechanical

studies of protein folding and protein/ligandinteractions have so far almost exclusively beencarried out in nonequilibrium and have thus beenrestricted to the study of unfolding and unbindingreactions (7–9). Hence, half of the energy land-scape (the part that controls re-folding or ligandre-binding) has been mostly inaccessible to theseexperiments. We have used a low-drift atomicforce microscope (AFM) (10) to observe folding/unfolding fluctuations of single calmodulin pro-teins (CaM) under equilibrium conditions in thepresence of Ca2+ and target peptides. Ligand-dependent folding/unfolding fluctuations, as wellas ligand binding/unbinding fluctuations, can beinduced mechanically and observed in real time.

CaM is the most prominent Ca2+ sensor ineukaryotic cells (11). When Ca2+ binds to CaM,

the flexible Ca2+-free apo conformation convertsinto a rigid and stable holo conformation (12).Holo CaM binds to specific target sequences indownstream regulatory proteins, thus alteringtheir signaling properties. To date, more thanone hundred target proteins for CaM have beendescribed (11). Little is known about the foldingof CaM, but temperature-jump experiments haveshown that folding of apo CaM occurs on thesubmillisecond time scale (13).

The experimental scheme, including a sketchof Ca2+-CaM, is shown in Fig. 1A. Followingpreviously described methods, we have sand-wiched a single CaMmolecule between immuno-globulin domains from Dictyostelium discoideumfilamin that serve as attachment points for theAFM tip and the surface. CaM consists of twostructurally similar domains, a ~75-residue N-terminal (DomN, red) and a ~70-residue C-terminal (DomC, blue) globular domain, thateach bind two Ca2+ ions. A typical force exten-sion trace of CaM at a pulling velocity of 1 nm/sand 10 mM Ca2+ is shown in Fig. 1B. At pullingvelocities generally used in force spectroscopy of~1 mm/s, unfolding events of CaM are buried inthe thermal noise (14). Low instrumental driftallows us to achieve the slow pulling velocity(1 nm/s) necessary for the required force reso-lution of ~2 pN (15). Two distinct peaks (markedin red and blue, respectively) correspond to theunfolding of the two globular CaM domains. Toallow for a structural interpretation of the un-folding traces, we introduced a disulfide bondbetween residues 128 and 144 of DomC toshorten the extensible backbone of this domain(CaM 128×144). Ligand binding properties ofthe shortened version were indistinguishablefrom the wild type. The length gain upon un-

1Physik Department E22, Technische Universität München,James-Franck-Strasse, 85748 München, Germany. 2MunichCenter for Integrated Protein Science, 81377 München,Germany.



REPORTS

...

......


folding of the domains can be measured usingthe wormlike chain model of polymer elasticity(16) (black solid lines in Fig. 1B, main graph).The average values are DLI = 25.7 T 0.4 nm andDLII = 17.6 T 0.7 nm. This result suggests that thered unfolding peak marks unfolding of DomN(expected value ∆L = 25.0 nm), whereas in thesecond peakDomC unfolds (expected value∆L=18.7 nm) (17). Closer inspection of the unfoldingpeaks reveals rapid transitions between the foldedand unfolded states in both unfolding peaks. Theinsets above the peaks in Fig. 1B show timetraces of the transition regions. Such transitionsare the hallmark of thermodynamic equilibrium,and a kinetic analysis now allows extracting bothequilibrium and nonequilibrium parameters and,hence, the energy landscape of CaM folding froma single molecule.

In agreement with experiments using single-domain CaM constructs, our results suggest thatthe two domains fold and unfold independentlyfrom each other (18). Therefore, we investigatedsingle-domain constructs of both DomN andDomC. A sample time trace of the unfoldingfluctuations of isolated DomC is shown in Fig.1C. Due to the large spring constant of the AFMprobe (6 pN/nm), the fluctuations do not occur atconstant load. Unfolding occurs at high loads (12pN), whereas refolding occurs from the relaxed

state (6 pN). The slope of the broken linesmarking the folded and unfolded levels, respec-tively, reflects the slow force ramp imposedduring the experiment. To analyze the fluctuationkinetics, we define the characteristic transitionrate as the averaged folding and unfolding rate inthe whole transition region (DomC: 11.7/s, num-ber of molecules analyzed n = 23). Using aMonte-Carlo simulation, we adjusted the unfold-ing and folding rates at zero force so that thecharacteristic transition rate matched our exper-iment [for details, see the supporting onlinemate-rial (SOM) text]. Independent measurements ofthe speed dependence of the nonequilibriumunfolding forces at higher pulling velocitiesyielded a distance of the transition state fromthe folded state of ∆xN-TS = 2 nm, which is atypical value for an all–a helical protein (SOMtext and fig. S1A). Combining these results, wecan now draw an energy landscape for the fold-ing of CaM at 10 mM Ca2+ (Fig. 1D). A zero-force folding rate of ~2 × 105/s under theseconditions makes CaM one of the fastest foldingproteins known to date (19).

To investigate the influence of ligand bindingon the folding dynamics of DomC in greaterdetail, we performed concentration scans of Ca2+

and mastoparan (Mas), a wasp venom peptideknown to bind to CaM (20). Figure 2A shows

equilibrium time traces of the transition region atvarying Mas concentrations with Ca2+ kept fixedat 10 mM. The most obvious influence of Masbinding is the decreased speed of the transitionkinetics with increasing Mas concentration. Weplotted the characteristic transition rate as a func-tion of Mas concentration in Fig. 2B. The effectof Mas on the midpoint unfolding force (theaverage value of the unfolding forces in thetransition region) is shown in the inset. As ex-pected, higher concentrations of Mas increase thestability of DomC and, hence, its midpoint un-folding force. The effect of Mas on both forceand kinetics can be readily explained by assum-ing that Mas binding exclusively slows the un-folding rate ku, as depicted in Fig. 2C for twosample concentrations. In the narrow force rangewhere we observe folding/unfolding transitions,both unfolding rates (green) and folding rates(purple) will depend in good approximation expo-nentially on the applied force (21–23). The slopesof the green and purple lines are determined bythe distance of the transition state from the nativeand unfolded states, respectively (21, 22). At1 mM Mas, the midpoint force of the transitionregion, where unfolding and folding occur at thesame rates, is defined by the intersection of thethin green line and the purple line. Because Masbinding acts on the unfolding rate, increasing the

Fig. 1. (A) Sketch ofthe experimental setup,showing CaM attachedto AFM cantilever tipand surface by meansof filamin domains thatserve as handles (not toscale). (B) Sample traceof CaM 128×144 at10mMCa2+. The folding/unfolding transitions ofDomN are shown inred, and those of DomCare shown inblue.Worm-like chain fit curves (blacktraces) allow determiningthe contour length in-crease upon unfolding∆L and thus identifyingthe individual domains.In the insets above thepeaks, time traces of thetransition region are de-picted. (C) Time tracesof the transition regionof isolated DomC at 10mM Ca2+. (D) Potentialenergy landscape forthe folding of DomC at10 mM Ca2+. N, nativestate; TS, transitionstate; U, unfolded state.

For

ce (

pN)

A

DExtension (nm)

Reaction coordinate x

free

ene

rgy

ku=0.02 s-1

kf=2·105 s-1

DomN

DomC

N TS U

8 nm2 nm

Time

For

ce (

pN)

LI=25.7 nm LII=17.6 nm

C

B


REPORTS

...

......


20 p

N

20 p

N

1 s 1 s

D

Time

For

ce

10 µM Mas

1 m

M C

a2+10

mM

Ca2+

100

mM

Ca2+

incr

easi

ng C

a2+ c

once

ntra

tion

A

Time

For

ce

10 mM Ca2+

100

µM M

as10

µM

Mas

0.3

µM M

as

E

Ca2+ concentration (M)

Ca2+ concentration (M)

Cha

ract

eris

tic

tran

sitio

n ra

te (

s-1)

For

ce (

pN)B

Mastoparan concentration (M)

Cha

ract

eris

tic

tran

sitio

n ra

te (

s-1)

Mas concentration (M)

For

ce (

pN)

F

Force

log

rate

s

1 mM Ca 2+ 100 mM Ca 2+

kf

ku

C

Force

log

rate

s

1 µM Mas

100 µM Mas

kf

ku

incr

easi

ng M

as c

once

ntra

tion

N TS U

free

ene

rgy

free

ene

rgy

N U

reaction coordinate reaction coordinate

0

0

TS

Fig. 2. (A) Time traces of isolated DomC at 10 mM Ca2+ and varying Masconcentrations. (B) Characteristic transition rate and unfolding force at differentMas concentrations (blue symbols). The ligand effect can be fully reproduced byMonte Carlo simulations (black symbols and black interpolation curve). (C) (Left)Force dependence of folding rate (purple) and unfolding rate (green). An increasein Mas from 1 mM (thin green line) to 100 mM (thick green line) moves theintersection point (transition region) to higher forces and lower rates. (Right)

Potential energy landscape at low (thin green line) and high (thick green line) Mas.(D) Time traces of isolated DomC at 10 mM Mas and varying Ca2+ concentrations.(E) Transition rate and unfolding force at different Ca2+ concentrations (bluesymbols). Results of a Monte Carlo simulation are shown in black. (F) (Left) IncreaseinCa2+ from1mMCa2+ (thin purple line) to 100mMCa2+ (thick purple line)movesthe intersection point to higher forces and higher rates. (Right) Potential energylandscape for folding at low (thin purple line) and high (thick purple line) Ca2+.


REPORTS

...

......


concentration to 100 mMwill shift the green linedown (thick green line). The new intersectionpoint now lies at an increased force as well as areduced characteristic transition rate. A detailedmodeling of ku and kf in dependence of ligandconcentration and force in a Monte Carlo simu-lation could reproduce the full effect of Masconcentration on midpoint unfolding forces andtransition rates (solid lines in Fig. 2B; for details,see SOM text). This simulation includes equilib-rium constants for Mas binding to the native, un-folded, and transition state, respectively.

Similarly, we investigated the effect of Ca2+

on midpoint unfolding forces and kinetics (Fig.2D). From the data traces taken at three sampleconcentrations, it can be readily seen that unlikeMas, Ca2+ increases the characteristic transitionrate of DomC. Forces and transition rates arequantified in Fig. 2E. Despite the increase intransition rates, the midpoint unfolding force (seeinset) also increases (24). Increased transitionrates, together with increased stability, can onlybe explained by Ca2+ raising predominantly the

folding rate (see the scheme in Fig. 2F). In-creasing the Ca2+ concentration from 1 mM to100 mM will shift up the folding rate (purplelines) but leave the unfolding rate unaffected. Thenew intersection point lies at both higher forcesand higher rates. Again, a Monte Carlo simu-lation incorporating the Ca2+ dependence of kuand kf can fully explain the data (solid lines inFig. 2E; for details, see SOM text).

Thus, the two CaM ligands, Ca2+ and Mas,act on folding in different ways (see energy land-scapes in Fig. 2, C and F). Mas stabilizes thealready folded form but does not interact with thetransition state structure or the unfolded proteinbecause it leaves the folding rate unaffected.Hence, the transition state has only negligible af-finity to Mas (Fig. 2C) (8). In contrast, at theconcentrations measured (100 mM to 100 mM),Ca2+ predominantly acts on the folding rate.Hence, Ca2+ binding stabilizes both the transitionstate and the folded state, indicating a dissocia-tion constant of Ca2+ binding to the transitionstate of ~1 mM (see energy landscape in Fig. 2F

and SOM text). At very lowCa2+, CaM folds intothe flexible and highly dynamic apo structure thatundergoes a large conformational change uponaddition of Ca2+ (12). Higher Ca2+ opens a newfolding pathway directly to the holo form via atransition state that has already bound Ca2+ ions.At high Ca2+, this pathway prevails over the apofolding route (see scheme in fig. S3 and SOMtext). Along this holo folding route, the Ca2+

binding sites might act as nucleation seeds forfolding because Ca2+ binding appears to occurearly on the folding pathway.

Our mechanical folding/unfolding assay nowoffers the possibility to study CaM-target peptideinteraction modes of full-length CaM directly onthe single-molecule level. Figure 3A (upper trace)shows a mechanical unfolding curve of CaM128×144 in the presence of 100 mM Mas. Incomparison to peptide-free CaM, a mechanicalstabilization of both DomN and DomC becomesobvious, whichmanifests itself in higher forces aswell as slower transition rates (Table 1). In thefull-length construct, the stabilizations of the in-

Fig. 3. (A) (Upper trace)Sample trace of CaM 128 ×144 at 10 mM Ca2+ and100 mMMas. DomN (red) andDomC (blue) are stabilizedapproximately equally byMas. (Lower traces) IsolatedDomN (red) and DomC (blue)at 100 mMMas. (B) Unfoldingforces and transition rates ofDomN of CaM 128×144 atdifferent Mas concentrations(red). Solid lines, concentra-tion dependence in the caseof a 1:2 stoichiometry; bro-ken lines, 1:1 stoichiometry(Monte Carlo simulations). (C)(Upper trace) Sample traceof CaM 128×144 at 10 mMCa2+ and 10 mM MLCK.DomN (red) is stabilized con-siderably more strongly thanDomC (blue). (Lower traces)Isolated DomN (red) andDomC (blue) at 10 mMMLCK.(D) Unfolding forces andtransition rates of DomN ofCaM 128×144 at differentMLCK concentrations (red).Solid lines, concentration de-pendence for a 1:2 stoichi-ometry; broken lines, 1:1stoichiometry. (E) Sampletrace of CaM with MLCK fusedto the C terminus at 10 mMCa2+. The protein construct isshown in the scheme. (Inset)Time traces of first unfoldingpeak. An intermediate state,corresponding to CaM withunbound peptide (black lev-el), appears.

For

ce (

pN)

10 µM MLCK

For

ce (

pN)

A C

E

For

ce (

pN)

30

20

10

0604020

1 s

Extension (nm)

20

10

0

Extension

100 µM Mas

For

ce (

pN)

BExtension

D

concentration (M) concentration (M) concentration (M) concentration (M)

For

ce (

pN)

For

ce (

pN)

Tran

sitio

n ra

te (

s-1)

Tran

sitio

nra

te (

s-1)


REPORTS

...

......


dividual domains obviously occur independentlyfrom each other because they behave identical tosingle-domain constructs at the same Mas con-centration (Fig. 3A, lower traces, and Table 1).This result is surprising, given that Mas bindingto CaM is discussed in the literature almost ex-clusively in the context of a binding stoichi-ometry of 1:1 (25). Cooperative binding of bothdomains simultaneously to a single Mas peptideshould occur with a higher energy than the sumofthe interaction energies of the individual do-mains. Our result indicates that full-length CaMbinds Mas noncooperatively in a 1:2 stoichiom-etry where each peptide interacts with only oneCaM domain. The 1:2 stoichiometry can be con-firmed by analyzing the single-molecule thermo-dynamics of ligand binding to CaM (SOM text).The Mas dependence of forces and transitionrates of the first unfolding peak (DomN) of CaM128×144 is shown in Fig. 3B. A 1:1 stoichiom-etry is not consistent with our data (dashed blackline obtained from the Monte Carlo simulation),whereas a 1:2 stoichiometry fits our data well(solid black line) (26).

Because binding of the venom peptideMas toCaM probably has a very different physiologicalrole than binding to target enzymes, we also in-vestigated binding of CaM to the interaction sitepeptide of its target enzyme skeletal musclemyosin light chain kinase (MLCK) (27). As ex-pected for a strongly and cooperatively bindingpeptide (28), the stabilization of full-length CaMupon MLCK peptide binding is dramatic at con-centrations of 10 mM (Fig. 3C, upper trace, andTable 1). The unfolding peak of DomN is in-creased to ~19 pN, whereas DomC is stabilizedto an extent similar to Mas. For comparison, theunfolding traces of the isolated domains areshown in the lower traces of Fig. 3C. In contrastto Mas, isolated DomN seems to exhibit almostno affinity for the MLCK peptide. In agreementwith the literature, we find a binding stoichiom-etry of 1:1 for this peptide (Fig. 3D).

The large interaction energy of CaM andMLCK peptide, together with a diffusion-limitedon-rate (29), suggests that the exchange of thebound peptide from the complex may be slow.Mechanical force can be used as control param-eter to affect the peptide off-rates directly andbring them into an observable regime while stillleaving the structure of CaM intact. To this end,we fused the MLCK peptide directly to the Cterminus of CaM (30), enabling us to apply forceto the peptide directly (see sketch in Fig. 3E).

Again, the observed traces exhibit near equilib-rium fluctuations. In the first unfolding peak,rapid transitions between three levels can be ob-served: peptide bound (green), peptide unbound(black), and DomN unfolded (red) (Fig. 3E). Forclarity, the inset shows time traces of thosefluctuations taken from three independent mole-cules. This data directly reveals the thermody-namic hierarchy of the folding/peptide-bindingprocess. The black level (CaM folded and pep-tide unbound) seems to be obligatory both on thefolding/binding pathway and on the unfolding/unbinding pathway. Even though the stabiliza-tion due to peptide binding (~20 kBTas estimatedfrom the area enclosed by the green and blacklevels) is higher than the folding free energyof the individual domains (~15 kBT), binding/unbinding and folding/unfolding are still separateprocesses.

At forces of ~20 pN, the peptide-bound statestill has a long lifetime on a time scale of seconds(green levels). Using a transition state positionof 1 nm for peptide unbinding, we estimated azero-force lifetime for the bound peptide of 102

to 103 s. Such a long lifetime may be essential forproper functioning of the CaM signaling path-way, specifically because typical free cellularCa2+-CaM concentrations are lower than theoverall concentration of target sites (31).

We demonstrated that single-molecule forcespectroscopy by AFM allows online observationof the interaction dynamics of a signaling proteinwith its target. How CaM binding kinetics andmechanics are affected by full-length target pro-teins and how their activity is modulated will bean important question for future experiments. Weanticipate that direct single-molecule measure-ments of equilibrium fluctuations, as presented inthis study, will provide an important tool to mea-sure protein-target interaction dynamics in realtime.

References and Notes1. W. J. Greenleaf, K. L. Frieda, D. A. N. Foster, M. T. Woodside,

S. M. Block, Science 319, 630 (2008); published online2 January 2008 (10.1126/science.1151298).

2. C. Cecconi, E. A. Shank, C. Bustamante, S. Marqusee,Science 309, 2057 (2005).

3. A. P. Wiita et al., Nature 450, 124 (2007).4. S. P. Ng et al., Proc. Natl. Acad. Sci. U.S.A. 104, 9633

(2007).5. H. Dietz, F. Berkemeier, M. Bertz, M. Rief, Proc. Natl.

Acad. Sci. U.S.A. 103, 12724 (2006).6. E. M. Puchner et al., Proc. Natl. Acad. Sci. U.S.A. 105,

13385 (2008).7. E.-L. Florin, V. T. Moy, H. E. Gaub, Science 264, 415

(1994).

8. Y. Cao, T. Yoo, H. Li, Proc. Natl. Acad. Sci. U.S.A. 105,11152 (2008).

9. Y. Cao, M. M. Balamurali, D. Sharma, H. Li, Proc. Natl.Acad. Sci. U.S.A. 104, 15677 (2007).

10. Materials and methods are available as supportingmaterial on Science Online.

11. A. P. Yamniuk, H. J. Vogel,Mol. Biotechnol. 27, 33 (2004).12. M. Zhang, T. Tanaka, M. Ikura, Nat. Struct. Biol. 2, 758

(1995).13. C. R. Rabl, S. R. Martin, E. Neumann, P. M. Bayley,

Biophys. Chem. 101–102, 553 (2002).14. M. Carrion-Vazquez et al., Prog. Biophys. Mol. Biol. 74,

63 (2000).15. The low instrumental drift now allows a force resolution

of AFM that comes close to optical traps. In ourexperience, AFM offers a better length resolution,whereas optical tweezers, due to the much softer springconstants, have advantages if constant force conditionsare required.

16. C. Bustamante, J. F. Marko, E. D. Siggia, S. Smith,Science 265, 1599 (1994).

17. The expected contour length increase upon unfolding of aprotein domain consisting of N amino acid residues canbe calculated as ∆L = 0.365N – dfolded(I) + dfolded(II),where dfolded(I) and dfolded(II) are the end-to-enddistances of the folded structure before and after theunfolding event, respectively (SOM text). Due to thea-helical linker between DomN and DomC, the domainboundaries of CaM are not precisely defined, which mayaccount for the small deviations from the expected values.

18. L. Masino, S. R. Martin, P. M. Bayley, Protein Sci. 9, 1519(2000).

19. V. Munoz, Annu. Rev. Biophys. Biomol. Struct. 36, 395(2007).

20. D. A. Malencik, S. R. Anderson, Biochem. Biophys. Res.Commun. 114, 50 (1983).

21. G. I. Bell, Science 200, 618 (1978).22. M. Schlierf, F. Berkemeier, M. Rief, Biophys. J. 93, 3989

(2007).23. R. B. Best, G. Hummer, J. Am. Chem. Soc. 130, 3706

(2008).24. The increase in force upon an increase in Ca2+ is, in fact,

stronger than with an increase in Mas (insets in Fig. 2, Band E). This difference directly reflects the binding of 2Ca2+ ions to CaM, whereas it only binds 1 molecule ofMas (SOM text).

25. S. Linse, T. Drakenberg, S. Forsen, FEBS Lett. 199, 28(1986).

26. A simple thermodynamic argument can readily explainwhy a 1:1 stoichiometry should lead to concentrationindependent forces and transition rates (dashed lines inFig. 3B). In our assay, the fully folded CaM is transformedto CaM with unfolded DomN in the first unfolding peak.Because at all Mas concentrations used (>0.3 mM), bothintact CaM and DomC alone are fully saturated with Mas,changes in concentration cannot affect the energetics ofthis transition any further (SOM text).

27. M. Ikura et al., Science 256, 632 (1992).28. P. M. Bayley, W. A. Findlay, S. R. Martin, Protein Sci. 5,

1215 (1996).29. S. E. Brown, S. R. Martin, P. M. Bayley, J. Biol. Chem.

272, 3389 (1997).30. A. Miyawaki et al., Nature 388, 882 (1997).31. M. N. Teruel, W. Chen, A. Persechini, T. Meyer, Curr. Biol.

10, 86 (2000).32. We thank P. M. Bayley for inspiring discussions and

A. Bausch, H. Gaub, M. Bertz, and M. Schlierf for helpfulcomments on the manuscript. This work was supported byDeutsche Forschungsgemeinschaft grant RI 990/3-1.J.P.J. was supported by the International Graduate School“Materials Science of Complex Interfaces.”



Table 1. Midpoint unfolding forces and characteristic transition rates (10 mM Ca2+).

Peptides Isolated DomN Isolated DomC DomN(CaM 128x144)

DomC(CaM 128x144)

Force Rate Force Rate Force Rate Force Rate

No peptide 12.6 pN 11.5 s−1 12.1 pN 11.7 s−1 11.8 pN 11.9 s−1 14.2 pN 19.5 s−1

100 mM Mas 14.0 pN 4.0 s−1 16.1 pN 2.1 s−1 14.2 pN 3.2 s−1 18.2 pN 2.8 s−1

10 mM MLCK 12.2 pN 12.2 s−1 14.9 pN 2.6 s−1 19.5 pN 4.9 s−1 15.5 pN 5.4 s−1


REPORTS

...

......


Stretching Single Talin RodMolecules Activates Vinculin BindingArmando del Rio,1 Raul Perez-Jimenez,1 Ruchuan Liu,2 Pere Roca-Cusachs,1Julio M. Fernandez,1 Michael P. Sheetz1*

The molecular mechanism by which a mechanical stimulus is translated into a chemical responsein biological systems is still unclear. We show that mechanical stretching of single cytoplasmicproteins can activate binding of other molecules. We used magnetic tweezers, total internal reflectionfluorescence, and atomic force microscopy to investigate the effect of force on the interactionbetween talin, a protein that links liganded membrane integrins to the cytoskeleton, and vinculin, afocal adhesion protein that is activated by talin binding, leading to reorganization of thecytoskeleton. Application of physiologically relevant forces caused stretching of single talin rods thatexposed cryptic binding sites for vinculin. Thus in the talin-vinculin system, molecularmechanotransduction can occur by protein binding after exposure of buried binding sites in the talin-vinculin system. Such protein stretching may be a more general mechanism for force transduction.

Force sensing by cells is critical for their prop-er function; however, how cells transforma force stimulus into a chemical response

remains unclear. Forces have been shown to openion channels, activate phosphorylation, and causecatch bond formation. These could all be mech-anisms for force transduction [reviewed in (1)].Another mechanism that has been postulated butnever demonstrated experimentally is that stretch-ing of single force-bearing cytoplasmic moleculesmight expose binding sites for other proteins. Wecharacterized at a single-molecule level the in-

teraction of two cytoskeletal proteins, talin, theadhesion force–bearing protein, and vinculin, itsbinding partner, in the absence and presence ofa physiologically relevant force.

Talin and vinculin are key players in cell sig-naling, adhesion, and migration (2) and are lo-calized at the sites of cell-matrix adhesion (3).They are ideal candidates for mechanotransduc-tion because the binding of vinculin to focal ad-hesion complexes was shown to increase uponthe application of force (4). Talin couples thecytoskeleton to the extracellular matrix through

membrane integrins and can activate signalingthrough vinculin binding, for assembly and reor-ganization of the actin cytoskeleton (3, 5). Theglobular head of talin contains a FERM (4.1/ezrin/radixin/moesin) domain that binds and ac-tivates b-integrin cytodomains (6), whereas thetalin rod (TR) contains up to 11 vinculin bindingsites (VBSs) (7). A portion of the TR spanningresidues 482 to 889 (Fig. 1A) contains a five-helix bundle with a cryptic VBS in helix four,followed by a seven-helix bundle containingfour additional VBSs (helices 6, 9, 11, and 12).The x-ray structures of VBSs reveal that theyare defined by six turns of an a helix and thatthey are often buried because of extensive hy-drophobic interactions with other amphipathichelices (8, 9). VBSs trigger conformationalchanges in the vinculin head (Vh), in a processof helical bundle conversion: a helices of Vh re-organize to surround the VBS from the TR, bury-ing 50% of its solvent accessible area (5, 10).Figure 1, B and C, shows the proposed sequentialunfolding of the helical bundles in the TR and Vhbinding to the helix 12, which is exposed insteered molecular dynamic simulations (11).

1Department of Biological Sciences, Columbia University, NewYork, NY 10027, USA. 2Department of Physics, National Uni-versity of Singapore, Singapore 117542, Singapore.


Fig. 1. (A) Structure of the 12 helices that form the TR 482 to 889[Protein Data Bank (PDB) 1xwx]. Color coding: blue, N terminus; green,middle; and red, C terminus. VBS helices are numbered and representedwith cartoon models, and the rest of the protein is shown with tube models.(B) Under the application of a force in the direction indicated by the

black arrows in (A), TR starts to unfold. Once helix 12 exposed its VBS, Vh(PDB 1u6h), represented in yellow, reorganizes to bind to it. (C) X-raystructure of the complex Vh-VBS-helix 12, PDB 1u6h. Images were gen-erated by using the VMD (Visual Molecular Dynamics) program (www.ks.uiuc.edu/research/vmd/).


REPORTS

...

......


Previous studies have shown that four of fiveVBSs in TR residues 482 to 889 are inactive forbinding in the native molecule (8, 12). Whatactivates the VBSs in TR is still unknown. Thefacts that mechanical force increases vinculinrecruitment to focal adhesions (4) and that talincan bind to the actin cytoskeleton suggest thatforce induced by actomyosin contraction couldstretch the TR, exposing the cryptic VBSs to Vh(2). The mechanical stretching of cytoskeletal-ly attached proteins by applied force is docu-mented for the case in which stretch activatestyrosine phosphorylation by Src family kinases(13). Moreover, it has been shown that stretch-ing of DNA leads to changes in the binding ofDNA-interacting proteins (14). Molecular dy-namics simulations have supported the abilityof force-induced conformational changes to ex-pose the cryptic residues of VBSs in the TR (15).In addition, another recent steered molecular dy-namic study provided a model in which mechan-ical force gradually breaks the 12 helices of theTR into three smaller sub-bundles (11).

To experimentally test the hypothesis thatforce-induced stretching of the TR can openthe helical bundles and enable Vh binding, we

used a combination of magnetic tweezers to en-able stretching of single TR molecules and totalinternal reflection fluorescence (TIRF) micros-copy to observe fluorescent Vh binding (Fig. 2).

Three DNA constructs encoding monomericTR, dimeric tandem TR, and full-length a-actininwere designed with His and Avi tags at their N-and C-termini, respectively. The dimeric tandemTR served as a positive control because it shouldhave twice the number of binding sites that are ina single TR. Because a-actinin has an activebinding site for Vh (10), it served as a negativecontrol; we would not expect to see any dif-ference in the number of Vh molecules bound toa-actininwith andwithout force-induced stretching.All expressed proteins had 6×His and a biotin attheir N- and C-termini, respectively (fig. S1). TheN-terminal was used to attach the protein to a Ni–nitrilotriacetic acid (NTA) glass surface, whereasavidinated magnetic beads were coupled to the bio-tinylated C-terminus. The molecule attached to theglass was stretched by applying a magnetic force tothe beads in a vertical direction with the magnetictweezers apparatus described previously (16).

The criterion for deciding that beads werebound to a single talin molecule was to observe

a random motion of the bead around a singlepoint close to the surface. The Vh was labeledwith the fluorophore Alexa 488 at a molar ratioof 1:1 and further incubated with the TR. Thisincubation was performed in presence or absenceof force depending on whether or not stretchingof TR was pursued. Then, all unbound Alexa488–Vh was washed, and TIRF intensity wasmeasured over time. The number of Alexa 488–Vh molecules bound to the TR was determinedby single-molecule fluorescence photobleachingevents. The fluorescence intensity abruptly de-creased in a standard step if a single fluorophorewas bleached. The same protocol was used tomeasure Vh binding to TR tandem fragment anda-actinin controls.

With monomeric TR in the absence of force,we observed (Fig. 3A) one or no photobleach-ing events of the Alexa 488–Vh fluorescencefor each bead attached to a TR. After the appli-cation of 12 pN of force to the beads, as manyas three photobleaching events were observed,each the size of a single bleaching event. Thehistogram next to each graph represents thedistribution of photobleaching events for allthe analyzed beads. Although fewer than halfof the diffusing beads showed binding and in-dividual bleaching events, the percentage ofvinculin bound to beads was greater upon ap-plication of force. Because each Alexa 488–Vhwas labeled with one fluorophore molecule,each photobleaching event was associated withthe binding of one Vh molecule to the TR. Wesuggest that force applied to the beads causedmechanical unfolding of TR, exposing up totwo additional VBSs. When 2 pN of force wasapplied to the beads, the percentage of vinculinbound to beads was larger than in the case ofno force but smaller than when 12 pN of forcewas applied, which supports the hypothesisthat the number of vinculin molecules boundto the TR depends on the stretching force ap-plied to the TR.

The analysis of the photobleaching eventsfor the experiments performed using the tandemTR as the positive control (Fig. 3B) revealed upto two Vh binding events to the tandem TR inthe absence of force. When 12 pN of force wasapplied to the beads, up to six binding eventswere observed. In all experiments performedusing a-actinin as the negative control (Fig. 3C),one or no binding event was observed with andwithout the application of 12 pN of force.

In order to measure the pattern of force-dependent stretching of TR, we designed a DNAconstruct encoding for I272-TR-I272 [TR flankedby two molecules of the I27 (27th immuno-globular domain of human cardiac titin), whichhas well-characterized mechanical properties(17) (Fig. 4A)]. The protein I272-TR-I272 wasmanipulated by using single-molecule forceextension spectroscopy (18). Figure 4B shows acharacteristic force extension curve correspondingto the unfolding of I272-TR-I272. In this trace, thefour equally spaced peaks at ~200 pN with

Fig. 2. Representation of the device used to measure the binding events. The Ni-NTA (labeled N)grafted slides containing the TR fixed through its 6His N terminus (labeled H) to the glass andwith the avidinated magnetic bead bound to its biotinylated C terminus (labeled B) was placed overthe objective. Alexa 488–Vh was added to the slides for the period of the incubation. The TR and Vhstructures are represented in green and yellow, respectively. The arrow shows the direction of themovement of the beads when they are pulled using the magnetic tweezers.


REPORTS

...

......


contour length increments of ~28.5 nm corre-spond to the unfolding of the four flanking I27handles (17). The largest peak at the right ofFig. 4B represents the detachment of the canti-lever from the polyprotein. Before the four forcepeaks for the I27 domains, there are severalsmaller peaks over a distance of about 160 nm,which corresponds closely to the expected lengthof the fully stretched 407-residue TR domain. Dis-tances between the unfolding peaks were deter-mined by using the contour length measuredby fits of the wormlike chain model (green linesin Fig. 4B). [The contour length of a fully stretchedprotein is about 0.4 nm per amino acid (19).] Ahistogram of positions for each of the five TRforce peaks (Fig. 4C, n = 120 traces) with theresting length of the protein as reference (163 nmminus the measured distance to the first I27 mod-ule) showed that the unfolding pattern was veryregular. Five unfolding peaks with average forcesof 29 T 19, 31 T 19, 36 T 22, 42 T 22, and 51 T26 pN were identified at 24 T 7, 81 T 9, 107 T 3,129 T 6, and 152 T 5 nm, respectively. His-tograms showing the force distribution for eachpeak can be found in fig. S2.

The force extension experiment provides amechanical description for TR, that is, unfoldingforces and contour lengths. To study the unfoldingrate of TR, we used force-clamp spectroscopy, atechnique that allows us to apply a constant cal-ibrated force to a single TR for determining theunfolding trajectory as a function of time (20). Adouble pulse protocol was used in order to sep-arate the unfolding of TR from those of I27 mod-ules. The first force pulse was set up at forcesranging from 20 to 50 pN to allow for the un-folding of TR. A second pulse of force of 150 pNwas applied to capture the unfolding of the I27modules, represented with the four identical stepsof 24 nm (Fig. 4D). Only those recordings show-ing at least three I27 unfolding events were ana-lyzed. The portion of the traces containing the TRunfolding was extracted, averaged, and fitted to asingle exponential with a time constant, t, whichprovides the unfolding rate, r = 1/t (Fig. 4E). Con-sidering the number of traces collected per force(~20 traces), the exponential fitting provides a goodapproximation for the unfolding rate that adequate-ly describes the process over the entire range offorces. The unfolding rate versus the applied forceis presented in Fig. 4F. The extrapolation at 12 pN,the unfolding force value used in the bindingexperiments, yields a constant rate of 2.7 s−1.

Taking together the results of the atomicforce spectroscopy and the binding experimentsof Vh to the TR, it is possible to make someconclusions. First, the fact that only one bind-ing event was observed in the absence of force-induced stretching indicates that, in its intactstate, TR contains one active VBS, as was pre-dicted in previous studies using several biophys-ical techniques (9, 12). This active VBS couldbe located in the helix 6, 9, 11, or 12, most likelyhelix 9, where the trypsin cleavage site is found(12) and which is likely the initial vinculin in-

teraction site in cells. Second, stretching the TRwith 12 pN of force can increase the Vh bindingevents from one to three. Thus, force-inducedstretching of the TR caused conformationalchanges in its helical bundle, exposing two ofthe four remaining buried VBSs, which becomeavailable to bind Vh.

Previous steered molecular dynamic calcula-tions potentially provide insights into the numberof VBSs activated by force-induced stretching(11). Because the mechanical stabilities of thedomains that form the TR are different (8, 12),force applied across the H1-H12 fragment willsequentially break the rod into smaller helixbundles. In the calculations, force first splits thefragments H1-H8 from H9-H12, and then H1-H8breaks into the sub-bundles H1-H5 and H6-H8.The formation of these two intermediate statesis in agreement with the unfolding pattern thatwe obtained by using force extension spec-troscopy, because the subsequent unfolding ofthe three domains would give five mechanicaltransition states as observed. The predicted frag-mentation pattern involves the sequential expo-sure of VBSs in TR, and this could explain whywe observed up to three instead of five Vh bind-ing events when force is applied. Although the12-pN force used in the binding experiments is

lower than the average unfolding forces in theatomic force microscopy (AFM) experiments, itis applied for a much longer time and could be aseffective. More experimentation is needed to de-fine the partial unfolding of the TR to interme-diate states, in which the cryptic VBSs becomeexposed but the secondary structure of helicesenables recognition by Vh. Similarly, lower un-folding forces have been used in AFM force-clampunfolding of polyubiquitins (21) as compared withAFM force extension experiments (22). Further-more, it is known that at constant low force (i.e.,12 pN), the unfolding and refolding processesmay compete with each other, leading to situa-tions such as that reported by Cecconi et al. (23)in which only up to three VBSs are available.Higher forces might stretch the VBS helices andreduce the binding probability; however, a newexperimental setup is needed to generate thehigher forces.

These findings establish that force-inducedstretching can expose previously buried bind-ing sites for vinculin in the TR. The consensussequence for the 11 VBSs in the TR indicatesthat they have extensive hydrophobic inter-actions with adjacent amphipathic helices (7).If this scenario is correct, the proposed mechano-transduction mechanism for the VBSs in TR

Fig. 3. Diagram of photobleaching events of Alexa 488–Vh bound to (A) TR, (B) dimeric tandem TR,and (C) a-actinin. Histograms show the number of beads per photobleaching event. In all cases, blue,gray, and green colors correspond with no force, 2-pN force, and 12-pN force applied, respectively. TheTR, talin dimeric tandem (positive control), and a-actinin (negative control) showed maximally 1 and 3,2 and 6, and 1 and 1 photobleaching events (black arrows) when no force and 12 pN force, respec-tively, was applied.


REPORTS

...

......


may be a general pathway used by the talinmolecule to amplify biochemical signals uponactivation by force. Indeed, talin head interactswith membrane integrins while its tail bindsactin filaments, so that force transmitted to thematrix from the actin cytoskeleton would log-ically be borne by talin. Actomyosin contractioncould develop force-dependent talin unfoldingthat may cause vinculin binding and activation,leading to several intracellular responses. Al-though the force per integrin is 1 to 2 pN onaverage, the bond between integrin and fibro-nectin is able to withstand a force of about 20 pNbefore breaking (24). Thus, forces used in thisstudy are in a physiological range.

Talin stretching may provide a prototype fora wide range of signaling events in cells becausethere are many multidomain proteins that linkcytoskeleton with membrane sites (25). Learn-ing the mechanism by which stretching forcesactivate binding may broaden our understand-ing of the fundamental workings of mechano-sensory systems in cells.

References and Notes1. V. Vogel, M. Sheetz, Nat. Rev. Mol. Cell Biol. 7, 265

(2006).2. D. R. Critchley, Biochem. Soc. Trans. 33, 1308 (2005).3. D. R. Critchley, Curr. Opin. Cell Biol. 12, 133 (2000).4. C. G. Galbraith, K. M. Yamada, M. P. Sheetz, J. Cell Biol.

159, 695 (2002).5. T. Izard, C. Vonrhein, J. Biol. Chem. 279, 27667 (2004).6. D. A. Calderwood, V. Tai, G. Di Paolo, P. De Camilli,

M. H. Ginsberg, J. Biol. Chem. 279, 28889 (2004).7. A. R. Gingras et al., J. Biol. Chem. 280, 37217 (2005).8. E. Papagrigoriou et al., EMBO J. 23, 2942 (2004).9. I. Fillingham et al., Structure 13, 65 (2005).10. T. Izard et al., Nature 427, 171 (2004).11. V. P. Hytonen, V. Vogel, PLOS Comput. Biol. 4, e24

(2008).12. B. Patel et al., J. Biol. Chem. 281, 7458 (2006).13. Y. Sawada et al., Cell 127, 1015 (2006).14. Y. Harada et al., Biophys. J. 76, 709 (1999).15. S. E. Lee, R. D. Kamm, M. R. Mofrad, J. Biomech. 40,

2096 (2007).16. M. Tanase, N. Biais, M. Sheetz, Methods Cell Biol. 83,

473 (2007).17. M. Carrion-Vazquez et al., Proc. Natl. Acad. Sci. U.S.A.

96, 3694 (1999).18. A. F. Oberhauser, P. E. Marszalek, H. P. Erickson,

J. M. Fernandez, Nature 393, 181 (1998).19. S. R. Ainavarapu et al., Biophys. J. 92, 225 (2007).

20. J. M. Fernandez, H. Li., Science 303, 1674 (2004).21. M. Schlierf, H. B. Li, J. M. Fernandez, Proc. Natl. Acad.

Sci. U.S.A. 101, 7299 (2004).22. M. Carrion-Vazquez et al., Nat. Struct. Biol. 10, 738

(2003).23. C. Cecconi, E. A. Shank, C. Bustamante, S. Marqusee,

Science 309, 2057 (2005).24. G. Jiang, G. Giannone, D. R. Critchley, E. Fukumoto,

M. P. Sheetz, Nature 424, 334 (2003).25. D. A. Calderwood, M. H. Ginsberg, Nat. Cell Biol. 5, 694

(2003).26. Materials and methods are available as supporting

material on Science Online.27. We thank D. Critchley and N. Bate for Vh and TR

monomer constructs, C. Badilla-Fernandez for advisementwith constructs, and V. Hytonen for contribution toFig. 1. This work was supported by NIH grants to J.M.F.and M.P.S., an NIH award to A.d.R., a grant from theSpanish Ministry of Science and Education to R.P.-J., anda Marie Curie international fellowship to P.R.-C.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/323/5914/638/DC1Materials and MethodsFigs. S1 to S4References

8 July 2008; accepted 10 December 200810.1126/science.1162912

Fig. 4. AFM experiments. (A) Diagram of the polyprotein designed forAFM, I272–TR–I272. (B) Force extension trace for I272-TR-I272. (C) His-togram of the contour length increments (DLc) for the unfolding peaks ofthe TR (120 traces). (D) Force-clamp trace obtained by stretching I272-TR-I272 at 20 and 150 pN of force. The red and black lines representextension and force, respectively. (E) Probability of unfolding versus time

for TR. Average recordings of TR unfolding at 20 pN (15 traces), 30 pN(14 traces), 40 pN (17 traces), and 50 pN (22 traces). A single expo-nential (colored smooth lines) is fitted to each average trace (in black).(F) The rate constant of unfolding as a function of force. Data are rep-resented as mean T SD; error bars were calculated by using the boot-strapping method (26).


REPORTS

...

......


Mechanically Activated IntegrinSwitch Controls a5b1 FunctionJulie C. Friedland,1,2* Mark H. Lee,1,2* David Boettiger1,2,3†

The cytoskeleton, integrin-mediated adhesion, and substrate stiffness control a common set of cellfunctions required for development and homeostasis that are often deranged in cancer. Theconnection between these mechanical elements and chemical signaling processes is not known.Here, we show that a5b1 integrin switches between relaxed and tensioned states in response tomyosin II–generated cytoskeletal force. Force combines with extracellular matrix stiffness togenerate tension that triggers the integrin switch. This switch directly controls the a5b1-fibronectinbond strength through engaging the synergy site in fibronectin and is required to generate signalsthrough phosphorylation of focal adhesion kinase. In the context of tissues, this integrin switchconnects cytoskeleton and extracellular matrix mechanics to adhesion-dependent motility andsignaling pathways.

Integrins are expressed on the surface of mostcells in the body, and their primary function isto maintain a dynamic adhesion between the

cell and its microenvironment (1). Integrin-mediated cell adhesion controls critical intra-cellular signals that regulate cell differentiation,proliferation, and survival (2–4). Recent analyseshave found that tissue stiffness and the ability ofcells to generate tension also control the samesignaling pathways (5, 6). How the mechanicalproperties of the cell’s microenvironment are

transmitted to intracellular biochemical pathwaysis not known. However, several intracellular sig-naling molecules, including src, Cas, and vincu-lin, show tension-dependent conformationalchanges that affect kinase activity, availability ofphosphorylation sites, or intracellular localization(7–9). Binding proteins to the extracellular andcytoplasmic domains of integrins forges a linkbetween the extracellular matrix and the actincytoskeleton that is tensioned by myosin II mo-tors acting on actin filaments (7). This complexcontains the major mechanical elements that havebeen associated with the mechano-sensing capa-bilities of the cell (10). Integrins are necessaryelements for most mechano-sensing models andlie at the beginning of the sensing pathway. Here,we investigate the mechano-responsive proper-ties of a5b1 integrin bound to surface-attachedfibronectin.

HT1080 cells plated on a stiff fibronectin-coated substrate attach through a5b1 integrin, andthese adhesive bonds are then tensioned by actinand myosin II (11). To analyze the role of tensionfor a5b1-fibronectin adhesive bonds, we usedpharmacological inhibitors of myosin II functionand activation (12). The effects of these inhibitorson adhesive a5b1-fibronectin bonds were mea-sured by chemical cross-linking. In previousanalyses, the fraction of total a5b1 integrin thatcross-linked was proportional to the number ofadhesive bonds (11). This assay demonstrated adose-dependent decrease in cross-linked a5b1integrin for cells treated with the myosin light-chain kinase inhibitor, ML-7 (Fig. 1A), and themyosin II inhibitors 2,3-butanedione monoxime(BDM) and blebbistatin (fig. S1, A to C) [sup-porting online text (SOM text), note S1]. Similardecreases were observed with the actin assemblyinhibitors cytochalasin D and latrunculin A (fig.S1, D and E). Thus, blocking the ability of thecell to apply tension reduced the number of cross-linkable a5b1-fibronectin bonds. In contrast,these inhibitors had no effect on the chemicalbinding of soluble fibronectin to a5b1 integrin oncells in suspension (fig. S2, A and B). Anotherapproach to measuring adhesive bonds is to mea-sure the force required to detach the cell. Wedeveloped a spinning disc device for this purposeand demonstrated that the cell detachment forcewas proportional to the number of adhesivebonds (11, 13, 14). This measure of adhesivebondswas not affected by inhibiting actin-myosin–induced cell tension (Fig. 1B). Hence, we hypoth-esized that there were two distinct forms of thea5b1-fibronectin adhesive bonds: one that formedin the absence of cell tension (relaxed bonds) thatcould not be cross-linked and one that formed in

1Institute for Medicine and Engineering, University ofPennsylvania, Philadelphia, PA 19104, USA. 2Departmentof Microbiology, University of Pennsylvania, Philadelphia,PA 19104, USA. 3Department of Pharmacology, Universityof Pennsylvania, Philadelphia, PA 19104, USA.

*These authors contributed equally to the work.†To whom correspondence should be addressed: E-mail:[email protected]

Fig. 1. Analysis of adhesive a5b1-fibronectin bonds by chemical cross-linkingbefore and after application of external force. (A) Proportion of cross-linked a5and b1 integrin in ML-7–treated cells. (B) Tension-inhibited cells were analyzed byusing the spinning disc: [ML-7] = 15 mM; [BDM] = 10 mM. (C) Distribution ofcross-linkeda5b1 integrin 60min after plating, before and after spinning T10mM

BDM. Nonspecific aggregates of secondary antibody provide an internal focus orexposure control. Scale bar, 25 mm. (D) Prespin represents tensioned bonds, andpost spin represents total bonds; the difference between the two representsrelaxed bonds. Data show percentage of total a5 and b1 cross-linked as a functionof time. (E) BDM-treated cells, the same as (D). Means T SEM; n = 3.


REPORTS

the presence of cell tension (tensioned bonds)that could be cross-linked. Tension involvesholding one end of the bond and pulling on theother. This can have two outcomes; either thebond separates or the bond switches to a new,stronger state that resists more tension.

To determine whether a state shift occurredwhen tension was applied, cells were allowed toadhere for 60 min (prespin) or incubated for 60

min and spun 5 min, which would apply tensionto adhesive bonds by pulling on the cell (postspin) (Fig. 1C). The control showed considerablecross-linked a5b1 integrin prespin and only aslight increase post spin, particularly toward thecenters of the cells. The BDM-inhibited cellsshowed no detectable cross-linking before spin-ning, in agreement with the earlier result, but postspin, the levels of cross-linking were similar tothose of the control. Thus, external force couldsubstitute for intracellular force. The resultingtension switched the relaxed (non–cross-linkable)a5b1-fibronectin bonds to the tensioned (cross-linkable) state. To expand this result, a quantita-tive analysis of cross-linkable a5b1 was performedat 15, 30, and 60 min after plating, both beforeand after a 5-min spin (Fig. 1, D and E, and fig.S2C). The prespin values represent tensionedbonds; the spinning would convert all bonds totensioned bonds, and hence, the post spin valuesrepresent total bonds; and the difference repre-sents the relaxed bonds. Control cells showed aprogressive increase in the level of tensionedbonds with time. In the BDM-treated cells, therewas very little increase in tensioned bonds, al-though the total number of bonds increased withsimilar kinetics and extent as the control. Thus, inthe absence of tension, adhesive a5b1-fibronectinbonds accumulated at a rate similar to that in thepresence of tension, but the bonds remained in arelaxed state. Because it is necessary to generatethe linkage from substrate througha5b1-fibronectinbonds to actin and myosin before tension can beexerted, the relaxed bond is a necessary, logicalprecursor to the tensioned bond. It is unlikely thatthe increase in bonds observed after spinning couldbe caused by the spinning itself because: (i) theextent of the increase depended on the time of theprespin incubation and not on the spinning time,and (ii) a5b1-fibronectin bonds accumulate rela-tively slowly because of issues of diffusion and therequirements for complex formation; 5 min is notsufficient time for the observed increases (fig. S2Dand Fig. 1C).

If tension converts a5b1 integrin–fibronectinbonds from the relaxed to the tensioned staterather than causing the bond to dissociate, theintermolecular contacts that form the two statesmust be distinct (15). The relaxed state can beapproximated by the chemical binding betweenfibronectin [the fragment containing the 7th to10th type II repeats (Fn7-10), including both theArg-Gly-Asp (RGD) and synergy sites (16) (Fig.2A)] and the purified extracellular domain ofa5b1 integrin, because these reactions were donewith purified components, the bonds could not betensioned (17). A combination of electron mi-croscopy and surface plasmon resonance analy-ses demonstrated that the binding interfaceincludes the RGD but not the synergy site. Todetermine the intermolecular contacts for ten-sioned bonds, the spinning disc was used to ana-lyze fibronectin mutants. Bonds were allowed toform for 60 min to approximate steady-state ad-hesion (fig. S2D). When RGD was deleted, therelaxed bonds could not form and no cells at-tached. Individual synergy site mutations inwhich Ala replaced Arg at residue 1374 and 1379(R1374A and R1379A, respectively) producedweaker bonds that required less force to rupture,and this effect was roughly additive in the doubleR1374/1379A mutant (Fig. 2B). Thus, the relaxedbonds involved only the RGD site, but thetensioned bonds required also R1374 and R1379in the synergy site. The reduction of ~90% inadhesion strength for the double R1374/1379Amutant implies that most of the bond strengthdepended on synergy-site contacts. Both moleculardocking and the presence of epitopes for inhibito-ry monoclonal antibodies suggest that the syner-gy site would contact the b-propeller in a5 (18).The differential cross-linking observed for therelaxed and tensioned states can be explained bythe resulting change in intermolecular distance.

The generation of tension involves pullingagainst a resistance, which is provided by thesubstrate to which the fibronectin is attached. Upto this point, the experiments have used very stiff

Fig. 3. Substrate stiffness controls the mechanical activation of a5b1 integrin and signaling to FAK. (A) Meancell detachment force per a5b1 integrin in the cell-gel interface. See fig. S2 calculations. (B) Proportion ofb1 integrin cross-linked to fibronectin-polyacrylamide gels 4 hours after plating. Background was subtractedand normalized to 400 Pa gels and for proportion of a5b1 integrin in the interface. (C) Effect of ML-7 on FAKY397 phosphorylation. Sus denotes cells in suspension; data were normalized to untreated controls. (D) Effectof Fn7-10 mutants on FAK Y397 phosphorylation normalized to wild type. T, Thr. (E) Effect of gel stiffness onFAK Y397 () and Y861 () phosphorylation. Means T SEM; n = 3.

Fig. 2. Tensioned a5b1-fibronectin bonds en-gage the synergy site of fibronectin. (A) Crystalstructure of fibronectin type III repeats 9 and 10.a5b1 integrin binds from the top overlapping Fn9and Fn10 surfaces. The RGD extended loop inFn10 acts as the recognition sequence. Synergy-site R1374 and R1379 (white) binds near theedge of the b-propeller of a5. Charged residues(black). (B) Spinning disc analysis of mutant Fn7-10. Means T SEM; n = 3.


REPORTS

substrates that provide strong resistance to forcesapplied to the cell or cytoskeleton. On these sub-strates, most a5b1 integrin–fibronectin bonds weretensioned, but softer substrates in the range ofthat found in tissues (0.2 to 20 kPa) provide lessresistance, and the proportion of adhesive bondsin the relaxed state should increase. Fibronectinwas attached to polyacrylamide gel substrateswithdifferent stiffness and a5b1 integrin–fibronectinbonds were measured by using both spinningdisc and chemical cross-linking methods (Fig. 3,A and B). Note that corrections were made forcell spreading and shape differences (see fig. S3,A and B, and SOM text, note S2). Spinning discanalysis showed that the total number of adhesivebonds that formed was independent of substratestiffness. In contrast, the cross-linking assayshowed a strong dependence of the number oftensioned bonds on substrate stiffness. Thus, asthe substrate stiffness increased, the proportion ofa5b1-fibronectin adhesive bonds in the tensionedstate increased. If only tensioned bonds generateddownstream signals, this would provide amecha-nism for cells to sense microenvironmental stiff-ness. Inhibitors of myosin II and actin that blockedthe conversion of relaxed to tensioned bondsexhibited a parallel dose-dependent reduction inphosphorylation of the focal adhesion kinaseFAK on Y397 (Fig. 3C and fig. S4, A and B).The R1379Amutant that had the largest effect onsynergy-site engagement was sufficient to blockFAK Y397 phosphorylation, but the weakerR1374A mutant was below the tension threshold(Fig. 3D). Finally, FAK Y397 exhibited a stiffness-dependent increase in phosphorylation, whereasFAK Y861, which is tethering-independent (11),showed no stiffness-dependent change in phos-phorylation in the same cells (Fig. 3E).

a5b1 integrin functions both to adhere cells toa fibronectin substrate and to mediate down-stream signals that control cell physiology andcell fates (1, 3). Previous analyses have asso-ciated those functions with integrin activationand clustering (19, 20). The current models forintegrin activation are based on analyses in whichtension was not applied (17, 21). This activationswitches the integrin into a state that can bindligand and is necessary for the clustering ofintegrins and the formation of focal complexes(22). Although these processes were required toform adhesion complexes, they were insufficientto generate either strong adhesion or downstreamsignals (fig. S5). The initial a5b1-fibronectin bondis equivalent to the previously described activated-bound state (17, 19). Application of force switchesthe relaxed state to a new tensioned state, whichhas increased bond strength. Bonds that undergotension-strengthening have been called “catch-bonds” and have been previously reported forleukocyte selectin and Escherichia coli FimHadhesion receptors (23, 24). A catch-bond canfunction as a molecular clutch that is engagedunder tension and that will release when tension isreleased. Thus, a5b1-fibronectin bonds providemechanical clutches for cell migration (25). For

a5b1 integrin, the catch-bond mechanism de-pended on the synergy-site engagement, whichraises the question of whether synergy sites existmore broadly in integrin ligands.

Mechano-responsive proteins undergo func-tional conformational transitions, in response totension, that alter binding properties and/or enzy-matic functions (8, 10). Intracellular proteins, likesrc, that have this property require cytoskeletalstructure and integrin-mediated adhesion formechanical activation (9). Thus, the mechanicalactivation of src would depend on the mechanicalactivation of integrins, which suggests a modelthat contains multiple intra- and extracellularmechano-sensitive nodes to generate intracellularsignaling responses. Control of intracellular sig-nals by mechanical triggers provides a mecha-nism for spatial distribution of signals withincells. This is important for the cells’ capability tosense and respond to differences in microen-vironmental stiffness and generates a need formechanisms to control the cell’s “tensional”homeostasis (5).

References and Notes1. R. O. Hynes, Cell 69, 11 (1992).2. A. S. Menko, D. Boettiger, Cell 51, 51 (1987).3. C. K. Miranti, J. S. Brugge, Nat. Cell Biol. 4, E83 (2002).4. A. K. Fournier et al., J. Cell Sci. 121, 226 (2008).5. M. J. Paszek et al., Cancer Cell 8, 241 (2005).6. D. E. Discher, P. Janmey, Y. L. Wang, Science 310, 1139

(2005).7. D. Riveline et al., J. Cell Biol. 153, 1175 (2001).8. Y. Sawada et al., Cell 127, 1015 (2006).9. S. Na et al., Proc. Natl. Acad. Sci. U.S.A. 105, 6626

(2008).

10. T. Shemesh, B. Geiger, A. D. Bershadsky, M. M. Kozlov,Proc. Natl. Acad. Sci. U.S.A. 102, 12383 (2005).

11. Q. Shi, D. Boettiger, Mol. Biol. Cell 14, 4306(2003).


13. A. J. Garcia, F. Huber, D. Boettiger, J. Biol. Chem. 273,10988 (1998).

14. D. Boettiger, Methods Enzymol. 426, 1 (2007).15. B. C. Cunningham, J. A. Wells, J. Mol. Biol. 234, 554

(1993).16. D. J. Leahy, I. Aukhil, H. P. Erickson, Cell 84, 155

(1996).17. J. Takagi, K. Strokovich, T. A. Springer, T. Walz, EMBO J.

22, 4607 (2003).18. M. J. Humphries, E. J. Symonds, A. P. Mould, Curr. Opin.

Struct. Biol. 13, 236 (2003).19. R. O. Hynes, Cell 110, 673 (2002).20. S. Miyamoto, S. K. Akiyama, K. M. Yamada, Science 267,

883 (1995).21. J. P. Xiong et al., Science 296, 151 (2002).22. C. Cluzel et al., J. Cell Biol. 171, 383 (2005).23. O. Yakovenko et al., J. Biol. Chem. 283, 11596

(2008).24. B. T. Marshall et al., Nature 423, 190 (2003).25. Y. L. Wang, Sci. STKE 2007, e10 (2007).26. This work was supported by grant GM57388 from the

National Institutes of Health to D.B. We thank S. Akiyama(NIH) for 13G12 antibody, H. Erickson (Duke) formutant fibronectin constructs, and Qi Shi for technicalassistance.


12 November 2008; accepted 24 November 200810.1126/science.1168441

A Human Telomerase HoloenzymeProtein Required for Cajal BodyLocalization and Telomere SynthesisAndrew S. Venteicher,1,2 Eladio B. Abreu,3* Zhaojing Meng,4* Kelly E. McCann,1,5Rebecca M. Terns,3 Timothy D. Veenstra,4 Michael P. Terns,3 Steven E. Artandi1,2,5†

Telomerase is a ribonucleoprotein (RNP) complex that synthesizes telomere repeats in tissueprogenitor cells and cancer cells. Active human telomerase consists of at least three principalsubunits, including the telomerase reverse transcriptase, the telomerase RNA (TERC), anddyskerin. Here, we identify a holoenzyme subunit, TCAB1 (telomerase Cajal body protein 1),that is notably enriched in Cajal bodies, nuclear sites of RNP processing that are important fortelomerase function. TCAB1 associates with active telomerase enzyme, established telomerasecomponents, and small Cajal body RNAs that are involved in modifying splicing RNAs.Depletion of TCAB1 by using RNA interference prevents TERC from associating with Cajalbodies, disrupts telomerase-telomere association, and abrogates telomere synthesis by telomerase.Thus, TCAB1 controls telomerase trafficking and is required for telomere synthesis in humancancer cells.

The telomerase reverse transcriptase (TERT)and the telomerase RNA component(TERC) comprise the minimal catalytic

core of the telomerase enzyme (1), whereasdyskerin is an RNA-binding protein that recog-

nizes the H/ACA sequence motif shared byTERC and two groups of noncoding RNAsinvolved in RNA modification: small Cajal bodyRNAs (scaRNAs) and small nucleolar RNAs(snoRNAs) (2, 3). Dyskerin functions in part


REPORTS

to support telomerase ribonucleoprotein (RNP)biogenesis and TERC stability (4, 5). TERT,TERC, and dyskerin are all components of ac-tive telomerase (6), and mutations in any ofthese genes can cause the human stem celldisorder dyskeratosis congenita (7). Other po-tential components of active telomerase includethree evolutionarily conserved dyskerin-associatedproteins, nucleolar protein 10 (NOP10), non-histone protein 2 (NHP2), and glycine/arginine–rich domain containing protein 1 (GAR1) (8–10),and ever-shorter telomeres 1A (EST1A), ahomolog of the yeast telomerase protein Est1p(11, 12). However, the size of active humantelomerase, estimated in the 0.65-to-2-MDarange (6, 13, 14), suggests the existence ofadditional components. We reasoned that otherdyskerin-associated proteins may be telomerase

components, and we therefore sought to purifydyskerin complexes.

To study dyskerin, we expressed tagged dyskerinprotein at endogenous levels in the absence ofcompeting endogenous protein (fig. S1) and iso-lated dyskerin complexes by using a dual-affinitychromatography strategy. Purified dyskerin com-plexes were analyzed by SDS–polyacrylamidegel electrophoresis (SDS-PAGE) and nano–liquidchromatography–tandem mass spectrometry(nanoLC-MS/MS) for identification of copurify-ing proteins (Fig. 1, A and B). Dense peptidecoverage was obtained for dyskerin and for thedyskerin-associated adenosine triphosphatases(ATPases) pontin and reptin (14). Each of theevolutionarily conserved dyskerin-binding pro-teins NHP2, NOP10, and GAR1 was detected,as were the dyskerin-associated proteins nucle-olar and coiled-body phosphoprotein of 140 kD.(Nopp140) and nuclear assembly factor 1 (NAF1),a nucleoplasmic factor required for the assemblyof H/ACA RNPs, including telomerase (Fig. 1B)(15). In addition, this approach identified WDrepeat domain 79 (WDR79) (Fig. 1B), a pro-tein that had not been previously implicated indyskerin or telomerase function.

We further characterized WDR79, hereafterreferred to as TCAB1 (telomerase Cajal body pro-tein 1) (fig. S2). Endogenous TCAB1 wasspecifically bound to Flag-dyskerin that wasimmunoprecipitated from Flag-dyskerin+shRNA

HeLa cells (shRNA, short hairpin RNA), as were

endogenous pontin, NAF1, TERT, and TERC(Fig. 1C). Interactions between TERT and dyskerinwere disrupted by ribonuclease A (RNase A) treat-ment of the extract, which degraded TERC. In con-trast, dyskerin interactions with TCAB1, NAF1, andpontin were not RNase A–sensitive, indicating thatthese associations occur through protein-proteincontacts (Fig. 1C). Reciprocal immunoprecipita-tion (IP) of Flag-TCAB1 from Flag-TCAB1+shRNA

HeLa cells showed that Flag-TCAB1 not onlyassociates with endogenous dyskerin but also withTERT and TERC, the catalytic core of telomerase(Fig. 1D). Like the TERT-dyskerin association,the binding of TERT to Flag-TCAB1 was RNaseA–sensitive, suggesting that the interaction ofTCAB1 with telomerase is dependent on TERC(Fig. 1D). Thus, TCAB1 interacts specifically withdyskerin, TERT, and TERC, all three known com-ponents of active telomerase.

Although TCAB1 associated with TERT,TERC, and dyskerin, it did not interact with theassembly factors NAF1, pontin, or reptin (Fig. 1D),suggesting that TCAB1 may be a componentof the enzymatically active telomerase complexrather than a pre-telomerase complex. To testthis hypothesis, we asked to what extenttelomerase activity in cell extracts was associatedwith overexpressed TCAB1. Flag-tagged TCAB1,dyskerin, and NAF1 were depleted from extractsby IP (Fig. 2A, lanes 14 to 21). Flag-TCAB1and Flag-dyskerin immunoprecipitates wereassociated with high telomerase activity (Fig. 2A,

Fig. 1. Identification of TCAB1 as a dyskerin- andtelomerase-interacting protein. (A) Dual affinity–purified dyskerin complexes fractionated by SDS-PAGE and silver-stained. (B) Unique and totalpeptides corresponding to dyskerin-associatedproteins identified by nanoLC-MS/MS. (C) Flag-dyskerin interactions with endogenous TCAB1 andtelomerase components and (D) Flag-TCAB1 andFlag-NAF1 interactions with endogenous telomerasecomponents. IP–Western blot using extracts fromFlag-dyskerin+shRNA cells, Flag-TCAB1+shRNA cells, orFlag-NAF1+shRNA cells is shown. IB, immunoblot; NB,Northern blot; parental, HeLa cells. Plus signindicates treatment of extracts with RNase A duringIP. Recovery control was exogenous RNA spiked intosamples after IP to control for differential RNArecovery.

1Department of Medicine, Stanford School of Medicine,Stanford, CA 94305, USA. 2Program in Biophysics, StanfordSchool of Medicine, Stanford, CA 94305, USA. 3Departmentof Biochemistry and Molecular Biology and Department ofGenetics, University of Georgia, Athens, GA 30602, USA.4Laboratory of Proteomics and Analytical Technologies, Ad-vanced Technology Program, Science Applications Interna-tional Corporation (SAIC)–Frederick, National Cancer Institute(NCI)–Frederick, Frederick, MD 21702, USA. 5Program inCancer Biology, Stanford School of Medicine, Stanford, CA94305, USA.



REPORTS

...

......


lanes 12 and 13). Moreover, purification of eitherFlag-TCAB1 or Flag-dyskerin depleted theseextracts of telomerase activity, indicating thatboth Flag-TCAB1 and Flag-dyskerin wereassociated with nearly all telomerase activity inthe extract. In contrast, Flag-NAF1 bound only asmall percentage of telomerase and did notdeplete telomerase activity from cell extracts(Fig. 2A, lanes 1 to 8). Furthermore, IP of eitherFlag-TCAB1 or Flag-dyskerin depleted TERC butnot unrelated RNAs such as U1 from cellextracts (Fig. 2A, lanes 14 to 21). Flag-TCAB1and Flag-dyskerin were associated with similaramounts of endogenous TERT and TERC,further suggesting that TCAB1 and dyskerinreside together in the same active telomerasecomplexes (Fig. 2A, lanes 24 to 25).

To understand the composition of endog-enous telomerase, IPs were performed usingantibodies to TCAB1 and NAF1, each of whichefficiently depleted its cognate protein from cellextracts (Fig. 2B, lanes 9 to 13). High telo-merase activity was associated with TCAB1immunoprecipitates, and antibodies to TCAB1quantitatively depleted telomerase activity fromcell extracts (Fig. 2B, lanes 1 to 8, and fig. S4).In contrast, antibodies to NAF1 pulled downonly a small percentage of telomerase activity.Assessing the TCAB1 immunoprecipitates fortelomerase components revealed that TCAB1interacts with endogenous dyskerin, TERT, andTERC (Fig. 2B, lanes 14 to 16). IP of TCAB1effectively depleted cell extracts of TERC byuse of Northern blot (Fig. 2B, lanes 9 to 13,and figs. S4 and S5). We conclude that TCAB1,like dyskerin, associates stably with a vast ma-jority of active telomerase and TERC and there-fore is a component of a human telomeraseholoenzyme.

Using immunofluorescence (IF), we foundthat stably overexpressed hemagglutinin (HA)–tagged TCAB1 (HA-TCAB1) was distributedweakly throughout the nucleoplasm but wasstrongly enriched within nuclear foci resemblingCajal bodies, sites of RNP processing shown tocontain dyskerin and TERC (16–18). IF showedthat the HA-TCAB1 foci overlapped with p80-coilin and were therefore Cajal bodies (Fig. 3A).Endogenous TCAB1 was also highly enriched inCajal bodies, with smaller amounts distributed inthe nucleoplasm (Fig. 3B). Although dyskerinaccumulates in both Cajal bodies and nucleoli,the Cajal body–restricted localization of TCAB1suggested that TCAB1 may function specificallywith telomerase and with other noncoding RNAsfound in Cajal bodies.

Immunoprecipitates of Flag-TCAB1 (Fig.3C) and endogenous TCAB1 (Fig. 3D) wereassayed for scaRNAs and representatives ofother classes of nuclear noncoding RNAs byNorthern blot. Both overexpressed TCAB1 andendogenous TCAB1 specifically bound H/ACAscaRNAs but associated with neither snoRNAsnor splicing RNAs. In contrast, Flag-dyskerinspecifically bound both H/ACA scaRNAs and

H/ACA snoRNAs, whereas NAF1 did notassociate stably with any noncoding RNAsstudied. For each H/ACA scaRNA tested, asubstantial proportion was depleted from theextract by IP of TCAB1 (Fig. 3, C, lanes 5 to 6,and D, lanes 1 to 5). Thus, TCAB1 specificallybinds scaRNAs, which share a Cajal body(CAB) box sequence that controls Cajal bodylocalization, and provides a potential mechanismto explain how scaRNAs, including TERC, areretained in Cajal bodies.

TCAB1 depletion by using retroviral-encoded shRNAs reduced neither telomeraseactivity nor TERC levels, in contrast to dyskerindepletion, which markedly diminished TERCand telomerase activity (Fig. 4A). Thus, TCAB1may be required in vivo for a step after theassembly of a catalytically competent telomerasecomplex. To determine the role of TCAB1 inCajal body localization of telomerase, TERC

localization was measured by RNA fluorescencein situ hybridization (FISH) in TCAB1-depletedHeLa cells. FISH revealed that TCAB1 knock-down substantially reduced the percentage ofcells in which TERC was found in Cajal bodies(Fig. 4, B and C) without affecting overall TERCRNA levels (Fig. 4A). Cajal bodies have beendirectly implicated in the delivery of TERC totelomeres during S phase (19–21). Therefore,we also examined the effect of TCAB1 knock-down on the localization of TERC to telomeres,using FISH for TERC and IF for telomere repeatbinding factor 2 (TRF2). Depletion of TCAB1substantially reduced the presence of TERC attelomeres during S phase (Fig. 4, B and C). Thus,TCAB1 is required for TERC localization inCajal bodies and for delivery of TERC totelomeres during S phase.

To determine whether TCAB1 is required fortelomere addition by telomerase, we first induced

Fig. 2. TCAB1 is a componentof active telomerase. (A) Flag-TCAB1 or Flag-dyskerin IP quan-titatively codepletes telomeraseactivity and TERC from extracts.Telomeric repeat amplificationprotocol (TRAP) assays wereperformed on extracts beforeand after IP (lanes 1 to 8) andon each immunoprecipitatedcomplex (lanes 10 to 13). De-pletion of tagged proteins anddepletion of endogenous TERCon extracts before and afterIP are shown in lanes 14 to21. IP of tagged proteins

and associated telomerase components are shown in lanes 22 to 25. U1-splicing RNA was the negativecontrol. (B) IP of endogenous TCAB1 codepletes telomerase activity and TERC. TRAP assays wereperformed on extracts before and after each IP (lanes 1 to 4) and on the IP (lanes 6 to 8). Depletion ofendogenous NAF1 or endogenous TCAB1 and depletion of TERC are shown in lanes 9 to 13. Associationof NAF1 and TCAB1 with telomerase components is shown in lanes 14 to 16. Recovery control wasexogenous RNA spiked into samples after IP to control for differential RNA recovery.


REPORTS

...

......


telomere elongation through TERC overexpres-sion in HTC75 fibrosarcoma cells. Overexpres-sion of wild-type TERC lengthened telomereswith cell passage, but telomere lengthening wasinhibited in cells overexpressing a CAB box–mutant TERC that fails to accumulate in Cajalbodies, as previously shown (19). Depletion ofTCAB1 in cells overexpressing wild-type TERCsubstantially inhibited telomere elongation,mimicking the effect of the CAB box mutation(Fig. 4D and fig. S6). To determine whetherTCAB1 is required for telomere synthesis byendogenous telomerase, we assayed telomerelengths with serial passage in HTC75 cellstransduced with TCAB1 shRNAs or with theempty vector. Both shRNAs targeting TCAB1led to progressive telomere shortening as

compared with that of the empty vector control(Fig. 4E and fig. S7), indicating that TCAB1 isrequired for telomere synthesis in human cancercells.

Our data identify TCAB1 as a Cajal body–enriched protein that associates with TERC andother scaRNAs and explains how TERC, andperhaps other scaRNAs, localize in Cajal bodies.TCAB1 functions as a telomerase holoenzymecomponent in the telomere synthesis pathway at astep after the assembly of a minimal telomerasecomplex containing TERT, TERC, and dyskerin.NAF1 and the ATPases pontin and reptin are re-quired for assembly of the catalytically competentcomplex (14, 15). In contrast, TCAB1 stably asso-ciates with active telomerase enzyme and directsit through Cajal bodies to telomeres. In this

manner, TCAB1 may act as a Cajal body–targeting or –retention factor, may facilitateadditional assembly steps of the enzyme in Cajalbodies, and/or may facilitate translocation oftelomerase to telomeres (Fig. 4F). The interactionof telomerase with telomeres and the activity ofthe telomerase holoenzyme may be enhanced bythe telomere-binding proteins TPP1 and protec-tion of telomeres 1 (POT1) (22, 23), as well asother factors that remain to be discovered.

References and Notes1. K. Collins, Nat. Rev. Mol. Cell Biol. 7, 484 (2006).2. U. T. Meier, Chromosoma 114, 1 (2005).3. A. G. Matera, R. M. Terns, M. P. Terns, Nat. Rev. Mol. Cell

Biol. 8, 209 (2007).4. J. R. Mitchell, E. Wood, K. Collins, Nature 402, 551

(1999).

Fig. 3. TCAB1 is enriched inCajal bodies and associates spe-cifically with scaRNAs. (A) HA-TCAB1 colocalizes with the Cajalbody marker p80-coilin. IF usingantibodies to HA (rat) and p80-coilin (mouse) on fixed HeLa cellsstably overexpressing HA-TCAB1is shown. (B) Colocalization of en-dogenous TCAB1 with p80-coilin.IF using antibodies to TCAB1(rabbit) and p80-coilin (mouse)on fixed HeLa cells. (C) Flag-TCAB1associates specifically with scaRNAsby IP–Northern blot. Cell extractfrom parental HeLa cells, Flag-NAF1+shRNA cells, Flag-TCAB1+shRNA

cells, or Flag-dyskerin+shRNA cellswas incubated with resin coatedwith antibodies to Flag. RNA wasisolated from extracts before (input)and after (depleted) antibodytreatment. Two micrograms oftotal RNA was fractionated byurea-PAGE (lanes 1 to 8), repre-senting 2% of the input. RNAsfrom IPs are shown in lanes 9 to12. (D) Endogenous TCAB1 asso-ciates specifically with scaRNAs.Antibodies to NAF1 or TCAB1,or the immunoglobulin G con-trol, were used in assays at thesame scale as in (C). RNAs frominput and depleted extracts areshown in lanes 1 to 5. RNAs fromIPs are shown in lanes 6 to 8.Recovery control was exogenousRNA spiked into samples after IPto control for differential RNArecovery.


REPORTS

...

......


5. Y. Mochizuki, J. He, S. Kulkarni, M. Bessler,P. J. Mason, Proc. Natl. Acad. Sci. U.S.A. 101, 10756(2004).

6. S. B. Cohen et al., Science 315, 1850 (2007).7. M. Kirwan, I. Dokal, Clin. Genet. 73, 103 (2008).8. D. Fu, K. Collins, Mol. Cell 28, 773 (2007).9. V. Pogacic, F. Dragon, W. Filipowicz, Mol. Cell. Biol. 20,

9028 (2000).10. F. Dragon, V. Pogacic, W. Filipowicz, Mol. Cell. Biol. 20,

3037 (2000).11. B. E. Snow et al., Curr. Biol. 13, 698 (2003).12. P. Reichenbach et al., Curr. Biol. 13, 568

(2003).13. G. Schnapp, H. P. Rodi, W. J. Rettig, A. Schnapp,

K. Damm, Nucleic Acids Res. 26, 3311 (1998).14. A. S. Venteicher, Z. Meng, P. J. Mason, T. D. Veenstra,

S. E. Artandi, Cell 132, 945 (2008).

15. C. Hoareau-Aveilla, M. Bonoli, M. Caizergues-Ferrer,Y. Henry, RNA 12, 832 (2006).

16. U. T. Meier, G. Blobel, J. Cell Biol. 127, 1505(1994).

17. B. E. Jady, E. Bertrand, T. Kiss, J. Cell Biol. 164, 647(2004).

18. Y. Zhu, R. L. Tomlinson, A. A. Lukowiak, R. M. Terns,M. P. Terns, Mol. Biol. Cell 15, 81 (2004).

19. G. Cristofari et al., Mol. Cell 27, 882 (2007).20. B. E. Jady, P. Richard, E. Bertrand, T. Kiss, Mol. Biol. Cell

17, 944 (2006).21. R. L. Tomlinson, T. D. Ziegler, T. Supakorndej, R. M. Terns,

M. P. Terns, Mol. Biol. Cell 17, 955 (2006).22. F. Wang et al., Nature 445, 506 (2007).23. H. Xin et al., Nature 445, 559 (2007).24. This work was supported by Medical Scientist Training

Program grant GM07365 (A.S.V.) and NCI grant

CA104676 (M.P.T. and R.M.T.), with a Supplement toPromote Diversity in Health-Related Research (E.B.A.).This project has been funded in part with federal fundsfrom the NCI under contract NO1-CO-12400 (T.D.V.).This work was supported by NCI grants CA111691 andCA125453 (S.E.A.) and by Leukemia and LymphomaSociety Specialized Centers of Research grant 7007-09(S.E.A.).



Fig. 4. TCAB1 is essen-tial for TERC localizationto Cajal bodies and fortelomere synthesis bytelomerase. (A) HeLa cellswere transduced with ret-roviruses expressing inde-pendent shRNA sequencestargeting the indicatedproteins or with emptyvector control. Telomeraseactivity was measured byTRAP assay. (B) TERC colo-calization with p80-coilinwas determined by RNAFISH for TERC (green) andIF for p80-coilin (red) (top).Cells synchronized in Sphase were assayed forTERC by RNA FISH (green)and for telomeres withantibody to TRF2 (red) toassess trafficking of TERCto telomeres (bottom). (C)Quantification of data in(B). (Top) Cells in whichTERC colocalized with p80-coilin (+) versus cells inwhich TERC was not de-tected in Cajal bodies (–)(P < 0.0001, Fisher’s exacttest). (Bottom) Cells inwhich TERC colocalizedwith telomeres (+) versuscells in which TERC wasnot detected at telomeres(–) (P < 0.0001, Fisher’sexact test). (D) Telomerelengths were measuredusing terminal restrictionfragment (TRF) Southernblot in HTC75 cells over-expressing wild-typeTERC (lanes 1 to 8) ormutant TERC-m1 (lanes 9to 12). Cells overexpress-ing wild-type TERC weretransduced with shRNAretroviruses targeting TCAB1 or with the empty vector. (E) Effect of TCAB1 de-pletion on endogenous telomerase was assessed by TRF Southern blot in pa-rental HTC75 cells that were transduced with the empty vector or retroviruses

expressing independent TCAB1 shRNAs. Cells were counted at each passageand population doublings are indicated. (F) Model for TCAB1 function in thetelomere synthesis pathway.


REPORTS

...

......


PAN1: A Receptor-Like ProteinThat Promotes Polarization of anAsymmetric Cell Division in MaizeHeather N. Cartwright,*† John A. Humphries,* Laurie G. Smith‡

Polarization of cell division is essential for eukaryotic development, but little is known abouthow this is accomplished in plants. The formation of stomatal complexes in maize involves thepolarization of asymmetric subsidiary mother cell (SMC) divisions toward the adjacent guardmother cell (GMC), apparently under the influence of a GMC-derived signal. We found that themaize pan1 gene promotes the premitotic polarization of SMCs and encodes a leucine-rich repeatreceptor-like protein that becomes localized in SMCs at sites of GMC contact. PAN1 has an inactivekinase domain but is required for the accumulation of a membrane-associated phosphoprotein,suggesting a function for PAN1 in signal transduction. Our findings implicate PAN1 in thetransmission of an extrinsic signal that polarizes asymmetric SMC divisions toward GMCs.

In plants, as in other eukaryotes, asymmetriccell divisions are associated with developmen-tal patterning events, formation of new cell

lineages and cell types, and maintenance of stemcell populations (1, 2). In all of these processes,developmental asymmetry is closely tied todivision polarity, which is thought to be orientedby either intrinsic or extrinsic cues. However,very little is known about such cues in plants orhow cells perceive and respond to them.

Extrinsic cues appear to orient asymmetriccell divisions in maize during the development ofepidermal stomatal complexes, each consistingof a pair of guard cells flanked by a pair of sub-sidiary cells that function in guard cell regulation(Fig. 1A). Stomatal complexes in grasses such asmaize develop through an invariant sequence ofcoordinated asymmetric cell divisions (fig. S1)(3). After an asymmetric division that forms aguard mother cell (GMC), its lateral neighbors,called subsidiary mother cells (SMCs), polarizetoward the GMC, positioning their nuclei and form-ing dense patches of cortical F-actin at the GMCcontact site (Fig. 1E) (4–6). SMCs subsequentlydivide asymmetrically to form subsidiary cellsflanking the GMC (Fig. 1E), and the GMC thendivides longitudinally to form a guard cell pair.Analysis of stomatal complex formation ingrasses led almost 50 years ago to the proposalthat SMC divisions are polarized by a signalemanating from GMCs (3), but the mechanismof interaction between the GMC and SMC hasremained unknown.

A recessive mutation in maize, pan1, causesdefects in the premitotic polarization of SMCsand the formation of abnormal subsidiary cells(5). An ethylmethane sulfonate (EMS)–induced,

non-allelic pan2 mutation caused a similar phe-notype (Fig. 1, B toD). Inwild-type (WT) leaves,92% of 317 SMCs analyzed in files of develop-ing stomata containing at least one divided SMChad polarized nuclei, and 80% of those had well-focused actin patches (Fig. 1, E and H). In con-trast, only 60% of 325 pan1 SMCs and 62% of392 pan2 SMCs had polarized nuclei, and mostof these SMCs either lacked an actin patch or hada delocalized patch (Fig. 1, F to H). The majorityof mutant SMCs with unpolarized nuclei lackedactin patches, but many of them had delocalized

or normal actin patches (Fig. 1, F to H). Thus,panmutant SMCs exhibit defects in both nuclearpolarization and actin-patch formation, but thesedefects are not closely correlated, suggesting thatthey are largely independent effects of the panmutations on SMC polarity.

We used amutant allele containing aMutator1(Mu1) transposon insertion to clone the pan1 gene(7). In this allele, Mu1 is inserted into the codonencoding amino acid 135 of a 662–amino acidleucine-rich repeat receptor-like kinase (LRR-RLK)belonging to the LRRIII subfamily (8). This pro-tein consists of a predicted extracellular domaincontaining five LRR motifs, a transmembrane do-main, and an intracellular serine/threonine kinasedomain. In plants carrying an EMS-induced pan1mutation (pan1-EMS), a premature stop codon inthe first exon truncates this protein at amino acid47, confirming the identity of this gene as pan1.Other LRR-RLKs function as receptors or co-receptors in diverse aspects of plant developmentand defense (9, 10). Extracellular LRRmotifs areimplicated in ligand binding; intracellular kinasedomains are implicated in signal transductionvia autophosphorylation and phosphorylationof target proteins.

To further investigate PAN1 function,we raiseda polyclonal antibody that recognized a protein ofaround 75 kD by immunoblotting, which is closeto the predictedmolecularmass of PAN1 (68.4 kD).This proteinwas depleted in extracts of both pan1-Mu and pan1-EMS mutants, confirming its iden-

Section of Cell and Developmental Biology, University ofCalifornia San Diego, 9500 Gilman Drive, San Diego, CA92093–0116, USA.

*These authors contributed equally to this work.†Present address: Stowers Institute for Medical Research,1000 East 50th Street, Kansas City, MO 64110, USA.‡To whom correspondence should be addressed. E-mail:[email protected]

Fig. 1. Stomatal complex formation in WT and pan mutant maize leaves. (A to C) Toluidine blueO–stained epidermal peels from adult leaves of WT (A), pan1 (B), and pan2 mutant (C). Guard cellsstain blue [black arrow in (A)]; subsidiary cells stain pink [black arrowheads in (A)]. White arrowheads in(B) and (C) indicate abnormal subsidiary cells in pan mutant leaves. Scale bar, 100 mm. (D) Quantitativeanalysis of abnormal subsidiary cells in mature leaf tissue. Error bars represent SEM (n = 5 plants pergenotype; >600 subsidiary cells were analyzed for each genotype). (E to G) F-actin (green) in developingstomata of WT (E), pan1 (F), and pan2 (G) with propidium iodide-stained nuclei in blue (single confocalplanes). In (E), arrowheads on GMCs point to normal actin patches in adjacent SMCs, and dashed linesindicate division planes in two divided SMCs. In (F) and (G), arrows indicate delocalized actin patches, andu indicates unpolarized nuclei. Scale bar, 10 mm. (H) Quantitative analysis of SMC polarity defects (P,polarized nucleus; U, unpolarized nucleus). Error bars represent SEM [n = 5 plants per genotype, the sameplants analyzed for the graph in (D); >300 SMCs were analyzed for each genotype].


REPORTS

...

......


tity as PAN1 (Fig. 2). This protein was found inthe microsomal fraction of extracts from the divi-sion zone ofWTmaize leaves (Fig. 2), consistentwith the expectedmembrane localization for PAN1.Immunoblot analysis indicated that in developingleaves, PAN1 was most abundant in the divisionzone, but the protein was also found in dividingregions of other tissues that lack stomata (fig. S2),suggesting that PAN1 may function in other pro-cesses besides stomatal development. However,pan1 mutants do not have obvious developmen-tal defects affecting other tissues or cell types.

In WT leaves, whole mount immunofluores-cence revealed PAN1 only in SMCs and newlyformed subsidiary cells at sites of contact withGMCs (Fig. 3A). These “PAN1 patches” were

absent in pan1mutants (Fig. 3B), confirming thespecificity of immunolabeling for PAN1. Specif-ic labeling with the PAN1-specific antibody to(anti-PAN1) was observed only at the cellperiphery, consistent with localization to theplasma membrane. Alternatively, although thecortical endoplasmic reticulum (ER) itself wasnot specifically enriched in SMCs at GMCcontact sites (fig. S3), some or all of the PAN1protein may be ER-localized. PAN1 protein waspresent in pan2 mutants at reduced levels andremained associated with the microsomal frac-tion (Fig. 2), but no specific PAN1 staining wasobserved in pan2 mutants (Fig. 3C).

Double labeling of PAN1 and actin was em-ployed to investigate the timing of PAN1 patch

formation relative to nuclear polarization and actin-patch formation in WT cells. 92% of 168 SMCswith fully polarized nuclei had both PAN1 andactin patches (Fig. 3, D to I). Among 119 SMCswith unpolarized or partially polarized nuclei(presumably in the process of becoming polar-ized), 27% had neither PAN1 nor actin patches,33% had only PAN1 patches, and 40% had bothPAN1 and actin patches (Fig. 3, D to I). None ofthese SMCs had actin patches only. GMC lengthprovides a measure of developmental age,increasing from inception, when flanking SMCshave unpolarized nuclei, through nuclear polar-ization and SMC division (Fig. 3M). The averagelengths of adjacent GMCs were lowest for theSMC class lacking both PAN1 and actin patches,intermediate for the PAN1+ actin– class, andhighest for the PAN1+ actin+ class (Fig. 3N).Thus, PAN1 and actin patches both appearedafter GMC formation and before nuclear polar-ization, but PAN1 patches were detectable beforeactin patches. The presence of PAN1 and actinpatches exclusively at sites of contact with GMCssupports the conclusion that these patches areinduced by GMCs rather than by intrinsic cues.Further supporting this conclusion, in unusualcases where a single SMC flanks two GMCs,PAN1 and actin patches were observed adjacentto both GMCs (n = 14) (Fig. 3, J to L).

Although the PAN1 kinase domain is wellconservedwith enzymatically active serine/threoninekinases, it lacks several amino acid residues con-served in active kinases, including a lysine res-idue in subdomain II that is critical for kinaseactivity (11) (fig. S4), suggesting that it is not anactive kinase. Consistent with this prediction, thekinase domain of PAN1 did not detectably phos-phorylate itself or the artificial substrate myelinbasic protein (MBP) in vitro under conditionswhere the kinase domain of the LRR-RLK BRI1phosphorylated both substrates robustly (Fig. 4A).Some LRR-containing transmembrane proteinslack intracellular kinase domains altogether andare thought to function in signal transduction pro-cesses by associating with an active kinase (9).To investigate whether PAN1might function sim-ilarly, microsomal fractions of extracts fromWT,pan1, and pan2 mutants were probed on immu-noblots with anti-phosphoamino acid antibodies.No differences between wild type and mutantwere observed with anti-phosphoserine or anti-phosphotyrosine, but a 52-kD protein (smallerthan PAN1) identified by anti-phosphothreoninewas almost undetectable in both pan1-Mu andpan1-EMS mutants (Fig. 4B). This protein wasdepleted in pan2 mutants and in WT extractstreated with lambda phosphatase (Fig. 4B andfig. S5). Thus, it appears that phosphorylation ofa 52-kD membrane protein of unknown identitydepends on PAN1, suggesting a role for PAN1 insignal transduction. Alternatively, the 52-kDphosphoprotein may depend on PAN1 andPAN2 for its synthesis or stability rather than itsphosphorylation. As demonstrated or proposedfor other LRR-RLKs with conserved but enzy-

Fig. 3. PAN1 and actinlocalization in developingstomata.White arrowheadson GMCs indicate PAN1and actin patches in ad-jacent SMCs. SMC nucleiare identified as unpolar-ized (u), partially polar-ized (pp), polarized (p),or divided (d) and arenumbered in (D) and (G)for reference. (A to C)PAN1 staining in greenand propidium iodide-stained nuclei in blue(single confocal planes).(D to L) Double stainingof PAN1 [monochrome in(D), (G), and (J); green in(F), (I), and (L)] and actin[monochrome in (E), (H),and (K); red in (F), (I),and (L)] in WT leaves (Zprojections of confocal stacks). In (D) to (I), SMCs 1, 2, and6 have polarized nuclei and both PAN1 and actin patches;the remaining SMCs have unpolarized or partially polar-ized nuclei with both PAN1 and actin patches (SMC3), aPAN1 patch only (SMC5), or neither PAN1 nor actin patches(SMC4). Scale bar, 10 mm. (M) Lengths of GMCs adjacentto SMCs with nuclei that were unpolarized (U, n = 119),polarized (P, n = 168), or divided (D, n= 36). (N) Lengthsof GMCs adjacent to SMCs with unpolarized nuclei havingneither PAN1 nor actin patches (n = 32), PAN1 patchesonly (n= 39), or both PAN1 and actin patches (n= 48). InM and N, error bars indicate SEM; all differences shown are significant (P < 0.01) by Student’s t test.

Fig. 2. Immunoblot of proteins extracted from the division zone of leaves of the indicated genotypesprobed with anti-PAN1 antibody. Extracts were separated into 110,000g supernatants (S) and pelletsrepresenting the microsomal fraction (P). Coomassie staining of the equivalent region of a duplicategel demonstrates equal loading.


REPORTS

...

......


matically inactive kinase domains (12, 13), thePAN1 kinase domain may mediate interaction(s)with other proteins that function in concert withor downstream of PAN1.

Our results implicate PAN1 as a receptor orco-receptor ofGMC-derived signals that promotethe polarization of SMCs in preparation for theirasymmetric division. Because SMC polarizationis not completely abolished in pan1 null mutants,it is clear that PAN1 is not solely responsible forthe reception of polarizing cues fromGMCs. Thus,parallel or cooperating pathways may work to-gether with the PAN1 pathway to promote SMCpolarization. The reduction in PAN1 protein levelsand lack of detectable PAN1 patches in pan2mutants suggest that PAN2 functions in the PAN1pathway, but the identity of PAN2 and its func-tional relationship to PAN1 remains to bedetermined. If PAN1 relays polarity cues fromGMCs, why is it already localized at sites ofGMC contact at its earliest appearance? PAN1may be initially distributed uniformly at the SMCsurface but detectable by immunolocalization onlyafter becoming locally concentrated as a result ofbinding to a GMC-derived ligand. Alternatively,PAN1 may be recruited directly to GMC contactsites in response to an upstream sensor of GMCposition, functioning there as a secondary sensor.The localization of PAN1 at a specific site on thecell surface, its enzymatically inactive kinasedomain, and its cellular function differentiate itfrom two other receptor-like proteins previouslyshown to play a role in polarized cell division orgrowth in plants. The Arabidopsis TMM protein,which contains LRR and transmembrane do-mains but lacks a kinase domain, functions in awell-characterized signal transduction pathwaythat controls the entry of epidermal cells into thestomatal lineage and the orientation of asym-metric, stomate-forming divisions (14). HoweverTMM does not promote the polarization of thesedivisions, nor is it asymmetrically localized(15). Tomato LRR-RLK proteins LePRK1 andLePRK2 promote the polarized growth of pollentubes (16). They have active kinase domains and

are localized uniformly at the pollen tube surfacebut appear to interact with rho of plants (ROP)guanine nucleotide exchange factors only at thetube tip to stimulate localized actin assembly andpolarized growth (17, 18).

References and Notes1. B. Scheres, P. N. Benfey, Annu. Rev. Plant Physiol. Plant

Mol. Biol. 50, 505 (1999).2. C. A. ten Hove, R. Heidstra, Curr. Opin. Plant Biol. 11, 34

(2008).3. G. L. Stebbins, S. S. Shah, Dev. Biol. 2, 477 (1960).4. S. Cho, S. M. Wick, Protoplasma 157, 154 (1990).

5. K. Gallagher, L. G. Smith, Curr. Biol. 10, 1229(2000).

6. E. Panteris, P. Apostolakos, B. Galatis, Cell Motil.Cytoskeleton 63, 696 (2006).


8. S. H. Shiu, A. B. Bleecker, Proc. Natl. Acad. Sci. U.S.A.98, 10763 (2001).

9. K. U. Torii, Int. Rev. Cytol. 234, 1 (2004).10. A. Dievart, S. E. Clark, Development 131, 251

(2004).11. S. K. Hanks, T. Hunter, FASEB J. 9, 576 (1995).12. B. Llompart et al., J. Biol. Chem. 278, 48105 (2003).13. D. Chevalier et al., Proc. Natl. Acad. Sci. U.S.A. 102,

9074 (2005).14. D. C. Bergmann, F. D. Sack, Annu. Rev. Plant Biol. 58,

163 (2007).15. J. A. Nadeau, F. D. Sack, Science 296, 1697 (2002).16. J. Muschietti, Y. Eyal, S. McCormick, Plant Cell 10, 319

(1998).17. P. Kaothien et al., Plant J. 42, 492 (2005).18. Y. Zhang, S. McCormick, Proc. Natl. Acad. Sci. U.S.A.

104, 18830 (2007).19. D. M. Friedrichsen, C. A. Joazeiro, J. Li, T. Hunter,

J. Chory, Plant Physiol. 123, 1247 (2000).20. We thank A. Wright and F. Delgado for help with

experiments and J. Bolado, J. Chory, and R. Bostonfor reagents. This work was supported by NSF grantIOB-0544226. pan1 gene and PAN1 protein sequencesare available at GenBank (accession number FJ231525).


11 June 2008; accepted 21 November 200810.1126/science.1161686

Calcineurin/NFAT Signaling IsRequired for Neuregulin-RegulatedSchwann Cell DifferentiationShih-Chu Kao,1,2 Hai Wu,1,2 Jianming Xie,1 Ching-Pin Chang,1,2 Jeffrey A. Ranish,3Isabella A. Graef,2* Gerald R. Crabtree1,2*

Schwann cells develop from multipotent neural crest cells and form myelin sheaths aroundaxons that allow rapid transmission of action potentials. Neuregulin signaling through the ErbBreceptor regulates Schwann cell development; however, the downstream pathways are not fullydefined. We find that mice lacking calcineurin B1 in the neural crest have defects inSchwann cell differentiation and myelination. Neuregulin addition to Schwann cell precursorsinitiates an increase in cytoplasmic Ca2+, which activates calcineurin and the downstreamtranscription factors NFATc3 and c4. Purification of NFAT protein complexes shows thatSox10 is an NFAT nuclear partner and synergizes with NFATc4 to activate Krox20, whichregulates genes necessary for myelination. Our studies demonstrate that calcineurin and NFAT areessential for neuregulin and ErbB signaling, neural crest diversification, and differentiation ofSchwann cells.

Rapid transmission of action potentials inthe vertebrate nervous system is achievedby insulating axons with myelin sheaths,

the membrane outgrowth from Schwann cells inthe peripheral nervous system (PNS) and oligo-dendrocytes in the central nervous system (1).Defects in myelination occur in many humandiseases; however, the mechanisms regulating

myelin formation are still not fully understood.Neuregulin-1 (NRG1) and its receptor tyrosinekinase, ErbB2/3 are required for Schwann celldevelopment (2–7). Genetic studies implicateSox10, Oct6 (SCIP or Tst1), Ets, Brn1 and 2,Krox20 (Egr2) and Nab1 and 2 in PNS myeli-nation (8–12). The expression of Sox10 is main-tained throughout Schwann cell development,

Fig. 4. AnalysisofPAN1-dependentprotein phospho-rylation. (A) In vi-tro kinase assays.Lanes 1 and 2show Coomassiestainingof thepro-teins used in eachreaction: MBP and the kinase domains of BRI1 and PAN1 fusedto glutathione S-transferase. Autoradiography in lanes 3 and 4reveals that only the BRI1 kinase domain phosphorylates itself(19) and MBP. (B) Immunoblot of microsomal fractions ofextracts from the division zones of leaves of the indicatedgenotypes was probed with anti-phosphothreonine antibody. Aduplicate blot was probed with anti-actin antibody to demonstrateequal loading.


REPORTS

...

......


whereas Oct6, Brn1/2, and Krox20 are inducedlater in immature and promyelinating Schwanncells after axonal contact (13–16). Krox20 is ofparticular interest because it coordinates the ex-pression of several myelin structural proteins.

In many cell types, Ca2+ influx activates thephosphatase calcineurin, which dephosphorylatesNFATc proteins, which in turn leads to theirnuclear entry and assembly into NFAT transcrip-tion complexes (17–20). We observed that NFATwas active in migrating mouse neural crest cellsat embryonic (E) day 9.0 (Fig. 1A) (21); there-fore, we deleted CnB1, the essential regulatorysubunit of calcineurin in neural crest, usingWnt1cre

(22, 23) (Fig. 1B). By E11.5, CnB1 protein wasalmost undetectable in dorsal root ganglia (DRG)and sympathetic ganglia (SG) (fig. S1A). The lossof CnB1 expression in DRG, SG, and Schwanncells persisted until death (within 20 hours afterbirth) (Fig. 1C and fig. S1, B and C). However,CnB1 was not deleted in axons of ventral rootsderived from motor neurons where Wnt1cre isinactive (Fig. 1C). CnB1 was deleted in vitro insensory neurons and Sox10-positive Schwanncell precursors (SCPs) (Fig. 1D and fig. S1D).NFATc3 and c4 were hyperphosphorylated, whichindicated loss of calcineurin phosphatase activity(Fig. 1E). DRG architecture, cell proliferation, andcell death were not changed in the CnB1 mutantembryos (fig. S2A), and peripheral nerve pro-jections were comparable to those of controls(fig. S2B). However, myelination of mutant sci-atic nerves was defective, and fewer axons weresorted into a 1:1 ratio with Schwann cells (Fig. 2,A and B, and fig. S3, A and B). Mutant nervesalso had a higher g ratio (axon diameter to totalmyelinated fiber diameter) (fig. S3C). They ex-pressed less of the promyelinating proteinKrox20,the early myelin protein MAG, and major com-pacted myelin components such as MBP and P0(myelin basic protein and myelin protein zero,respectively) (Fig. 2, C and D). Furthermore,NFATc3 and c4 were hyperphosphorylated inmutant Schwann cells (Fig. 2C). These obser-vations indicate that mice lacking CnB1 havedefective myelination.

Investigation of the Schwann cell lineage re-vealed that SCPs were found in both control andmutant peripheral nerves at E11.5 (fig. S4A). AtE13.5, control andmutant DRG contained similarnumbers of sensory neurons and SCPs (fig. S4B).Unlike neuregulin or erbB mutant mice (2, 5),SCPs from CnB1 mutant embryos showed nodifferences in proliferation or apoptosis in vitro inresponse to NRG1 stimulation (fig. S4, C and D).We observed comparable numbers of proliferat-ing Schwann cells in control andmutant newbornsciatic nerves (fig. S5A).

Survival of SCPs has been shown to be de-pendent on trophic support from sensory neurons(24). To rule out the possibility that hypomyeli-nation was due to dysfunction of mutant sensoryneurons, we coculturedmutant SCPswith controlsensory neurons under conditions that supportedsensory neuron survival. Fewer MBP-positiveSchwann cells were found in mutant SCP cocul-tures, although comparable numbers of sensoryneurons were present in both control and mutantcocultures (fig. S5, B to D). ErbB2 and 3 expres-sion and phosphorylation levels were normal inmutant SCPs, and sensory neurons expressed com-parable amounts of pro-NRG1 and the cleavedform of NRG1 (NTF), which suggested thatBACE1, an enzyme involved in NRG1 process-ing, functioned normally in mutant sensory neu-rons (fig. S5E).

To investigate cell autonomy of the myelina-tion defects, we took advantage of selective dele-tion of CnB1 in Schwann cells, but not in motorneurons of CnB1D/f; Wnt1cre mice and found thataxonal sorting was reduced inmutant ventral rootsand phrenic nerves (Fig. 2, E and F, and fig. S6A).This was similar to the defect seen in dorsal rootsfrom CnB1D/f; Wnt1cre mutants, where CnB1 isdeleted in both sensory neurons and Schwanncells (fig. S6B). In a second approach, we exam-ined CnB1f/f; Nestincre mice, where CnB1 is de-leted in sensory neurons, but not in Schwanncells (fig. S7, A and B). CnB1 deletion did notaffect the numbers of MBP-positive Schwanncells or axonal sorting (fig. S7, C to F). These twolines of evidence indicate that the myelinationdefects inCnB1mutant mice are due to a Schwanncell–autonomous mechanism.

In our analysis of signaling pathways that ac-tivated calcineurin/NFAT in DRG cocultures, wefound that NRG1 induced phospholipase C–g(PLC-g)–dependent Ca2+ influx in SCPs (Fig. 3A)and CnB1-dependent dephosphorylation ofNFATc3 and c4 (Fig. 3B). NRG1 also stimulatedCa2+ influx in sensory neurons (fig. S8A). CnB1deletion had no effect on NRG1-induced phos-phorylation of the ErbB3 receptor, which indi-cated that calcineurin functions downstream ofErbB (Fig. 3C). ErbB kinase inhibitors specifi-cally blocked NRG1-stimulated NFATc3 and c4dephosphorylation, whereas neither phosphati-dylinositol 3-kinase nor mitogen-activated proteinkinase kinase inhibitors had an inhibitory effect(fig. S8, B and C).

NRG1not only induced changes inNFATcphos-phorylation, but also activated NFAT-dependenttranscription, measured with two NFAT reporters(25, 26) (Fig. 3D). NRG1-stimulated reporter ac-tivitywas not detected inmutant cocultures, whichindicated that calcineurin functions downstreamof ErbB and upstream of NFAT. The observationthat Krox20 expression was reduced in CnB1mutant Schwann cells led us to examine the roleof NFAT in Krox20 regulation. We found thatboth the promoter and the myelin-specific en-hancer (MSE, +35 kilobases) (27) required cal-cineurin for their activity (Fig. 3D). Three NFATcconsensus sites were present in the MSE thatbound NFAT complexes (fig. S9, A and B). Be-cause Krox20 is expressed selectively in Schwanncells, but not in DRG neurons (16), these resultsprovide a third line of evidence that calcineurin/NFAT functions in Schwann cells to directly pro-mote myelination.

1HowardHughesMedical Institute, StanfordUniversity, Stanford,CA 94305, USA. 2Department of Pathology and Department ofDevelopmental Biology, StanfordUniversity, Stanford, CA 94305,USA. 3Institute for Systems Biology, Seattle, WA 98103, USA.

*To whom correspondence should be addressed. E-mail:[email protected] or [email protected]

D E

X-g

al

Rosa26; Wnt1creDSCR-lacZ

B C

CnB1wt/f

E10.5

Cn

B1

C

nB

1/β

III-

Tub

E9.0

30µm 800µm

CnB1∆/f; Wnt1cre

300µm

A

P0

Fig. 1. CnB1 is deleted in the PNS byWnt1cre. (A) X-gal staining of E9.0 neural tube from a NFAT-reportermouse line harboring LacZ under control of the DSCR1 enhancer region counterstained with nuclear fastred. Arrows, migrating neural crest cells; arrowheads, interneuron progenitors. (B) Whole-mount X-galstaining of E10.5 Rosa26; Wnt1cre mouse embryo. Arrows, sensory ganglia. (C) Immunostaining ofnewborn DRG and peripheral nerves at thoracic limb region by using CnB1-specific antibody. Arrows,CnB1-positive motor axons of peripheral nerves; arrowheads, ventral roots. (D and E) Immunoblots ofCnB1 and NFATc3 and c4 from E13.5 DRG cocultures at day 0 (D) and day 12 (E). Heat shock protein 90(Hsp90) is a loading control. Arrows, NFATc4 proteins. *Cross-reactive protein.


REPORTS

...

......


The specificity of NFAT complexes on targetgenes arises from assembly of NFATc familymem-bers with nuclear partner proteins (18–20). Toidentify nuclear partners, we purified NFATcom-plexes from E11.5 mouse neural tubes (as suf-ficient quantities of SCPs were not available)using oligonucleotide affinity columns.Mass spec-trometry analysis identified the B group of the Sox

family proteins (Sox1, 2, 3, 14, and 21) (fig. S10,A and B). Because Sox10 has a similar bindingsequence and is expressed throughout Schwanncell development (10), we tested whether Sox10interacts with NFATc using a Schwann cell lineexpressing NFATc4 and Sox10. Sox10 copuri-fied with NFATc4 using control but not mutantoligonucleotides containing the NFATc-binding

sites of the DSCR1 (Down syndrome criticalregion 1) enhancer or Krox20 MSE (NRE4-s)(figs. S10C and S13). Because these oligo-nucleotides do not contain Sox10-binding sites,the isolation of Sox10 by control, but not mu-tant, oligonucleotides supports the direct in-teraction of NFATc4 and Sox10. Proteindomain deletion studies indicated that Sox10

A

D

sciatic nerves

ventral roots

F

C

B

CnB1∆/f; Wnt1creCnB1wt/f

**

a

10µm

b

**

P0

E

MBP/DAPI

* *

3µm

*

CnB1∆/f; Wnt1creCnB1wt/f

80µm

*

*

Fig. 2. Schwann cell differentiation is defective inCnB1 mutant mice. (A) Schwann cells in newbornmutant sciatic nerves fail to establish one-to-onerelations with axons. Low-power electron microscopicimages showing overall structure of sciatic nerves.Asterisks, sorted axons; arrows, unsorted axon bundles.(B and F) Quantification of axonal sorting (number ofaxons in bundles) of sciatic nerves and ventral motorroots of spinal cord (n = 4 per group; error bars, +SEM;P < 0.001; two-tailed t test). (C) Immunoblots ofnewborn sciatic nerves showing expression levels ofmyelin-specific proteins. Results from two differentlitters of animals are shown. Arrows, NFATc3 and c4proteins. *Cross-reactive protein. Hsp90 is a loadingcontrol. (D) Immunostaining of cryosections ofnewborn sciatic nerves using MBP-specific antibody(green). (E) Electron microscopic images of ventralmotor roots. Asterisks, sorted axons; arrows, unsortedaxon bundles. All experiments were performed with atleast four littermate pairs of mice.

Fig. 3. Calcineurin/NFAT is required for NRG1 and ErbBsignaling. (A) Ca2+ influx stimulated by NRG1. DRGcocultures were preloaded with the Ca2+ sensor Fura-2AM in the presence or absence of PLC-g inhibitorU73122, 30 min before stimulation. Ca2+ influx wasdetected, and average Ca2+ levels were measured. (B)Dephosphorylation of NFATc3 and c4 is not induced inmutant cocultures after NRG1 stimulation, as examinedby immunoblotting (ib). Arrows, dephosphorylated(active) NFATc3 and c4. (C) Tyrosine phosphorylationlevels of ErbB3 following the activation by NRG1. ErbB3were isolated by immunoprecipitation (ip). Tyrosinephosphorylation was detected by immunoblotting usingphosphotyrosine-specific antibody. The blot wasreprobed with ErbB3-specific antibody to verify proteinexpression levels. (D) Calcineurin/NFAT-dependenttranscription is stimulated by NRG1. E13.5 DRGcocultures were transfected with luciferase reporterconstructs containing the DSCR1 enhancer, NFATc4promoter, Krox20 promoter, or Krox20 MSE. Luciferaseactivities were measured after 20 hours of NRG1stimulation. Fold activation is relative to unstimulatedcells (n = 6; error bars, +SEM; **P < 0.01; ***P < 0.001;t test).

ctrl ctrl ctrlmut mut mutDSCR1

enhancer

D DSCR1 enhancer (1kb)

NFATc4 promoter (3kb)

Luciferase

Luciferase

Krox20 promoter (1.3kb)

Krox20 Myelin Specific Enhancer (1.3kb)

Luciferase

Luciferase

ctrl mutctrl mutKrox20

promoterKrox20MSE

B

Fol

d S

timul

ated

by

NR

G1

Time (min)

A

C

NFATc4promoter

pGL3

Ave

rage

Ca2

+ L

evel

s


REPORTS

...

......


bound to the Rel homology domain of NFATc4(fig. S10D).

Because several conserved NFATc-bindingsites lie next to putative Sox10-binding sites inthe Krox20 MSE (fig. S9A), we tested coactiva-tion of the Krox20 MSE by NFATc4 and Sox10(Fig. 4A). NFATc4 and Sox10 synergistically ac-tivatedKrox20MSE-driven luciferase expressionin DRG cocultures. Brn2 and Oct6 are also im-portant regulators of Krox20 transcription (28).The functional interaction between calcineurin/NFAT and Sox10 might provide NRG1 respon-siveness to the Krox20 MSE and might assistBrn2 and Oct6 in its regulation.

Synergy between NFATc4 and Sox10 wasalso observed with the P0 promoter (Fig. 4A andfig. S11) but not with the Krox20 or NFATc4promoters (Fig. 4A and fig. S12B). Sox10 andKrox20 have been shown to regulate the P0

intron1 (29). We found that calcineurin/NFATalso stimulated the P0 intron1 (fig. S12). Thus,calcineurin/NFAT signaling likely regulates P0

expression directly through the P0 promoter bycooperating with Sox10 and indirectly throughstimulating intron1 activity by inducing Krox20expression. Sox2 blocked transcriptional acti-vation (Fig. 4A), consistent with the observationthat it is an inhibitor of Schwann cell differen-tiation (30). Using oligonucleotide affinity pu-rification, we found that NFATc4 facilitatedSox10 binding to the NRE4 region of Krox20MSE (NRE4-L), because mutation of the NFATc-

binding sequence led to reduced cobinding ofSox10 (fig. S13). Together, these data indicatecooperation between NFATc4 and Sox10 in my-elin gene expression.

Our data indicate that calcineurin/NFAT sig-naling regulates the development of Schwanncells at two steps (Fig. 4B). One is by actingdownstream of neuregulin and ErbB to directlyactivate Krox20 at the promyelination stage,which in turn activates later myelin genes.Another is by activating genes encoding myelinstructural proteins, such as P0, at the myelinationstage. NFATc4 interacts with Sox10 and providesspecificity to the calcineurin/NFAT pathway forthe Schwann cell lineage.

References and Notes1. K. R. Jessen, R. Mirsky, Nat. Rev. Neurosci. 6, 671

(2005).2. D. Meyer, C. Birchmeier, Nature 378, 386 (1995).3. Z. Dong et al., Neuron 15, 585 (1995).4. K. A. Nave, J. L. Salzer, Curr. Opin. Neurobiol. 16, 492

(2006).5. M. T. Woldeyesus et al., Genes Dev. 13, 2538 (1999).6. P. Maurel, J. L. Salzer, J. Neurosci. 20, 4635 (2000).7. M. Rahmatullah et al., Mol. Cell. Biol. 18, 6245 (1998).8. S. Britsch et al., Genes Dev. 15, 66 (2001).9. C. Paratore, D. E. Goerich, U. Suter, M. Wegner,

L. Sommer, Development 128, 3949 (2001).10. D. W. McCauley, M. Bronner-Fraser, Nature 441, 750

(2006).11. D. B. Parkinson, K. Langner, S. S. Namini, K. R. Jessen,

R. Mirsky, Mol. Cell. Neurosci. 20, 154 (2002).12. R. Srinivasan et al., BMC Mol. Biol. 8, 117 (2007).13. R. P. Friedrich, B. Schlierf, E. R. Tamm, M. R. Bosl,

M. Wegner, Mol. Cell. Biol. 25, 1821 (2005).

14. M. Jaegle et al., Genes Dev. 17, 1380 (2003).15. M. Jaegle et al., Science 273, 507 (1996).16. P. Topilko et al., Nature 371, 796 (1994).17. J. P. Shaw et al., Science 241, 202 (1988).18. W. M. Flanagan, B. Corthesy, R. J. Bram, G. R. Crabtree,

Nature 352, 803 (1991).19. H. Wu, A. Peisley, I. A. Graef, G. R. Crabtree, Trends Cell

Biol. 17, 251 (2007).20. J. Jain, P. G. McCaffrey, V. E. Valge-Archer, A. Rao,

Nature 356, 801 (1992).21. Materials and methods are available as supporting

material on Science Online.22. J. R. Neilson, M. M. Winslow, E. M. Hur, G. R. Crabtree,

Immunity 20, 255 (2004).23. P. S. Danielian, D. Muccino, D. H. Rowitch, S. K. Michael,

A. P. McMahon, Curr. Biol. 8, 1323 (1998).24. K. R. Jessen et al., Neuron 12, 509 (1994).25. J. R. Arron et al., Nature 441, 595 (2006).26. H. Wu et al., J. Biol. Chem. 282, 30673 (2007).27. J. Ghislain et al., Development 129, 155 (2002).28. J. Ghislain, P. Charnay, EMBO Rep. 7, 52 (2006).29. S. E. LeBlanc, S. W. Jang, R. M. Ward, L. Wrabetz,

J. Svaren, J. Biol. Chem. 281, 5453 (2006).30. N. Le et al., Proc. Natl. Acad. Sci. U.S.A. 102, 2596

(2005).31. We thank B. A. Borres for insightful comments; K. J. Liu,

Q. Lin, and J. Perrino for technical assistance; andJ. I. Wu for critically reading this manuscript. G.R.Cis an Investigator of HHMI. These studies were supportedby grants from NIH (NS046789, HD55391, and AI60037)to G.R.C.



Fig. 4. Synergisticeffects of NFATc4and Sox10 on my-elin gene expres-sion. (A) NFATc4andSox10synergis-tically activate theKrox20MSE and P0promoter. DRGcocultures weretransfected with lu-ciferase reporterconstructs contain-ing the Krox20MSE, P0 promoter

A

B

NFATc4 NFATc4

P

CnA B

Ca /Calmodulin

Neural Crest Cells

Schwann cellPrecursors

ImmatureSchwann cells

MyelinatingSchwann cells

axon

ErbB2ErbB3

2+PLCγ

PI-3KMAPK

ProliferationSurvival

NFATc4

Krox20 Myelin Specific Enhancer (1.3kb)

Luciferase

CnA*/c4Sox

--

- - -++ + +Sox10 Sox2 Sox21

Fol

d A

ctiv

atio

n

NFATc4

Sox10

Krox20 promoter (1.3kb)

Luciferase

CnA*/c4Sox

--

- - -++ + +Sox10 Sox2 Sox21

Krox20, P0

P0 promoter (1.1kb)

Luciferase

CnA*/c4Sox

--

- - -++ + +Sox10 Sox2 Sox21

Sox10

NRG1or Krox20 promoter. Expression constructs encoding constitutively active calcineurinA (CnA*), NFATc4, and Sox family members were co-transfected. Luciferase activitieswere measured. Fold activation is relative to reporter baseline activities (n = 6, errorbars, +SEM, ***P < 0.001; t test). (B) Model of myelination process regulated bycalcineurin/NFAT signaling pathway. Solid arrows, calcineurin/NFAT signalingpathway; dotted arrows, pathways defined by previous literature.


REPORTS

...

......


New Productslife science technologiesAAAS/Science Business Office

Newly offered instrumentation, apparatus, and laboratory materials of interest to researchers in all disciplines in academic, industrial, and governmental organizations are featured in this space. Emphasis is given to purpose, chief characteristics, and availabilty of products and materials. Endorsement by Science or AAAS of any products or materials mentioned is not implied. Additional information may be obtained from the manufacturer or supplier.

Electronically submit your new product description or product literature information! Go to www.sciencemag.org/products/newproducts.dtl for more information.

655 www.sciencemag.org/products science VOL 323 30 JANUARY 2009

Portable Light BoxThe ViewOne is a portable light box designed to fit into limited labo-ratory space. Measuring 20.5 cm by 14 cm by 0.6 cm and weighing 160 g, it is slightly larger than a passport. The ViewOne LED provides a consistent white light source for hours of dependable operation. It has three selectable light intensities and fixed retainers molded in to hold petri dishes and enzyme-linked immunosorbent assay plates in place. Its smooth, water-resistant surface makes it easy to wipe down and clean. Embi Tec For information 858-684-3190www.embitec.com

Pipetting SystemAn integrated pipetting system with epT.I.P.S. LoRetention features “Pearl Effect” technology to ensure maximum sample recovery when dispensing detergent solutions. The low-retention tips are designed for users engaged in polymerase chain reaction (PCR), real-time PCR, proteomics, and molecular biology applications. When liquids that contain detergents wet polypropylene, they leave a film on a standard tip’s surface that can lead to loss of precious sample, increased consumption of reagents, and poor precision and reproducibility. The “Pearl Effect” technology involves no coating or additive that might affect or bleed into the sample, but instead renders the tip surface ultrahydrophobic through a modification at the molecular level so detergent solutions roll off completely. Eppendorf For information +49-40-53801-0wwww.eppendorf.com

Electronic Laboratory NotebookThe Symyx Notebook is an electronic laboratory notebook (ELN) that is configurable to meet the needs of biologists and analytical chemists. It can simplify operations and lower costs by enabling research and development organizations to replace multiple discipline-specific ELNs with a single, multidiscipline application that is deployable across the enterprise. Built on an enterprise-scale

informatics platform, Symyx Notebook consolidates experimental data from multiple domains into shareable and searchable documents controlled by customizable document workflows with secure document versioning, electronic signature, and audit trails. Symyx Technologies For information +41-22-884-1331www.symyx.com

Protein AssayCB-X Protein Assay is designed to be compatible with all commonly used buffers and reagents. For samples in simple aqueous buffers, CB-X is a sensitive, single-reagent assay that can be performed in five minutes. For complicated protein samples, CB-X is supplied with reagents to clean up the samples and remove all reagents and chemicals that interfere with accurate protein estimation. These reagents include detergents, chaotropes, reducing agents, alkylating agents, sugars, high salt concentrations, buffering agents, and chelating agents. The cleanup stage and subsequent protein assay is performed in a single tube to ensure no protein loss and to maintain the accuracy of the assay. G-Biosciences For information 314-991-6034www.GBiosciences.com

Antigen StandardsVerify Antigen Standard is a collection of tagged overexpression cell lysates for use as protein immunoblot standards. The line covers 5,000 overexpressed human proteins, each with a built-in C-terminal Myc/DDK tag for easy expression verification and antibody validation. These initial 5,000 Verify Antigen Standards will not only provide scientists with easy access to a validated antigen of interest, they can also be easily quantified with an Myc/DDK tag standard to provide definitive detection sensitivity of a particular antibody. OriGene Technologies For information 240-620-0270www.origene.com

Robotic PlatformAutoMAP is a robotic platform for high throughput microplate, tube, and vial processing in a self-contained environment for a variety of drug discovery applications. The platform can be operated as a stand-alone system or integrated with additional platforms, including the MiniStore (an automated sample management system) or MACCS (an automated cell culture system). Temperature, humidity, carbon dioxide, nitrogen, or other gases or gas mixtures can be regulated internally, ensuring optimized experimental conditions. It offers a biological safety cabinet level 1 classification, minimizing potential hazards to laboratory personnel and the environment by isolating dangerous biological agents in an enclosed facility. Its list of applications is almost limitless, including chemical and biological sample management, polymerase chain reaction purification and preparation, cycle sequencing preparation, sample normalization, chemical assays, plate replication, cell-based assays, and more.

MatriCal For information 509-343-6244

www.matrical.com

New Products

...

......


science 2009-01-30

Documents

van der marel

parallelresistor formalism 1 rab

cole normale suprieure

fuels anoxygenic photosynthesis

premature stop codon

max planck inst

supporting online materialwww

pandemic flu vaccine