Science - October 14th 2011 _True PDF_

Document Sample
Science - October 14th 2011 _True PDF_ Powered By Docstoc
					14 October 2011 | $10
The BD Accuri® C6
personal flow cytometer

Flow cytometry within reach.®
                                   The BD Accuri® C6 is a personal flow cytometer                Setup and maintenance are also simplified to
                                   that puts flow cytometry within reach. It gives you           increase availability and up-time. In the event service
                                   4-color cell analysis in an affordable, transportable,       is required, just pack up the system and return it.
                                   and easy-to-use format.                                      A loaner system will be provided while your system
                                   Measuring just 11 x 14.75 x 16.5 inches (27.9 x              is serviced.
                                   37.5 x 41.9 centimeters) and weighing just 30                For more information about how you can more
                                   pounds (13.6 kilograms), the BD Accuri C6 gives              easily access the power of flow cytometry in your
                                   both novice and experienced researchers the power            lab, visit
                                   of multicolor analysis, when and where you need
                                   it. BD CFlow® software’s intuitive interface guides          Flow cytometry within reach.®
                                   you through workflows, making it easy to begin
                                   collecting and analyzing data—even if you have
                                   little flow cytometry knowledge or know-how.

BD flow cytometers are Class 1 Laser Products.                                                                                 BD Biosciences
For Research Use Only. Not for use in diagnostic or therapeutic procedures.                                                   2350 Qume Drive
BD, BD Logo and all other trademarks are property of Becton, Dickinson and Company. © 2011 BD                                 San Jose, CA 95131
here's a good reason
to smile
You can depend on NuPAGE® gels to deliver straight, sharp bands
Life Technologies offers you a reason to smile with Novex® NuPAGE® precast gels for protein
separation. Accurate and easy separation is at your fingertips. Enjoy the convenience of
consistent, superior performance.

                  Go to
                  Get the free mobile app for your phone at

©2011 Life Technologies Corporation. All rights reserved. The trademarks mentioned herein are the property of Life Technologies Corporation
or their respective owners. For research use only. Not intended for any animal or human therapeutic or diagnostic use. CO19167 0511
GE Healthcare
Life Sciences

Cytiva Cardiomyocytes
Relevant. Reliable. Confident.
A new era in safety screening with human cell models.
Cytiva Cardiomyocytes from GE Healthcare Life Sciences are derived from human
embryonic stem (hES) cells and provide a biologically-relevant alternative to current
cell models and primary cells, for predictive toxicity testing.

•   Highly characterized
•   Functionally verified
•   Representative myocyte population
•   Karyotype verified
•   Scalable
•   Cryopreserved for ease of use

               imagination at work                                                  GE, imagination at work, and GE monogram are trademarks of General Electric Company.
                                                                                    Cytiva is a trademark of GE Healthcare Companies.
                                                                                    © 2011 General Electric Company—All rights reserved. First published Aug. 2011.
                                                                                    GE Healthcare Bio-Sciences AB, Björkgatan 30, SE-751 84 Uppsala, Sweden.

CONTENTS                                                                                                                Volume 334 Issue 6053

EDITORIAL                                              BOOKS ET AL.
157  Peruvian Highlands, Fume-Free                     178 Redirect
      Pilar Nores Bodereau                                    T. D. Wilson, reviewed by G. L. Cohen
      >> Policy Forum p. 180                           179    Browsings
NEWS OF THE WEEK                                       POLICY FORUM
160 A roundup of the week’s top stories                180  A Major Environmental Cause of Death
                                                              W. J. Martin II et al.
NEWS & ANALYSIS                                               >> Editorial p. 157
163 Panel Draws Ambitious Road Map
    for Gulf Restoration                               PERSPECTIVES
164   Teasing Out Cause and Effect                     182  Keeping Bacteria at a Distance
      in Macroeconomics                                       M. E. V. Johansson and G. C. Hansson
165   Once-Ridiculed Discovery Redefined                       >> Report p. 255
                                                                                                                        page 173
      the Term Crystal                                 183    Self-Assembly Enters the Design Era
166   HIV Study Renews Scrutiny of                            A. Travesset
      Hormonal Contraception                                  >> Report p. 204

167   African Data Bolster New View                    184    The Costs of Breathing
      of Modern Human Origins                                 N. Lane

168   Gut Bacteria Lend a Molecular Hand               186    SevERing Mitochondria
      to Viruses                                              A. S. Rambold and J. Lippincott-Schwartz
      >> Reports pp. 245 and 249                       187    Shining Light on Diabolic Points
                                                              B. J. Whitaker
NEWS FOCUS                                                    >> Report p. 208
169 Beyond the Data                                    188    Grass Trumps Trees with Fire
      The Community Weighs In on Broader Impacts              A. L. Mayer and A. H. Khalyani
      >> Science Podcast                                      >> Reports pp. 230 and 232
173   Vital Details of Global Warming
      Are Eluding Forecasters                          REVIEW
                                                       190    The Diets of Early Hominins
LETTERS                                                       P. S. Ungar and M. Sponheimer
176  Partial Retraction
      R. H. Silverman et al.                           BREVIA
      Chemical Elements: What’s in a Name?             194    Replication-Dependent Loss of
      J. P. Leal                                              5-Hydroxymethylcytosine in
                                                              Mouse Preimplantation Embryos
      Biosecurity and the Politics of Fear
                                                              A. Inoue and Y. Zhang
      P. E. Hulme                                             The oxidation product of methylated cytosine              page 178
177   TECHNICAL COMMENT ABSTRACTS                             is passively lost from DNA in the zygote
                                                              as cell division progresses.

                                                              CONTENTS continued >>

                                         COVER                                                                          DEPARTMENTS
                                         Whiskers at the snout are instrumental for the rat to explore                  153   This Week in Science
                                         the external world. During development, sensory information                    158   Editors’ Choice
                                         provided by the whiskers is critical for the formation of their                159   Science Staff
                                         representation in the brain. Minlebaev et al. show how this                    259   New Products
                                         process is controlled by gamma oscillations in developing                      260   Science Careers
                                         neuronal networks. See page 226.
                                         Photo: Henrik Sorensen/Getty Images

                                         SCIENCE           VOL 334        14 OCTOBER 2011                                149

                                     RESEARCH ARTICLE                                      226   Early Gamma Oscillations Synchronize
                                                                                                 Developing Thalamus and Cortex
                                     195   Synthesized Light Transients                          M. Minlebaev et al.
                                           A. Wirth et al.
                                                                                                 Thalamic gamma rhythms help develop
                                           Light spanning the near infrared to the               highly spatially and laminar-specific
                                           ultraviolet has been confined in pulses                ascending cortical projections.
                                           shorter than a single optical cycle.
                                                                                           230   The Global Extent and Determinants
                                     REPORTS                                                     of Savanna and Forest as Alternative
                                                                                                 Biome States
                                     200   Observation of Correlated Particle-Hole               A. C. Staver et al.
                                           Pairs and String Order in Low-Dimensional             Savanna and forest are alternative states
                                           Mott Insulators                                       governed by fire at intermediate rainfall
                                           M. Endres et al.                                      levels.
           pages 182 & 255                 Parity correlations in a one-dimensional
                                           Bose gas in an optical lattice reveal a
                                                                                           232   Global Resilience of Tropical Forest
                                           hidden “string order.”                                and Savanna to Critical Transitions
                                                                                                 M. Hirota et al.
                                     204   Nanoparticle Superlattice Engineering                 Tree distributions across continents indicate
                                           with DNA                                              three distinct stable states in tree cover—
                                           R. J. Macfarlane et al.                               forest, savanna, and treeless.
                                           Design rules allow the synthesis of                   >> Perspective p. 188
                                           nanoparticle-DNA superlattices in
           pages 183 & 204                 nine different lattices.                        235   The Escherichia coli Replisome Is
                                           >> Perspective p. 183                                 Inherently DNA Damage Tolerant
                                                                                                 J. T. P. Yeeles and K. J. Marians
                                     208   Conical Intersection Dynamics in NO2                  The core DNA replication machinery of
                                           Probed by Homodyne High-Harmonic                      Escherichia coli has an inherent ability
                                           Spectroscopy                                          to squeeze past DNA damage on the
                                           H. J. Wörner et al.                                   leading strand.
                                           Coincident vibrational and electronic
                                                                                           238   Sequential Establishment of Stripe
                                           rearrangements in a photoexcited
                                           molecule are tracked in fine detail.
                                                                                                 Patterns in an Expanding Cell Population
                                                                                                 C. Liu et al.
                                           >> Perspective p. 187
                                                                                                 A synthetic circuit implementing density-
                                     213   Linear Alkane Polymerization on                       controlled bacterial motility autonomously
                                           a Gold Surface                                        produces a tunable stripe pattern.
                                           D. Zhong et al.                                 242   NMR Detection of Structures in the
                                           The confining channel geometry of a
                                                                                                 HIV-1 5′-Leader RNA That Regulate
                                           gold surface induces selective end-to-end
                                           linking of hydrocarbon chains.
                                                                                                 Genome Packaging
                                                                                                 K. Lu et al.
                                     216   Flash Heating Leads to Low Frictional                 An RNA structural switch regulates whether
                                           Strength of Crustal Rocks at Earthquake               the HIV genome is translated or dimerized
                                           Slip Rates                                            and packaged.
                 page 219                  D. L. Goldsby and T. E. Tullis                  245   Successful Transmission of a Retrovirus
                                           Extreme temperatures generated over
                                                                                                 Depends on the Commensal Microbiota
                                           short distances may weaken faults during
                                                                                                 M. Kane et al.
                                                                                           249   Intestinal Microbiota Promote Enteric Virus
                                     219   A 100,000-Year-Old Ochre-Processing
                                                                                                 Replication and Systemic Pathogenesis
                                           Workshop at Blombos Cave, South Africa                S. K. Kuss et al.
                                           C. S. Henshilwood et al.
                                                                                                 Commensal microflora promote the pathogens
                                           Early humans mixed and stored ochre
                                                                                                 of mucosally acquired viruses.
                                           pigments in shells 100,000 years ago,
                                                                                                 >> News story p. 168
                                           an indication of the emergence of
                                           higher planning.                                252   MED12, the Mediator Complex Subunit
                                           >> Science Podcast                                    12 Gene, Is Mutated at High Frequency
                                     222   The Dynamic Architecture of Hox                       in Uterine Leiomyomas
                                                                                                 N. Mäkinen et al.
                                           Gene Clusters
                                           D. Noordermeer et al.                                 Uterine fibroids frequently harbor mutations
                                                                                                 in a specific gene that has been implicated
                                           Sequential activation of Hox genes correlates
                                                                                                 in transcriptional regulation.
                                           with a transition of negative to positive
                                           three-dimensional chromosome structure.         255   The Antibacterial Lectin RegIIIγ Promotes
                                                                                                 the Spatial Segregation of Microbiota and
                                                                                                 Host in the Intestine
                                                                                                 S. Vaishnava et al.
                                                                                                 Innate immune signaling and antimicrobial
                                                                                                 peptide activity maintain separation of the
                                                                                                 microbiota and intestinal epithelium.
                                                                                                 >> Perspective p. 182

150                          14 OCTOBER 2011     VOL 334       SCIENCE 

SCIENCEXPRESS                                          SCIENCENOW                                              RESEARCH ARTICLE: Human TH17 Cells                                                              Are Long-Lived Effector Memory Cells
A Potent and Broad Neutralizing Antibody               Highlights From Our Daily News Coverage                 I. Kryczek et al.
                                                                                                               Human TH17 cells function as long-lived effector
Recognizes and Penetrates the HIV Glycan Shield        Snails Ship Out on Scrambled Eggs
                                                                                                               memory cells in the context of chronic disease.
R. Pejchal et al.                                      Bobbing invertebrates build their mucus life rafts
An HIV antibody achieves potency and breadth           from ancestral reproductive structures.                 RESEARCH ARTICLE: Diabetes Impairs
by binding simultaneously to two conserved                                  Hematopoietic Stem Cell Mobilization
glycans on the viral envelope protein.                                                                         by Altering Niche Function
                                                       Sex-Crazed Astrologer Was a Stellar
10.1126/science.1213256                                                                                        F. Ferraro et al.
                                                       Records Keeper
Correction of Sickle Cell Disease in Adult Mice        17th-century medical files are most extensive            Impaired mobilization of hematopoietic stem cells
                                                       from that period.                                       in diabetic mice is due to sympathetic nervous
by Interference with Fetal Hemoglobin Silencing
                                                                           system dysregulation of CXCL12 distribution.
J. Xu et al.
Manipulation of a transcriptional repressor promotes   Deciphering the Brain’s Autofocus Mechanism             RESEARCH ARTICLE: Vitamin D Is Required for
expression of protective fetal globin genes.           This finding could improve eye surgery, digital          IFN-λ–Mediated Antimicrobial Activity of Human
10.1126/science.1211053                                cameras.                                                Macrophages
Degradation of Paternal Mitochondria                                          M. Fabri et al.
                                                                                                               Vitamin D is required for both innate and adaptive
by Fertilization-Triggered Autophagy
in C. elegans Embryos                                  SCIENCESIGNALING                                        immunity to tuberculosis.
M. Sato and K. Sato                          
                                                       The Signal Transduction Knowledge Environment           SCIENCECAREERS
Maternal inheritance of mitochondrial DNA results
from autophagy-dependent clearance of paternal         11 October issue:     
mitochondria.                                                                                                  Free Career Resources for Scientists
                                                       RESEARCH ARTICLE: Load-Induced Modulation
                                                       of Signal Transduction Networks                         Careers for Scientists in the Patenting World
Torsional Carbon Nanotube Artificial Muscles            P. Jiang et al.                                         E. Pain
J. Foroughi et al.                                     The presence of downstream partners that interact       Some scientists who leave the research bench end up
Carbon nanotube yarns are used to make fast,           with enzymatically modified or unmodified signaling       in patent-related careers, including patent law.
multirotational torsional actuators.                   proteins changes the dynamics of signal transduction.
10.1126/science.1211220                                RESEARCH ARTICLE: Phosphorylation of                    In Person: A Career in Biotech Patent Law
>> Science Podcast                                                                                             W. J. Simmons
                                                       Mad Controls Competition Between Wingless
                                                                                                               A former immunologist discusses his transition into
Twin Matter Waves for Interferometry                   and BMP Signaling                                       a patent law career in biotech.
Beyond the Classical Limit                             E. Eivers et al.                              
B. Lücke et al.                                        The transcription factor Mad participates in two
An entangled state of up to 10,000 atoms is used to    different developmentally important signaling           Seeding Scientists
enhance the resolution of an atomic interferometer.    pathways depending on its phosphorylation status.       M. Price
10.1126/science.1208798                                                                                        Companies are building academic partnerships to
                                                       PODCAST: Science Signaling Podcast—                     train their future scientific employees.
                                                       11 October 2011                               
TECHNICALCOMMENTS                                      E. M. De Robertis and A. M. VanHook
Comment on “Changes in Climatic Water                  Phosphorylation of the transcription factor             SCIENCEPODCAST
Balance Drive Downhill Shifts in                       Mad determines whether it mediates Wingless
                                                       or BMP signaling.
Plant Species’ Optimum Elevations”                                                                             Free Weekly Show
A. Wolf and W. R. L. Anderegg                          PERSPECTIVE: De-AMPylation Unmasked—                    On the 14 October Science Podcast: Middle Stone Age
Full text at           Modulation of Host Membrane Trafficking                  artisans, NSF’s “broader impacts” criteria, carbon
full/334/6053/177-a                                    H. Ham and K. Orth                                      nanotube muscles, and more.
Comment on “Changes in Climatic Water                  The reversible modification of a protein that
Balance Drive Downhill Shifts in                       regulates membrane trafficking promotes                  SCIENCE (ISSN 0036-8075) is published weekly on Friday, except the last
                                                       the replication of a bacterial pathogen.                week in December, by the American Association for the Advancement of
Plant Species’ Optimum Elevations”                                                                             Science, 1200 New York Avenue, NW, Washington, DC 20005. Periodicals Mail
R. J. Hijmans                                                                                                  postage (publication No. 484460) paid at Washington, DC, and additional mailing
Full text at           SCIENCETRANSLATIONAL MEDICINE                           offices. Copyright © 2011 by the American Association for the Advancement of
                                                                                                               Science. The title SCIENCE is a registered trademark of the AAAS. Domestic individual
full/334/6053/177-b                                              membership and subscription (51 issues): $149 ($74 allocated to subscription).
                                                                                                               Domestic institutional subscription (51 issues): $990; Foreign postage extra: Mexico,
                                                       Integrating Medicine and Science
Comment on “Changes in Climatic Water                                                                          Caribbean (surface mail) $55; other countries (air assist delivery) $85. First class,
                                                       12 October issue:              airmail, student, and emeritus rates on request. Canadian rates with GST available
Balance Drive Downhill Shifts in                                                                               upon request, GST #1254 88122. Publications Mail Agreement Number 1069624.
Plant Species’ Optimum Elevations”                     EDITORIAL: A Rattle Bag of Science and the              Printed in the U.S.A.
                                                                                                               Change of address: Allow 4 weeks, giving old and new addresses and 8-digit account
N. L. Stephenson and A. J. Das                         Art of Translation                                      number. Postmaster: Send change of address to AAAS, P.O. Box 96178, Washington,
Full text at           E. R. Edelman and G. A. FitzGerald                      DC 20090–6178. Single-copy sales: $10.00 current issue, $15.00 back issue prepaid
                                                                                                               includes surface postage; bulk rates on request. Authorization to photocopy
full/334/6053/177-c                                    Science Translational Medicine’s new Co–Chief           material for internal or personal use under circumstances not falling within the fair
                                                       Scientific Advisors define translational science          use provisions of the Copyright Act is granted by AAAS to libraries and other users
Response to Comments on “Changes in                                                                            registered with the Copyright Clearance Center (CCC) Transactional Reporting Service,
                                                       and articulate their vision for the journal.
Climatic Water Balance Drive Downhill Shifts                                                                   provided that $25.00 per article is paid directly to CCC, 222 Rosewood Drive, Danvers,
                                                                                                               MA 01923. The identification code for Science is 0036-8075. Science is indexed in the
in Plant Species’ Optimum Elevations”                  COMMENTARY: Medical Faculty Development—                Reader’s Guide to Periodical Literature and in several specialized indexes.
S. Z. Dobrowski et al.                                 A Modern-Day Odyssey
Full text at           M. C. Beckerle et al.
full/334/6053/177-d                                    Academic medical centers must make systematic
                                                       changes to improve the quality of life—and thus
                                                       productivity—of their faculty.

                                          SCIENCE         VOL 334         14 OCTOBER 2011                                                                                   151
Built for
DNA Ligases from
New England Biolabs.
Why compromise any aspect of performance in your
cloning experiments? DNA ligases from NEB set the
industry standard for function and purity. Continuous
product development and extensive quality controls
ensure success in your ligation reactions.

DNA ligases from NEB – building a strong foundation
for your cloning experiments.

                                                        Experience extreme purity with
                                                        NEB’s T4 DNA Ligase
                                                                     NEB     NEB
                                                                     Lot     Lot                   Supplier
                                                             M       101     102        A         B       C           D

Purchase NEB 5-alpha Competent Cells
with NEB’s Quick Ligation Kit and
                                                        Equivalent amounts of protein were loaded and analyzed by
receive the item of lesser value for FREE               Coomassie gel and silver stained using SilverXpress™. Marker M
                                                        is NEB’s Broad Range Protein Marker (NEB #P7702).
For more information, visit                                                                 SilverXpress™ is a trademark of Invitrogen.
                                                                                                           Monarch caterpillar
                                                                                                    © Maryellen Baker/Botanica
                                                                                                                                                                                                    EDITED BY STELLA HURTLEY

                                                                                                                                                                                   Earth’s crust, which eventually leads to earth-
                                                                                                                           String Order                                            quakes. Various mechanisms involving increased
                                                                                                                           One of the first successes of cold atomic gases          temperatures with friction have been invoked
                                                                                                                           as quantum simulators was the achievement of            to explain the initial stages of earthquake
                                                                                                                           a transition between the so-called Mott insulator       nucleation and rupture, including lubrication
                                                                                                                           to a superfluid phase: Optical lattices formed           of the contact area by the formation of a melt
                                                                                                                           with lasers were populated with atoms that              layer. However, many typically require slip to
                                                                                                                           stayed stuck on the lattice sites in the insulat-       occur over relatively large distances in order for
                                                                                                                           ing phase, and delocalized in the superfluid             friction to become sufficiently reduced. Goldsby
                                                                                                                           phase. Now, Endres et al. (p. 200) used the             and Tullis (p. 216) demonstrated, through a
                                                                                                                           newly available ability to observe individual           series of friction experiments with a variety of
                                                                                                                           lattice sites to track the transition in more detail,   silicate rock types, that the coefficient of friction
                                                                                                                           and observed correlated pairs of doubly- and            can also drop dramatically across short distanc-
                                                                                                                           unpopulated sites, which represent the excita-          es—just a few centimeters—at high slip rates
                                                                                                                           tions of the insulating phase in which each site        through a process known as “flash” heating.
                                                                                                                           is populated by a single atom. A consequence            Flash heating may contribute to the weakening
                                                                                                                           of the correlated pairs, a hidden “string order,”       of faults during small earthquakes and at the
                                                                                                                           was found to mark the transition.                       beginning stages of large earthquakes.

                                                                                                                           Rules for Positioning                                   Hydrocarbons Get Hitched
                                                                                                                                                                                   Saturated hydrocarbons are among the least
                                                                        Pieces of Light                                    Nanoparticles                                           reactive organic compounds. Once unsaturated
                                                                        Waves consist of peaks and troughs, whether        When atoms and molecules crystallize, the lattice       (that is, deprived of hydrogen atoms at adjacent
                                                                        the oscillating medium is water in the ocean       that forms is usually fixed by temperature and           carbon centers) by heating on a catalyst, they
                                                                        or electromagnetic energy in light. If waves       pressure. For larger nanoparticles and colloids,        are subject to polymerization, but the resulting
                                                                        with different wavelengths overlap pre-            more control can be exerted over the assem-             product distribution tends to be complex. Zhong
                                                                        cisely, the pattern gets more complex and          bly into lattices because particle coatings and         et al. (p. 213) now
                                                                        less repetitive until eventually, with enough      solvation can be varied, in addition to thermo-         show that confining
                                                                        components, there is no cycle left—just an         dynamic parameters. For example, DNA coatings           long-chain linear
                                                                        isolated blip pointing in some discrete direc-     with “sticky ends” have been used to assemble           hydrocarbons in
                                                                        tion. Wirth et al. (p. 195, published online 8     gold nanoparticles into body-centered cubic             the one-dimension-
                                                                        September) have achieved this subcycle state       (bcc) and face-centered cubic (fcc) superlattices.      al channels of a
                                                                        of light by packing a vast spectrum of light       Macfarlane et al. (p. 204; see the Perpective by        gold surface leads
                                                                        (from the near-infrared through the visible        Travesset) now show how control over hydrody-           to selective end-to-
                                                                        and into the ultraviolet) with a well-controlled   namic radii of particles and assembly kinetics can      end catenation of
                                                                        phase into a single tight pulse. They further      be used to create bcc and fcc lattices, and seven       the chains, following thermally induced loss
                                                                        demonstrate the potential of the light pulses      other lattices, with a wide variation in nanopar-       of hydrogen.
                                                                        in probing atomic electron dynamics at the         ticle size and lattice parameters.
                                                                        fastest time scales.
                                                                                                                           Ancient Ochre Workshop                                  A Time and a Place
                                                                                                                                                                                   for Hox Genes
                                                                        Talking Teeth                                      Ochre—essentially, colored earth—was com-

                                                                                                                           monly used by early humans for simple art work,         Patterning of the mammalian body relies on the
                                                                        Human evolution has often been closely tied to     body pigmentation, or protection. The use of            stepwise transcriptional activation of Hox genes.
                                                                        diet, and teeth are some of the most common        ochre is well documented after about 60,000             Noordermeer et al. (p. 222) show that this pro-
                                                                        fossils available. Ungar and Sponheimer (p.        years ago, although there is evidence of earlier        cess involves a dynamic transition in the global
                                                                        190) review recent developments in understand-     use. Henshilwood et al. (p. 219) now describe           architecture of Hox gene clusters, with each
                                                                        ing the diets of early humans, focusing on         an ochre processing workshop at Blombos Cave,           gene transitioning, one after the other, from a
                                                                        two emerging approaches that have provided         South Africa, dated to about 100,000 years ago,         negative three-dimensional (3D) compartment
                                                                        new perspectives. The first is examination of       considerably earlier than other such workshops.         to an active compartment. This bimodal configu-
                                                                        microwear on teeth, which in part reflect the       The workshop includes hammers and grind-                ration parallels the distribution of distinct chro-
                                                                        abrasiveness and hardness of foods. The second     stones for making the ochre powder and two              matin marks, suggesting the existence of a link
                                                                        is stable isotope analysis; particularly, carbon   shells where the ochre was stored.                      between the presence of chromatin domains and
                                                                        isotopes that reflect the proportion of grasses                                                             the formation of 3D chromosomal structures.
                                                                        versus fruits and nuts eaten directly, or con-                                                             This model for Hox gene activation would ensure
                                                                        sumed by an animal eaten by an early hominin.
                                                                                                                           Flash Drive                                             the proper sequence in the transcriptional acti-
                                                                        The diets of early humans appear to have been      The rate at which rocks slide past one another          vation of Hox genes within each gene cluster.
                                                                        more diverse than previously assumed.              controls the stress accumulated along faults in                                         Continued on page 155

                                                                                                       SCIENCE VOL 334                    14 OCTOBER 2011                                                      153
Simple Western is here.
                                                                                                     This Week in Science

                      Continued from page 153

                      Forest Tipping Points
                      The existence of tipping points in natural systems is increasingly recognized (see the Perspective
                      by Mayer and Khalyani). Hirota et al. (p. 232) analyzed a global data set on tree distribution
                      showing that forest, savanna, and a treeless state represent alternative stable states on large scales.
                      This result implies that, when drivers such as climate or logging reach a tipping point, there is a
                      potential for irreversible shifts. Staver et al. (p. 230) explore the bimodalities in tree cover across

                      sub-Saharan Africa, South America, and Australia and evaluate the potential role of fire in the dis-
                      tribution and dynamics of savanna and forest. Rainfall and seasonality are global predictors of tree
                      cover but, at intermediate rainfall with mild seasonality, tree cover is bimodal. Within this climate
                      envelope, fire differentiates between savanna and forest.

                      Skipping Replication                                                                                      polysyllabic
                      The DNA replication machinery replicates DNA in a 5’ to 3’ direction. Thus, the leading (5’ to 3’)
                      strand of the DNA double helix is thought to be replicated continuously, and the lagging (3’ to 5’)
                      strand discontinuously. DNA lesions on the lagging strand cause little impediment to replisome prog-
                      ress because of the discontinuous nature of replication. Lesions on the leading strand were thought
                      to require DNA polymerase to dissociate from DNA and reinitiation downstream of the damage.
                      Yeeles and Marians (p. 235) studied the in vitro replication of an Escherichia coli plasmid carrying      Ardipithecus
                      a single cyclobutane pyrimidine dimer lesion, on the leading strand. On encountering the damaged
                      template, a minimal E. coli replisome remained associated with the DNA, skipped across the lesion,        ramidus.
                      and reinitiated leading-strand synthesis downstream of the damage, independent of any replication
                      restart proteins.                                                                                         One more data point on why
                                                                                                                                you should spend more time
                      How Escherichia coli Got Its Stripes                                                                      at
                      How living organisms develop regular anatomic patterns can be diffi-                                       There you can enjoy
                      cult to unravel when studied in the context of complex physiology.                                        members-only downloads,
                      Synthetic biology provides a bottom-up approach to identify
                                                                                                                                videos, webinars, blogs,
                      minimal circuits that can drive patterning. Liu et al.
                      (p. 238) describe a synthetic genetic circuit that                                                        discounts, and other content
                      couples cell density and motility in order to program                                                     geared for people who aren’t
                      the formation of periodic stripes in a growing
                      E. coli population. The system could be adjusted
                                                                                                                                afraid of footnotes.
                      to modulate the patterning, and a mathematical
                      model was able to predict the experimental results.

                      A Genetic Clue to Fibroids
                      Uterine fibroids (leiomyomas) are estimated to affect more than half of the female population over
                      the age of 50. Although benign, these tumors nevertheless can cause serious health complications
                      and are one of the most common reasons for hysterectomy. Mäkinen et al. (p. 252, published
                      online 25 August) explored the pathogenetic basis of uterine fibroids by performing gene sequence
                      analysis on 225 tumor samples derived from 80 patients. Remarkably, somatic mutations in a gene
                      called MED12 were found in over 70% of the tumors. The protein encoded by MED12 is a subunit of
                      the mediator complex, a 26-subunit transcription factor that is thought to regulate gene expres-
                      sion through its interactions with RNA polymerase II and DNA regulatory elements.

                      Keep Your Distance
                      Nearly one trillion bacteria reside in our gut, but amazingly, our immune system does not wage a
                      war against them. How is such a peaceful coexistence achieved? There is likely a battery of mecha-

                      nisms; however, one is necessary to keep the bacteria in the small intestine and colon physically
                      separated from the immunologically active mucosal epithelium. Vaishnava et al. (p. 255; see the
                      Perspective by Johansson and Hansson) used a combination of genetic approaches to demonstrate
                      that signaling downstream of Toll-like receptors (TLR), which recognize conserved signatures of
                      microorganisms to initiate immune responses, is required to maintain this separation.
                              SCIENCE VOL 334                   14 OCTOBER 2011
                                                                                                New Promotion Sept. 1 – Dec. 31, 2011.
                                                                                                        Offers may vary by country.

                                                                                                                                                          Falcon is a registered trademark of Becton, Dickinson and Co., USA. Eppendorf and Eppendorf Thermomixer are registered trademarks
                                                                                                                                                          of Eppendorf AG, Germany. Eppendorf Advantage, Eppendorf Safe-Lock Tubes, Thermomixer comfort and ThermoStat plus are
                                                                                                                                                          trademarks of Eppendorf AG, Germany. All rights reserved, including graphics and images. © Copyright 2011 by Eppendorf AG.
                      ‡ New ThermoStat
                        plus Bundles:
                        Save 23 %!*

                      ‡ New Thermomixer
                        comfort Bundle:
                        Save ~ 18 %!*

                      ‡ Refrigerated
                        Save up to 16 %!*

                                                                                                                                         Limited offer
                                                                                                                                        Sept.1–Dec. 31,

                         Hot Offers, Cool Solutions!
 Save big on ThermoStat plus™, Thermomixer comfort™ and refrigerated microcentrifuges

Working with precious and expensive samples requires                       Hot offers, cool solutions for a variety of tasks!
your utmost concentration. The last thing you want to be                   Check out our new limited offers* on ThermoStat plus,
thinking about or questioning is the reliability of the equip-             Thermomixer comfort and refrigerated microcentrifuges
ment you are using.                                                        5418 R, 5424 R and 5430 R. To take advantage, contact
                                                                           your participating Eppendorf Partner today!
We offer you the perfect partnership!
Your priority is your sample; it’s our priority too.                       More product details and dealer contact information at
Let Eppendorf take care of your high-quality instrument          
needs; you take care of the research – it’s the perfect
                                                                           * Savings compared to individual purchases at list prices.
partnership! Eppendorf’s reliable and robust, versatile and                  Promotion bundles available as long as supplies last.
intuitive products leave you free to concentrate on what
really matters – your sample, your research, your results!

                   Your local distributor · Application Support E-Mail:
               Eppendorf AG · Hamburg · Germany · Tel: +49 40 538 01-0 · Eppendorf North America, Inc. · USA · Tel.: +1 800 645 3050
                                       Eppendorf Asia Pacific Headquarters · Malaysia · Tel: +60 3 8023 5430

                                                                                              Peruvian Highlands, Fume-Free
                                                                       Pilar Nores Bodereau   IN THE ANDEAN HIGHLANDS OF PERU, A TYPICAL SINGLE-ROOM HOME WILL BURN APPROXIMATELY
                                                                       is the founder of      3.6 tons of wood a year, not just for heating but for cooking indoors. Three billion people world-
                                                                       Sembrando and          wide cook indoors over open fires with solid fuels. This use has a detrimental effect on human
                                                                       a former first lady     health. According to the World Health Organization (WHO), nearly two million people die
                                                                       of Peru.               annually from open-fire cooking, with women and young children the most affected. The haz-
                                                                                              ards related to these practices include acute respiratory infections, insufficient weight at birth,
                                                                                              and chronic obstructive pulmonary disease. Despite WHO estimates that cookstove smoke is
                                                                                              one of the top five threats to public health in poor developing countries, the effects of expo-
                                                                                              sure to it have received limited funding and research attention. The good news is that since the
                                                                                              announcement last year of the Global Alliance for Clean Cookstoves [launched by the United
                                                                                              Nations (UN) Foundation], coordinated efforts are successfully under way to support the use of
                                                                                              clean cookstoves in the developing world.*
                                                                                                  The alliance involves a wide scope of public and private partners,
                                                                                              from international nonprofit organizations, academic institutions,
                                                                                              corporate leaders, governments, and UN agencies to local commu-
                                                                                              nity groups. It is critical to emphasize the impact of complemen-
                                                                                              tary local-level efforts, as evident in Peru. In Peru, 10 million people
                                                                                              (about 30% of the population) live in the Andean highlands. They
                                                                                              cook over open fires inside their homes of about 200 square feet with
                                                                                              no windows, causing indoor air pollution at a rate 30 times higher
                                                                                              than permitted by WHO. In these rural areas, about 60% of children
                                                                                              suffer from chronic malnutrition mainly caused by poor hygiene (the
                                                                                              lack of clean water) and relentless respiratory diseases. More than
                                                                                              40% of the women suffer from chronic obstructive lung diseases and/
                                                                                              or from a cardiovascular disease, apparently related to their house-
                                                                                              hold work conditions. The Peruvian Andean population is highly dis-
                                                                                              persed over more than 70,000 small communities, many of which are remote, making the
                                                                                              provision of basic services from the government impossible. There, traditional models of aid
                                                                                              for development have failed to achieve their many objectives because they tackle each prob-
                                                                                              lem separately. A developmental program suitable for these rural areas needs to incorporate
                                                                                              a comprehensive assessment of the problems and a multisectorial approach to the solutions.
                                                                                                  Sembrando, an initiative of the private organization Instituto Trabajo y Familia (Work and
                                                                                              Family Institute), supports Andean communities to improve their productivity and social
                                                                                              development, including a decrease in child mortality, the promotion of maternal health and
                                                                                              gender equality, and environmental preservation—all UN Millennium Goals. Sembrando
                                                                                              provides tools to diminish indoor air pollution and general contamination by helping the
                                                                                              community build improved cookstoves, latrines, and family orchards for every household. It

                                                                                              also offers extensive training at both the family and community levels, with special attention
                                                                                              to pregnant women, helping them achieve empowerment and develop the abilities needed to
                                                                                              overcome their own problems associated with underdevelopment.
                                                                                                  Since 2006, Sembrando has served 92,000 families (approximately 500,000 people)
                                                                                              in 2800 communities in the Andean region of Peru. Preliminary results show a substantial
                                                                                              decrease in bronchopulmonary diseases and a clear increase in the height/weight ratio of chil-
                                                                                              dren under 5 years old. Sembrando has also yielded positive results toward the reduction of
                                                                                              extreme poverty and chronic malnutrition in the rural Andean context, using simple technol-
                                                                                              ogy with immediate and low-cost results (U.S. $200 per family). This has inspired the Peru-
                                                                                              vian government to start a campaign to build 500,000 clean cookstoves nationwide.
                                                                                                  Strong local efforts such as Sembrando can be sustainable, helping to meet the Global Alli-
                                                                                              ance goal of clean cookstoves in 100 million households by 2020, while improving the eco-
                                                                                              nomic and social conditions of families that live in extreme poverty.      – Pilar Nores Bodereau


                                                                                     SCIENCE VOL 334 14 OCTOBER 2011                                                                    157


      A Smart Merger
      The identification of factors that have driven the
      evolution of cognition not just between humans
      and nonhumans, but across taxa, is of interest
      to comparative psychologists. This interest has
      arisen because of the recognition that high-level
      cognitive function is not limited to primate
      lineages and that cognition, like many other
      traits, is probably shaped by selection imposed
      by ecological and environmental demands.
      MacClean et al. now propose that the merger
      of the fields of comparative psychology and
      phylogenetics will greatly improve our ability to
      understand the forces that drive cognitive evolu-
      tion. By using examples of comparative data
      on inhibitory function, a measure of cognitive
      ability, they highlight how the comparative phy-                DEVELOPMENT
      logenetic method will expand our understanding
      by testing for correlations between cognition                   Myocytes, Mitochondria, and Maturation
      and specific life history, morphological, or social
      ecological traits; measuring how well phyloge-
      netic relatedness predicts similarity in cognitive              Understanding the factors and events required for normal heart development will hope-
      function; and estimating ancestral levels of                    fully provide insight into how to treat heart disease and other cardiac abnormalities.
      cognitive function based on measured levels                     Mitochondrial function is essential in cardiac cells, with mitochondria transitioning from an
      in extant related taxa. This approach will also                 immature to mature state in the embryo. How this occurs, however, is not well understood.
      allow for an improved ability to select appropri-               During early embryonic development, Hom et al. find that mitochondria alter their length,
      ate pairs of species for comparison. The new                    network complexity, and area, becoming more mature. This corresponds with a decrease
      level of analysis afforded by the merger of these               in reactive oxygen species in cardiomyocytes and their differentiation. Genetic and phar-
      two fields will enable a move away from simple                   macological inhibition showed that these changes are probably the result of the closure of
      “cognitive model” species and will provide                      mitochondrial permeability transition pores (mPTPs). mPTP closure increased ATP produc-
      insight into how cognitive abilities evolved and                tion needed for myocyte function and drove myocyte differentiation through redox signal-
      operate in a wide array of species, perhaps even                ing. Thus, the maturation of mitochondria signals cardiomyocyte differentiation. — BAP
      our own. — SNV                                                                                                                         Dev. Cell 21, 469 (2011).
                                         Anim. Cogn. 14, 10.1007/
                                       s10071-011-0448-8 (2011).

      M AT E R I A L S S C I E N C E                                cubes. At the interface between the nucleus and       noticeable in comparison with the dominant but
                                                                    the amorphous phase, they observed a series of        conventionally conducting bulk state. Chemical
      Time for Change                                               quasi-ordered rings, an arrangement facilitated       doping and electrical gating have been used
      Phase-change materials, such as Ge2Sb2Te5                     by the near 90° bond angles found in GST. They        to control the conductivity of the bulk state in
      (GST), show exceptionally fast phase transi-                  propose that it is the similarity between the         binary TIs such as Bi2Se3. With the same aim,
      tions, which renders them useful for optical                  bonds found in the amorphous state and the            Kong et al. vary the Bi and Sb content in thin
      data storage and nonvolatile memory applica-                  distorted rocksalt structure, common to most          nanoplates of a ternary compound (BixSb1-x)2Te3     CREDIT: THOMAS DEERINCK, NCMIR/SCIENCE PHOTO LIBRARY
      tions. Ab initio molecular dynamics simulations               phase-change materials, that allows such rapid        between the two extremes, Bi2Te3 and Sb2Te3,
      can explore the early stages of nucleation and                crystallization. — MSL                                which are both TIs. They find that the TI char-
      crystallization in such materials. Lee and Elliott                           Phys. Rev. Lett. 107, 145702 (2011).   acter is maintained across the doping range,
      used models with 180 atoms that were held at                                                                        while the carrier density of the bulk state
      constant volume to simulate capped films and                   APPLIED PHYSICS                                       varies dramatically in magnitude, reaching a
      annealed from a quenched glassy state. They                                                                         minimum at equal amounts of Bi and Sb (x =
      defined the smallest structural element in the
                                                                    TI Tuning                                             0.5). For this optimal composition and at an
      metastable rocksalt phase of GST as a fourfold                On the list of exotic materials on the verge of       optimal thickness of the plate, applying a gate
      ring (or square), six of which are required to                becoming technologically useful, topologi-            voltage leads to pronounced tuning of the car-
      make a cube. Both these elements are observed                 cal insulators (TIs) rank high. They possess a        rier density and type, going from electron- to
      in the amorphous phases as transient structures.              favorable electronic surface state in which a         hole-dominated transport, similar to what has
      By tracking the number of rings and cubes, they               certain type of scattering is suppressed because      been observed in graphene. — JS
      identified four stages in the crystallization pro-             of symmetry; however, in almost all materials                                Nat. Nanotechnol. 10.1038/
      cess, with a critical nucleus size of 5 to 10 GST             identified to date, this surface state is barely                               NNANO.2011.172 (2011).

158                                                   14 OCTOBER 2011          VOL 334      SCIENCE
                                                                              EDITOR-IN-CHIEF Bruce Alberts                                                       EXECUTIVE PUBLISHER Alan I. Leshner
                                                                                               EXECUTIVE EDITOR                   NEWS EDITOR                                                       PUBLISHER Beth Rosner
                                                                                          Monica M. Bradford                     Colin Norman
                                                                                   MANAGING EDITOR, RESEARCH JOURNALS Katrina L. Kelner                                 FULFILLMENT SYSTEMS AND OPERATIONS (; CUSTOMER
                                                                                   DEPUTY EDITORS R. Brooks Hanson, Barbara R. Jasny,                   Andrew          SERVICE SUPERVISOR Pat Butler; SPECIALISTS LaVonda Crawford, Vicki
                                                                                   M. Sugden                                                                            Linton, April Marshall; DATA ENTRY SPECIALISTS Shirlene Hall, Tarrika Hill,
                                                                                   EDITORIAL  SENIOR EDITORS/COMMENTARY Lisa D. Chong, Brad Wible; SENIOR               William Jones
                    1200 New York Avenue, NW                                       EDITORS Gilbert J. Chin, Pamela J. Hines, Paula A. Kiberstis (Boston), Marc          BUSINESS OPERATIONS         AND   ADMINISTRATION         Deborah Rivera-
                      Washington, DC 20005                                         S. Lavine (Toronto), Beverly A. Purnell, L. Bryan Ray, Guy Riddihough,               Wienhold;        BUSINESS SYSTEMS AND FINANCIAL ANALYSIS DIRECTOR Randy Yi;
              Editorial: 202-326-6550, FAX 202-289-7562                            H. Jesse Smith, Phillip D. Szuromi (Tennessee), Valda Vinson, Jake                   MANAGER, FULFILLMENT SYSTEMS Frits Buningh; SYSTEMS ANALYST Nicole
               News: 202-326-6581, FAX 202-371-9227                                S. Yeston, Laura M. Zahn (San Diego); ASSOCIATE EDITORS Kristen L.                   Mehmedovich; MANAGER , BUSINESS ANALYSIS Eric Knott; MANAGER ,
                                                                                   Mueller, Jelena Stajic, Sacha Vignieri, Nicholas S. Wigginton; BOOK REVIEW           BUSINESS OPERATIONS Jessica Tierney; BUSINESS ANALYSTS Priti Pamnani,
                Bateman House, 82-88 Hills Road                                    EDITOR Sherman J. Suter; ASSOCIATE LETTERS EDITOR Jennifer Sills; EDITORIAL          Celeste Troxler; Christine Wehrli; RIGHTS AND PERMISSIONS :
                    Cambridge, UK CB2 1LQ                                          MANAGER Cara Tate; SENIOR COPY EDITORS Jeffrey E. Cook, Cynthia Howe,                ADMINISTRATOR Emilie David; ASSOCIATE Elizabeth Sandler;
           +44 (0) 1223 326500, FAX +44 (0) 1223 326501                            Harry Jach, Lauren Kmec, Barbara P. Ordway, Trista Wagoner; COPY EDITOR              MARKETING DIRECTOR Ian King; MARKETING MANAGERS Allison Pritchard, Alison
                                                                                   Chris Filiatreau; SENIOR EDITORIAL COORDINATORS Carolyn Kyle, Beverly                Chandler, Julianne Wielga, Samantha Smith; MARKETING ASSOCIATES
SUBSCRIPTION SERVICES For change of address, missing issues, new                   Shields; EDITORIAL COORDINATORS Joi S. Granger, Anita Wynn; PUBLICATIONS             Aimee Aponte, Mary Ellen Crowley; SENIOR MARKETING EXECUTIVE Jennifer
orders and renewals, and payment questions: 866-434-AAAS (2227)                    ASSISTANTS Ramatoulaye Diop, Aneera Dobbins, Jeffrey Hearn,                          Reeves; DIRECTOR, SITE LICENSING Tom Ryan; DIRECTOR, CORPORATE RELATIONS
or 202-326-6417, FAX 202-842-1065. Mailing addresses: AAAS, P.O.                   Michael Hicks, Lisa Johnson, Scott Miller, Jerry Richardson,                         Eileen Bernadette Moran; SENIOR PUBLISHER RELATIONS SPECIALIST
Box 96178, Washington, DC 20090-6178 or AAAS Member Services,                      Teresa R. Sakon, Brian White; EDITORIAL ASSISTANT Patricia M. Moore; EXECUTIVE       Kiki Forsythe; PUBLISHER RELATIONS MANAGER Catherine Holland; PUBLISHER
1200 New York Avenue, NW, Washington, DC 20005                                     EDITORIAL ASSISTANT Yolanda O'Bannon (San Francisco); EXECUTIVE ASSISTANT            RELATIONS, EASTERN REGION Phillip Smith; PUBLISHER RELATIONS, WESTERN REGION
I NSTITUTIONAL S ITE L ICENSES please call 202-326-6755 for any                    Alison Crawford; ADMINISTRATIVE SUPPORT Maryrose Madrid; EDITORIAL FELLOW            Ryan Rexroth; CUSTOMER RELATIONS MANAGER Iquo Edim; MARKETING MAN-
questions or information                                                           Melissa R. McCartney                                                                 AGER Christina Schlecht; MARKETING ASSOCIATE Chad Johnson; ELEC-
REPRINTS: Author Inquiries 800-635-7181                                            E DITORIAL DIRECTOR , WEB AND NEW MEDIA Stewart Wills; RESEARCH                      TRONIC MEDIA : DIRECTOR Lizabeth Harman; ASSISTANT MANAGER Lisa
           Commercial Inquiries 803-359-4578                                       ASSOCIATE Corinna Cohn; WEB DEVELOPMENT MANAGER Martyn Green;                        Stanford; SENIOR PRODUCTION SPECIALIST Ryan Atkins; PRODUCTION SPE-
                                                                                   WEB DEVELOPER Andrew Whitesell; INTERNS Andrew Green, Kerry Klein                    CIALISTS Antoinette Hodal, Nichele Johnston, Kimberly Oster; DIREC-
PERMISSIONS 202-326-7074, FAX 202-682-0816
                                                                                   NEWS DEPUTY NEWS EDITORS Robert Coontz, David Grimm (Online), Eliot                  TOR, WEB AND NEW MEDIA Will Collins; PROJECT MANAGER Trista Snyder;
MEMBER BENEFITS AAAS/Barnes& bookstore;                   Marshall, Jeffrey Mervis, Leslie Roberts, John Travis; CONTRIBUTING EDITORS          SENIOR PRODUCTION SPECIALIST Christopher Coleman; COMPUTER SPECIALISTS
AAAS Online Store code MKB6; AAAS                          Elizabeth Culotta, Polly Shulman; NEWS WRITERS Yudhijit Bhattacharjee,               Walter Jones, Kai Zhang
Travels: Betchart Expeditions 800-252-4910; Apple Store www.                       Adrian Cho, Jennifer Couzin-Frankel, Carolyn Gramling, Jocelyn Kaiser,; Bank of America MasterCard 1-800-833-6262                                                                                                      ADVERTISING DIRECTOR, WORLDWIDE AD SALES Bill Moran
                                                                                   Richard A. Kerr, David Malakoff, Greg Miller, Elizabeth Pennisi,
priority code FAA3YU; Cold Spring Harbor Laboratory Press                                                                                                               COMMERCIAL EDITOR Sean Sanders: 202-326-6430
                                                                                   Robert F. Service (Pacific NW), Erik Stokstad; WEB DEVELOPER Daniel
Publications; GEICO Auto                     Berger; INTERN Daniel Strain; CONTRIBUTING CORRESPONDENTS Jon Cohen (San             ASSISTANT COMMERCIAL EDITOR Tianna Hicklin 202-326-6463
Insurance;                           Diego, CA), Daniel Ferber, Ann Gibbons, Sam Kean, Andrew Lawler,                     P R O D U C T (;           M I D W E S T Rick
Hertz 800-654-2200 CDP#343457; Office Depot https://bsd.                           Mitch Leslie, Charles C. Mann, Virginia Morell, Gary Taubes; COPY EDITORS            Bongiovanni: 330-405-7080, FAX 330-405-7081; EAST COAST /; Seabury & Smith Life Insurance 800-                 Linda B. Felaco, Melvin Gatling, Melissa Raimondi; ADMINISTRATIVE SUPPORT            E . CANADA Laurie Faraday: 508-747-9395, FAX 617-507-8189;
424-9883; Subaru VIP Program 202-326-6417; VIP Moving Services                     Scherraine Mack; BUREAUS San Diego, CA: 760-942-3252, FAX 760-942-                   WEST COAST/W. CANADA Lynne Stickrod: 415-931-9782, FAX 415-520-; Other Benefits: AAAS                       4979; Pacific Northwest: 503-963-1940                                                 6940; UK/EUROPE/ASIA Roger Goncalves: TEL/FAX +41 43 243 1358;
Member Services 202-326-6417 or                                PRODUCTION DIRECTOR Wendy K. Shank; ASSISTANT MANAGER Rebecca                        JAPAN ASCA Corporation, Makiko Hara: +81 (0) 3 6802 4616, FAX              (for general editorial queries)              Doshi; SENIOR SPECIALISTS Steve Forrester, Chris Redwood, Anthony                    +81 (0) 3 6802 4615;; CHINA/TAIWAN Ruolei Wu:              (for queries about letters)                  Rosen; PREFLIGHT DIRECTOR David M. Tompkins; MANAGER Marcus Spiegler;                +86 1367 1015 294              (for returning manuscript reviews)           SPECIALISTS Jason Hillman, Tara Kelly                                                WORLDWIDE ASSOCIATE DIRECTOR OF SCIENCE CAREERS Tracy Holmes: +44 (0)             (for book review queries)                    ART DIRECTOR Yael Fitzpatrick; ASSOCIATE ART DIRECTOR Laura Creveling;               1223 326525, FAX +44 (0) 1223 326532
                                                                                   SENIOR ILLUSTRATORS Chris Bickel, Katharine Sutliff; ILLUSTRATOR Yana
Published by the American Association for the Advancement of Science                                                                                                    CLASSIFIED (; U.S.: MIDWEST/WEST COAST/
(AAAS), Science serves its readers as a forum for the presentation and             Hammond; SENIOR ART ASSOCIATES Holly Bishop, Preston Huey,                           SOUTH CENTRAL/CANADA Tina Burks: 202-326-6577; EAST COAST/INDUSTRY
discussion of important issues related to the advancement of science,              Nayomi Kevitiyagala, Matthew Twombly; ART ASSOCIATE Kay Engman;                      Elizabeth Early: 202-326-6578; SALES ADMINISTRATOR Marci Gallun;
including the presentation of minority or conflicting points of view,              PHOTO EDITOR Leslie Blizard
                                                                                                                                                                        E U R O P E / R OW SA L E S Simone Jux, Alex Palmer; SA L E S A S S I S -
rather than by publishing only material on which a consensus has been                                     SCIENCE INTERNATIONAL
                                                                                                                                                                        TANT Lisa Patterson; JAPAN Makiko Hara +81 (0) 90 9853-
reached. Accordingly, all articles published in Science—including                  EUROPE ( EDITORIAL: INTERNATIONAL MANAGING
                                                                                   EDITOR Andrew M. Sugden; SENIOR EDITOR/COMMENTARY Julia Fahrenkamp-
                                                                                                                                                                        9982, FAX +81 (0)3-6369-4491;;
editorials, news and comment, and book reviews—are signed and reflect                                                                                                    CHINA / TAIWAN Ruolei Wu: +86 1367 1015 294;
the individual views of the authors and not official points of view adopted         Uppenbrink; SENIOR EDITORS Caroline Ash, Stella M. Hurtley, Ian S. Osborne,
                                                                                                                                                                        ADVERTISING SUPPORT MANAGER Karen Foote: 202-326-6740; ADVERTISING
by AAAS or the institutions with which the authors are affiliated.                  Peter Stern; ASSOCIATE EDITOR Maria Cruz; LOCUM EDITOR Helen Pickersgill;
                                                                                                                                                                        PRODUCTION OPERATIONS MANAGER Deborah Tompkins; SENIOR PRODUCTION
                                                                                   EDITORIAL SUPPORT Samantha Hogg, Alice Whaley; ADMINISTRATIVE
AAAS was founded in 1848 and incorporated in 1874. Its mission is to               SUPPORT John Cannell, Janet Clements, Louise Hartwell; NEWS : DEPUTY
                                                                                                                                                                        SPECIALIST / GRAPHIC DESIGNER Amy Hardcastle; PRODUCTION SPECIALIST
advance science, engineering, and innovation throughout the world for                                                                                                   Yuse Lajiminmuhip; SENIOR TRAFFIC ASSOCIATE Christine Hall; SALES
                                                                                   NEWS EDITOR , U . K . Daniel Clery; NEWSWRITER Sara Reardon; CONTRIBUTING
the benefit of all people. The goals of the association are to: enhance                                                                                                  COORDINATOR Shirley Young
                                                                                   EDITOR, EUROPE Martin Enserink; CONTRIBUTING CORRESPONDENTS Michael Balter
communication among scientists, engineers, and the public; promote and                                                                                                  AAAS B OARD OF D IRECTORS RETIRING PRESIDENT , CHAIR Alice Huang;
defend the integrity of science and its use; strengthen support for the            (Paris), John Bohannon (Vienna), Gretchen Vogel (Berlin)
                                                                                   ASIA Japan Office: Asca Corporation, Tomoko Furusawa, Rustic Bldg. 7F,                PRESIDENT Nina Fedoroff; PRESIDENT - ELECT William Press; TREASURER
science and technology enterprise; provide a voice for science on societal                                                                                              David E. Shaw; CHIEF EXECUTIVE OFFICER Alan I. Leshner; BOARD Nancy
issues; promote the responsible use of science in public policy; strengthen        77 Tenjin-cho, Shinjuku-ku, Tokyo 162-0808, Japan; +81 3
and diversify the science and technology workforce; foster education in            6802 4616, FAX +81 3 6802 4615,;                               Knowlton, Stephen Mayo, Raymond Orbach, Julia M. Phillips, Sue V.
science and technology for everyone; increase public engagement with               ASIA NEWS EDITOR Richard Stone (Beijing:; CONTRIBUTING              Rosser, David D. Sabatini, Inder Verma, Thomas A. Woolsey
science and technology; and advance international cooperation in science.          CORRESPONDENTS Dennis Normile [Japan: +81 (0) 3 3391 0630, FAX
                                                                                   +81 (0) 3 5936 3531;]; Hao Xin [China:
 INFORMATION FOR AUTHORS                                                 ]; Mara Hvistendahl [China: mara@
 See pages 784 and 785 of the 11 February 2011 issue or                  ]; Pallava Bagla [South Asia: +91 (0) 11 2271
 access                                           2896;]

   SENIOR EDITORIAL BOARD                                      Tom Daniel, Univ. of Washington                              Daniel Kahne, Harvard Univ.                                       Barbara A. Romanowicz, Univ. of California, Berkeley
                                                               Stanislas Dehaene, Collège de France                         Bernhard Keimer, Max Planck Inst., Stuttgart                      Jens Rostrup-Nielsen, Haldor Topsoe
   A. Paul Alivisatos, Lawrence Berkeley Nat'l. Laboratory     Emmanouil T. Dermitzakis, Univ. of Geneva Medical School     Joel Kingsolver, Univ. of North Carolina at Chapel Hill           Edward M. Rubin, Lawrence Berkeley National Lab
   Cori Bargmann, The Rockefeller Univ.                        Robert Desimone, MIT                                         Robert Kingston, Harvard Medical School                           Mike Ryan, Univ. of Texas, Austin
   Ernst Fehr, Univ. of Zurich                                 Claude Desplan, New York Univ.                               Alberto R. Kornblihtt, Univ. of Buenos Aires                      Shimon Sakaguchi, Kyoto Univ.
   Richard Losick, Harvard Univ.                               Ap Dijksterhuis, Radboud Univ. of Nijmegen                   Leonid Kruglyak, Princeton Univ.                                  Miquel Salmeron, Lawrence Berkeley National Lab
   Michael S. Turner, University of Chicago                    Dennis Discher, Univ. of Pennsylvania                        Mitchell A. Lazar, Univ. of Pennsylvania                          Jürgen Sandkühler, Medical Univ. of Vienna
                                                               Jennifer A. Doudna, Univ. of California, Berkeley            David Lazer, Harvard Univ.                                        Randy Seeley, Univ. of Cincinnati
   BOARD OF REVIEWING EDITORS                                  Julian Downward, Cancer Research UK                          Virginia Lee, Univ. of Pennsylvania                               Vladimir Shalaev, Purdue Univ.
   Adriano Aguzzi, Univ. Hospital Zürich                       Bruce Dunn, Univ. of California, Los Angeles                 Ottoline Leyser, Cambridge Univ.                                  Joseph Silk, Univ. of Oxford
   Takuzo Aida, Univ. of Tokyo                                 Christopher Dye, WHO                                         Olle Lindvall, Univ. Hospital, Lund                               Denis Simon, Univ. of Oregon
   Sonia Altizer, Univ. of Georgia                             David Ehrhardt, Carnegie Inst. of Washington                 Marcia C. Linn, Univ. of California, Berkeley                     Alison Smith, John Innes Centre
   Sebastian Amigorena, Institut Curie                         Michael B. Elowitz, Calif. Inst. of Technology               John Lis, Cornell Univ.                                           Davor Solter, Inst. of Medical Biology, Singapore
   Angelika Amon, MIT                                          Tim Elston, Univ. of North Carolina at Chapel Hill           Jianguo Liu, Michigan State Univ.                                 John Speakman, Univ. of Aberdeen
   Kathryn Anderson, Memorial Sloan-Kettering Cancer Center    Gerhard Ertl, Fritz-Haber-Institut, Berlin                   Richard Losick, Harvard Univ.                                     Allan C. Spradling, Carnegie Institution of Washington
   Siv G. E. Andersson, Uppsala Univ.                          Barry Everitt, Univ. of Cambridge                            Jonathan Losos, Harvard Univ.                                     Jonathan Sprent, Garvan Inst. of Medical Research
   Peter Andolfatto, Princeton Univ.                           Paul G. Falkowski, Rutgers Univ.                             Ke Lu, Chinese Acad. of Sciences                                  Elsbeth Stern, ETH Zürich
   Meinrat O. Andreae, Max Planck Inst., Mainz                 Ernst Fehr, Univ. of Zurich                                  Laura Machesky, CRUK Beatson Inst. for Cancer Research            Ira Tabas, Columbia Univ.
   John A. Bargh, Yale Univ.                                   Tom Fenchel, Univ. of Copenhagen                             Andrew P. MacKenzie, Univ. of St Andrews                          Yoshiko Takahashi, Nara Inst. of Science and Technology
   Ben Barres, Stanford Medical School                         Alain Fischer, INSERM                                        Anne Magurran, Univ. of St Andrews                                John Thomas, Duke Univ.
   Jordi Bascompte, Estación Biológica de Doñana, CSIC         Wulfram Gerstner, EPFL Lausanne                              Oscar Marin, CSIC & Univ. Miguel Hernández                        Herbert Virgin, Washington Univ.
   Facundo Batista, London Research Inst.                      Karl-Heinz Glassmeier, TU Braunschweig                       Charles Marshall, Univ. of California, Berkeley                   Bert Vogelstein, Johns Hopkins Univ.
   Ray H. Baughman, Univ. of Texas, Dallas                     Diane Griffin, Johns Hopkins Bloomberg School of              Martin M. Matzuk, Baylor College of Medicine                      Cynthia Volkert, Univ. of Gottingen
   David Baum, Univ. of Wisconsin                                   Public Health                                           Graham Medley, Univ. of Warwick                                   Bruce D. Walker, Harvard Medical School
   Yasmine Belkaid, NIAID, NIH                                 Elizabeth Grove, Univ. of Chicago                            Yasushi Miyashita, Univ. of Tokyo                                 Douglas Wallace, Leibniz Inst. of Marine Sciences
   Philip Benfey, Duke Univ.                                   Taekjip Ha, Univ. of Illinois at Urbana-Champaign            Richard Morris, Univ. of Edinburgh                                Ian Walmsley, Univ. of Oxford
   Stephen J. Benkovic, Penn State Univ.                       Christian Haass, Ludwig Maximilians Univ.                    Edvard Moser, Norwegian Univ. of Science and Technology           David A. Wardle, Swedish Univ. of Agric Sciences
   Gregory C. Beroza, Stanford Univ.                           Steven Hahn, Fred Hutchinson Cancer Research Center          Sean Munro, MRC Lab. of Molecular Biology                         David Waxman, Fudan Univ.
   Peer Bork, EMBL                                             Gregory J. Hannon, Cold Spring Harbor Lab.                   Thomas Murray, The Hastings Center                                Detlef Weigel, Max Planck Inst., Tübingen
   Bernard Bourdon, Ecole Normale Superieure de Lyon           Martin Heimann, Max Planck Inst., Jena                       Naoto Nagaosa, Univ. of Tokyo                                     Jonathan Weissman, Univ. of California, San Francisco
   Ian Boyd, Univ. of St. Andrews                              Isaac Held, NOAA                                             James Nelson, Stanford Univ. School of Med.                       Sue Wessler, Univ. of California, Riverside
   Paul M. Brakefield, Univ. of Cambridge                       James A. Hendler, Rensselaer Polytechnic Inst.               Timothy W. Nilsen, Case Western Reserve Univ.                     Ian A. Wilson, The Scripps Res. Inst.
   Christian Büchel, Universitätsklinikum Hamburg-Eppendorf    Janet G. Hering, Swiss Fed. Inst. of Aquatic                 Pär Nordlund, Karolinska Inst.                                    Timothy D. Wilson, Univ. of Virginia
   Joseph A. Burns, Cornell Univ.                                   Science & Technology                                    Helga Nowotny, European Research Advisory Board                   Jan Zaanen, Leiden Univ.
   William P. Butz, Population Reference Bureau                Ray Hilborn, Univ. of Washington                             Luke O'Neill, Trinity College, Dublin                             Kenneth Zaret, Univ. of Penn. School of Medicine
   Gyorgy Buzsaki, Rutgers Univ.                               Michael E. Himmel, National Renewable Energy Lab.            Stuart H. Orkin, Dana-Farber Cancer Inst.                         Mayana Zatz, University of Sao Paolo
   Mats Carlsson, Univ. of Oslo                                Kai-Uwe Hinrichs, Univ. of Bremen                            Christine Ortiz, MIT                                              Jonathan Zehr, Univ. of California, Santa Cruz
   Mildred Cho, Stanford Univ.                                 Kei Hirose, Tokyo Inst. of Technology                        Elinor Ostrom, Indiana Univ.                                      Huda Zoghbi, Baylor College of Medicine
   David Clapham, Children’s Hospital, Boston                  David Hodell, Univ. of Cambridge                             Andrew Oswald, Univ. of Warwick                                   Maria Zuber, MIT
   David Clary, Univ. of Oxford                                Ove Hoegh-Guldberg, Univ. of Queensland                      Jane Parker, Max-Planck Inst. of Plant Breeding Research
   J. M. Claverie, CNRS, Marseille                             David Holden, Imperial College                               Donald R. Paul, Univ. of Texas at Austin
   Jonathan D. Cohen, Princeton Univ.                          Lora Hooper, UT Southwestern Medical Ctr at Dallas           P. David Pearson, Univ. of California, Berkeley                   BOOK REVIEW BOARD
   Robert Cook-Deegan, Duke Univ.                              Jeffrey A. Hubbell, EPFL Lausanne                            Reginald M. Penner, Univ. of California, Irvine                   John Aldrich, Duke Univ.
   Alan Cowman, Walter & Eliza Hall Inst.                      Steven Jacobsen, Univ. of California, Los Angeles            John H. J. Petrini, Memorial Sloan-Kettering Cancer Center        David Bloom, Harvard Univ.
   Robert H. Crabtree, Yale Univ.                              Kai Johnsson, EPFL Lausanne                                  Simon Phillpot, Univ. of Florida                                  Angela Creager, Princeton Univ.
   Wolfgang Cramer, Medit. Inst. for Ecology & Paleoecology    Peter Jonas, Universität Freiburg                            Philippe Poulin, CNRS                                             Richard Shweder, Univ. of Chicago
   F. Fleming Crim, Univ. of Wisconsin                         William Kaelin, Dana-Farber Cancer Inst.                     Colin Renfrew, Univ. of Cambridge                                 Ed Wasserman, DuPont
   Jeff L. Dangl, Univ. of North Carolina                      Barbara B. Kahn, Harvard Medical School                      Trevor Robbins, Univ. of Cambridge                                Lewis Wolpert, Univ. College London

                                                                               SCIENCE             VOL 334               14 OCTOBER 2011                                                                                    159
           NEWS OF THE WEEK




      Rockville, Maryland 1                                slots slated for 2017 and 2019. The two mis-
                                                           sions, dubbed Solar Orbiter and Euclid, were      Guinea gone. Jimmy Carter observes worm treatment.
      Task Force: Prostate Cancer Test                     selected 4 October and will now move for-
      Does More Harm Than Good                             ward to construction. There were three mis-       Mali, Ethiopia, Chad, and newly indepen-
      A test that millions of U.S. men take in the         sions in the running for the two launches; the    dent South Sudan, which accounts for 98%
      hope of nabbing cancer early—the prostate-           runner-up is PLATO, an observatory to study       of cases.
      specific antigen (PSA) test—got a bad (and            planet formation and the conditions for life,         Behavioral changes such as filtering
      highly controversial) grade this week. Based         which will be retained as a possibility for       drinking water and discouraging people
      on recent data, the U.S. Preventive Services         future launches.                                  with an emerging worm from walking into
      Task Force concluded that more harm than                 Solar Orbiter (pictured) will go closer       ponds and lakes have reduced cases from
      good comes from asking healthy men to get            to the sun than any previous mission, about       3.5 million in 1986 to 1797 in 2010. The
      a PSA test—and it shouldn’t be routinely             42 million kilometers away at its closest         Carter Center’s current goal is worldwide
      offered to patients. While a positive result         point, and will study how the sun interacts       eradication—defined as three consecutive
      can signal life-threatening cancer, it can also      with its environment. Euclid is a space tele-     years of no reported cases—by 2015.
      be a false alarm.                                    scope that will map the large-scale struc-
          Positive PSA results can lead not just to        tures of the universe with extreme accu-
      bleeding and infections from biopsies, but to        racy, trying to elucidate why the universe is     New Delhi 4
      “increased risk of incontinence and impo-            expanding at an ever-increasing rate.
      tence from prostatectomy and radiation ther-
                                                                                                             Center to Battle Future Food Crises
      apy,” and even “a small risk of death,” says         London 3                                          Hoping to kick-start another green revolu-
      Roger Chou, lead author of a review paper in                                                           tion, the Indian government on 5 October
      the Annals of Internal Medicine that backed
                                                           U.K. Pledges £20 Million                          announced the creation of a research cen-
      the task force. After The Cancer Letter              To Finish Off Guinea Worm                         ter to develop wheat and maize varieties
      leaked a draft of the Chou review on 6 Octo-         Twenty-five years after health workers             that thrive in warmer temperatures and on
      ber, the Annals posted it online. The review         started a campaign to rid the world of the        degraded land. Launched in partnership with
      noted that a large percentage of men in one          guinea worm, cases have been reduced by           the International Maize and Wheat Improve-
      trial had false positive results, but benefits        over 99%. But the 1800 or so cases that still     ment Centre (CIMMYT) in Mexico, the

                                                                                                                                                                  CREDITS (TOP TO BOTTOM): LOUISE GUBB/THE CARTER CENTER; ESA/SOHO
      were small or undetectable. The task force           occur annually are the hardest to address.        Borlaug Institute for South Asia will employ
      invited comment on its draft on 11 October.          At a press conference on 5 October, U.K.          300 researchers at three sites in India.
                                                           international development minister Stephen            The center’s establishment is a “momen-
      Paris 2                                              O’Brien announced that the government will        tous event in the history of global food secu-
                                                           donate about £20 million ($31 million) over       rity,” claims CIMMYT Director General
      Probing Dark and Light                               4 years to finish the eradication effort, led by   Thomas A. Lumpkin. South Asia’s popula-
      The European                                         the Carter Center in Atlanta—provided other       tion is expected to swell from 1.6 billion
      Space Agency                                         donors come forward with the remaining            today to 2.4 billion by 2050. By that time,
      (ESA) will launch                                    £40 million needed for the campaign.              CIMMYT predicts, almost a quarter of
      two missions to                                          The guinea worm is spread when people         South Asia’s wheat yield could be wiped out
      study the sun and                                    ingest its larvae through contaminated            by global warming.
      the mysterious                                       drinking water; the larvae incubate inside            The center honors the 1970 Peace Nobel
      dark energy that                                     the human host and worms up to 1 meter            laureate Norman Borlaug, a renowned wheat
      is speeding the                                      long emerge painfully through the skin a          breeder who helped lead the first green revo-
      universe’s expan-                                    year later. Once abundant across Africa and       lution in the 1960s and avert widespread
      sion in launch                                       South Asia, the parasite is now confined to        famine in India. The $125 million center will

160                                         14 OCTOBER 2011            VOL 334     SCIENCE
                                                                                                            take root near agricultural universities in the
                                                                                                            states of Punjab, Bihar, and Madhya Pradesh
                                                                                                            and is expected to open in 2 years.

                                                                                                            Malabo, Equatorial Guinea 5
                                                                                                            Dictator’s New Try for UNESCO
                                                                                                            Award Rebuffed
                                                                                                            The U.N. Educational, Scientific and Cul-
                                                                                                            tural Organization (UNESCO) has again
                                                                                                            deferred a decision on a proposal to establish
                                                                                                            an award for the life sciences funded by and
                                                                                                            named after Teodoro Obiang, the dictator of
                                                                                                            Equatorial Guinea. The prize was put on ice
                                                                                                                              indefinitely last year after
                                                                                                                              a global outcry by human
                                                                                                                              rights activists and scientists
                                                                                                                              protesting Obiang’s poor
                                                                                                                                                                   Glimpse of a Frozen World
                                                                                                                              human rights record. But it          British explorer Robert Falcon Scott’s arduous, ill-fated trek across Antarctica to the South
                                                                                                                              was up for discussion again          Pole is a famous saga. Less well known is the story of the last photographs he took, which
                                                                                                              Obiang          at UNESCO’s 58-member                have had a lengthy journey of their own. Last week, after nearly a century of being exhib-
                                                                                                                              Executive Board meeting              ited, fought over, auctioned off, and quietly tucked away in a photographic archive, many
                                                                                                            in Paris on 4 October, and this time Obiang            of Scott’s final photographs have been restored to the public eye with the publication of
                                                                                                            had the support of the African Union, which            The Lost Photographs of Captain Scott, a collaboration between the photographs’ current
                                                                                                            he currently chairs. In an uncharacteristically        owner, antiquarian Richard Kossow, and polar historian David Wilson, the great nephew of
                                                                                                            outspoken speech, UNESCO’s Director-                   one of Scott’s team members, Edward Wilson.
                                                                                                            General Irina Bokova called on Obiang to                   Scott, Wilson, and three other team members reached the South Pole in January 1912
                                                                                                            stop pushing for the award, which she said             only to find that Norwegian explorer Roald Amundsen had arrived there first. Dejected,
                                                                                                            could put UNESCO at “war with the scien-               they began the journey back across the Ross Ice Shelf. They didn’t make it: By the end of
                                                                                                            tific commmunity.” The issue will be decided            March 1912, all five men were dead. Their bodies and records—including Scott’s final rolls
                                                                                                            at the board’s next meeting in April.                  of film—were found in November of that year.

                                                                                                            Pyongyang 6
                                                                                                                                                                Neureiter, a senior policy adviser at AAAS        than 90%, and traps didn’t yield a single
                                                                                                            North Korean University Hosts                       (Science’s publisher). “For several Ameri-        infected mosquito, researchers reported last
                                                                                                            International Science Conference                    cans and Europeans to be in a room and            week in The American Journal of Tropical
                                                                                                            A novel experiment in higher education in           speak directly to some 250 North Korean           Medicine and Hygiene.
                                                                                                            North Korea celebrated its second anniver-          students is an unprecedented event,” he says.
                                                                                                            sary last week with a fete featuring a Nobel
                                                                                                            laureate, a former astronaut, and a member of       Cotonou, Benin 7
                                                                                                            U.K. House of Lords. The first international
                                                                                                            conference of Pyongyang University of Sci-
                                                                                                                                                                Backup Insecticide Shows
                                                                                                            ence and Technology (PUST) in the North             Promise Against Malaria
                                                                                                            Korean capital was an academic milestone.           Indoor spraying of a compound called ben-
                                                                                                            And as it turns out, PUST is in a unique posi-      diocarb may provide a backup now that
                                                                                                            tion: Other universities in Pyongyang are said      malaria mosquitoes are becoming resistant to
                                                                                                            to have cancelled classes for several months        mainstay insecticides, a large study in Benin
                                                                                                            to free students to work on city-beautifying        suggests. But scientists say more alternatives
                                                                                                            projects to commemorate the 100th anni-             are urgently needed—if only because of ben-       Net gain. A treated mosquito net drying in Benin.
                                                                                                            versary of the birth of the country’s founder,      diocarb’s toxicity.
                                                                                                            Kim Il Sung, next April.                               Bendiocarb is a candidate replacement              Bendiocarb, which inhibits a brain
                                                                                                                Since opening its doors 2 years ago             for pyrethroids, compounds whose wide-            enzyme called acetylcholinesterase, is
                                                                                                            (Science, 25 September 2009, p. 1610),              spread use in indoor spraying and bednets         among a dozen insecticides approved by the
                                                                                                            PUST has enrolled fewer students than               is causing resistance in Anopheles mos-           World Health Organization for malaria con-
                                                                                                            expected and the campus is still working            quitoes, the vector for malaria, especially       trol, but safety concerns led manufacturers
                                                                                                            to provide unfettered Internet access. Still,       in West Africa. After government teams            to voluntarily withdraw it from the U.S. mar-
                                                                                                            to the two dozen scholars who trekked to            carried out two rounds of spraying in 2008        ket in 1999. Beninese teams did not spray
                                                                                                            Pyongyang last week, PUST’s existence is            and 2009 in an area in Benin where 350,000        in low-lying areas prone to flooding, where
                                                                                                            grounds for optimism, says attendee Norman          people live, Anopheles bites fell by more         they feared toxic run-off into local waters.

                                                                                                                                               SCIENCE       VOL 334     14 OCTOBER 2011                                                  161
           NEWS OF THE WEEK

         21 Researchers awarded $50,000                   Random Sample
         grants in the first round of the
         National Science Foundation’s                    Survival of the Palimpsest
                                                          The Greek genius Archimedes was a pro-
         (NSF’s) new Innovation Corps Pro-                lific inventor and writer: He designed screw
         gram. NSF hopes to help 100 inves-               pumps and heat rays, approximated pi
         tigators a year assess the commer-                                     using a method that
         cial potential of their research, start-                               resembles modern
                                                                                integral calculus, and
         ing with an entrepreneur boot camp                                     wrote mathematical        Natural light.
         this week at Stanford University.                                      treatises on spheres,
                                                                                cylinders, and levers.
         17.4% Early estimate of the                                            But a large body of
         “success rate” for reviewed research                                   his work, known as the
         grant proposals at the National                                        Archimedes Palimp-
                                                                                sest, until recently
         Institutes of Health for the year that
                                                          remained tantalizingly just out of reach.
         ended 30 September. Last year’s rate                 In 1906, classical scholar Johan
         was 21%.                                         Heiberg discovered that beneath the Latin
                                                          writing of a 1229 prayer book was a much
                                                          older Greek text by Archimedes. Heiberg
      FINDINGS                                            transcribed what he could, but much of the
                                                          writing was indecipherable.
      Dark Ice on a Hot Planet                                Enter modern technology. In 1998, a           Ultraviolet light.
      Despite basking in the sun’s fiery glow,             private collector bought the manuscript
      tiny Mercury, the innermost planet in our           and lent it to the Walters Art Museum in
      solar system, is probably home to extensive         Baltimore, Maryland, where scholars began to unravel its secrets. Researchers led by
      ice fields. Twenty years ago, radar obser-           William Noel, the curator of manuscripts at the museum, used multispectral analysis and
      vations from Earth revealed small, highly           other techniques to delve through layers of Latin writing and even 20th century forgeries
      reflective areas close to Mercury’s poles,           to reveal the Greek texts, imaging the pages through 14 different wavelengths of light
      suggesting the presence of ice. Now, NASA’s         ranging from infrared to ultraviolet. They also took the text to the Stanford Synchrotron
      MESSENGER spacecraft, which has orbited             Radiation Lightsource for x-ray fluorescence, which helped reveal much of the iron-laden
      Mercury since March, has confirmed that              ink of the “under” text.
      these radar-bright patches coincide with                The prayer book had been written over seven treatises by Archimedes, including two
      deep craters near the poles that never receive      found only in the Palimpsest: The Stomachion, the earliest known study in a branch of

                                                                                                                                                            CREDITS (TOP TO BOTTOM): © THE OWNER OF THE ARCHIMEDES PALIMPSEST (3); NANCY CHABOT/JHU-APL
      sunlight. This color-coded photo mosaic             mathematics called combinatorics; and The Method of Mechanical Theorems, in which
      of Mercury’s south polar region, presented          Archimedes “was essentially calculating with infinity,” Noel says.
      5 October at a joint meeting of the European            The Walters Art Museum’s exhibit Lost and Found: The Secrets of Archimedes will
      Planetary Science Congress and the Divi-            display the results of this decade-long project from 16 October through 1 January 2012.
      sion for Planetary Sciences of the American
      Astronomical Society, shows these “freezer”
      areas as dark blotches. According to
      MESSENGER instrument scientist Nancy                                                               trodes into a monkey’s brain: one set in the
                                                       Monkeys Control Virtual Limbs
      Chabot of the Johns Hopkins University’s                                                           motor control center, and the other in the
      Applied Physics Laboratory in Laurel, Mary-      With Their Minds                                  part of the somatosensory cortex that pro-
      land, one-fifth of the region within 200 kilo-    Brain-controlled prosthetics that enable a        cesses the sensation of physical touch either
      meters of Mercury’s south pole is in perma-      person to, say, pick up a pencil continue to      from the left hand or the leg. Using the first
      nent shadow. “It’s all consistent with there     improve for amputees, but limbs that can          set, the monkey could control a virtual mon-
      being water ice,” she says.                      actually feel touch sensations have remained      key arm on a computer screen and sweep
                                                       a challenge. Now researchers led by neuro-        the hand over virtual disks with different
                                                       scientist Miguel Nicolelis of Duke Univer-        “textures.” The second set of electrodes fed
                                                       sity in Durham, North Carolina, have created      a series of electrical pulses (low frequency
                                                       a virtual prosthetic arm that monkeys con-        for rough texture, high frequency for fine
                                                       trol using only their minds, and that enables     texture) into the touch center of its brain.
                                                       them to feel virtual textures.                    With little training, the monkeys could con-
                                                           To “close the loop” between control-          sistently distinguish the textures as if the arm
                                                       ling a limb and feeling physical touch, the       was their own, the team reported 5 October
                                                       researchers implanted two sets of tiny elec-      in Nature.

162                                       14 OCTOBER 2011        VOL 334      SCIENCE
                                                                                                                                       NEWS & ANALYSIS

                                                                                                                                                                  Big push. Gulf Coast lawmakers want oil spill
                                                                                                                                                                  fines used for restoration, such as efforts to return
                                                                                                                                                                  Mississippi sediment (yellow, above) to wetlands.

                                                             GULF OIL SPILL                                                                                       siana, Mississippi, and Texas, the panel held a
                                                                                                                                                                  series of public meetings and tapped govern-
                                                             Panel Draws Ambitious Road                                                                           ment scientists involved in previous efforts to
                                                                                                                                                                  prepare gulf restoration road maps (Science,
                                                                                                                                                                  25 June 2010, p. 1618).
                                                             Map for Gulf Restoration                                                                                 The result is a set of sweeping but often
                                                                                                                                                                  general recommendations aimed at bolster-
                                                                                                                                                                  ing both the science and the political sup-
                                                             In Bayou Dupont, a windswept marshland            gulf—and are getting a generally warm              port needed to tackle restoration challenges,
                                                             about 15 kilometers south of New Orleans,         reception from environmentalists and               including reversing wetland loss and shrink-
                                                             Louisiana, engineers have been using a giant      researchers. The plan recognizes “the vital        ing the size of the oxygen-poor “dead zone”
                                                             pipe to spill mud mined from the bottom of        need for funding to advance restoration,”          that forms off the mouth of the Mississippi
                                                             the nearby Mississippi River into a shal-         says Angelina Freeman, a coastal scientist         each year. “It is more of a plan for a plan rather
                                                             low stretch of open water. The goal of the        with the Environmental Defense Fund in             than an action plan,” says Larry McKinney,
                                                             2-year-old, nearly $30 million state- and fed-    Washington, D.C. But “it takes tremendous          director of the Harte Research Institute for
                                                             erally funded experiment: to rebuild nearly       effort to go from a report to actual projects,”    Gulf of Mexico Studies at Texas A&M Uni-
                                                             200 hectares of wetlands that were starved        says Robert Twilley, an ecologist at Louisi-       versity in Corpus Christi. That’s “understand-
                                                             of nourishing sediment after engineers built      ana State University in Baton Rouge, who is        able,” he says, given “the scale of the task.”
                                                             flood-control levees along the river decades       involved in wetland restoration.                       One overarching need, the report con-
                                                             ago. Now, a presidential task force is calling       Some worry that the plan, which includes        cludes, is to place ecosystem restoration on an
                                                             for dramatically expanding such aggressive        recommendations for 19 “major actions,”            “equal footing” with other gulf region priori-
                                                             efforts to restore battered Gulf of Mexico        isn’t specific enough and could create the         ties, such as improving navigation and reduc-
                                                             ecosystems—and using billions of dollars          kind of bureaucratic quagmire that has hob-        ing floods. It also calls for steps to increase
                                                             in fines resulting from the 2010 Deepwater         bled other megarestoration projects. Others        the “resilience” of both human and ecological
                                                             Horizon oil spill to pay for them.                are concerned that the complex funding for-        communities, such as making better “strategic

                                                                 “Unless bold and broad-scale measures         mula envisioned by Congress—which would            use” of the millions of tons of sediment that
                                                             are taken soon, the health and future of the      apportion money to the five Gulf Coast states       flow down the Mississippi. That could mean
                                                             Gulf will remain in jeopardy,” the Gulf Coast     based on factors such as their distance from       cutting new outlets in levees to allow the river
                                                             Ecosystem Restoration Task Force con-             the doomed Deepwater well—could end                to return to old floodplains, or ramping up
                                                             cluded in a draft restoration strategy released   up funneling too much money to economic            efforts—like the one in Bayou Dupont—to
                                                             5 October. The same day, key members of           development efforts, such as tourist facilities,   pump mud to where it is needed.
                                                             Congress introduced bipartisan legislation        that don’t address the gulf’s underlying eco-          The need for better science gets plenty of
                                                             that, if passed, would steer 80% of oil spill     logical problems.                                  ink, including calls for more comprehensive
                                                             fines—which could add up to more than                The White House created the restoration         gulf monitoring. But “the dire state of many
                                                             $21 billion—to ecological and economic            panel last October at the recommendation of        elements of the Gulf ecosystem cannot wait
                                                             restoration efforts along the Gulf Coast and      the blue-ribbon committee that investigated        for scientific certainty and demand immedi-
                                                             set up five new regional science centers.          the spill. Composed of senior officials from        ate action,” the report says. To avoid delays, it
                                                                 Both developments have added momen-           federal environmental agencies and the five         proposes an “adaptive management” process
                                                             tum to efforts to do something big in the         Gulf Coast states of Alabama, Florida, Loui-       of “learning by doing, wherein flexibility is

                                                                                              SCIENCE       VOL 334      14 OCTOBER 2011                                                     163

      built into projects” to change course based on          Privately, some environmental activ-            which could range from $5.4 billion to $21.1
      new science.                                        ists are unhappy that the task force stopped        billion—into restoration. Both would send
          That’s a solid strategy, says ecologist Lance   short of identifying—and using its clout to         35% of the money directly to the five states and
      Gunderson of Emory University in Atlanta,           jump-start—specific restoration projects.           60% to the new restoration council, although
      who has closely studied other large, expensive      Instead, it recommends that Congress create         the House bill would give states a greater say
      efforts to restore complex ecosystems such as       a new regional restoration council that would       in how the council uses its funding. A 5%
      the Florida Everglades and the Colorado River.      approve project plans and hand out some             share would go to the National Oceanic and
      But “adaptive management is not a panacea,”         funding. Such a council would be established        Atmospheric Administration to set up research
      he says, adding that it appears the task force      by the legislation (H.R. 3096) introduced last      centers in each state, but the House bill would
      has adopted a relatively “weak” definition of        week in the House of Representatives by key         close the centers when the money runs out.
      the concept that may not allow enough room          gulf state lawmakers; a nearly identical bill (S.       It’s not yet clear if both the House and
      for trying bold experiments and learning from       1400) was approved last month by the Sen-           the Senate will be able to pass their bills and
      failure. Many government agencies “say that         ate Committee on Environment and Public             agree on a single version this year. Comments
      they are doing [adaptive management] when           Works. Both versions would funnel 80% of            on the task force report, meanwhile, are due
      they don’t have a clue,” he says.                   the Clean Water Act fines levied for the spill—      26 October.                  –DAVID MALAKOFF


      Teasing Out Cause and Effect in Macroeconomics
      Did the U.S. government’s decision to keep of how other parties will behave in the future. tion before changing to policies aimed at
      interest rates down lead to increased home For example, what companies expect policy- lowering inflation. Sargent’s work showed
      buying in the United States in the middle makers to do can influence how much they how the public and banks adjusted to the
      of the past decade? Or did the boom in the invest in a new business venture. Similarly, change through a gradual learning process,
      housing market during those years result when governments institute policy measures explaining why it took a long time for infla-
      from people’s expectations that the govern- such as a tax increase, they are influenced by tion to come down.
      ment would not raise interest rates? Sort- how they expect markets, investors, and con-                Soon after the prize was announced, both
      ing out such tangles of cause and effect in sumers to react.                                       laureates found themselves fielding ques-
      the world of macroeconomics is a messy             Sims and Sargent built on this idea of tions about lessons their research might
      business. But if you do it well, you might “rational expectations” to develop methods hold for a world engulfed by a growing eco-
      win a Nobel Prize in eco-                                                                                                nomic crisis. Both had a
      nomics—as Christopher                                                                                                    guarded response, refus-
      Sims and Thomas Sargent
      have done.                                                         N BEL                                                 ing to be drawn into poli-
                                                                                                                               tics. “There is no simple
          Sims is a professor at
      Princeton University; Sar-
                                                                         PRIZE                                                 answer; it requires a lot of
                                                                                                                               slow work looking at the
      gent is a professor at New                                             2011                                              data,” Sims told report-
      York University (NYU).                                                                                                   ers, speaking by phone
      They were once fellow                                           ECONOMICS                                                from his home in Princ-
      graduate students at Har-                                                                                                eton, New Jersey, during
      vard University and each                                                                                                 a press conference held in
      earned a Ph.D. in 1968;                                                                                                  Stockholm to announce
                                     CHRISTOPHER SIMS                                        THOMAS SARGENT
      later, they spent several                                                                                                the prize. “The meth-
      years as colleagues at the Chicken or egg? Sims and Sargent have independently studied the interplay between govern- ods that I have used and
      University of Minnesota. ment policies and the forecasts of economic actors—work with immediate practical applications. that Tom has developed
      But they did their award-                                                                                                are central to finding our
      winning work independently. Sargent and that help identify causal links between pol- way out of this mess.” In an interview with
      Sims will share the $1.5 million prize (for- icy changes and economic events on the one the Nobel Foundation, Sargent described
                                                                                                                                                              CREDITS: DENISE APPLEWHITE/PRINCETON UNIVERSITY

      mally known as the Sveriges Riksbank Prize hand, and macroeconomic indicators such as himself and his co-winner thus: “We’re just
      in Economic Sciences) for their “empirical gross domestic product, inflation, and unem- bookish types that look at numbers and try to
      research on cause and effect in the macro- ployment on the other.                                  figure out what’s going on.”
      economy,” according to an announcement             Sims focused on how the economy is                  NYU economist David Backus says
      from the Nobel Foundation.                     affected by short-term changes in economic Sargent and Sims have had a profound influ-
          Most causal relationships in the physical policy and events such as an increase in the ence on hundreds of students. Both are known
      world are unambiguous: Push a ball and it interest rate or a spike in oil prices. Sargent for their devotion to mentoring, Backus says.
      moves. By contrast, the relationship between helped understand how systematic changes Even today, their “lines of research are at the
      policy measures such as tax cuts and the in economic policy have led to long-term forefront of modern economics, something
      health of the economy is often a two-way impacts. For instance, he studied how, in the of a rarity when prizes are awarded for work
      street: Individuals, businesses, and govern- decades following World War II, many coun- done decades ago.”
      ments make decisions based on expectations tries implemented policies that raised infla-                                 –YUDHIJIT BHATTACHARJEE

164                                          14 OCTOBER 2011          VOL 334     SCIENCE

                                   N O B E L P R I Z E I N C H E M I ST RY

                                   Once-Ridiculed Discovery Redefined the Term Crystal
                                   Daniel Shechtman has the last laugh. In 1982,        cule. His claims caused such embarrassment            That set a number of chemists thinking
                                   Shechtman, of the Technion-Israel Institute of       that his boss asked him to leave the research     about whether atoms could adopt a similar
                                   Technology in Haifa, discovered an alloy of          group. “That was the atmosphere at [NIST],”       pattern. Crystallographer Alan Mackay built
                                   aluminum and manganese that appeared to              Shechtman says. But he persevered with the        a model with circles representing atoms at the
                                   have fivefold symmetry: that is, the atoms in it      help of a few colleagues, and when he finally      corners of Penrose’s tiles and calculated that
                                   formed a pattern that appeared essentially the       published in Physical Review Letters in           it would produce a diffraction pattern with
                                   same when rotated by a fifth of a turn, or 72˚.       November 1984, “then all hell broke loose.”       10-fold symmetry. Paul Steinhardt, then at the
                                   Other researchers scoffed, as such an arrange-       It was a simple experiment that other labs        University of Pennsylvania, and his student
                                   ment was thought to be mathematically                could repeat in a matter of days, and his phone   Dov Levine had also been devising theoreti-
                                   impossible. Yet scientists eventually realized       started ringing off the hook.                     cal structures based on Penrose tiling. When
                                   that atoms in a solid can achieve such symme-            The diffraction patterns may have been        a colleague showed Steinhardt a preprint of
                                   try by arranging themselves in a pattern that        real, but how were the atoms arranged? The        Shechtman’s first paper in autumn 1984, “I
                                   almost but never quite repeats—a “quasicrys-         answer came from mathematicians who had           leapt up in the air. The two matched by eye
                                   tal.” Shechtman’s discovery has now gone full                                                          beautifully,” says Steinhardt, who is now at
                                   circle, from ridicule to ultimate accolade: It                                                         Princeton University. Steinhardt and Levine
                                   has netted this year’s Nobel Prize in chemistry.                                                       published a paper shortly after Shechtman’s
                                       “He does deserve a Nobel Prize for ush-                                                            linking his observations to Penrose-like struc-
                                   ering in this new kind of phase in chemistry:                                                          tures and coined the term “quasicrystal.” Not
                                   crystals that are not crystals,” says mathe-                                                           everyone was convinced. X-ray crystallog-
                                   matician Roger Penrose of the University of                                                            raphers did not accept it for 3 years, until a
                                   Oxford in the United Kingdom, who played                                                               quasicrystal could be grown big enough to
                                   an indirect role in explaining the materials.                                                          perform x-ray diffraction. “That was the turn-
                                       Prior to Shechtman’s discovery, a crystal                                                          ing point,” Shechtman says, that led the Inter-
                                   was defined as a material in which atoms are                                                            national Union of Crystallography in 1992 to
                                   arranged in a regular pattern that repeats itself.                                                     change its definition of a crystal from a regu-
                                   That definition puts limits on the symmetry                                                             lar repeating array of atoms to “any solid hav-
                                   a crystal can have, as a simple child’s game                                                           ing an essentially discrete diffraction pattern.”
                                   shows. Suppose you want to cover a tabletop                                                                But double Nobelist Linus Pauling, a dom-
                                   with identical tiles. A pattern of triangles does     DANIEL SHECHTMAN                                 inant figure among U.S. chemists who died in
                                   the trick, so it’s possible to make crystals with                                                      1994, never accepted quasicrystals, despite
                                   threefold symmetry. Squares or hexagons also                                                           Shechtman traveling to his lab in Palo Alto
                                   work, so crystals with fourfold and sixfold
                                   symmetry can also be made. But pentagons
                                                                                                                     N BEL                and giving him a personal hourlong lecture.
                                                                                                                                              Much remains mysterious about quasi-
                                   won’t work; there will always be gaps between                                     PRIZE                crystals, including how such complex long-
                                   them. Thus, fivefold symmetry is impossible.
                                       Nevertheless, Shechtman saw what he
                                                                                                                         2011             range structures can form from single atoms.
                                                                                                                                          “They can’t be produced simply with local
                                   saw. On 8 April 1982, while working at the                                                             rules; there has to be some subtle kind of
                                   National Institute of Standards and Technol-                                        CHEMISTRY          production,” Penrose says. Predicting their
                                   ogy (NIST) in Gaithersburg, Maryland, he                                                               properties is also hard, Steinhardt says: “The
                                   quickly cooled a sample of aluminum and                                                                mathematical techniques we use on crystals
                                   manganese alloy to keep it from crystallizing        Irregular beauty. Atomic model of an aluminum-    don’t work on quasicrystals.”
                                   and then fired a beam of electrons into it. If        palladium-manganese quasicrystal.                     Quasicrystals have been found in nature,
                                   there was an orderly arrangement of atoms in                                                           in a mineral reported to have come from
                                   the material, the electrons would “diffract” off     been thinking about patterns of tiles. Some       the Koryak Mountains in eastern Russia
                                   the various planes of atoms in it and emerge         had been puzzling over curious mosaics that       (Science, 5 June 2009, p. 1306). They have
                                   at specific angles to produce a recognizable          have a limited number of different-shaped         also been found in one of the world’s most
                                   pattern in a detector. Shechtman saw a dif-          tiles and that fitted together in patterns that    durable steels, made by a company in Sweden
                                   fraction pattern unlike any he’d seen before:        never repeated themselves. Such mosaics           for razor blades and surgical needles. They
                                   concentric circles of 10 bright dots. The tal-       were used by Arabic artists as early as the

                                                                                                                                          are beginning to find other industrial appli-
                                   lies pointed to an impossible symmetry. “10          13th century to decorate buildings such as the    cations, such as nonstick coatings in pans,
                                   fold???” he recorded in his notebook.                Alhambra Palace in Granada, Spain. Math-          heat insulation in engines, and thermoelec-
                                       Shechtman says he was convinced on the           ematicians in the 1960s and ’70s strove to        tric materials to salvage waste heat. “If there
                                   first day, but he checked and rechecked his           find the smallest number of tiles that could       is one particular lesson we are taking from
                                   experiment and tried others over the next            produce such a pattern. In the mid-1970s,         [Shechtman’s] research, it is not to underesti-
                                   week to investigate the material further.            Penrose came up with a set of just two rhom-      mate the imagination of nature herself,”Andrew
                                   When he finally told colleagues about his dis-        buses that did the job. Penrose’s pattern had     Goodwin of the University of Oxford said in
                                   covery, he was met with dismissal and ridi-          plenty of pentagons and decagons.                 a statement.                   –DANIEL CLERY

                                                                       SCIENCE       VOL 334     14 OCTOBER 2011                                                  165


      HIV Study Renews Scrutiny of Hormonal Contraception
      In the media world, the most precious real
      estate is page one, above the fold of The New
      York Times, which on 4 October featured this
      headline: “Contraceptive Said to Double
      Risks of H.I.V.: Study’s Findings Pose Quan-
      dary in Africa.” In the first quote in the story,
      a foreign policy expert said, in essence, that
      if the new findings about injectable hor-
      monal contraceptives used by many African
      women were true, “we have a major health
      crisis on our hands.”
          The Times story posed a quandary for
      some HIV/AIDS researchers who study                                                                   Top story. An HIV transmission study conducted
      female contraceptives and HIV risks, a                                                                at sites in Kenya and in six other African coun-
                                                                                                            tries again questions the safety of Depo-Provera.
      research niche that has a long history of con-
      flicting reports. For many poor women in
      sub-Saharan Africa, HIV and pregnancy are          by a team from the University of Washing-              Charles Morrison, an epidemiologist
      “competing risks,” as having a child presents      ton (UW), Seattle, the study aimed to assess       at FHI360 (formerly Family Health Inter-
      both health and socioeconomic downsides.           whether treating HSV-2 with acyclovir could        national) in Durham, North Carolina,
      Evaluating the contribution that injectable        thwart HIV transmission—it did not—and             co-authored a commentary that said the risk
      hormones make to HIV transmission is also          couples used whatever contraception they           of HIV transmission and hormonal contra-
      fraught with confounding variables that com-       chose. As the authors note in the paper pub-       ception remains “an unanswered question.”
      plicate analyses of cause and effect.              lished online by The Lancet Infectious Dis-        The commentary notes that five of 12 pub-
          Although the Times attempted to put the        eases on 4 October, the contraceptive use          lished studies—all in women at high risk for
      new findings in context and noted that they         was “observational”: It was not a definitive        infection—have linked hormonal contra-
      had limitations, some researchers worry            trial of women randomly assigned to differ-        ception to transmission. So did a persuasive
      that the global influence of the newspaper’s        ent contraceptives. The researchers note that      experiment in monkeys published in Nature
      account will overshadow the many complexi-         self-reporting could also bias results: Women      Medicine 15 years ago. He characterizes the
      ties in the data that leave open critical ques-    using hormonal contraceptives, for exam-           new study as a “decent secondary analysis”
      tions about whether injected contraceptives        ple, may have used condoms less often than         and says it “does add to our suspicion level,
      truly pose an HIV risk.                            claimed and thus had more exposure to HIV     .    but it’s by no means conclusive.”
          Epidemiologists Maria Wawer and                    Roughly two-thirds of the HIV-infected             Morrison’s own large study—which doc-
      Ronald Gray, a husband-and-wife team at            partners at the study’s outset were men, and of    umented 76 infections in women who used
      the Johns Hopkins Bloomberg School of              the 1314 uninfected women, about 15% used          injectable hormones—found an increased
      Public Health in Baltimore, Maryland, for          an injectable contraceptive (mainly Depo-          risk only when his team did a fine-grained
      the most part applaud the study featured by        Provera) at some point during the 6-year           analysis and discovered higher transmis-
      the Times. “They’re a good group, and it’s a       study. Only 10 of the women using injectable       sion in younger and HSV-2 negative women.
      good study,” says Wawer, who with Gray has         hormones became infected, which did not            “This is a tough area,” he says. “Sexual behav-
      evaluated injectable hormonal contracep-           reach statistical significance until research-      ior, contraception, pregnancy, and HIV trans-
      tives and HIV risks in Uganda. But their own       ers adjusted the data for variables including      mission are all correlated, and it becomes
      study is one of many that found no increased       age, HIV concentration in infected partners,       quite hard to tease them all apart.”
      risk from these contraceptives in women            and pregnancy, which itself increases the              Baeten and his team well recognize the
      they followed for 5 years. They further note       risk of infection. (Oral contraceptive use was     pregnancy risks. “I wouldn’t want our results
      that the Times story did not point out that        extremely low and did not lead to significant       to be interpreted as an excuse to dimin-
      the new study had very small numbers of            results.) In the adjusted analysis, the women      ish contraceptive use in the world,” he says.
      infected women, analytic issues, and possi-        who used the injectable hormones had twice         “They should reinforce something that may
      bly misleading information because of self-        the risk of becoming infected than those who       not be pushed as hard as it should be: Counsel
      reporting by participants about frequency          did not, with an annual incidence of infection     women that surely hormonal contraception
      of sex and condom use. “I’m afraid this is         of 6.85% versus 3.78%. A separate, adjusted        doesn’t decrease risk and may increase it and
      going to set family planning in Africa back        analysis of uninfected men found that having       should be used along with condoms.”
      a decade,” says Wawer, who worries that            sex with an HIV-infected woman who used                In response to the new findings, the World
      African media and health ministers will dis-       the injectable hormones at least doubled the       Health Organization has convened a “techni-
                                                                                                                                                                CREDIT: NELLY MUGO

      cuss the Times report more than the new data.      risk of transmission. “It was really hard news     cal consultation” to begin 31 January 2012
          The study followed 3790 couples in seven       for me to hear, and it’s really hard to report,”   with experts to “re-examine the totality of the
      African countries in which one partner at the      says the study’s leader, UW Seattle epidemi-       evidence” to see whether it needs to revise
      outset had a known infection with both HIV         ologist Jared Baeten. “That said, wishing it       recommendations about hormonal contra-
      and herpes simplex virus-2 (HSV-2). Led            away doesn’t help the situation.”                  ceptive use in women.              –JON COHEN

166                                         14 OCTOBER 2011          VOL 334      SCIENCE

                                         H U M A N E VO L U T I O N                                                                                 Just what archaic human did the ances-
                                                                                                                                                tors of these Africans mate with? Geneti-

                                         African Data Bolster New View                                                                          cists can’t say because they have no ancient
                                                                                                                                                DNA from African fossils. And there are no
                                                                                                                                                fossils of Neandertals or Denisovans on the
                                         Of Modern Human Origins                                                                                continent. But fossil hunters have long puz-
                                                                                                                                                zled over a few strange African bones that
                                         It’s impossible to pinpoint the moment          Asia with Neandertals, and a separate inter-           carry both archaic and modern features. In
                                         when a paradigm shifts. But when it comes       mingling with Denisovans may have taken                a PLoS ONE paper published a week after
                                         to views of the origin of Homo sapiens,         place in Asia (Science, 23 September, 1689).           Hammer’s, a team led by Stringer and
                                         last month may be as good a time as any:            Now, a team led by Hammer has found                Katerina Harvati of the Eberhard Karls Uni-
                                         That’s when two papers independently sug-       that some hunter-gatherers in Africa also              versity of Tübingen in Germany took a new
                                         gested that early H. sapiens interbred with     carry unusual segments of DNA that the                 look at the cranium of a modern human found
                                         now-extinct forms of humans in Africa, so       researchers propose are archaic, as reported           in 1965 at Nigeria’s Iwo Eleru rock shelter.
                                         that some living Africans carry genes from      last month in the Proceedings of the National              Stringer had studied these bones for
                                         archaic people, just as all Europeans and       Academy of Sciences (PNAS). The team                   his Ph.D. thesis and had always wondered
                                         Asians recently have been shown to do.          screened 61 regions of DNA from three rela-            about them, for they didn’t fit the theory of
                                         The new data imply that there were at least     tively isolated groups in sub-Saharan Africa,          replacement: They had a thick, archaic brow
                                         three fruitful encounters between H. sapi-      examining noncoding DNA that was less                  ridge, but the radiocarbon dating of the time
                                         ens and archaic species. And one of the new     likely to be influenced by natural selection.           had put them at only about 13,000 years
                                         papers, on African fossils, was co-authored     They found three regions whose pattern of              old. With respect to replacement, “it was an
                                         by paleoanthropologist Chris Stringer of        alleles varied widely among individuals—a              anomaly,” Stringer says. “We didn’t expect to
                                         the Natural History Museum in London,           sign of great antiquity. But the variants were         have archaic people hanging around so long
                                         once the leading advocate of a conflicting       linked to each other across extended regions           after the appearance of modern humans.”
                                         hypothesis that H. sapiens simply replaced,     of the chromosome, which suggested they                    Harvati analyzed the skull quantita-
                                         rather than mated with, archaic peoples in      were inherited recently, because such asso-            tively and found that in addition to the thick
                                                                                                                                                brow ridge and robust jaw, the cranium is
                                                                                                                                                elongated from front to back, transitional
                                                                                                                                                between the shape seen in archaic humans
                                                                                                                                                like Neandertals and the globular H. sapi-
                                                                                                                                                ens skulls. New uranium-series dating on
                                                                                                                                                a piece of the bone, done by Rainer Grun
                                                                                                                                                of the Australian National University in
                                                                                                                                                Canberra, confirmed it is only 11,700 to
                                                                                                                                                16,300 years old.
                                                                                                                                                    The Iwo Eleru bones “could be a relic of
                                                                                                                                                ancient people” who went extinct, Stringer
                                                                                                                                                says, or a descendant of mating between
                                                                                                 Ancient head lines. The shape of this mod-     modern and archaic humans. The speci-
                                                                                                 ern human’s skull is elongated from front to   mens haven’t yielded any ancient DNA. But
                                                                                                 back (left), a trait seen in archaic humans.   researchers are already studying archaic
                                                                                                                                                features in other H. sapiens fossils. “I think
                                         Africa, Asia, and Europe. Says Stringer         ciations among chromosomal regions break               there’s more to come,” Hammer says.
                                         now: “Africa doesn’t have a simple story of     down over time. So the overall pattern sug-                Indeed, “if interbreeding happened out-
                                         modern humans appearing and everything          gests that ancient segments of DNA recently            side of Africa,” as the complete genomes
                                         else disappearing.”                             entered the H. sapiens genome, as would be             of Neandertals and Denisovans suggest,
                                            That marks a turning point in views          the case if the DNA was inherited from an              “it is quite likely it also happened within
                                         of modern human origins, says geneti-           archaic human.                                         Africa,” says population geneticist Laurent
                                         cist Michael Hammer of the University of           The team screened for these three vari-             Excoffier of the University of Bern in Swit-
                                         Arizona in Tucson, lead author of another       ants in an expanded sample of 500 Africans             zerland. But he says Hammer’s team needs
                                         paper on African genetics. “There is a para-    and found them in pygmy and nonpygmy                   to do more modeling to see if conditions

                                         digm shift,” he says. “The burden of proof is   populations from Zaire, Cameroon, and                  such as population bottlenecks could pro-
                                         now on those who want to claim complete         the Democratic Republic of the Congo. By               duce the same genetic patterns without
                                         replacement. Two years ago, it was on those     using computational and statistical meth-              interbreeding. Geneticist Sarah Tishkoff of
                                         who wanted to claim [mixing].”                  ods to simulate the distribution of the vari-          the University of Pennsylvania would like
                                            After sequencing the genome of Nean-         ants, Hammer’s team concluded that modern              whole-genome analyses, which Hammer
                                         dertals and that of mysterious hominins         humans mated with archaic humans perhaps               says are in the works. Still, Tishkoff says, “it
                                         from Denisova Cave in Siberia, research-        as recently as 35,000 years ago in central             looks increasingly possible that there’s been
                                         ers concluded last year that our ancestors      Africa. Hammer stresses that it was a small            some low level of admixture” in Africa as
                                         interbred with both Neandertals and Deniso-     amount of interbreeding, “or we would have             well as Europe and Asia.
                                         vans. One meeting may have been in western      discovered this long ago.”                                                            –ANN GIBBONS

                                                                        SCIENCE        VOL 334        14 OCTOBER 2011                                                  167

      M I C R O B I O LO G Y

      Gut Bacteria Lend a Molecular Hand to Viruses
      Ten years ago, the Jackson Laboratory had a       wise, germ-free pregnant mice failed to pass             infectivity of retroviruses and lentiviruses.”
      problem: A line of mice wouldn’t develop the      virus to their offspring.                                    Pfeiffer’s team has uncovered a remark-
      expected cancer when exposed to a certain            Further test-tube experiments and studies             ably similar story. Three years ago, like
      tumor-causing virus. Tatyana Golovkina,           with mice bred to lack certain immune sys-               many others, Pfeiffer was convinced that gut
      then a molecular geneticist at the Maine lab,     tem signaling proteins ultimately revealed               bacteria help limit viral infection. But when
      was called in to find out why. Researchers         that MMTV gains a foothold in the body                   her graduate student, Sharon Kuss, tested
      knew that this mouse strain had a defective       by binding to copies of a bacterial cell wall            this idea by introducing the poliovirus to
      version of a mouse immune system receptor         molecule called lipopolysaccharide (LPS),                mice treated with four antibiotics and mice
      called TLR4, and Golovkina showed that the        a natural trigger of TLR4. When the LPS-                 with their full complement of gut bacteria,
      virus depended on this protein to establish       covered virus binds to TLR4 on white blood               twice as many mice with intact microflora
      an infection. The finding was puzzling, how-       cells, it stimulates their production of an              died. And when the researchers reintroduced
      ever, because viruses were not supposed to                                                                 mouse gut bacteria into the antibiotic-treated
      interact with that protein.                                                                                mice, the mice also became sick from the
          On page 245, Golovkina, now at the                                                                     virus. Further experiments showed that the
      University of Chicago in Illinois, and her                                                                 poliovirus replicated much more readily in
      colleagues provide a solution: The virus                                                                   mice with normal gut microbes, whereas in
      covers itself with molecules from natural                                                                  mice with those bacteria depleted, “the virus
      gut bacteria, and those molecules interact                                                                 hardly replicated at all,” Pfeiffer says.
      with TLR4 to make viral infection possible.          LPS in bacterial                                          The researchers then did the same experi-
                                                           cell wall
      Nor is this pathogen, the mouse mammary                                                                    ments with a reovirus, which impairs bile duct
      tumor virus (MMTV), the only virus that                                                                    function. It, too, made use of the gut micro-
      co-opts components of the microflora. At                  Free LPS                                          flora, they found. The researchers discovered,
                                                               in mouse
      least two others, including poliovirus, rely             digestive tract                                   as Golovkina’s team had, that LPS was the
      on those same molecules from gut bacteria                                       LPS-bound MMTV             key. Viruses incubated with LPS were better
      to invade their host, Julie Pfeiffer, a virolo-                                                            able to infect cultured cells normally. Another
      gist at the University of Texas Southwest-                                                                 bacterial wall component called peptidogly-
      ern Medical Center in Dallas and her team                                                  TLR4            can also promoted viral replication.
      report in another study, on page 249, that                                                                     Pfeiffer’s team has not, however, seen
      echoes Golovkina’s finding.                                                                                 evidence that the binding of LPS by viruses
                                                                    Dendritic cell,
          “The studies uncover an unexpected                        macrophage                                   induces an antiviral response via IL-10, as
      aspect of microbiome contribution to viral                                                                 Golovkina found. Instead, their test-tube
      infection cycles,” says Ruslan Medzhitov, an                                                               experiments indicate that the viruses’ abil-
      immunologist at Yale University. “The work                                                                 ity to attach to the host gut cells is somehow
      adds to a growing appreciation of the impor-                                                               enhanced, improving infectivity immensely.
      tance of the microbiome.”                                                                                  It could be that the binding keeps the
          Since the 1950s, researchers have rec-                                                                 viruses from being inactivated in the gut. Or
      ognized that the body’s bacteria help fight                                                                 the virus may use the bacteria as transport,
      infection by crowding out potential patho-                                                                 making it easier for it to get to the host cell,
      gens. In the past few years, scientists have                                                               Pfeiffer says.
      even begun to see our microbiota as staunch                                                                    The two papers are “of great significance
      allies. Gut bacteria help shape the develop-                                     response                  for understanding the infection biology of
      ing immune system, for example, and mice                                                                   these viruses,” says Wolf-Dietrich Hardt, a
      lacking their natural gut bacteria, known as                                                               microbiologist at the Swiss Federal Institute
      germ-free mice, are more susceptible to sev-      Viral trick. The MMTV virus uses LPS from gut bacte-     of Technology Zurich in Switzerland. Other
      eral viruses. But with MMTV, a different          ria to avoid the immune system.                          viruses may use similar mechanisms, he says:
      story emerged.                                                                                             “Thus the relevance may be quite broad.”
          MMTV is transmitted from mother to            immune system signaling molecule, called                     Pfeiffer wonders whether knowing that
                                                                                                                                                                    CREDIT: ADAPTED FROM TATYANA GOLOVKINA

      young in milk and invades the body from           IL-10, which suppresses the body’s antiviral             poliovirus relies on gut bacteria might help
      the gut. In one experiment with a normal          reaction. Infection then proceeds unim-                  improve vaccine efficacy. Others think there
      mouse strain, Golovkina’s Chicago graduate        peded, Golovkina and her team report.                    are implications for treating viral infec-
      student Melissa Kane treated pregnant mice            “This is a particularly striking example             tions. The work “will make us reevaluate
      with antibiotics that killed their gut bacte-     of interdependence” among viruses, bacte-                the impact antibiotics will have on viral
      ria and tested their offspring as adults. These   ria and the gut, says Herbert Virgin, a viral            disease,” says Richard Grencis, an immu-
      offspring carried no viruses in their spleens     immunologist at Washington University                    nologist at the University of Manchester in
      and produced antibodies to MMTV when              School of Medicine in St. Louis, Missouri.               the United Kingdom. “This could have far-
      injected with the virus, signs they had never     “It means that retrovirologists have to now              reaching clinical implications.”
      been infected, the researchers report. Like-      think about bacteria when they think about                                        –ELIZABETH PENNISI

168                                         14 OCTOBER 2011             VOL 334       SCIENCE 

            The latest attempt to clarify how NSF assesses grant proposals for possible impacts beyond
            the expected scientific results has not ended a long-running debate

THE NATIONAL SCIENCE FOUNDATION                    business by explaining once and for all what               “So many people have interpreted the
(NSF) uses two criteria to judge the 55,000        NSF means by broader impacts. In June, a broader-impacts criterion as requiring some-
grant proposals it receives each year. One is      board task force took a stab at drafting a new thing involving K–12 education,” says Cora
straightforward enough: intellectual merit.        set of principles for reviewers and applicants Marrett, deputy NSF director and former head
But the other, known as “broader impacts,”         to follow. Instead of clearing the air, however, of NSF’s education directorate. “But that’s not
is so confusing that a cottage industry has        the draft guidelines generated a fresh wave of our position at all.” Pinning down the scope
sprung up to help scientists figure it out.         protest. (See page 170 for excerpts                                 of broader impacts could actu-
    For only $79—marked down from $197,            from a few of the 265 e-mailed com-                                 ally be counterproductive, she
according to one recent e-mail—a company
in Florida offers a CD that teaches appli-
                                                   ments. NSF also gathered feedback
                                                   in meetings with and surveys of
                                                                                                                       adds. “Efforts to find activities
                                                                                                                       that are universal may be totally
cants how “to successfully identify, distill,      stakeholders.)                                    Podcast interview inappropriate for a particular
and communicate your project’s broader                 The task force has taken that                 with author       field, or project, or individual PI
impacts to NSF reviewers, improving your           criticism to heart and is busy revis-      Jeffrey Mervis.          [principal investigator],” Mar-
chances of funding.” For those scientists on       ing the guidelines. “It’s our attempt                               rett says, “because people may
a tight budget, there are plenty of free tips in   to be frighteningly clear,” task force member not have the special expertise needed.”
the open literature.                               Alan Leshner (CEO of AAAS, which pub-                      Working to improve STEM (science,
    The phrase is intended to help NSF             lishes Science) explained during a meeting technology, engineering, and mathematics)
determine whether the cutting-edge science         last month of the task force. “But if we’re education at all levels is certainly one way
being proposed is also addressing an impor-        not, we need to know that.”                           to demonstrate a commitment to broader
tant societal issue. It was adopted in 1997                                                              impacts. But so are activities aimed at
to rebut criticism that the basic-research         An elusive definition                                  attracting and retaining more women and
agency cared more about the interests of           NSF has been anything but clear over the minorities and, more generally, training the
the academic community it serves than the          years about the meaning of broader impacts. next generation of scientists. Strengthening
needs of the taxpayers who ultimately foot         Officials steadfastly refused to provide a pre- the nation’s scientific infrastructure through
the bill for its programs.                         cise definition, for what they believed to be increased collaborations and the sharing of
    The National Science Board, NSF’s              a good reason: Such a statement might stifle equipment and facilities would qualify, as
presidentially appointed oversight body,           innovative thinking about the myriad ways would many types of public outreach. Over
would like to put that cottage industry out of     research can benefit society.                          the years, NSF has also looked favorably

                           SCIENCE VOL 334                   14 OCTOBER 2011                                                169

      upon researchers who hope to commercial-          that are directly related to specific research     hunt for an economic payoff. Another group,
      ize their discoveries with the goal of boosting   projects, or through activities supported by      less vocal but equally critical, told the task
      the economy, improving national security,         the project but ancillary to the research.”       force that it had lost sight of the importance of
      protecting the environment, and enhancing             The list of national goals drew sharp and     basic research. Some even worried that ask-
      the general well-being of society.                bimodal criticism. One segment of the com-        ing applicants to describe how they planned
          Since the second criterion was adopted,       munity complained vociferously that the new       to satisfy the broader-impacts criterion might
      however, NSF officials have complained            language would undermine efforts to pro-          corrupt the merit-review process itself.
      that significant portions of the community         mote NSF’s historical commitment to diver-            Recognizing that its first draft had fallen
      are either ignoring or not taking it seriously.   sity, improving instruction, and raising public   short, the task force scrapped it and began
      In 2002, a frustrated NSF Director Rita           literacy. They feared that those goals, previ-    working on a new version. Although the task
      Colwell tried to crack the whip. “NSF will        ously the centerpiece of the broader-impacts      force has not made public a copy of its lat-
      return without review proposals that do not       criterion, would become subservient to the        est draft, John Bruer, president of the James
      separately address both merit review criteria
      within the Project Summary,” she declared
      in a bulletin to the community.
          Confusion persisted, however. Merit              THE COMMUNITY WEIGHS IN ON
      review at NSF is a complex process, involv-
      ing panels made up of thousands of volunteer
      reviewers guided by hundreds of NSF pro-
      gram officers. And scientists say that no two
                                                           BROADER IMPACTS
      panels follow exactly the same definition of            “The new criteria shift from                     “Making the stated [and
      broader impacts.                                        a strong, clear expression                       laudable] national goals an
          So in 2007, NSF tried a new tack. It
      described five “representative activities” that          of a commitment to preparing                     explicit funding criterion
      would qualify as having a broader impact.               and engaging a diverse                           amounts to apologizing for
      Four of them were familiar: promoting                   domestic scientific workforce                     scientific research rather than
      teaching, training, and learning at all levels;
      broadening participation of groups under-               to the downgrading of diver-                     leading the fight in support of
      represented in science, notably women, non-             sity to be one of nine possible                  it. … Proposals for funding of
      Asian minorities, persons with disabilities,            methods of demonstrating                         basic science will only address
      and those at institutions outside the scientific
      mainstream; enhancing the research and edu-
                                                              broad impact. … I have no                        these criteria if the proposer
      cation infrastructure through increased col-            doubt that this change will                      is insincere but convincing
      laborations; and disseminating knowledge of             dilute the impact of NSF on                      and the responsible program
      science and technology to the general public.
      The fifth activity, listed under the heading
                                                              access and representation in                     director is insincere in reading
      Benefits to Society, serves as a catch-all cat-          science and engineering.”                        that part of the proposal.”
      egory for activities such as advising govern-                —JANET WEISS, DEAN AND VICE                  —BORIS HASSELBLATT, CHAIR OF THE
      ment agencies, working with industry, and                  PROVOST FOR ACADEMIC AFFAIRS,                         MATHEMATICS DEPARTMENT,
      putting data into a format that nonscientists                    UNIVERSITY OF MICHIGAN,                     TUFTS UNIVERSITY IN MEDFORD,
      can understand. Last year, Congress waded                                     ANN ARBOR                                   MASSACHUSETTS
      into the fray, asking NSF to explain what it
      means by broader impacts and how it plans to
      ensure compliance.                                                                                      “We recommend that the third
                                                             “The broader impacts should                       national goal be removed
      No-go on national goals
      The science board’s June draft tried to answer          carry weight only as an extra                    from the list and included as
      that question. It began by laying out four              benefit, an additional posi-                      a separate principle [to wit]:
      principles underpinning merit review, the               tive feature of a proposal. A
      most notable being a statement that “NSF                                                                 Broadening participation is a
      projects should help to advance a broad set             person should be allowed to                      key to achieving the national
      of important national goals.” That asser-               say that there are no broader                    goals, and investigators must
      tion is followed by a list of nine examples             impacts envisioned and still
      of activities that would qualify as demon-
                                                                                                               include mechanisms in their
      strating broader impacts. The list began with
                                                              get funding if the science is                    broader impacts.”
      “increased economic competitiveness” and                visionary and outstanding.”                              —ANN GATES, ASSOCIATE VICE
      included “increased partnerships between                  —STEPHEN LEONE, PROFESSOR OF                               PRESIDENT OF RESEARCH,
      academia and industry” and “increased                            CHEMISTRY AND PHYSICS,                         UNIVERSITY OF TEXAS, EL PASO,
      national security.” Another principle declares                 UNIVERSITY OF CALIFORNIA,                        ON BEHALF OF THE COMPUTING
      that “broader impacts may be achieved                                          BERKELEY                              ALLIANCE FOR HISPANICS
      through the research itself, through activities

170                                       14 OCTOBER 2011         VOL 334 SCIENCE

S. McDonnell Foundation and chair of the           actually intended to describe a set of out-       Some types of NSF-funded activities and pro-
task force, agreed to talk with Science about      comes,” Bruer says. “Listing them as a guid-      grams, such as large centers or collaborations
its key provisions.                                ing principle created some concerns within        with industry, “are explicit about how broader
    A major change is a reconfiguration of          the community that we have addressed in the       impacts will be evaluated.” But that’s not the
the section that lays out the overriding prin-     new version.”                                     case for standard grants to individuals. “Prin-
ciples for merit review. Instead of four prin-         The emphasis on top-quality research in       cipal investigators have tried to respond in
ciples, with the list of national goals featured   the first principle, Bruer says, should assuage    good faith and devote appropriate resources to
prominently, there are three governing ideas,      those who fear that the broader-impacts cri-      these activities,” he says. “But it may be more
none as controversial as the goals list. The       terion “would dilute NSF’s commitment to          appropriate to measure them at the portfolio
first declares that all NSF projects “should        excellence in research.” The second principle,    level of an NSF directorate, or perhaps across
be of the highest quality.” The second says        he adds, explains “how broader impacts might      an entire university department or institution.”
that, in the aggregate, projects should con-       be achieved. We want to make clear that it’s         Bruer didn’t rule out soliciting another
tribute more broadly to advancing societal         not seen simply as an additional activity on a    round of public comments. But he hopes that
goals. The third states that any assessment        grant but rather that it is an essential compo-   the full board will embrace the task force’s
of a project’s broader impacts should be           nent of the research activities.”                 report at its next meeting, in December, and
“scaled” to the size of the activity.                  Bruer said it may turn out that broader       he says that posting another draft “might push
    “The original list of national goals … was     impacts are best measured “in the aggregate.”     back that time frame.”        –JEFFREY MERVIS

  “The criteria as modified are                         “It is very difficult to do                        “Much of the research in my
   troubling. Those supported                           first-rate science, and very                       field [computer sciences] has
   by the NSF must be held                              time-consuming. It is therefore                   significant impact on other
   accountable to the society                           unreasonable that NSF demand                      fields, on national security, and
   that supports them, not just                         its PIs spend time and energy                     on economic competitiveness.
   to those that look like the                          on activities peripheral to the                   Yet my field has a terrible
   majority of scientists [white                        research they are proposing                       record for the inclusion of
   and male] awarded grants                             as a condition for receiving                      women and minorities, and
   from mostly research-                                support.”                                         a terrible record for K–12
   intensive universities.”                                     —PETER MOORE, PROFESSOR                   impact. To include [those goals]
                                                                  EMERITUS OF CHEMISTRY,
               —MARIA ELENA ZAVALA,
                                                                         YALE UNIVERSITY
                                                                                                          as broader impacts gives
                                                                                                          researchers in my field an
                         NORTHRIDGE                                                                       unconscionably cheap way
                                                                                                          to satisfy this criterion.”
                                                                                                                           —ED LAZOWSKA,
                                                       “I have always valued the                          PROFESSOR OF COMPUTER SCIENCES,
                                                        older construal of the                           UNIVERSITY OF WASHINGTON, SEATTLE
  “Forcing PIs to devote time,                          broader impacts criterion,
   energy, and resources toward                         i.e., the value of outreach,
   outreach activities, one                             the inclusion of different                       “It is not clear to me how a
   NSF grant at a time, in an                           educational institutions,                         merit-review panel could
   uncoordinated fashion with                           and the purposeful attention                      decide how one PI’s goals
   other efforts, is highly waste-                      to supporting and bringing in                     and aspirations for broader
   ful of resources. Many PIs are                       women and members of                              impacts are more likely to
   not well-qualified, well-suited,                      underrepresented groups.                          succeed than another’s. …
   or highly motivated for K–12                         I believe that important focus                    This uncertainty contrasts
   outreach, public outreach, et                        is hidden, and even lost, in                      sharply with the clarity with
   cetera.”                                             the proposed guidelines.”                         which intellectual merit can
                                                                                                          be evaluated.”
      MECHANICAL ENGINEERING AND                                COUNSELING PSYCHOLOGY,                       —GORDON LOGAN, PROFESSOR OF
               APPLIED MECHANICS,                              ARIZONA STATE UNIVERSITY,                  PSYCHOLOGY, VANDERBILT UNIVERSITY
       UNIVERSITY OF PENNSYLVANIA                                                 TEMPE                                       IN NASHVILLE.

                            SCIENCE VOL 334                  14 OCTOBER 2011                                                171
Quantify, verify,
   In science there are always essential steps in your workflow. QA/QC of DNA,
   RNA and protein samples is critical for confidence in qPCR, sequencing,
   microarray analysis or bioproduction, but there’s a better alternative to the
   time and complexity of conventional methods. Using just a minimal sample
   and almost no sample prep, the Thermo Scientific NanoDrop and its built-
   in methods make QA/QC so incredibly easy, and so much faster, you won’t

   notice this step on the way to your ultimate discovery.

                                                                                                 • learn about our free trial program •

                                                                                                                                                                           NanoDrop™ 2000 and 2000C
                                                                                                                                                                           UV-Vis Spectrophotometers
                                                                                                                                                                           Measure samples as small
                                                                                                                                                                           as 0.5 μL up to 2 μL
owned by a third party and licensed for limited use only to Thermo Fisher Scientific by iStock.
© 2011 Thermo Fisher Scientific Inc. All rights reserved. Copyright to the protein image are

                                                                                                                                                                           NanoDrop 8000
                                                                                                                                                                           UV-Vis Spectrophotometer
                                                                                                                                                                           Measure up to 8 microvolume
                                                                                                                                                                           samples at a time

                                                                                                                                                                           NanoDrop 3300
                                                                                                                                                                           Measure fluorescent samples
                                                                                                                                                                           as small as 1 μL

                                                                    P R E D I C T I N G C L I M AT E C H A N G E                                                                  They are already locked into climate change.
                                                                                                                                                                                      So scientists have been doing what they
                                                                    Vital Details of Global Warming                                                                               can for decision-makers. Early on, it wasn’t
                                                                                                                                                                                  much. A U.S. government assessment
                                                                                                                                                                                  released in 2000, Climate Change Impacts on
                                                                    Are Eluding Forecasters                                                                                       the United States, relied on the most rudimen-
                                                                                                                                                                                  tary regional forecasting technique (Science,
                                                                    Decision-makers need to know how to prepare for inevitable climate change, but climate                        23 June 2000, p. 2113). Expert commit-
                                                                    researchers are still struggling to sharpen their fuzzy picture of what the future holds                      tee members divided the country into eight
                                                                                                                                                                                  regions and then considered what two of their
                                                                    Seattle Public Utilities officials had a ques-     comparison of global-regional model combi-                  best global climate models had to say about
                                                                    tion for meteorologist Clifford Mass. They        nations to sort out the uncertainties, although             each region over the next century. The two
                                                                    were planning to install a quarter-billion dol-   that won’t help Seattle’s storm-drain builders.             models were somewhat consistent in the far
                                                                    lars’ worth of storm-drain pipes that would                                                                   southwest, where the report’s authors found
                                                                    serve the city for up to 75 years. “Their ques-   Most humble origins                                         it was likely that warmer and drier condi-
                                                                    tion was, what diameter should the pipe be?       Policymakers have long asked for regional                   tions would eliminate alpine ecosystems and
                                                                    How will the intensity of extreme precipita-      forecasts to help them adapt to climate                     shorten the ski season.
                                                                    tion change?” Mass says. If global warming        change, some of which is now unavoidable.                       But elsewhere, there was far less consis-
                                                                    means that the past century’s rain records are    Even immediate, rather drastic action to curb               tency. Over the eastern two-thirds of the con-
                                                                    no guide to how heavy future rains will be,       emissions of greenhouse gases would not                     tiguous 48 states, for example, the two models
                                                                    he was asked, what could climate modeling         likely limit warming globally to 2°C, gener-                couldn’t agree on how much moisture soils
                                                                    say about adapting to future climate change?      ally considered the threshold above which                   would hold in the summer. Kansas corn would
                                                                    “I told them I couldn’t give them an answer,”     “dangerous” effects set in. And nothing at                  either suffer severe droughts more frequently,
                                                                    says the University of Washington (UW),           all can be done to reduce the global warming                as one model had it, or enjoy even more mois-
                                                                    Seattle, researcher.                              effects expected in the next several decades.               ture than it currently does, as the other indi-
                                                                        Climate researchers are quite comfort-                                                                    cated. But at least the uncertainties were plain
                                                                    able with their projections for the world under   Winter Temperature Change                                   for all to see.
                                                                                                                                  Global Model
                                                                    a strengthening greenhouse, at least on the                                                                       The uncertainties of regional projec-
                                                                    broadest scales. Relying heavily on climate                                                                   tions nearly faded from view in the next U.S.
                                                                    modeling, they find that on average the globe                                                                  effort, Global Climate Change Impacts in the
                                                                    will continue warming, more at high northern                                                                  United States. The 2009 study drew on not
                                                                    latitudes than elsewhere. Precipitation will                                                                  two but 15 global models melded into single
                                                                    tend to increase at high latitudes and decrease                                                               projections. In a technique called statistical
                                                                    at low latitudes.                                                                                             downscaling, its authors assumed that local
                                                                        But ask researchers what’s in store for the                                                               changes would be proportional to changes on
                                                                    Seattle area, the Pacific Northwest, or even the                                                               the larger scales. And they adjusted regional
                                                                    western half of the United States, and they’ll                                                                projections of future climate according to
                                                                    often demur. As Mass notes, “there’s tremen-                                                                  how well model simulations of past climate
                                                                    dous uncertainty here,” and he’s not just talk-                                                               matched actual climate.
                                                                    ing about the Pacific Northwest. Switching                                                                         Statistical downscaling yielded a broad
                                                                    from global models to models focusing on a                                                                    warming across the lower 48 states with less
                                                                    single region creates a more detailed forecast,                                                               warming across the southeast and up the
                                                                    but it also “piles uncertainty on top of uncer-                                                               West Coast. Precipitation was mostly down,

                                                                    tainty,” says meteorologist David Battisti of                                                                 especially in the southwest. But discussion
                                                                    UW Seattle.                                                                                                   of uncertainties in the modeling fell largely
                                                                        First of all, there are the uncertainties                                                                 to a footnote (number 110), in which the
                                                                    inherent in the regional model itself. Then                                                                   authors cite a half-dozen papers to support
                                                                    there are the global model’s uncertainties                                                                    their assertion that statistical downscaling
                                                                    at the regional scale, which it feeds into the                                                                techniques are “well-documented” and thor-
                                                                    regional model. As the saying goes, if the                                                                    oughly corroborated.
                                                                    global model gives you garbage, regional                                                                          The other sort of downscaling, known as
                                                                    modeling will only give you more detailed                                                                     dynamical downscaling or regional model-
                                                                    garbage. And still more uncertainties are cre-                                                                ing, has yet to be fully incorporated into a
                                                                    ated as data are transferred from the global to                                                               U.S. national assessment. But an example of
                                                                    the regional model.                               Degrees C                                                   state-of-the-art regional modeling appeared
                                                                        Although uncertainties abound, “uncer-                                                                    30 June in Environmental Research Let-
                                                                                                                        -2 -1.5 -1 -0.5 0   0.5 1   1.5   2 2.5   3   4   5   7
                                                                    tainty tends to be downplayed in a lot of                                                                     ters. To investigate what will happen in
                                                                    [regional] modeling for adaptation,” says         Sharp but true? Feeding a global climate model’s            the U.S. wine industry, regional modeler
                                                                    global modeler Christopher Bretherton of          prediction for midcentury (top) into a regional             Noah Diffenbaugh of Purdue University in
                                                                    UW Seattle. But help is on the way. Regional      model gives more details (bottom), but modelers             West Lafayette, Indiana, and his colleagues
                                                                    modelers are well into their first extensive      aren’t sure how accurate the details are.                   embedded a detailed model that spanned

                                                                                                 SCIENCE VOL 334                              14 OCTOBER 2011                                        173

      the lower 48 states in a climate model that        Summer Precipitation Change                                 building global models and regional models
      spanned the globe. The global model’s rela-                   Global Model                                     can lead to mismatches that create phantom
      tively fuzzy simulation of evolving climate                                                                    atmospheric circulations, Tripoli says. “It’s
      from 1950 to 2039—calculated at points                                                                         not straightforward you’re going to get any-
      about 150 kilometers apart—then fed into                                                                       thing realistic,” he says.
      the embedded regional model, which calcu-
      lated a sharper picture of climate change at                                                                   Redeeming regional modeling
      points only 25 kilometers apart.                                                                               “You could say all the global and regional
          Closely analyzing the regional model’s                                                                     models are wrong; some people do say that,”
      temperature projections on the West Coast,                                                                     notes regional modeler Filippo Giorgi of the
      the group found that the projected warm-                                                                       Abdus Salam International Centre for Theo-
      ing would decrease the area suitable for pro-                                                                  retical Physics in Trieste, Italy. “My personal
      duction of premium wine grapes by 30% to                                                                       opinion is we do know something now. A few
      50% in parts of central and northern Califor-                                                                  reports ago, it was really very, very difficult to
      nia. The loss in Washington state’s Columbia                                                                   say anything about regional climate change.”
      Valley would be more than 30%. But adap-                                                                           But Giorgi says that in recent years he has
      tation to the warming, such as the introduc-                                                                   been seeing increasingly consistent regional
      tion of heat-tolerant varieties of grapes, could                                                               projections coming from combinations of
      sharply reduce the losses in California and                                                                    many different models and from successive
      turn the Washington loss into a 150% gain.                                                                     generations of models. “This means the pro-
                                                                                                                     jections are more and more reliable,” he says.
      Not so fast                                                                                                    “I would be confident saying the Mediterra-
      A rapidly growing community of regional                                                                        nean area will see a general decrease in pre-
      modelers is turning out increasingly detailed                                                                  cipitation in the next decades. I’ve seen this
      projections of future climate, but many                                                                        in several generations of models, and we
      researchers, mostly outside the downscaling                                                                    understand the processes underlying this phe-
      community, have serious reservations. “Many                                                                    nomenon. This is fairly reliable information,
      regional modelers don’t do an adequate job                                                                     qualitatively. Saying whether the decrease will
      of quantifying issues of uncertainty,” says        Percent                                                     be 10% or 50% is a different issue.”
      Bretherton, who is chairing a National                                                                             The skill of regional climate forecasting
      Academy of Sciences study committee on a             -50 -40 -30 -20 -10   -5   0   5   10   20   30 40   50   also varies from region to region and with
      national strategy for advancing climate mod-       A tougher nut. Predicting the details of precipi-           what is being forecast. “Temperature is much,
      eling. “We’re not confident predicting the very     tation using a regional model (bottom) fed by a             much easier” than precipitation, Giorgi notes.
      things people are most interested in being pre-    global model (top) is even more uncertain than              Precipitation depends on processes like atmo-
      dicted,” such as changes in precipitation.         projecting regional temperature change.                     spheric convection that operate on scales too
          Regional models produce strikingly                                                                         small for any model to render in detail. Trouble
      detailed maps of changed climate, but they         medium-size weather systems that they                       simulating convection also means that higher-
      might be far off base. “The problem is that        should be sending into any embedded regional                latitude climate is easier to project than that of
      precision is often mistaken for accuracy,”         model. Tripoli, a meteorologist and modeler                 the tropics, where convection dominates.
      Bretherton says. Battisti just doesn’t see the     at the University of Wisconsin, Madison, cites                  Regional modeling does have a clear
      point of downscaling. “I would never use one       the case of summertime weather disturbances                 advantage in areas with complex terrain such
      of these products,” he says.                       that churn down off the Rocky Mountains                     as mountainous regions, notes UW’s Mass,
          The problems start with the global models,     and account for 80% of the Midwest’s sum-                   who does regional forecasting of both weather

                                                                                                                                                                          CREDIT: NORTH AMERICAN REGIONAL CLIMATE CHANGE ASSESSMENT PROGRAM
      as critics see it. Regional models must fill in     mer rainfall. If a regional model forecasting               and climate. In the Pacific Northwest, the
      the detail in the fuzzy picture of climate pro-    for Wisconsin doesn’t extend to the Rockies,                mountains running parallel to the coast direct
      vided by global models, notes atmospheric          Wisconsin won’t get the major weather events                onshore winds upward, predictably wring-
      scientist Edward Sarachik, professor emeritus      that add up to be climate. And some atmo-                   ing rain and snow from the air without much
      at UW Seattle. But if the fuzzy picture of the     spheric disturbances travel from as far away                difficult-to-simulate convection.
      region is wrong, the details will be wrong as      as Thailand to wreak havoc in the Midwest,                      The downscaling of climate projections
      well. And global models aren’t very good at        he says, so they could never be included in the             should be getting a boost as the Coordinated
      painting regional pictures, he says. A glaring     regional model.                                             Regional Climate Downscaling Experiment
      example, according to Sarachik, is the way             Even the things the global models get right             (CORDEX) gets up to speed. Begun in 2009,
      global models place the cooler waters of the       have a hard time getting into regional models,              CORDEX “is really the first time we’ll get a
      tropical Pacific farther west than they are in      critics say. “There are a lot of problems match-            handle on all these uncertainties,” Giorgi says.
      reality. Such ocean temperature differences        ing regional and global models,” Tripoli says.              Various groups will take on each of the world’s
      drive weather and climate shifts in specific        In one problem area, global and regional mod-               continent-size regions. Multiple global mod-
      regions halfway around the world, but with the     els usually have different ways of accounting               els will be matched with multiple regional
      cold water in the wrong place, the global mod-     for atmospheric processes such as individual                models and run multiple times to tease out
      els drive climate change in the wrong regions.     cloud development that neither model can                    the uncertainties in each. “It’s a landmark for
          Gregory Tripoli’s complaint about the          simulate directly, creating further clashes.                the regional climate modeling community,”
      global models is that they can’t create the        Even the different philosophies involved in                 Giorgi says.                 –RICHARD A. KERR

174                                        14 OCTOBER 2011            VOL 334 SCIENCE
                                                             Best ways to                                                              Smoke pollutes
                                                          change our ways                                                                    indoors

                                                                      178                                                                      180
                                                     LETTERS I BOOKS I POLICY FORUM I EDUCATION FORUM I PERSPECTIVES

      edited by Jennifer Sills

      Partial Retraction                                                                                                                                nium (7, 11), and americium was ekairidium
                                                                                                                                                        (7). Some alternative names remain in use:
      IN OUR 23 OCTOBER 2009 REPORT, “DETECTION OF AN INFECTIOUS RETROVIRUS, XMRV, IN                                                                   Metallurgists still use columbium for nio-
      blood cells of patients with chronic fatigue syndrome” (1), two of the coauthors, Silverman                                                       bium (12). The original names of many of the
      and Das Gupta, analyzed DNA samples from chronic fatigue syndrome (CFS) patients and                                                              elements have slipped into the past, but they
      healthy controls. A reexamination by Silverman and Das Gupta of the samples they used                                                             represent an important part of chemistry’s
      shows that some of the CFS peripheral blood mononuclear cell (PBMC) DNA preparations                                                              history (one that crossword puzzle aficiona-
      are contaminated with XMRV plasmid DNA (2). The following figures and table were based                                                             dos would do well to study).
      on the contaminated data: Figure 1, single-round PCR detection of XMRV sequences in CFS                                                                                                             J. P. LEAL
      PBMC DNA samples; table S1, XMRV sequences previously attributed to CFS patients; and                                                             Department of Chemistry and Biochemistry, Faculty of Sci-
      figure S2, the phylogenetic analysis of those sequences. Therefore, we are retracting those                                                        ences, University of Lisbon, 1149-016 Lisbon, Portugal and
                                                                                                                                                        UCQR, Instituto Tecnológico Nuclear, 2686-953 Sacavém,
      figures and table.                                                                                                                                 Portugal. E-mail:
          MAX A. PFOST,2 KATHRYN S. HAGEN,2 DANIEL L. PETERSON,2† SANDRA K. RUSCETTI,4 RACHEL K. BAGNI,5                                                     References
                                                                                                                                                         1. International Year of Chemistry (www.chemistry2011.
                                  CARI PETROW-SADOWSKI,6 BERT GOLD,3 MICHAEL DEAN,3 JUDY A. MIKOVITS2
       Department of Cancer Biology, The Lerner Research Institute, The Cleveland Clinic Foundation, Cleveland, OH 44195, USA.                           2. W. C. Fernelius, K. Loening, R. M. Adams, J. Chem. Educ.
       Whittemore Peterson Institute, Reno, NV 89557, USA. 3Laboratory of Experimental Immunology, National Cancer Institute–                               52, 583 (1975).
      Frederick, Frederick, MD 21701, USA. 4Laboratory of Cancer Prevention, National Cancer Institute–Frederick, Frederick, MD                          3. D. W. Ball, J. Chem. Educ. 62, 787 (1985).
      21701, USA. 5Advanced Technology Program, National Cancer Institute–Frederick, Frederick, MD 21701, USA. 6Basic Research                           4. V. Ringnes, J. Chem. Educ. 66, 731 (1989).
      Program, Scientific Applications International Corporation, National Cancer Institute–Frederick, Frederick, MD 21701, USA.                          5., Encyclopedia: Darmstadtium
      *To whom correspondence should be addressed. E-mail:
                                                                                                                                                         6. Royal Society of Chemistry, Chemistry in Its Element:
      †Present address: Sierra Internal Medicine, 865 Tahoe Boulevard no. 306, Incline Village, NV 89451, USA.
                                                                                                                                                            Americium (
             References and Notes                                                                                                                           Interactive_Periodic_Table_Transcripts/Americium.asp).
          1. V. C. Lombardi, F. W. Ruscetti, J. Das Gupta, M. A. Pfost, K. S. Hagen, D. L. Peterson, S. K. Ruscetti, R. K. Bagni, C. Petrow-Sadowski,    7. Elementymology and Elements Multidict, Names That Did
             B. Gold, M. Dean, R. H. Silverman, J. A. Mikovits, Science 326, 585 (2009).                                                                    Not Make It (
          2. Supporting online material showing the reanalysis is available at                  8. D. Mendeleev, Ann. Chem. Pharm., (suppl. VIII), 133
             Published online 22 September 2011; 10.1126/science.1212182                                                                                    (1871).
                                                                                                                                                         9. C. A. Winkler, B. Deut. Chem. Ges. 19, 210 (1886).
                                                                                                                                                        10. P.-E. Le coq de Boisbeaudran, Ann. Chim. 10, 100
                                                                                                                                                        11. Early History of LBNL, Transcript of lecture by Glenn T.
      Chemical Elements:                                                        the name because the telephone number of
                                                                                                                                                            Seaborg, 26 August 1996, Elements 93 and 94 (www.lbl.
                                                                                                                                                        12. S. Luidold, H. Antrekowitsch, R. Ressel, Int. J. Refract.
      What’s in a Name?                                                         the police in Germany is 110 (5). Today we
                                                                                                                                                            Metals Hard Mat. 25, 423 (2007).
                                                                                refer to this element as darmstadtium. What
      IN 2011, THE INTERNATIONAL YEAR OF                                        about pandemonium and delirium? Seaborg
      Chemistry, we pause to celebrate the achieve-                             and co-workers proposed these names for                                 Biosecurity and
      ments of chemistry and the field’s contribu-                               the elements known today as americium and
      tions to the well-being of humankind (1). In                              curium, respectively, because of the tremen-                            the Politics of Fear
      doing so, it seems appropriate to recognize                               dous difficulties they faced in preparing them                           SINCE THE 2001 ANTHRAX MAILINGS
      the history of the chemical elements, the                                 (6, 7). Mendeleev suggested the prefixes eka                             shocked the public, the United States has
      building blocks of chemistry. The commonly                                and dwi (meaning one and two in Sanskrit,                               substantially increased its funding for
      used names of chemical elements have been                                 to indicate that a given element is one or two                          research and development of biodefense
      presented and explained in detail over the                                rows below the stated element) (8). If the pre-                         countermeasures (“Taking stock of the bio-
      years (2–4). However, some erstwhile names                                fixes had been adopted, germanium would be                               defense boom,” J. Kaiser, News Focus, 2
      are much more difficult to find. Who remem-                                 known as ekasilicon (9) and gallium would be                            September, p. 1214). These funds could be
      bers the name policium, proposed by the dis-                              ekaaluminium (10). Accordingly, Otto Hahn                               better spent. Biodefense research focuses on
      coverers of element 110? They suggested                                   and co-workers named neptunium ekarhe-                                  pathogens of high biodefense value but low

176                                                            14 OCTOBER 2011                    VOL 334 SCIENCE
                                                                                       Variable threshold                                          Organelle contact                        TECHNICAL COMMENT ABSTRACTS

                                                                                                                                                                                      Comment on “Changes in Climatic
                                                                                       184                                                         186                                Water Balance Drive Downhill
                                                                                                                                                                                      Shifts in Plant Species’ Optimum
                                                                                                                                                                                      Adam Wolf and William R. L. Anderegg
                                                                                                                                                                                      Crimmins et al. (Reports, 21 January 2011, p. 324)
                                                                                                                                                                                      presented a study that purports to show that plants in
                                                                                                                                                                                      California are shifting downslope to maintain a constant
                                                                                                                           APHIS receives each year. Contrary                         water deficit. We argue that the results are limited in
                                                                                                                           to perceptions, it is probably anthrac-                    scope to just a handful of woody species in one part of
                                                                                                                           nose—a group of plant diseases—                            the state and are confounded by methodological errors.
                                                                                                                           rather than anthrax that poses the                         Full text at
                                                                                                                           greater threat to the United States                        6053/177-a
                                                                                                                           economy. Unfortunately, this reality
                                                                                                                           is not reflected in either the priorities                   Comment on “Changes in Climatic
                                                                                                                           or performance of the Department                           Water Balance Drive Downhill
                                                                                                                           of Homeland Security (12), under                           Shifts in Plant Species’ Optimum
                                                                                                                           whose jurisdiction APHIS border                            Elevations”
                                                                                                                           inspections have operated since                            Robert J. Hijmans
                                                                                                                           2003. Threats to agriculture may be                        Crimmins et al. (Reports, 21 January 2011, p. 324)
                                                              Anthracnose. Agroterrorism risks are overlooked.             less terrifying than those to human                        reported that plant species moved downhill between
                                                                                                                           health, but proportionally more                            1935 and 2005. They compared plot data for two time
                                                              public health significance, such as tularemia,      investment in better border biosecurity and                          periods, ignoring that the modern plots were farther
                                                                                                                                                                                      north than the historical plots. I contend that there is no
                                                              anthrax, plague, glanders, melioidosis, and        emergency incursion response has the poten-                          support for a general downhill shift after correcting for
                                                              brucellosis (1). Because the technological         tial to bring greater dividends to society than                      this geographic bias.
                                                              challenges of producing weaponized patho-          much of the current investment in biodefense                         Full text at
                                                              gens limit their development by rogue orga-        countermeasures (13).                                                6053/177-b
                                                              nizations (2), the primary risks to the United                                                 PHILIP E. HULME
                                                              States stem from national research establish-      The Bio-Protection Research Centre, Lincoln University, Can-         Comment on “Changes in Climatic
                                                              ments, either through malevolent insiders (3)      terbury, New Zealand. E-mail:             Water Balance Drive Downhill
                                                              or inadvertent leaks (4). National biosecu-                                                                             Shifts in Plant Species’ Optimum
                                                              rity faces greater risks from low-technology,        1.   S. Altman et al., Science 307, 1409 (2005).                   Elevations”
                                                              high-impact threats targeting plant and ani-         2.   B. Balmer, Science 326, 1635 (2009).                          Nathan L. Stephenson and Adrian J. Das
                                                              mal, rather than human, health (5).                  3.   Y. Bhattacharjee, Science 323, 1282 (2009).
                                                                                                                                                                                      Crimmins et al. (Reports, 21 January 2011, p. 324)
                                                                                                                   4.   J. Kaiser, Science 317, 1852 (2007).
                                                                  Agroterrorism—deliberate attack with             5.   O. S. Cupp, D. E. Walker, J. Hillison, Biosecur. Bioterror.   attributed an apparent downward elevational shift
                                                              arthropods, viruses, bacteria, or fungi on                2, 97 (2004).                                                 of California plant species to a precipitation-induced
                                                              commercial crops or livestock populations—           6.   P. E. Hulme, in Invasion Ecology, D. M. Richardson, Ed.       decline in climatic water deficit. We show that the
                                                                                                                        (Wiley-Blackwell, Oxford, 2011), pp. 301–314.                 authors miscalculated deficit, that the apparent decline
                                                              can have a considerable impact on the global
                                                                                                                   7.   M. Wheelis, Nature 395, 213 (1998).                           in species’ elevations is likely a consequence of geo-
                                                              market (6). Deliberate threats to plant and          8.   F. Suffert, E. Latxague, I. Sache, Food Secur. 1, 221         graphic biases, and that unlike temperature changes,
                                                              animal health have a long history (7), and a              (2009).                                                       precipitation changes should not be expected to cause
                                                              wide range of countries (including the United        9.   J. Lockwood, Six-Legged Soldiers: Using Insects as            coordinated directional shifts in species’ elevations.
                                                                                                                        Weapons of War (Oxford Univ. Press, Oxford, 2009).            Full text at
                                                              States) are known or suspected to have been        10.    D. Pimentel, R. Zuniga, D. Morrison, Ecol. Econ. 52, 273
                                                              involved in anticrop programs or agroterror-                                                                            6053/177-c

                                                              ist acts (8, 9).                                   11.    U.S. Department of Agriculture, Fiscal Year 2012 Budget
                                                                  Conclusive evidence of agroterrorist acts             Summary and Annual Performance Plan (www.obpa.usda.           Response to Comments on “Changes
                                                              is hindered by the fact that such actions are
                                                                                                                        gov/budsum/FY12budsum.pdf).                                   in Climatic Water Balance Drive
                                                                                                                 12.    Y. Bhattacharjee, Science 332, 783 (2011).
                                                              dwarfed by the agricultural impact of unin-        13.    J. Cohen, Science 333, 1216 (2011).                           Downhill Shifts in Plant Species’
                                                              tentionally introduced pests and diseases                                                                               Optimum Elevations”
                                                              (10). Given the scale and low predictability                                                                            Solomon Z. Dobrowski, Shawn M. Crimmins,
                                                              of such unintentional threats, it is surprising       Letters to the Editor                                             Jonathan A. Greenberg, John T. Abatzoglou,
                                                              that the frontline agency in agricultural bios-                                                                         Alison R. Mynsberge
                                                                                                                    Letters (~300 words) discuss material published in
                                                              ecurity, the Animal and Plant Health Inspec-          Science in the past 3 months or matters of gen-                   Wolf and Anderegg, Hijmans, and Stephenson and Das
                                                              tion Service (APHIS), receives a fraction                                                                               suggest that our findings of changes in climatic water
                                                                                                                    eral interest. Letters are not acknowledged upon
                                                              of the resources devoted to biodefense and                                                                              balance driving downhill shifts in plant species distribu-
                                                                                                                    receipt. Whether published in full or in part, Let-               tions are flawed. We demonstrate that the conclusions
                                                              has suffered progressive cuts in its operating        ters are subject to editing for clarity and space.                these authors make are subject to the selection of meth-
                                                              budget (11). The value of the damage pre-             Letters submitted, published, or posted elsewhere,                ods they apply and do not provide sufficient evidence to
                                                              vented and mitigated annually as a result of          in print or online, will be disqualified. To submit a              reject our original findings.
                                                              plant and animal health monitoring and sur-           Letter, go to                             Full text at
                                                              veillance roughly matches the federal funds                                                                             6053/177-d

                                                                                           SCIENCE VOL 334                                14 OCTOBER 2011                                                               177
      BOOKS ET AL.
  PSYCHOLOGY                                                                                                        improve intertribal trust in Rwanda by mod-
                                                                                                                    eling cooperative intergroup relations through
  Social Psychology and Social Change                                                                               radio soap operas. In the United States, inter-
                                                                                                                    ventions that defuse blacks’ and whites’ fear
                                                                                                                    of interracial rejection increase their likeli-
  Geoffrey L. Cohen                                                                                                 hood of becoming friends. And reminiscent of
                                                                                                                    Lewin, there are studies that cleverly manipu-

           ver the past several decades one of              have large consequences. For example, to late social norms to reduce teen alcohol use
           the quieter social sciences, social              encourage families to eat cheap-cut meats and encourage energy conservation.
           psychology, has made breakthroughs               like sweetbreads during the war (because the                What these interventions share is that they
  in interventions to solve social problems.                finer cuts had limited supply), Lewin showed are grounded in science, found effective in
  Unlike the other branches of psychology,                  the importance of the gatekeeper, the person randomized experiments, have surprisingly
  social psychology was borne out of an inter-              who controls the behavioral channel—in this large and durable effects—and, by and large,
  est in remedying society’s ills. When the                 case, the housewife. He also                                              aren’t used. Over and over,
  father of the field, German refugee Kurt                  demonstrated the impotence                                                Wilson writes, schools,
  Lewin, conducted his seminal studies, the                 of persuasion and the power          Redirect                             government agencies, and
  problems of World War II preoccupied him:                 of the small group. Bring            The Surprising New Science           workplaces opt for inter-
  the power of leaders to shape citizens’ behav-            housewives together into             of Psychological Change              ventions that not only have
  ior for good and ill, intergroup conflict and              a new group supportive of            by Timothy D. Wilson                 never been subjected to
  aggression, minority groups’ belongingness                change, freeing them from            Little, Brown, New York, 2011.       experimental test but also,
  and adjustment, de-Nazification and cultural               the grip of their old familial       287 pp. $25.99, C$28.99.             when they finally are, often
  transformation (“nation building” in today’s              norms, and they would try            ISBN 9780316051880.                  yield null and even nega-
  parlance), and so on (1, 2). At the heart of              the novel foods far more fre-                                             tive effects. These interven-
  Lewin’s approach rested a novel idea: social              quently than if they were lec-                                            tions are usually based on a
  problems are amenable to experimentation.                 tured to. Time and again, Lewin showed that combination of intuition, ideology, and good
  “The best way to understand something is                  what often seem problems of bad attitudes, intentions. Wilson critiques several popular
                                                                      lack of information, or economic but unwise interventions: Drug Abuse Resis-
                                                                      incentives were instead problems of tance Education (D.A.R.E., implemented
                                                                      group influence, identity, and social in 75% of the school districts in the United
                                                                      perception. But most revolutionary States), “scared straight,” certain forms of
                                                                      was Lewin’s method. There was a posttraumatic grief counseling, many com-
                                                                      combination of optimism and folly monplace diversity training programs, and
                                                                      in the idea that researchers could, the self-help and positive thinking industry
                                                                      through the experimental method, in general [The Secret (4) receives sustained
                                                                      change reality and improve social criticism]. These are analogous, Wilson
                                                                      conditions for the better (3).                writes, to the practices of leeching and blood-
                                                                          In Redirect: The Surprising New letting before the scientific method took hold
                                                                      Science of Psychological Change, in medicine. The amounts of money spent by
                                                                      Timothy Wilson reviews much of this schools, workplaces, and other institutions
                                                                      history and revisits the field of social on ineffective programs, and the degree to
                                                                      psychology 70 years after Lewin’s which intuition and ideology determine how
                                                                      pioneering work. He focuses on the taxpayer money is spent, are astonishing. But
                                                                      contributions of social psychology to all of us are culpable. When it comes to the
                                                                      understanding and remedying social mini-interventions that we naturally inflict
                                                                      problems. Wilson is a social psychol- on our family and co-workers, commonplace
  To change direction.                                                ogist at the University of Virginia practices that seem obviously salutary, like
                                                                      who has made groundbreaking dis- praising your child’s intelligence or reward-
  to try to change it,” he was fond of saying.              coveries in the study of intuition and introspec- ing people for their freely chosen good
  Beyond descriptive and correlational studies,             tion. In clear prose that does not trivialize the behavior, can backfire, sometimes dramati-
  Lewin championed experimental manipula-                   science, he reviews the many success stories cally. Wilson discusses how such unwise
                                                                                                                                                                      CREDIT: FRANCKREPORTER/ISTOCKPHOTO.COM

  tion: Introduce an exogenous shock to the                 in social psychology. There are interventions practices often arise from inaccurate theories
  system, and see how it responds.                          that harness the power of expressive writing of human nature.
      Lewin also advocated a diagnosis stage                and volunteerism to improve happiness and                   Redirect also helpfully reviews larger-
  in what he dubbed “action research”: First                health and to lessen rates of teen pregnancy. scale programs that reliably achieve posi-
  assess the relationships among variables in               There are interventions that reduce student tive results in randomized trials (such as Big
  a system. In doing so, one could identify the             failure and close gaps between minority and Brothers/Big Sisters) and various programs
  pressure points where a small nudge might                 nonminority students by inculcating in them to reduce teen violence and substance abuse
                                                            core positive beliefs that sustain them through (such as LifeSkills Training).
  The reviewer is at the School of Education and the        hardship, such as the belief that intelligence is           Wilson wants society to adopt more of
  Department of Psychology, 520 Galvez Mall, CERAS, Stan-
  ford University, Stanford, CA 94305, USA. E-mail: glc@    not a fixed entity but rather like a muscle that an experimental approach to solving social                                              grows with effort. There are interventions that problems—putting interventions to the test

178                                            14 OCTOBER 2011 VOL 334 SCIENCE
                                                                                                                                                                                                                    BOOKS ET AL.

                                                               with randomized controlled trials. This is a          own words to younger students. Seldom do             problem. Why are such interventions ignored
                                                               good idea, at least when the ambition is to           beneficiaries of these interventions feel sin-        in favor of ideology and intuition? What can
                                                               disseminate the interventions widely. How-            gled out as in need of help. The research-           we do to prevent this? What interventions
                                                               ever, one problem that Redirect does not              ers anticipate that stigmatizing people as           should we be implementing today? Nonethe-
                                                               explicitly address concerns limitations in the        “in need” could do more harm than provid-            less, Wilson’s account should not leave read-
                                                               experimental method itself. There is nothing          ing no help at all. Likewise, high-pressure          ers with the belief that social-psychological
                                                               better than an experiment for testing causal-         messages to teenagers about the risks of             experiments have failed to affect real-world
                                                               ity, whether an intervention A affects a social       poor nutrition do not work as well as, say,          practices and policies. When social psychol-
                                                               problem B. However, a positive experimental           getting them to participate in fun extracur-         ogy affects practice and policy, it usually
                                                               result risks deluding us into believing that A        ricular athletic activities or social-political      does so through a delayed, trickle-down pro-
                                                               is both necessary and sufficient to solve B (5).       causes that have hidden but beneficial side           cess (3, 10). For example, the medical estab-
                                                               But as Lewin taught us, the effect of A will          effects for health (7). The stealth, attention       lishment’s emphasis on patient-centered care
                                                               depend on the context into which it is intro-         to detail, and human touch found in many             follows in part from social-psychological
                                                               duced—the preexisting system of variables.            of these interventions could be lost in large-       research on the importance of predictability
                                                               Encourage students to see their academic              scale efforts to scale up (8, 9).                    and control in health. The currently popular
                                                               fates as within their own control, and they               Wilson uses the thought-provoking met-           notion that good social policy nudges people
                                                               will thrive, provided that they inhabit a class-      aphor of “story editing” to describe the             to make decisions consistent with their long-
                                                               room that provides them with opportunities            ingredient common to many of the success-            term self-interests (for instance, by making
                                                               for growth, such as committed teachers and            ful interventions he reviews. They alter the         retirement saving a default for employees,
                                                               quality instruction (6). Many of the interven-        narratives people tell themselves about their        which they are free to opt out of) similarly
                                                               tions Wilson reviews act like catalysts. They         world and their place in it: Is it safe or threat-   follows from decades of social-psychological
                                                               will not teach a student who cannot spell to          ening? Do I belong or not? Am I capable or           research. The book points to many interven-
                                                               spell, but they will encourage the student to         not? During sensitive periods, people’s story-       tions that could trickle down as well, inform-
                                                               seize opportunities to learn how. Because             telling can be redirected and the change can         ing education, health, and social policy.
                                                               the effects of interventions are context-             build on itself over time. Amend the opening             How common sense and ideology lead
                                                               dependent, there will be no silver bullets.           sentence of the story of your transition to col-     us astray in our attempts to fix social prob-
                                                                   Another question concerns how to scale            lege, or to a new job, and the arc of your story     lems, how surprisingly difficult it is to dis-
                                                               up the interventions to reach more people.            may be entirely different from what it would         cern whether a program works without a true
                                                               Many (though not all) of the interventions are        have been otherwise. This helps explain why          randomized experiment, and how sometimes
                                                               highly psychologically leveraged and care-            seemingly simple interventions, such as writ-        subtle social-psychological processes con-
                                                               fully crafted. The medium is as important as          ing about a traumatic experience, or volun-          tribute to big social problems constitute the
                                                               the message: You can’t simply tell students           teering for a humanitarian cause, improve            lessons of Redirect. As the scientist Paul C.
                                                               that their intelligence is expandable and that        health and well-being. They give people an           Stern once wrote, a policy objective of sci-
                                                               success is possible and then expect positive          organizing narrative that puts their lives in an     ence is to “separate common sense from
                                                               results. The message needs to be conveyed             optimistic context.                                  common nonsense and make uncommon
                                                               vividly, impactfully, and sometimes stealth-              Wilson compellingly argues that effec-           sense more common” (11). Wilson’s book
                                                               ily: for instance, by recruiting beneficiaries         tive interventions validated by social-science       does science and society a great service by
                                                               to deliver the message of optimism in their           research are rarely implemented. This is a           accomplishing precisely this.

                                                                                                                                                                               References and Notes
                                                                                                                                                                           1. K. Lewin, Resolving Social Conflicts: Selected Papers on
                                                                 BROWSINGS                                                                                                    Group Dynamics, G. W. Lewin, Ed. (Harper and Row,
                                                                                                                                                                              New York, 1948).
                                                                                                                                                                           2. K. Lewin, Field Theory in Social Science: Selected
                                                                 A Bee in a Cathedral: And 99 Other Scientific                                                                 Theoretical Papers, D. Cartwright, Ed. (Harper and Row,
                                                                 Analogies.                                                                                                   New York, 1951).
                                                                                                                                                                           3. L. Ross, R. Nisbett, The Person and the Situation:

                                                                 Joel Levy. Firefly, Richmond Hill, Ontario, 2011.
                                                                                                                                                                              Perspectives of Social Psychology (Pinter and Martin,
                                                                 224 pp. $29.95, C$29.95. ISBN 9781554079599.                                                                 London, ed. 2, 2011).
                                                                 Educators, science communicators, and research-                                                           4. R. Byrne, The Secret (Atria, New York, 2006).
                                                                 ers have long used the “x is like y” construction                                                         5. M. Woodhead, Am. Psychol. 43, 443 (1988).
                                                                                                                                                                           6. V. H. Menec et al., J. Appl. Soc. Psychol. 24, 675
                                                                 with familiar objects and actions to make
                                                                 scientific ideas clear to their audiences. Writer                                                          7. T. N. Robinson, in Obesity Prevention: The Role of Brain
                                                                 and journalist Levy presents 100 such easy-to-                                                               and Society on Individual Behavior, L. Dube et al., Eds.
                                                                 understand comparisons that illuminate facts                                                                 (Elsevier, New York, 2011), pp. 319–327.
                                                                 and principles from physical sciences, biology,                                                           8. L. Schorr, Common Purpose: Strengthening Families
                                                                                                                                                                              and Neighborhoods to Rebuild America (Doubleday,
                                                                 human anatomy and physiology, and technology.
                                                                                                                                                                              New York, 1998).
                                                                 These range from Johannes Kepler’s clockwork                                                              9. D. S. Yeager, G. M. Walton, Rev. Educ. Res. 81, 267 (2011).
                                                                 cosmos to John Searle’s “Chinese room” thought                                                           10. L. Ross, M. Lepper, A. Ward, in Handbook of Social Psy-
                                                                 experiment on artificial intelligence. For each analogy, the author supplements his concise description       chology, D. Gilbert, S. T. Fiske, D. T. Gilbert, G. Lindzey,
                                                                 with a set of related facts and figures. Designer Lindsey Johns’s two-page spreads effectively organize       Eds. (Wiley, New York, ed. 5, 2010), pp. 3–50.
                                                                                                                                                                          11. P. C. Stern, Science 260, 1897 (1993).
                                                                 the short text and accompanying colorful graphics. These falling cards (above), for example, flank a      12. I thank E. Ponin and D. Sherman for feedback.
                                                                 comparison of entropy with the disorder introduced by shuffling a freshly unwrapped pack of cards.

                                                                                             SCIENCE VOL 334 14 OCTOBER 2011                                                                                     179

      A Major Environmental                                                                                       Exposure to indoor air pollution from
                                                                                                                  household burning and solid fuels affects
                                                                                                                  nearly half of the world’s population.
      Cause of Death
      William J. Martin II,* Roger I. Glass,* John M. Balbus, Francis S. Collins

             he World Health Organi-                                                                                       Cookstove pollution in a Peruvian
             zation (WHO) lists indoor                                                                                     home. Better understanding of the health
             air pollution (IAP) ( 1)                                                                                      risks could drive demand for cleaner stoves.
      from primitive household cooking
      fires as the leading environmen-                                                                                        tion. By freeing up time, efficient
      tal cause of death in the world, as                                                                                    stoves can even expand the oppor-
      it contributes to nearly 2.0 mil-                                                                                      tunities for education and eco-
      lion deaths annually (2)—more                                                                                          nomic development of women and
      deaths than are caused each year                                                                                       girls in these impoverished areas.
      by malaria. Almost half of the                                                                                             Promoting sustained changes in
      planet lives in poverty, and those                                                                                     the way food is cooked to reduce
      households generally use biomass                                                                                       IAP requires a fundamental under-
      (wood, crop residues, charcoal, or                                                                                     standing of traditions, social inter-
      dung) or coal as fuel for cooking                                                                                      actions, and family dynamics,
      and heating. The primitive fires                                                                                       which differ widely across cul-
      typically fill homes with dense                                                                                        tures. Successful implementation
      smoke, blackening walls and ceilings and                  lem, limited research into the health risks,      invariably involves women in designing the
      sickening those within.                                   lack of affordable improved stoves or fuels       stove, training, use in the home, and follow-
          Women and children living in extreme                  that reduce exposures to safer levels, and the    up in the community.
      poverty are at highest risk for adverse health            logistical challenges of solving a problem
      outcomes from IAP. Whereas men tend to be                 that affects almost 3 billion of the poorest      Role of the Market
      physically removed from household smoke                   people on the planet.                             A successful strategy to enable poor individ-
      exposures during the day, women and chil-                     To address these needs, the United Nations    uals or households to adopt stoves or fuels
      dren suffer high exposures, which lead to                 Foundation launched the Global Alliance for       requires the creation of market demand. A
      many of the same disease risks as if they were            Clean Cookstoves (http://cleancookstoves.         stove purchased by the consumer is inher-
      lifelong smokers of tobacco. Mortality esti-              org). It is a public-private partnership aimed    ently more valued than one that is received
      mates from IAP are primarily based on risk                at creating a global market for clean and effi-    without charge, especially if the free stove
      for acute pneumonia in children under age 5               cient cookstoves and fuels in the developing      was designed without consumer input. Prior
      and chronic obstructive pulmonary disease                 world. The interim target of the alliance is      large-scale implementation programs have
      (COPD) (3) (see the chart, page 181).                     “100 by 20,” for 100 million homes to adopt       frequently been supply-driven, with decisions
          The consequences of primitive household               clean and efficient stoves and fuels by 2020,      about need and stove design driven from the
      cooking also extend to the global environ-                with the ultimate goal of universal adoption.     top. This has begun to change, as evidenced
      ment. For those at the bottom of the energy               The U.S. government has committed more            by recent successes, such as the national stove
      ladder, reliance on biomass fuels and coal                than $50 million, including about $25 million     program in Peru (6).
      contributes to local and regional environmen-             of the NIH’s ongoing research funds, to miti-         Ideally, a thriving global market in
      tal degradation and deforestation. A 2011                 gate the impacts from IAP. Secretary of State     improved stoves and fuels would spur local
      World Bank report underscores health ben-                 Hillary Clinton has made the Global Alliance      economic development. Such a market-driven
      efits as a rationale for cookstove interven-               a centerpiece of her Global Partnerships Ini-     strategy is likely the only feasible way to meet
      tions but also emphasizes evidence for envi-              tiative. The Alliance has enrolled more than      the enormous task of supplying improved
      ronment and climate benefits (4). Improved                 175 countries, foundations, corporations,         stoves or fuels, ultimately to as many as 600
      and efficient stoves reduce fuel use (reduc-               and other nongovernmental organizations           to 800 million households. But there are
      ing CO2 release) and, if sufficiently advanced,            (NGOs) as partners.                               challenges in creating sufficient demand.
      decrease black carbon emissions.                                                                            There are many competing needs for those in
          Although implementation of improved                   Role of Women                                     extreme poverty, and awareness of the risks
      stoves or fuels has been tried for decades, suc-          Women and girls typically gather fuel, a task     posed by cooking methods is limited.
      cess has been limited by a number of factors,             that is time-consuming and often places them          An important factor to the user in pur-
      including a lack of awareness of the prob-                at risk of gender-based violence, as they are     chasing an improved stove is that it can con-
                                                                                                                                                                          CREDIT: PILAR NORES

                                                                forced to walk several miles from the safety      sume much less fuel, offering an economic
      National Institutes of Health, Bethesda, MD 20892, USA.   of their villages (5). More efficient stoves can   return that can eventually offset the cost of a
      *Author for correspondence. E-mail: wjmartin@mail.nih.    reduce fuel consumption and thereby decrease      stove (4, 7). Whether this and other motivat-
      gov (W.J.M.); (R.I.G.)                the attendant risks associated with its collec-   ing factors will be enough to create the global

180                                                  14 OCTOBER 2011 VOL 334 SCIENCE

demand for hundreds of millions of stoves           can achieve enough exposure
is a hypothesis that the Global Alliance is         reductions to improve health, yet
                                                                                                                    872,000 children under age 5 from
willing to test. Carbon financing or govern-         remain affordable and accept-                                   acute lower respiratory infections
mental subsidies can encourage demand by            able to the poor. Simply provid-
                                                                                                                    1,057,000 adults from COPD
offsetting some of the initial costs of stoves.     ing stoves to a user is not enough.
                                                                                                                    36,000 adults from lung and upper
But there is a need to balance the efforts to       The user must also receive train-
                                                                                                                    respiratory cancers
reach the “bottom billion,” who require some        ing about and support for stove and
mechanism of subsidy, with those who are            fuel use to assure proper and sus-
less poor, for whom the reduced fuel costs          tained operation of the stoves and Annual deaths attributed to IAP (3).
will be sufficient motivation.                       attendant, risk-reducing behavioral
                                                    changes among all members of the family.      tially share interest in IAP health research.
Need for More Health Research                           We need both: more research trials to Access to such information might be facili-
Beyond the three diseases associated by             document rigorously the amount of reduc- tated by creating an interactive global map
WHO with IAP mortality (3), there are addi-         tion in IAP necessary to improve health, and showing project sites or partners, much like
tional known and suspected health risks asso-       also new approaches to evaluate the health the one already developed for participating
ciated with IAP as diverse as low birth weight,     benefits of major implementation programs NGO’s by the Partnership for Clean Indoor
burn injuries, cataracts, cardiovascular dis-       already under way.                            Air (
ease, asthma, and tuberculosis (8). There is a
great opportunity to determine whether inter-       New Approaches                                     Next Steps
ventions that reduce IAP can prevent them.          New approaches to program evaluation from          To be successful in reaching 100 million
Major implementation programs under way             other disciplines are being examined for pos-      households by 2020, governments, funders,
around the world offer opportunities to delin-      sible adaptation to the study of the proposed      manufacturers, and NGOs must commit to
eate the critical exposure-response relations       stove implementation programs. These are           implement large-scale solutions and also to
that underlie disease risk, and such research       as diverse as matched-pair randomization           support essential research. The challenges
is essential to provide evidence for the health     design used in the universal health insur-         are great, but the potential to use a relatively
benefits of improved stove or fuel strategies.       ance program in Mexico (11), causal infer-         low cost intervention to save millions of lives,
    Although problems of IAP are well estab-        ence assessment of a preexisting water and         improve the environment, and encourage
lished, it remains unclear just how much emis-      hygiene intervention study in India (12), and      economic development is compelling.
sions must be reduced by cleaner cookstoves         the use of the national evaluation platform
and fuels to provide substantial health benefits.    for assessment of large-scale implementation           References and Notes
                                                                                                        1. The term “indoor air pollution” is being replaced by
This has led to some controversy about how          programs (13).                                         “household air pollution.” An expert panel of the WHO is
best to proceed. One view is that the health            An important related need is for contin-           expected to make recommendations.
benefits need to be well documented before           ued technology advancements in exposure             2. WHO, Global Health Risks: Mortality and Burden of Dis-
major implementation programs move for-             assessment. Virtually all of the proposed              ease Attributable to Selected Major Risks (WHO, Geneva,
ward. The alternative view is that cookstove        research studies require access to personal,           GlobalHealthRisks_report_Front.pdf.
and fuel implementation programs by NGOs,           household, and outdoor exposure monitor-            3. WHO, Quantifying Environmental Health Impacts: Global
multinational companies, and governments            ing instruments that are accurate for moni-            Estimates of Burden of Disease Caused by Environmental
                                                                                                           Risks (WHO Geneva, 2009);
are increasingly under way, driven in part by       toring air pollutants associated with disease          ing_ehimpacts/global/globalair2004/en/index.html.
social and environmental concerns and by the        risk. Such instruments must be affordable and       4. World Bank, Household Cookstoves, Environment, Health
recent availability of new and improved tech-       adaptable for use in community settings of             and Climate Change: A New Look at an Old Problem
                                                                                                           (63217, World Bank Washington, DC, 2011); http://
nologies on the market. These programs are          lower-income countries. Field testing of these
moving forward whether or not the research is       monitors to measure the very high exposure             cookstoves-report.
available to document the health benefits of a       levels from IAP is needed, as well as the con-      5. E. Patrick, Forced Migration Rev. 27, 40 (2007).
specific type of intervention.                       tinued advancement of the technology that is        6. P. Nores, Science 334, 157 (2011).
                                                                                                        7. For example, in Guatemala, the total cost of US$122
    To date, the only completed randomized          available for investigators conducting both            includes the stove with table top, community demonstra-
controlled trial (RCT) using improved stoves        RCTs and program evaluation.                           tion, training, delivery, and a follow-up visit. The mini-
with chimneys to study the impact of reduced                                                               mum wage is US$280/month; however, if there is a 50%
                                                                                                           reduction in fuel use with a more efficient stove, a family
IAP on child pneumonia is the RESPIRE               Leveraging Existing Infrastructure                     can pay back the cost within 8 months (www.helpsintl.
study in Guatemala (9). Preliminary expo-           The costs for the high-priority research related       org).
sure-response data from RESPIRE suggest             to health and IAP would be roughly between          8. D. G. Fullerton, N. Bruce, S. B. Gordon, Trans. R. Soc.
that exposure reductions of as much as 90%          $150 and $200 million (14). Although such              Trop. Med. Hyg. 102, 843 (2008).
                                                                                                        9. N. Bruce et al., Bull. World Health Organ. 85, 535
are needed to achieve substantial reductions        numbers may seem high for an emerging                  (2007).
in pneumonia risk; even modest risk reduc-          field, they are typical of investments required     10. J. McCracken et al., Epidemiology 18(5), S185 (2007).
tion requires exposures to be lowered by at         to develop the evidence base to combat lead-       11. K. Imai, G. King, C. Nall, Stat. Sci. 24, 29 (2009).
                                                                                                       12. B. F. Arnold et al., Proc. Natl. Acad. Sci. U.S.A. 107,
least 50% (10).                                     ing sources of global mortality. Use of exist-         22605 (2010).
    It is not known if other health risks require   ing centers or networks that dovetail with the     13. C. G. Victora, R. E. Black, J. T. Boerma, J. Bryce, Lancet
such a marked reduction in exposures. If the        mission of IAP health research could reduce            377, 85 (2011).
                                                                                                       14. Health Burden of Indoor Air Pollution (IAP) on Women
results of RESPIRE are replicated in future         costs and improve the quality of research.
                                                                                                           and Children in Developing Countries, NIH Workshop,
RCTs for pneumonia and other health out-                There is value in establishing a global            Washington, DC, 9 to 11 May 2011.
comes, one of the challenges will be to imple-      inventory of relevant research sites and
ment improved stove and fuel programs that          implementation programs that would poten-                                             10.1126/science.1213088

                            SCIENCE VOL 334 14 OCTOBER 2011                                                                                 181
                                                                                                                        An antibacterial peptide is essential for

      Keeping Bacteria at a Distance                                                                                    restricting contact between the intestinal
                                                                                                                        microbiota and host.

      Malin E. V. Johansson and Gunnar C. Hansson

              he human intestine harbors enor-
              mous amounts of bacteria that have
              an essential role in host metabolism,
      but how this mutualistic balance is main-
      tained is unclear. The current understanding
      has focused on the concept that bacteria con-
      tinuously interact with the intestinal immune
      system in a balanced proinflammatory and
      tolerogenic way. The discovery of a protec-
      tive inner mucus layer in the colon that sepa-
      rates bacteria from the epithelium has broad-
      ened this view (1). On page 255 of this issue,
      Vaishnava et al. (2) show that the antibacterial
      protein RegIIIγ secreted by specialized epi-
      thelial cells is involved in limiting the epithe-
      lial contact with bacteria in the small intes-
      tine. This observation further substantiates
      the role of intestinal epithelial cells and the
      mucus that covers them as important parts of
      the innate immune defense.
          The small intestine and colon are quite
      different organs, especially when it comes                                                                                                               Inner mucus layer
      to the relation between bacteria and host.
      The colon is protected by an inner mucus
      layer that is firmly attached to the epithe-                       Goblet cells               Epithelium
      lium and protects it from bacteria and
      from mechanical stress. This layer is con-
      tinuously converted to an outer, less dense
                                                                                                  Paneth cells
      mucus layer, which is the habitat for the
      commensal bacteria (see the figure) (1, 3).                                 Small intestine crypt                                           Colon crypt
      By contrast, such a compact inner mucus
      layer in the small intestine would be detri-              Intestinal protection. The intestinal epithelium is covered by mucus. In the colon, the inner stratified mucus
      mental as the small intestine must absorb                 layer acts as a physical barrier separating bacteria from the epithelium. The commensal bacteria are kept at
      nutrients. Fast peristaltic propulsion in the             a distance in the penetrable outer mucus layer. The small intestine is only covered with penetrable mucus,
      distal direction, fluid and mucus secretion,               where antibacterial molecules such as RegIIIγ generate a bacteria-depleted region close to the epithelium.
      and antibacterial proteins instead maintain
      homeostasis in this intestinal region. The                these organs. In the small intestine, the               the epithelium and trigger an overt immune
      type of mucus that covers the small intes-                mucus gel acts as a mesh that retains the               reaction and colitis (1, 5).
      tine, unlike that of the colon, is not attached           antibacterial proteins capable of killing                   In the small intestine, invaginations of the
      to the epithelium and is permeable to bacte-              the trapped bacteria. The mucus gel is also             epithelium called crypts contain Paneth cells
      ria (4). Yet, Vaishnava et al. show that bac-             continuously replenished by secretion from              and the epithelial stem cells (6). The Paneth
      teria are kept at a distance from the epithe-             the mucus-secreting goblet cells, provid-               cells constitutively produce high amounts
      lium by the antibacterial protein RegIIIγ,                ing a rapidly renewable system protecting               of antibacterial peptides, lysozyme (hydro-
      which is secreted into the small intestinal               the small intestine. Vaishnava et al. show              lyzes bacterial cell wall peptidoglycans), and
      mucus by the enterocytes (2).                             that in RegIIIγ-deficient mice, more bac-                MUC2 mucin (7, 8). These antibacterial pep-
          The mucus of both the small intestine                 teria reach the small intestinal epithelium             tides (such as defensins) become trapped in
      and colon is organized around the net-like                and trigger the adaptive immune response                the intestinal mucus. Vaishnava et al. show
      polymer formed by the MUC2 mucin (3).                     with increased immunoglobulin A (pro-                   that the more common cells in the intesti-
                                                                                                                                                                                   CREDIT: B. STRAUCH/SCIENCE

      It is unclear why the same protein gives                  duced by B cells located beneath the epi-               nal epithelium, the enterocytes, have a regu-
      rise to mucus with different properties in                thelial cells, throughout the intestine) and T          lated feedback system through which bacte-
                                                                helper 1 cells. RegIIIγ is not produced in the          ria are sensed, probably by Toll-like receptors
                                                                colon, where instead the inner mucus layer              (TLRs). These receptors signal through the
      Department of Medical Biochemistry, University of
      Gothenburg, Gothenburg, Sweden. E-mail: gunnar.hansson@   provides protection. In the absence of the              adaptor protein MyD88 and induce RegIIIγ                                              MUC2 mucin in the colon, bacteria reach                 secretion. The enterocytes thereby control

182                                                 14 OCTOBER 2011 VOL 334 SCIENCE

                                     and limit the bacterial load at the epithelial              responds by secreting RegIIIγ, but also relays               sense and coordinate the host response to
                                     surface. RegIIIγ specifically affects Gram-                  information to the host’s adaptive immune                    intestinal bacteria.
                                     positive bacteria, and only the mucus-asso-                 system about microbial penetration of the
                                     ciated bacteria showed reduced numbers as                   mucus (10). Reciprocally, the intestinal epi-                     References
                                                                                                                                                               1. M. E. V. Johansson et al., Proc. Natl. Acad. Sci. U.S.A.
                                     compared to RegIIIγ-deficient mice. How                     thelium also responds to signals from the                        105, 15064 (2008).
                                     the number of commensal Gram-negative                       immune system to alter the mucus proper-                      2. S. Vaishnava et al., Science 334, 255 (2011).
                                     bacteria in the small intestine is controlled               ties and turnover as suggested from studies of                3. M. E. V. Johansson, J. M. Larsson, G. C. Hansson, Proc.
                                                                                                                                                                  Natl. Acad. Sci. U.S.A. 108 (suppl. 1), 4659 (2011).
                                     is not understood. No RegIIIγ-dependent                     parasite infections (8).                                      4. C. Atuma, V. Strugula, A. Allen, L. Holm, Am. J. Physiol.
                                     alterations in the luminal flora were evident                    The separation of bacteria and epithe-                       280, G922 (2001).
                                     in RegIIIγ-deficient mice suggesting that                   lium has emerged as a new concept under-                      5. M. Van der Sluis et al., Gastroenterology 131, 117 (2006).
                                                                                                                                                               6. N. Barker et al., Nature 449, 1003 (2007).
                                     the antibacterial protein was inactivated,                  lying host-microbiota homeostasis in the                      7. N. H. Salzman, D. Ghosh, K. M. Huttner, Y. Paterson,
                                     degraded, or sufficiently diluted when it                   small intestine and colon, although by dif-                      C. L. Bevins, Nature 422, 522 (2003).
                                     reached the intestinal lumen.                               ferent mechanisms. The different ways these                   8. M. A. McGuckin, S. K. Lindén, P. Sutton, T. H. Florin, Nat.
                                                                                                                                                                  Rev. Microbiol. 9, 265 (2011).
                                         The findings of Vaishnava et al. empha-                  organs solve the challenge of bacterial colo-                 9. M. Swamy, C. Jamora, W. Havran, A. Hayday, Nat. Immunol.
                                     size the role of intestinal epithelium and its              nization must have evolved to meet the dif-                      11, 656 (2010).
                                     enterocytes in orchestrating the intestinal                 ferent physiological needs. However, there                   10. A. Bas et al., Proc. Natl. Acad. Sci. U.S.A. 108, 4376
                                     defense system (9). The epithelium not only                 is still much to learn about how mucus prop-
                                     senses the intestinal bacterial milieu and                  erties are controlled and how enterocytes                                                        10.1126/science.1213909

                                     MATERIALS SCIENCE

                                                                                                                                                              A set of simple rules is used to design
                                     Self-Assembly Enters the Design Era                                                                                      and control the self-assembly of nanoparticles
                                                                                                                                                              into complex structures.
                                     Alex Travesset

                                               anotechnology holds the key to the                eral periodic structures of nanoparticles. The                  There are many open questions as well
                                               discovery of materials with the elec-             work of Macfarlane et al. is likely to elevate               as exciting opportunities following from this
                                               tronic, optical, mechanical, or trans-            DNA-programmed self-assembly into a tech-                    work. For example, the crystalline structures
                                     port properties required to overcome many                   nique for the design of nanoparticle structures              reported are limited in size, typically involv-
                                     of today’s technological challenges. The past               a la carte. They report nanoparticle crystals                ing a few thousand nanoparticles. The ori-
                                     two decades have seen impressive progress                   with nine different space symmetries, show                   gin of this limitation is unclear. The authors
                                     in the synthesis of nanoparticles with aston-               how to precisely tune the lattice constant to                have gone to impressive lengths to show the
                                     ishing properties (1). Bringing the poten-                  basically any arbitrary value, and provide six               reproducibility of the results under a wide
                                     tial of nanotechnology to reality requires an               simple rules that allow them to predict the                  range of diverse conditions; hence, there
                                     in-depth understanding of how to precisely                  lattice structure into which the nanoparticles               must be a physical effect behind this size
                                     assemble nanoparticles into structures and/                 will self-assemble (see the figure).                          limitation. Experimental studies focusing on
                                     or phases over multiple length scales, but this
                                     has proven enormously difficult to accom-
                                     plish. Assembling nanoparticles into peri-                    Nanoparticles in solution:                                              Attach complementary DNA strands:
                                     odic structures, for example, has only been
                                     achieved for a handful of systems under very                                 +          = no ordered structures
                                     specific conditions (2). On page 204 of this
                                     issue, Macfarlane et al. (3) show that the use
                                     of single-stranded DNA as linkers provides a
                                     general strategy to program the self-assembly
                                     of almost any nanoparticle into a wide range
                                     of different periodic structures and do so with
                                                                                                                        +              =
                                     an exquisite control over their properties.
                                         In the mid-1990s the method of using
                                     complementary DNA strands as linkers to
                                     direct the self-assembly of nanoparticles was                                    Sticky ends                                        AB2 lattice

                                     pioneered (4, 5). A decade later saw the major                                                            CsCl lattice
                                     breakthrough of programming gold nanopar-                        Lattice controllable by:                                                                   CS6C60 lattice
                                     ticles to self-assemble into lattices (6, 7), thus                 •Size
                                                                                                        •Length of sticky ends                                       + Other lattices
                                     opening the door for a rational design of gen-
                                                                                                 Putting it together. Assembling nanoparticles into structures presents extraordinary difficulties. Attaching
                                     Department of Physics and Astronomy and Ames Labora-        them with complementary single-stranded DNA linkers (or “sticky ends”) overcomes all these difficulties,
                                     tory, Iowa State University, Ames, IA 50011, USA. E-mail:   resulting in the controlled self-assembly of nanoparticles with exquisite control over their properties. AB2 is a
                                                                     lattice isostructural with aluminum diboride.

                                                                        SCIENCE VOL 334 14 OCTOBER 2011                                                                                             183

      the dynamics of self-assembly (8) suitably         an elegant solution to this crucial problem in                      5. A. P. Alivisatos et al., Nature 382, 609 (1996).
                                                                                                                             6. S. Y. Park et al., Nature 451, 553 (2008).
      complemented with theoretical models of            materials science.                                                  7. D. Nykypanchuk, M. M. Maye, D. van der Lelie, O. Gang,
      the dynamics (9, 10) may clarify this issue.          Being able to assemble nanoparticles with                           Nature 451, 549 (2008).
      The ability to extend the size of crystals and     such control represents a major accomplish-                         8. R. J. Macfarlane et al., Proc. Natl. Acad. Sci. U.S.A. 106,
                                                                                                                                10493 (2009).
      even realize single crystals would dramati-        ment in our quest to manipulate matter. There                       9. C. Knorowski, S. Burleigh, A. Travesset, Phys. Rev. Lett.
      cally enhance the already huge value of these      are immediate important applications related                           106, 215501 (2011).
      materials. Other studies will be necessary to      to catalysis, medical sensing, new optical                         10. C. Knorowski, A. Travesset, Curr. Opin. Solid State Mater.
                                                                                                                                Sci. 10.1016/j.cossms.2011.07.002 (2011).
      characterize, control, or adjust the rheologi-     materials or metamaterials, and others that                        11. M. R. Jones et al., Nat. Mater. 9, 913 (2010).
      cal properties of these materials.                 will follow from these studies. Most likely,                       12. A. V. Tkachenko, Phys. Rev. Lett. 106, 255501 (2011).
          With such an impressive inventory of           however, many other applications will arise                        13. M.-P. Valignat, O. Theodoly, J. C. Crocker, W. B. Russel, P. M.
                                                                                                                                Chaikin, Proc. Natl. Acad. Sci. U.S.A. 102, 4225 (2005).
      nanoparticles available (1), one direction is to   as we dig deeper, understand better, expand                        14. N. C. Seeman, Annu. Rev. Biochem. 79, 65 (2010).
      classify the self-assembled structures result-     further, and tinker with the opportunities pro-                    15. S. K. Kumar, R. Krishnamoorti, Annu. Rev. Chem. Biomol.
      ing from the different properties of these         vided by these materials.                                              Eng. 1, 37 (2010).
                                                                                                                            16. R. Sknepnek, J. A. Anderson, M. H. Lamm, J. Schmalian,
      nanoparticles. Some preliminary work focus-                                                                               A. Travesset, ACS Nano 2, 1259 (2008).
                                                             References and Notes
      ing on nanoparticle geometry has already            1. S. C. Glotzer, M. J. Solomon, Nat. Mater. 6, 557 (2007).       17. I am indebted to my collaborators in this project, J.
                                                                                                                                Anderson, R. Sknepnek, and especially C. Knorowski.
      appeared (11), but many more possibilities          2. E. V. Shevchenko, D. V. Talapin, N. A. Kotov, S. O’Brien, C.
                                                                                                                                Supported by the U.S. Department of Energy through the
      remain to be explored.                                 B. Murray, Nature 439, 55 (2006).
                                                          3. R. J. Macfarlane et al., Science 334, 204 (2011).                  Ames Lab under contract DE-AC02-07CH11358.
          So far, nanoparticles with only a single        4. C. A. Mirkin, R. L. Letsinger, R. C. Mucic, J. J. Storhoff,
      species of DNA linker have been considered.            Nature 382, 607 (1996).                                                                              10.1126/science.1213070
      With the current sophistication in experi-
      mental methods, nanoparticles consisting of
      patches with different linkers can be synthe-
      sized, thus endowing them with an anisotro-        EVOLUTION
      pic interaction or discrete valence (12). This
      possibility is briefly explored in the work of
      Macfarlane et al., where a given nanoparti-        The Costs of Breathing
      cle is endowed with two different linker types
      resulting in NaCl and simple cubic lattices.       Nick Lane
      DNA-programmed self-assembly has also
      been extended to particles with diameters          Selection for respiratory function has implications for organism fitness, fertility, and life span.
      on the micrometer scale (13). However, the

      resulting structures do not show the same                  ukaryotic cell respiration depends on                      respiratory chains composed of numer-
      degree of order as in the nanometer-scale                  the interactions of proteins encoded                       ous interacting protein subunits encoded
      experiments. The largest nanoparticles con-                by two genomes, mitochondrial and                          in both genomes. Electrons normally pass
      sidered by Macfarlane et al. are only about        nuclear, which evolve in radically different                       swiftly down the full length of the respira-
      one order of magnitude smaller, so it seems        ways. Mitochondrial genes evolve asexually                         tory chain, and ultimately reduce oxygen
      likely that smooth interpolation between           (mitochondrial DNA is generally passed                             safely to water, the energy released driving
      these diverse size scales is in sight. There has   from mother to offspring without recom-                            adenosine 5′-triphosphate (ATP) synthesis.
      been amazing progress in building structures       bination), unlike nuclear genes, and their                         But if the progress of electrons down the full
      with DNA, ranging from DNA origami to              mutation rate can be orders of magnitude                           length of this chain is blocked for any rea-
      three-dimensional periodic structures where        faster than the nuclear average (1). Despite                       son, they are more likely to escape and react
      nanoparticles can be attached (14)—provid-         these differences, the two genomes coad-                           directly with oxygen to form reactive, par-
      ing even more exciting opportunities.              apt to each other over evolutionary time                           tially reduced intermediates known as reac-
          The implications of the work extend to         (2): Mutations in one genome are offset by                         tive oxygen species (ROS). In the same way,
      less obvious situations. A major unresolved        changes in the other, preserving respiratory                       a partially dammed stream is more likely
      problem, with very far-reaching applica-           function and possibly adapting it to changes                       to burst its banks. Any impedance to elec-
      tions in materials science, has been how to        in diet and climate (3). The details of selec-                     tron flux increases ROS leak, and—because
      design materials consisting of polymers and        tion may hold surprising implications for fit-                      fewer electrons make it down the whole
      nanoparticles such that the nanoparticles are      ness, fertility, and aging.                                        respiratory chain in a given time—reduces
      homogeneously dispersed (do not cluster)              Outcrossing between different species or                        the amount of ATP produced. This combina-
      and display long-range order. There are very       genetically distinct populations can lead to                       tion of high ROS leak and low ATP synthesis
      few experimental examples, if any, where this      hybrid breakdown—that is, poorly viable or                         is the classic trigger for the release of cyto-
      has been successfully accomplished (15). It        inviable offspring—through the loss of mito-                       chrome c (a component of the respiratory
      has been predicted theoretically (10, 16) that     nuclear coadaptation (2, 4). The problem is                        chain) from the mitochondria, an event that
      generalizations of the work of Macfarlane et       that cellular respiration (which takes place in                    initiates programmed cell death (apoptosis)
      al. polymers with attached single-stranded         mitochondria) depends on the flow of elec-                          in most eukaryotic cells.
      DNA complementary to nanoparticle link-            trons to oxygen or other acceptors through                             This is exactly the problem that occurs
      ers would lead to polymer nanocomposites                                                                              if the mitochondrial and nuclear genes are
      with homogeneous nanoparticle dispersion           Department of Genetics, Evolution and Environment, Uni-
                                                                                                                            mismatched through outcrossing between
      and displaying long-range order with a wide        versity College London, Gower Street, London WC1E 6BT,             species or genetically distant populations.
      range of lattice symmetries, thus resulting in     UK. E-mail:                                    The two sets of genes encode proteins that

184                                           14 OCTOBER 2011 VOL 334 SCIENCE

                                                                                                  must operate together with nanoscopic pre-                                                                 ROS leak in pigeons is a fraction of that in
                                                                                                                                                        Variable threshold
                                                                                                  cision to convey electrons down the respira-             for ROS leak                                      rats, and pigeons live up to 10 times longer.
                                                                                                  tory chain, and given the major disparities                                                                The apoptotic threshold view is a possible
                                                                                                                                                                         Apoptosis; decreased
                                                                                                  in their tempo and mode of evolution, only                              organism survival                  explanation for why birds and bats should
                                                                                                  selection for coadaptation can ensure that                                                                 have low ROS leak: They require a high aero-
                                                                                                  they do indeed cooperate properly together.                                                                bic capacity and good mitonuclear match, so
                                                                                                  Outcrossing can undermine this exquisite                                                                   only individuals with low ROS leak make it
                                                                                                  coadaptation. Proteins encoded by maternal                                                                 through embryonic development. Thus, the
                                                                                                  mitochondrial genes do not interact properly                                                               link between low ROS leak and life span may
                                                                                                  with proteins encoded by genetically distant                                                               not necessarily be the result of direct selec-
                                                                                                  paternal nuclear genes. This physical mis-                                                                 tion, but a side-effect of selection for mito-
                                                                                                  match impedes electron flow, compromising                                                                   nuclear match, optimizing fitness and fertility
                                                                                                  respiration. ROS leak rises, ATP synthesis                                                                 over generations.
                                                                                                  falls, and cytochrome c is released, triggering                                                                The apoptotic threshold could also oper-
                                                                                                  apoptosis. This in turn undermines develop-                                                                ate within each generation, affecting aging
                                                                                                  ment, leading to low fitness, sterility, inviabil-                                                          and disease. The problem is that mutations
                                                                                                  ity, or developmental abnormalities (4), out-                                                              in mitochondrial DNA arise over time as a
                                                                                                  comes that potentially are a step en route to                                                              result of simple usage; and mitochondrial
                                                                                                  speciation (5).                                                                                            heteroplasmy (a mixture of mutant and nor-
                                                                                                      Thus, the involvement of respiratory pro-                                                              mal DNA in the mitochondria of the same
                                                                                                  teins like cytochrome c in apoptosis makes                                                                 cell) is indeed common in aging tissues and
                                                                                                  good sense as a form of selection for cells                                                                many tumors (8). Heteroplasmy potentiates
                                                                                                  and organisms with functional respiration                                                                  mitonuclear mismatch, increasing ROS leak
                                                                                                  (6). Mitonuclear mismatch leads to high                                                                    and the likelihood of apoptosis, which can
                                                                                                  ROS leak and loss of cytochrome c, so cells                                                                lead to tissue loss. As the most metaboli-
                                                                                                  with compromised respiration are eliminated                                                                cally active tissues—notably brain and mus-
                                                                                                  by apoptosis. Because mitochondrial genes           The apoptotic threshold. In this model, if the         cle—approach the apoptotic threshold faster,
                                                                                                                                                      threshold for ROS leak from the respiratory chain
                                                                                                  operate against a new nuclear background                                                                   they are preferentially lost, which can lead
                                                                                                                                                      in mitochondria is low, even a small increase in ROS
                                                                                                  every generation, selection for mitonuclear         leak during cellular respiration triggers apoptosis.   to degenerative diseases. Some cells escape
                                                                                                  coadaptation must normally take place every         A low tolerance for ROS leak selects for good mito-    that fate. Senescent cells survive by glycoly-
                                                                                                  generation. And because the new nuclear             nuclear match; high tolerance raises the threshold     sis, often with high mitochondrial ROS leak,
                                                                                                  background is only resolved after zygote for-       for apoptosis and relaxes selection for mitonuclear    which gives rise to oxidative stress. This
                                                                                                  mation, this selection must take place during       match. See SOM text for suggested resources related    drives epigenetic changes and the expression
                                                                                                  embryonic development or after birth.               to mitonuclear coadaptation.                           of proinflammatory factors (3) associated
                                                                                                      But how good does mitonuclear match                                                                    with chronic inflammatory conditions such
                                                                                                  need to be? To attain optimal respiration,          tility, life span, and age-related diseases (see       as diabetes, and cancer (9).

                                                                                                  electrons must flow freely down the full respi-      the figure). For example, if there is a require-            Mitonuclear coadaptation can explain the
                                                                                                  ratory chain to oxygen, and any impedance           ment for good mitonuclear match (fast elec-            relationship between fitness, fertility, and
                                                                                                  must be eliminated. That could mean elimi-          tron flux, hence low ROS leakage during res-            life span across species through a simple
                                                                                                  nating even slight mitonuclear mismatches           piration), then organisms with poor matches            biophysical mechanism, ROS leak. If true,
                                                                                                  via apoptosis, which in turn would mean sac-        should be eliminated during development                this view has implications for how we might
                                                                                                  rificing even mildly substandard embryos.            as a result of high sensitivity to ROS leak.           tackle aging and age-related disease. Empiri-
                                                                                                  Such sacrifice could only make sense if respi-       If all embryos with even modest ROS leak               cal investigation will reveal if genome match-
                                                                                                  ratory demand is high. High energy demand,          are eliminated during embryonic develop-               ing has played, and continues to play, a role in
                                                                                                  as in flight in bats and birds, certainly requires   ment, then the few individuals that do sur-            eukaryotic evolution.
                                                                                                  fast electron flux, and so stricter selection        vive development should have low mitochon-
                                                                                                  for mitonuclear match could be beneficial           drial ROS leak. The prediction is that fertility        1. W. Fan et al., Science 319, 958 (2008).
                                                                                                  despite the high cost. From this point of view,     and ROS leak should covary across species.              2. D. M. Rand, R. A. Haney, A. J. Fry, Trends Ecol. Evol. 19,
                                                                                                  even a small increase in ROS leak would             Flighted birds and bats are predicted to have              645 (2004).
                                                                                                                                                                                                              3. D. C. Wallace, Proc. Natl. Acad. Sci. U.S.A. 107 (suppl. 2),
                                                                                                  betray mitonuclear mismatch, and might              low ROS leak (which they do), corresponding                8947 (2010).
                                                                                                  therefore be penalized by apoptosis during          to low fertility, and small litter sizes, whereas       4. C. K. Ellison, R. S. Burton, Evolution 62, 631 (2008).
                                                                                                  embryonic development. On the other hand,           rats should have higher ROS leak, higher fer-           5. J. B. W. Wolf, J. Lindell, N. Backström, Philos. Trans. R.
                                                                                                  such a stringent penalty could hardly benefit        tility, and larger litter sizes. ROS leak is there-        Soc. Lond. B Biol. Sci. 365, 1717 (2010).
                                                                                                                                                                                                              6. N. W. Blackstone, D. R. Green, Bioessays 21, 84 (1999).
                                                                                                  animals with less exacting energetic require-       fore, according to the model, critical to selec-        7. G. Barja, Rejuv. Res. 10, 215 (2007).
                                                                                                  ments. Presumably, animals with lower aero-         tion: Tolerance of high ROS leak can be ben-            8. Y. He et al., Nature 464, 610 (2010).
                                                                                                  bic demands, like rats, should tolerate more        eficial and positively selected; it is not simply        9. D. R. Green, L. Galluzzi, G. Kroemer, Science 333, 1109
                                                                                                  ROS leak during embryonic development.              an unwelcome side-effect of respiration.
                                                                                                  If so, that would imply a variable “apoptotic            That has an important corollary. The rate             Supporting Online Material
                                                                                                  threshold” across species.                          of ROS leak corresponds closely to life span     
                                                                                                      A variable apoptotic threshold model            across species (7). Pigeons and rats have sim-             SOM Text
                                                                                                  allows testable predictions about fitness, fer-      ilar body size and resting metabolic rate, but                                              10.1126/science.1214012

                                                                                                                               SCIENCE VOL 334 14 OCTOBER 2011                                                                                      185

                                                                                                                          The endoplasmic reticulum is an active

      SevERing Mitochondria                                                                                               participant in the division of another organelle,
                                                                                                                          the mitochondrion.
      Angelika S. Rambold and Jennifer Lippincott-Schwartz

                 itochondria are the cell’s meta-                       in a process that depends on the membrane         constrictions at sites of ER-mitochondrial
                 bolic headquarters, fueling oxi-                       protein mitochondrial fission factor (Mff         contact were still observed in the absence of
                 dative phosphor ylation for                            in mammals) (11). Whereas Dnm1/Drp1               Drp1 or Mff, which suggests that constric-
      adenosine 5′-triphosphate production, and                         assemblies are always found at mitochon-          tion preceding full fission is independent
      driving reactions to manufacture core metab-                      drial fission sites, not all Dnm1/Drp1 sites       of both proteins and is potentially mediated
      olites for the biosynthesis of fats, DNA,                         define sites of division (7). Friedman et al.      by the ER-mitochondria contact itself. This
      and proteins. In addition to their metabolic                      discover that ER-mitochondria interaction         might occur if ER-mitochondrial tethering
      roles, these organelles regulate various cel-                     sites mark the sites of mitochondrial fission      proteins interact as ER tubules wrap around
      lular processes, including proliferation (1),                     and suggest that the ER may be an active par-     mitochondria, and could actively contribute
      immune responses (2–4), and apoptotic cell                        ticipant in mitochondrial division.               to preconstriction, and/or to the fission pro-
      death (5). Mitochondrial function in many                             The ER and mitochondria exhibit tightly       cess itself.
      of these processes is coupled to their spe-                       coupled dynamics and have extensive con-             The findings of Friedman et al. raise
      cific morphology, which ranges from small                          tacts ( 12). Using sophisticated imaging          questions about whether the ER actively
      individual mitochondrial elements to large                        approaches, Friedman et al. found that mito-      contributes to mitochondrial fission and
      interconnected networks (6). These diverse                        chondrial fission occurs at regions of contact     what proteins are required. The authors
      shapes result from fission into smaller forms                      between mitochondria and ER that are char-        hypothesize that partially constricting mito-

                                                                                                          Mitochondrial fission

                                                                         Mff                              Mff


                                                                      Drp1                    ER tubule

      or fusion into larger structures (7, 8), events                   acterized by ER tubules crossing and nearly       ER-dependent mitochondrial fission. In this
      thought to occur autonomously. However, a                         enwrapping mitochondria. At, or close to,         model, the major mitochondrial fission protein Drp1
      study by Friedman et al. in Science Express                       these ER-mitochondrial contact sites, the         is recruited from cytosolic or mitochondrial spots
      (9) reveals that mitochondrial division is                        mitochondria first become partially con-          to ER-mitochondria contacts sites (containing the
                                                                                                                          Drp1-receptor Mff). At sites of interorganellar con-
      intimately coupled to another organelle, the                      stricted, and then completely divide. Notably,
                                                                                                                          tact, Drp1 stabilizaton and/or oligomerization into
      endoplasmic reticulum (ER).                                       the ER-mitochondrial contact sites involve        helices might be facilitated by ER tubule–dependent
          Mitochondrial division is driven by the                       the activity of ER tubules; sheetlike regions     mitochondrial constriction.
      fission protein Dnm1 in yeast, and its homo-                       of the ER did not appear to be important.
      log Drp1 in higher eukaryotes. Dnm1 and                           Even shifting ER morphology toward sheet-         chondria could physically be important for
      Drp1 belong to a family of dynamin-related                        like architecture (through deletion of the        assembling Drp1 helices around mitochon-
      proteins that are thought to regulate mem-                        yeast ER-shaping proteins, Rtns/Yop1) did         dria, because Drp1 ring width is smaller
      brane fission by forming contractile helices                       not prevent small tubules from extending off      than the mitochondrial diameter. ER-
      that wrap around membrane to constrict it                         sheetlike ER to interact with mitochondria.       induced mitochondrial constriction could
      (requiring guanosine 5′-triphosphate hydro-                           Time-lapse imaging further identified         therefore facilitate the stabilization and/or
      lysis) (7, 10). Most Drp1 in higher eukary-                       the sequential nature of ER-based mito-           oligomerization of Drp1 at division sites.
      otes is dispersed throughout the cytosol; only                    chondrial division. The appearance of ER-         But proteins residing on the ER or mito-
                                                                                                                                                                                 CREDIT: B. STRAUCH/SCIENCE

      a small fraction is recruited onto mitochon-                      mitochondria contact sites coincides with         chondria could also contribute to mitochon-
      dria and assembles into oligomeric structures                     constriction of mitochondria at these sites.      drial constriction and/or fission. Potential
                                                                        Drp1 is then recruited to these sites, either     candidates include the tethering proteins
                                                                        from the cytosol or from other mitochon-          Mfn2 in higher eukaryotes or a complex in
      The Eunice Kennedy Shriver National Institute of Child Health
      and Human Development, National Institutes of Health,             drial sites, and is followed by mitochondrial     yeast called the ER-mitochondria encounter
      Bethesda, MD 20892, USA. E-mail:            fission (see the figure). Initial mitochondrial     structure (13, 14). Other ER-mitochondria

186                                                     14 OCTOBER 2011 VOL 334 SCIENCE

functions such as calcium or lipid transfer           of mitochondrial fission, such as neurons.                            3. R. Zhou, A. S. Yazdi, P. Menu, J. Tschopp, Nature 469,
(12, 15) also could play a role during fission.        Even a slight reduction in mitochondrial                                221 (2011).
                                                                                                                           4. K. Yasukawa et al., Sci. Signal. 2, ra47 (2009).
ER-mediated release of calcium could spa-             fission in these cells may contribute to so-                          5. G. Kroemer, L. Galluzzi, C. Brenner, Physiol. Rev. 87, 99
tially activate Drp1-dependent fission close           called ER-architecture diseases, such as                                (2007).
to ER-mitochondria contacts (7). Further-             hereditary spastic paraplegia (16, 17). Fur-                         6. E. Braschi, H. M. McBride, Bioessays 32, 958 (2010).
                                                                                                                           7. L. L. Lackner, J. M. Nunnari, Biochim. Biophys. Acta
more, changes in mitochondrial lipid compo-           thermore, the connection between ER and                                 1792, 1138 (2009).
sition, as a consequence of ER-mitochondria           mitochondrial fission may be related to                              8. V. Soubannier, H. M. McBride, Biochim. Biophys. Acta
lipid transfer, could contribute to fission.           other processes, such as the segregation                                1793, 154 (2009).
                                                                                                                           9. J. R. Friedman et al., Science, 1 September 2011
    Clearly, tight ER-mitochondrial cou-              or replication of mitochondrial DNA, and
pling could contribute to mitochondrial fis-           other cellular functions, such as apoptotic                         10. G. J. Praefcke, H. T. McMahon, Nat. Rev. Mol. Cell Biol. 5,
sion in several ways, but it is unclear whether       cell death. There may even be bidirectional                             133 (2004).
mitochondria-ER contacts are essential for            coupling between mitochondrial fission                              11. H. Otera et al., J. Cell Biol. 191, 1141 (2010).
                                                                                                                          12. O. M. de Brito, L. Scorrano, EMBO J. 29, 2715 (2010).
fission. Friedman et al. extend the complex-           and ER activities, in which ER-facilitated                          13. O. M. de Brito, L. Scorrano, Nature 456, 605 (2008).
ity of mitochondrial fission, with potential           mitochondrial fission produces feedback to                           14. B. Kornmann, P. Walter, J. Cell Sci. 123, 1389 (2010).
implications for diseases of ER- or mito-             affect ER properties.                                               15. A. R. English, N. Zurek, G. K. Voeltz, Curr. Opin. Cell Biol.
                                                                                                                              21, 596 (2009).
chondrial-shaping proteins. Although the                                                                                  16. J. Hu et al., Cell 138, 549 (2009).
authors report that changes in ER architec-                  References                                                   17. S. H. Park, P. P. Zhu, R. L. Parker, C. Blackstone, J. Clin.
                                                          1. K. Mitra, C. Wunder, B. Roysam, G. Lin, J. Lippincott-
ture do not appreciably affect mitochon-                     Schwartz, Proc. Natl. Acad. Sci. U.S.A. 106, 11960 (2009).
                                                                                                                              Invest. 120, 1097 (2010).
drial fission in yeast, specific mammalian                  2. O. Kepp, L. Galluzzi, G. Kroemer, Nat. Immunol. 12, 199
cells could be sensitive to malfunctions                     (2011).                                                                                          10.1126/science.1214059


Shining Light on Diabolic Points                                                                                          A direct probe of the electronic dynamics where
                                                                                                                          excited states intersect has been achieved by
                                                                                                                          using high-harmonic spectroscopy.
Benjamin J. Whitaker

        hemical reactions are described at a          used high-harmonic generation, a technique                          nuclear coordinates (2, 3). Conical intersec-
        fundamental level in terms of how             for converting optical laser pulses into higher                     tions play a crucial role in the photochemis-
        the potential energy of molecules             frequency radiation, to resolve the electronic                      try of polyatomic molecules because they act
changes as a function of distance between             state of molecules at diabolic points.                              as funnels that convert the electronic energy
the constituent atoms. These potential energy             An example of the simplest conical inter-                       of excited states into nuclear kinetic energy.
surfaces, which describe the barriers that            section, where two surfaces meet at a single                        Normally, such transitions between elec-
must be overcome as reactants form prod-              point, is shown in the lower right in the fig-                       tronic states lose energy by emitting photons,
ucts, are calculated with molecular orbital           ure. More generally, for higher dimensional                         but at conical intersections, these processes
theory, which approximates the real wave              potential surfaces, the singularity is described                    are nonradiative and waste no energy. Coni-
functions of molecules as combinations (con-          by a conical intersection seam in the space of                      cal intersections are the key to understanding
figurations) of many one-electron wave
functions. This approach is accepted by                                                                                                Brief but important meetings. Coni-
chemists because it accounts for the rates                                                                                             cal intersections, or diabolic points, are
of many reactions, but can it be directly                                                                                              critical points on the molecular poten-
verified? On page 208 of this issue,                                                                                                   tial energy landscape where states of the
                                                                                                                                       same symmetry meet. They play a crucial
Wörner et al. (1) test the validity of these
                                                                                                                                       role in photochemistry because they act
descriptions of electronic structure.                                                                                                  as conduits to convert electronic energy
    Their experiment probed the elec-                                                                                                  into nuclear kinetic energy. A simple ver-
tron dynamics of a molecule at a coni-                                                                                                 sion of such a point is shown in the lower
cal intersection, or diabolic point, where                                                                                             right. The main panel shows a photon of
two different electronic states cross at                                                                                               energy ħω exciting a molecule from the
the same energy and create strong cou-                                                                                                 ground-state electronic surface (red) to
pling between nuclear and electronic                                                                                                   the first excited-state surface (green). The
motions. The name diabolic comes from                                                                                                  trajectory shown in blue depicts a vibra-
the resemblance of the intersection to                                                                                                 tional motion induced by the transition.
                                                                                                                                       The molecule crosses back to the ground
the juggler’s Diabolo top on a string and             h
                                                                                                                                       state at the diabolic point depicted with
to conjure the notion that the molecular                                                                                               the arrow. Wörner et al. ( 1) literally
dynamics close to these points is dev-                                                                                                 shined light on these points by using
ilishly tricky to calculate. The authors                                                                                               ultrafast laser pulses to probe the elec-
                                                                                                                                       tronic dynamics directly as the electronic
School of Chemistry, University of Leeds, Leeds LS2                                                                                    character of an NO2 molecule changes on
9JT, UK. E-mail:                                                                                              passing through a diabolic point.

                                    SCIENCE VOL 334 14 OCTOBER 2011                                                                                               187

      how electronic energy is converted into heat         uration is largely blind to the configuration             on the excited-state surface. Although the
      and work and for understanding photochem-            interactions in the true, many-electron wave             nuclear motion across a conical intersection
      ical processes such as photosynthesis, the           function (8, 9).                                         has been observed before (4, 5), this study
      mechanism of vision, electron transfer reac-             This framework is important for inter-               reports the first direct observation of the
      tions, and the photostability of DNA.                preting the results of Wörner et al. (1). They           dynamic rearrangement of electronic con-
          Along conical intersections, small               used laser pulses to pump a molecule, nitro-             figuration in the vicinity of a conical inter-
      changes in molecular geometry can have dra-          gen dioxide (NO2), into an excited electronic            section. Thus, the use of extremely fast light
      matic effects on the electronic configuration         state before observing the HHS with a sec-               pulses—on close to attosecond time scales—
      of a molecule. These effects of the nuclear          ond intense pulse. They created a spatial dis-           has allowed the observation of events trig-
      dynamics can be picked up by measuring the           tribution of excited-state molecules by creat-           gered by electronic transitions in molecules
      photoelectron spectrum in a time-resolved            ing an interference pattern with two identical           and shone light on the description of their
      experiment in which a short-duration laser           pump pulses, which allowed them to separate              electronic structure.
      pulse excites the molecule (called the pump          the spectrum of ground-state molecules from
      pulse) and, after a short time delay, a second       the electronically excited ones. The interest-               References and Notes
                                                                                                                     1. H. J. Wörner et al., Science 334, 208 (2011).
      probe pulse ionizes it. However, the results         ing observation is that the HHS of the excited            2. Technically, for an N-atom molecule, the energy is plot-
      are ambiguous in terms of which electronic           state is sensitive to the time delay between the             ted as a 3N – 6 dimensional hypersurface, and a conical
      state emitted the photoelectron (4, 5).              pump field and the intense probe pulse that                   intersection occurs along a seam in 3N – 8 dimensions
          Wörner et al. (1) attacked this ambigu-          created the HHS.                                             where two surfaces with the same spatial and spin sym-
                                                                                                                        metry meet.
      ity by investigating the dynamical rearrange-            Wörner et al. (1) argued that this time               3. D. R. Yarkony, Rev. Mod. Phys. 68, 985 (1996).
      ment of the electrons directly as a molecule         dependence reflects the electronic dynamics                4. A. Stolow, J. G. Underwood, Adv. Chem. Phys. 139, 497
      moves through a conical intersection by mea-         that results when the bending-mode vibra-                    (2008).
                                                                                                                     5. T. Suzuki, Annu. Rev. Phys. Chem. 57, 555 (2006).
      suring the high-harmonic spectrum (HHS)              tional state created after electronic excita-             6. J. P. Marangos et al., Phys. Chem. Chem. Phys. 10, 35
      induced when an electronically excited mol-          tion moves along the excited-state surface                   (2008).
      ecule interacts with an intense laser field.         and reaches a geometry where it intersects                7. W. H. E. Schwarz, Angew. Chem. Int. Ed. 45, 1508
      High-harmonic generation, the production of          with the ground-state surface (see the fig-
                                                                                                                     8. It has recently been reported that in certain circum-
      radiation at many times the frequency of the         ure). Because the electronic configuration in                 stances high-harmonic spectroscopy on a ground-state
      driving ultrafast laser pulse, occurs in three       the electronically excited state is more easily              atom is sensitive to multielectron dynamics (9).
      stages. First, the intense electric field of the      ionized than the ground-state configuration,               9. A. D. Shiner et al., Nat. Phys. 7, 464 (2011).
      laser pulse removes an electron from an atom         a modulation was observed with a frequency
      in a process known as field ionization. The           characteristic of this vibrational frequency                                               10.1126/science.1212327
      free electron is then accelerated in the electric
      field of the laser. However, the electric field
      in a light pulse oscillates, so in half an optical
      cycle (1.3 femtoseconds for a pulse centered
      at 800 nm in the near infrared), the sign of
      the field reverses. The electron is accelerated
      back toward its source, where the resulting
                                                           Grass Trumps Trees with Fire
      collision releases energy and generates light        Audrey L. Mayer and Azad Henareh Khalyani
      at harmonics of the driving laser field.
          The relative intensity of the HHS of mol-
                                                           Feedbacks involving rainfall, fire, and vegetation govern transitions between forests, savannas,
      ecules is sensitive to their electronic struc-
                                                           and grasslands.
      ture, and this property has been used to
      reconstruct the electronic wave function of

      the highest occupied molecular orbital in                    cologists have long assumed that for-            that global climate change will be substan-
      some simple molecules (6). The interpreta-                   ests, savannas, and grasslands change            tially influenced by nonlinear behaviors and
      tion of this experiment has been controver-                  gradually over space and time, with              feedbacks between biophysical and human
      sial for two reasons. First, molecules other         tree cover responding linearly to gradients              systems.
      than H2+ have many electrons, and the con-           in precipitation, aridity, fire disturbance, and              Sudden transitions between forests,
      cept of a “one-electron” wave function is            grazing pressure. However, a growing body                savannas, and grasslands, and the feedbacks
      merely a theoretical construction to make            of evidence suggests that these biomes are               that drive them, have been observed at local
      electronic structure calculations easier to          self-reinforcing and that transitions between            to continental scales (6, 7). Sudden transi-
      handle in a computer. Second, a fundamen-            them can be nonlinear, governed by feed-                 tions at global scales were thought not to be
      tal tenet of quantum mechanics is that the           backs at local and regional scales (1–3). Two            observable, because local heterogeneity and
      wave function, which is usually described by         reports in this issue, by Staver et al. on page          small-scale transitions would appear as a
      complex numbers, is not in itself an observ-         230 (4) and by Hirota et al. on page 232 (5),            more gradual change at larger scales (8).
      able quantity (7). This debate was resolved          find evidence for these feedbacks and transi-                 Staver et al. used tree cover derived from
      when it was realized that the observed HHS           tions at the global scale. These results suggest         satellite data, together with global data for
      resulted from a coherent interaction between                                                                  annual and seasonal fire frequency and pre-
      the ionized photoelectron with the hole (the                                                                  cipitation, to examine whether forests and
                                                           School of Forest Resources and Environmental Science,
      empty level) it left behind, and so the experi-      Michigan Technological University, Houghton, MI 49931,   savannas are alternative stable states and, if
      ment on a ground state with a single config-          USA. E-mail:                             so, which mechanisms maintain these states.

188                                            14 OCTOBER 2011 VOL 334 SCIENCE

                                                                                                                                                                                                           Transitions between biomes. Staver
                                                                                                      Tree cover (%)                                                                                       et al. (4) and Hirota et al. (5) have
                                                              100                     60     50                                    5    1               0                                                  analyzed global data sets to iden-
                                                                                                                                                            Low                                            tify transitions between biomes. The
                                                                                                                                                                                                           results show that areas with regular
                                                                                                                                                                                                           rainfall above 2500 mm/year are for-
                                                                                                                                                                                                           est (>60% tree cover). Areas receiv-
                                                                                                                                                                                                           ing rainfall between 1000 and 2500
                                                           2500                                                                                                                                            mm/year can persist in either a for-
                                                                                                                                                                                                           est or savanna state, depending upon
                                                                                                                                                                                                           the strength of the fire-grass feedback
                                                                                                                                                                                                           ( 4, 5). Tree cover between 50 and

                                                                                                                                                                                    Rainfall seasonality
                                      Precipitation (mm)

                                                                                                                                                                                                           60% is an unstable state and demar-
                                                                                                                                                                                                           cates forest from savanna. Areas
                                                                                                                                                                                                           receiving highly seasonal rainfall
                                                                                                                                                                                                           between ~750 and 1500 mm/year

                                                                                                                                                            ~7-month dry season
                                                           1500                                                                                                                                            can persist as either savanna or grass-
                                                                                                                                                                                                           land, depending on fire frequency (5,
                                                                                                                                                                                                           9). The effects of grazing and human
                                                           1000           Forest
                                                                                                                                                                                                           intervention are not shown. The rain-
                                                                                                                                                                                                           fall axis is not linear.
                                                                            Unstable state               Savanna
                                                                                                                                                                     raphy may also influence micro-
                                                                                                                       Unstable state                                climates and thus fire spread and
                                                                                                                                                          High       vegetation (9). Finally, prehis-
                                                            200                                                                                                      toric and historic human activi-
                                                                    Low                                                                          High                ties had a sizable influence on the
                                                                                                      Fire frequency                                                 forests, savannas, and grasslands
                                                                                                                                                                     that exist today (13).
                                                                                                                                                                         Humans will continue to
                                     They excluded high-elevation areas, areas                            Adding these global results to previous       influence the distributions and resilience
                                     with high human impacts, and areas where                         studies at local and continental scales (3, 6,    of these ecosystems on multiple scales; fire
                                     most rainfall occurs during the nongrowing                       9–11) presents a strong argument that for-        suppression, grazing of domesticated ani-
                                     season. A forested state (more than 60% tree                     ests, savannas, and grasslands are alternative    mals, forest harvests, restoration efforts,
                                     cover) was predominant in areas receiving                        stable states at the global scale, maintained     and contributions to climate change all have
                                     more than 2500 mm/year rainfall and where                        by three main mechanisms: a strong feed-          effects (5, 14). Future studies should exam-
                                     rainfall was temporally uniform (see the fig-                     back between vegetation and precipitation;        ine the universality of the feedbacks driv-
                                     ure). Areas receiving rainfall between 1000                      a strong feedback between rainfall season-        ing biome transitions in both hemispheres,
                                     and 2500 mm/year were either forest or                           ality and grass; and a very strong feedback       as well as the impacts of human activities on
                                     savanna, but few areas supported tree cover                      between grass and fire (see the figure).            these feedbacks, to assist the development
                                     between 50 and 60%. Savannas existed                                 Both reports identify an unstable state at    of better-informed management and resto-
                                     where rainfall was highly seasonal and the                       50 to 60% tree cover; either trees take hold      ration plans.
                                     fire-grass feedback was strong. In Australia,                     and promote their own growth hydrologically
                                     extreme rainfall seasonality precluded for-                      (and suppress fire), or grasses take hold and                                References
                                                                                                                                                         1. J. A. Foley, M. T. Coe, M. Scheffer, G. Wang, Ecosystems
                                     ests even in areas receiving up to 2500 mm/                      promote their expansion through fire. This             6, 524 (2003).
                                     year of rain.                                                    work has implications for the resilience of        2. M. Rietkerk, S. C. Dekker, P. C. de Ruiter, J. van de Kop-
                                        Hirota et al. used the same tree-cover data                   these biomes in the Southern Hemisphere.              pel, Science 305, 1926 (2004).
                                     as Staver et al., but with different precipita-                  Most notably, large areas of savanna in Africa     3. D. J. Beerling, C. P. Osborne, Glob. Change Biol. 12,
                                                                                                                                                            2023 (2006).
                                     tion data. They only excluded parts of the                       could shift to forest (if fire and grazing are      4. A. C. Staver, S. Archibald, S. A. Levin, Science 334, 230
                                     Saharan and Australian deserts. By relating                      suppressed), and large areas of forest in South       (2011).
                                     tree cover to mean annual precipitation, the                     America could convert to savanna [although         5. M. Hirota, M. Holmgren, E. H. Van Nes, M. Scheffer,
                                                                                                                                                            Science 334, 232 (2011).
                                     authors identified three normal distributions                     degraded forest does not necessarily transi-       6. M. Rietkerk et al., Ecol. Complex. 8, 223 (2011).
                                     that were best explained if forests, savannas,                   tion to savanna (11)], as climate change and       7. P. deMenocal et al., Quat. Sci. Rev. 19, 347 (2000).
                                     and treeless models were included. Thresh-                       local human impacts such as logging interact       8. M. Scheffer, Critical Transitions in Nature and Society
                                                                                                                                                            (Princeton Univ. Press, Princeton, NJ, 2009), pp. 86 and 95.
                                     olds were sharp between these states, with                       with rainfall seasonality and fire.

                                                                                                                                                         9. R. C. Anderson, J. Torrey Bot. Soc. 133, 626 (2006).
                                     forests (>60% cover) found mainly above                              The two reports show convincingly that        10. S. Kröpelin et al., Science 320, 765 (2008).
                                     2500 mm/year rainfall, savannas (5 to 50%                        forests, savannas, and grasslands are distrib-    11. J. Ratnam et al., Glob. Ecol. Biogeogr. 20, 653 (2011).
                                     tree cover) most common between 750 and                          uted discontinuously at the global scale, but     12. C. M. Janis, J. Damuth, J. M. Theodor, Palaeogeogr.
                                                                                                                                                            Palaeoclimatol. Palaeoecol. 177, 183 (2002).
                                     2000 mm/year rainfall, and treeless areas                        the authors do not analyze several impor-         13. K. J. Willis, H. J. B. Birks, Science 314, 1261 (2006).
                                     dominating below 750 mm/year rainfall. The                       tant mechanisms. Large herbivores (such as        14. K. N. Suding, R. J. Hobbs, Trends Ecol. Evol. 24, 271
                                     extreme rarity of areas with intermediate tree                   horses and antelope) evolved in concert with          (2009).
                                     cover suggests that transitions between these                    savannas and grasslands (12), and their feed-
                                     biomes are rapid.                                                backs with grasses are known (3, 4, 9). Topog-                                                                       10.1126/science.1213908

                                                                                    SCIENCE VOL 334 14 OCTOBER 2011                                                                                                      189
                                                                                                                               considered the quintessence of this “nut-cracking”
                                                                                                                               morphology (4) (Fig. 1).
      The Diets of Early Hominins                                                                                                  The earliest members of our own genus are
                                                                                                                               believed to have had tools to acquire and process
                                                                                                                               a broad range of foods, such as meat and under-
      Peter S. Ungar1* and Matt Sponheimer2*
                                                                                                                               ground storage organs, bespeaking a generalized
      Diet changes are considered key events in human evolution. Most studies of early hominin                                 and versatile diet (5). Morphological evidence sug-
      diets focused on tooth size, shape, and craniomandibular morphology, as well as stone tools                              gests that early Homo had smaller cheek teeth,
      and butchered animal bones. However, in recent years, dental microwear and stable isotope                                thinner dental enamel, and greater occlusal relief
      analyses have hinted at unexpected diversity and complexity in early hominin diets. Some                                 than did their Australopithecus predecessors or
      traditional ideas have held; others, such as an increasing reliance on hard-object feeding and                           their Paranthropus contemporaries (6–8) (Fig. 1).
      a dichotomy between Australopithecus and Paranthropus, have been challenged. The first known                             This may indicate changing selective pressures
      evidence of C4 plant (tropical grasses and sedges) and hard-object (e.g., seeds and nuts)                                due to extraoral food processing with tools, but
      consumption dates to millions of years after the appearance of the earliest probable hominins, and                       also suggests that early Homo teeth could more
      there are no consistent trends in diet change among these species through time.                                          efficiently shear tough foods (such as leaves and
                                                                                                                               meat) than could those of the australopiths. The
                iet is fundamental to an organism’s ecol-             mandibles relative to living apes) reflects an           possession of larger brains in some cases has also

      D         ogy and, unsurprisingly, changes in diet
                have been hailed as key milestones in
      human evolution. Our understanding of the diets
                                                                      adaptive shift from diets dominated by soft, sug-
                                                                      ary forest fruits to hard, brittle nuts or seeds, or
                                                                      to those with adherent abrasives, such as under-
                                                                                                                               been used to argue that Homo required high-
                                                                                                                               energy–yielding foods (9, 10).
                                                                                                                                   Much research on early hominin diets has fo-
      of our distant forebears, the early hominins, has               ground storage organs that were readily available        cused on archaeological and morphological da-
      been honed in recent decades as a result of new                 in increasingly open Plio-Pleistocene landscapes         ta, but like all lines of evidence for subsistence of
      methods for dietary inference, the discovery of                 (2, 3) (Fig. 1). The larger teeth with well-buttressed   fossil species, they have limitations. Stone tools
      new fossil species and additional specimens, and                skulls and massive chewing muscles of Paran-             and butchered bones tell us little about the plant
      improved reconstructions of the environments in                 thropus have led to the notion that “robust”             foods that likely dominated early hominin diets.
      which they evolved. Here, we focus on recent                    australopiths relied more heavily on hard foods          Moreover, although the earliest known stone tools
      contributions from dental microwear and stable                  than did Australopithecus. The eastern African           and cut-marked bones date to at least 2.6 Ma and
      isotope analyses, two approaches that have chal-                “hyper-robust” Paranthropus boisei has been              possibly earlier (11, 12), they still postdate the earliest
      lenged traditional thinking about early hominin
      dietary ecology over the past few years.
          There are four principal groups of interest
      in early hominin evolution: the Mio-Pliocene
      probable hominins (Sahelanthropus, Orrorin,
      Ardipithecus); the Plio-Pleistocene “gracile” aus-
      tralopiths (Australopithecus); the “robust” austra-
      lopiths (Paranthropus); and the earliest members
      of our own genus, Homo. The first group dates
      from about 7 million years ago (Ma), although
      the best-known species, Ardipithecus ramidus,
      lived about 4.4 Ma. The earliest recovered Aus-
      tralopithecus dates to approximately 4.2 Ma,
      whereas Paranthropus and Homo have their first
      known appearances shortly before and after
      2.5 Ma, respectively, presumably from Australo-
      pithecus or Australopithecus-like ancestors. All
      of these groups are represented in eastern Africa,
      the first two are also known from Chad, and the
      latter three are found in South Africa (1).
      What We Thought About Early Hominin Diets
      The earliest probable hominins are not all well
      known, but in some cases their molars are smaller
      and more thinly enameled than those of later
      australopiths and more like those of extant chim-
      panzees (1), suggesting a diet of fleshy fruits
      and soft, young leaves. According to conven-
      tional wisdom, the craniodental morphology of
      later Australopithecus (e.g., thickly enameled,
      large, flat cheek teeth; heavily built crania and               Fig. 1. Composite skulls (left) and specimen sketches (right) of the crania, maxillae, and mandibles of
       Department of Anthropology, University of Arkansas, Fayette-   Australopithecus africanus (Sts 5, Sts 52a, and Sts 52b), Paranthropus boisei (KNM‑ER 406, OH 5, Peninj),
      ville, AR 72701, USA. 2Department of Anthropology, University   and Homo habilis (OH 24, KNM-ER 1813, OH 13). Differences in craniodental size and shape underscore
      of Colorado, Boulder, CO 80309, USA.                            the importance of diet for understanding hominin diversity and evolution. [Composites and specimen
      *To whom correspondence should be addressed. E-mail:            sketches are modified and reproduced from (50) and (51) with permission from the publishers and (P.S.U.); (M.S.)          authors (©1988 Academic Press and ©1991 Waveland Press)]

190                                                  14 OCTOBER 2011              VOL 334        SCIENCE

      Food            Tooth movement                  Microwear texture

  Hard, brittle             Crushing                    Complex, isotropic
                                                                                                 P. robustus

                                                                                                    P. boisei

                                                                                                   H. habilis

                                                                                                  H. erectus

                                                                                              Au. anamensis

                                                                                               Au. africanus

                                                                                               Au. afarensis

                                                                                                                0         1        2      3           4         5        6
   Soft, tough              Shearing                     Simple, isotropic

Fig. 2. Microwear textures of early hominins. Left: A model for microwear for-            foods are sheared between opposing teeth that slide past one another, causing
mation, wherein hard and brittle foods are crushed between opposing teeth,                parallel scratches and simpler, anisotropic surfaces (18). Right: Microwear texture
causing pitting with complex, isotropic surface textures; in contrast, soft and tough     complexity values for individual fossil hominins by species [data from (21–24)].

hominins by millions of years. And although tooth              consume a more limited variety of foods. Several           73 specimens of Australopithecus (Au. anamensis,
and jaw size, shape, and structure offer important             integrated metrics of microwear have proven                Au. afarensis, Au. africanus), Paranthropus
clues to the fracture properties of foods and as-              useful: Surface fractal complexity, or change in ap-       (P. boisei, P. robustus), and early Homo (H. habilis,
sociated masticatory stresses and strains to which             parent roughness with scale of observation, is used        African H. erectus, Homo specimens from
a species is adapted (13–16), they indicate what               as a proxy for food hardness; anisotropy, or direc-        Sterkfontein Member 5 and Swartkrans Mem-
early hominins were capable of eating and sug-                 tionality of the wear fabric, is used as a proxy for       ber 1) (21–24) (Fig. 2, Fig. 3, and table S1). None
gest the selective pressures faced by their ances-             food toughness. High complexity and anisotropy             of the Australopithecus specimens have the high
tors, rather than what specific individuals ate. For           values correspond roughly to surfaces with heavy           complexity values or heavily pitted surfaces of a
direct evidence of the diets of fossil specimens               pitting and highly aligned scratches, respectively.        hard-object feeder, as originally expected given
recovered, we need other sources of information,                    Dental microwear texture data have not yet            their morphology. And the two eastern African
such as dental microwear and stable light isotope              been collected for the earliest probable hominins,         species, Au. anamensis and Au. afarensis, have
analyses of teeth.                                             but results have been published for cheek teeth of         similar and homogenous microwear complexity

Dental Microwear
Mammals show a strong and consist-              A                                                                     B
ent association between dental mi-                   1.90 μm                                                        2.24 μm
crowear pattern and food fracture
properties. Those that crush hard, brit-
tle foods (e.g., nuts, bones) typically
have occlusal microwear dominated
by pits, whereas those that shear tough
items (e.g., leaves, meat) more often           C                                                                     D
show long, parallel striations on their              2.91 μm                                                        9.17 μm
wear surfaces (17, 18). These pits and
scratches are traces of actual chewing
events; they record activities during a
moment in the life of an individual,
much like footprints. And like foot-
prints, microwear is fleeting; individ-
                                                 E                                                                    F
ual features turn over and are replaced
by others as a tooth wears down (19).                3.38 μm                                                        7.71 μm
Indeed, microwear textures reflect diet
in the days or weeks before death.
This “last-supper” phenomenon (20)
can be an asset, as large samples pro-
vide a sense of variation in diet within
a species. Taxa with more catholic           Fig. 3. (A to F) Photosimulations of microwear surfaces representing (A) Au. afarensis (AL 333w-1a), (B) Au. africanus
diets should evince a broader range of       (Sts 61), (C) P. boisei (KNM-CH1B), (D) P. robustus (SK 16), (E) H. habilis (OH 16), and (F) H. erectus (KNM-ER 807).
microwear textures than those that           Each represents an area of 102 mm by 139 mm on facet 9; vertical scales are as indicated.

                                                SCIENCE        VOL 334       14 OCTOBER 2011                                                          191
      (within and between taxa) despite a                                                                                         foods consumed by their prey (28)
      sample distribution spanning more                    C3 plant (Ficus)                    C4 plant (Panicum)                 (Fig. 4). This allows a variety of
      than 1 million years and 1500 km                                                                                            questions to be addressed. For
      and habitats as different as closed                                                                                         instance, did the early members of
      woodland and grassland (25). The                                                                                            our lineage have diets similar to
      South African Au. africanus has high-                                                                                       those of our closest living kin, the
      er average complexity, but this is still                                                                                    chimpanzee (Pan troglodytes)? We
      lower than that expected of a hard-                                                                                         know that chimpanzees have diets
      object feeder. Australopithecus spp.                                                                                        dominated by C3 tree foods (es-
      also have low to moderate anisot-                                                                                           pecially fleshy fruits), whether in
      ropy, with few values extending into                                                                                        their preferred forest habitats or in
      the upper ranges of living folivorous                                                                                       fairly open savannas with abundant
      primates. This indicates that they did                                                                                      grasses (32, 33). Carbon isotope ra-
      not shear tough leaves as do modern                                                                                         tios (13C/12C) have been analyzed
      folivores, perhaps because such foods                                                                                       for more than 75 hominin specimens
      were not an important part of their                                                                                         from sites in Ethiopia, Kenya, Tanza-
      diet. However, it is also possible that                                                                                     nia, and South Africa, ranging in age
      they ground tough foods in the man-                                                                                         from about 4.4 to possibly 0.8 Ma
      ner of a mortar and pestle, as their                                                                                        (Fig. 4 and table S1). The broad view
      flat teeth might have posed fewer                  1.0                                                                      of these data is that early hominins
      masticatory constraints than those                                                                                          did not have diets like those of ex-
      of modern folivores (23).                          0.9                                                                      tant African apes, but this conclusion
          The eastern African “robust”                   0.8                                                                      belies the complexity of the varied
      australopith, P. boisei, has low mi-                                                                                        results. For instance, the earliest tax-
      crowear texture complexity and low                 0.7                                                                      on analyzed to date, Ar. ramidus, had
      to moderate anisotropy values, sug-                                                                                         in aggregate a C3 diet much like that
                                                     Fraction of data

      gesting a diet dominated by foods                                                                                           of savanna chimpanzees (34). Other
      with fracture properties similar to                0.5                                                                      taxa, such as Au. africanus, P. robustus,
      those eaten by Au. anamensis and                                                                                            and early Homo, were more middling,
      Au. afarensis (22, 23). The South                                                                        Ar. ramidus
                                                                                                                                  as they ate more than 50% C3 foods
      African P. robustus, on the other hand,            0.3                                                   Au. africanus      but also consumed substantial quan-
      has the highest average complexity                                                                       Early Homo         tities of C4 foods (33, 35–38) that
      and lowest anisotropy of any early                 0.2                                                   P. robustus        became increasingly available in the
      hominin (21). Complexity in South                                                                        P. boisei          Plio-Pleistocene (39, 40). In marked
      African “robust” australopiths also                                                                                         contrast, P. boisei had a diet of about
      shows high variance, with a distribu-              0.0                                                                      75 to 80% C4 plants, unlike that of
                                                           -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1
      tion most comparable to hard-object                         Carbon isotope composition/δ13C (% )                            any other fossil hominin but similar
      fallback feeders such as gray-cheeked                                                                                       to that of grass-eating warthogs, hip-
      mangabeys (Lophocebus albigena) Fig. 4. Carbon isotope compositions (13C/12C) of early hominins. Top: Carbon flows pos, and zebras (18, 37, 41). Carbon
      and brown capuchins (Cebus apella), from C3 and C4 plants (blue and pink arrows, respectively) into the tooth enamel of isotopic variability between these taxa
      which tend to “fall back” on harder the consumer (in this case P. robustus, SK 1), and its resulting carbon isotope is also marked, with Au. africanus
      items when softer, more preferred composition reveals the proportions of these plant types consumed. Bottom: ranging from pure C3 to nearly pure
      foods are unavailable (21). This could Quantile plot with carbon isotope ratio data for all early hominins analyzed to date C4 diets, whereas other taxa such as
      indicate that their anatomy evolved [data from (34–38, 49)]. Darker shading indicates a greater degree of C3 plant P. boisei have much reduced ranges.
      to cope with infrequently eaten but consumption. Each data point reflects a hominin’s diet for a period ranging from             Thus, carbon isotopes suggest
      fracture-resistant foods [see (15)].     months to years depending on the sampling procedure used (red rectangles marked dietary diversity within the
                                                                                                           13 12
          Microwear textures of early Homo represent hypothetical sampling areas). Carbon isotope ratios ( C/ C) are expressed hominins. This is not surprising, given
      suggest that all species had fairly as d values in parts per thousand relative to the PeeDee Belemnite standard.            their temporal and ecogeographic
      generalized diets lacking special-                                                                                          ranges and the variation in mastica-
      ization for either extremely hard or tough foods tissues then acquire an isotopic composition related tory morphology they manifest; however, the iso-
      (24). Of note, H. erectus has substantially more to that of the source food that can reveal much about tope data also suggest enormous and unanticipated
      variation in microwear complexity values than paleodiets (28–30). Although dietary studies can differences between contemporaneous taxa with
      H. habilis, or indeed than that of any other be undertaken with bone mineral in some cases, strong morphological similarities, notably the
      hominin examined to date except P. robustus. dental enamel is preferred as it is more highly “robust” australopiths P. robustus and P. boisei.
      This suggests that H. erectus had a comparatively mineralized and thus less susceptible to postdepo- Despite their attribution to the same genus, there is
      broad-based diet, spanning a range of fracture sitional chemical alteration than bone (31); it is also no overlap in their carbon isotope compositions
      properties including some hard and perhaps an incremental tissue that may allow investiga- (41), which is a rarity for congeners among extant
      tough foods, which may also produce small pits tion of intra-individual diet change through time. mammals. All told, the early hominins analyzed to
      through adhesive wear (26).                                 Carbon isotopes are particularly useful for date fall roughly into three groups: (i) those with
                                                              hominin paleodietary studies because they tell carbon isotope compositions indicating strong C3
      Stable Isotopes                                         us about the relative proportions of plants using diets similar to those of savanna chimpanzees, (ii)
      Stable isotope analysis of ancient tissues is based the C3 (trees, bushes, forbs) and C4 (tropical those with variably mixed C3/C4 diets, and (iii)
      on the principle “you are what you eat” (27). Stable grasses and some sedges) photosynthetic path- those with carbon isotope compositions indicat-
      isotopes in foodstuffs become incorporated into the ways that were consumed by herbivores, or, in ing diets of chiefly C4 vegetation, as is typically
      growing teeth and bones of consumers. These the case of faunivores, the proportions of these seen for grass-eating ungulates in tropical climes.

192                                           14 OCTOBER 2011             VOL 334       SCIENCE
Integrating and Moving Forward                              tory of teeth, also reveals considerable variation               16. P. W. Lucas, Dental Functional Morphology: How Teeth
Microwear and stable carbon isotope studies have            in within-tooth carbon isotope compositions of                       Work (Cambridge Univ. Press, New York, 2004).
                                                                                                                             17. P. S. Ungar, Mammal Teeth: Origin, Evolution, and
challenged long-held assumptions about early hom-           P. robustus at inter- and intra-annual time scales                   Diversity ( Johns Hopkins Univ. Press, Baltimore, MD, 2010).
inin diets. The simple textbook model—in which              (44). In contrast, the teeth of Au. afarensis show               18. See supporting material on Science Online.
hominin craniodental functional morphology                  little variance in microwear texture complexity                  19. M. F. Teaford, O. J. Oyen, Am. J. Phys. Anthropol. 75,
evolved for increasing consumption of hard, brit-           despite a range of samples across time and space.                    279 (1988).
                                                                                                                             20. F. E. Grine, J. Hum. Evol. 15, 783 (1986).
tle foods as savannas spread—is incorrect, or at least      In this case, a model involving increased foraging               21. R. S. Scott et al., Nature 436, 693 (2005).
too simplistic. None of the Australopithecus or even        ranges for foods with given fracture or nutritional              22. P. S. Ungar, F. E. Grine, M. F. Teaford, PLoS ONE 3,
Paranthropus specimens examined from eastern                properties, such as observed for some chimpan-                       e2044 (2008).
Africa show microwear patterns of a hard-object             zees (45), might be more appropriate.                            23. P. S. Ungar, R. S. Scott, F. E. Grine, M. F. Teaford,
                                                                                                                                 Philos. Trans. R. Soc. London Ser. B 365, 3345 (2010).
feeder (18). And whereas the Ardipithecus carbon                 The above evidence challenges certain as-                   24. P. S. Ungar, K. L. Krueger, R. J. Blumenschine, J. Njau,
isotope composition is consistent with a diet sim-          pects of our understanding of hominin biology,                       R. S. Scott, J. Hum. Evol. 10.1016/j.jhevol.2011.04.006 (2011).
ilar to those of savanna chimpanzees (as might be           biogeography, and evolution. For instance, if                    25. F. E. Grine, P. S. Ungar, M. F. Teaford, S. Afr. J. Sci.
expected), that of P. boisei indicates that C4 plant        P. boisei was a C4 sedge consumer (37), its dis-                     102, 301 (2006).
                                                                                                                             26. M. F. Teaford, J. A. Runestad, Am. J. Phys. Anthropol.
foods such as grasses or sedges provided the vast           tribution was likely limited to the periphery of
                                                                                                                                 88, 347 (1992).
majority of dietary energy for this taxon. This was         permanent sources of water. Its eventual extinc-                 27. M. J. Deniro, S. Epstein, Geochim. Cosmochim. Acta
almost completely unanticipated [but see (42)]              tion might then be linked to the difficulty of                       42, 495 (1978).
and raises the intriguing possibility that earlier east-    dispersing away from water sources despite the                   28. J. A. Lee-Thorp, J. C. Sealy, N. J. van der Merwe,
ern African australopiths may have had a similar            vaunted energetic efficiency of bipedalism (46).                     J. Archaeol. Sci. 16, 585 (1989).
                                                                                                                             29. P. L. Koch, K. A. Hoppe, S. D. Webb, Chem. Geol.
penchant for C4 foods, especially given the sim-            On the other hand, if P. boisei was a C4 grass                       152, 119 (1998).
ilarities of their dental microwear to that of P. boisei.   consumer, it might have thrived in the emerging                  30. T. E. Cerling, J. M. Harris, M. G. Leakey, Oecologia
     Both the microwear and carbon isotope data             savannas of the Pleistocene, demanding an expla-                     120, 364 (1999).
offer other surprises. First, there seems to be a           nation other than habitat change for its extinction.             31. J. A. Lee-Thorp, N. J. van der Merwe, J. Archaeol. Sci.
                                                                                                                                 18, 343 (1991).
geographic influence on australopith diets; the                  Our understanding of the paleoecology of                    32. M. J. Schoeninger, J. Moore, J. M. Sept, Am. J. Primatol.
microwear texture complexity of eastern African             these organisms is in flux, and a great deal of                      49, 297 (1999).
Australopithecus and Paranthropus is lower than             directed, integrative research remains to be done                33. M. Sponheimer et al., J. Hum. Evol. 51, 128 (2006).
that of their South African congeners. Likewise,            [e.g., (47)]. Microwear and stable carbon isotope                34. T. D. White et al., Science 326, 67 (2009).
                                                                                                                             35. J. Lee-Thorp, J. F. Thackeray, N. van der Merwe,
P. boisei and P. robustus have different carbon             analyses are needed for all relevant species, and                    J. Hum. Evol. 39, 565 (2000).
isotope compositions, with the South African “ro-           these results must be integrated with data on                    36. N. J. van der Merwe, J. F. Thackeray, J. A. Lee-Thorp,
bust” australopiths consuming a much higher frac-           masticatory biomechanics, plant distributions and                    J. Luyt, J. Hum. Evol. 44, 581 (2003).
tion of C3 foods, like most other early hominins            nutritional/mechanical properties, and primate ecol-             37. N. J. van der Merwe, F. T. Masao, M. K. Bamford, S. Afr.
                                                                                                                                 J. Sci. 104, 153 (2008).
(although not to the extent seen in Ardipithecus).          ogy and digestive physiology. An important role
                                                                                                                             38. J. A. Lee-Thorp, N. J. van der Merwe, C. K. Brain,
This might indicate that a specialized morphologi-          for microwear and isotope analyses within con-                       J. Hum. Evol. 27, 361 (1994).
cal complex can serve more than one function and            temporary paleodietary research is to focus on                   39. P. B. deMenocal, Earth Planet. Sci. Lett. 220, 3 (2004).
reflect more than one type of diet; perhaps “robust”        underlying processes rather than outcomes, as                    40. R. Bobe, J. Arid Environ. 66, 564 (2006).
                                                            well as to recognize evolutionary novelties, such                41. T. E. Cerling et al., Proc. Natl. Acad. Sci. U.S.A. 108,
morphology functioned in high-stress hard-object
                                                                                                                                 9337 (2011).
feeding for P. robustus but in repetitive loading           as grazing giraffes (48) and grass- or sedge-eating              42. C. J. Jolly, Man 5, 5 (1970).
during grinding of tough foods for P. boisei.               apes (37). When these behavioral proxies are                     43. P. S. Ungar, in Comparative Dental Functional Morphology,
     The apparent continuity of microwear pattern           linked to morphological and paleoenvironmental                       T. Koppe, G. Meyer, K. W. Alt, Eds. (Karger, Basel, 2009),
through the putative lineage Au. anamensis–Au.              data sets through time, yoking habitat and dietary                   pp. 38–43.
                                                                                                                             44. M. Sponheimer et al., Science 314, 980 (2006).
afarensis–P. boisei could even suggest that mor-            change to morphological response, our under-                     45. W. C. McGrew, P. J. Baldwin, C. E. G. Tutin, J. Hum. Evol.
phological changes reflect increasing efficiency            standing of the patterns and processes of hominin                    10, 227 (1981).
for grinding large quantities of tough food. Al-            evolution will be greatly augmented.                             46. P. S. Rodman, H. M. McHenry, Am. J. Phys. Anthropol.
though living primates that eat tough items typ-                                                                                 52, 103 (1980).
                                                                                                                             47. N. Dominy, E. R. Vogel, J. D. Yeakel, P. Constantino,
ically have sharp shearing crests, eastern African              References and Notes                                             P. W. Lucas, Evol. Biol. 35, 159 (2008).
australopiths and especially P. boisei may have              1. B. Wood, N. Lonergan, J. Anat. 212, 354 (2008).              48. N. Solounias, M. Teaford, A. Walker, Paleobiology 14,
evolved a different solution for processing such             2. M. F. Teaford, P. S. Ungar, Proc. Natl. Acad. Sci. U.S.A.        287 (1988).
foods, given the flattened, thickly enameled teeth              97, 13506 (2000).                                            49. M. Sponheimer et al., J. Hum. Evol. 48, 301 (2005).
                                                             3. G. Suwa et al., Science 326, 94 (2009).                      50. J. G. Fleagle, Primate Adaptation and Evolution
of their close ancestors (23). Natural selection must                                                                            (Academic Press, New York, 1988).
                                                             4. B. Wood, P. Constantino, Am. J. Phys. Anthropol. 50
work with the raw materials available to it. Thus,              (suppl. 45), 106 (2007).                                     51. C. S. Larsen, R. M. Matter, D. L. Gebo, Human Origins:
the present-day ecomorphological diversity within            5. P. S. Ungar, F. E. Grine, M. F. Teaford, Annu. Rev.              The Fossil Record (Waveland, Prospect Hills, IL, 1991).
the primates may not be sufficient for making some              Anthropol. 35, 209 (2006).                                   Acknowledgments: Supported by NSF, the Leakey
                                                             6. A. D. Beynon, B. A. Wood, Am. J. Phys. Anthropol. 70,            Foundation, and the Wenner-Gren Foundation.
paleoecological inferences, which is not surprising
                                                                177 (1986).                                                      We thank C. Campbell, J. Leichliter, O. Paine,
given that the vast majority of all primates, espe-          7. P. S. Ungar, J. Hum. Evol. 46, 605 (2004).                       Y. Rahman, C. Ross, and P. Sandberg for comments;
cially apes, that have ever lived are now extinct.           8. H. M. McHenry, K. Coffing, Annu. Rev. Anthropol. 29,             T. Cerling, D. Codron, D. de Ruiter, F. Grine, J. Lee-Thorp,
     The microwear and isotope evidence also                    125 (2000).                                                      R. Scott, M. Teaford, and N. van der Merwe for discussions
gives insight into food choices and foraging strat-          9. L. C. Aiello, P. Wheeler, Curr. Anthropol. 36, 199 (1995).       over the years; and D. de Ruiter for the image of SK 1.
                                                            10. K. Milton, Evol. Anthropol. 8, 11 (1999).                        All raw data referred to in this paper are published in
egies. The P. robustus microwear complexity                                                                                      (21–24, 34–38, 49).
                                                            11. S. Semaw, J. Archaeol. Sci. 27, 1197 (2000).
distribution suggests that individuals ate hard ob-         12. S. P. McPherron et al., Nature 466, 857 (2010).
jects only on occasion, perhaps in a manner akin            13. D. J. Daegling, F. E. Grine, Am. J. Phys. Anthropol. 86,
                                                                                                                             Supporting Online Material
to the lowland gorilla’s (Gorilla gorilla) falling              321 (1991).
back on lower-quality, tough foods during times             14. W. L. Hylander, in Evolutionary History of the “Robust”
                                                                Australopithecines, F. E. Grine, Ed. (Aldine de Gruyter,     SOM Text
when preferred soft, sugar-rich items are unavail-              New York, 1988), pp. 55–83.                                  Table S1
able (43). Laser ablation analysis, which allows            15. D. S. Strait et al., Proc. Natl. Acad. Sci. U.S.A. 106,      References
isotopic sampling along the rough growth trajec-                2124 (2009).                                                 10.1126/science.1207701

                                                SCIENCE           VOL 334          14 OCTOBER 2011                                                                     193
                                                                                                                         chromosomes, with little staining of 5hmC anti-
                                                                                                                         bodies (fig. S2A). This result is consistent with
      Replication-Dependent Loss of                                                                                      previous studies demonstrating relative enrich-
                                                                                                                         ment of 5hmC in the sperm-derived chromosomes

      5-Hydroxymethylcytosine in Mouse                                                                                   (5, 6). Similar staining of mitotic chromosomes at
                                                                                                                         the two-cell stage revealed that only one of the two
                                                                                                                         sister chromatids of sperm-derived chromosomes
      Preimplantation Embryos                                                                                            is enriched for 5hmC (Fig. 1B), indicating that the
                                                                                                                         5hmC mark is not maintained during DNA rep-
      Azusa Inoue1,2 and Yi Zhang1,2*                                                                                    lication. Further analysis of the four-cell– and eight-
                                                                                                                         cell–stage embryo blastomeres revealed that the
                lthough DNA methylation is a relative-         possibilities, including base excision repair (BER),      chromosomes containing 5hmC are gradually re-

      A         ly stable epigenetic modification, global
                DNA demethylation has been observed
      in two stages during embryogenesis. One takes
                                                               further oxidation, and replication-dependent di-
                                                               lution, have been proposed for the fate of 5hmC
                                                               (2, 4, 7). Indeed, two recent studies have dem-
                                                                                                                         duced (Fig. 1, C and D).
                                                                                                                             We note that, at the four- and eight-cell stages,
                                                                                                                         5hmC appears to be present only in part of the chro-
      place in zygotes, when the paternal genome is            onstrated that Tet proteins can oxidize 5mC and           matids, likely because of sister chromatid exchange
      preferentially demethylated (1). Despite great ef-       5hmC further to 5-formylcytosine (5fC) and                (Fig. 1, C and D; note the enlarged images on right
      forts in identifying the responsible enzymes, the        5-carboxylcytosine (5caC) (8, 9) and that 5caC can        of the panels). Although the exact numbers of
      identity of the putative DNA demethylase has re-         be excised by thymine-DNA glycosylase (TDG)–              5hmC-positive chromatids vary among individual
      mained elusive (2). Recent demonstration that the        mediated BER (8).                                         blastomeres, the total length of the 5hmC-enriched
      Tet (ten eleven translocation) family proteins are           To distinguish between these possibilities, we        part in chromatids in each blastomere is very close
      capable of converting 5-methylcytosine (5mC) of          prepared mitotic chromosome spreads at various            to the theoretical number of 5hmC-positive chro-
      DNA to 5-hydroxymethylcytosine (5hmC) raises             stages of preimplantation embryos. Co-staining            matids expected from the original 5hmC-containing
      the possibility that Tet proteins might be involved      of the chromosome spreads with antibodies against         chromatids present in the zygote (table S1).
      in this process (3, 4).                                  5mC or 5hmC revealed that, at the one-cell stage,             The above results, together with previous
          To address this possibility, we performed im-        the two sets of chromosomes are compartmental-            studies (5, 6), support a model by which Tet3 con-
      munostaining that demonstrated, similar to recently      ized, with 5hmC specifically enriched in the sperm-       verts 5mC to 5hmC in the male pronucleus in
      published results (5, 6), that loss of 5mC staining in   derived chromosomes and 5mC specifically enriched         zygotes followed by replication-dependent dilution
      the male pronucleus correlates with the appearance       in the egg-derived chromosomes (Fig. 1A). Similar         during preimplantation development (fig. S3).
      of 5hmC (fig. S1), indicating the conversion of          staining with parthenogenetic one-cell embryos
      5mC to 5hmC at the male pronucleus. Several              revealed uniform 5mC staining in the two sets of               References and Notes
                                                                                                                          1. W. Mayer, A. Niveleau, J. Walter, R. Fundele, T. Haaf,
                                                                                                                             Nature 403, 501 (2000).
                                                                                                                          2. S. C. Wu, Y. Zhang, Nat. Rev. Mol. Cell Biol. 11, 607
                                                                                                                          3. S. Ito et al., Nature 466, 1129 (2010).
                                                                                                                          4. M. Tahiliani et al., Science 324, 930 (2009); 10.1126/
                                                                                                                          5. K. Iqbal, S. G. Jin, G. P. Pfeifer, P. E. Szabó, Proc. Natl.
                                                                                                                             Acad. Sci. U.S.A. 108, 3642 (2011).
                                                                                                                          6. M. Wossidlo et al., Nat. Commun. 2, 241 (2011).
                                                                                                                          7. J. U. Guo, Y. Su, C. Zhong, G. L. Ming, H. Song, Cell 145,
                                                                                                                             423 (2011).
                                                                                                                          8. Y.-F. He et al., Science 333, 1303 (2011); 10.1126/
                                                                                                                          9. S. Ito et al., Science 333, 1300 (2011); 10.1126/
                                                                                                                         Acknowledgments: We thank S. Wu for critical reading of the
                                                                                                                             manuscript. This work was supported by HHMI and NIH
                                                                                                                             (GM68804). Y.Z. is an Investigator of HHMI.

                                                                                                                         Supporting Online Material
                                                                                                                         Materials and Methods
                                                                                                                         SOM Text
                                                                                                                         Figs. S1 to S3
                                                                                                                         Table S1

                                                                                                                         10 August 2011; accepted 6 September 2011
                                                                                                                         Published online 22 September 2011;

      Fig. 1. Replication-dependent loss of 5-hmC during mouse preimplantation development. Representative                Howard Hughes Medical Institute (HHMI), Lineberger Compre-
                                                                                                                         hensive Cancer Center, University of North Carolina at Chapel
      confocal microscopy images of mitotic chromosome spreads derived from in vitro fertilization mouse
                                                                                                                         Hill, Chapel Hill, NC 27599, USA. 2Department of Biochemistry
      embryos co-stained with 5hmC (red) or 5mC (green) antibodies or DAPI (4´,6´-diamidino-2-phenylindole,              and Biophysics, Lineberger Comprehensive Cancer Center, Uni-
      blue) at one-cell (A), two-cell (B), four-cell (C), and eight-cell (D) stages. Shown are the images of             versity of North Carolina at Chapel Hill, Chapel Hill, NC 27599,
      chromosomes from one blastomere at each developmental stage. As a result of sister chromatid exchange,             USA.
      only part of the sister chromatids stained positively for 5hmC at the four-cell and eight-cell stages (C and D).   *To whom correspondence should be addressed. E-mail:
      The dotted squares represent enlarged paternal (Pat.) or maternal (Mat.) chromosomes.                    

194                                            14 OCTOBER 2011              VOL 334        SCIENCE
                                                                                                RESEARCH ARTICLE
                                                                                                             information technologies. Electronic processes
                                                                                                             on the atomic scale typically evolve on a few-
                                                                                                             femtosecond to sub-femtosecond time scale.
Synthesized Light Transients                                                                                 Time-domain access to these dynamics requires
                                                                                                             the extension of electric field control to optical
A. Wirth,1 M. Th. Hassan,1,2 I. Grguraš,1 J. Gagnon,1 A. Moulet,1 T. T. Luu,1                                frequencies.
S. Pabst,3,4 R. Santra,3,4 Z. A. Alahmed,2 A. M. Azzeer,2 V. S. Yakovlev,1,5                                     As a first step to this end, measurement (2, 3)
V. Pervak,5 F. Krausz,1,5 E. Goulielmakis1*                                                                  and control (4–9) of the phase of field oscilla-
                                                                                                             tions relative to their envelope [carrier-envelope
Manipulation of electron dynamics calls for electromagnetic forces that can be confined to                   phase (CEP)] yielded reproducible few-cycle light
and controlled over sub-femtosecond time intervals. Tailored transients of light fields can provide          waveforms (10). Attosecond metrology (11) was
these forces. We report on the generation of subcycle field transients spanning the infrared, visible,       further advanced by the reproducible generation
and ultraviolet frequency regimes with a 1.5-octave three-channel optical field synthesizer and              and measurement of isolated attosecond pulses
their attosecond sampling. To demonstrate applicability, we field-ionized krypton atoms within
a single wave crest and launched a valence-shell electron wavepacket with a well-defined                     1
                                                                                                              Max-Planck-Institut für Quantenoptik (MPQ), Hans-Kopfermann-
initial phase. Half-cycle field excitation and attosecond probing revealed fine details of                   Strasse 1, D-85748 Garching, Germany. 2Department of Physics
atomic-scale electron motion, such as the instantaneous rate of tunneling, the initial charge                and Astronomy, King Saud University, Riyadh 11451, King-
distribution of a valence-shell wavepacket, the attosecond dynamic shift (instantaneous                      dom of Saudi Arabia. 3Center for Free-Electron Laser Science,
ac Stark shift) of its energy levels, and its few-femtosecond coherent oscillations.                         Deutsches Elektronen Synchrotron, Notkestrasse 85, 22607
                                                                                                             Hamburg, Germany. 4Department of Physics, University of Ham-
                                                                                                             burg, Jungiusstrasse 9, 20355 Hamburg, Germany. 5Depart-
                                                                                                             ment für Physik, Ludwig-Maximilians-Universität (LMU), Am
       he generation and measurement (briefly,           cuits with sub-picosecond temporal resolution

                                                                                                             Coulombwall 1, D-85748 Garching, Germany.
       synthesis) of electric field transients per-      (1) and constitutes a base technology for advanc-   *To whom correspondence should be addressed. E-mail:
       mitted the characterization of electric cir-      ing high-speed electronics and electron-based

Fig. 1. Apparatus for infrared-visible-ultraviolet
field synthesis. (A) Schematic representation of a
prototypical three-channel light field synthesizer.
(B) Spectrum of the coherent radiation at the exit
of the hollow-core fiber (dashed line). Spectra ex-
iting the individual channels (not to scale) are
shown in red for ChNIR (700 to 1100 nm), yellow
for ChVIS (500 to 700 nm), and blue for ChVIS-UV
(350 to 500 nm). (C) Temporal intensity (solid
lines) and phase profiles (dashed curves) of the
respective pulses. The thin black lines depict the
intensity profiles of the corresponding bandwidth-
limited pulses, with durations of tCh(NIR) = 6.8 fs,
tCh(VIS) = 5 fs, and tCh(VIS-UV) = 4.5 fs. Insets show
photos of the respective beam profiles taken at the
exit of the apparatus.

                                          SCIENCE       VOL 334      14 OCTOBER 2011                                                             195
      (12–14). Control (15) and real-time observation           trol, we field-ionized atoms within a single wave    near the edges of the spectral bands. As a re-
      (16–21) of electronic processes would greatly             crest and triggered valence electron motion on       sult, the pulses in the individual channels are
      benefit from sub-femtosecond sculpting and con-           a sub-femtosecond scale. By providing a sub-         compressed close to their bandwidth-limited
      finement of strong light fields. We are able to           femtosecond optical field trigger and a robust       durations (Fig. 1C).
      demonstrate this capability along with some of            attosecond probe, subcycle light transients estab-       The chirp, the CEP, and the delay of the pulses
      its consequences.                                         lish sub-femtosecond pump-probe spectroscopy.        formed in ChNIR, ChVIS, and ChVIS-UV can be
          Tailoring light fields on the electronic time             1.5-octave optical field synthesizer. We         precisely controlled by wedges and nanometer-
      scale requires the coherent superposition and ma-         produced coherent supercontinua by propagat-         precision delay stages, respectively. The adjust-
      nipulation of frequencies over more than an oc-           ing ~ 0.8 mJ, ~25-fs pulses carried at a wave-       ment of the beam size in each channel—via an
      tave in the visible and flanking spectral ranges.         length of l0 ~ 780 nm in a hollow-core fiber         iris—allows control of the pulse’s energy [sup-
      So far, this demand could only be met through             filled with neon gas. Spectral broadening was        porting online material (SOM) text, section 1].
      the technique of molecular modulation (22–27).            enhanced with respect to previous experiments        These control knobs offer both subcycle shaping
      This approach recently allowed the subcycle               (34) by raising the gas pressure to ~3.5 bar,        of the generated fields and compression close
      shaping of optical fields via the superposition           resulting in a nearly uniform (to within 20 dB)      to their bandwidth limit. Owing to the high ef-
      of quasi-monochromatic waves in the infrared-             energy distribution over a bandwidth of >0.6         ficiency of the chirped multilayer optics used
      visible range (28). Although these periodic               PHz (330 to 1100 nm) (Fig. 1B). Manipulation         (fig. S2), the device transmits some ~83% of the
      waveforms are highly relevant to advancing mod-           of individual spectral components requires their     incident continuum beam (Fig. 1B, dotted line),
      ern electronics, time-domain access to electronic         spatial separation and subsequent recombina-         resulting in a pulse energy of ~0.3 mJ/pulse at the
      phenomena calls for the temporal confinement              tion. The conventional approach, based on prisms     exit of the apparatus (ChNIR ~ 250 mJ, ChVIS ~ 35 mJ,
      of the sculpted waveform to a single cycle or             and liquid crystal modulators (31), is hardly        and ChVIS-UV ~ 15 mJ). The setup is assembled
      just a few oscillation cycles. We refer to such           scalable for super-octave–spanning operation;        on a monolithic aluminum base plate, with active
      super-octave optical waveforms as light transients.       therefore, we instead implemented chirped mul-       thermal and interferometric path-length stabiliza-
      Recent experiments have paved the way toward              tilayer mirror technology, proposed in (34), which   tion (fig. S3).
      the synthesis of light transients (29–33), but they       offers scalability to several octaves in the visi-       By analogy with femtosecond electro-optic
      have not yet achieved the goal of subcycle field          ble and nearby spectral regions. Our prototyp-       sampling of THz transients (1), we can use atto-
      shaping and measurement.                                  ical three-channel device (Fig. 1A) subdivides       second streaking (12, 35) for sampling the
          Here, we report on the shaping, confine-              the aforementioned ~ 0.6-PHz spectral range          electric field of PHz transients. To this end, the
      ment, and attosecond sampling of the fields of            into three bands of nearly equal width—ChNIR,        PHz transients exiting our three-channel synthe-
      intense light transients within their carrier wave        700 to 1100 nm; ChVIS, 500 to 700 nm; and            sizer were gently focused (to a peak intensity
      cycle (~2.4 fs) over the frequency band of 0.3            ChVIS-UV, 350 to 500 nm—with the help of di-         of ~1014 W/cm2) into a neon gas jet, where they
      to 0.9 PHz. A variety of on-demand waveforms              chroic beamsplitters DBSVIS-NIR and DBSUV-VIS        generated broadband extreme ultraviolet (XUV)
      with controlled subcycle field evolution, yield-          (fig. S1). Dispersive chirped mirrors CMVIS/UV,      radiation (14) emitted in a near-diffraction-limited
      ing sub-femtosecond rise times or subcycle con-           CMVIS, and CMNIR compensate for the chirp            beam collinear with the driving radiation (SOM
      finement of instantaneous intensity, demonstrate          carried by the pulse as well as that introduced      text, section 2). Bandpass filtering (width of
      the power of PHz field synthesis. As an appli-            by the thin fused silica wedge pairs incor-          ~13 eV centered at ~85 eV) near the cut-off en-
      cation of enhanced atomic-scale electron con-             porated in each channel and the beamsplitters        ergy (~ 90 eV), implemented with multilayer

      Fig. 2. Synthesis of petahertz light field transients.
      (A to F) Attosecond streaking spectrograms com-
      posed of photoelectron spectra normalized to their
      integral (left) and the respective retrieved electric
      fields (middle) and instantaneous intensity (right).
      Relative intensities for the most intense field crests—
      normalized to the maximum—are given in brackets.
      From (A) to (C), the delay of ChVIS-UV is varied in
      steps of 200 as (~p/4). Dashed black lines in (B) and
      (C) show the field transients calculated from the
      reference waveform of (A). (D) Relative delays and
      CEPs of the individual channels are adjusted so as to
      create twin transients with a field minimum in
      between them. (E) ChNIR is delayed by 1.45 fs (~p),
      resulting in a high-frequency leading transient
      followed by a low-frequency tail. The dashed line
      in (E) shows the field transients calculated from
      the reference waveform of (D). Transients in (A)
      to (C), (E), and (F) carry less than one cycle within
      the FWHM of their temporal intensity profile. (F)
      tFWHM ~ 2.1 fs, incorporating only ~0.88 field
      cycles at the carrier wavelength of l0 ~ 710 nm.

196                                             14 OCTOBER 2011            VOL 334       SCIENCE
                                                                                                                                        RESEARCH ARTICLE
optics and thin metal foils, isolated a single at-       ization of the properties of the apparatus and          The instantaneous intensity shown in Fig. 2,
tosecond pulse (12, 14). Both pulses were then           subsequently on-demand synthesis of pre-            right, reveals substantial variations of the strength
focused into a second neon gas jet placed near           scribed fields. To this end, we used the wave-      of consecutive wave crests upon these transfor-
the entrance of a time-of-flight electron spec-          form in Fig. 2A as a reference and retrieved        mations. The reference waveform (Fig. 2A) ex-
trometer (TOF) for measuring the XUV-induced,            the values of the control parameters of the sys-    hibits half cycles with relative intensities of (0.64,
laser-field–streaked photoelectron spectra ver-          tem such as the field amplitudes, phase delays,     1, 0.46) underpinning the subcycle character of
sus delay (an attosecond streaking spectro-              and the CEPs of the three channels through          the transient. Delaying ChVIS-UV gradually trans-
gram) (35).                                              the numerical band-pass filtering of the mea-       forms the field into the highly asymmetric tran-
    Attosecond streaking spectrograms of subcycle        sured output waveform within the spectral           sient of Fig. 2C, with the temporal extension left
waveforms synthesized from near-bandwidth–               ranges defined by ChNIR, ChVIS, and ChVIS-UV.       almost unchanged. This transient carries its most
limited fields exiting ChNIR, ChVIS, and ChVIS-UV        The field transients shown in Fig. 2, B and C,      intense field crest right at its leading edge, followed
are shown in Fig. 2, left. The spectrograms are          were then synthesized by delaying ChVIS-UV          by half cycles of decreasing intensity (1, 0.89,
composed of a series of laser-field–streaked             in steps of p/4 (~200 as) with respect to the       0.57), resulting in a sub-femtosecond rise time of
XUV photoelectron spectra recorded as a func-            reference waveform. A more complex, nonsi-          its instantaneous intensity. On the other hand, a
tion of delay between the XUV pulse and the              nusoidal transient, which is generated by de-       transient with its two most intense field crests sep-
subcycle field. A delay step of 0.2 fs was used,         laying ChNIR so that the fields from the three      arated by ~4.5 fs and a half-cycle virtually an-
which safely allows sampling up to the highest           channels cancel each other at the center of the     nihilated in between is revealed in Fig. 2D. The
frequency components (~0.9 PHz) in the wave-             waveform, is shown in Fig. 2D. The transient        transient with the largest degree of temporal en-
form. Remarkably, each spectrogram reveals an            shown in Fig. 2E is generated via delay of          ergy confinement is shown in Fig. 2F, with field
isolated <200-as XUV pulse. This is a direct con-        ChNIR by ~ p with respect to the waveform           crest intensities of (0.38, 1, 0.62) and with ~35%
sequence of the substantial subcycle variation           shown in Fig. 2D, resulting in a single intense     of its energy carried in a single wave crest.
of field amplitude (Fig. 2, right), which provides       field crest pointing in the opposite direction to       Field ionization and its real-time sampling.
an efficient temporal gate—via ionization con-           the peak field in Fig. 2A. The full red and the     We used transients with a central field crest ~1.7
finement and/or energy filtering—for isolating           dashed black lines in Fig. 2, middle, depict, re-   times more intense than the adjacent half cy-
a single attosecond burst in the XUV radiation           spectively, waveforms measured or calculated        cles (Fig. 3B) to ionize Kr atoms enclosed in a
emitted by the ionizing atoms. The feasibility           from the constituent ChNIR, ChVIS, and ChVIS-UV     quasi-static gas cell (length l ≈ 0.74 mm) at a
of generating a robust isolated attosecond probe         fields retrieved from the reference waveform        density of ≈5.6 × 1018 cm−3. The gas cell was
for a wide range of waveforms not only is re-            (Fig. 2A), with control parameters changed by       positioned at the laser focus, replacing the neon
quired for the sampling of the transients but is         known amounts with respect to those of the ref-     gas jet previously used for recording the streak-
most important also for the interrogation of pro-        erence waveform. The agreement between pre-         ing spectrograms shown in Fig. 3A (the experi-
cesses triggered and/or controlled by the op-            diction and measurement demonstrates controlled     mental setup is shown in fig. S4). This procedure
tical field transients.                                  sub-femtosecond shaping, complete characteri-       reveals the absolute timing of any process ini-
    Any of the retrieved electric field wave-            zation, and reproducibility (synthesis) of peta-    tiated or affected by the field transients with at-
forms (Fig. 2, middle) permits full character-           hertz field transients.                             tosecond precision (SOM text, section 8).

Fig. 3. Ionization with a subcycle light field transient. (A) Streaking spectrogram
and (B) retrieved electric field of the transient used for field ionization of Kr
atoms. (C) Energy-level diagram and attosecond absorption spectrogram of Kr
ions. The plotted absorbance is defined as A(ℏw,t) = −ln[I(ℏw,t)/I0 (ℏw)],
where I0 (ℏw) is the spectral density recorded with attosecond pulses averaged
over delays between –20 and –5.7 fs (preceding the pump), and I(ℏw,t) is the
spectral density recorded at a pump-probe delay t. The delay is varied in steps
of 100 as in the range of (–3 fs, 3 fs) and 300 as in the ranges of (–6 fs, –3 fs)
and (3 fs, 6 fs). (D) Absorption spectra for different delays. The dots and error
bars represent the mean value and SE evaluated from 12 spectra recorded at
each delay step.

                                           SCIENCE        VOL 334    14 OCTOBER 2011                                                       197
          We increased the intensity with an iris to          in the same apparatus, this evolution can be di-           femtosecond confinement of field ionization. It is
      ≈4.8 × 1014 W/cm2 and ionized ~16% of Kr                rectly timed and contrasted with the evolution             this confinement to a single field crest that allows
      atoms in the gas cell. We then probed strong-           of the ionizing field [ jEL ðtÞj2 , shown by the           quantitative evaluation of the time-dependent rate
      field ionization with a time-delayed attosecond         dashed line].                                              of optical field ionization and state-selective pop-
      pulse by measuring transient absorption spectra             The buildup of the retrieved ionic populations         ulation dynamics. For the populations depicted,
      (19). Krypton ions created in their 4p−1 ground-
                                             j¼3=2            exhibits steps that are in synchrony with the field        we evaluate a peak production rate for reff ,3=2 ðtÞ
      state manifold and in the 4p−1   j¼1=2 excited-state    crests of the transient. This becomes even more                                                    ðT1=2Þ
                                                                                                                         of Gpeak = (0.12 T 0.01) fs–1 and for r1=2,1=2 ðtÞ of
      manifold, comprising four (mj=3/2 = –3/2, –1/2,         evident from the population rates obtained by
      1/2, 3/2) and two (mj=1/2 = –1/2, 1/2) states,          taking the time-derivative of the ionic popula-            Gpeak = (0.059 T 0.009) fs–1, which is in excellent
      respectively, are promoted to the 3d–1 core-            tions in Fig. 4A. The ionization rate, which is            agreement with results obtained by numerically in-
      hole excited states by absorption of XUV photons        estimated as dt reff ,3=2 ðtÞ and shown in Fig. 4B
                                                                               3=2                                       tegrating the Schrödinger equation of a single-active
      from the attosecond pulse (Fig. 3C, left) (36).         (dots and red line), exhibits three main features at       electron model in three dimensions (SOM text,
      The resulting spectra transmitted through the gas       the crests of the ionizing field. The main ioniza-         section 6), yielding Gpeak ðreff ,3=2 Þ ¼ 0:13 fsÀ1
      cell were recorded as a function of the delay be-       tion burst is responsible for approximately 80%                           ðT1=2Þ
                                                                                                                         and Gpeak ðr1=2,1=2 Þ ¼ 0:059 fsÀ1 , which also
      tween the attosecond probe (pulse duration of           of the ion population and has a full width at half
      ~200 as, centered at ~85 eV) and the ionizing           maximum (FWHM) of <0.7 fs, indicating a sub-               well reproduces details of the temporal evolution
      field transient (pump), yielding the absorption
      spectrogram shown in Fig. 3C, right. The rele-
      vant transitions, 4p−1 → 3d5=2 , 4p−1 → 3d3=2 ,
                           3=2              1=2
               −1        −1
      and 4p3=2 → 3d3=2 , are indicated with arrows
      in the Fig. 3C level diagram. The absorption
      lines recorded at the leading edge of the ioniz-
      ing field transient (Fig. 3D, i and ii, for ex-
      ample) have a characteristic profile revealing
      negative absorbance at photon energies be-
      low the resonances and positive absorbance
      above them (37). These transient line shapes
      gradually evolve to quasi-steady-state profiles
      toward the trailing edge of the ionizing field
      (Fig. 3D, iii).
          State-selective sub-femtosecond tracing of
      field ionization. In order to relate the attosecond
      transient absorption spectra with ion population
      dynamics, we have performed numerical simu-
      lations by treating the generated Kr ions within
      a simplified, three-level model using the density
      matrix formalism (SOM text, section 3). The sim-
      ulations reproduce well the experimental data
      (fig. S5) and demonstrate—in agreement with
      previous studies (38, 39)—that the distortions
      of the transient absorption spectra are due to
      the synthesized pump field acting on the XUV-
      initiated, time-dependent ionic polarization re-
      sponse. Moreover, our simulations (based on
      an adiabatic tunnel-ionization calculation) suggest
      that—under the conditions of our experiments—
      the emerging absorption lines coincide with the
      population dynamics of the relevant ionic states,
      permitting retrieval of their transient evolution
      from the peak absorbances with good (~10%)
      accuracy (fig. S9). Figure 4A shows the sub-
      femtosecond evolution of the effective transient
      population in the ground state manifold 4p−1   j¼3=2
      (black dots), defined asreff ,3=2ðtÞ ¼ r3=2,3=2 ðtÞ þ
      ar3=2,3=2 ðtÞ, where a = 2/3 reflects the higher        Fig. 4. Attosecond ionization and Stark effect dynamics in Kr+. (A) Population dynamics in the ground-
                                                                                                                                                                     ( / 2)
                                                              state 4p−1 3/2 manifold (dots), reff2,3/2 (t), and in the excited-state 4p−1 1/2 manifold (diamonds), r1T1,1/2 (t),
      transition cross-section for the (mj = T1/2) transi-             j=                      3/                                       j=                            /2
      tions between 4p−1 and3d5=2 states; the 4p−1
                       j¼3=2                         j¼1=2    retrieved from the absorption spectrogram of Fig. 3C and contrasted with the instantaneous intensity
                              ðT1=2Þ                          (dashed line) as well as with the prediction of numerical simulations convolved with the XUV probe pulse
      manifold population r1=2,1=2 ðtÞ is represented by
                                                              duration (green and purple lines). (B) Ionization (population) rate dreff2,3/2 (t)/dt evaluated from the data
      black diamonds. These populations are retrieved
                                                              in (A) (dots and red line) in comparison with the theoretical prediction (green line). (C) Shift of the central
                        −1               −1
      from the 4p−1 → 3d5=2 and 4p−1 → 3d3=2 ab-
                 3=2              1=2
                                                              energy of the 4p−12 → 3d5/2 transition evaluated by fitting a Lorentzian profile to the attosecond
      sorbances versus pump-probe delay, respective-          transient absorption spectra shown in Fig. 3C. Red curves in (A) to (C) and magenta curve in (A) are guides
      ly. Because attosecond streaking is performed           to the eye obtained by three adjacent point fast Fourier transform–smoothing.

198                                            14 OCTOBER 2011            VOL 334         SCIENCE
                                                                                                                                       RESEARCH ARTICLE
                                                                                                                                                         ðm Þ
of the ionic populations (Fig. 4A, green and         ulates the frequency of the respective coherent        the form of the diagonal matrix elements rj, jj0 ¼j of
purple lines, and B, green line).                    dipole emission. Because the decay of the emis-                                       ð3=2Þ       ð−3=2Þ
                                                                                                            the reduced density matrix—r3=2,3=2 þ r3=2,3=2 ¼
    Observation of the instantaneous optical         sion lasts several field cycles (t = t3d ~ 7.5 fs),
                                                                                                                              ð1=2Þ     ð−1=2Þ
Stark shift. The shift DE of quantum energy lev-     the Stark effect does not merely shift but also        0:315 T 0:024, r3=2,3=2 þ r3=2,3=2 = 0:400 T 0:024;
els of atoms, molecules, or solids induced by an     reshapes the transient absorption lines shown           ð1=2Þ        ð−1=2Þ
optical field EL(t)—the ac Stark shift (40)—plays    in Fig. 3D, i and ii. Thanks to the nearly instan-     r1=2,1=2 þ r1=2,1=2 ¼ 0:285 T 0:004—and a de-
a central role in fundamental dynamical pro-         taneous triggering of the polarization oscillations    gree of coherence of g = 0.85 T 0.06, which
cesses. So far, only cycle-averaged Stark effects    and their subsequent rapid decay, signatures of        exceeds that measured in the previous exper-
have been accessible to experiments (41). If the     the instantaneous ac Stark shift come to light         iment (19) by ~1.4 times and is unparalleled for
laser frequency wL is much smaller than atomic       in our transient absorption spectra (Fig. 4C),         long-lived (>1 fs) coherences in the valence
resonance frequencies, the Stark shift of a non-     which is in agreement with our simulations             shell. Our simulations [based on a state-of-the-art
degenerate atomic level is expected to instant-      (figs. S8 and S9).                                     three-dimensional time-dependent configuration-
ly follow variations of the laser field: DEðtÞ ¼         Valence wavepacket with well-defined quan-         interaction singles approach (43) that includes
−1=2mEL ðtÞ, where m = m(wL) is the atomic
                                                     tum phase. Attosecond probing of few-cycle–            correlation dynamics between the field-generated
polarizability (42). When an XUV pulse cre-          driven field ionization of Kr atoms has revealed       hole and the photoelectron and has been ex-
ates a coherent superposition of two states with     the emergence of a valence electron wavepacket         tended to include spin-orbit interaction (44)] well
a difference in their respective polarizabilities    in the 4p subshell of the Kr+ ensemble—as a con-       predict the measured coherence as well as frac-
Dm(wL), the induced polarization oscillations        sequence of liberation of electrons from the 4pj=3/2   tional populations in the 4p−1 and 4p−1 mani-
                                                                                                                                        j¼1=2       j¼3=2
experience a phase shift, which is approximately     as well as the 4pj=1/2 manifolds—separated in          folds (SOM text, section 7).
given by                                             energy by spin-orbit coupling in the Kr atoms              The nearly perfect coherence is, once again,
                                                     (19). By repeating this attosecond absorption          a direct consequence of the sub-femtosecond
                            Dm t 2 ′ ′               spectroscopic experiment with our subcycle tran-       width of the ionization gate. This confinement,
         Dϕdipole ðtÞ ≈ −       ∫ E ðt Þdt    ð1Þ
                            2ℏ t0 L                  sient shown in Fig. 3B, and supplementing it           along with sub-femtosecond absolute timing in-
                                                     with attosecond streaking, we can now launch           formation from streaking, has far-reaching con-
where t0 denotes the moment of arrival of the        a valence wavepacket within a sub-femtosecond          sequences. The former allows launching of the
attosecond XUV pulse (with tXUV ≪ 2p/wL).            interval and with sub-femtosecond absolute tim-        wavepacket with a well-defined initial phase,
The instantaneous Stark shift detunes the ener-      ing accuracy.                                          whereas the latter permits reliable determina-
gy at which the atom most efficiently absorbs            From the recorded absorption spectrogram           tion of this initial quantum phase. The retrieved
photons from the XUV probe pulse—in our case,        (Fig. 5A), we retrieve (SOM text, section 4) the       phase f(t) [equation 2 in (19)] of the quantum
by the 3d → 4p transition in Kr+ ions—and mod-       fractional populations of the six ionic states, in     superposition is shown in Fig. 5, B and C, along
                                                                                                            with representative snapshots of the generated
                                                                                                            ensemble-averaged hole density distributions,
                                                                                                            as evaluated from our data in Fig. 5A. Linear
                                                                                                            extrapolation of f(t) to “time zero,” the birth of
                                                                                                            the hole at the peak of the ionizing field tran-
                                                                                                            sient (Fig. 5B, blue line), yields f(t0) = (0.99 T
                                                                                                            0.04)p, which is in very good agreement with
                                                                                                            the prediction of our configuration-interaction–
                                                                                                            based simulations: f(t0) = 1.06p. This initial
                                                                                                            quantum phase implies an elongated initial hole-
                                                                                                            density distribution aligned with the ionizing field
                                                                                                            vector, which is commensurate with our intui-
                                                                                                            tive expectation.
                                                                                                                Outlook. Subcycle engineering of optical
                                                                                                            field transients opens new prospects for steer-
                                                                                                            ing the atomic-scale motion of electrons (15)
                                                                                                            with the electric force of light and for driving
                                                                                                            complex valence-shell dynamics in molecules
                                                                                                            (45). As a simple manifestation of enhanced
                                                                                                            control over valence shell dynamics, they al-
                                                                                                            low sub-femtosecond temporal confinement of
                                                                                                            ionization and precise associated triggering of
                                                                                                            a wealth of subsequent electronic phenomena.
                                                                                                            They also provide an isolated attosecond pho-
                                                                                                            ton probe for interrogating the unfolding elec-
                                                                                                            tronic and—in molecules—nuclear motions by
                                                                                                            means of attosecond absorption and/or photo-
                                                                                                            electron spectroscopy, as well as an isolated
Fig. 5. Initial quantum phase and density distribution of a valence electron wavepacket. (A) Attosecond     electron probe for tracing these dynamics via
XUV transient absorption spectrogram of Kr atoms field-ionized by a subcycle field transient shown by       electron diffraction (46) or high-harmonic in-
the blue line in (B). Linear extrapolation of the retrieved quantum phase f(t) [shown by the red line in    terferometry (47). This constitutes a substantial
(B)] to time zero as determined through attosecond streaking, allows access to the initial quantum          extension of the repertoire of attosecond sci-
phase f(t0) = (0.99 T 0.04)p of the valence electron wavepacket. (C) Ensemble-averaged initial hole         ence, which was previously restricted to either
density distribution in the valence shell at the instant of ionization and its subsequent evolution, as     triggering or probing electronic processes within
evaluated from (A).                                                                                         a sub-femtosecond time window. The feasibility

                                       SCIENCE       VOL 334        14 OCTOBER 2011                                                      199
      of sub-femtosecond pump-probe interrogation                 20. S. Zherebtsov et al., Nat. Phys., published online            40. S. H. Autler, C. H. Townes, Phys. Rev. 100, 703
      of strong-field phenomena opens exciting re-                    24 April 2011 (10.1038/nphys1983).                                (1955).
                                                                  21. P. Eckle et al., Science 322, 1525 (2008).                    41. B. J. Sussman, Am. J. Phys. 79, 477 (2011).
      search prospects. Real-time insight into multiple           22. S. E. Harris, A. V. Sokolov, Phys. Rev. Lett. 81, 2894        42. N. B. Delone, V. P. Krainov, Phys. Uspekhi 169, 753
      ionization in multi- to single-cycle laser fields,              (1998).                                                           (1999).
      or into strong-field–induced electron correlations          23. A. V. Sokolov, D. R. Walker, D. D. Yavuz, G. Y. Yin,          43. L. Greenman et al., Phys. Rev. A 82, 023406 (2010).
      in atoms, molecules, or solids (48, 49), are but a              S. E. Harris, Phys. Rev. Lett. 85, 562 (2000).                44. S. Pabst, L. Greenman, P. J. Ho, D. A. Mazziotti, R. Santra,
                                                                  24. J. Q. Liang, M. Katsuragawa, F. L. Kien, K. Hakuta,               Phys. Rev. Lett. 106, 053003 (2011).
      few examples.                                                   Phys. Rev. Lett. 85, 2474 (2000).                             45. F. Remacle, M. Nest, R. D. Levine, Phys. Rev. Lett. 99,
                                                                  25. A. V. Sokolov, D. R. Walker, D. D. Yavuz, G. Y. Yin,              183902 (2007).
          References and Notes                                        S. E. Harris, Phys. Rev. Lett. 87, 033402 (2001).             46. M. Meckel et al., Science 320, 1478 (2008).
       1. J. F. Whitaker et al., Microelectron. Eng. 12, 369      26. T. Suzuki, M. Hirai, M. Katsuragawa, Phys. Rev. Lett. 101,    47. O. Smirnova et al., Nature 460, 972 (2009).
          (1990).                                                     243602 (2008).                                                48. M. Gertsvolf, M. Spanner, D. M. Rayner, P. B. Corkum,
       2. L. Xu et al., Opt. Lett. 21, 2008 (1996).               27. Z.-M. Hsieh et al., Phys. Rev. Lett. 102, 213902                  J. Phys. B 43, 131002 (2010).
       3. T. Brabec, F. Krausz, Rev. Mod. Phys. 72, 545               (2009).                                                       49. M. Durach, A. Rusina, M. F. Kling, M. I. Stockman,
          (2000).                                                 28. H.-S. Chan et al., Science 331, 1165 (2011).                      Phys. Rev. Lett. 105, 086803 (2010).
       4. J. Reichert et al., Opt. Commun. 172, 59 (1999).        29. R. K. Shelton et al., Science 293, 1286 (2001).               Acknowledgments: We acknowledge the development of
       5. D. J. Jones et al., Science 288, 635 (2000).            30. M. Yamashita, K. Yamane, R. Morita, IEEE J. Sel. Top.             dedicated attosecond soft x-ray optics by M. Hofstetter
       6. S. A. Diddams et al., Phys. Rev. Lett. 84, 5102             Quantum Electron. 12, 213 (2006).                                 and U. Kleineberg (LMU, MPQ). This work was supported
          (2000).                                                 31. S. Rausch, T. Binhammer, A. Harth, F. X. Kärtner,                 by the Max Planck Society, the European Research
       7. A. Apolonski et al., Phys. Rev. Lett. 85, 740 (2000).       U. Morgner, Opt. Express 16, 17410 (2008).                        Council grant (Attoelectronics-258501), the Deutsche
       8. T. Udem, R. Holzwarth, T. W. Hänsch, Nature 416, 233    32. G. Krauss et al., Nat. Photonics 4, 33 (2010).                    Forschungsgemeinschaft Cluster of Excellence: Munich
          (2002).                                                 33. K. Okamura, T. Kobayashi, Opt. Lett. 36, 226 (2011).              Centre for Advanced Photonics (,
       9. S. T. Cundiff, J. Ye, Rev. Mod. Phys. 75, 325 (2003).   34. E. Goulielmakis et al., Science 317, 769 (2007).                  the King Saud University–MPQ collaboration, and the
      10. A. Baltuška et al., Nature 421, 611 (2003).             35. E. Goulielmakis et al., Science 305, 1267 (2004).                 European Research Training Network ATTOFEL.
      11. M. Hentschel et al., Nature 414, 509 (2001).            36. This indicates that relative to the ground state
      12. R. Kienberger et al., Nature 427, 817 (2004).               configuration, a hole (or electron vacancy) is created in     Supporting Online Material
      13. G. Sansone et al., Science 314, 443 (2006).                 the nl−1 nl subshell, where j denotes the total angular
      14. E. Goulielmakis et al., Science 320, 1614 (2008).           momentum and mj its projection on the z axis, the latter      SOM Text
      15. M. F. Kling et al., Science 312, 246 (2006).                being aligned with the laser polarization.                    Figs. S1 to S10
      16. G. Sansone et al., Nature 465, 763 (2010).              37. B. R. Mollow, Phys. Rev. A 5, 1522 (1972).                    References (50–59)
      17. M. Schultze et al., Science 328, 1658 (2010).           38. T. Unold, K. Mueller, C. Lienau, T. Elsaesser, A. D. Wieck,
      18. J. Mauritsson et al., Phys. Rev. Lett. 105, 053001          Phys. Rev. Lett. 92, 157401 (2004).                           24 June 2011; accepted 30 August 2011
          2010).                                                  39. C. H. B. Cruz, J. P. Gordon, P. C. Becker, R. L. Fork,        Published online 1 September 2011;
      19. E. Goulielmakis et al., Nature 466, 739 (2010).             C. V. Shank, Int. J. Quant. Elec. 24, 261 (1988).             10.1126/science.1210268

                                                                                                                                    coherence (9), and lie at the heart of superexchange-
      Observation of Correlated Particle-Hole                                                                                       mediated spin interactions that form the basis of
                                                                                                                                    quantum magnetism in multicomponent quan-

      Pairs and String Order in Low-Dimensional                                                                                     tum gas mixtures (10–12).
                                                                                                                                        In a one-dimensional system, the appearance
                                                                                                                                    of correlated particle-hole pairs at the transition
      Mott Insulators                                                                                                               point from a SF to a MI is intimately connected to
                                                                                                                                    the emergence of a hidden string-order param-
                                                                                                                                    eter OP (13, 14):
      M. Endres,1* M. Cheneau,1 T. Fukuhara,1 C. Weitenberg,1 P. Schauß,1 C. Gross,1 L. Mazza,1
      M. C. Bañuls,1 L. Pollet,2 I. Bloch,1,3 S. Kuhr1,4

      Quantum phases of matter are characterized by the underlying correlations of the many-body
                                                                                                                                        O2 ¼ lim O2 ðlÞ ¼ lim
                                                                                                                                                  P                   〈     ∏
                                                                                                                                                                 l→∞ k ≤ j ≤ kþl
                                                                                                                                                                                    eiπδn j   〉   ð1Þ

      system. Although this is typically captured by a local order parameter, it has been shown that a
      broad class of many-body systems possesses a hidden nonlocal order. In the case of bosonic                                                 %    %
                                                                                                                                        Here, δnj ¼ nj − n denotes the deviation in
      Mott insulators, the ground state properties are governed by quantum fluctuations in the form of                              occupation of the jth lattice site from the
      correlated particle-hole pairs that lead to the emergence of a nonlocal string order in one                                   average background density, and k is an arbi-
      dimension. By using high-resolution imaging of low-dimensional quantum gases in an optical                                    trary position along the chain. In the simplest
      lattice, we directly detect these pairs with single-site and single-particle sensitivity and                                  case of a MI with unity filling (n ¼ 1), relevant
      observe string order in the one-dimensional case.                                                                             to our experiments, each factor in the product of
                                                                                                                                    operators in Eq. 1 yields −1 instead of +1 when a
             he realization of strongly correlated quan-          interaction energy vanishes, particle fluctua-                    single-particle fluctuation from the unit back-

      T      tum many-body systems using ultracold
             atoms has enabled the direct observation
      and control of fundamental quantum effects
                                                                  tions are completely suppressed and the lattice
                                                                  sites are occupied by an integer number of par-
                                                                  ticles. However, at a finite tunnel coupling but                  1
                                                                                                                                     Max-Planck-Institut für Quantenoptik, 85748 Garching,
                                                                                                                                    Germany. 2Theoretische Physik, Eidgenössische Technische
      (1–3). A prominent example is the transition                still in the Mott insulating regime, quantum fluc-
      from a superfluid (SF) to a Mott insulator (MI),            tuations create correlated particle-hole pairs                    Hochschule (ETH) Zurich, 8093 Zurich, Switzerland. 3Ludwig-
                                                                                                                                    Maximilians-Universität, 80799 Munich, Germany. 4University
      occurring when interactions between bosonic                 on top of this fixed-density background, which                    of Strathclyde, Scottish Universities Physics Alliance, Glasgow
      particles on a lattice dominate over their ki-              can be understood as virtual excitations. These                   G4 0NG, UK.
      netic energy (4–8). At zero temperature and in              particle-hole pairs fundamentally determine the                   *To whom correspondence should be addressed. E-mail:
      the limit where the ratio of kinetic energy over            properties of the MI, such as its residual phase        

200                                               14 OCTOBER 2011               VOL 334          SCIENCE 
ground density is encountered. In the SF, particle      To detect particle-hole pairs, we evaluated           particle-hole pairs extend over arbitrary distances
and hole fluctuations occur independently and        two-site parity correlation functions (25)               and are therefore uncorrelated. Their presence
are uncorrelated, such that OP ¼ 0. However,                                                                  leads to a reduction of the correlation signal. We
in the Mott insulating phase, density fluctua-                         % %           % %                      found no correlations when performing the same
                                                              CðdÞ ¼ 〈 sk skþd 〉 − 〈 sk 〉 〈 skþd 〉      ð2Þ
tions always occur as correlated particle-hole                                                                analysis perpendicular to the one-dimensional
pairs, resulting in OP ≠ 0. For a homogeneous                                                                 tubes (blue circles in Fig. 2B), showing that the
system, OP is expected to follow a scaling of                 %
                                                     where sk ¼ eiπδnk is the parity operator at site k       coupling between the tubes was negligible.
Berezinskii-Kosterlitz-Thouless (BKT) type (15).     and d is the distance between the lattice sites. For         Our data show very good agreement with
Nonlocal correlation functions, like the string-                          %
                                                     the case of n ¼ 1, sk yields +1 for an odd occu-         ab initio finite-temperature matrix product state
order parameter defined above, have been in-         pation number nk and −1 for an even nk. If a             (MPS) calculations (28, 29) at temperature T =
troduced in the context of low-dimensional quantum   particle-hole pair exists on sites k and k + d, the      0.09 U/kB (where kB is Boltzman’s constant)
systems. They classify many-body quantum             same parity s(nk) = s(nk+d) = −1 is detected             (Fig. 2A, solid line) that also take into account
phases that are not amenable to a description        (Fig. 1). The existence of correlated particle-hole      our harmonic trapping potential with frequency
through a local order parameter, typically used                                                   % %
                                                     pairs therefore leads to an increase of 〈 sk skþd 〉      w/(2π) = 60(1) Hz. Compared with a homoge-
in the Landau paradigm of phase transitions.                                        % %
                                                     above the factorized form 〈 sk 〉 〈 skþd 〉, which re-     neous system at T = 0 (dashed line), the ex-
Examples include spin-1 chains (16) and spin-        sults from uncorrelated fluctuations, for exam-          perimental signal is reduced, especially around
1/2 ladders (17), fermionic Mott and band            ple, because of thermal excitations. We obtained         the maximum. This reduction can be attributed in
insulators (18), and Haldane insulators in one-      C(d ) from our deconvolved images by an aver-            equal parts to the finite temperature of our system
dimensional Bose gases (13, 14). Recently, the       age over many experimental realizations and              and the averaging over different local chemical
intimate connection of string order and local        by an additional average over k in a central             potentials. The latter is especially severe in the
symmetries has been uncovered (19), and wide-        region of interest.                                      one-dimensional case owing to the narrow width
ranging classification schemes for quantum               We first analyzed two-site parity correlations       of the Mott lobe for n ¼ 1 close to the critical
phases using such symmetry principles have           in one-dimensional systems (Fig. 2, A and B). To         point (15). Interestingly, the growth of particle-
been introduced (20, 21). We show that corre-        create isolated one-dimensional tubes, we kept           hole correlations º J 2/U 2 expected from first-
lated particle-hole pairs and string order can be    the lattice axis along y at a constant depth of Vy =     order perturbation theory (24) is limited to very
directly detected by using single-atom–resolved      17(1) Er , where Er ¼ h2 /(8ma2 ) denotes the re-
                                                                                      lat                     small values, J/U < 0.05, before deviations in
images of strongly correlated ultracold quantum      coil energy and m is the atomic mass of 87Rb.            the experiment and the numerical simulations
gases (22, 23).                                      Experimental uncertainties are marked as terms           are observed.
    We prepared a two-dimensional degenerate         in parentheses following numerical values. We                Because the dimensionality of the system
gas of ultracold 87Rb atoms before shining in        recorded the nearest-neighbor correlations C(d = 1)      plays an important role in its correlation properties,
a two-dimensional square optical lattice (lattice    for different values of J/U along the direction of       we also measured the two-site parity correlations
spacing alat = 532 nm) with variable lattice         the one-dimensional tubes (red circles in Fig. 2B),      across the two-dimensional SF-MI transition by
depths in x and y directions (23, 24). A micro-      where J and U are the tunneling matrix element           simultaneously varying J/U along both lattice axes
scope objective with a resolution comparable         and the on-site interaction energy in the Bose-          (Fig. 2C). In contrast to the one-dimensional case,
to the lattice spacing was used for fluorescence     Hubbard model, respectively (24). For small J/U,         we now observe the same nearest-neighbor cor-
detection of individual atoms. Because inelas-       the nearest-neighbor correlations vanish, because        relations within our error bars along both axes.
tic light-assisted collisions during the imaging     only uncorrelated thermal excitations exist deep         The maximum correlations are smaller than in
lead to a rapid loss of atom pairs, our scheme       in the MI regime. Because particle-hole pairs            one dimension, and the peak value is now reached
detects the parity of the atom number. We used       emerge with increasing J/U, we observe an in-            around the critical value (J /U )2d ≈ 0:06 (30).
an algorithm to deconvolve the images, yielding      crease of nearest-neighbor correlations until a          We compared our data with quantum Monte
single-site–resolved information of the on-site      peak value is reached, well before the critical value    Carlo (QMC) simulations for a homogeneous
parity. Typically, our samples contained 150 to      (J /U )1d ≈ 0:3 (26, 27) for the one-dimensional
                                                            c                                                 system at T = 0.1 U/kB (solid line in Fig. 2C) and
200 atoms in order to avoid MIs of occupation        SF-MI transition. The observed signal is a gen-          found good quantitative agreement. Here, the
numbers n > 1.                                       uine quantum effect because thermally induced            broader shape of the Mott lobe leads to a weaker

Fig. 1. Quantum-correlated particle-hole pairs in one-dimensional MIs. (A) In    ellipses) along the x direction of the one-dimensional MIs. (B) Such correlated
a two-dimensional array of atoms (blue circles), decoupled one-dimensional                                         %
                                                                                 fluctuations in the occupation nj are detected in the experiment as correlated
systems were created by suppressing tunneling along the y direction. Quantum                                %
                                                                                 fluctuations in the parity sj. The light red bar in (A) marks the one-dimensional
fluctuations then only induce correlated particle-hole excitations (yellow       chain chosen in (B) for further explanations.

                                       SCIENCE         VOL 334           14 OCTOBER 2011                                                   201

      Fig. 2. Nonlocal parity correlations. (A) (Top) Typical experimental fluo-
      rescence images for J/U = 0.06 (1), J/U = 0.11 (2), and J/U = 0.3 (3) for the
      one-dimensional geometry. (Bottom) Reconstructed on-site parity. Particle-
      hole pairs are emphasized by a yellow shading in (1). For increased J/U, the
      pairs start to proliferate, and an identification in a single experimental image
      becomes impossible (2 and 3). (B) One-dimensional nearest-neighbor
      correlations C(d = 1) as a function of J/U along the x (red circles) and y
      directions (blue circles). The curves are first-order perturbation theory
      (dashed-dotted line), DMRG calculations for a homogeneous system at T =
      0 (dashed line), and finite-temperature MPS calculations including
      harmonic confinement at T = 0.09 U/kB (solid line). (C) Parity correlations
      in two dimensions. Symbols have the same meaning as in (B). The curves
      are first-order perturbation theory (dashed-dotted line) and QMC
      calculations for a homogeneous system at T = 0.01 U/kB (dashed line)
      and at T = 0.1 U/kB (solid line). Each data point is an average over the
      central nine-by-seven lattice sites from 50 to 100 pictures. The error bars
      denote the 1s statistical uncertainty. The light blue shading highlights the SF phase.

      averaging effect over different local chemical
      potentials. The increased strength of the cor-
      relations and the larger shift of the maximum of
      the correlations relative to the critical point in
      the one-dimensional case directly reflect the
      more prominent role of quantum fluctuations in
      lower dimensions. This can also be seen from the
      on-site fluctuations C(d = 0) at the critical point,
      which are increased in the one-dimensional case
      (24). In both the one-dimensional and the two-
      dimensional systems, two-site correlations are
      expected to decay strongly with distance. Our
      data for the next-nearest-neighbor correlations
      C(d = 2) are consistent with this predicted be-
      havior (24).
          In addition to two-site correlations, we eval-
      uated string-type correlators O2 (l ) ¼ 〈 ∏kþl sj 〉
                                          P          j¼k
      (Eq. 1 and Fig. 1B) in our one-dimensional sys-
      tems, where the product is calculated over a chain
      of length l + 1. In the simplest case of a zero-       Fig. 3. Numerical calculation of the string-order parameter. O 2 (l) is shown as a function of J/U calculated
      temperature MI at J/U = 0, no fluctuations exist       with DMRG for a homogeneous chain (n = 1, T = 0) of total length 216. Lines show O 2 (l) for selected l
                                                                                                       ¯                                                   P
      and therefore O2 (l ) ¼ 1. As J/U increases, fluc-
                                                             (black to red colors). (Inset) Extrapolated values of O P = liml→∞ O 2 (l) together with a fit (black line) of the
      tuations in the form of particle-hole pairs appear.
                                                             form O 2 º exp f− A [ ( J/U )c − ( J/U ) ]      g, characteristic for a transition of BKT type (24).
                                                                                          1d              −1/2
      Whenever a certain number of particle-hole pairs              P
      lies completely within the region covered by
      the string correlator, the respective minus signs      ability to cut particle-hole pairs becomes larger.          neous system at T = 0. We show O2 ðlÞ in Fig. 3
      cancel pairwise. However, there is also the pos-       Finally, at the transition to the SF phase, the             for selected distances of l together with the extra-
      sibility that a particle-hole pair is cut by one end   pairs begin to deconfine and overlap, resulting             polated values of OP = liml→∞ OP ðlÞ (Fig. 3 inset),
                                                                                                                                               2            2

      of the string correlator, for example, when a          in a completely random product of signs and                 which we computed by using finite-size scaling
      particle exists at position < k and the corre-         O2 ¼ 0.
                                                               P                                                         (24). We performed a fit to the extrapolated values
      sponding hole has a position ≥ k, resulting in an          To support this intuitive argument, we calcu-           close to the critical point with an exponential scal-
      unpaired minus sign. As a consequence, O2 (l )   P     lated O2 (l ) numerically by using density-matrix
                                                                     P                                                   ing O2 º expf−A[(J /U )1d − (J /U )]−1=2 g char-
                                                                                                                                P                     c
      decreases with increasing J/U because the prob-        renormalization group (DMRG) in a homoge-                   acteristic for a transition of BKT type (Fig. 3)

202                                           14 OCTOBER 2011            VOL 334         SCIENCE 
                                                                                                                        References and Notes
                                                                                                                    1. D. Jaksch, P. Zoller, Ann. Phys. 315, 52 (2005).
                                                                                                                    2. M. Lewenstein et al., Adv. Phys. 56, 243 (2007).
                                                                                                                    3. I. Bloch, J. Dalibard, W. Zwerger, Rev. Mod. Phys. 80,
                                                                                                                       885 (2008).
                                                                                                                    4. M. P. A. Fisher, P. B. Weichman, G. Grinstein, D. S. Fisher,
                                                                                                                       Phys. Rev. B 40, 546 (1989).
                                                                                                                    5. D. Jaksch, C. Bruder, J. I. Cirac, C. Gardiner, P. Zoller,
                                                                                                                       Phys. Rev. Lett. 81, 3108 (1998).
                                                                                                                    6. M. Greiner, O. Mandel, T. Esslinger, T. W. Hänsch, I. Bloch,
                                                                                                                       Nature 415, 39 (2002).
                                                                                                                    7. T. Stöferle, H. Moritz, C. Schori, M. Köhl, T. Esslinger,
                                                                                                                       Phys. Rev. Lett. 92, 130403 (2004).
                                                                                                                    8. I. B. Spielman, W. D. Phillips, J. V. Porto, Phys. Rev. Lett.
                                                                                                                       98, 080404 (2007).
                                                                                                                    9. F. Gerbier et al., Phys. Rev. Lett. 95, 050404 (2005).
                                                                                                                   10. A. B. Kuklov, B. V. Svistunov, Phys. Rev. Lett. 90, 100401
                                                                                                                   11. L.-M. Duan, E. Demler, M. D. Lukin, Phys. Rev. Lett. 91,
                                                                                                                       090402 (2003).
                                                                                                                   12. S. Trotzky et al., Science 319, 295 (2008); 10.1126/
                                                                                                                   13. E. G. Dalla Torre, E. Berg, E. Altman, Phys. Rev. Lett. 97,
                                                                                                                       260401 (2006).
                                                                                                                   14. E. Berg, E. Dalla Torre, T. Giamarchi, E. Altman,
                                                                                                                       Phys. Rev. B 77, 245119 (2008).
                                                                                                                   15. T. Kühner, H. Monien, Phys. Rev. B 58, R14741 (1998).
                                                                                                                   16. M. den Nijs, K. Rommelse, Phys. Rev. B 40, 4709
Fig. 4. String correlators. (A) Experimental values of O 2 (l) for 0 ≤ l ≤ 8. (B) In-trap MPS calculations
                                                         P                                                             (1989).
at T = 0.09 U/kB. (C) Experimentally determined string correlator O 2 (l) as defined in Eq. 3 for                  17. E. Kim, G. Fáth, J. Sólyom, D. Scalapino, Phys. Rev. B 62,
lengths 1 ≤ l ≤ 8 (D) In-trap MPS calculations at T = 0.09 U/kB. The l axes in (C) and (D) have been                   14965 (2000).
inverted.                                                                                                          18. F. Anfuso, A. Rosch, Phys. Rev. B 75, 144420 (2007).
                                                                                                                   19. D. Pérez-García, M. M. Wolf, M. Sanz, F. Verstraete,
                                                                                                                       J. I. Cirac, Phys. Rev. Lett. 100, 167202 (2008).

that one expects in one dimension (14, 15). From                       ˜2
                                                           Second, OP (l ) ≈ O2 (l ) for long distances l
                                                                                                                   20. X. Chen, Z.-C. Gu, X.-G. Wen, Phys. Rev. B 83, 035107
                                                                                  P                                    (2011).
the fit, we find (J /U )1d = 0:295 − 0:320, which
                                                           because ∏k≤j≤kþl 〈 sj 〉 eventually decays to zero       21. N. Schuch, D. Pérez-García, J. I. Cirac, (2010); http://
is compatible with previously computed values              (except for the singular case J/U = 0 and T = 0). 
(26, 27, 24). The fact that OP ¼ 0 in the SF and                                        ˜2
                                                           The correlation function OP (l) can therefore be        22. W. S. Bakr et al., Science 329, 547 (2010); 10.1126/
OP > 0 in the MI as well as the agreement with             understood as an extension of the two-site cor-         23. J. F. Sherson et al., Nature 467, 68 (2010).
the expected scaling show that OP serves as an             relation function that essentially captures the phys-   24. Materials and methods are available as supporting
order parameter for the MI phase in one dimen-             ics behind string order in one-dimensional MIs.             material on Science Online.
sion (14). Additionally, the simulations demon-                Experimental and theoretical values for             25. E. Kapit, E. Mueller, Phys. Rev. A 82, 013644
strate that OP (l) is well suited to characterize the       ˜2
                                                           OP (l) are shown in Fig. 4, C and D. For small
                                                                                                                   26. T. D. Kühner, S. R. White, H. Monien, Phys. Rev. B 61,
SF-MI transition even for finite l.                               ˜2
                                                           J/U, OP (l) is reduced compared with O2 (l) be-             12474 (2000).
    Our experimentally obtained values of O2 (l)   P       cause few particle-hole pairs exist and O2 (l) is
                                                                                                          P        27. V. Kashurnikov, A. Krasavin, B. Svistunov, JETP Lett. 64,
for string length l ≤ 8 (Fig. 4A) agree qual-              close to its factorized form for short values of l.         99 (1996).
                                                                                                                   28. F. Verstraete, J. J. García-Ripoll, J. I. Cirac, Phys. Rev. Lett.
itatively well with in-trap MPS calculations at            In the case of vanishing J/U, we even expect
T = 0.09 U/kB (Fig. 4B). We observe a stronger              ˜2
                                                           OP (l) = 0 because all sites are completely de-
                                                                                                                       93, 207204 (2004).
                                                                                                                   29. M. Zwolak, G. Vidal, Phys. Rev. Lett. 93, 207205
decay of O2 (l ) with l compared to the T = 0              coupled. For intermediate J/U ≈ 0.1, OP (l)      ˜2         (2004).
case, because at finite temperature thermal fluc-          grows rapidly with l showing a strong deviation         30. B. Capogrosso-Sansone, S. Söyler, N. Prokof'ev,
                                                                                                                       B. Svistunov, Phys. Rev. A 77, 015602 (2008).
tuations lead to minus signs at random posi-               from the factorized form. Lastly, in the SF re-
tions of the chain and reduce the average value                    ˜2
                                                           gime, OP (l) becomes indiscernible from zero for
                                                                                                                   31. M. Saffman, T. Walker, K. Mølmer, Rev. Mod. Phys. 82,
                                                                                                                       2313 (2010).
of O2 (l ). Despite that, we still see a strong growth
     P                                                     large lengths in contrast to the nearest-neighbor       Acknowledgments: We acknowledge helpful discussions with
of O2 (l ), once the transition from the SF to the
      P                                                    two-site correlation function. Furthermore, our             E. Altman, E. Dalla Torre, M. Rizzi, and I. Cirac. This work
MI is crossed, with a similar behavior as in               data show a genuine three-site correlation, which           was supported by Max-Planck-Gesellschaft, Deutsche
                                                                                                                       Forschungsgemeinschaft, European Union (NAMEQUAM,
Fig. 3.                                                    we revealed after subtracting all two-site correla-         AQUTE, and Marie Curie Fellowship to M.C.), and Japan
    For a completely uncorrelated state, O2 (l )   P       tors in addition to local terms (24).                       Society for Promotion of Science (Postdoctoral Fellowship
factorizes to ∏k≤ j ≤ k þl 〈 sj 〉, and in a homoge-            We have shown direct measurements of non-               for Research Abroad to T.F.). L.P. is supported by the
neous system we would expect a decay with string           local parity-parity correlation functions on the            Swiss National Science Foundation under grant
                                                                                                                       PZ00P2-131892/1. DMRG simulations were performed
length of the form 〈 sj 〉lþ1 , which can be slow           single–lattice-site and single-atom level, and              with use of code released within the PwP project
provided the mean on-site parity 〈 sj 〉 is close           we demonstrated that a one-dimensional MI is                ( QMC calculations were performed on
to one. To rule out that our experimental data             characterized by nonlocal string order. A nat-              the Brutus cluster at ETH Zurich.
shows only such a trivial behavior, we define a            ural extension of our work would be to reveal,
new quantity OP (l) that more naturally reflects           for example, topological quantum phases such as         Supporting Online Material
the underlying correlations:                               the Haldane insulator of bosonic atoms (13, 14).
                                                                                                                   Materials and Methods
                                                           A Haldane insulator exhibits a hidden antiferro-        SOM Text
         OP (l) = O2 (l) −       ∏            %
                                            〈 sj 〉   ð3Þ   magnetic ordering and is expected to occur in           Figs. S1 to S3
                             k ≤ j ≤ k þl                  one-dimensional quantum gases in the presence           Table S1
                                                           of longer ranged interactions, which could be           References (32–42)
    First, we notice that OP (l ) for l = 1 is equal       realized in our experiment by using Rydberg             2 June 2011; accepted 15 August 2011
to the two-site correlation function C(d = 1).             atoms (31).                                             10.1126/science.1209284

                                             SCIENCE        VOL 334       14 OCTOBER 2011                                                                      203
                                                                                                                                  ly reduces the crystal lattice parameters as de-
      Nanoparticle Superlattice                                                                                                   termined with in situ SAXS measurements.
                                                                                                                                      In a typical experiment, we assembled DNA-
      Engineering with DNA                                                                                                        NP superlattices using oligonucleotide linker
                                                                                                                                  strands that, upon binding to a DNA-NP, present
      Robert J. Macfarlane,1,2 Byeongdu Lee,3 Matthew R. Jones,2,4 Nadine Harris,1,2                                              a short, single-stranded DNA “sticky end” at a
      George C. Schatz,1,2 Chad A. Mirkin1,2,3*                                                                                   controllable distance from the nanoparticle sur-
                                                                                                                                  face (19) (fig. S1 and table S1). This distance
      A current limitation in nanoparticle superlattice engineering is that the identities of the                                 dictates the interparticle spacing in a program-
      particles being assembled often determine the structures that can be synthesized. Therefore,                                mable manner (16). Because of the polyvalent
      specific crystallographic symmetries or lattice parameters can only be achieved using specific                              nature of the DNA-NPs, each NP hybridizes to
      nanoparticles as building blocks (and vice versa). We present six design rules that can be                                  multiple linker strands and subsequently forms
      used to deliberately prepare nine distinct colloidal crystal structures, with control over lattice                          tens to hundreds of sticky end duplexes to adja-
      parameters on the 25- to 150-nanometer length scale. These design rules outline a strategy to                               cent NPs, enabling the construction of lattices that
      independently adjust each of the relevant crystallographic parameters, including particle size                              are indefinitely stable under ambient conditions.
      (5 to 60 nanometers), periodicity, and interparticle distance. As such, this work represents                                However, because individual sticky end connec-
      an advance in synthesizing tailorable macroscale architectures comprising nanoscale materials                               tions are weak (a single sticky end duplex is not
      in a predictable fashion.                                                                                                   stable on its own at room temperature) and there-
                                                                                                                                  fore transient, upon thermal annealing, DNA-NPs
              he crystallographic lattice adopted by a                      Herein, we describe a set of rules for using          can shift positions within the material to ulti-

      T       given set of atomic and molecular com-
              ponents is often difficult to predict and
      control and is dependent on a large number of
                                                                       programmable oligonucleotide interactions, ele-
                                                                       ments of both thermodynamic and kinetic con-
                                                                       trol, and an understanding of the dominant forces
                                                                                                                                  mately form ordered lattices (15). Although all
                                                                                                                                  of the structures we describe are made with gold
                                                                                                                                  NPs, the assembly process should also be ap-
      factors. For ionic solids, Pauling developed rules               that are responsible for particle assembly to de-          plicable to any other NP that can be densely func-
      that explain the relative stabilities of different               sign and deliberately make a wide variety of crystal       tionalized with oligonucleotides.
      lattices of simple salts, but these rules do not                 types. Like the rules for atomic lattices developed            We determined structural characteristics for a
      allow for structure control (1). This is because                 by Pauling, these are guidelines for determining           total of 41 crystals that adopted one of nine crys-
      parameters such as size and charge for atoms                     relative nanoparticle superlattice stability, rather       tal lattices. In addition to fcc and bcc structures,
      (and small molecules) are not tunable; chang-                    than rigorous mathematical descriptions. How-              we also prepared the following lattices (19) (figs.
      ing an atom’s size or charge inherently changes                  ever, unlike Pauling’s rules, the set of rules below       S2 to S21 and S29 to S31): hexagonal close-
      the electronic properties that affect relative lat-              can be used not only to predict crystal stability          packed (hcp); AB, isostructural with cesium
      tice stability. In principle, nanoparticle-based                 but also to deliberately and independently control         chloride (CsCl); AB2, isostructural with alumi-
      superlattice materials should allow for more                     the nanoparticle sizes, interparticle spacings, and        num diboride; AB3, isostructural with Cr3Si;
      control over the types of crystal lattice that they              crystallographic symmetries of a superlattice (Fig.        AB6, isostructural with the alkali-fullerene com-
      adopt, given that one can tune multiple varia-                   1A). This methodology represents a major advance           plex Cs6C60; AB, isostructural with sodium
      bles (such as nanoparticle size or the presence of               toward nanoparticle superlattice engineering, as           chloride (NaCl); and simple cubic (sc). For each
      different organic molecule layers on the nano-                   it effectively separates the identity of a particle core   structure, we could tune lattice parameters by
      particle surface) to control superlattice stability              (and thereby its physical properties) from the var-        means of independent modifications to both oli-
      (2–14). Although advances have been made using                   iables that control its assembly.                          gonucleotide interconnect length and nanoparticle
      a variety of electrostatic forces (7–9), covalent                     We used polyvalent conjugates of DNA and              size. Rather than discuss each group of structures
      and noncovalent molecular interactions (6, 11),                  gold nanoparticles (DNA-NPs) as the basic build-           in turn (19), we describe a set of rules that consti-
      and biologically driven assembly strategies                      ing blocks for assembling superlattices, for               tute a design strategy for synthesizing a particular
      (2–5, 12), predictable architectural control re-                 which programmable recognition and hybridiza-              choice of one of the nine distinct crystallographic
      mains an elusive goal, regardless of the type of                 tion interactions between DNA strands drive the            symmetries.
      particle interconnect strategy chosen. In 1996,                  assembly process (Fig. 1B). The key hypothesis                 Rule 1: When all DNA-NPs in a system possess
      the use of oligonucleotides as particle-directing                in this work is that the maximization of DNA               equal hydrodynamic radii, each NP in the ther-
      motifs to synthesize amorphous polymeric ma-                     hybridization events between adjacent particles            modynamic product will maximize the number
      terials from polyvalent particles modified with                  is a more important factor in determining lat-             of nearest neighbors to which it can form DNA
      nucleic acids was demonstrated (2). Subse-                       tice stability than all other forces in the system.        connections. This occurs because maximizing
      quent work showed that crystallization and                       Synthetically controllable variations in nucleo-           the number of nearest neighbors in these sys-
      lattice control were possible for face-centered                  tide sequence allowed us to change the overall             tems in turn maximizes the number of poten-
      cubic (fcc) and body-centered cubic (bcc) crys-                  hydrodynamic size and coordination environ-                tial DNA connections between nanoparticles,
      tal structures simply by taking advantage of the                 ment (and thus the hybridization behavior) of the          which we have hypothesized to be the driving
      programmable nature of DNA (both in base                         particles, without the need to alter the structure of      force in forming ordered crystals. When using
      sequence and in overall oligonucleotide length)                  the inorganic nanoparticle core (2–5, 15–18). We           linkers with self-complementary sticky ends,
      (3–5, 15–18).                                                    used synchrotron-based small-angle x-ray scat-             where all particles can bind to all other parti-
                                                                       tering (SAXS) to characterize all lattices reported        cles in solution, the observed thermodynamic
       Department of Chemistry, Northwestern University, Evanston,     herein, because it allows for in situ analysis of          product is always an fcc lattice (Fig. 1C), a
      IL 60208, USA. 2International Institute for Nanotechnology,      highly solvated materials. We also have devel-             conclusion supported by theory (3). When two
      Northwestern University, Evanston, IL 60208, USA. 3X-ray Sci-    oped a complementary method to embed these                 sets of nanoparticles are functionalized with
      ence Division, Advanced Photon Source, Argonne National          superlattices in a resin, which enables their char-        linkers that contain different but complemen-
      Laboratory, Argonne, IL 60439, USA. 4Department of Materials
      Science and Engineering, Northwestern University, Evanston, IL   acterization by transmission electron microscopy           tary sticky ends, particles can only bind to par-
      60208, USA.                                                      (TEM) (17). However, we note that the embed-               ticles of the opposite type. A bcc lattice is
      *To whom correspondence should be addressed. E-mail:             ding process results in a slight deformation and           therefore the most stable for these binary sys-                                        disordering of the lattices, and that it significant-      tems (Fig. 1D), rather than an fcc lattice, as

204                                                   14 OCTOBER 2011               VOL 334        SCIENCE
each NP in a bcc lattice possesses more nearest               fcc lattices, and thus any hcp crystals observed             particle) and annealing at 25° to 30°C, one can
neighbors of the opposite particle type. Note                 would likely be kinetic products (20). Indeed, we            preferentially stabilize the growth of initial hcp-
that this rule holds for a wide range of nano-                have observed hcp lattices in these systems, but             like lattices that form during early time points
particle diameters and oligonucleotide lengths,               only as metastable structures that reorganize into           of the assembly process (15). It is important to
and it can therefore be used to make many fcc                 fcc lattices upon annealing (15). Stable hcp lat-            note that although this process can consistently
and bcc lattices with well-defined and predict-               tices can be realized by annealing at lower so-              be used to produce large (>1 mm) hcp lattices
able lattice parameters over the 25- to 150-nm                lution temperatures and decreasing the local DNA             that are stable for extended periods of time (sev-
range (figs. S2 and S3).                                      density around a NP surface (Fig. 1E). These two             eral weeks after formation), these structures are
   Rule 2: When two lattices are of similar                   changes both slow the DNA linker sticky end re-              still kinetic products. Annealing hcp lattices at
stability, the kinetic product can be produced by             lease and rehybridization rates necessary for crys-          higher temperatures for several hours always re-
slowing the rate at which individual DNA linkers              tallization, and promote lattice growth over lattice         sults in the lattices reorganizing to an fcc structure
dehybridize and subsequently rehybridize. For                 reorganization, thereby stabilizing initial kinetic          (fig. S23).
example, theoretical predictions show that, al-               products. For example, by using long DNA strands                  Rule 3: The overall hydrodynamic radius of a
though they possess the same number of nearest                (~30 nm) and NPs bearing a small number of                   DNA-NP, rather than the sizes of its individual
neighbors, hcp lattices are slightly less stable than         linkers (7.2-nm NPs, 20 T 3 DNA strands per                  NP or oligonucleotide components, dictates its

Fig. 1. (A) Nanoparticle superlattice engineering with DNA, unlike conventional             The superlattices reported herein are isostructural with (C) fcc, (D) bcc, (E) hcp, (F)
particle crystallization, allows for independent control of three important design          CsCl, (G) AlB2, (H) Cr3Si, and (I) Cs6C60 lattices. From left to right, each panel contains
parameters (particle size, lattice parameters, and crystallographic symmetry) by            a model unit cell (not to scale), 1D and 2D (inset) x-ray diffraction (SAXS) patterns,
separating the identity of the particle from the variables that control its assembly. (B)   and a TEM image of resin-embedded superlattices, along with the unit cell viewed
The DNA strands that assemble these nanoparticle superlattices consist of (i) an alkyl-     along the appropriate projection axis (inset). Lines in the model denote edges of the
thiol moiety and 10-base nonbinding region, (ii) a recognition sequence that binds to       unit cell; individual DNA connections are omitted for clarity. SAXS data are plots of
a DNA linker, (iii) a spacer sequence of programmable length to control interparticle       nanoparticle superlattice structure factor S(q) (y axis, arbitrary units) versus scattering
distances, and (iv) a “sticky end” sequence that drives nanoparticle assembly via DNA       vector q (x axis, Å−1). Black traces are experimental data; blue traces are modeled
hybridization interactions. Although only a single linkage is shown schematically           SAXS patterns for perfect lattices. All scale bars in the TEM images are 50 nm. See
here, DNA-NPs typically contain tens to hundreds of DNA linkers per particle. (C to I)      (19) for a complete list of particle sizes and lattice parameters.

                                                 SCIENCE         VOL 334         14 OCTOBER 2011                                                              205
      assembly and packing behavior. An important
      aspect of DNA-NP design is that the overall hy-
      drodynamic radius of a DNA-NP is a combina-
      tion of the NP diameter and the DNA length. As
      each of these parameters is independently con-
      trollable, one can easily synthesize two DNA-
      NPs with the same overall hydrodynamic radius
      but different NP core sizes (Fig. 2A). Thus, we
      could assemble NPs into three-dimensional (3D)
      structures with lattice parameters and interparticle
      distances that are not dictated solely by the sizes
      of the inorganic particle cores.
           This rule is well illustrated by the synthesis of
      CsCl lattices (Fig. 1F), which exhibit the same
      DNA-NP arrangement and connectivity as a bcc
      lattice but use two different NP core sizes. To
      create a range of CsCl lattices, we systematically
      changed the lengths of oligonucleotide linkers to
      obtain DNA-NPs with the appropriate hydro-
      dynamic radii (Fig. 2B). Note that by simply
      changing the length of the oligonucleotide link-
      ers, the inorganic particle radius and interpar-
      ticle distance were independently programmed
      for nanoparticles ranging from 5 to 60 nm in
      diameter, with lattice parameters ranging from
      ~ 40 to ~140 nm. The inorganic NP core sizes in
      these lattices differed by as much as 30 nm and
      still exhibited equivalent packing and assembly
      behavior.                                                Fig. 2. (A) Two particles with the same hydrodynamic radius exhibit the same assembly behavior,
           Rule 4: In a binary system, the size ratio and      regardless of the sizes of the inorganic nanoparticle cores. (B) SAXS patterns for CsCl lattices in binary
      DNA linker ratio between two particles dictate           systems where two particles have the same hydrodynamic radii but different inorganic core sizes. The
      the thermodynamically favored crystal structure.         inset and model show the relative sizes of the nanoparticles, DNA linkers, and assembled lattices, all
      For this rule, the “size ratio” is defined as the        drawn to scale. From top to bottom, the nanoparticle sizes are 60 and 40 nm, 40 and 20 nm, and 30
      ratio of the DNA-NPs’ hydrodynamic radii (a              and 10 nm. (C) SAXS patterns for AlB2 lattices, demonstrating that crystallographic symmetry and
      sum of the inorganic particle radius and DNA             lattice parameters can be controlled independently of the sizes and size ratios of the inorganic
                                                               nanoparticle cores (inset and model, both drawn to scale). From top to bottom, the inorganic core sizes
      linker length), and the DNA linker ratio is the
                                                               of the “big” and “small” nanoparticles (as defined by their overall hydrodynamic radii) are 10 and 10 nm,
      ratio of the number of DNA linkers on the two
                                                               20 and 10 nm, and 5 and 10 nm. See (19) for exact interparticle distances and lattice parameters for
      different types of DNA-NPs. Size ratio can be            all structures.
      predicted to affect the stability of different crystal
      symmetries because it determines the packing
      parameters of DNA-NPs within a lattice (i.e., the        and S17 to S19). By applying rule 3, one can          namic product. Consequently, the application of
      number and positions of adjacent particles to            tune the hydrodynamic radii of particles (and         rule 5 as a guiding principle in superlattice as-
      which a given DNA-NP can bind). The DNA                  thus the hydrodynamic size ratio) to position par-    sembly enables a large number of lattices to be
      linker ratio can also be expected to affect crystal      ticles into a specific crystallographic symmetry      synthesized without necessitating a complete re-
      stability, as it determines the number of DNA            without being restricted to specific inorganic        analysis of the forces involved in assembly for
      sticky ends available to form DNA connections            particle sizes or even to specific inorganic par-     each specific nanoparticle size or DNA length.
      with these adjacent particles. For example, by ad-       ticle size ratios. Indeed, the hydrodynamic radii     Further, this result implies that one could con-
      justing the size ratio of the DNA-NP components,         of the particles can even be tuned such that in       struct a phase diagram that would predict the
      lattices isostructural with AlB2 can be obtained         a given system, the DNA-NP with the larger            most stable crystal structure as a function of
      (Fig. 1G; size ratio 0.64). By varying both the          inorganic core size possesses the smaller hydro-      these two variables. As previously mentioned, the
      size ratio and the DNA linker ratio, lattices iso-       dynamic radius. In this way, one can position         main hypothesis of this work states that the ther-
      structural with Cr3Si can be made (Fig. 1H; size         a given nanoparticle at any of the occupied           modynamic products in this assembly method
      ratio 0.37, DNA linker ratio ~2)—an unusual              Wyckoff positions within a given lattice type’s       are the ones that maximize DNA duplex forma-
      example of a NP superlattice with this lower             unit cell, regardless of the inorganic particle’s     tion. However, experimental verification of this
      crystallographic symmetry. Finally, by using a           size (Fig. 2C).                                       hypothesis (and thus the development of a phase
      DNA linker ratio of ~3, we synthesized a lattice             Rule 5: Two systems with the same size ratio      diagram) is challenging, as it is difficult to ex-
      that has no mineral equivalent but is isostructural      and DNA linker ratio exhibit the same thermo-         perimentally determine the number of DNA
      with the alkali-fullerene complex Cs6C60 (21) (Fig.      dynamic product. Note that crystal stability is       duplexes formed in a given lattice. Therefore,
      1I; size ratio ~0.35).                                   determined by the ratio of the two variables dis-     we have constructed a model that is based on
           Note that the lattices in Fig. 1 are only indi-     cussed in rule 4, not their absolute values. A        the predictable and well-established properties
      vidual examples of the many AlB2, Cr3Si, and             comparison of the lattices created with rule 4        of both DNA (persistence length, rise per base
      Cs6C60 crystals synthesized with this method.            shows that, regardless of the absolute values of      pair) (22) and DNA-NPs (number of DNA strands
      These structures also have been constructed using        DNA-NP size or the number of DNA linkers per          per particle, the hybridization behavior of sticky
      multiple particle sizes (5 to 30 nm) and hy-             particle, two systems with the same size ratios       ends) (16) and used this model to calculate rela-
      drodynamic radii (10 to 50 nm) (figs. S6 to S8           and DNA linker ratios form the same thermody-         tive crystal stabilities.

206                                            14 OCTOBER 2011            VOL 334       SCIENCE
Fig. 3. (A) Surface plot
of modeled data, in
which the percentage of
DNA sticky ends that form
duplexes (z axis) is calcu-
lated for different crystal-
lographic arrangements
as a function of exper-
imentally controllable de-
sign parameters (DNA
linker ratio, x axis; DNA-
NP size ratio, y axis). (B)
Phase diagram constructed as a top-down view of (A), where each dot on the            This plot demonstrates the relative stability of both lattices that have been
graph represents a lattice that was synthesized experimentally. The color of each     constructed with DNA-programmed assembly (color traces) and other lattices that
experimental data point denotes the identity of the lattice obtained. (C) Two-        have been theoretically predicted or synthesized using other assembly method-
dimensional “slice” through the plot in (B), at a constant DNA linker ratio of 1.0.   ologies (black traces). The inset indicates where this slice was taken from (B).

Fig. 4. (A) More com-                                                                                            spheres should possess the greatest number of
plex nanoparticle assem-                                                                                         DNA duplexes and therefore should be the
blies can be created when                                                                                        most stable phase for a given set of variables.
programming multivalent                                                                                          Although the CCM is not intended to provide
DNA-NP interactions. For                                                                                         an explicit solution for determining the most
example, by encoding mul-                                                                                        stable crystal structure for a given set of design
tiple distinct sticky end se-                                                                                    parameters, it should provide a suitable means
quences on the same particle,                                                                                    to test both rule 5 and the hypothesis that max-
both self-complementary                                                                                          imization of DNA hybridization is the driving
andnon–self-complementary                                                                                        force for forming ordered crystals.
interactions can be used
                                                                                                                     A comparison of the modeled phase diagram
to assemble lattices. (B
                                                                                                                 to experimentally obtained data shows that the
and C) This strategy can
be used to create a NaCl                                                                                         model correctly predicts the structures obtained
lattice (B) when using                                                                                           for a wide range of DNA-NP size ratios and
two particles with differ-                                                                                       DNA linker ratios, confirming the predominant
ent inorganic core sizes,                                                                                        hypothesis of this work as well as rules 4 and 5
or a simple cubic lattice                                                                                        (Fig. 3B). The model was also used to confirm
(C) when using two par-                                                                                          that the lattices obtained experimentally are more
ticles with the same in-                                                                                         stable than a number of other structures that have
organic core size. From                                                                                          been predicted by previous theoretical calcula-
left to right, each panel                                                                                        tions or that have been assembled with other
shows a model unit cell,                                                                                         methodologies (Fig. 3C) (8, 23). Although there
1D and 2D (inset) SAXS                                                                                           are limitations to the predictive nature of the
data, and a TEM image                                                                                            CCM as it currently is constructed (19) (figs. S24
with the unit cell viewed                                                                                        to S28), the vast majority of the data generated by
along the appropriate                                                                                            the model are in complete agreement with the
projection axis (inset). In (B), the SAXS data correspond to a NaCl lattice with 15-nm and 10-nm AuNPs           synthesized lattices. Given that all experimentally
and the TEM image is of a NaCl lattice with 30-nm and 15-nm AuNPs. In (C), the SAXS data correspond to           generated data points validate the six rules de-
a simple cubic lattice with 10-nm AuNPs and the TEM image shows a simple cubic lattice with 15-nm                veloped in this work, it is reasonable to assume
AuNPs. Scale bars, 50 nm.                                                                                        that simplifications used to develop the CCM are
                                                                                                                 the result of this discrepancy. Nonetheless, the
    The foundations for this model (hereafter            duplexes being formed. By using the physical            strong agreement between experiment and the-
referred to as the complementary contact model,          characteristics of the DNA-NP building blocks           ory demonstrates that the CCM should provide
or CCM) are the assumptions that (i) DNA linker          mentioned above, one can design a model lattice         a solid basis for future computational work in
sticky ends must be able to physically contact           of arbitrary symmetry that has the appropriate          this area. As a result, the control over experi-
one another to hybridize, and (ii) any sticky ends       lattice parameters (as dictated by a given set of       mental design parameters (hydrodynamic size
that can come into contact will eventually form a        particle sizes and DNA lengths), and then use           ratios, inorganic particle radii, and DNA lengths)
DNA duplex. The DNA linker strands on the                the CCM to determine how many complemen-                afforded by this DNA-based assembly method
surface of a DNA-NP are dynamic and can there-           tary sticky ends are able to contact one another        and coupled with the predictive nature of this
fore be treated as a single collective entity (16).      and subsequently hybridize. This process enables        phase diagram, allows one to determine the ex-
This allows one to represent a DNA-NP as a               the prediction of the number of DNA duplexes            perimental variables necessary to create a diverse
“fuzzy sphere” rather than a particle with a dis-        in a given crystal structure as a function of size      array of lattices a priori, with independent control
crete set of DNA linkers. Because DNA linker             ratio and DNA linker ratio (Fig. 3A) (19). If the       over crystallographic symmetry, lattice parame-
sticky ends must physically contact one another          main hypothesis of this work is correct, a larger       ters, and nanoparticle sizes (figs. S2 to S10 and
to form a DNA duplex, it is therefore assumed            number of DNA duplexes formed for a given               S13 to S21).
that a greater amount of surface contact between         crystal structure should correlate to a more                Rule 6: The most stable crystal structure will
adjacent spheres that contain complementary              stable lattice. Thus, the lattice with the most sur-    maximize all possible types of DNA sequence–
sticky ends correlates to a larger number of DNA         face contact between adjacent complementary             specific hybridization interactions. The examples

                                            SCIENCE       VOL 334        14 OCTOBER 2011                                                     207
      above examine relatively simple binary systems,             References and Notes                                           25. K. L. Kelly, E. Coronado, L. L. Zhao, G. C. Schatz, J. Phys.
      where only a single type of DNA sticky end du-           1. L. Pauling, The Nature of the Chemical Bond (Cornell               Chem. B 107, 668 (2002).
                                                                  Univ. Press, Ithaca, NY, ed. 3, 1960).                         26. J. A. Fan et al., Science 328, 1135 (2010).
      plex is created. However, because of the polyvalent      2. C. A. Mirkin, R. L. Letsinger, R. C. Mucic, J. J. Storhoff,    27. K. J. Stebe, E. Lewandowski, M. Ghosh, Science 325, 159
      nature of the DNA-NPs and the base sequence                 Nature 382, 607 (1996).                                            (2009).
      programmability of DNA, one is not necessarily           3. S.-J. Park, A. A. Lazarides, J. J. Storhoff, L. Pesce,         28. A. T. Bell, Science 299, 1688 (2003).
      restricted to a single type of favorable particle in-       C. A. Mirkin, J. Phys. Chem. B 108, 12375 (2004).              29. J. Grunes, J. Zhu, E. A. Anderson, G. A. Somorjai, J. Phys.
                                                               4. S. Y. Park et al., Nature 451, 553 (2008).                         Chem. B 106, 11463 (2002).
      teraction in a given lattice. By cofunctionalizing a     5. D. Nykypanchuk, M. M. Maye, D. van der Lelie, O. Gang,         Acknowledgments: Supported by the Defense Research
      nanoparticle with different linkers that contain dif-       Nature 451, 549 (2008).                                            & Engineering Multidisciplinary University Research
      ferent base sequences, multiple sequence-specific        6. C. J. Kiely, J. Fink, M. Brust, D. Bethell, D. J. Schiffrin,       Initiative of the Air Force Office of Scientific Research and
      DNA duplex interactions are possible (Fig. 4A).             Nature 396, 444 (1998).                                            by the U.S. Department of Energy Office of Basic
                                                               7. A. M. Kalsin et al., Science 312, 420 (2006); 10.1126/             Energy Sciences [award DE-SC0000989; Northwestern
      This is an inherent distinction and potential ad-                                                                              University (NU) Non-equilibrium Energy Research Center]
      vantage of using a sequence-programmable linker          8. E. V. Shevchenko, D. V. Talapin, N. A. Kotov, S. O’Brien,          (C.A.M. and G.C.S.); a National Security Science and
      such as DNA, as opposed to entropy- or charge-              C. B. Murray, Nature 439, 55 (2006).                               Engineering Faculty Fellowship from the U.S. Department
      dominated assembly processes.                            9. S. Srivastava et al., Science 327, 1355 (2010); 10.1126/           of Defense (C.A.M.); a NU Ryan Fellowship (R.J.M.); and a
                                                                  science.1177218.                                                   NU Ryan Fellowship and a NSF Graduate Research
           This rule was tested by cofunctionalizing a
                                                              10. S. Wong, V. Kitaev, G. A. Ozin, J. Am. Chem. Soc. 125,             Fellowship (M.R.J.). Portions of this work were carried out
      nanoparticle with two different linkers: one that           15589 (2003).                                                      at the DuPont-Northwestern-Dow Collaborative Access
      bore a self-complementary sticky end, and one           11. Y. Zhao et al., Nat. Mater. 8, 979 (2009).                         Team (DND-CAT) beamline located at Sector 5 of the
      that bore a sticky end sequence complementary           12. C.-L. Chen, N. L. Rosi, Angew. Chem. Int. Ed. 49, 1924             Advanced Photon Source (APS). DND-CAT is supported
      to the sticky ends of a second particle. In this sys-       (2010).                                                            by E. I. DuPont de Nemours & Co., Dow Chemical
                                                              13. Z. Nie, A. Petukhova, E. Kumacheva, Nat. Nanotechnol.              Company, and the state of Illinois. Use of the APS
      tem, the cofunctionalized particle (blue particle,          5, 15 (2010).                                                      was supported by the U.S. Department of Energy,
      Fig. 4A) exhibited an attractive force with respect     14. M. R. Jones, K. D. Osberg, R. J. Macfarlane, M. R. Langille,       Office of Science, Office of Basic Energy Sciences,
      to all particles encountered in the system, where-          C. A. Mirkin, Chem. Rev. 111, 3736 (2011).                         under contract DE-AC02-06CH11357. The transmission
      as the second particle (red particle, Fig. 4A) was      15. R. J. Macfarlane et al., Proc. Natl. Acad. Sci. U.S.A. 106,        electron microscope work was carried out in the EPIC
                                                                  10493 (2009).                                                      facility of the NU Atomic and Nanoscale Characterization
      only attracted to the first particle type. When the     16. R. J. Macfarlane et al., Agnew. Chem. Int. Ed. 49, 4589            Experimental Center, which is supported by NSF-NSEC,
      hydrodynamic radius size ratio of the two NPs               (2010).                                                            NSF-MRSEC, Keck Foundation, the state of Illinois,
      was ~0.3 to 0.4, the sticky ends were presented at      17. M. R. Jones et al., Nat. Mat. 9, 913 (2010).                       and NU. Ultrathin sectioning was carried out at the
      the correct distances from the particle surface to      18. H. Xiong, D. van der Lelie, O. Gang, Phys. Rev. Lett. 102,         NU Biological Imaging Facility, supported by the
                                                                  015504 (2009).                                                     NU Office for Research.
      form a NaCl lattice (Fig. 4B); that is, the self-
                                                              19. See supporting material on Science online.
      complementary and non–self-complementary                20. L. V. Woodcock et al., Nature 385, 141 (1997).                 Supporting Online Material
      linkers were both at a position to form duplexes        21. O. Zhou et al., Nature 351, 462 (1991).              
                                                                                                                                 Materials and Methods
      in this crystallographic arrangement. Furthermore,      22. V. A. Bloomfield, D. M. Crothers, I. Tinoco, Nucleic Acids:
                                                                                                                                 SOM Text
      when the inorganic core sizes were the same on              Structures, Properties, and Functions (University Science
                                                                  Books, Sausalito, CA, 2000).                                   Figs. S1 to S31
      both DNA-NPs, the particles formed a simple             23. A. V. Tkachenko, Phys. Rev. Lett. 89, 148303                   Tables S1 and S2
      cubic lattice, as defined by the positions of the           (2002).                                                        References (30–44)
      inorganic cores (Fig. 4C). Although NaCl and            24. M. I. Bodnarchuk, M. V. Kovalenko, W. Heiss, D. V.             29 June 2011; accepted 25 August 2011
      simple cubic structures are presented as the first          Talapin, J. Am. Chem. Soc. 132, 11967 (2011).                  10.1126/science.1210493
      examples of this multivalent strategy, one can
      envision even more sophisticated and complex
      systems (such as lattices with three or more
      nanoparticle components) using multiple DNA-
      programmed NP interactions.
                                                              Conical Intersection Dynamics in
           We have presented a set of basic design rules
      for synthesizing a diverse array of nanopar-
                                                              NO2 Probed by Homodyne
      ticle superlattices using DNA as a synthetically
      programmable linker. These rules provide ac-            High-Harmonic Spectroscopy
      cess to an easily tailorable, multifaceted design
      space in which one can independently dictate            H. J. Wörner,1,2* J. B. Bertrand,1 B. Fabre,3 J. Higuet,3 H. Ruf,3 A. Dubrouil,3
      the crystallographic symmetry, lattice parameters,      S. Patchkovskii,1 M. Spanner,1 Y. Mairesse,3 V. Blanchet,4 E. Mével,3 E. Constant,3
      and particle sizes within a lattice. This in turn       P. B. Corkum,1 D. M. Villeneuve1
      enables the synthesis of many different nano-
      particle superlattices that cannot be achieved          Conical intersections play a crucial role in the chemistry of most polyatomic molecules, ranging
      through other methodologies. Indeed, super-             from the simplest bimolecular reactions to the photostability of DNA. The real-time study of
      lattices that do not follow the well-known hard-        the associated electronic dynamics poses a major challenge to the latest techniques of ultrafast
      sphere packing parameter rules defined by               measurement. We show that high-harmonic spectroscopy reveals oscillations in the electronic
      Schiffrin and co-workers (6) and Murray and             character that occur in nitrogen dioxide when a photoexcited wave packet crosses a conical
      co-workers (8, 24) can easily be assembled as           intersection. At longer delays, we observe the onset of statistical dissociation dynamics. The
      thermodynamically stable structures over a range        present results demonstrate that high-harmonic spectroscopy could become a powerful tool to
      of nanoparticle sizes and lattice parameters. The       highlight electronic dynamics occurring along nonadiabatic chemical reaction pathways.
      understanding gained from the use of these
      rules will both inform and enable future assem-                he outcome of chemical reactions is de-                     in probing valence electron dynamics include
      bly efforts, allowing for the construction of new
      crystallographic arrangements that have emer-
      gent properties for use in the fields of plasmonics
                                                              T      termined by the valence electronic struc-
                                                                     ture of molecules. Therefore, the elucidation
                                                              of elementary reaction mechanisms requires an
                                                                                                                                 attosecond transient absorption (1), extreme ul-
                                                                                                                                 traviolet photoelectron spectroscopy (XUV-PES)
                                                                                                                                 (2), high-order harmonic spectroscopy (HHS)
      (14, 25, 26), photonics (27), catalysis (28, 29),       understanding of the valence electron dynamics.                    (3–5) and strong-field ionization (6). Both time-
      and potentially many others.                            Recently developed techniques that are efficient                   resolved PES (7) and time-resolved HHS are

208                                            14 OCTOBER 2011              VOL 334          SCIENCE  
sensitive to valence electron dynamics through
the molecular photoionization matrix elements.
     Electronic dynamics in molecules are par-
ticularly challenging to observe when they are
strongly coupled to nuclear dynamics. Such situ-
ations often arise in polyatomic molecules where
conical intersections between the potential en-
ergy surfaces induce very rapid radiationless tran-
sitions at particular nuclear configurations (see
inset of Fig. 1) (8, 9). These features channel
electronic excitation into atomic motion in such
diverse contexts as the primary steps of vision
(10) and the dynamics underlying electron trans-
fer and the photostability of DNA bases (11).
     Here, we show that high-harmonic spectros-
copy reveals the variations in electronic charac-
ter during the conical intersection dynamics and
the subsequent dissociation of nitrogen dioxide
(NO2). We chose NO2, a radical, because of its
                                                                                                                                                      ∼                 ∼
model status for theories of unimolecular disso-                      Fig. 1. Schematic representation of the potential energy surfaces of the ground X 2A1 and excited A 2B2
ciation (12–14) and conical intersection dynam-                       electronic states of NO2. The dominant electronic configuration in the two highest-lying molecular
ics (15–19). Our results translate the previously                     orbitals is shown for each state on the left. The orbitals are represented by isoamplitude surfaces of
recognized sensitivity of HHS to electronic struc-                    the wave function with color-coding of the sign. After excitation by a 400-nm pump pulse (blue
ture into a tool for elucidating chemical reaction                    arrow), the wave packet initially moves along the bending coordinate, crosses the conical intersection
dynamics.                                                             (shown in the top left inset) several times during the first 100 fs, and spreads along the asymmetric-
     High-harmonic spectroscopy can be factored                       stretch coordinate. Wave-packet population that has returned to the ground electronic state and
into three steps: removal of an electron by an                        possesses an energy above 3.1155 eV (green dashed line) dissociates on the picosecond time scale
intense femtosecond laser field, acceleration of                      (dashed arrow).
the electron in the laser field, and photorecombi-
nation (20, 21). Each step contributes an ampli-
tude and a phase to the emitted XUV radiation
(22–24, 20, 25). The measurement relies on a
coherent detection scheme in a transient grating
geometry, using unexcited molecules as a local
oscillator (4, 5). It is thus sensitive to both ampli-
tude and phase of the photorecombination matrix
elements, a quantity that has recently attracted
a lot of interest (26, 27). Time-resolved HHS is
thus related to time-resolved PES but differs in
its sensitivity to the continua associated with dif-
ferent ionic states. PES projects the molecular
wave packet onto a set of ionic states, influenced
by resonances, Franck-Condon factors, and dis-
sociative ionization. HHS involves recombination
from an energetic continuum electron with one
or a few of the lowest ionic states that were
selected by tunneling ionization.
     A schematic representation of the potential
energy surfaces of NO2 is shown in Fig. 1. In the
X 2A1 electronic ground state, NO2 possesses a
bent equilibrium geometry and the dominant
electronic configuration in the two highest occu-
pied orbitals is (b2)2(a1)1. Single-photon absorp-
tion at 400 nm excites the molecule to the A 2B2
 Joint Laboratory for Attosecond Science, National Research
Council of Canada and University of Ottawa, 100 Sussex Drive,
Ottawa, Ontario, Canada K1A 0R6. 2Laboratorium für Physikalische
Chemie, Eidgenössische Technische Hochschule Zürich, Wolfgang-        Fig. 2. Experimental setup for high-harmonic transient grating spectroscopy as first described in (4).
Pauli-Strasse 10, 8093 Zürich, Switzerland. 3Centre Lasers Intenses   The transient grating creates a spatially modulated population of excited molecules accompanied by
et Applications, Université de Bordeaux, CEA, CNRS, UMR5107,
                                                                      a depletion of the unexcited molecules. The periodic structure results in a modulation of amplitude
351 Cours de la Libération, 33405 Talence, France. 4Laboratoire
Collisions Agrégats Réactivité (IRSAMC), UPS, Université de           and phase of the XUV emission in the near field that leads to first-order diffraction in the far field. An
Toulouse, F-31062 Toulouse, France and CNRS, UMR 5589,                XUV grating disperses the radiation in one dimension while the beam freely diverges in the other
F-31062 Toulouse, France                                              dimension. With r(cos(kx) + 1) being the spatially modulated excitation fraction, the signal in m = 0 is
*To whom correspondence should be addressed. E-mail:                  given by Im=0 = j(1 − r)dg eifg + rde eife j2 and that in m = T 1 by Im=T1 = r4 jde eif e − d g eif g j2 ,                                             where the symbols are defined in the text relating to Eq. 1.

                                                       SCIENCE        VOL 334       14 OCTOBER 2011                                                    209
      state of dominant configuration (b2)1(a1)2. The        signature of any regular dynamics. Figure 3, C        place below threshold, the radiated XUV field
      A 2B2 excited state forms a conical intersection       and D, however, show exponential growth or de-        can be described as in Eq. 1
      with the ground state (see inset of Fig. 1). Wave-     cay on the picosecond time scale, following the
      packet calculations have shown that within the         step-like variations.                                       EXUV ðWÞ ¼ ð1 − rÞdg eifg þ rde eife       ð1Þ
      first femtoseconds after excitation, the nuclear           The combined information from Fig. 3 shows
      wave packet moves along the bending coordinate         that the step-like response to excitation below       where r is the spatially modulated fraction of ex-
      toward the conical intersection, where it can ei-      threshold characterizes electronic excitation with-   cited molecules, and dg, de and fg, fe are the high-
      ther cross the intersection and remain on the same     out dissociation (Fig. 3, A and B), whereas the       harmonic amplitudes and phases of the ground or
      diabatic surface or else stay in the upper cone        exponential variation of the signal in Fig. 3, C      excited molecular states, respectively. When the
      of the intersection and thus change the diabatic       and D shows the unimolecular decomposition            excitation frequency exceeds threshold, the ex-
      surface (gray arrows in Fig. 1) (15–18). After a       of NO2. To quantify these observations, we in-        cited molecules can undergo dissociation into
      few hundred femtoseconds, the nuclear wave             troduce a simple model. When excitation takes         NO(2P)+O(3P) that, together, emit harmonics
      packet returns to the electronic ground state
      through internal conversion. If the energy of the
      absorbed photon lies above the first dissociation      Table 1. Molecular parameters for strong-field ionization and high-harmonic generation, determined
      limit at 3.1155 eV (397.95 nm) (all quoted wave-       by fitting Eqs. 1 and 2 to the experimental data shown in Fig. 3, A to D. The excitation fraction r has
      lengths are vacuum values), the molecule dis-          been determined from the experimental parameters as described in the text.
      sociates into NO (X 2PΩ) and O (3PJ) on the
                                                                  Species          r     i/ig             H13                     H15                H17
      picosecond time scale (dotted arrow in Fig. 1).
                                                                                                d/dg     |f – fg|(rad)      d/dg |f – fg|(rad) d/dg |f – fg|(rad)
      Previous studies using laser-induced fluorescence
      have characterized the picosecond dissociation         NOÃ (407 nm)
                                                               2             0.15 4.5           3.0           2.03          2.3        2.02        1.2       2.24
      in detail (14, 28). However, the femtosecond           NOÃ (397 nm)
                                                               2             0.15 4.5           3.2           1.94          2.1        1.92        1.1       4.22
      conical intersection dynamics has been largely         NO + O (397 nm) 0.15 2.3           6.1           2.09          4.1        2.00        1.4       1.87
      obscured by competing multiphoton processes                                                      t = 2.71 T 0.15 ps
      (19), requiring elaborate coincidence detection
      methods (29).
          The experimental setup is illustrated in Fig. 2.   A                                                      C
      We excite NO2 in a transient grating formed from
      two synchronized 400-nm laser pulses and probe
      its dynamics by high-harmonic generation from
      an 800-nm, 32-fs laser pulse (4, 30). The ex-
      citation pulses are generated either in a 2-mm-
      thick b-barium borate (BBO) crystal, providing
      160-fs pulses of 1-nm spectral width tunable from
      395 to 407 nm or in a 100-mm-thick BBO crys-
      tal, giving 40-fs pulses of 5-nm spectral width.
      The combination of the transient grating with
      an XUV monochromator allows us to spectral-
      ly resolve the high harmonics (H11 to H21 in
      this experiment) and to measure both the undif-
      fracted (m = 0) and diffracted (m = T1) components
      of each harmonic order. The signal observed in
      m = 0 is equivalent to a measurement done in a
      collinear pump-probe geometry, whereas the dif-        B                                                      D
      fracted signal results from an interference be-
      tween equal populations of excited and unexcited
      molecules (4).
          We first discuss the picosecond photo-
      dissociation dynamics, which show that our
      measurement is dominated by single-photon ab-
      sorption. The measurements were done with the
      160-fs pulses, but we have obtained fully con-
      sistent results with the 40-fs pulses. The dynam-
      ics observed after excitation by pump pulses
      centered at 407 or 397.2 nm are shown in Fig. 3,
      A and C, respectively. Figure 3A shows a step-
      like decrease of the undiffracted XUV radiation        Fig. 3. High-harmonic and ion yields as a function of delay between two synchronized near-UV pump
      and a corresponding increase of the diffracted         pulses setting up a transient grating and an 800-nm probe pulse generating high harmonics in the
      intensity. The total ion yield, measured simulta-      excited sample. (A) The yield of diffracted (red dots) and undiffracted (blue dots) high-harmonic
      neously and shown in Fig. 3B, increases, whereas       signals, normalized to the undiffracted signal at negative pump-probe delays, for excitation by 407-nm
      the high-order harmonic signal decreases; this in-     pump pulses. The full lines represent the results of the theoretical model described in the text. (B) H13,
      dicates a destructive interference between har-        m = 0, and m = 1 from (A) together with the total ion yield measured in parallel to the experiment
      monics emitted by the ground state and those           (green dots) and the theoretical model (green line). (C and D) The same observables as (A and B), but
      emitted by the excited state (4, 5). In this scan,     for a pump pulse centered at 397.2 nm. Polarization of pump and probe are parallel. The typical pump
      performed with a 200-fs delay step, there is no        energy is 10 mJ to minimize multiphoton processes.

210                                           14 OCTOBER 2011           VOL 334        SCIENCE
with a resultant amplitude df and phase ff. The                                                                                                             time constant t = 2.71 T 0.15 ps, in agreement        with increasing harmonic order, as expected
radiated XUV field is then given by Eq. 2                                                                                                                   with the 2.78 ps measured previously at room          from the lower cutoff of the harmonic emission
                                                                                                                                                            temperature (14). The deep modulation of the          from NO. We thus conclude that the observed
    EXUV ðW,tÞ ¼ ð1 − rÞdg eifg þ re−t=t de eife þ                                                                                                          signals demonstrates that the probed dynamics         high-harmonic signal is dominated by single-
                                                                                                                                                            are dominated by one-photon absorption, which         photon excitation, in contrast to previous femto-
                                                              rð1 − e−t=t Þdf eiff                                                           ð2Þ
                                                                                                                                                            is in general difficult to achieve in femtosec-       second time-resolved measurements that reported
where t is the time elapsed since excitation                                                                                                                ond time-resolved measurements on molecules           oscillatory components of periods in the range
and t is the time constant of the unimolecular                                                                                                              because of multiphoton processes (19). The            of 500 to 850 fs (31–33). The latter were indeed
dissociation.                                                                                                                                               strong-field ionization probability of vibration-     attributed to multiphoton excitations to higher-
    To extract the relevant parameters from the                                                                                                             ally excited NO2 molecules in the electronic          lying electronic states that would not emit high
measurement, both the diffracted and undif-                                                                                                                 ground state is larger than that of unexcited         harmonics owing to their low binding energies.
fracted high-harmonic signals are normalized                                                                                                                molecules [vertical ionization potential (Ip) =            In the following, we exploit this property to
by the signal measured in the absence of exci-                                                                                                                                                                  investigate the hitherto unobserved femtosec-
tation (namely, |dg|2). We calculate the excited                                                                                                            11.2 eV] by a factor of iig ≈ 4:5. The ionization     ond dynamics of NO2 in the A 2B2 state. These
state fraction from the measured pulse energy,                                                                                                                                                                    measurements were done with 40-fs excitation
focal spot size, and the known absorption cross                                                                                                             rate of NO + O (dominated by NO because the           pulses centered at 401 nm. The experimental re-
section of NO2 and determine the unknown pa-                                                                                                                vertical Ip values are 9.2 and 13.8 eV, respective-   sults measured with cross-polarized laser pulses
rameters in a global nonlinear least-squares fit.                                                                                                           ly) exceeds that of the unexcited molecules by a      are shown in Fig. 4, A and B. Fig. 4A shows
The determined parameters are given in Table 1,                                                                                                             factor of ≈ 2.3. The relative high-harmonic am-       the undiffracted and diffracted signals measured
and the corresponding fit is shown as full lines in                                                                                                         plitudes are larger for the excited molecules         in harmonics 11 through 17 (blue and red dots,
Fig. 3, A to D. The total ion yield is represented                                                                                                          than for the unexcited molecules, especially for      respectively), and Fig. 4B shows the same quan-
by equations similar to Eqs. 1 and 2 but with                                                                                                               low harmonic orders, and the phase shift is           tities for H15 and H16 (green line). The latter
phases set to zero.                                                                                                                                         substantial, as expected from the observed de-        allows an accurate determination of the zero time
    The global fit to all high-harmonic orders                                                                                                              structive interference. The relative amplitudes       delay and the cross-correlation function (34).
and ion signals in Fig. 3, B and D, provides a                                                                                                              for the NO + O pair decrease particularly fast        Whereas the m = 0 order decreases smoothly
                                                                                                                                                                                                                  over the duration of the cross correlation, the
                                                                                                                                                                                                                  diffracted order (m = 1) increases and reaches a
                                   A                                                                                                               D                                                              maximum at a pump-probe delay of 35 fs. The
                                                                                                                                                                                                                  diffracted signal subsequently decreases and
                                          0. 8                                                                                                                                                                    reaches a minimum around 70 fs, followed by
                                          0. 6
                                                                                                                                                                                                                  another maximum at 130 fs. Further modulations
                                                                                                                                                                                                                  with decreasing contrast are observed at longer
                                          0. 4
                                                                                                                                                                                                                  pump-probe delays. As we show and discuss in
                                          0. 2
                                                                                                                                                                                                                  fig. S2 and the accompanying text, no modu-
                                               0                                                                                                                                                                  lations are observed in parallel polarization.
                                                                                                                                                                                                                       These oscillations, observed in the diffracted

                                                                                                                                                                                                                  XUV radiation, are a fingerprint of the electron-
                                                                                                                                                                                                                  ic dynamics of the molecule taking place around
                                                                         50    100                         150       200       250    300
                                                        −50    0
                                                                                                                                                                                                                  the conical intersection, as illustrated schemat-
                                                                                                                                                                                                                  ically in Fig. 4D. In the bright zones of the tran-
                                     1                                                                      0. 4                                                                                                  sient grating, the electronic character of the
                                                                              B                                                                   C                                                                                                       ∼
                                                                                                                                                                                                                  excited molecules oscillates between X 2A1 and
 normalized high−harmonic signal

                                                                                                           0. 35
                                   0. 8                                                                                                                                                                            ∼ 2                                   ∼ 2
                                                                                                            0. 3                                                                                                  A B2. When the molecule is in the X A1 state,
                                                                                     diabatic population

                                   0. 6                                                                    0. 25                                                                                                  the near-field variation of the high-harmonic
                                                                                                            0. 2                                                                                                  emission is much smaller for most molecular
                                   0. 4
                                                                                                           0. 15                                                                                                  geometries than when it is in the A 2B2 state,
                                                              H15, m=0
                                   0. 2
                                                              H15, m=1 (x5)                                 0. 1                                                                                                  which we detect as a variation of the intensity
                                                                                                           0. 05                                                                                                  of diffracted radiation [see section VI of the
                                           0            100        200        300
                                                                                                                           0          100   200       300
                                                                                                                                                                                                                  supporting online material (SOM)].
                                                                                                                                                                                                                       To rationalize these observations, we intro-
                                                                                                                                                                                                                  duce a simple model based on diabatic electronic
                                                                                                                                                                                                                  states and coordinate-independent transition
Fig. 4. (A) The high-harmonic yields after excitation by a pair of 10-mJ, 401-nm laser pulses in the                                                                                                              moments (more detailed calculations are given
undiffracted (blue) and diffracted orders (red) of the odd harmonics (H11 to H17). Experimental                                                                                                                   in the SOM). The total radiated XUV field is
data points appear as dots and a three-point-smoothed version as lines. (B) The undiffracted (blue)
                                                                                                                                                                                                                  the coherent sum of contributions from the un-
and diffracted orders of H15 (red) and H16 (green line). (C) The diabatic excited state population
from (17) convoluted with a 50-fs Gaussian cross-correlation function. (D) Illustration of how high-                                                                                                              excited molecules (subscript g) and excited mol-
                                                                                                                                                                                                                                                 ∼             ∼
harmonic transient-grating spectroscopy probes the electronic character of the molecule during the                                                                                                                ecules in the two diabatic A 2B2 and X 2A1
conical intersection dynamics. The top illustrates the spatial intensity structure of the transient                                                                                                               states (Eq. 3)
grating. The bottom shows schematically the lowest two potential energy surfaces of NO2. When the                                                                                                                         EXUV ðW,tÞ ¼ ½1 − rðtފdg eifg þ
molecule is in the excited diabatic state (represented as a blue wave packet), the high-harmonic
emission differs significantly from the unexcited (2A1) molecules, leading to a large variation of                                                                                                                                       rAðtÞd AeifA þ
                                                                                                                                                                                                                                          ∼     ∼
high-harmonic amplitude and phase across the transient grating and thus to strong diffraction.                                                                                                                                                      ∼
                                                                                                                                                                                                                                         rXðtÞd XeifX
                                                                                                                                                                                                                                          ∼     ∼                ð3Þ
When population is transferred into the ground diabatic state (red wave packet), the modulation
depth of high-harmonic amplitude and phase across the transient grating, and therefore the                                                                                                                        where rðtÞ ¼ rAðtÞ þ rXðtÞ is the total fraction
                                                                                                                                                                                                                                ∼       ∼
diffracted intensity, decreases.                                                                                                                                                                                  of excited molecules before dissociation takes

                                                                                                                                             SCIENCE        VOL 334       14 OCTOBER 2011                                                   211
      place. The intensity of the diffracted light is thus    diabatic state population. The first maximum oc-         ical calculations. We anticipate that this prop-
      given by Eq. 4 (4)                                      curs at a delay of 26 fs, the first minimum at           erty will be of great value to femtochemistry and
                                                              68 fs, and the second maximum at 106 fs. The             ultrafast imaging.
                     1             ∼                          modulations in the diffracted signal thus reflect
         Im¼1 ðW,tÞ ¼ jrAðtÞðd AeifA − dg eifg Þ þ
                        ∼      ∼
                     4                                        the diabatic state population dynamics. Con-                  References and Notes
                                                              sidering the complexity of the problem and the            1. E. Goulielmakis et al., Nature 466, 739 (2010).
                        rXðtÞðd XeifX − dg eifg Þj2
                         ∼      ∼
                                                              simplicity of our model, the agreement is re-             2. L. Nugent-Glandorf et al., Phys. Rev. Lett. 87, 193002
           The high-harmonic amplitude is determined          markable. To exclude the possibility that the                (2001).
                                                                                                                        3. W. Li et al., Science 322, 1207 (2008).
      by the probabilities of ionization and recombi-         observed modulations result from a change of              4. H. J. Wörner, J. B. Bertrand, D. V. Kartashov,
      nation. The phase is determined by the phase            the strong-field-ionization rate of the molecule             P. B. Corkum, D. M. Villeneuve, Nature 466, 604 (2010).
      accumulated by the bound state (Ip t), and the          as a function of the nuclear coordinates, we have         5. H. J. Wörner, J. B. Bertrand, P. B. Corkum, D. M.
      recombination phase (24), where t stands for            also measured the total ion yield in parallel to the         Villeneuve, Phys. Rev. Lett. 105, 103002 (2010).
                                                                                                                        6. M. Meckel et al., Science 320, 1478 (2008).
      the transit time of the electron in the continuum       high-harmonic yield and have not observed any             7. C. Z. Bisgaard et al., Science 323, 1464 (2009).
      (~1 to 1.7 fs). To the first order, ionization from     modulation on top of the smooth increase (see             8. W. Domcke, D. R. Yarkony, H. Köppel, Eds., Conical
               ∼          ∼
      either A 2B2 or X 2A1 involves the removal and          fig. S3 and accompanying text).                              Intersections: Electronic Structure, Dynamics and
      recombination of an a1 electron in the cross-               Comparing the experimental and theoretical               Spectroscopy, vol. 15, Advanced Series in Physical
                                                                                                                           Chemistry (World Scientific, Singapore, 2004).
      polarized case (Fig. 1). The main difference is         results, we can draw a qualitative picture of the         9. P. H. Bucksbaum, Science 317, 766 (2007).
      due to the Ip t phase. The ground-state channel         evolution of the electronic structure of the mole-       10. D. Polli et al., Nature 467, 440 (2010).
       A1 ← 2A1 has an Ip of 11.2 eV, whereas the             cule as it crosses the conical intersection. Photo-      11. T. Schultz et al., Science 306, 1765 (2004).
      main excited-state ionization channel 3B2 ←             excitation prepares the wave packet on the upper         12. G. E. Busch, K. R. Wilson, J. Chem. Phys. 56, 3626 (1972).
      2                                                                                                                13. M. Quack, J. Troe, Ber. Bunsenges. Phys. Chem 78, 240
        B2 has an Ip that varies between 9.8 eV and           diabatic state as shown in Fig. 4D. When it first
      13.2 eV as a function of the bending coordinate.        approaches the conical intersection, it has little       14. S. I. Ionov, G. A. Brucker, C. Jaques, Y. Chen, C. Wittig,
      The relative phase difference ΔIp t stays close to      expansion along the asymmetric stretch coordi-               J. Chem. Phys. 99, 3420 (1993).
      zero for the X 2A1 state, whereas it is signif-         nate (the b2 mode responsible for vibronic cou-          15. F. Santoro, C. Petrongolo, J. Chem. Phys. 110, 4419 (1999).
      icantly larger (between 1 and 3 radians) for most       pling), and thus most of the amplitude traverses         16. S. Mahapatra, H. Köppel, L. S. Cederbaum, P. Stampfufl,
                              ∼                                                                                            W. Wenzel, Chem. Phys. 259, 211 (2000).
      geometries of the A 2B2 state. Hence, in Eq. 4,         the intersection and remains on the same diabatic        17. Y. Arasaki, K. Takatsuka, Chem. Phys. 338, 175 (2007).
      jd ∼ e − dg e j≫jd ∼ eifX − dg eifg j and the time
            ifA       ifg
                                                              state [80% according to the wave packet cal-             18. Y. Arasaki, K. Takatsuka, K. Wang, V. McKoy, J. Chem.
      dependence of the diffracted signal will be dom-        culation (17)]. This fraction, the diabatic wave             Phys. 132, 124307 (2010).
      inated by HHG emission from the A 2B2 state. It         packet, returns to the conical intersection with a       19. I. Wilkinson, B. J. Whitaker, Annu. Rep. Prog. Chem. Sect. C:
                                                                                                                           Phys. Chem. 106, 274 (2010).
      is thus sensitive to the temporal variation of the      significant spread along the bond-stretching co-         20. A.-T. Le, R. R. Lucchese, M. T. Lee, C. D. Lin, Phys. Rev. Lett.
      population in the A 2B2 state, as illustrated in Fig.   ordinate, resulting in a strong transfer to the ground       102, 203001 (2009).
      4D. This conclusion is also supported by the            diabatic state. This leads to the first minimum in       21. M. V. Frolov et al., Phys. Rev. Lett. 102, 243901 (2009).
      detailed calculations described in the SOM.             the diabatic state population around 60 fs. After        22. J. Itatani et al., Nature 432, 867 (2004).
                                                                                                                       23. T. Morishita, A.-T. Le, Z. Chen, C. D. Lin, Phys. Rev. Lett.
           The observed polarization dependence of            two or three periods of motion along the bending             100, 013903 (2008).
      the oscillations is a consequence of the electronic     coordinate, wave packet components from dia-             24. H. J. Wörner, H. Niikura, J. B. Bertrand, P. B. Corkum,
      symmetries. Photoexcited molecules have their           batic and adiabatic traversals interfere with each           D. M. Villeneuve, Phys. Rev. Lett. 102, 103901 (2009).
      y axis (O-O axis) parallel to the polarization of       other and extend so significantly along the sym-         25. H. J. Wörner, J. B. Bertrand, P. Hockett, P. B. Corkum,
                                                                                                                           D. M. Villeneuve, Phys. Rev. Lett. 104, 233904 (2010).
      the exciting field (Fig. 1). In the cross-polarized     metric and asymmetric stretch coordinates that
                                                                                                                       26. M. Swoboda et al., Phys. Rev. Lett. 104, 103003 (2010).
      experiment, the emission from excited molecules         no appreciable motion of the wave packet av-             27. M. Schultze et al., Science 328, 1658 (2010).
      is thus dominated by those probed along their           erage position can be defined for times longer           28. B. Abel, B. Kirmse, J. Troe, D. Schwarzer, J. Chem. Phys.
      z axis (C2 axis). The same orientation also dom-        than 200 fs (15, 17).                                        115, 6522 (2001).
                                                                                                                       29. A. Vredenborg, W. G. Roeterdink, M. H. M. Janssen,
      inates the emission from the unexcited molecules,           High-harmonic spectroscopy is a powerful                 J. Chem. Phys. 128, 204311 (2008).
      resulting in a sensitivity to the diabatic electronic   probe of electronic dynamics in nonadiabatic             30. Y. Mairesse et al., Phys. Rev. Lett. 100, 143903 (2008).
      state of the excited molecule (2B2 versus 2A1). In      processes. The homodyne interference of species          31. N. T. Form, B. J. Whitaker, L. Poisson, B. Soep, Phys.
      the case of parallel polarizations, the photoex-        in different electronic states has enabled us to             Chem. Chem. Phys. 8, 2925 (2006).
                                                                                                                       32. D. Irimia, I. D. Petsalakis, G. Theodorakopoulos,
      cited molecules are being probed along their y          distinguish multiple photochemical pathways—                 M. H. M. Janssen, J. Phys. Chem. A 114, 3157 (2010).
      axis, whereas the unexcited molecules are probed        electronic excitation to bound states versus exci-       33. J. B. Hamard, R. Cireasa, B. Chatel, V. Blanchet,
      along their z axis. Therefore, the emission from        tation followed by dissociation. The coherence of            B. J. Whitaker, J. Phys. Chem. A 114, 3167 (2010).
      excited molecules in both diabatic states differs       the high-harmonic emission also enabled us to            34. J. B. Bertrand et al., Phys. Rev. Lett. 106, 023001 (2011).
                                                                                                                       Acknowledgments: We thank A. Stolow for fruitful
      significantly from that of the unexcited molecules,     extract amplitudes and phases of the various spe-            discussions and B. Whitaker and K. Takatsuka for permission
      and the amount of diffracted light is sensitive         cies occurring in the photochemical transforma-              to use information underlying Figs. 1 and 4C. We
      only to the total population of excited molecules.      tion and to learn how to interpret them. Temporal            acknowledge financial support from the Swiss National
           Because the electronic dynamics between            variations in the dominant electronic configura-             Science Foundation (PP00P2_128274), Agence Nationale
       ∼            ∼                                                                                                      de la Recherche (ANR-08-JCJC-0029 HarmoDyn), Centre
      X 2A1 and A 2B2 have not been observed exper-           tion of the photoexcited wave packet are man-                National de la Recherche Scientifique (PICS: Imagerie
      imentally before, we compare the measurements           ifested in a polarization dependence of the pump-            moléculaire par impulsions attosecondes), National Sciences
      to recent quantum dynamical calculations on             probe signal, which is expected to be a powerful             and Engineering Research Council of Canada, Canadian
      NO2 (17, 18). These three-dimensional wave              property in future studies of electronic dynamics.           Institute for Photonic Innovations, and Air Force Office of
                                                                                                                           Scientific Research.
      packet calculations have predicted characteristic           We have thus demonstrated how to use high-
      oscillations in the diabatic populations over the       harmonic spectroscopy to elucidate a complex             Supporting Online Material
      first few hundred femtoseconds. The diabatic            photochemical process from the first femto-    
       ∼                                                                                                               Materials and Methods
      A 2B2 population rAðtÞ, convoluted with a 50-fs
                              ∼                               seconds that are governed by a conical inter-
                                                                                                                       Figs. S1 to S5
      Gaussian cross-correlation function, is shown           section to the picosecond time scale where               Tables S1 to S6
      in Fig. 4C. Both the overall behavior and the           dissociation proceeds statistically. Our results         References (35–43)
      distinct features observed in the diffracted high-      on the femtosecond dynamics may be used in               20 May 2011; accepted 15 August 2011
      harmonic signal are present in the calculated           the future to check high-level quantum dynam-            10.1126/science.1208664

212                                            14 OCTOBER 2011            VOL 334        SCIENCE
                                                                                                                          molecular chains could cross the atomic step
Linear Alkane Polymerization                                                                                              edges of the surface (fig. S1D). At high surface
                                                                                                                          coverage, ends of the chains were rarely ob-
on a Gold Surface                                                                                                         served. The dark spots on the chains (marked by
                                                                                                                          arrows in the inset of Fig. 1B) correspond to Au
                                                                                                                          vacancies in the atomic rows rather than breaks in
Dingyong Zhong,1,2 Jörn-Holger Franke,1,2 Santhosh Kumar Podiyanachari,3 Tobias Blömker,3                                 molecular chains, as confirmed by removing the
Haiming Zhang,1,2 Gerald Kehr,3 Gerhard Erker,3* Harald Fuchs,1,2* Lifeng Chi1,2*                                         chains and rendering the vacancies in the Au
                                                                                                                          channels visible by STM (fig. S1). We repeated
In contrast to the many methods of selectively coupling olefins, few protocols catenate saturated                         the experiment more than 20 times in a controlled
hydrocarbons in a predictable manner. We report here the highly selective carbon-hydrogen (C–H)                           manner and concluded that the transition of the
activation and subsequent dehydrogenative C–C coupling reaction of long-chain (>C20) linear                               surface reconstruction and the formation of polym-
alkanes on an anisotropic gold(110) surface, which undergoes an appropriate reconstruction by                             erized molecular chains are thermally activated.
adsorption of the molecules and subsequent mild annealing, resulting in nanometer-sized channels                          Other stimuli such as light or electron irradiation
(1.22 nanometers in width). Owing to the orientational constraint of the reactant molecules in                            could be excluded, because these sources were
these one-dimensional channels, the reaction takes place exclusively at specific sites (terminal                          not used in the experimental setup.
CH3 or penultimate CH2 groups) in the chains at intermediate temperatures (420 to 470 kelvin)                                 By STM tip manipulation, sections of the
and selects for aliphatic over aromatic C–H activation.                                                                   molecular chains could be pulled out from the
                                                                                                                          channels (Fig. 1C), providing evidence that al-
         lkanes are the principal components in                  for alkyl C–H activation. The 1.22-nm-wide atomic        kane monomers were covalently bonded to form

A        petroleum, natural gas, and biogas and are
         widely used as a feedstock in chemical
production. It has long been desirable to use the
                                                                 grooves resulting from the missing-row recon-
                                                                 struction of Au(110)-(1×3) surface efficiently
                                                                 confine the diffusion of adsorbed molecules and
                                                                                                                          polymers. Usually polymer sections released from
                                                                                                                          the channels were longer than the evacuated
                                                                                                                          channel section visible in the STM images, in-
relatively inert alkanes directly as much cheaper                constrain the molecules into a 1D pathway, their         dicating that the polymer chains could slide along
precursors than their commonly employed func-                    interactions restricted to neighboring molecules         the channels during STM manipulation. The re-
tionalized derivatives for polymerization processes.             in the easy-diffusion direction within the atomic        sult of manipulation further proved that the step
In the past decades, strategies have been developed              channels. For instance, the hydrogen atoms on            edges of the substrate surface did not break the
for activating C–H bonds under mild conditions                   terminal methyl groups can dissociate and desorb         polymer chains. Most of them are longer than
and with high selectivity by using transition metal              from the surface, leading to C–C bond formation          200 nm or 50 monomer units (for example, as
centers, resulting in bond formation between metal               between the residual chains (reaction 1)                 shown in fig. S1E). Further STM manipulation
centers and alkyl groups (1–5). C–H bond ac-                                                                              results are shown in fig. S2. The periodicities of
                                                                                       1D constraint
tivation can also be realized at alkanes adsorbed                                            D                            the gold atomic rows and the polymer chains
on certain transition metal surfaces. In such het-                                ÀÀÀ
                                                                 R–CH3 þ CH3 –R′ À À À À→                                 were 0.29 and 0.25 nm, respectively (Fig. 1, D
erogeneous catalytic approaches, the molecules                                    R–CH2 –CH2 –R′ þ H2                     and E). The mismatch between the two perio-
may undergo various competing processes simul-                                                                     ð1Þ    dicities results in a Moiré pattern with a larger
taneously with poor selectivity (6–8). C–H bond                                                                           periodicity (that is, six gold atoms or 14 CH2
activation and direct C–C coupling reactions                         Experimentally, we deposited n-dotriacontane         units). The Moiré pattern, as well as the (1×3)
resulting in the conversion of low– to higher–                   (C32H66) on an Au(110)-(1×2) surface at 300 K            reconstruction of the substrate surface, was fur-
molecular-weight hydrocarbons can be achieved                    under ultrahigh-vacuum (UHV) conditions (27).            ther confirmed by low-energy electron diffraction
under superacidic conditions as well (9–11), or at               Compared with shorter linear alkanes, the rela-          (fig. S3).
high temperatures by using catalysts supported                   tively strong interaction of n-dotriacontane mono-           We conducted temperature-programmed de-
by porous materials (12–15). However, the latter                 mers with the surface inhibits their desorption at       sorption (TPD) to monitor the generation of H2
processes are not energetically economical due                   elevated temperatures. The molecules physisorbed         during thermally activated polymerization. A pro-
to the high processing temperatures.                             on the surface with their long axis along the gold       nounced H2 signal appeared at a temperature of
    Bulk gold has long been considered minimal-                  atomic rows, as observed by scanning tunnel-             529 K on heating the C32H66 adsorbed Au(110)
ly reactive, exhibiting poor catalytic activity. Sub-            ing microscopy (STM) at 78 K. At low coverage            sample up to 543 K at a constant power of 1.2 A
stituent gold atoms on Ni(111) surfaces even                     (≤0.3 nm–2) the molecules were located in the            by 10.25 V (Fig. 1F, see figs. S4 and S5 for fur-
inhibit methane C-H bond dissociative adsorp-                    grooves of the (1×2) reconstruction. At increased        ther TPD results). We also performed the anal-
tion (16, 17). In contrast, supported gold nano-                 coverage (0.5 nm–2), they packed on the surface          ogous measurement with the deuterated alkane
particles and ultrathin films possess strong activity            with five molecules in three reconstruction units        C30D62. Because of the lower background of D2
as heterogeneous catalysts (18–23). However,                     (Fig. 1A). On annealing at a temperature of 440 K        in the UHV chamber, the relative D2 signal was
recent studies show that bulk Au(111) surfaces                   for 30 min, the substrate surface underwent a            even more prominent. No signal was observed
can, in fact, serve as a platform to catalyze dehalo-            phase transition from (1×2) to (1×3) reconstruc-         on heating the clean Au(110) substrate or the
genating polymerization and dehydrocyclization                   tion. Such (1×n) reconstructions (n ≥ 3) at Au(110)      polymerized sample during a second and third
(24–26). Here, we report that the Au(110) surface                and Pt(110) surfaces, which have been previously         control heating cycle (30 min and 24 hours after
which exhibits a one-dimensional (1D) constraint,                observed (28, 29), are attributed to the interplay       the first heating process, respectively). Moreover,
acts as the platform and heterogeneous catalyst                  of the surfaces and the adsorbates. In our case,         STM characterization on the samples after heat-
                                                                 two of every three atomic rows along the [1–10]          ing to different temperatures during TPD indicated
  Physikalisches Institut, Universität Münster, Wilhelm-Klemm-   direction are missing, thus resulting in broader chan-   a direct correlation between polymerization and
Strasse 10, 48149 Münster, Germany. 2Center for Nanotech-        nels (1.22 nm) than those in the (1×2) reconstruc-       H2 generation. We also found that the peak signal
nology, Heisenbergstrasse 11, 48149 Münster, Germany.            tion (0.81 nm). Notably, we found that, after            shifted toward lower temperatures on decreasing
  Organisch-Chemisches Institut, Universität Münster, Correns-
strasse 40, 48149 Münster, Germany.                              annealing, the alkane monomers were bonded end-          the heating power (at 510 K for a power of 1.0 A
*To whom correspondence should be addressed. E-mail:
                                                                 to-end, forming linear molecular chains within           by 7.60 V and at 496 K for 0.9 A by 6.40 V). (G.E.);             the channels (Fig. 1B). Although some of them                For further confirmation of the dehydrogen-
(H.F.); (L.C.)                               terminated at a step edge at low coverage, the           ative polymerization process, we explored 1,4-

                                                   SCIENCE        VOL 334        14 OCTOBER 2011                                                    213
      Fig. 1. Dehydrogena-
      tive polymerization of A                                                                                                                                       Temperature (K)
                                                                     B                                                  F                          363            458
      n-dotriacontane (C32H66)                                                                                                                                          488      518         540
      on Au(110). (A) STM im-

                                                                                                                         Intensity (arb. unit)
      age of n-dotriacontane                                                                                                                     2.5                                     529 K
      monomers deposited on                                                                                                                                      H2 (mass = 2)
      Au(110)-(1×2) at 300 K                                                                                                                     2.0
      with monolayer coverage.
      –1 V, 0.1 nA, 20 by 15 nm2.                                                                                                                1.5
      (B) Parallel polyethyl-
      ene chains (bright lines) C                                                                                                                1.0
      formed in the Au(110)-                                                                                                                           0          25        50       75      100
      (1×3)reconstructiongrooves                                                                                                                                       Heating time (s)
      by heating at 440 K for
      30 min. 0.5 V, 0.5 nA,
      50 by 50 nm2. (Inset) 9                                                                                           E
      by 13 nm2. Red arrows, Au
                                                                                                                                                 0.20                          ( )

                                                                                                                         Height (nm)
      vacancies in atomic rows                                                                                                                   0.15                       14x(CH2)
      underneath the polymer                                         D
      chains. (C) Several poly-                                                                                                                  0.10
      ethylene chains on a sur-
      face with low-coverage
      coexisting (1×2) and (1×3)                                                                                                                 0 .00
      reconstruction grooves, par-                                                                                                                         0.0      0.5       1.0    1.5     2.0
      tially released from the                                                                                                                                            Distance (nm)
      grooves by the STM tip.
      White arrows denote the empty (1×3) missing-row grooves after removing           Au atomic row (0.29 nm) and the polymer chain (0.25 nm). –0.02 V, 2 nA, 6 by
      the polymer chains. –1 V, 0.3 nA, 50 by 70 nm2. (D) High-resolution STM          2.5 nm2. (E) Height profiles at the dotted lines in (D). (F) H2 signal detected by
      image of Au atomic rows and polyethylene chain. The contrast modulation with     mass spectrometry during the annealing process. C32H66 was deposited at 300 K
      a periodicity of 1.74 nm originates from the periodicity mismatch between the    with coverage of 0.2 nm2. Heating power: 1.2 A by 10.25 V.

                                                                                                                   di(eicosyl)benzene (DEB), which comprises a
                                                                                                                   phenylene moiety connecting two alkyl chains,
                                                                                                                   as a monomer. In the STM image of the as-
                                                                                                                   deposited sample (Fig. 2A), the middle phenyl-
                                                                                                                   ene groups (exhibiting a bright contrast) were
                                                                                                                   adsorbed on the uppermost gold rows, whereas the
                                                                                                                   alkyl chains, oriented along the gold atomic rows,
                                                                                                                   were adsorbed in the neighboring channels.
                                                                                                                   Polymerization took place at 420 K, resulting in
                                                                                                                   the formation of long polymeric chains in the
                                                                                                                   (1×3) reconstruction grooves (Fig. 2B). The phenyl-
                                                                                                                   ene rings were clearly observed by STM in either
                                                                                                                   anti- or syn-conformational arrangements along
                                                                                                                   the chains (circled in Fig. 2B). Additional bright
                                                                                                                   spots appeared close to the alkyl chains, roughly
                                                                                                                   in the middle between the two neighboring phenyl-
                                                                                                                   ene groups (marked by arrows in Fig. 2B). We
                                                                                                                   assign these to methyl branches resulting from
                                                                                                                   polymerization, based on the following experimen-
                                                                                                                   tal indication: The distances of the bright spots to
                                                                                                                   the two neighboring phenylene groups exhibit a
                                                                                                                   slight difference of about 0.2 to 0.3 nm (Fig. 2C),
                                                                                                                   which corresponds to a difference in length of
                                                                                                                   the two subsections of ~2 CH2 groups. There-
                                                                                                                   fore, we conclude that C–C coupling can occur
      Fig. 2. Polymerization and dimerization of hydrocarbons. (A) DEB monomers deposited on Au(110)-              not only between the terminal methyl groups, but
      (1×2) at 300 K. –0.5 V, 0.5 nA, 12 by 12 nm2. (B) Polymerized DEB chains located in the Au(110)-(1×3)        also between a terminal methyl group and a pe-
      missing-row channels on heating at 420 K for 10 hours. The C–C coupling takes place between the              nultimate methylene group, from each of which
      terminal carbon atom of one molecule and the penultimate carbon atom of a neighboring molecule.              a hydrogen atom dissociates (reaction 2)
      Circles denote the phenylene groups; arrows denote the methyl side groups. –1 V, 2 nA, 17.5 by 6 nm2.
                                                                                                                                                                             1D constraint
      (C) A section of DEB polymer chain overlapped with the molecular model. The newly formed C–C bonds                                                                           D
      are shown in red. –1 V, 2 nA, 14.4 by 1.6 nm2. (D) Dimerization of EB on heating at 440 K for 1 hour.                               ÀÀÀÀ
                                                                                                                    R–CH3 þ CH3 –CH2 –R′ À À À À →
      –1 V, 0.5 nA, 17.5 by 10 nm2. (E and F) Two types of dimerization: (E) terminal C–C coupling (a) and (F)                                                   R–CH2 –CH(CH3 )–R′ þ H2
      terminal/penultimate C–C coupling (b) with a methyl side group. The newly formed C–C bonds are shown
      in red. –1 V, 0.5 nA, 7 by 1.6 nm2.                                                                                                                                                     ð2Þ

214                                          14 OCTOBER 2011           VOL 334        SCIENCE

Fig. 3. Calculated energy diagrams for C–H activation of n-hexane at the           defect site (adatom-assisted process) is also shown. In general, the C–H ac-
terminal and penultimate methylene (C-2 and C-3) groups on gold surfaces. (A)      tivation is energetically more favorable on the Au(110)-(1×3) surface than on
Energy diagram on Au(110)-(1×3). (B) Energy diagram on Au(111). On the             Au(111). Furthermore, there are only small differences between the activation
Au(110)-(1×3) surface, the energy trajectory for terminal methyl activation at a   sites on Au(110), with activation at C-2 being the energetically lowest pathway.

    Further evidence for the terminal-penultimate      role in the selective C–H activation and C–C bond       ences (Fig. 3A). In contrast to the activation en-
C–C coupling is the dependence of molecular            coupling.                                               ergies on the Au(110)-(1×3) reconstructed surface,
configurations on the alkyl chain length (see              Although elucidation of the mechanism of the        the corresponding processes on Au(111) face a
fig. S6 and accompanying text). The terminal-          alkane coupling and polymerization reaction on          higher barrier, as shown in Fig. 3B (C-1, 1.69 eV;
penultimate C–C coupling, which was also ob-           the Au(110) surface needs further experimental          C-2, 1.75 eV; C-3, 1.91 eV). Therefore, we at-
served in the case of dotriacontane as a minor         work, we carried out density functional theory          tribute the preferential C–H activation and C–C
reaction pathway, is the dominant one in the case      (DFT) calculations to achieve some preliminary          coupling at the terminal methyl or penultimate
of DEB, probably due to the asymmetrical lo-           insight. The activation barrier for the formation       methylene groups on Au(110) to the 1D physical
cation of the two alkyl chains induced by the          of the alkyl intermediate by hydrogen atom dis-         constraint of the reactants. The anisotropic sur-
middle phenylene ring.                                 sociation from the terminal methyl group of the         face has two major effects on the reaction: First,
    On the basis of these results, we reasoned that    model compound n-hexane located in the groove           compared with the close-packed (111) surface,
dimerization could take place with molecules such      of the Au(110)-(1×3) surface was calculated to be       high-coordination sites on the (1×3) reconstructed
as eicosylbenzene (EB) that carries a single alkyl     1.33 eV (Fig. 3A). The distance between the carbon      Au(110) surface strengthen the interaction be-
chain blocked at one end by a phenyl group. On         atom of the resulting terminal methylene group          tween the surface and adsorbed species and there-
annealing at 440 K for 1 hour, EB molecules ad-        and the underlying gold atom decreases from             by decrease the activation energy. Second and
sorbed on the Au(110) surface formed two               0.336 to 0.216 nm during the reaction, indicating       more importantly, the diffusion of the molecules
types of dimers, which we designate type a (C–C        a major interaction with a binding energy of            in the reconstruction grooves is confined in the
coupling via the terminal methyl groups) and           –2.26 eV, much higher than the metal interaction        [1–10] direction, and the molecules can only in-
type b (coupling of terminal methyl and penul-         with the intact molecule (–0.93 eV). The C–H            teract with their neighbors in the same groove,
timate methylene groups) (Fig. 2, D to F). The         bond length is extended to 0.176 nm from its initial    whereas the interactions between the molecules
type b dimers contain a methyl branch and so are       value of 0.111 nm in the transition state, before the   lying in different grooves, which are separated by
slightly shorter than the type a dimers. These ex-     bond finally breaks. The overall calculated activ-      the top gold atomic rows, are negligible.
periments indicate that on reconstructed Au(110)       ation energy is much lower than the thermal methyl          The surface-assisted hydrogen dissociation
surfaces, methyl C–H bonds are easier to activate      C–H bond-dissociation energy (>4 eV) and is             process is endothermic by ~1 eVon the Au(110)-
than phenyl C–H bonds, in contrast to earlier ob-      consistent with the mild processing temperatures        (1×3) surface, according to our calculations.
servations that transition metal–induced activation    (420 to 470 K) required to induce/trigger polym-        Therefore, the process is evidently driven by the
of the stronger aryl C–H bonds is often kinet-         erization in our experiments. In addition to the        desorption of hydrogen molecules, which are
ically preferred in solution (30–33).                  surface-assisted dissociation process, we also in-      formed from dissociated hydrogen atoms and
    For comparison, we performed similar ex-           vestigated the role of gold adatoms in C–H activ-       readily leave from the surface under UHV con-
periments with n-dotriacontane on an Au(111)           ation. In the adatom-assisted dissociation process,     ditions (35). The loss of available hydrogen atoms
surface. Cai et al. recently reported that much        the C–H activation is assisted by a gold adatom         on the surface then tilts the equilibrium toward
higher temperatures are required for activating        on which both the dissociated hydrogen atom             formation of alkyl intermediates. Once these in-
aromatic C–H bonds on Au(111) surfaces (26).           and the resulting alkyl intermediate are bound,         termediates are formed on the surface, the sub-
In our case, no C–C coupling reaction was ob-          similar to a previous report on nickel-catalyzed        sequent C–C coupling should be an exothermic
served on heating up to 470 K. At higher temper-       methane dissociation (34). Compared with the            process with a relatively low activation barrier.
atures, branched islands formed on the surface         surface-assisted process discussed above, our cal-
due to possible hydrogen dissociation, C–C bond        culations indicate a higher activation barrier of           References and Notes
breaking and coupling at undefined positions           1.51 eV for the adatom-assisted process, imply-          1. K. I. Goldberg, A. S. Goldman, Eds., Activation and
on the linear alkane molecules, accompanied            ing that this process is unimportant in our case.           Functionalization of C–H Bonds ACS Symposium Series,
                                                                                                                   Vol. 885 (American Chemical Society, Washington, DC,
by the desorption of H2 (fig. S4). This result             A comparison of the activation energies at the          2004).
implies that the constraint in a 1D channel on         terminal methyl (C-1) and penultimate methylene          2. X. Chen, K. M. Engle, D.-H. Wang, J.-Q. Yu, Angew. Chem.
the anisotropic Au(110) surface plays a key            (C-2 and C-3) groups shows only minor differ-               Int. Ed. 48, 5094 (2009).

                                         SCIENCE        VOL 334       14 OCTOBER 2011                                                             215
       3. V. Ritleng, C. Sirlin, M. Pfeffer, Chem. Rev. 102, 1731       18. M. Turner et al., Nature 454, 981 (2008).                                                                   34. M. R. A. Blomberg, U. Brandemark, P. E. M. Siegbahn,
          (2002).                                                       19. M. Haruta, Catal. Today 36, 153 (1997).                                                                         J. Am. Chem. Soc. 105, 5557 (1983).
       4. H. Chen, S. Schlecht, T. C. Semple, J. F. Hartwig, Science    20. A. S. K. Hashmi, G. J. Hutchings, Angew. Chem. Int. Ed.                                                     35. A. G. Sault, R. J. Madix, C. T. Campbell, Surf. Sci. 169,
          287, 1995 (2000).                                                 45, 7896 (2006).                                                                                                347 (1986).
       5. S. E. Bromberg et al., Science 278, 260 (1997).               21. R. Zhao et al., Chem. Commun. 2004, 904 (2004).                                                             Acknowledgments: This work was financially supported
       6. G. A. Somorjai, Introduction to Surface Chemistry and         22. B. N. Zope, D. D. Hibbitts, M. Neurock, R. J. Davis,                                                            by the Deutsche Forschungsgemeinschaft through
          Catalysis (Wiley, New York, 1994).                                Science 330, 74 (2010).                                                                                         grants SFB858 and TRR61. H.F. thanks the Institute
       7. G. A. Somorjai, A. L. Marsh, Philos. Trans. R. Soc. London    23. M. S. Chen, D. W. Goodman, Science 306, 252 (2004);                                                             of Nanotechnology of the Karlsruhe Institute of
          Ser. A 363, 879 (2005).                                           10.1126/science.1102420.                                                                                        Technology for support. DFT computations were
       8. J. H. Lunsford, Catal. Today 63, 165 (2000).                  24. L. Grill et al., Nat. Nanotechnol. 2, 687 (2007).                                                               performed at the ZIVSMP and ZIVCLUSTER of the
       9. G. A. Olah, R. H. Schlosberg, J. Am. Chem. Soc. 90, 2726      25. L. Lafferentz et al., Science 323, 1193 (2009).                                                                 Univ. of Münster. A patent has been filed on the work
          (1968).                                                       26. J. Cai et al., Nature 466, 470 (2010).                                                                          presented herein.
      10. G. A. Olah, Makromol. Chem. 175, 1039 (1974).                 27. Materials and methods are available as supporting
      11. M. Herlem, Pure Appl. Chem. 49, 107 (1977).                       material on Science Online.
      12. J. Oró, J. Han, Science 153, 1393 (1966).                     28. X. Zhao, H. Yan, R. G. Zhao, W. S. Yang, Langmuir 18,
                                                                                                                                                                                  Supporting Online Material
      13. L. Wang, Y. Xu, L. Tao, Sci. China B 40, 161 (1997).              3910 (2002).
                                                                                                                                                                                  Materials and Methods
      14. L. Wang, Y. Xu, S.-T. Wong, W. Cui, X. Guo, Appl. Catal. A    29. P. Häberle, P. Fenter, T. Gustafsson, Phys. Rev. B 39,
                                                                                                                                                                                  SOM Text
          152, 173 (1997).                                                  5810 (1989).
                                                                                                                                                                                  Figs. S1 to S6
      15. Y. Xu, L. Lin, Appl. Catal. A 188, 53 (1999).                 30. J. Halpern, Inorg. Chim. Acta 100, 41 (1985).
                                                                                                                                                                                  References (36–51)
      16. F. Besenbacher et al., Science 279, 1913 (1998).              31. W. D. Jones, F. J. Feher, Acc. Chem. Res. 22, 91 (1989).
      17. P. Kratzer, B. Hammer, J. K. Nørskov, J. Chem. Phys. 105,     32. R. H. Crabtree, J. Organomet. Chem. 689, 4083 (2004).                                                 28 July 2011; accepted 1 September 2011
          5595 (1996).                                                  33. Z. Lin, Coord. Chem. Rev. 251, 2280 (2007).                                                           10.1126/science.1211836

                                                                                                                                                                                  Macroscopic frictional behavior results from the
      Flash Heating Leads to Low Frictional                                                                                                                                       interactions of these highly stressed microscopic
                                                                                                                                                                                  contacts. The total heat input Qc at a contact dur-
      Strength of Crustal Rocks at                                                                                                                                                ing its lifetime q is Qc = fcscVAcq, where fc is
                                                                                                                                                                                  the friction coefficient of the contact (i.e., the con-
                                                                                                                                                                                  tact shear stress t c divided by contact normal stress
      Earthquake Slip Rates                                                                                                                                                       sc), V is sliding velocity, and Ac the area of the con-
      David L. Goldsby* and Terry E. Tullis                                                                                                                                       tact. At slow slip rates, the heat generated at con-
                                                                                                                                                                                  tacts can diffuse away appreciably over the contact
      The sliding resistance of faults during earthquakes is a critical unknown in earthquake physics. The                                                                        lifetime, resulting in a small temperature rise and
      friction coefficient of rocks at slow slip rates in the laboratory ranges from 0.6 to 0.85, consistent with                                                                 negligible effect on contact strength. At suffi-
      measurements of high stresses in Earth’s crust. Here, we demonstrate that at fast, seismic slip rates,                                                                      ciently high slip rates, however, there is insuffi-
      an extraordinary reduction in the friction coefficient of crustal silicate rocks results from intense “flash”                                                               cient time for heat generated at contacts to diffuse
      heating of microscopic asperity contacts and the resulting degradation of their shear strengths. Values                                                                     away appreciably, resulting in increased contact
      of the friction coefficient due to flash heating could explain the lack of an observed heat flow anomaly                                                                    temperature and decreased contact strength. Thus,
      along some active faults such as the San Andreas Fault. Nearly pure velocity-weakening friction due                                                                         high contact stresses and high sliding velocities
      to flash heating could explain how earthquake ruptures propagate as self-healing slip pulses.                                                                               may induce intense, transient heating, even melt-
                                                                                                                                                                                  ing, of contacts that results in low shear strengths
                lthough high friction in laboratory ex-                periments have inferred (12, 13) or measured                                                               and thus low macroscopic friction. Extreme weak-

      A         periments at slow slip rates and high
                stresses measured in boreholes in the
      Earth’s crust (1) support the idea that faults are
                                                                       (14) average contact sizes of ~1 to 25 mm, result-
                                                                       ing in contact stresses of ~10 GPa for framework
                                                                       silicates, even for modest nominal normal stresses.
                                                                                                                                                                                  ening via this so-called “flash” heating can occur
                                                                                                                                                                                  even while the average temperature of the slip
                                                                                                                                                                                  surface is barely above ambient.
      statically strong, the emerging view from fault
      mechanics is that major faults are dynamically
      weak during earthquakes. This is based on inte-                  A 1.0                                                                   0.4                            B 1.0                                                              0.7
                                                                                                                               Quartzite                                                                                        Quartzite
      grated observations of the lack of pronounced heat                                      0.8                                                                                                    0.8
                                                                       friction coefficient

                                                                                                                                                                              friction coefficient
                                                                                                                                                     sliding velocity (m/s)

                                                                                                                                                                                                                                                       sliding velocity (m/s)
      flow anomalies associated with seismic slip (2, 3),                                                                                      0.3
                                                                                                    fo                                  fo                                                                                 fo
      the nearly fault-normal orientation of maximum                                          0.6                                                                                                    0.6
      stresses and therefore low resolved shear stresses                                                                                       0.2
      on mature faults (4), and a virtual explosion in the                                    0.4                                                                                                    0.4                                         0.3
      number of laboratory studies in the past decade
                                                                                              0.2                                                                                                    0.2                                         0.2
      demonstrating that rocks are weakened by various                                                                                         0.1
      dynamic-weakening mechanisms [e.g., melt lu-                                                                                                                                                   0.0
      brication (5, 6), gel lubrication (7, 8), and powder                                                                                     0.0                                                                                               0.0
      lubrication (9); see (10, 11) for reviews] when                                                    0   10     20    30      40         50                                                            0   10    20    30       40      50

      sheared at the high slip rates and large sliding                                                            slip (mm)                                                                                         slip (mm)
      displacements characteristic of earthquakes.                     Fig. 1. (A) Friction coefficient (black trace) and sliding velocity (red trace) plotted against sliding
          Two rock surfaces brought together touch                     displacement, for a high-speed friction experiment on quartzite. The friction versus slip curve is essentially
      over only a small fraction of their nominal con-                 a mirror image (across a horizontal mirror plane) of the velocity versus slip curve above the weakening
      tact area. Typical specimens for rock-friction ex-               velocity Vw. The friction coefficient at low slip rates, fo, obtains nearly identical values before and after
                                                                       sliding at rapid slip rates. (B) Friction coefficient (black trace) and sliding velocity (red trace) plotted
      Department of Geological Sciences, Brown University, Provi-      against slip for a VS test on quartzite. As in (A), above a characteristic weakening velocity Vw, the friction
      dence, RI 02912, USA.                                            curve is essentially a mirror image of the velocity curve. The friction coefficient at low slip rates, fo , obtains
      *To whom correspondence should be addressed. E-mail:             nearly identical values before and after rapid slip. On acceleration from 0.06 to 0.13 m/s, weakening is                                          not observed until a slip of ~3 mm has accrued above V = 0.10 m/s (24).

216                                                  14 OCTOBER 2011                                     VOL 334     SCIENCE 
    Here, we report results of high-velocity fric-                                                       ments ensure that the average temperature of the                                                celeration from the peak velocity to 0.1 m/s (Fig.
tion experiments conducted on a variety of rocks                                                         sliding surface remains barely above room tem-                                                  2). Repeat tests on the same sample duplicate the
found in the seismogenic zone of the Earth’s                                                             perature. The experimental conditions thus iso-                                                 same hysteresis loop. Discrete friction versus ve-
crust. The experiments explore seismic slip ve-                                                          late flash heating from the spectrum of behaviors                                               locity data points from VS tests (e.g., Fig. 2.),
locities, up to ~0.4 m/s, over relatively small dis-                                                     associated with meters of slip at high slip rates,                                              overlaid on data from the CVV test for com-
placements, ~ 45 mm. High velocities activate                                                            such as viscous shearing of a frictional-heating–                                               parison, agree well with the data from the decel-
dynamic-weakening mechanisms not observed at                                                             generated melt layer [melt lubrication (5, 6, 15)]                                              eration portion of the CVV experiment (Fig. 2).
low slip rates, whereas relatively small displace-                                                       or a silica gel-like layer [gel lubrication (7, 8)] on                                          The matching friction coefficient versus veloc-
                                                                                                         the slip surface.                                                                               ity trends from both types of test in Fig. 2 reveal
                                                                                                             The friction coefficient during the low-                                                    a 1/V dependence of friction on slip rate above
                                                                                                         velocity portion of continuously varying-velocity                                               ~0.1 m/s (Fig. 3).
                                                                                   Quartzite             (CCV) tests on quartzite reaches a value of ~0.6                                                     CVV tests were also conducted for novacu-
                            0.8                                                                          (Fig. 1A), consistent with values obtained in many                                              lite, albite rock, granite, and gabbro. The tests
                                                                                                         previous studies on quartz rocks at low slip rates                                              yield qualitatively results identical to those de-
     friction coefficient

                            0.6                                                                          [e.g., (16, 17)] and with “Byerlee’s law” (18).                                                 scribed above for quartzite. Data from these tests
                                                                                                         The minimum value of the friction coefficient,                                                  are plotted and compared with models for flash
                            0.4                                                                          ~0.2 (an approximate “average” of the oscillating                                               heating (19, 20) (Fig. 3). The plot of the no-
                                                                                                         friction values), is obtained at the peak sliding                                               vaculite data (Fig. 3B) also includes data from
                                                                                                         velocity of ~0.36 m/s. The friction coefficient                                                 another study on novaculite, which explored high
                                              CVV test
                                              VS tests                                                   increases modestly with the modest decrease in                                                  slip rates (up to 2 to 4 m/s) over short sliding
                                                                                                         velocity over the next 35 mm of slip, followed by                                               displacements (~1 mm) (21). Those experiments
                                                                                                         an abrupt increase in friction accompanying the                                                 explored much higher normal stresses (65 to
                                                                                                 0       abrupt decrease in velocity near the end of the                                                 70 MPa) than in our study; however, contact
                                                          sliding velocity (m/s)                         experiment. The value of the friction coefficient                                               stresses and average contact sizes are expected
                                                                                                         at low velocities at the end of the tests is nearly                                             to be similar in both studies (22, 23).
Fig. 2. Friction coefficient versus sliding velocity
from the CVV test (solid black line) of Fig. 1 and                                                       identical to that obtained at low velocity at the                                                    Inspection of the sample surface after each
several subsequent VS tests (open blue-gray circles)                                                     beginning of the tests (Fig. 1A). Results of the                                                test indicated that a very thin layer of gouge formed
on the same quartzite sample. Arrows indicate the                                                        velocity-stepping (VS) test on the identical                                                    on the initially bare rock surface. The gouge layer
sequence in which the data were acquired in the                                                          quartzite sample corroborate the CVV findings                                                   thickness is ≤30 mm, as seen with transmitted
CVV test. A hysteresis in the friction coefficient from                                                  (Fig. 1B).                                                                                      light microscopy of petrographic thin sections
the CVV experiment results from higher friction on                                                           Data from the CVV test reveal a marked hys-                                                 through the slip surface interface [fig. S4 (24)].
acceleration compared to lower friction on deceler-                                                      teresis, with the friction measured during the                                                  Attempts to determine contact sizes from scanning
ation, attributed to the effects of shear localization                                                   initial acceleration from slow slip rates to the                                                electron microscope analyses of the slip surface
and delocalization above and below, respectively,                                                        peak velocity of 0.36 m/s having higher values at                                               were inconclusive; however, friction experiments
the weakening velocity Vw (24).                                                                          a given slip rate than data obtained during de-                                                 on silicate rock samples of comparable surface

A                           1.0                                                                          B                      1.0                                                                 C 1.0
                                         Quartzite                                                                                         Novaculite                                                                                 Albite rock
                            0.8                                                                                                 0.8                                                                                        0.8
                                                                                                                                                                                                    friction coefficient
friction coefficient

                                                                                                         friction coefficient

                            0.6                                                                                                                                                                                            0.6

                                                                                                                                0.4                                                                                        0.4

                            0.2               CVV test                                                                                          Yuan-Prakash (2008)
                                              VS tests                                                                          0.2                                                                                        0.2
                                              f = fw+ (f 0 - fw )*Vw / V
                                                                                                                                0.0                                                                                        0.0
                                    -3                -2                 -1                 0                                         -3                -2             -1              0                                         -3                 -2           -1          0
                               10                10                 10                 10                                          10              10             10              10                                          10               10           10          10
                                                 sliding velocity (m/s)                                                                           sliding velocity (m/s)                                                                      sliding velocity (m/s)

D                           1.0                                                                          E                                                                                                   Fig. 3. Comparison of friction versus velocity data
                                                                                                                                                                                                             with predictions of the Rice model for flash heating
                            0.8                                                                                                                                                                              (19), shown as red curves (24); dashed portions of
friction coefficient

                                                                                                         friction coefficient

                                                                                                                                                                                                             curves are extrapolations to a slip rate of 1 m/s or
                            0.6                                                                                                                                                                              larger. Plots are for (A) quartzite (black symbols,
                                                                                                                                                                                                             CVV test; blue-gray symbols, VS test), (B) novacu-
                            0.4              Granite                                                                                            Gabbro                                                       lite, (C) Tanco albite rock, (D) Westerly granite, and
                                                                                                                                                                                                             (E) India gabbro.
                            0.2                                                                                                 0.2

                            0.0                                                                                                 0.0
                                  -3                      -2                  -1                     0                                     -3                -2              -1                 0
                               10                    10               10                        10                                    10             10                 10                 10
                                                sliding velocity (m/s)                                                                              sliding velocity (m/s)

                                                                                                                    SCIENCE           VOL 334         14 OCTOBER 2011                                                                                217
      roughness [e.g., (13)] yield values of the critical       [table S2 (24)], and the similar roughnesses of                 3. J. N. Brune, T. L. Henyey, R. F. Roy, J. Geophys. Res.
      slip weakening distance Dc, a proxy for average           the samples, serve to yield similar values of Vw .                 74, 3821 (1969).
                                                                                                                                4. V. S. Mount, J. Suppe, Geology 15, 1143 (1987).
      contact size, of 1 to 25 mm. The reproducibility              The dramatic effects of flash heating on                    5. A. Tsutsumi, T. Shimamoto, Geophys. Res. Lett. 24,
      of the data for identical successive tests on a           natural faults could be moderated by (i) ho-                       699 (1997).
      given sample indicates that contact sizes did not         mogeneously distributed shearing of fault gouge                 6. G. Di Toro, T. Hirose, S. Nielsen, G. Pennacchioni,
      evolve significantly with accumulated slip [fig.          (ground-up rock in faults), leading to a gouge                     T. Shimamoto, Science 311, 647 (2006).
                                                                                                                                7. D. L. Goldsby, T. E. Tullis, Geophys. Res. Lett. 29, 1844
      S3 (24)].                                                 particle-scale slip rate that might be less than                   (2002).
          The experimental results agree both qualita-          the far-field rate by a factor of L/p, where L is               8. G. Di Toro, D. L. Goldsby, T. E. Tullis, Nature 427,
      tively and quantitatively with theory (19). Con-          the layer thickness and p the gouge particle size;                 436 (2004).
      tacts are heated to the temperature appropriate for       and (ii) ultracomminution of gouge, such that                   9. Z. Reches, D. A. Lockner, Nature 467, 452 (2010).
                                                                                                                               10. T. E. Tullis, in Treatise on Geophysics, H. Kanamori,
      the ambient sliding velocity over slip distances on       contacts become small enough that the value of                     Ed. (Elsevier, Amsterdam, 2007), vol. 4, pp. 131–152.
      the order of the contact size (25, 26), 1 to 25 mm        Vw is greater than seismic slip rates (Eq. 2).                 11. G. Di Toro et al., Nature 471, 494 (2011).
      for rocks with roughnesses comparable to those            Clay minerals, common in faults, would, if load-               12. J. H. Dieterich, Pure Appl. Geophys. 116, 790 (1978).
      tested here (12–14), such that the macroscopic            supporting, bear less contact stress than stron-               13. J. H. Dieterich, J. Geophys. Res. 84, 2161 (1979).
                                                                                                                               14. J. H. Dieterich, B. D. Kilgore, Tectonophysics 256,
      friction coefficient is a nearly pure function of         ger framework silicates, possibly shifting Vw to
                                                                                                                                   219 (1996).
      sliding velocity above a characteristic weakening         beyond-seismic slip rates (Eq. 2). Flash heating               15. J. G. Spray, J. Struct. Geol. 9, 49 (1987).
      velocity Vw [fig. S5 (24)], where                         is nevertheless expected to be a dominant or                   16. J. H. Dieterich, G. Conrad, J. Geophys. Res. 89, 4196
                                                                contributing dynamic fault-weakening mech-                         (1984).
                         pa rCðTw − Tf Þ 2                      anism in many earthquakes, primarily due to lo-                17. N. M. Beeler, T. E. Tullis, J. Geophys. Res. 102, 22595
                  Vw ¼                                ð1Þ                                                                          (1997).
                         D         tc                           calization of slip (24). For relatively small slips            18. J. D. Byerlee, Pure Appl. Geophys. 116, 615 (1978).
                                                                in small earthquakes, and in the initial slip stages           19. J. R. Rice, J. Geophys. Res. 111, B05311 (2006).
      Here, a is thermal diffusivity, D is contact diam-        of larger earthquakes, flash heating likely dom-               20. N. M. Beeler, T. E. Tullis, D. L. Goldsby, J. Geophys. Res.
      eter, rC is thermal capacity (where r is density          inates the strength of the fault. With continued                   113, B01401 (2008).
                                                                                                                               21. F. Yuan, V. Prakash, Int. J. Solids Struct. 45, 4247
      and C heat capacity), Tw is the characteristic            slip in sufficiently large earthquakes, other dy-                  (2008).
      weakening temperature, and Tf the average                 namic fault-weakening mechanisms may com-                      22. J. A. Greenwood, J. B. P. Williamson, Proc. R. Soc. Lond. A
      temperature of the slip surface. For V > Vw , a           bine with or dominate over flash heating, such                     Math. Phys. Sci. 295, 300 (1966).
      contact spends fractions of its lifetime in its un-       as thermal pore-fluid pressurization (27–29),                  23. An increase in normal stress causes an increase in the
                                                                                                                                   size of existing contacts, as well as the formation of new,
      weakened and weakened states, with the overall            melt lubrication (5, 6), or weakening due to gel                   smaller contacts, such that the average contact size is
      strength a weighted sum of these. Assuming that           formation (7, 8). The low friction and therefore                   independent of normal stress. This behavior has been
      macroscopic shear stress t and nominal normal             modest rise in Tf of faults from flash heating,                    demonstrated for surfaces in elastic contact only in
      stress s on a fault satisfy t/s = (t c)ave /sc = f,       thermal pore-fluid pressurization, and/or silica-                  the referenced paper; similar behavior for plastically
                                                                                                                                   deformed contacts is assumed to underlie the common
      where (t c)ave is the average contact shear stress        gel lubrication could explain the lack of pro-
                                                                                                                                   observation from rock-friction studies that the critical
      and f is the friction coefficient (19), and that          nounced heat flow anomalies along active faults                    sliding displacement Dc , a proxy for average contact size,
      thermal weakening (or even melting) of contacts           known to have slipped seismically, like the San                    is independent of normal stress.
      is confined to a layer that is thin compared to the       Andreas Fault, as well as the relatively rare ob-              24. Materials, methods, and other supporting data are
      contact size (i.e., sc is constant), f as a function of   servance of pseudotachylites (30).                                 available on Science Online.
                                                                                                                               25. H. Blok, Proceedings of the General Discussion on
      slip velocity for V > Vw is                                   Our results have important implications for                    Lubrication and Lubricants (Institute of Mechanical
                                                                the style of dynamic rupture—i.e., whether rup-                    Engineers, London, 1937).
                   f ¼ ð fo − fw Þ      þ fw             ð2Þ    ture occurs as a self-healing slip pulse [Heaton               26. J. C. Jaeger, J. Proc. R. Soc. N.S.W. 76, 203 (1942).
                                     V                          pulse (31)] or a conventional crack. A slip pulse              27. R. H. Sibson, Nature 243, 66 (1973).
                                                                                                                               28. A. Lachenbruch, J. Geophys. Res. 85, 6097 (1980).
      where fo is the friction coefficient at low slip          is a narrow band of self-healing slip that pro-                29. C. W. Mase, L. Smith, Pure Appl. Geophys. 122, 583
      rates and fw that in the weakened state (19).             pagates away from the earthquake focus, such                       (1985).
      Equation 2 is compared with our experimental              that a point on the fault surface slips for only a             30. R. H. Sibson, V. G. Toy, Geophys. Monograph Ser. 170,
      data in Fig. 3, with published values of thermal          small fraction (~10%) of the total earthquake                      153 (2006).
                                                                                                                               31. T. H. Heaton, Phys. Earth Planet. Inter. 64, 1 (1990).
      diffusivity a and thermal capacity ( rC) for the          duration. By contrast, in a cracklike rupture, a               32. H. Noda, E. M. Dunham, J. R. Rice, J. Geophys. Res. 114,
      various minerals composing the rocks (24). We             point on the fault surface experiences slip for                    B07302 (2009).
      assume that (i) sc is equal to the indentation hard-      nearly as long as the earthquake duration. Two                 33. G. Zheng, J. R. Rice, Bull. Seismol. Soc. Am. 88, 1466
      ness H of the most abundant mineral(s) in each rock,      necessary conditions for slip-pulse behavior                       (1998).
                                                                                                                               Acknowledgments: Funding from the U.S. Geological
      such that t c = fcH, where fc is equal to the mac-        are that (i) faults are characterized by velocity-
                                                                                                                                   Survey–National Earthquake Hazards Reduction
      roscopic friction coefficient f measured at slow          weakening friction (32, 33) and (ii) frictional                    Program, Awards G10AP00067, G09AP00053, and
      slip rates; and (ii) contact shear strength is se-        strength recovers on time scales shorter than the                  08-HQGR0067, and the Southern California Earthquake
      verely degraded above Tw ~1000°C. Despite a               rupture duration (31). Both conditions are amply                   Center, from its National Science Foundation grant
      relatively large range of uncertainty (due to con-        satisfied by flash heating, characterized by extreme-              EAR-0529922, is gratefully acknowledged. The data in
                                                                                                                                   this paper are available as supporting material on
      siderable uncertainties in materials parameters),         ly strong velocity-weakening friction and very                     Science Online and archived in the Rock Friction
      Eq. 2 predicts well the values of Vw exhibited by         rapid healing with decreasing slip rate. Dynamic-                  Database in the Rock Deformation Laboratory at Brown
      the data for estimated contact sizes in the range         rupture models incorporating constitutive equa-                    University, Department of Geological Sciences.
      1 to 10 mm. The close fit of Eq. 2 to our data and        tions for flash heating (like Eq. 2) yield slip-pulse
      those from Yuan and Prakash (21) in Fig. 3B               behavior [e.g., (32)]. Flash heating is therefore              Supporting Online Material
      reinforces the extrapolation of our data and ap-          expected to readily give rise to slip-pulse rupture            Materials and Methods
      plication of the Rice theory to higher slip rates.        in nature.                                                     SOM Text
      The similar value of Vw for the different rocks,                                                                         Figs. S1 to S6
      ~ 0.1 m/s, might appear surprising; however, the                                                                         Tables S1 and S2
                                                                    References and Notes                                       References (34–48)
      similar values of indentation hardness and other           1. M. D. Zoback et al., Science 238, 1105 (1987).
      materials parameters (thermal diffusivity, heat            2. A. H. Lachenbruch, J. H. Sass, J. Geophys. Res. 85, 6185   4 May 2011; accepted 19 August 2011
      capacity, and density) for the dominant minerals              (1980).                                                    10.1126/science.1207902

218                                             14 OCTOBER 2011              VOL 334         SCIENCE 
                                                                                                                           and upper M2 phases (78 to 72 ka) contain Still
A 100,000-Year-Old Ochre-Processing                                                                                        Bay–type bifacial foliate points, engraved ochre
                                                                                                                           and bone, bone tools, and Nassarius kraussianus
Workshop at Blombos Cave, South Africa                                                                                     shell beads (2). The M3 phase contains many
                                                                                                                           shellfish, in situ hearths, faunal remains, stone
Christopher S. Henshilwood,1,2* Francesco d’Errico,3,1 Karen L. van Niekerk,1 Yvan Coquinot,4                              tools, and many modified ochre pieces. The ochre-
Zenobia Jacobs,5 Stein-Erik Lauritzen,6 Michel Menu,4 Renata García-Moreno3                                                processing toolkits were recovered in the lower
                                                                                                                           M3 phase. They were found within layer CP,
The conceptual ability to source, combine, and store substances that enhance technology or                                 composed mainly of aeolian dune sand, and lay
social practices represents a benchmark in the evolution of complex human cognition. Excavations                           on a thin orange sand layer, CPA (Fig. 2 and fig.
in 2008 at Blombos Cave, South Africa, revealed a processing workshop where a liquefied                                    S1A). Within layer CP, there are few artifacts
ochre-rich mixture was produced and stored in two Haliotis midae (abalone) shells 100,000                                  apart from the toolkits. Using single-grain opti-
years ago. Ochre, bone, charcoal, grindstones, and hammerstones form a composite part of this                              cally stimulated luminescence (OSL) dating (10),
production toolkit. The application of the mixture is unknown, but possibilities include                                   we estimated the time of deposition of the quartz
decoration and skin protection.                                                                                            sediments in which the ochre containers were
                                                                                                                           buried to be 101 T 4 ka (weighted mean and
         rinding or scraping ochre to produce a                     siltstone). Ochre may have been applied with sym-      standard error for all three samples collected from

G        powder for use as a pigment was com-
         mon practice in Africa and the Near East
after 100,000 years ago (ka) (1–4). Ochre is the
                                                                    bolic intent as decoration on bodies and clothing
                                                                    during the Middle Stone Age (MSA) (2). Archae-
                                                                    ological evidence for the procedures that MSA
                                                                                                                           layer CP). This age is stratigraphically consistent
                                                                                                                           with younger ages for the overlying sediments
                                                                                                                           estimated by OSL dating and by thermolumines-
colloquial term used by archaeologists to describe                  people followed during their handling, prepa-          cence (TL) dating of burnt lithics (11, 12) (Fig. 2
an earth or rock containing red or yellow oxides                    ration, storage, and application of ochre is limited   and table S4). We also dated calcium carbonate
or hydroxides of iron (for example, ferruginous                     (2, 3, 5). Here, we report on the in situ discov-      concretions formed within the layer-CP sediments,
                                                                    ery and subsequent analysis of two coeval, spa-        using uranium-series methods and an isochron
                                                                    tially associated toolkits (Figs. 1 and 2) used for    approach to deal with detrital thorium contam-
 Institute for Archaeology, History, Culture and Religion, Uni-     the production and storage, in shell containers,       ination (13, 14) (figs. S43 to S45 and tables S5
versity of Bergen, Øysteinsgate 3, 5007 Bergen, Norway. 2In-        of an ochre-rich compound at Blombos Cave,             and S6). The most reliable isochron age estimate
stitute for Human Evolution, University of the Witwatersrand,
                                                                    South Africa, 100 ka. Other pigment workshops          of >92 ka is consistent with carbonate formation
Johannesburg, South Africa. 3Université de Bordeaux, UMR 5199
CNRS, Avenue des Facultés, 33405 Talence, France. 4Centre de        and containers date to about 60 ka [e.g. (6)],         after sediment deposition, and should be regarded
Recherche et de Restauration des Musées de France, UMR 171          although older grindstones and hammerstones            as a minimum age for layer CP and its associated
CNRS, Palais du Louvre, 14 Quai François Mitterrand, 75001 Paris,   used for ochre processing have been recovered          artifacts. The most accurate estimate of age for
France. 5Centre for Archaeological Science, School of Earth and     [e.g. (7, 8)].                                         the ochre toolkits is ~100 ka.
Environmental Sciences, University of Wollongong, Wollongong
2522, Australia. 6Department of Earth Science, University of             Blombos Cave is situated on the southern Cape         Toolkit Tk1 comprises a stack of artifacts (Figs.
Bergen, Allégaten 41, 5007 Bergen, Norway.                          coast, 300 km east of Cape Town. The MSA levels        1A and 3) above and below a Haliotis midae
*To whom correspondence should be addressed. E-mail:                at the site are divided into three phases: M1, M2      (abalone) shell (Tk1-S1) (figs. S2 and S3). A                                 (upper and lower), and M3 (9) (Fig. 2). The M1         quartzite cobble (Tk1-L1), tightly fitted within

Fig. 1. Ochre-processing toolkits in situ showing Tk1 (A) and Tk2 (B). [Images: G. Moéll Pedersen]

                                                     SCIENCE        VOL 334        14 OCTOBER 2011                                                     219
      the shell aperture, has use-wear marks consistent    (i) microflakes and microchips of two ferrugi-       the compound, and of crushed compact bone;
      with its application as a percussor and grinder.     nous siltstone (ochre) types (FS1 and FS2) that      (iii) charcoal fragments; (iv) quartz and quartzite
      The upper face is stained with red ochre and         are predominantly present inside the shell, with     microflakes with ochre on some of the striking
      encrusted with fragments of trabecular bone          only minute quantities in the CP layer matrix;       platforms; and (v) quartz grains coated with ochre
      (figs. S7 and S8). Removal of the cobble revealed    FS1 and FS2 are composed of quartz, hematite,        powder and in some instances coated with a micrite
      a 5-mm-thick red compound adhering to the shell      muscovite/illite, and goethite but differ in their   containing hematite, illite/muscovite, quartz, and
      nacre, overlain by a khaki-colored aeolian sand      petrographic structure and elemental compo-          calcium phosphate (figs. S40 and S41).
      (fig. S1C). Microscopic and chemical analysis        sition; (ii) fragments, some apparently burnt, of         A khaki-colored aeolian sand (figs. S2 and S3),
      (15) of the red compound (figs. S5, S6, S36, and     crushed trabecular (spongy) bone, once rich in fat   overlying the red compound, consists of quartz
      S37 and table S3) shows that it is composed of:      and marrow, that may have acted as a binder in       grains, glauconitic grains, vertebrate microfauna

      Fig. 2. South section of Blombos Cave showing layers, phases, and ages. The ages shown here were determined with the OSL, TL, and uranium/thorium (U/Th)
      methods. The ochre-processing toolkits (Tk1 and Tk2) came from layer CP in the M3 phase and are shown in situ. [Image: G. Moéll Pedersen and K. van Niekerk]

220                                         14 OCTOBER 2011           VOL 334       SCIENCE
remains, marine gastropods, ostracods, foram-         above the cobble and below the quartzite slab              face of the second flake (fig. S23). The quartz
inifera, urchin spines, lamellibranches, and cal-     (Tk1-L4) (figs. S13 to S15). Longitudinal streaks          and quartzite microchips found in the red com-
citic worm tubes from annelids (fig. S34). The        of red ochre are visible on one face of this slab.         pound in the Haliotis shell Tk1-S1 are of the same
CP layer matrix (fig. S1) corresponds to this upper   The coarse surface of the slab contains micro-             raw material as Tk1-L5, Tk1-L6, and Tk1-L7.
sand layer (figs. S33 and S35 and tables S1 and       scopic traces of ochre powder, suggesting that it              The second toolkit (Tk2), located 16 cm
S2), and the aeolian quartz grains within the red     was used as a grinder.                                     west of Tk1 (Figs. 1B and 2), comprises a
compound originally derive from the CP layer              Lying below the Haliotis Tk1-S1 were (i) the           Haliotis midae shell (Tk2-S1) (fig. S24), broken
matrix. When the red and aeolian sand layers were     distal portion of a canid ulna (Tk1-B1) with ochre         postdepositionally, with a red compound on the
removed, an orange ring-stain residue consisting      residues on the broken tip and close to the epiph-         nacre of the inner surface of the Haliotis (Fig. 1B
of calcium phosphate mixed with traces of hematite    ysis (fig. S16); (ii) a seal scapula (Tk1-B2) with         and fig. S24). The components of this red com-
and calcite was visible on the shell’s inner lip,     numerous microspots of red ochre on its lateral            pound (fig. S26) are the same as described for
base (fig. S3D), and outer lip (figs. S3C and S6).    surface (fig. S17); (iii) a broken bovid vertebra          Tk1-S1, with the addition of fragments of coarse
The calcium phosphate may result from diage-          (Tk1-B3) (fig. S18); (iv) a quartz flake (Tk1-L5)          silcrete probably originating from a grindstone.
netic microbial activity (fig. S41). A small piece    with red ochre residues and a microchipped edge            The ferruginous siltstone FS2 was not detected in
of FS1 red ochre (Tk1-P1), rubbed on one face,        indicating its use as a grinder (fig. S19); and (v) two    this red compound (table S2). On the ventral side
was found on the inner lip of the shell (figs. S9     quartzite flakes (Tk1-L6 and Tk1-L7) (figs. S20 to         and close to the outer lip of Tk2-S1, ancient stri-
and S10). A quartzite flake fragment (Tk1-L2),        S22). The striking platform of the first flake is cov-     ations are present on the nacre (fig. S25). Mi-
with ochre powder on all faces, was adhering to it    ered with red ochre powder and is microchipped             croscopic observation suggests that they were
(figs. S11 and S12). A small quartzite flake (Tk1-    on one edge, suggesting its use as a grinder (fig. S21).   produced when quartz or ochre grains were gently
L3) with red ochre on the striking platform lay       The same ochre powder is present on the cortical           moved across the nacre surface during mixing.

Fig. 3. Artifacts making up Tk1 and their relative spatial locations. [Image: C. Henshilwood and F. d’Errico]

                                         SCIENCE         VOL 334        14 OCTOBER 2011                                                    221
      After the shell containing the red compound was         reuse to grind red ochre. The ochre FS2 is present                    9. C. S. Henshilwood et al., J. Archaeol. Sci. 28, 421 (2001).
      abandoned, it was covered with aeolian dune sand        only in Tk1, and the stone tools found in close                      10. Z. Jacobs, R. G. Roberts, Evol. Anthropol. Issues News Rev.
                                                                                                                                       (Melb.) 16, 210 (2007).
      derived from layer CP, as was the case for Tk-S1.       association with each shell may have been ex-                        11. Z. Jacobs, G. A. T. Duller, A. G. Wintle, C. S. Henshilwood,
          A small quartzite core (Tk2-L1) rested on the       clusive to the processing related to that shell.                         J. Hum. Evol. 51, 255 (2006).
      shell nacre close to the anterior edge (figs. S24       However, the close proximity of the two toolkits                     12. C. Tribolo et al., Archaeometry 48, 341 (2006).
      and S27). The core was used first to grind ferru-       suggests that they were used contemporaneously.                      13. H. P. Schwarcz, Archaeometry 22, 3 (1980).
                                                                                                                                   14. D. A. Richards, J. A. Dorale, Rev. Mineral. Geochem. 52,
      ginous lutites, composed in one case of goethite,       Because both toolkits were left in situ, and be-                         407 (2003).
      calcite, and quartz, and in the other of hematite,      cause there are few other archaeological remains                     15. Information on materials and methods is available on
      calcite, and quartz (figs. S28 and S29 and table        in the CP layer, it seems that the site was used                         Science Online.
      S2). Several flakes were then removed from the          primarily as a workshop and was abandoned                            16. J. S. Compton, Quat. Sci. Rev. 30, 506 (2011).
                                                                                                                                   17. B. M. Henn et al., Proc. Natl. Acad. Sci. U.S.A. 108,
      utilized area, and the object was again used to         shortly after the compounds were made. Aeolian                           5154 (2011).
      grind a type FS1 red ferruginous siltstone. A           sand then blew into the cave from the outside,                       18. Q. D. Atkinson, Science 332, 346 (2011).
      large fragment of ochre (Tk2-P1) composed of            encapsulating the toolkits (Fig. 2).                                 Acknowledgments: C.S.H. and F.D. received funding from
      ferruginous siltstone FS1 lay 5 cm southwest of             Recent support for a southern African origin                         the European Research Council (ERC) under the European
                                                                                                                                       Union's Seventh Framework Programme (FP7/2007-2013)/
      the shell (fig. S30). It was knapped to produce         for Homo sapiens comes from genomic and phe-
                                                                                                                                       ERC grant agreement no. 249587, the PROTEA
      small flakes, similar to those in the red compound      nomic diversity studies (17, 18). The recovery of                        French–South African research program, and the Groupe
      from Tk1-S1 and Tk2-S1. The piece was also              these toolkits at Blombos Cave adds evidence for                         de Recherche International STAR of the CNRS. C.S.H. was
      rubbed against a hard stone to produce ochre            early technological and behavioral developments                          also funded by a National Research Foundation/
      powder (figs. S31 and S32).                             associated with H. sapiens and documents their                           Department of Science and Technology–supported chair
                                                                                                                                       at the University of the Witwatersrand, South Africa, and
          The two Haliotis shells derive from the infra-      deliberate planning, production, and curation of a                       by a joint Norwegian Research Council/ South African
      tidal zone, at that time a few hundred meters from      pigmented compound and the use of containers.                            National Research Foundation grant; and Z.J. was funded
      the cave (16). Before their use as containers, the      H. sapiens thus also had an elementary knowl-                            by Australian Research Council Discovery project grant
      respiratory holes of the Haliotis were possibly         edge of chemistry and the ability for long-term                          DP1092843. The data described in the paper are
                                                                                                                                       presented in the supporting online material. C.S.H. and
      plugged. When recovered, these holes were filled        planning.                                                                K.vN. excavated the toolkits; C.S.H., F.D., K.vN., and
      with detritus (fig. S6), but this could have occurred                                                                            R.G.-M. wrote the text; Z.J. and S.-E.L. wrote the
      postdepositionally. The ochre and silcrete were             References and Notes                                                 dating text; and all authors were involved in aspects
      sourced from at least several kilometers away (9),       1. E. Hovers, S. Ilani, O. Bar-Yosef, B. Vandermeersch,                 of the data analysis. All the authors discussed the
                                                                  Curr. Anthropol. 44, 491 (2003).                                     results and implications and commented on the
      and the rest of the objects that make up the tool-       2. C. S. Henshilwood, F. d’Errico, I. Watts, J. Hum. Evol.              manuscript at all stages. The authors declare no
      kits were available in the immediate environment.           57, 27 (2009).                                                       competing interests.
          We infer that manufacturing proceeded as             3. I. Watts, J. Hum. Evol. 59, 392 (2010).
      follows: Pieces of ochre (FS1 and FS2) were              4. F. d'Errico, H. Salomon, C. Vignaud, C. Stringer,
                                                                  J. Archaeol. Sci. 37, 3099 (2010).
                                                                                                                                   Supporting Online Material
      rubbed on quartzite slabs to produce a fine red          5. L. Wadley, T. Hodgskiss, M. Grant, Proc. Natl. Acad.
      powder, and some were knapped with large lithic             Sci. U.S.A. 106, 9590 (2009).                                    Materials and Methods
                                                                                                                                   SOM Text
      flakes. The ochre chips resulting from the latter        6. L. Wadley, J. Archaeol. Sci. 37, 2397 (2010).
                                                                                                                                   Figs. S1 to S45
      were crushed with quartz, quartzite, and silcrete        7. A. Šajnerová-Dušková, J. Fridrich, I. Fridrichová-Sýkorová,
                                                                                                                                   Tables S1 to S6
                                                                  in Non-flint Raw Material Use in Prehistory: Old Prejudices
      hammerstones/grinders. Quartzite grinders were              and New Directions, F. Sternke, L. J. Costa, L. Eigeland, Eds.
                                                                                                                                   References (19–50)
      used to crush goethite or hematite-rich lutite.             (Archeopress, Oxford, 2009), pp. 145–151.                        21 July 2011; accepted 31 August 2011
      Medium-sized mammal bone was crushed, prob-              8. P. Van Peer et al., J. Hum. Evol. 45, 1 (2003).                  10.1126/science.1211535
      ably with a stone hammer. The red or reddish
      brown color and cracked, flaky texture of some
      of the trabecular bone suggest that it was heated
      before crushing, probably to enhance the extrac-
      tion of the marrow fat. The hematite powder,
                                                              The Dynamic Architecture of
      charcoal, crushed trabecular bone, stone chips, and
      quartz grains and a liquid were then introduced
                                                              Hox Gene Clusters
      into the Haliotis shells and gently stirred (figs.
      S5, S25, and S26). Charcoal is rare in the layer-       Daan Noordermeer,1 Marion Leleu,1 Erik Splinter,2 Jacques Rougemont,1,3
      CP matrix, suggesting that it was a deliberate          Wouter De Laat,2 Denis Duboule1,4*
      addition to the mix. The quartz and quartzite chips,
      produced during the action of crushing the ochre,       The spatial and temporal control of Hox gene transcription is essential for patterning the vertebrate
      and the quartz grains may have been incidentally        body axis. Although this process involves changes in histone posttranslational modifications,
      incorporated.                                           the existence of particular three-dimensional (3D) architectures remained to be assessed in vivo.
          The application or use of the compound is not       Using high-resolution chromatin conformation capture methodology, we examined the spatial
      self-evident. No resins or wax were detected that       configuration of Hox clusters in embryonic mouse tissues where different Hox genes are active.
      might indicate it was an adhesive for hafting.          When the cluster is transcriptionally inactive, Hox genes associate into a single 3D structure
      Possible uses could include painting a surface in       delimited from flanking regions. Once transcription starts, Hox clusters switch to a bimodal 3D
      order to decorate or protect it, or to create a de-     organization where newly activated genes progressively cluster into a transcriptionally active
      sign. Ochre residues on the bone Tk1-B1 show            compartment. This transition in spatial configurations coincides with the dynamics of chromatin
      that it was possibly used as a stirrer and also to      marks, which label the progression of the gene clusters from a negative to a positive transcription
      transfer some of the compound out of the shell.         status. This spatial compartmentalization may be key to process the colinear activation of these
      At least some of the components of the toolkit          compact gene clusters.
      were reused, suggesting that production was not
      a one-time event. An example is the first use of                  uring mammalian development, Hox genes                     (HoxA to HoxD). As a result, this process leads to
      the grinder Tk2-L1 to grind yellow goethite, its
      subsequent reduction by flaking, and then its           D         are activated sequentially relative to their
                                                                        positions along the four genomic clusters
                                                                                                                                   a corresponding distribution of transcripts along
                                                                                                                                   the rostral-to-caudal body axis. This process of

222                                            14 OCTOBER 2011              VOL 334           SCIENCE  
colinear activation is essential for the organiza-                the HoxD cluster in forebrain cells, where all Hox                modal profiles of association, dividing the gene
tion of the body plan (1, 2). Accompanying this                   genes are inactive. Seven different viewpoints                    cluster into two distinct 3D compartments. How-
process, a dynamic transition occurs in the chro-                 revealed comparable domains of 3D association,                    ever, the boundaries between these two compart-
matin microenvironment, from a repressive (his-                   spanning from Evx2 to a few kilobases down-                       ments were located at different positions in anterior
tone H3K27me3) to a transcription-permissive                      stream of Hoxd1 (Fig. 1C and fig. S7). Likewise,                  versus posterior trunk samples. In anterior trunk,
(histone H3K4me3) state (3). Changes in higher-                   three viewpoints within each of the other Hox                     transcribed genes like Hoxd4 no longer contacted
order chromatin organization have been reported                   clusters uncovered association domains covering                   the silent (centromeric) part of the cluster, thus
to accompany the transcription of developmen-                     the entire clusters plus a few kilobases on either                forming an “active domain” (Fig. 2A). Accord-
tally relevant genes (4), and the 3D organization                 side (fig. S8). Therefore, silent Hox clusters form               ingly, Hoxd13, the most centromeric gene, no
of the HoxA cluster is changed upon gene ac-                      3D compartments with discrete separation from                     longer contacted the telomeric part of the clus-
tivation in mammalian cultured cells (5–7). Fur-                  flanking DNA regions. Little specific organiza-                   ter (Fig. 2C, “inactive domain”). The transition of
thermore, the HoxB and HoxD clusters adopt a                      tion was scored within these domains, suggesting                  genes from an inactive to an active 3D domain
decondensed conformation along with gene ac-                      mostly random contacts. Furthermore, these as-                    was best exemplified by Hoxd9, expressed strong-
tivation (8, 9) accompanied by modifications of                   sociation domains precisely matched the distri-                   ly in posterior trunk cells but only weakly in the
the Polycomb repressive complex 1, as shown                       bution of H3K27me3 marks decorating these loci                    anterior sample (Fig. 2B and fig. S1). Contacts
in cultured cells (10). We analyzed the architec-                 (Fig. 1C and figs. S7 to S9), both in the positions               of Hoxd9 were stronger with the centromeric
tures of these genomic loci in embryonic tissues                  of the borders and in the organization within each                part of the cluster in anterior cells (gene mostly
at different stages of the colinear transcriptional               cluster, supporting a functional interplay between                off; in the “inactive domain”), but they clearly
activation and describe a bimodal state, where                    these two parameters in vivo (10).                                shifted toward the telomeric part of the cluster
active and inactive genes are found in distinct                       We examined these architectures in “anterior”                 in posterior trunk cells (gene on, “active domain”;
three-dimensional (3D) domains and genes pro-                     and “posterior” embryonic tissue samples, where                   Fig. 2B, see ratio). The same was observed for
gressively lose their interactions with the repres-               different HoxD genes are transcribed at this stage                Hoxd11, which strongly contacted the negative
sive domain to associate with a transcriptionally                 of development (Fig. 2 and fig. S10). In contrast                 domain in anterior trunk cells (off state), whereas
active structure.                                                 to the single interaction domain observed in brain                most contacts were with the positive domain in
     We used gene expression microarrays to com-                  cells (Fig. 2, A to C, green), both anterior (in red)             posterior trunk cells (on state; fig. S10). This bi-
pare Hox gene activity in three tissue samples                    and posterior (in blue) trunk cells generated bi-                 modal organization also applied to the other three
obtained from embryonic day 10.5 (E10.5) mouse
embryos: “anterior” dorsal trunk cells (from upper
forelimb to upper hindlimb levels), “posterior”                   A                                                        C     42k
dorsal trunk cells (from upper forelimb level to                    Forebrain
tailbud), and forebrain cells. The latter cells do
not express any Hox genes and were used as                                                                                  Fore-
negative control (Fig. 1A, fig. S1, and table S1).
We determined which genes were either tran-
scribed or silent in these samples and positioned                   Anterior                                                       0
                                                                    trunk                                                        55k
the dissection limit between the two trunk sam-
ples approximately at the level of the Hoxd10
expression boundary (Fig. 1A, arrowheads; fig.                                                                              Fore-
S1 and table S2).                                                                                d9          d11
                                                                    Posterior                          d10
     Using these tissue samples, we examined
the 3D architecture by high-resolution chromatin                                                                                   0
conformation capture [Multiplex 4C-seq (11, 12);                  B                                                              52k
fig. S2 and table S3] with multiple genes as an-
chor points (“viewpoints”), taken in all four Hox                                                                           Fore-
clusters (figs. S3 to S5 and table S4). Although                  Fore-                                                     brain
most of the intrachromosomal associations were
restricted to the Hox clusters themselves (Fig. 1B                                                                               0
and fig. S6), a statistical algorithm reliably iden-                    0
tified nondynamic long-range interaction land-                    Interacting                                               me3
                                                                  regions                                                        0
scapes surrounding each locus and extending                                             Atp5g3        HoxD    Hnrnpa3                                 d13 d11   d9   d4 d3      d1
slightly past the flanking gene deserts (13), which               HoxD                                                      HoxD
may reflect a generic organization of Hox clus-                    Chr 2                                                       Chr 2
                                                                                73.0       74.0      75.0        76.0                            74.5                          74.6
ters and their surroundings in these cells (Fig. 1B,                                   Chromosomal Position (Mb)                                        Chromosomal Position (Mb)
fig. S6, and table S5).
                                                                  Fig. 1. The inactive HoxD cluster forms a discrete 3D compartment. (A) Schematized E10.5 mouse
     We quantified intracluster 3D organization at
                                                                  embryo highlighting tissue samples used in this work: forebrain (green), anterior trunk (red), and
highest resolution [figs. S5 and S7 (11)], using
                                                                  posterior trunk (blue). Approximate positions of expression boundaries for either Hoxd9 (d9), Hoxd10
                                                                  (d10), or Hoxd11 (d11) are indicated with arrowheads. (B) Running-mean 4C-seq interaction patterns
 National Research Centre “Frontiers in Genetics,” School of      of Hoxd9 in a 4-Mb-large genomic region surrounding the HoxD cluster in forebrain tissue. Significant
Life Sciences, Ecole Polytechnique Fédérale (EPFL), Lausanne,     interactions are depicted below (see also fig. S6). The position of the HoxD cluster is shown in red,
CH-1015, Switzerland. 2Hubrecht Institute and University Med-
                                                                  flanked by two gene deserts (below). (C) Quantitative local 4C-seq interactions reveal the local 3D
ical Center, Utrecht, 3584 CT, Netherlands. 3Swiss Institute of
Bioinformatics, Lausanne, CH-1015, Switzerland. 4Department       domain of the inactive HoxD cluster (forebrain). Below, H3K27me3 signal is aligned. Dashed lines
of Genetics and Evolution, University of Geneva, CH-1211,         emphasize the discrete borders of the local 3D domain and the coincidence with the H3K27me3
Switzerland.                                                      domain. Three viewpoints are used (Hoxd13, Hoxd9, and Hoxd4, from top to bottom) and are indicated
*To whom correspondence should be addressed. E-mail:              with arrowheads. Excluded regions around these viewpoints are depicted with vertical light gray boxes. or                   The locations of Hoxd genes (red) and of other transcripts (black) are shown below.

                                                       SCIENCE               VOL 334        14 OCTOBER 2011                                                     223
      A                                              Hoxd4                       B                                Hoxd9                                  C               Hoxd13
                   52k                                                                55k                                                                    42k


                   0                                                                    0                                                                     0

                    8                                                                   8                                                                     8
       FB / AT      0                                                                   0                                                                     0
                    -8                                active domain                    -8            inactive domain                                          -8         inactive domain
                   52k                                                                55k                                                                    42k


                   0                                                                    0                                                                     0
                   30                                                                  30                                                                    30
                    0                                                                   0                                                                     0
                   30                                                                  30                                                                    30
                    0                                                                   0                                                                     0
                    8                                                                   8                                                                     8
       AT / PT      0                                                                   0                                                                     0
                    -8                                                                 -8                                   active domain                     -8
                   52k                                                                55k                                                                    42k


                   0                                                                    0                                                                     0

                                      d13 d11   d9     d4 d3          d1                                    d13 d11    d9        d4 d3       d1                                 d13 d11    d9    d4 d3    d1
                               Evx2                                                                  Evx2                                                                Evx2
            HoxD                                                                   HoxD                                                                  HoxD

                         cen    74.5                              74.6     tel              cen       74.5                                  74.6   tel             cen    74.5                           74.6   tel

      Fig. 2. Dynamic architectures of the HoxD cluster at different stages of                                        active; red, inactive) of Hoxd genes (bottom line) is schematized below each
      colinear activation. The frequencies of associations are shown with Hoxd4                                       profile. For the anterior trunk sample (in red), the corresponding H3K27me3
      (A), Hoxd9 (B), or Hoxd13 (C) as viewpoints in forebrain (profiles in green),                                   and H3K4me3 signals are indicated just below (profiles in black). Residence of
      anterior trunk (profiles in red), or posterior trunk (profiles in blue) tissues.                                viewpoints in the active [(A), anterior trunk Hoxd4 and (B), posterior trunk
      The ratios of interactions are indicated between the profiles for the same                                      Hoxd9] and inactive [(B), anterior trunk Hoxd9 and (C), anterior trunk Hoxd13]
      viewpoint in different embryonic tissues. The colinear expression status (blue,                                 domains is indicated with arrowheads.

      Hox clusters, with slight variations in the location
      of the internal boundary, depending on the pro-
      gression of gene activation within each cluster
      (fig. S11).
          In the trunk samples, not only did we observe
      a coincidence between the inactive 3D domain
      and the extent of H3K27me3 modifications, as in
      the brain sample, but the active compartments
      also matched the presence of H3K4me3 chroma-
      tin domains (Fig. 2, A to C, and figs. S9 to S11).
      The distribution of these chromatin marks cor-
      related with the 3D organization at these Hox
      clusters. In this context, the HoxB cluster was                                             HoxA, HoxC, HoxD                                                                        HoxB
      particularly interesting because an 80-kb, repeat-                         Fig. 3. Model of the 3D organization of Hox gene clusters, at various stages of colinear gene activation.
      rich intergenic region separates Hoxb13 from                               Transcriptionally inactive genes are depicted in red and active genes in blue. Gene activation is
      the rest of the cluster (14). Using viewpoints in                          paralleled by a transition from one 3D domain, matching the presence of H3K27me3, to another
      the cluster and within this intergenic region, we                          domain of active transcription (marked with H3K4me3). Although the same dynamics are observed for
      observed a weak association only (if any) be-                              the HoxA, HoxC, and HoxD clusters (left), the HoxB cluster (right) shows a slight variation with a large
      tween this region and the rest of the HoxB clus-                           piece of intergenic DNA looping out from these two domains.
      ter (fig. S12), suggesting that it loops out from
      this bimodal architecture. The same interruption
      was seen in the distribution of H3K27me3 marks                             3 Mb and both heavily decorated with H3K27me3                           Hox gene activities are differentially perturbed
      (fig. S12), illustrating again the precise corre-                          marks (fig. S13). In contrast, such interactions                        (15). Deletion of the Hoxd8 to Hoxd10 DNA
      spondence between chromatin marks and 3D                                   were not scored with the active part of the HoxD                        fragment [Del(8–10)] does not severely change
      architecture and showing that these spatial do-                            cluster.                                                                the expression of Hoxd11 (fig. S14). However,
      mains do not necessarily involve an uninterrupted                              From these data sets, we propose a model                            the additional deletion of the intergenic region
      linear chromatin fiber. This latter conclusion                             whereby Hox genes move stepwise from an in-                             i [Del(i8-10)] results in a strong activation of
      was further illustrated by strongly increased and                          active compartment marked by H3K27me3 to                                Hoxd11 in anterior tissues (fig. S14). These over-
      targeted associations between the inactive do-                             another, transcriptionally active domain labeled                        lapping deletions thus have distinct transcriptional
      main of the HoxD cluster (Hoxd13) and the                                  with H3K4me3 marks (Fig. 3). We challenged                              outcomes, with Hoxd11 ectopically activated in
      Dlx1 locus, which are separated by a distance of                           this view by using two deletions in vivo where                          the anterior trunk sample of Del(i8-10) embryos

224                                                       14 OCTOBER 2011                    VOL 334             SCIENCE       
A                                      Forebrain                                     B                                 Anterior trunk                               C                         Anterior trunk
        48k                                                                                    46k                                                                    65k

          0                                Del(i8-10)
                                                                                                   0                          Del(i8-10)                                0                            Del(i8-10)
Del(i8-10)                                                                                         8                                                                    8
                                           Del(i8-10)                                 Ratio
          0                                                                           Del(i8-10) /
        48k                                                                                        0                                                                    0
                                                                                                 -8                                                                    -8
                                                                                               46k                                                                    65k

          0                                                                                      0                                                                      0
        48k                                                                                      8                                                                      8
                                            Del                                       WT /
                                                                                                 0                                                                      0
                                           (8-10)                                     Del(8-10)
        56k                                                                           (log2)
                                                                                                -8                                                                     -8
Del(8-10)                                                                                     46k                                                                     65k
             0                             (8-10)
                                 d13 d11     d9         d4   d3    d1
                          Evx2                                                                                                 Del                                                                    Del
    HoxD                                                                                                                      (8-10)                                                                 (8-10)
                                                                                                   0                                                                    0
                                                                                                                    d13 d11     d9         d4   d3     d1                                  d13 d11     d9         d4   d3    d1
                 cen          74.5                                74.6      tel                              Evx2                                                                   Evx2
                                                                                          HoxD                                                                      HoxD
                 Del(i8-10)                  Wildtype                    Del(8-10)
                                                                                                       cen    74.5                                    74.6   tel            cen       74.5                                  74.6   tel

Fig. 4. Ectopic activation of Hoxd11 in anterior tissue increases its asso-                                              mutant anterior tissue, but not in the Del(8-10) mutant (see figs. S14 to
ciation frequency with other active Hoxd genes. (A) Three-dimensional or-                                                S17). Accordingly, increased association frequencies are observed between
ganization of the inactive HoxD cluster in wild-type forebrain (blue profiles,                                           Hoxd11 and the active part of the cluster in the Del(i8-10) mutant (purple
middle) or in forebrain tissues carrying two distinct internal deletions into                                            box and ratio), as compared to both wild-type (in blue) and the Del(8–10)
the HoxD cluster [Del(i8-10), profiles in purple on the top, and Del(8-10),                                              mutant embryos (yellow box and ratio). (C) Same experiment as in (B), but
profiles in yellow, bottom]. The deletions are indicated between small ar-                                               with Hoxd4 as a viewpoint (arrowhead). Again, interactions are increased
rowheads and, for each sample, both Hoxd11 and Hoxd4 are used as view-                                                   between Hoxd4 and the posterior part of the HoxD cluster (purple box and
points (large arrowheads). Regardless of cluster size, the 3D inactive domains                                           ratio) containing Hoxd11, which is ectopically expressed in the mutant
remain demarcated by the same outside borders. (B) Three-dimensional                                                     Del(i8-10) anterior tissue. In contrast, the Del(8-10) mutant tissue, where
organization of mutant HoxD clusters in anterior trunk with Hoxd11 as a                                                  Hoxd11 is not expressed anteriorly, does not show such increased inter-
viewpoint (arrowhead). Hoxd11 is expressed ectopically in the Del(i8-10)                                                 actions (yellow box and ratio).

only. We first assessed whether such deletions                                    sequence of it, it is noteworthy that the former                                 13. A. P. Lee, E. G. Koh, A. Tay, S. Brenner,
had changed the overall cluster architecture in                                   possibility would provide a mechanistic solu-                                        B. Venkatesh, Proc. Natl. Acad. Sci. U.S.A. 103,
                                                                                                                                                                       6994 (2006).
brain cells (Fig. 4A and figs. S15 and S16) and                                   tion to three crucial problems encountered dur-                                  14. L. Zeltser, C. Desplan, N. Heintz, Development 122,
observed that the inactive domains maintained                                     ing the activation of this gene family: (i) to ensure                                2475 (1996).
the same borders on both sides, indicating that                                   a proper colinear sequence in gene activation,                                   15. P. Tschopp, B. Tarchini, F. Spitz, J. Zakany, D. Duboule,
the mechanism underlying this 3D compartmen-                                      such that axial morphologies are respected [see                                      PLoS Genet. 5, e1000398 (2009).
                                                                                                                                                                   16. D. M. Wellik, Dev. Dyn. 236, 2454 (2007).
talization is likely intrinsic to the gene cluster.                               e.g., (16)]; (ii) to prevent the most posterior genes                            17. T. Young et al., Dev. Cell 17, 516 (2009).
We then studied the interaction profiles using                                    from being activated too early, which leads to                                   Acknowledgments: We thank B. Mascrez for assistance
Hoxd11 and Hoxd4 as viewpoints in anterior trunk                                  deleterious phenotypes (17); and (iii) to fix and                                    with mouse handling and genotyping, P. Descombes
samples (Fig. 4, B and C, and fig. S17). In the                                   memorize transcriptional states at various body                                      and members of the National Research Centre genomics
                                                                                                                                                                       platform for high-throughput sequencing, and members of
Del(i8-10) mutant, where Hoxd11 is ectopically                                    levels. These critical constraints are well addressed                                the Duboule laboratories for discussion. Computations were
activated (15), the association between Hoxd11                                    by our cis-acting model, whereas other potential                                     performed at the Vital-IT Center for high-performance
and the “positive” compartment was strongly in-                                   mechanisms, such as relying upon trans-acting                                        computing ( at the Swiss Institute of
creased. This was scored either by using Hoxd11                                   interactions, may not allow the same level of pre-                                   Bioinformatics. This work was supported by funds from
                                                                                                                                                                       the Ecole Polytechnique Fédérale (Lausanne), the Uni-
as a viewpoint (Fig. 4B, shaded purple/blue ra-                                   cision and reliability.
                                                                                                                                                                       versity of Geneva, the Swiss National Research Fund, the
tio), or Hoxd4 (Fig. 4C, shaded purple/blue ra-                                                                                                                        National Research Centre “Frontiers in Genetics,” and the
tio). However, contacts remained as in wild-type                                          References and Notes                                                         European Research Council grant (to D.D.).
embryos when the shorter Del(8–10) deletion                                           1. M. Kmita, D. Duboule, Science 301, 331 (2003).                                Data are all based on ENSEMBL Mouse assembly
was analyzed with the same viewpoints (Fig. 4, B                                      2. R. Krumlauf, Cell 78, 191 (1994).                                             NCBIM37. 4C-seq patterns can be obtained from
                                                                                      3. N. Soshnikova, D. Duboule, Science 324, 1320                         or
and C, shaded blue/yellow ratios). Ectopic activa-                                                                                                                     page-66605-en.html. Microarray and ChIP-seq data
tion, rather than a deletion per se, was thus par-                                    4. R. J. Palstra et al., Nat. Genet. 35, 190 (2003).                             have been submitted to the Gene Expression Omnibus
alleled by enhanced association between Hoxd11                                        5. M. A. Ferraiuolo et al., Nucleic Acids Res. 38, 7472                          (GEO) repository ( under
and the “active” anterior domain.                                                        (2010).                                                                       accession no. GSE31570.
    This work suggests that the colinear activa-                                      6. J. Fraser et al., Genome Biol. 10, R37 (2009).
                                                                                      7. K. C. Wang et al., Nature 472, 120 (2011).                                Supporting Online Material
tion of Hox genes involves a stepwise transition                                      8. S. Chambeyron, W. A. Bickmore, Genes Dev. 18,                   
of each gene from a negative to a positive com-                                          1119 (2004).                                                              Materials and Methods
partment, which display different biochemical                                         9. C. Morey, N. R. Da Silva, M. Kmita, D. Duboule,                           Figs. S1 to S17
properties and thus results in a physical separa-                                        W. A. Bickmore, J. Cell Sci. 121, 571 (2008).                             Tables S1 to S6
                                                                                     10. R. Eskeland et al., Mol. Cell 38, 452 (2010).                             References (18–24)
tion of their regulatory modalities. Although it                                     11. Material and methods are available as supporting
remains to be fully demonstrated whether such a                                          material on Science Online.                                               19 April 2011; accepted 17 August 2011
process underlies colinear activation or is a con-                                   12. M. Simonis et al., Nat. Genet. 38, 1348 (2006).                           10.1126/science.1207194

                                                                            SCIENCE           VOL 334                    14 OCTOBER 2011                                                                     225
                                                                                                                         single whisker is processed in the corresponding
      Early Gamma Oscillations Synchronize                                                                               developing cortical column.
                                                                                                                             We found that in postnatal days 2 to 7 (P2 to
      Developing Thalamus and Cortex                                                                                     P7), rats brief single principal whisker (PW) de-
                                                                                                                         flections evoke an oscillatory local field potential
      Marat Minlebaev,1,2 Matthew Colonnese,1,2 Timur Tsintsadze,1,2                                                     (LFP) response in the gamma band in the corre-
      Anton Sirota,3* Roustem Khazipov1,2*†                                                                              sponding cortical barrel (peak frequency 55 T2
                                                                                                                         Hz; n = 45 rats) (Fig. 1 and fig. S1). Multiunit
      During development, formation of topographic maps in sensory cortex requires precise                               activity (MUA), gamma oscillation power, and
      temporal binding in thalamocortical networks. However, the physiological substrate for such                        the current sinks of these early gamma oscillations
      synchronization is unknown. We report that early gamma oscillations (EGOs) enable precise                          (EGOs) were maximal in the granular (Gr) layer
      spatiotemporal thalamocortical synchronization in the neonatal rat whisker sensory system.                         (Fig. 1, B to D, and Fig. 2C). EGOs were time-
      Driven by a thalamic gamma oscillator and initially independent of cortical inhibition, EGOs                       locked to the stimulus and were apparent in the
      synchronize neurons in a single thalamic barreloid and corresponding cortical barrel and support                   stimulus-triggered average LFP and MUA time
      plasticity at developing thalamocortical synapses. We propose that the multiple replay of                          histograms (Fig. 1B). Gr MUA was strongly
      sensory input in thalamocortical circuits during EGOs allows thalamic and cortical neurons                         phase-modulated by EGOs and occurred during
      to be organized into vertical topographic functional units before the development of horizontal                    the descending phase and troughs of gamma
      binding in adult brain.
                                                                                                                          INSERM U901, 163 Avenue de Luminy, B.P. 13, 13273
             ensory cortex is organized as a topographic          corresponding whisker (2). Development of the          Marseille, France. 2Université Aix-Marseille, 163 Avenue de

      S      map consisting of arrays of columns that
             each receive an input from a particular re-
      gion of sensory space (1). In the rodent “barrel”
                                                                  topographic thalamocortical connections depends
                                                                  critically on activity driven by the whiskers (2–4).
                                                                  During development, sensory input triggers vari-
                                                                                                                         Luminy, 13273 Marseille, France. 3University of Tuebingen
                                                                                                                         Center for Integrative Neuroscience, Paul-Ehrlich Strasse 15,
                                                                                                                         Tuebingen 76072, Germany.
                                                                                                                         *These authors contributed equally to this work.
      cortex, each cortical barrel column receives a              ous cortical activity patterns (5–8). However, it      †To whom correspondence should be addressed. E-mail:
      specific input, conveyed via the thalamus, from a           remains unknown how the sensory input from a 

      Fig. 1. Brief single-whisker deflection evokes gam-
      ma oscillations in the corresponding cortical barrel
      of a P6 neonatal rat. (A) Scheme of the experi-
      mental setup, with principal barrel location detected
      by intrinsic optical signal imaging (IOS) shown be-
      low. (B) (Top left) Recording sites of a multielectrode
      array overlaid on a Ctip2-stained cortical slice. IG, in-
      fragranular. (Top right) Sensory response evoked by
      C2 PW stimulation at different depths of the C2 cor-
      tical barrel column. LFP, black traces; MUA, red bars
      overlaid on color-coded current-source density (CSD)
      plot. (Bottom) The stimulus-triggered averages (n =
      100) for Gr layer (red asterisk) wavelet spectrogram,
      average Gr layer LFP, and MUA peristimulus time
      histograms (PSTHs) across layers. (C) Power spectral
      density of Gr layer LFP during a 200-ms time win-
      dow [white horizontal bar in (B)] following PW or
      AW sensory evoked potential (SEP) or during a prestim-
      ulus epoch. Shading and error bars hereafter show
      jackknife standard deviation. (D) Gamma-trough trig-
      gered average Gr LFP and CSD map.

226                                              14 OCTOBER 2011              VOL 334       SCIENCE
cycles (figs. S1 and S2C). Day-by-day analysis           and were maximal in SG layers. Simultaneous            P7 (n = 30), PW stimulation evokes stimulus-
revealed that power in the gamma band increased          recordings from neighboring barrel columns in          locked gamma-rhythmic excitatory postsynaptic
from birth to attain maximal values at P2 to P7          P19 to P33 rats (n = 6) revealed strong hori-          currents (EPSCs) tightly locked to the field EGO’s
and abruptly declined at P8, followed by a grad-         zontal gamma synchronization of units (Fig. 2,         troughs (Fig. 3A and fig. S4, A and B). These
ual increase both in spontaneous and sensory-            A and B, and fig. S2). Thus, during develop-           gamma-rhythmic EPSCs were not present in P10
evoked activity in the gamma frequency range             ment, sensory-evoked gamma activity switches           to P11 rats (n = 3), in agreement with the dis-
(Fig. 2D and figs. S1 and S3). Simultaneous record-      from transient input-specific “vertical” EGOs to       appearance of field EGOs after P8. The engage-
ings in Gr layer of two neighboring barrel col-          “horizontal” gamma oscillations synchronizing          ment of inhibitory circuits into EGOs followed a
umns revealed that EGOs are restricted to the            activity in neighboring cortical columns. The          different age-dependent track. At P2 to P3, only
principal barrel (n = 13) (P5 to P7) (Fig. 2, A, C,      switch is characterized by a developmental gap         three of five cells displayed any inhibitory post-
and D, and fig. S2).                                     between these two forms of gamma activity:             synaptic currents (IPSCs), and these occurred at
    We further addressed the development of              EGOs disappear around P8, whereas adult gam-           the end of EGOs. By P5 to P7, IPSCs showed
activity transfer from Gr to the downstream su-          ma activity gradually builds up along with a de-       gamma rhythmicity during EGOs (Fig. 3, A and
pragranular (SG) layers and of horizontal synchro-       velopment of the active cortical state, manifested     B, and fig. S4, A and B), which was also reflected
nization by gamma oscillations in the neighboring        by an increase in background activity (Fig. 2, D       in firing pattern and gamma-rhythmic EPSCs in a
columns. During the first postnatal week, PW-            and E, and figs. S1 and S3) (8, 11, 12) and with       fast-spiking interneuron at P7 (fig. S5). Cells in
evoked responses in SG layers were weak and              the development of explorative behaviors and ac-       SG layers received only weak and largely sub-
virtually no response was seen to adjacent whisker       tive whisking that starts at P11 to P13 (13). That     threshold excitatory input during the first post-
(AW) stimulation in either Gr or SG layers (Fig.         EGOs are local and are barely seen at the cortical     natal week. In contrast, at P19 to P20, SG neurons
1, Fig. 2, A, C, and E, and fig. S2). This is in         surface may explain an apparent contradiction to       were activated during gamma oscillations, and the
keeping with the emergence of suprathreshold             the generally accepted idea, based on scalp electro-   gamma rhythmicity of IPSCs was superior to that
Gr to SG and horizontal connections only from            encephalogram recordings, that gamma oscillations      of EPSCs, a reversal from earlier ages (n = 5
the second postnatal week (9, 10). From the end          emerge relatively late in development (14, 15).        cells) (Fig. 3B and fig. S4, C and D) (22, 23).
of the second postnatal week onward, power in                The existence of robust EGOs during the neo-           In keeping with the whole-cell data, which
the gamma frequency band in Gr and SG layers             natal period is surprising because feed-forward        suggested a limited contribution of IPSCs to EGOs
progressively increased (Fig. 2E and fig. S3),           perisomatic inhibition, known to be central for        in the youngest animals, blockade of cortical
and gamma oscillations started to synchronize            the generation of cortical gamma oscillations          inhibition by gabazine did not modify EGOs at
units in the neighboring columns (fig. S2). Unlike       (15–19), is not present in Gr layer until P6 to P7     P2 to P4 (n = 5) but strongly reduced them at P6
EGOs, these later-emerging “adult” gamma os-             (20, 21). Using whole-cell recordings, we found        (n = 4) (Fig. 3, E and F). Thus, EGOs primarily
cillations were not time-locked to the stimulus          that in all Gr neurons recorded between P2 and         result from a gamma-rhythmic excitatory input to

Fig. 2. Developmental switch from input-specific
early gamma oscillations to horizontal adult gamma
oscillations. (A) IOS-guided multishank recordings of
responses evoked by a single-whisker deflection in
SG and Gr layers of the principal and adjacent barrel
columns in a P5 (top traces) and P32 (bottom traces)
rat. Color-contour plots show stimulus-triggered av-
erages for Gr (P5) and SG (P19) LFP wavelet spectro-
grams. (B) Expanded gamma-filtered (20- to 100-Hz
bandpass) LFP and MUA in SG and Gr layers for the
principal column (PC, red) and adjacent column (AC,
blue) to a single-whisker deflection recorded simulta-
neously at P32. (C) PW deflection evoked averaged
sensory potentials overlaid on a color CSD map (left,
red vertical lines indicate the time of the stimulus
onset) and mean power spectrum of the PW (middle)
and AW (right) evoked responses throughout the
cortical depth in a P5 (top) and P33 (bottom) rat. (D
and E) Postnatal changes in PW-evoked (red), AW-
evoked (blue), and baseline (gray) gamma power
(peak between 40 and 60 Hz in 200-ms window) in
Gr (D) and SG (E) layers. Pooled data from 67 rats.

                                          SCIENCE        VOL 334        14 OCTOBER 2011                                                  227
      Gr cells, which constitutes the only drive for EGOs     evoked IPSCs was delayed by >100 ms from the            Gr units after the first peak in stimulus-triggered
      during the first postnatal days. Local inhibitory       onset of EPSCs at P2 to P3. This temporal               MUA histogram (dip) was unchanged at P2 to
      circuits are progressively recruited into EGOs          “integration window” rapidly shortened during           P4 (n = 5) but was strongly reduced at P6 (n =
      only at the end of the first postnatal week.            the first postnatal week (Fig. 3, C and D) to attain    4) after blockade of cortical inhibition (Fig. 3,
          In agreement with previous results from             5.9 T1.5 ms (n = 6) at P7 to P11 [compared with         E and G). Thus, the two facets of perisomatic
      thalamocortical slices (20, 21), the onset of PW-       ~1 ms in adults (24)]. Accordingly, suppression of      inhibition—feed-forward inhibition and gamma
                                                                                                                      synchronization—show remarkably similar develop-
                                                                                                                      mental profiles during the first postnatal week.
                                                                                                                           Simultaneous recordings from topographi-
                                                                                                                      cally aligned loci in the ventral posterior medial
                                                                                                                      (VPM) nucleus of the thalamus and corresponding
                                                                                                                      cortical column (n = 9) (P5 to P7) (Fig. 4A) re-
                                                                                                                      vealed PW-evoked gamma-rhythmic VPM MUA
                                                                                                                      responses (peak frequency 48 T 1 Hz), strongly
                                                                                                                      coherent with cortical EGOs (Fig. 4, B to E).
                                                                                                                      Thalamic units showed strong phase modulation
                                                                                                                      relative to the cortical EGOs (resultant vector =
                                                                                                                      0.24) such that they fired 7 T1 ms ahead of Gr
                                                                                                                      cells (Fig. 4, D and E). This thalamocortical
                                                                                                                      binding was maintained for eight EGO cycles,
                                                                                                                      indicating a multiple replay of a sensory input in
                                                                                                                      topographic thalamocortical microcircuits (Fig.
                                                                                                                      4F). Interestingly, blockade of intrathalamic in-
                                                                                                                      hibition suppressed cortical EGOs at P2 to P4
                                                                                                                      (n = 5) (fig. S9), suggesting that the generation of
                                                                                                                      thalamic gamma activity involves synchroniza-
                                                                                                                      tion driven by the reticular nucleus (25).
                                                                                                                           Our findings suggest the following network
                                                                                                                      model (Fig. 4G). Sensory input from a whisker
                                                                                                                      activates an inhibition-based gamma oscillator in
                                                                                                                      the thalamic barreloid, which imposes topographic
                                                                                                                      feed-forward synchronization in the corresponding
                                                                                                                      cortical barrel (Fig. 4G, 1) (26, 27). Cortical in-
                                                                                                                      terneurons then become involved in EGOs in an
                                                                                                                      age-dependent manner (Fig. 4G, 2). Until ~P5,
                                                                                                                      EGOs are independent of cortical inhibition. Start-
                                                                                                                      ing from P5, along with the development of feed-
                                                                                                                      forward inhibition, interneurons are recruited and
                                                                                                                      support EGOs by controlling runaway recurrent
                                                                                                                      cortical excitation. Thus, during the first postnatal
                                                                                                                      week, EGOs undergo evolution from a primitive
                                                                                                                      form of cortical activity passively following a
                                                                                                                      thalamic oscillator, to a more complex interac-
                                                                                                                      tive model in which an active cortical oscillator,
                                                                                                                      by virtue of emerging inhibition, starts to support
                                                                                                                      gamma oscillations.
                                                                                                                           The first postnatal week is also characterized
                                                                                                                      by an enhanced plasticity of thalamocortical con-
                                                                                                                      nections (28). From a synaptic plasticity stand-
                                                                                                                      point, EGOs provide repetitive synchronization
                                                                                                                      of thalamic and cortical neurons, thus creating
                                                                                                                      conditions for potentiation of the topographic
                                                                                                                      thalamocortical connections (29). Indeed, 30 arti-
                                                                                                                      ficial EGOs (aEGOs), mimicked by pairing sub-
                                                                                                                      threshold gamma-rhythmic (50 Hz) thalamic input
                                                                                                                      with action potentials in Gr neurons in thalamo-
      Fig. 3. Synaptic mechanisms of early gamma oscillations. (A) Whole-cell responses evoked by PW stim-
      ulation in Gr neurons at different ages recorded in voltage clamp to separate IPSCs (blue top traces at 0 mV)   cortical slices, resulted in long-lasting potentiation
      and EPSCs (red bottom traces at –75 mV). (B) Age dependence of the IPSC/EPSC gamma power ratio in               of thalamocortical EPSPs by 23 T 1% (n = 9) (P4
      Gr cells (P2 to P11) (n = 33 cells) and SG cells (P19 to P20) (n = 5 cells). (C) Age dependence of EPSCs and    to P6) (Fig. 4H). In contrast, 30 artificial spindle-
      IPSCs onset delay in Gr cells from PW stimulus onset. (D) Age dependence of the integration window,             bursts (10 Hz) or rhythmic pairing at 0.5 Hz induced
      defined as the difference between EPSCs and IPSCs onset delay. (E) (Top) PW-evoked Gr LFP responses in          a depression or no change in EPSPs, respective-
      P4 and P6 rats before (black traces) and after (red traces) epipial application of the g-aminobutyric acid      ly (Fig. 4H).
      type A receptor blocker gabazine (50 mM). (Bottom) Corresponding Gr MUA PSTHs. (F and G) Summary                     In summary, we show that EGOs are a char-
      plots of the effects of gabazine on (F) PW-evoked Gr LFP gamma power at P2 to P4 (n = 5) and P6 (n = 4),        acteristic activity pattern transiently expressed in
      and (G) MUA dip-to-peak ratio [as shown in (E)].                                                                the developing rat barrel cortex during the critical

228                                           14 OCTOBER 2011             VOL 334       SCIENCE
Fig. 4. Thalamocortical binding during early gam-
ma oscillations. (A) (Top) Experimental setup for
simultaneous recordings of single-whisker evoked
responses in the corresponding VPM barreloid and
cortical barrel column. (Bottom) A coronal section
showing the location of the DiI-labeled recording
electrode (arrow). (B) (Top) Three sequential re-
sponses to C5 whisker deflections in the C5 bar-
reloid of VPM thalamus. (Bottom) Average Gr layer
LFP (red) and VPM MUA PSTH (black histogram)
from 100 deflections on the same time scale. (C)
Power spectral density for VPM MUA evoked by PW
and AW stimulations and for baseline activity. (D)
Autocorrelogram for PW-evoked VPM units and cross-
correlogram for Gr versus VPM units. (E) Cortical
gamma-phase modulation of Gr and VPM units and
the Rayleigh’s resultant vectors (R). (F) Latency in
VPM versus Gr layer MUA for each consecutive cycle
of the PW-evoked EGO. [(C) to (F) show pooled data
from nine P5 to P7 rats.] (G) Proposed network EGOs
model. (H) (Top left) Scheme of whole-cell recordings
of thalamocortical EPSPs in thalamocortical slices.
(Top right) Subthreshold EPSPs in Gr neurons (con-
trol, black trace) were potentiated after 30 episodes
of artificial EGO-like pairing of thalamic inputs with
spikes in postsynaptic neurons (after pairing, red
trace). (Bottom) Time course (left) and averages (TSE)
(right) of normalized EPSP slopes after 150 pair-
ings, organized in 30 aEGOs (50 Hz), 30 spindle
bursts (10 Hz), and rhythmic pairing at 0.5 Hz (n = 9
P4 to P6 cells for each condition).

period for activity-dependent plasticity in thala-         6. M. Minlebaev, Y. Ben-Ari, R. Khazipov, J. Neurophysiol.           22. A. Hasenstaub et al., Neuron 47, 423 (2005).
mocortical synapses. In contrast to the inhibition-           97, 692 (2007).                                                   23. B. V. Atallah, M. Scanziani, Neuron 62, 566 (2009).
                                                           7. J. W. Yang, I. L. Hanganu-Opatz, J. J. Sun, H. J. Luhmann,        24. L. Gabernet, S. P. Jadhav, D. E. Feldman, M. Carandini,
based “adult” gamma oscillations that emerge at               J. Neurosci. 29, 9011 (2009).                                         M. Scanziani, Neuron 48, 315 (2005).
the end of the second postnatal week and enable            8. M. T. Colonnese et al., Neuron 67, 480 (2010).                    25. D. Pinault, M. Deschênes, Neuroscience 51, 245
horizontal synchronization, EGOs are primarily             9. I. Bureau, G. M. Shepherd, K. Svoboda, Neuron 42, 789                 (1992).
driven by gamma-rhythmic excitatory thalamic                  (2004).                                                           26. I. Timofeev, M. Steriade, J. Physiol. 504, 153 (1997).
                                                          10. A. J. Borgdorff, J. F. Poulet, C. C. Petersen, J. Neurophysiol.   27. M. Castelo-Branco, S. Neuenschwander, W. Singer,
input and provide vertical synchronization be-                97, 2992 (2007).                                                      J. Neurosci. 18, 6395 (1998).
tween topographically aligned thalamic and cor-           11. P. Golshani et al., J. Neurosci. 29, 10890 (2009).                28. D. E. Feldman, R. A. Nicoll, R. C. Malenka, J. Neurobiol.
tical neurons. Multiple replay of sensory input           12. N. L. Rochefort et al., Proc. Natl. Acad. Sci. U.S.A. 106,            41, 92 (1999).
in the thalamocortical synapses during EGOs                   15049 (2009).                                                     29. Y. Dan, M. M. Poo, Physiol. Rev. 86, 1033 (2006).
                                                          13. M. Landers, H. Philip Zeigler, Somatosens. Mot. Res. 23,          Acknowledgments: We thank S. Beguin and D. Zaynutdinova
(“repetitio est mater studiorum”) may allow                                                                                         for slice staining and morphology and G. Buzsaki,
                                                              1 (2006).
thalamic and cortical neurons to be woven into            14. P. J. Uhlhaas, F. Roux, E. Rodriguez, A. Rotarska-Jagiela,            A. Destexhe, R. Cossart, Y. Ben-Ari, C. Hammond,
vertical topographic functional units before the              W. Singer, Trends Cogn. Sci. 14, 72 (2010).                           D. Debanne, R. Miles, S. Feldt, and N. Burnashev for
development of horizontal binding and other in-           15. M. A. Whittington, M. O. Cunningham, F. E. LeBeau,                    comments. Financial support was provided by the Agence
                                                              C. Racca, R. D. Traub, Dev. Neurobiol. 71, 92 (2011).                 Nationale de la Recherche (ANR_09MNPS006) and the
tegrative cortical functions (14–19) subserved by
                                                          16. P. Fries, Annu. Rev. Neurosci. 32, 209 (2009).                        Fondation pour la Recherche Médicale (FRM_DEq.
“adult” gamma oscillations in mature brain.                                                                                         20110421301).
                                                          17. G. Buzsaki, Rhythms of the Brain (Oxford Univ. Press,
                                                              Oxford, 2006).
    References and Notes                                  18. M. Bartos, I. Vida, P. Jonas, Nat. Rev. Neurosci. 8, 45           Supporting Online Material
 1. V. B. Mountcastle, J. Neurophysiol. 20, 408 (1957).
 2. H. Van der Loos, T. A. Woolsey, Science 179, 395      19. X. J. Wang, Physiol. Rev. 90, 1195 (2010).                        Materials and Methods
                                                          20. A. Agmon, D. K. O’Dowd, J. Neurophysiol. 68, 345                  Figs. S1 to S10
 3. K. Fox, J. Neurosci. 12, 1826 (1992).                     (1992).                                                           References
 4. D. J. Simons, P. W. Land, Nature 326, 694 (1987).     21. M. I. Daw, M. C. Ashby, J. T. Isaac, Nat. Neurosci. 10, 453       1 July 2011; accepted 1 September 2011
 5. R. Khazipov et al., Nature 432, 758 (2004).               (2007).                                                           10.1126/science.1210574

                                               SCIENCE            VOL 334            14 OCTOBER 2011                                                                229
                                                                                                                                             cover. Seasonality, although globally constrained
      The Global Extent and Determinants of                                                                                                  by total rainfall, varies substantially, as exempli-
                                                                                                                                             fied in the extreme by the monsoon in Australia.
      Savanna and Forest as Alternative                                                                                                      If seasonality does affect tree cover, it may pro-
                                                                                                                                             foundly affect savanna and forest distributions.

      Biome States                                                                                                                           Mechanisms are largely unknown, but the effects
                                                                                                                                             of seasonality have been attributed to effects on
                                                                                                                                             tree physiology and/or fire spread (2). Direct phys-
      A. Carla Staver,1* Sally Archibald,2 Simon A. Levin1                                                                                   iological limitations to tree growth (2, 14) might
                                                                                                                                             prevent forest establishment in seasonal environ-
      Theoretically, fire–tree cover feedbacks can maintain savanna and forest as alternative stable states.                                 ments, whereas indirect positive effects of long
      However, the global extent of fire-driven discontinuities in tree cover is unknown, especially accounting                              dry seasons on the likelihood of fire spread (15)
      for seasonality and soils. We use tree cover, climate, fire, and soils data sets to show that tree cover is                            could limit either savannas to seasonal environ-
      globally discontinuous. Climate influences tree cover globally but, at intermediate rainfall (1000 to                                  ments (if seasonality is necessary for fire spread)
      2500 millimeters) with mild seasonality (less than 7 months), tree cover is bimodal, and only fire                                     or forests to aseasonal ones (if seasonality makes
      differentiates between savanna and forest. These may be alternative states over large areas, including                                 fire so likely that forest cannot persist).
      parts of Amazonia and the Congo. Changes in biome distributions, whether at the cost of savanna (due to                                    A comprehensive understanding of tree-cover
      fragmentation) or forest (due to climate), will be neither smooth nor easily reversible.                                               distributions and of the potential for fire feed-
                                                                                                                                             backs to maintain savanna and forest as distinct
              ire is a strong predictor of the global dis-           tree cover or on the potential distribution of fire                     states requires more extensive, global evalua-

      F       tribution of the savanna biome (1, 2) and of
              tree cover within savannas (3–5). Experi-
      mental work shows that fire can impact tree cov-
                                                                     effects. Locally, big differences in soil texture can
                                                                     have substantial effects on tree cover (11, 12),
                                                                     whereas at the continental scale, soil texture and
                                                                                                                                             tion. Incorporating not only tree cover, mean an-
                                                                                                                                             nual rainfall, and fire frequency, but also rainfall
                                                                                                                                             seasonality and soils into this analysis would
      er and can maintain savanna where climate can                  fertility have limited effects on tree cover (2, 3, 5).                 provide additional insights into whether fire is
      support forest (6–8). Meanwhile, fire spread de-               Marked rainfall seasonality is also associated                          a primary driver of biome distributions world-
      pends on a continuous grass layer, to which tree               with savannas and tends to decrease tree cover in                       wide. We analyzed spatial patterns of tree cover
      cover acts as a barrier; tree cover has little effect          the tropics/subtropics [(2, 13), but see (4)], al-                      [from the Moderate Resolution Imaging Spec-
      on fire spread, frequency, or size until it reaches            though continental analyses have not yet identi-                        troradiometer Satellite (MODIS)] with respect
      a threshold (45 to 50%) at which fire can no lon-              fied seasonality as driving bimodalities in tree                        to rainfall [from the Tropical Rainfall Measuring
      ger spread (1, 9, 10). Thus, fire can theoretically
      act as a positive feedback within savannas that                                                     Africa                        Australia                      South America
      maintains open canopies, which, in turn, promote                                            A                              B                             C

      fire spread. These effects depend on climatic

      context. In Africa, low rainfall deterministically

      results in savanna and high rainfall in forest (1, 2).
      At intermediate rainfall, forests and savannas
      both persist and tree cover is bimodal, indicating

      that savanna is a distinct and possibly alternative
                                                                                                  0     25    50    75    100     0    25    50    75   100        0    25    50     75   100
      stable state to forest (1).
           Fire feedbacks provide a plausible mechanism                                                                               % tree cover

      to explain observed bimodalities in tree cover, but
                                                                      % tree cover

                                                                                                  D                                                                F
      questions remain as to how globally widespread
      they are and about potential alternative drivers.

      The prevailing wisdom is that, whereas Africa is
      characterized by variable, bimodal tree cover at

      intermediate rainfall, tree cover in Australia is
      more tightly constrained by rainfall (8, 11). Aus-                                          0          2000        4000     0         2000        4000       0         2000         4000
      tralian savannas may have a unique ecology, driven                                                                        mean annual rainfall (mm)
      by, for example, the distinct physiology of euca-
                                                                     Fig. 1. Frequency distribution of tree cover (A to C) and relation of tree cover to mean annual rainfall
      lypts (8). Alternatively, determinants of savanna              (D to F). Gray zones denote intermediate rainfall [1000- to 2500-mm mean annual rainfall (MAR)].
      distributions may be poorly understood. Little
      is known about tree-cover distributions in South
                                                                                                        Africa                        Australia                    South America
                                                                      dry season (months)

      America, although constraints appear to be less

      deterministic than in Australia (2).
           The universality of fire feedbacks as primary

      drivers of the distribution of savanna in areas of
      intermediate rainfall is also uncertain. Two major

      additional factors—soils and rainfall seasonality—
      may also have strong impacts, either directly on

       Department of Ecology and Evolutionary Biology, Princeton
                                                                                                  0          2000        4000     0         2000        4000       0         2000         4000
      University, Princeton, NJ 08544, USA. 2Natural Resources and                                                              mean annual rainfall (mm)
      Environment, Council for Scientific and Industrial Research,
      Pretoria 0001, South Africa.                                   Fig. 2. Dry season length versus mean annual rainfall for areas with forest (>55% tree cover, yellow
      *To whom correspondence should be addressed. E-mail:           crosses) and savanna (≤55% tree cover, black circles). Gray zones denote intermediate rainfall (1000- to                                          2500-mm MAR) with mild seasonality (<7 months).

230                                                 14 OCTOBER 2011                                   VOL 334      SCIENCE
Mission (TRMM)], rainfall seasonality (from              Intermediate tree cover between 50% and 55 to         that long dry seasons prevent the establishment
TRMM), soil texture (from the Food and Agri-             60% was scarce (Fig. 1, A and C). Mean annual         of forests and act as a rigid limitation on global
culture Organization’s Harmonized World Soils            rainfall was strongly related to tree cover on all    tree cover, although this large-scale correlative ap-
Database), and fire frequency (from MODIS),              continents (table S1). In Africa and South America,   proach can offer no insights into mechanism. It
using satellite-derived data sets with complete          tree cover increased predictably with increasing      also severely limits the extent of forest in Australia;
spatial coverage of tropical and subtropical Af-         rainfall until 1000 mm, when a closed canopy          aseasonal to mildly seasonal environments were
rica, Australia and Southeast Asian islands, and         became possible (1). At this point, bimodalities in   common in the 1000- to 2500-mm rainfall range
South America (16). These data permit evaluations        tree cover appeared (Fig. 1, D and F, and fig. S1).   on other continents, but these climatic conditions
of not only the global prevalence of bimodalities        The maximum rainfall at which savanna persisted       did not occur in Australia. Geographical patterns of
in tree cover but also the extent to which fire,         was difficult to define in Africa because areas       rainfall and seasonality are largely driven by the
versus climatic and edaphic factors, is globally re-     with high rainfall were uncommon, but South           prevalence of the monsoon in northern Australia.
sponsible for differentiating savanna from forest.       American data suggest that savannas persisted         Monsoonal rainfall is characterized by heavy pre-
    Herbivory, both grazing and browsing, may            up to rainfall of 2500 mm (fig. S1). High rainfall    cipitation during only a few months of the year,
also play a role in shaping tree cover at the con-       (>2500 mm) deterministically resulted in closed       resulting in intermediate rainfall systems with
tinental scale. We have largely excluded systems         canopies. Meanwhile, in Australia, tree-cover pat-    severe rainfall seasonality. Thus, differences in tree-
where the impacts of modern humans and do-               terns were not bimodal, not because intermediate      cover patterns and their relation to rainfall in Aus-
mesticated grazers are intensive by excluding crop-      tree cover was abundant but because high tree         tralia were largely explained by differences in
land and converted pasture (16). However,                cover was scarce, and tree cover increased more       geography, rather than by intrinsic continental dif-
experimental and observational work has dem-             predictably with rainfall (Fig. 1, B and E). At       ferences in the Australian ecology (table S1). The
onstrated that wild herbivores can be direct pri-        intermediate rainfall, where both savanna and         Australian example highlights the pivotal role that
mary drivers of local tree cover (17, 18) and can        forest occurred on other continents, Australia was    extreme seasonality plays in determining the glob-
even influence characteristic fire regimes (19).         dominated by savanna (8, 11). Either the Austra-      al extent of savanna and forest, but the climatic
An appropriate global herbivory data set would           lian flora has different environmental constraints,   processes that shape tree cover appear to be global.
provide a critical tool for evaluating the global        or some other factor prevented forest occurrence.         Although strong seasonality constrained for-
relevance of those impacts.                                  Adding estimates of seasonality improved pre-     est distributions, the data showed little indication
    Tree cover across sub-Saharan Africa and             dictions of tree cover (table S1). Globally, forest   that mild seasonality constrained savanna occur-
South America was distinctly bimodal, with an            (>55% tree cover) only occurred where dry sea-        rence. Savannas occurred in areas with rainfall of
abundance of low (<50%) and high tree cover              sons were shorter than ~7 months (and rainfall        up to 2500 mm, regardless of dry season length,
(>60% in Africa and >55% in South America).              >1000 mm) (Fig. 2 and fig. S1). This suggests         although savannas occurred only rarely in Afri-
                                                                                                               can areas with dry seasons shorter than 2 months
                                                                                                               (Fig. 2 and fig. S1) (2). This suggests that long
                       Africa                       Australia                    South America                 dry seasons are unnecessary for savanna persist-


                                                                                                               ence. The result is a climate envelope, with in-

                                                                                                               termediate rainfall and mild seasonality, in which

                                                                                                               both savanna and forest are common.

                                                                                                                   As in Africa (1), fire was a powerful predictor


                                                                                                               of tree cover (table S1) and strongly differentiated
                                                                                                               between savanna and forest within the climate
                                                                                                               envelope in which either could persist (Fig. 3 and
                                                                                                               table S1). On all three continents, fire was ubiq-


                   0   25   50   75 100         0   25   50    75 100            0   25    50    75 100        uitously present in savanna, and forest occurred
                                                                                                               more commonly where fires were absent. How-
                                                    % tree cover                                               ever, patterns were weaker in South America,
Fig. 3. Frequency distributions of tree cover in areas of intermediate rainfall (1000- to 2500-mm MAR)         where fire occurred in forests more frequently.
and mild seasonality (dry season <7 months), with fire present (gray bars) and with fire absent (hashed        Soil texture (like seasonality) did not distinguish
black bars).                                                                                                   between areas with high versus low tree cover at

Fig. 4. Distributions of
biome types across sub-
Saharan Africa, South
America, and Southeast
Asia/Australia. Biome
types are defined as areas
where climate (i) deter-
ministically supports low
tree cover (low rainfall,
high seasonality); (ii) sup-
ports biome bistability
(intermediate rainfall,
mild seasonality), current-
ly savanna; (iii) supports                                                                            Deterministic low tree cover
biome bistability, current-                                                                           Bistable, currently low tree cover
ly forest; and (iv) deter-                                                                            Bistable, currently forest
ministically supports forest                                                                          Deterministic forest
(high rainfall).

                                          SCIENCE        VOL 334       14 OCTOBER 2011                                                       231
      intermediate rainfall (fig. S2), although it was weak-          Ongoing human-driven global change is like-                12. P. G. Cruz Ruggiero, M. A. Batalha, V. R. Pivello,
      ly predictive of tree cover (table S1). It is possible      ly to have major impacts on those distributions                    S. T. Meirelles, Plant Ecol. 160, 1 (2002).
                                                                                                                                 13. S. P. Good, K. K. Caylor, Proc. Natl. Acad. Sci. U.S.A.
      that global soils data sets are insufficiently accu-        for a variety of reasons. Historical biome dis-                    108, 4902 (2011).
      rate or fine-scaled, but this analysis suggests that        tributions were certainly affected by humans in                14. E. Veenendaal, O. Kolle, J. Lloyd, Glob. Change Biol.
      soils do not provide an alternative mechanism               some areas (24, 25), but modern human activ-                       10, 318 (2004).
      for explaining global bimodalities in tree cover.           ities are much more extensive and intensive.                   15. D. Cahoon Jr., B. Stocks, J. Levine, W. Cofer III, K. O’Neill,
                                                                                                                                     Nature 359, 812 (1992).
          Patterns of fire, tree cover, and climate are           Fire exclusion experiments (6–8) and documen-                  16. Materials and methods, as well as other supporting
      consistent with the idea that fire can function in          tation of forest encroachment from around the                      material, is available on Science Online.
      savanna systems as a positive feedback, wherein             world (26, 27) provide evidence that changing                  17. H. Prins, H. van der Jeugd, J. Ecol. 81, 305 (1993).
      fire suppresses tree cover (6–8) and low tree cov-          burning practices and patterns through landscape               18. A. C. Staver, W. J. Bond, W. D. Stock, S. J. Van Rensburg,
                                                                                                                                     M. S. Waldram, Ecol. Appl. 19, 1909 (2009).
      er promotes fire spread (1, 9, 10). This feedback           fragmentation and management policy have re-                   19. S. Archibald, W. Bond, W. Stock, D. Fairbanks, Ecol. Appl.
      affects observed tree cover patterns by expanding           sulted in widespread encroachment of forest into                   15, 96 (2005).
      the range of savanna beyond areas with climates             savanna. The weakening of the monsoon could                    20. M. A. Cochrane et al., Science 284, 1832 (1999).
      that directly limit tree cover and by maintaining           compound these effects (28). Meanwhile, recent                 21. J. Balch et al., Glob. Change Biol. 14, 2276 (2008).
                                                                                                                                 22. J. Keeley, P. Rundel, Ecol. Lett. 8, 683 (2005).
      the bimodalities that define savanna and forest as          work has suggested that global climate change
                                                                                                                                 23. T. Desjardins, A. C. Filho, A. Mariotti, C. Girardin,
      distinct states. However, the not-infrequent inci-          could put Amazonia at risk of severe drying (29)                   A. Chauvel, Oecologia 108, 749 (1996).
      dence of fires in high–tree-cover areas of South            and that increased fire risk could compound these              24. K. J. Willis, L. Gillson, T. M. Brncic, Science 304, 402
      America suggests that some dynamic change be-               effects (30, 31), potentially resulting in widespread              (2004).
      tween states may be possible. Under hot, dry con-           transitions from forest to savanna. Depending on               25. F. E. Mayle, R. P. Langstroth, R. A. Fisher, P. Meir,
                                                                                                                                     Philos. Trans. R. Soc. London Ser. B 362, 291 (2007).
      ditions, forest litter can carry fires; forests have only   climatic context, massive areas of South Amer-                 26. D. Goetze, B. Horsch, S. Porembski, J. Biogeogr. 33,
      a limited capacity to recuperate from such events           ica and possibly of Africa that are currently char-                653 (2006).
      (20, 21). These South American forests may be more          acterized by savanna and forest are potentially at             27. E. Mitchard, S. Saatchi, F. Gerard, S. Lewis, P. Meir,
      subject to an ongoing process of encroachment               risk for changes in biome state (Fig. 4). Those                    Earth Interact. 13, 1 (2009).
                                                                                                                                 28. K. K. Kumar, B. Rajagopalan, M. A. Cane, Science 284,
      of fire and savanna than are African forests.               changes, whether at the cost of savanna or for-                    2156 (1999).
          The global impacts of these fire feedbacks              est, are unlikely to be smooth or easily reversible.           29. O. L. Phillips et al., Science 323, 1344 (2009).
      on savanna and forest distributions are extensive                                                                          30. Y. Malhi et al., Proc. Natl. Acad. Sci. U.S.A. 106, 20610
      (Fig. 4). If savanna and forest are alternative sta-            References and Notes                                           (2009).
                                                                   1. A. C. Staver, S. Archibald, S. Levin, Ecology 92, 1063     31. G. P. Asner, A. Alencar, New Phytol. 187, 569 (2010).
      ble states over large parts of their range, as we                                                                          Acknowledgments: We thank A. Wolf, S. Batterman, and
      argue, explicit considerations of transitions be-            2. C. E. Lehmann, S. A. Archibald, W. A. Hoffmann,                S. Rabin for manuscript feedback. Funding was
      tween the two are fundamental to understanding                  W. J. Bond, New Phytol. 191, 197 (2011).                       provided by the Andrew W. Mellon Foundation. All data
      historical and future changes in biome distribu-             3. M. Sankaran et al., Nature 438, 846 (2005).                    are freely available at,
                                                                   4. G. Bucini, N. Hanan, Glob. Ecol. Biogeogr. 16, 593   , and
      tions globally. C4 grass evolution and the global               (2007).                                                        soils/harmonized-world-soil-database/en/.
      expansion of savanna occurred during the Mio-                5. M. Sankaran, J. Ratnam, N. Hanan, Global Ecol. Biogeogr.
      cene, a period marked not only by aridity but also              17, 236 (2008).
      by increased rainfall seasonality (22). Paleo-               6. M. Swaine, W. Hawthorne, T. Orgle, Biotropica 24, 166      Supporting Online Material
      ecological work suggests that contemporary                   7. A. Moreira, J. Biogeogr. 27, 1021 (2000).                  Materials and Methods
      savannas occurring in areas now wet enough to                8. W. Bond, Annu. Rev. Ecol. Syst. 39, 641 (2008).            Figs. S1 and S2
      support forest were probably established dur-                9. S. Archibald, D. Roy, B. van Wilgen, R. Scholes,           Table S1
                                                                      Glob. Change Biol. 15, 613 (2009).                         References (32–37)
      ing drier periods in Earth’s history (23), in line          10. S. Pueyo et al., Ecol. Lett. 13, 793 (2010).
      with the idea that fire feedbacks promote the               11. R. Williams, G. Duff, D. Bowman, G. Cook, J. Biogeogr.     28 June 2011; accepted 6 September 2011
      persistence of historical biome distributions.                  23, 747 (1996).                                            10.1126/science.1210465

      Global Resilience of Tropical Forest                                                                                       cover will respond smoothly to climatic change
                                                                                                                                 and other stressors (5) or exhibit sharp transitions
                                                                                                                                 between contrasting stable states at tipping points
      and Savanna to Critical Transitions                                                                                        (6). In some regions, forest, savanna, and treeless
                                                                                                                                 (barren or grassy) states have been suggested to
                                                                                                                                 represent alternative attractors (7–9). However,
      Marina Hirota,1 Milena Holmgren,2* Egbert H. Van Nes,1 Marten Scheffer1                                                    the case for multiple stable states of tree cover
                                                                                                                                 is largely based on models and on local obser-
      It has been suggested that tropical forest and savanna could represent alternative stable states,                          vations of sharp transitions (6–9). Systematic
      implying critical transitions at tipping points in response to altered climate or other drivers.                           studies of tree-cover distributions could help dis-
      So far, evidence for this idea has remained elusive, and integrated climate models assume smooth                           tinguish between hypotheses (1) but have been
      vegetation responses. We analyzed data on the distribution of tree cover in Africa, Australia,                             largely restricted to particular continents or biome
      and South America to reveal strong evidence for the existence of three distinct attractors:                                types (4–6, 10, 11). To explore whether global
      forest, savanna, and a treeless state. Empirical reconstruction of the basins of attraction indicates                      patterns of tree abundance suggest gradual re-
      that the resilience of the states varies in a universal way with precipitation. These results allow                        sponses or, rather, alternative stable states, we
      the identification of regions where forest or savanna may most easily tip into an alternative
      state, and they pave the way to a new generation of coupled climate models.                                                 Department of Aquatic Ecology and Water Quality Man-
                                                                                                                                 agement, Wageningen University, Post Office Box 47, NL-6700
                                                                                                                                 AA, Wageningen, Netherlands. 2Resource Ecology Group,
             ree cover is one of the defining variables           and disturbances on tree growth and survival

      T      of landscapes, their ecological function-
             ing, and their impact on climate. Despite
      insights into the effects of resource availability
                                                                  (1–4), our understanding of the mechanisms de-
                                                                  termining global patterns of tree cover remains
                                                                  fragmented. A major question is whether tree
                                                                                                                                 Wageningen University, Post Office Box 47, NL-6700 AA,
                                                                                                                                 Wageningen, Netherlands.
                                                                                                                                 *To whom correspondence should be addressed. E-mail:

232                                               14 OCTOBER 2011              VOL 334          SCIENCE

Fig. 1. Tree cover distribution in tropical and subtropical regions of Africa,     frequency of the three vegetation types in rainfall classes. Curves represent
Australia, and South America. (A) Tri-modal relative frequency distribution of     logistic regression models (table S3). (C) Frequency distributions of tree
tree cover (T) indicating three distinct underlying states: forest, savanna, and   cover at different precipitation levels, illustrating the fact that distinct veg-
treeless, defined for further analyses as T ≥ 60%, 5% ≤ T < 60%, and T <           etation types persist despite the overall shift from treeless to forest with
5%, respectively (12). (B) The probability of being in each of the states as       increasing precipitation. In (A) and (C), the percent of tree cover has been
a function of the mean annual precipitation. Bars represent the relative           arcsine-transformed.

Fig. 2. Tree cover (untransformed) in 1-km2 grid
cells as a function of the mean annual precipitation
for (A) Africa, (B) Australia, (C) South America, and
(D) intercontinental data sets [between 35°S and
15°N (12)]. Although the precipitation distribution
and forest abundance vary between continents, the
statistical relationships of tree cover to precipitation
are quite similar (fig. S2).

                                         SCIENCE        VOL 334       14 OCTOBER 2011                                                      233
      analyzed the MODIS (Moderate-Resolution Im-            of average tree cover with precipitation, the dis-         traction basin around the savanna state is grow-
      aging Spectroradiometer) set of remotely sensed        tinct character of the three states remains. Indeed,       ing. As a consequence, the critical loss of tree
      estimates of tree cover in 1-km2 blocks in tropical    the characteristic tree cover of savanna (around           cover beyond which forest is expected to shift
      and subtropical zones of Africa, Australia, and        20%) and of forest (around 80%) remains re-                toward a savanna state becomes smaller in the
      South America. We relate these patterns to precipi-    markably constant over a wide range of rainfall            vicinity of this bifurcation point.
      tation (12), which is a major driver of past (13)      levels. Thus, rather than a gradual increase in tree            The inferred shapes of the basins of attraction
      and recent (14) shifts in the extension of tropical    cover with precipitation, we see a shifting prob-          show how the resilience of the different states
      forests and savannas.                                  ability of being in either of the three distinct           changes with precipitation. Another obvious in-
          Taking the entire data set together, the fre-      states. Defining cutoff levels of 5 and 60% tree           dicator of resilience of a state for a given annual
      quency distribution of tree cover (Fig. 1A) is         cover for treeless and forest states, respectively         precipitation level is simply the fraction of sites
      strikingly tri-modal (table S1), implying that there   (fig. S1), we can compute the probability of be-           that are in that state. Our logistic regression mod-
      are three distinct underlying states: forest, sa-      ing in each of the states as a function of precip-         els predict that fraction as a function of rainfall
      vanna, and treeless. A closer look at tree distri-     itation (Fig. 1B). Although continents differ widely       (Fig. 1B). One straightforward application is the
      butions for different precipitation ranges (Fig. 1C)   in their rainfall range and forest abundance               mapping of estimated resilience for current biomes.
      shows that although there is an overall increase       (Fig. 2), we find little difference between the con-       In this interpretation, a currently forested site, for
                                                             tinents in the way in which the probability of             instance, has a low resilience if, given its rainfall,
                                                             finding forest or savanna varies with precipitation        the site would be much more likely to be in a
                                                             (fig. S2), suggesting that the probability of find-        savanna or treeless state. One can imagine that
                                                             ing the three alternative states varies in a fairly        such forests most easily shift to a contrasting state
                                                             universal way with rainfall for tropical and sub-          with fewer trees in response to perturbations such
                                                             tropical regions. The most conspicuous deviation           as drought or logging. As an example of such a
                                                             is a relative lack of forest in high-rainfall places in    resilience map, consider the estimated forest re-
                                                             Australia, which is most likely explained by their         silience for the South American continent (Fig. 4).
                                                             relatively long dry season (15).                           The low predicted resilience of forest in the central-
                                                                  The patterns suggest a double hysteresis of           eastern and southeastern Amazon basin coincides
                                                             tree cover in response to rainfall (Fig. 3). This can      with human pressures in the arc of deforestation.
                                                             be inferred in a formal way from the data, as-             These areas are precisely the regions with the high-
                                                             suming the distribution of states to result from the       est risk of enhanced drought in future climate
                                                             interplay between stochastic processes and the             scenarios (19). For savanna, resilience decreases
                                                             dynamics of an underlying deterministic system             toward the drier end, where the chances of turn-
                                                             (16). The diminishing frequency at which the               ing into a treeless state are larger, but also toward
                                                             three states occur toward the end of their pre-            the wetter end, where a forested state seems a
                                                             cipitation range corresponds to a decrease in the          more likely alternative. At the latter sites it seems
                                                             size of the basin of attraction of the states toward       reasonable to assume that forests could be re-
                                                             the bifurcation points (B), where the stability ends       stored relatively easily. A full set of resilience maps
                                                             through a collision with the unstable (dashed)             for the three continents can be found in (12).
                                                             equilibria. The resilience, defined as the capacity             To interpret these results in terms of vulner-
                                                             to recover from perturbations (17, 18), declines           ability and management options it is important to
                                                             toward such tipping points. For instance, approach-        understand the causes of patterns. Many studies
                                                             ing the bifurcation point BF,S at which a forest-          have related tree cover to abiotic and biotic driv-
                                                             savanna transition is inevitable, the basin of             ers (1–11) as well as social drivers (19–22), all of
                                                             attraction of the forest state shrinks while the at-       which will probably affect the resilience of the

      Fig. 3. Relationship between the resilience of trop-
      ical forest, savanna, and treeless states and mean
      annual precipitation (in millimeters per year). (A)
      The tree cover data (percent, bottom plane) sug-
      gest a double catastrophe-fold. Stable states cor-
      respond to solid parts of the curve on the bottom
      plane and to minima in the stability landscapes.
      Unstable equilibria correspond to the dashed parts
      of the curve and to hilltops in the stability land-
      scapes. At bifurcation points (B), stable equilibria
      disappear through collision with unstable equilib-
      ria. Resilience measured as the width of the basin
      of attraction around a stable state diminishes
      toward such bifurcation points. (B) Potential land-
      scapes as computed directly from the data. Stable      Fig. 4. Forest resilience for South America. (A) Resilience of remaining forest expressed as the probability of
      states (solid dots) are minima and the unstable        finding forest at the local mean annual precipitation level, computed with the global logistic regression model
      equilibria (open dots) are maxima at a given pre-      depicted in Fig. 2B. Forest with low resilience (yellow dots) is predicted to be most likely to turn into a savanna
      cipitation level. A three-dimensional animation is     or treeless state. (B) Current distribution of tree density obtained from remote sensing (12). Resilience maps
      available at (12).                                     of forest, savanna, and treeless states for South America, Africa, and Australia can be found in (12).

234                                           14 OCTOBER 2011            VOL 334         SCIENCE
different states. An important pattern is the con-        our analysis does not detect situations of low                       11. R. J. Williams, G. A. Duff, D. M. J. S. Bowman, G. D. Cook,
spicuous rarity of places with roughly 60 or 5%           resilience related to a possible collapse of a rain-                     J. Biogeogr. 23, 747 (1996).
                                                                                                                               12. See supporting material on Science Online.
tree cover. Our interpretation is that these specific     fall pattern that depends on tree cover. However,                    13. F. E. Mayle, R. P. Langstroth, R. A. Fisher, P. Meir,
situations are unstable, and this can only be ex-         coupled vegetation-climate models focusing on                            Philos. Trans. R. Soc. London Ser. B 362, 291 (2007).
plained from positive feedbacks (17). Fire might          this issue have so far neglected the possibility of                  14. D. M. J. S. Bowman, B. P. Murphy, D. S. Banfai,
play a role in the instability around 60% tree            intrinsic vegetation hysteresis. Our results pave                        Landscape Ecol. 25, 1247 (2010).
                                                                                                                               15. A. C. Liedloff, G. D. Cook, Ecol. Modell. 201, 269 (2007).
cover. It can promote the openness of savanna             the way for a new generation of such models,                         16. V. N. Livina, T. M. Lenton, Geophys. Res. Lett. 34,
when grasses produce sufficient flammable fuel            combining local and regional nonlinearities. Con-                        L03712 (2007).
(3, 21, 23), but once tree cover becomes suffi-           straining the character of the local hysteresis re-                  17. M. Scheffer, Critical Transitions in Nature and Society
ciently dense, inhibition of grass growth and the         quires further work, because our current data                            (Princeton Univ. Press, Princeton, NJ, 2009).
                                                                                                                               18. M. Scheffer et al., Nature 461, 53 (2009).
resulting fires could lead to a self-propagating          combine sites that differ in aspects such as to-                     19. Y. Malhi et al., Science 319, 169 (2008).
shift to a closed forest (3, 4, 9, 15). The instability   pography, rainfall variability, soil characteristics,                20. W. F. Laurance et al., J. Biogeogr. 29, 737 (2002).
at very low tree cover might be related to a pos-         and human pressures, thus enlarging the range of                     21. D. C. Nepstad, C. M. Stickler, B. Soares-Filho, F. Merry,
itive feedback, promoting further tree coloniza-          rainfall levels over which states are found.                             Philos. Trans. R. Soc. London Ser. B 363, 1737 (2008).
                                                                                                                               22. C. A. Nobre, L. D. Borma, Curr. Opin. Sust. 1, 28 (2009).
tion due to facilitative amelioration of the stressors        Although more-detailed studies are essential
                                                                                                                               23. S. I. Higgins et al., Ecology 88, 1119 (2007).
that would prevent tree seedling establishment in         to understand dynamics in different regions, our                     24. M. Holmgren, M. Scheffer, M. A. Huston, Ecology 78,
an entirely grassy or barren state (1, 24).               analysis shows that global patterns may be used                          1966 (1997).
    The multistability suggested by our results has       to infer resilience. Such empirical approaches are                   25. M. Scheffer, E. H. van Nes, M. Holmgren, T. Hughes,
practical implications. For instance, deforestation to    essential because it is becoming clear that accu-                        Ecosystems (N. Y.) 11, 226 (2008).
                                                                                                                               26. M. Holmgren, M. Scheffer, Ecosystems (N. Y.) 4,
the unstable threshold of 60% tree cover might in-        rate mechanistic models to predict tipping points                        151 (2001).
duce a self-propagating shift to an open savanna          are currently beyond our reach (18), and de-                         27. E. A. B. Eltahir, R. L. Bras, Adv. Water Resour. 17,
over a range of rainfall levels. On the other hand,       termining the resilience of complex systems to                           101 (1994).
exceptionally wet years related, for example, to          critical transitions remains one of the most chal-                   28. R. Avissar, D. Werth, J. Hydrometeorol. 6, 134 (2005).
                                                                                                                               29. M. D. Oyama, C. A. Nobre, J. Clim. 17, 3203 (2004).
strong El Niño–Southern Oscillation events could          lenging problems in environmental science today                      30. V. Brovkin, M. Claussen, V. Petoukhov, A. Ganopolski,
enhance the probability of a shift to higher tree cover   (17, 18).                                                                J. Geophys. Res. Atmos. 103, 31613 (1998).
(25), implying a potential window of opportunity                                                                               Acknowledgments: This research was partly funded by
for restoration efforts to help tip the balance (26).         References and Notes                                                 the European Research Council–Early Warning grant and
                                                           1. J. I. House, S. Archer, D. D. Breshears, R. J. Scholes,              Spinoza award received by M.S. The data reported in this
Insights into the particular ways in which drivers
                                                              J. Biogeogr. 30, 1763 (2003).                                        paper are extracted as described in the supporting online
such as herbivores, fire, soil biota, and nutrients                                                                                material from the publicly available sites of MODIS
                                                           2. F. I. Woodward, M. R. Lomas, C. K. Kelly, Philos. Trans.
interact to affect tree dynamics (1–11, 23, 25, 26)           R. Soc. London Ser. B 359, 1465 (2004).                              ( and the Climatic Research
will be essential to design practical guidelines for       3. W. J. Bond, Annu. Rev. Ecol. Evol. Syst. 39, 641 (2008).             Unit (
managing these complex systems.                            4. M. J. Hill, N. P. Hanan, Eds., Ecosystem Function in
    Because tree cover can substantially enhance              Savannas: Measurements and Modeling at Landscape                 Supporting Online Material
                                                              to Global Scales (CRC Press, Boca Raton, FL, 2011).    
rainfall in some parts of the world (22, 27–30), an        5. M. Sankaran et al., Nature 438, 846 (2005).                      SOM Text
interactive vegetation-climate system may some-            6. A. C. Staver, S. Archibald, S. Levin, Ecology 92, 1063 (2011).   Figs. S1 to S3
times have alternative stable states even if veg-          7. H. T. Dublin, A. R. E. Sinclair, J. Mcglade, J. Anim. Ecol.      Tables S1 to S3
etation by itself does not (8, 22, 29, 30). Because           59, 1147 (1990).                                                 References (31–33)
                                                           8. L. D. L. Sternberg, Glob. Ecol. Biogeogr. 10, 369 (2001).        Movie S1
our analysis refers to current rainfall distributions,     9. L. Warman, A. T. Moles, Landscape Ecol. 24, 1 (2009).
it will not reveal such hysteresis resulting from         10. G. Bucini, N. P. Hanan, Glob. Ecol. Biogeogr. 16,                4 July 2011; accepted 6 September 2011
vegetation-climate coupling. A corollary is that              593 (2007).                                                      10.1126/science.1210657

The Escherichia coli Replisome Is                                                                                              lagging-strand template damage with little im-
                                                                                                                               pediment because of the discontinuous nature
                                                                                                                               of lagging-strand synthesis (7–9). Leading-strand
Inherently DNA Damage Tolerant                                                                                                 lesions may be overcome by reinitiating leading-
                                                                                                                               strand synthesis downstream of damage, as the
Joseph T. P. Yeeles and Kenneth J. Marians*                                                                                    primase, DnaG, can prime the leading-strand
                                                                                                                               template on a model fork structure after origin-
The Escherichia coli DNA replication machinery must frequently overcome template lesions under                                 independent replisome assembly (10). This mod-
normal growth conditions. Yet, the outcome of a collision between the replisome and a leading-strand                           el assumes that the replisome dissociates from
template lesion remains poorly understood. Here, we demonstrate that a single, site-specific, cyclobutane                      the template after collision with the lesion and
pyrimidine dimer leading-strand template lesion provides only a transient block to fork progression in                         therefore requires the replication restart pro-
vitro. The replisome remains stably associated with the fork after collision with the lesion. Leading-strand                   teins to reload the replisome (11). However,
synthesis is then reinitiated downstream of the damage in a reaction that is dependent on the primase,                         the fate of the replisome after a collision with
DnaG, but independent of any of the known replication-restart proteins. These observations reveal that                         leading-strand template damage has not yet been
the replisome can tolerate leading-strand template lesions without dissociating by synthesizing the                            determined.
leading strand discontinuously.                                                                                                    To investigate how the replisome may over-
                                                                                                                               come leading-strand template damage, we con-
       ncounters between the DNA replication              DNA replication continues past multiple le-                          structed a 10.4-kbp plasmid containing the E. coli

E      machinery and obstacles such as DNA
       damage and RNA polymerase transcrip-
tion complexes are thought to occur frequent-
                                                          sions, generating single-stranded DNA (ssDNA)
                                                          gaps in the newly synthesized DNA (5, 6), im-
                                                          plying that replication is reinitiated downstream
                                                                                                                               Molecular Biology Program, Memorial Sloan-Kettering Cancer
                                                                                                                               Center, New York, NY 10065, USA.
ly under normal growth conditions (1–4). After            of lesions in both the leading- and lagging-                         *To whom correspondence should be addressed. E-mail:
ultraviolet irradiation of Escherichia coli cells,        strand templates. Indeed, the replisome bypasses           

                                               SCIENCE            VOL 334           14 OCTOBER 2011                                                                  235

           A                                                         B      oriC                                             PvuI
                                                                                                                                                         ~1 kbp                                                                           CPD template
                                terB    oriC PvuI                                                                 terB
                        EcoRI                                        terB
                                                                                             Replicate          EcoRI                                                                 32
                                                                                                                                                                                  [α- P]dATP
                                                                                               4 min                                                                  (i)
                                      10.4 kbp                                                no topo                                                                                                                         PvuI
           4.2 kbp                                                                                                                                                                                              Quench
                                                                                                           positive                                                   (ii)
                                                           5.3 kbp                                         supercoils                                                                                                                     Undamaged template
                                                                                                                                          CPD                                 Gel              EcoRI
                                                                                                                                                                             filter         [α-32P]dATP                                                    +
                                                                                                               Replisome-associated ERI
           C Bulk reaction                                                                                                                        D Column-isolated reaction
                        Template                     CPD                    Undamaged                                                                                Template                                CPD                          Undamaged
                        Time (min)        1      2    3     4   6     1     2                3     4       6                                                         Time (min)             1         2        3      4       6   1        2    3      4          6
                                                                                                                kbp                                                                                                                                                     kbp
                        Stalled fork                                                                                                                                 Stalled fork
                                                                                                                23.1                                                                                                                                                    23.1

                        Full length                                                                             9.4                                                  Full length       *                                                                                9.4
                                                                                                                6.6                                                                                                                                                     6.6

                        Broken fork                                                                                                                                  Broken fork
                                                                                                                4.4                                                                                                                                                     4.4
                                                                                                                kb                                                                                                                                                      kb
                        Full length                                                                             9.4                                                  Full length                                                                                        9.4
                                                                                                                6.6                                                                                                                                                     6.6
                        Stall                                                                                                                                        Stall
                                                                                                                4.4                                                                                                                                                     4.4


                        Restart                                                                                                                                      Restart

                                                                                                                2.3                                                                                                                                                     2.3
                                                                                                                2                                                                                                                                                       2

                        Okazaki                                                                                                                                                        *
                        fragments                                                                               0.6                                                  fragments                                                                                          0.6

      Fig. 1. A single leading-strand template CPD provides only a transient block                                              and undamaged templates. (C and D) Time courses for replication reactions
      to replisome progression. (A) Diagram of the CPD-containing plasmid used as                                               conducted in bulk (C) or using column-isolated, replisome-associated ERIs
      template for replication reactions. (B) Illustration showing the staging of                                               (D). Reaction products were digested with Pvu I and analyzed by both native
      typical bulk (i) and column-isolated (ii) oriC-dependent replication reac-                                                and denaturing gel electrophoresis as indicated. (D) The asterisk (*) denotes
      tions using Eco RI to release topologically stalled forks. The predicted rep-                                             products formed from ERIs where the leading strand was labeled but not
      lication products after postreplication Pvu I digestion are shown for the CPD                                             extended after Eco RI cleavage.

      Fig. 2. Restart products are associated with the                           A           Bulk reaction                                                                                      B            Column-isolated reaction
      production of full-length duplex DNA. (A and B)
                                                                                                                                                                                                                                                    Full length
                                                                                                                                    Full length

      Two-dimensional gel electrophoretic analysis of

      the replication products from both bulk (A) and



      column-isolated (B) Eco RI–dependent reactions                                                                                                                                                         Native


      using the CPD template. Reactions were incubated                                       Native                                                                                                                                             *

      for 4 min before quenching and Pvu I cleavage.                                                                                                                                                                                                                           kb

                                                                                                                                                                                      kb                                                                                       6.6
      Samples of the same reactions were run solely                                                                                                                                   6.6                     Stall
      through native and alkaline gels to serve as mark-                                         Stall
      ers for the major reaction products. The asterisk (*)                                                                                                                                                   Restart                                                          4.4
                                                                                                 Restart                                                                              4.4
      is as in Fig. 1D.
                                                                                                                                                                                      2.3                                                                                      2.3
                                                                                                                                                                                      2                                                                                        2

                                                                                              Okazaki                                                                                                         Okazaki
                                                                                              fragments                                                                                                       fragments
                                                                                                                                                                                      0.6                                                                                      0.6

                                                                                                                      23.1      9.4                            6.6           4.4 kbp                                                  23.1          9.4        6.6        4.4 kbp

236                                                         14 OCTOBER 2011                                VOL 334            SCIENCE                              
chromosomal origin of replication, oriC, and a                      migrated to a position near the top of the native gel,                        tiated replication in bulk and isolated ERIs and
single DNA lesion in the form of a cyclobutane                      indicating the presence of a stalled replication                              their associated replisomes free from unbound
pyrimidine dimer (CPD) (Fig. 1A and fig. S1A).                      fork (Fig. 1C, native). The stalled fork resumed rep-                         proteins by gel-filtration chromatography (fig. S4)
To generate an essentially unidirectional repli-                    lication to generate full-length duplex DNA (figs.                            (15). The proteins that act distributively during rep-
cation system, we inserted two terB sites ~400                      S2 and S5A, B, and C, native), despite the absence                            lication (16)—DnaG, the b subunit of the Pol III
base pairs (bp) to the counterclockwise side of                     of any of the known origin-independent, replisome-                            HE, and SSB—were added back to the reactions
oriC. In the presence of the Tus protein, terB                      loading proteins. Alkaline gel analysis of the CPD-                           and forks were released by Eco RI cleavage (Fig.
blocks replication (12), enabling us to monitor                     template reaction (Fig. 1C, denaturing) revealed the                          1D). The products generated in column-isolated
the progression of the clockwise moving fork that                   presence of a truncated leading strand (stall) and                            reactions, on both the undamaged and CPD tem-
will encounter the CPD in the leading-strand                        Okazaki fragments. In addition to these anticipated                           plates, were very similar to those observed in
template. oriC-dependent replication reactions                      products, a zone of nascent single strands that mi-                           bulk (Fig. 1, C and D, and figs. S5 and S6). One
were initiated in the absence of a topoisomerase                    grated between 3.5 and 4 kb were generated (re-                               difference was a CPD-template reaction product
to generate stable early-replication intermedi-                     start). The upper length estimate for these products                          that migrated faster than full-length DNA on na-
ates (ERIs) by stalling the clockwise moving                        (4 to 4.2 kb) corresponds to the distance (4.2 kb)                            tive gels and was resistant to cleavage by restric-
replisome ~1 kbp from the origin because of                         from the CPD to the end of the template (Eco RI                               tion enzymes that map downstream of the CPD
the accumulation of positive supercoils (Fig. 1B)                   site), and restriction mapping confirmed that they                            (fig. S5, B and D), indicating that it may be the
(13). Topologically stalled replisomes were re-                     were generated by replication downstream of the                               product of uncoupled replication. The kinetics with
leased by Eco RI cleavage (3, 14) (fig. S1b).                       CPD (fig. S5, B and C, denaturing). Furthermore,                              which full-length duplex DNA and restart products
Replication of an undamaged template in bulk,                       the kinetics with which restart products appeared                             were formed in both bulk and column-isolated re-
with only the minimum proteins required to sus-                     closely mirrored those of the transition from the                             actions were almost indistinguishable, strongly sug-
tain oriC-dependent replication [DnaA, DnaB,                        stalled fork to full-length duplex DNA (compare                               gesting that the oriC-loaded replisome can catalyze
DnaC, DnaG, HU, SSB (single-stranded DNA                            Fig. 1C, native and denaturing). Comparable results                           their formation without the need to recruit addi-
binding protein), and the DNA polymerase III                        were also obtained when the CPD was replaced                                  tional DNA polymerase or helicase from solution.
holoenzyme (Pol III HE; as Pol III* and the b                       with an abasic site (fig. S3). These data are consist-                            Our Pol III* preparation is likely to be com-
subunit)], generated full-length duplex products                    ent with a model whereby leading-strand synthesis                             posed of complexes that predominantly contain two
of the predicted 9.6 kbp (Fig. 1C, native), consist-                is reinitiated downstream of template damage.                                 DNA polymerases, whereas a recent study has sug-
ing of unit-length leading strands and 1- to 2-kb                       To determine if the replisome that initiated rep-                         gested that the active replication machinery in
Okazaki fragments (Fig. 1C, denaturing). Dur-                       lication at oriC was responsible for generating                               E. coli possesses three DNA polymerases (17). We
ing the first 2 min of the reaction with the CPD                    the full-length duplex DNA and restart products                               therefore repeated experiments, both in bulk and
template, almost all newly synthesized products                     observed in the CPD-template reaction, we ini-                                after column isolation with a Pol III* preparation

                             A        Tus / terB                                                                                                                 DnaG                   +
                                                       50-100 bases                                                                                              EagI               +       +   kbp
                                                                  EcoRI                                                                                          Stalled fork
                                                                cleavage                       +
                                                                                                                                                    Native                                      23.1

                                        Leading    Lagging                         Leading         Lagging
                                                                                                                                                                 Full length    *               9.4
B             Undamaged template                                      C CPD template                                                                             Uncoupled






              Native                                                               Native


                                                           kb                                                                                                                                    kb
                                                                                                                                            6.6                  Stall
               Full length                                 9.4                      Stall
                                                           6.6                      Restart                                                 4.4

                                                                                                                                            2.3                                                  2.3
                                                                                                                                            2                                                    2

               Okazaki                                                             Okazaki
               fragments                                   0.6                                                                                                   Okazaki                         0.6

                                                                                                             23.1    9.4   6.6        4.4   kbp
                             23.1      9.4   6.6 kbp
                                                                                                                                                  Fig. 4. DnaG reprimes the leading-strand tem-
Fig. 3. Restart products are generated by leading-strand synthesis. (A) Inhibition of replication by                                              plate downstream of the CPD. Column-isolated CPD-
the Tus-ter complex generates a 50- to 100-base region of ssDNA on the lagging-strand template                                                    template replication reactions were initiated by Eco
(12). Postreplication Eco RI cleavage will release leading-strand products as full-length duplex DNA                                              RI cleavage in the presence or absence of DnaG for
because the restriction site is located 32 bp to the 5′ side of terB. (B and C) Replication reactions                                             6 min. Reactions were quenched and digested with
were conducted in bulk on undamaged (B) and CPD (C) templates in the presence of DNA gyrase.                                                      either Pvu I, or Pvu I and Eag I, which maps to the
After 6-min incubations, products were digested with Eco RI and Pvu I before analysis by two-                                                     region of the template downstream of the CPD (fig.
dimensional gel electrophoresis, with additional samples run solely through either native or                                                      S5B). Samples were analyzed by both native and
alkaline gels serving as markers. “Uncut” refers to the positions of products (stalled forks and                                                  denaturing gel electrophoresis. The asterisk (*) is
products of uncoupled replication) that are not released by Eco RI–Pvu I cleavage.                                                                as in Fig. 1D.

                                                                   SCIENCE              VOL 334          14 OCTOBER 2011                                                       237
      that should predominantly contain three DNA              maining associated with the DNA and reinitiat-                5. W. D. Rupp, P. Howard-Flanders, J. Mol. Biol. 31, 291
      polymerases (fig. S6 and supporting methods).            ing leading-strand synthesis downstream of                       (1968).
                                                                                                                             6. V. N. Iyer, W. D. Rupp, Biochim. Biophys. Acta 228, 117
      No appreciable differences were observed in either       the damage via DnaG-dependent leading-strand                     (1971).
      the efficiency, or kinetics, with which replication      repriming. The reaction is independent of the                 7. K. Higuchi et al., Genes Cells 8, 437 (2003).
      was reinitiated downstream of the CPD.                   replication restart proteins, demonstrating that              8. V. Pagès, R. P. Fuchs, Science 300, 1300 (2003).
          We next analyzed the products of bulk and            it is an inherent property of the replisome. Bypass           9. P. McInerney, M. O’Donnell, J. Biol. Chem. 279, 21543
      column-isolated, Eco RI–dependent CPD-template           of lesions in this manner will generate ssDNA                10. R. C. Heller, K. J. Marians, Nature 439, 557 (2006).
      reactions by two-dimensional gel electrophore-           gaps behind the replication fork (6) that are                11. R. C. Heller, K. J. Marians, Nat. Rev. Mol. Cell Biol. 7, 932
      sis to establish if the restart products detected in     predominantly repaired by the faithful process                   (2006).
      denaturing gels were associated with the tran-           of RecA-dependent recombination (18). Because                12. T. M. Hill, K. J. Marians, Proc. Natl. Acad. Sci. U.S.A. 87,
                                                                                                                                2481 (1990).
      sition from stalled fork to full-length duplex           the SOS response will not be induced, this mech-             13. H. Hiasa, K. J. Marians, J. Biol. Chem. 269, 16371
      DNA (Fig. 2, A and B). As seen for both re-              anism should be particularly advantageous                        (1994).
      actions, the stalled fork was composed of the            under normal growth conditions when low lev-                 14. K. J. Marians, H. Hiasa, D. R. Kim, C. S. McHenry,
      5.3-kb stall product and a smear of Okazaki              els of damage are present, thereby avoiding cell-                J. Biol. Chem. 273, 2452 (1998).
                                                                                                                            15. K. M. Carr, J. M. Kaguni, J. Biol. Chem. 276, 44919
      fragments. Restart fragments were exclusively            division arrest and the induction of mutagenic
      associated with full-length duplex products,             translesion polymerases. However, the leading-               16. C. A. Wu, E. L. Zechner, K. J. Marians, J. Biol. Chem. 267,
      supporting the hypothesis that restart-product           strand reinitiation that we observe is not complete              4030 (1992).
      formation is coupled to the transition from stalled      and therefore replisome dissociation may oc-                 17. R. Reyes-Lamothe, D. J. Sherratt, M. C. Leake, Science
      fork to full-length DNA. Additionally, the puta-         cur, the chances of which will be increased when                 328, 498 (2010).
                                                                                                                            18. A. Berdichevsky, L. Izhar, Z. Livneh, Mol. Cell 10, 917
      tive uncoupled product of the CPD-template reac-         there are multiple lesions in the template, such                 (2002).
      tion was composed solely of leading-strand stall         as under conditions of replication stress. Should            Acknowledgments: We thank S. Bahng for establishing
      products (Fig. 2B), as would be predicted if, in some    the replisome dissociate, origin-independent repli-              the protocol for replicating M13 DNA primed with the
      cases, template unwinding and lagging-strand             some loading and SOS-inducible systems such as                   CPD oligonucleotide and C. Gabbai, S. Gupta, and
                                                                                                                                S. Shuman for advice on the manuscript. These studies
      synthesis continued to the end of the template           nuclear excision repair and translesion polymer-                 were supported by NIH grant GM34557.
      but leading-strand reinitiation did not occur (7, 8).    ases are likely to become central to cell survival.
          To investigate whether restart products were                                                                      Supporting Online Material
      the result of leading-strand synthesis, we exploited         References and Notes                           
      the manner in which the Tus-ter complex arrests           1. H. Merrikh, C. Machón, W. H. Grainger, A. D. Grossman,   Materials and Methods
                                                                   P. Soultanas, Nature 470, 554 (2011).                    Figs. S1 to S8
      replication by using DNA gyrase to relax positive         2. M. M. Cox et al., Nature 404, 37 (2000).                 References
      supercoiling (fig. S7). Under these conditions,           3. C. P. Guy et al., Mol. Cell 36, 654 (2009).
      the clockwise moving fork is replicated to the Tus-       4. P. McGlynn, R. G. Lloyd, Nat. Rev. Mol. Cell Biol. 3,    31 May 2011; accepted 1 September 2011
      ter complex. The leading strand will extend to the           859 (2002).                                              10.1126/science.1209111
      first nucleotide of terB, whereas the lagging strand
      terminates 50 to 100 bp upstream, generating a
      region of ssDNA specifically on the lagging-strand
      template (12). Postreplication Eco RI–Pvu I cleav-
      age should therefore only release leading-strand
                                                               Sequential Establishment of
      products as full-length duplex DNA (Fig. 3A).
      Figure 3B shows that the full-length products of
                                                               Stripe Patterns in an Expanding
      an undamaged-template replication reaction were
      composed exclusively of leading-strand products.         Cell Population
      The equivalent result was obtained with the CPD
      template, in both bulk (Fig. 3C) and column-             Chenli Liu,1* Xiongfei Fu,2* Lizhong Liu,1 Xiaojing Ren,3 Carlos K.L. Chau,1 Sihong Li,2
      isolated reactions (fig. S8). However, the full-length   Lu Xiang,1 Hualing Zeng,2 Guanhua Chen,3 Lei-Han Tang,4 Peter Lenz,5 Xiaodong Cui,2
      duplex products were composed of two nascent             Wei Huang,1,2† Terence Hwa,6† Jian-Dong Huang1†
      chains—the stall and restart—unequivocally show-
      ing that restart products were generated by leading-     Periodic stripe patterns are ubiquitous in living organisms, yet the underlying developmental
      strand synthesis.                                        processes are complex and difficult to disentangle. We describe a synthetic genetic circuit
          To determine if these products were primase          that couples cell density and motility. This system enabled programmed Escherichia coli cells
      dependent, column-isolated, replisome-associated         to form periodic stripes of high and low cell densities sequentially and autonomously.
      ERIs were released by Eco RI cleavage in the             Theoretical and experimental analyses reveal that the spatial structure arises from a recurrent
      presence or absence of DnaG. In the absence of           aggregation process at the front of the continuously expanding cell population. The number
      DnaG, there was virtually no full-length duplex          of stripes formed could be tuned by modulating the basal expression of a single gene. The
      DNA generated (Fig. 4, native). The major reac-          results establish motility control as a simple route to establishing recurrent structures without
      tion product migrated at the position of the un-         requiring an extrinsic pacemaker.
      coupled band and was resistant to Eag I cleavage,
      demonstrating that it was single stranded beyond               iving organisms display an amazing array               strategies of pattern formation (6–8). Recently,
      the CPD. Consistent with these findings, the de-
      naturing gel revealed that restart-product forma-
      tion was entirely dependent on DnaG, supporting
                                                               L     of regular spatial patterns (1–4). Tradi-
                                                                     tionally, elucidation of their developmental
                                                               mechanisms has been pursued through forward
                                                                                                                            efforts have been made to emulate patterning sys-
                                                                                                                            tems with predeposited positional cues (9, 10).
                                                                                                                            These studies do not address spatial or temporal
      the finding that DnaG can reinitiate leading-            or reverse genetics (3, 5). However, essential               self-organization, which is key to morphogenesis.
      strand synthesis by repriming the leading-strand         components required for pattern formation and                Indeed, coordinated cell movement in response
      template (10).                                           control are often buried in the overwhelmingly               to self-generated cues is important in embryonic
          Our data show that the E. coli replisome can         complex physiological context. Synthetic biology             development (11–13). Here, we investigate strik-
      tolerate leading-strand template lesions by re-          provides an engineering approach to examine                  ing patterns that emerge in growing bacterial

238                                            14 OCTOBER 2011             VOL 334         SCIENCE

Fig. 1. Spatiotemporal patterns formed by engineered E. coli strains. (A)                    regulation) (18). Relative diffusion coefficient values are normalized by the
Illustration of the desired cell behavior. (B) Design of the genetic circuit; see            mean value of CL3GFP at 0.4 × 108 cells ml−1. Results are representative data
text. (C) Relative cheZ and cI mRNA level in strain CL3 as a function of cell                from three independent experiments. Error bars represent the SD of three
density in bulk culture by qRT-PCR. Data are normalized by the mean value                    replicates. Time-lapsed photographs of typical patterns obtained for the en-
of cells at 0.3 × 108 cells ml−1. (D) The relative diffusion coefficient as a                gineered strain CL3 (E) and the control strain CL4 (F); see also movies S1
function of cell density. CL3GFP, strain CL3 carrying superfolder GFP. CL4(cn),              and S2. (G) Spatiotemporal diagram of (E). All experiments were carried out
strain CL4 as a control (carrying density-sensing module but with native cheZ                at 37°C. Scale bars, 1 cm.

populations when cell motility is controlled by                   Fig. 2. Kinetic model of autonomous pe-
cell density, construed as one of the simplest ex-                riodic stripe pattern formation. (A) The model
amples of morphogenic cues (14).                                  comprises three key ingredients: cell density
    We constructed a genetic circuit to suppress                  (r), AHL concentrations (h), and nutrient lev-
the motility of Escherichia coli cells at high cell               els (n); see text and (18). The key feature of
densities (Fig. 1A). The circuit is composed of                   this model is that the AHL-dependent cell
two modules: a density-sensing module and a                       motility m(h), in light of the data in Fig. 1D,
motility-control module (Fig. 1B). The quorum-                    is modeled by a steep Hill function (red line)
sensing system in Vibrio fischeri was adopted                     with an abrupt transition between the val-
                                                                  ues Dr and Dr,0 at h ≈ Kh (vertical dashed
as the density-sensing module to signal the local
                                                                  line). (B) Time-lapse plots of the simulated
cell density. This system synthesizes and excretes
                                                                  relative cell-density profiles r(x, t) (blue),
a small-molecule acyl-homoserine lactone (AHL),                   AHL concentrations h(x, t) (red), and nutri-
which, at high extracellular concentrations (re-                  ent levels n(x, t) (green shade); see also
flecting high cell density), accumulates intra-                   movie S4. The AHL threshold of motility
cellularly and activates a constitutively expressed               regulation Kh is shown as the horizontal
regulator, LuxR (15). The motility-control mod-                   dashed line; the vertical axis is rescaled
ule was devised to modify bacterial motility by                   to cell density (108 cells ml−1) (18). The pe-
regulating the transcription of cheZ. cheZ dele-                  riodic pattern is seen as a recurrent process
tion causes cells to tumble incessantly, result-                  involving the formation of a bud (B); the
ing in a nonmotile phenotype in semisolid agar                    bud subsequently grows into a stationary mound (M), and a cleft (C) structure forms in a region between
(16). Reintroducing cheZ restores cell motility                   the bud and the expanding front (F).
(17). In our circuit, coupling of the two modules
was achieved via the lambda repressor (CI):                       (Fig. 1B). On the basis of this design, we created    8.5-cm Petri dish containing 10 ml of LB me-
The LuxR-AHL complex drove the expression                         an engineered strain, CL3 (18).                       dium and 0.25% agar, a pattern of alternating
of CI that in turn repressed cheZ transcription                       Gene expression of the strain CL3 at vari-        white (high cell density) and dark (low cell den-
                                                                  ous cell densities was quantitatively characterized   sity) stripes developed overnight (Fig. 1E and
 Department of Biochemistry, The University of Hong Kong,         with quantitative reverse transcription–polymerase    movie S1) (18). The stripes formed sequentially,
Pokfulam, Hong Kong, China. 2Department of Physics, The           chain reaction (qRT-PCR). As designed, cI ex-         at a spacing of ~ 0.5 cm once every ~2 hours.
University of Hong Kong, Pokfulam, Hong Kong, China. 3De-         pression level (red) increased more than 40-fold      These stripes were stable for days until the agar
partment of Chemistry, The University of Hong Kong, Pokfulam,
                                                                  as cell density increased, whereas cheZ expres-       completely dried up. As a control, strain CL4
Hong Kong, China. 4Department of Physics, Hong Kong Baptist
University, Kowloon Tong, Kowloon, Hong Kong, China. 5De-         sion (blue) decreased sharply to 5% of the peak       (carrying the density-sensing module but with
partment of Physics and Center for Synthetic Microbiology,        level (Fig. 1C). Cell motility was measured in        the native cheZ regulation) behaved like the wild
University of Marburg, 35032 Marburg, Germany. 6Center for        semisolid agar with a modified continuous flu-        type (16): From the position of the inoculum, two
Theoretical Biological Physics, University of California at San   orescence photobleaching (CPB) method (18).           traveling waves successively moved radially out-
Diego, La Jolla, CA 92093–0374, USA.
                                                                  It dropped abruptly at a density of ~4 × 108 cells    wards, followed by a uniform expansion of cell
*These authors contributed equally to this work.
†To whom correspondence should be addressed. E-mail:
                                                                  ml−1 (Fig. 1D, purple symbols).                       density without stripes (Fig. 1F and movie S2). ( J.D.H.); (T.H.); huangwei@              When a suspension of exponentially grow-          We further verified that each element in the de- (W.H.)                                                     ing CL3 cells was inoculated at the center of a       signed genetic circuit is necessary for the stripe

                                                   SCIENCE        VOL 334       14 OCTOBER 2011                                                  239
      pattern formation (fig. S1). We also surveyed the      cell proliferation: There, a “bud” (marked by B)       behind, and the budding within the frontal region,
      effect of nutrient content, agar thickness, humid-     shoots up in the density profile. Meanwhile, the       result from an aggregation phenomenon driv-
      ity, starting cell number, volume, and growth          front (marked by F), which is unaffected by            en by density-dependent motility discussed in
      phase, and found the pattern to be robust to most      what happens behind because of its low density,        recent theoretical studies (21, 24) and illustrated
      conditions tested (fig. S2). Notably, the stripe       continues to propagate laterally at the same           in Fig. 3A. Cells can diffuse freely in semisolid
      pattern persists even in cheB−cheR− double mu-         speed. Later (T = 450 min), the bud in the den-        agar when the cell density is low (top panel). As
      tants incapable of chemotaxis [fig. S2J (19)] but      sity profile grows into a “mound” (marked by M),       cells proliferate and the local AHL level exceeds
      still able to swim and tumble randomly, indicating     separated from the frontal region by a “cleft”         the threshold, cell motility slows down as pro-
      that the stripes are not a result of chemotactic       (marked by C). The continued cell proliferation        grammed (middle panel, with the nonmotile cells
      signaling (20). Also, no stripe pattern resulted if    in the region just behind the front drives up the      in teal). These cells cannot move away, but neigh-
      CL3 cells were spotted on agar premixed with a         AHL level and subsequently produces a new bud          boring cells may continue to move into this high-
      low level of CL3 cells in otherwise identical con-     (B′ at T = 600 min), while the lateral movement        density region and become nonmotile (bottom
      ditions (fig. S2K), indicating that the pattern is     of the mound and cleft eventually stops (M and         panel), leading to a net cell flow toward the high-
      not a result of spontaneous amplification of den-      C at T = 600 min). The new bud starts a new stripe     density region.
      sity fluctuations, as would follow from the par-       of low-motility cells, which again expands into            Wild-type E. coli cells may aggregate when
      adigm of Turing instability (21, 22).                  a mound (M′ at T =750 min) delimited by a cleft        grown in specific poor-nutrient media owing to
           We monitored the spatiotemporal cell density      (C′ at T = 750 min) from the propagating front,        the secretion of potent chemoattractants (25). To
      profiles from optical absorbance of the plates         and the process repeats itself. Well behind the        show that cell aggregation in our system is con-
      [fig. S3 and (18)]. Figure 1G displays the time        front, cell proliferation in both the mound and        trolled by the synthetic circuit, we created two
      course of light intensities along a diagonal that      cleft eventually stops as the nutrient is depleted     strains: the senders (CL6) that synthesize AHL
      goes through the inoculum, with the color codes        (height of green shade).                               but are nonmotile, and the receivers (CL8), capable
      indicating the intensity as in Fig. 1E. It shows           The development of a cleft that separates the      of receiving AHL and regulating motility but in-
      that once formed, the pattern of low and high cell     propagating front from the dense zone of cells         capable of AHL synthesis. The receiver cells
      densities was frozen in space. Furthermore, new
      stripes of high cell density emerged at regular time
                                                             Fig. 3. Cell aggrega-
      and distance intervals. The same phenomenon            tion driven by density-
      was observed in a rectangular geometry when            dependent motility. (A)
      bacterial growth was initiated along a thin line in    Illustration of aggregation
      the middle of the plate (18). Stripes developed        via density-dependent
      and propagated toward the opposite edges for           motility; see text. (B) Ex-
      CL3 (fig. S4A and movie S3) but not for the wild       perimental evidence of
      type or CL4 cells. Fourier analysis revealed high      effective aggregation. Re-
      spatial periodicity, with a wavelength of 0.5 cm       ceiver cells were uniform-
      (fig. S4B).                                            ly mixed with 0.25% agar.
           To gain a quantitative understanding of the       Five microliters of send-
      patterning process, we developed a mathematical        er cells (CL6, producing
      model based on the characterized properties of         AHL but nonmotile, GFP-
      the engineered circuit (Fig. 2) (18). The stochas-     expressing) were spotted at the center of receiver-cell–agar mixture, followed by 12-hours’ incubation at
      tic swim-and-tumble motion of bacterial cells is       37°C. See (18) for details. (Upper panels) Strain CL8 (carrying the entire genetic circuit but lacking the
      described at the population level by a diffusion-      AHL-producing gene) as receiver (movie S5). (Lower panels) Strain CL10 (CL8 lacking cI) as receiver. (Right
      like equation (Eq. S1) for the cell density r(x, t).   panels) Fluorescence photographs indicate the positions of the sender cells.
      The synthesis, diffusion, and turnover of AHL
      are described by Eq. S2, while the consumption         Fig. 4. Tunable stripe
      and diffusion of the nutrient n(x, t) are described    patterns. (A) Phase dia-
      by Eq. S3. The AHL-dependent motility enters           gram; see (18, 26). (B)
      through a phenomenological diffusion coefficient       Genetic circuit design for
      m(h) that drops abruptly where the local AHL           tuning stripe number.
      concentration h(x, t) exceeds a threshold level Kh     An aTc-inducible module
      (Fig. 2A). Numerical simulations of Eqs. S1 to         was added to vary the ex-
      S3 (23) with realistic parameter values (table S3)     pression of cheZ. (C) Rel-
      (18) generated patterns similar to the experimental    ative cheZ and cI mRNA
      observations in both one- and two- dimensional         level of CL5 in bulk cul-
      geometries (figs. S5 and S20). The model also          ture containing various
                                                             concentrations of aTc
      captures the more elaborate patterns formed by
                                                             (0.1 to 3 ng ml−1). Cells
      multiple seeding (fig. S6).
                                                             were cultured to an ab-
           The mathematical model reveals an intrigu-        sorbance at 600 nm of
      ing sequence of events at the propagating front        ~0.05 and harvested for
      (Fig. 2B and movie S4). Early on (T = 150 min),        measurement. Data were
      cell density (blue line) grows and propagates lat-     normalized by the mean
      erally just as the Fisher wave front (2) because       value of CL3 at 0.3 ×
      the AHL concentration (red line) is uniformly          108 cells ml−1. (D) Ex-
      below the threshold level Kh (gray dashed line)        perimental patterns of
      due to the low initial cell density. A while later     CL3 (also see fig. S2G)
      (T = 300 min), the AHL level exceeds the thresh-       or CL5 inoculated at 0.25% agar containing various concentrations of aTc. Agar plates were incubated
      old in a region behind the front owing to local        at 37°C for 40 hours.

240                                           14 OCTOBER 2011            VOL 334       SCIENCE
were uniformly mixed with semisolid agar. Subse-       involving chemotaxis, swarming, and differen-                     20. H. C. Berg, Annu. Rev. Biophys. Bioeng. 4, 119
quently, a drop of sender cells was spotted at the     tiation (4). Using synthetic circuits, we demon-                      (1975).
                                                                                                                         21. M. E. Cates, D. Marenduzzo, I. Pagonabarraga, J. Tailleur,
center of the hardened cell-agar mixture. The non-     strate that precise and robust spatiotemporal                         Proc. Natl. Acad. Sci. U.S.A. 107, 11715 (2010).
motile sender cells could not move away from           patterns can be generated autonomously from                       22. A. M. Turing, Philos. Trans. R. Soc. Lond. B Biol. Sci. 237,
where they were spotted, but were expected to          a very simple interaction—motility control by                         37 (1952).
synthesize and excrete AHL as they proliferated at     density. A recurrent mechanism enables struc-                     23. We believe that Eqs. S1 to S3 comprise the minimalistic,
                                                                                                                             experimentally faithful model needed to generate the
the center of the plate. After a 12-hour incuba-       tures to form periodically and sequentially. Im-                      observed patterns. A simplified model in which the
tion, a high-density stripe of receiver cells aggre-   portant features of the pattern such as the number                    explicit description of nutrient dynamics was replaced
gated around the spotted sender cells (Fig. 3B,        of stripes can be manipulated by tuning com-                          by a growth saturation term (Eq. S11) could initiate but
upper panels, and movie S5). No aggregation            ponents of the circuit, such as the basal expres-                     not maintain the patterns, because cell density eventually
                                                                                                                             took on the saturation value everywhere in space (fig.
was observed in a control with receiver cells in-      sion level of a single gene.                                          S17A). A related density-only model (21) generated
capable of reducing motility in response to high            The strategy of sequential stripe formation                      stripe patterns by balancing cell aggregation with cell
AHL concentration (CL10) (Fig. 3B, lower pan-          is likely not limited to cellular motility control.                   death [see also (18)]; it is not applicable to our
els), or in other controls (fig. S7). Thus, an ef-     In the kinetic equations S1 to S3 that generated                      experiments because cells stop growing but do not die
                                                                                                                             on the relevant time scales. The density-only model
fective aggregation phenomenon was mediated            the stripe patterns, the cell density r(x, t) is mere-
                                                                                                                             also predicted patterns to form from small initial density
by density-dependent motility.                         ly an example of a space- and time-dependent                          fluctuation, in contrast to our observation (fig. S18)
     Detailed analysis of the mathematical model       variable. Another example could be the concen-                        and model output (18).
indicates that the aggregation effect alone is not     tration of a diffusible molecule, which stimulates                24. J. Tailleur, M. E. Cates, Phys. Rev. Lett. 100, 218103
sufficient to generate stripes. Budding in the fron-   its own synthesis to mimic the effect of “cell                        (2008).
                                                                                                                         25. E. O. Budrene, H. C. Berg, Nature 349, 630 (1991).
tal region requires the AHL level to exhibit a lo-     growth.” Thus, Eqs. 1 to 3 may be taken as a                      26. Equations S1 to S3 can produce a variety of stripe
cal maximum, which can happen only when the            generalized reaction-diffusion system with the                        patterns depending on the values of two key parameters,
diffusion length of AHL molecules is sufficiently      key feature that the mobility or transport of one                     maximum motility Dr and AHL half-life ln2/b, as
short (so that the AHL profile closely follows         of the regulators between adjacent cells is regu-                     summarized in the phase diagram (Fig. 4A); see (18)
                                                                                                                             for details. To the right of the solid red line, the system
the cell density profile) and the cleft in the cell    lated by another mobile regulator. As regulated                       exhibits the periodic stripe phase with an infinite
density profile just behind the front is suffi-        mobility of signaling molecules is a common                           number of stripes. Moving to the left of the solid line,
ciently deep. The latter is determined in turn by      strategy in metazoan development (30, 31), the                        there is a regime where the system can still generate a
the parameters, such as the maximum cell mo-           lessons from this work may stimulate new in-                          limited number of stripes for some initial conditions.
                                                                                                                             Far away from the solid line, no stripes can form. The
tility (18, 26). A key output of the model is sum-     sights and inspire new directions in the studies                      separation between the latter two regimes is not
marized by the phase diagram (Fig. 4A), which          of developmental systems. For example, the se-                        clear-cut and is indicated by the dashed green line.
predicts that in addition to the periodic stripe       quential establishment of a periodic somite struc-                    Additional effects such as chemotaxis (Eq. S17) have
phase, the engineered strain may exhibit a qual-       ture during vertebrate embryonic development                          been investigated. They modify the detailed appearances
                                                                                                                             of the patterns as well as the location of the phase
itatively different behavior with no stripes, by       is commonly assumed to be controlled by an en-
                                                                                                                             boundaries (fig. S21). The general occurrence of the
passing through a transition region exhibiting a       abling clock coupled to a sweeping morphogen                          periodic stripes in this class of models does not require
limited number of stripes.                             gradient (32). Our results suggest that a spatially                   such effects.
     A direct test of the occurrence of the phase      periodic structure can be formed autonomously                     27. H. Fujikawa, Physica A 189, 15 (1992).
transition predicted by the phase diagram (Fig.        without the need for a clock.                                     28. T. Matsuyama et al., J. Bacteriol. 182, 385 (2000).
                                                                                                                         29. D. E. Woodward et al., Biophys. J. 68, 2181 (1995).
4A) is to vary the maximum cell motility (Dr in                                                                          30. L. Niswander, Nat. Rev. Genet. 4, 133 (2003).
Fig. 2A), thereby changing the number of stripes.          References and Notes                                          31. M. Affolter, K. Basler, Nat. Rev. Genet. 8, 663 (2007).
Within our experimental design in which motili-         1. L. I. Held, Models for Embryonic Periodicity (Karger,         32. O. Pourquié, Science 301, 328 (2003).
ty is set by cheZ expression, tuning of Dr can be          Basel, 1992).                                                 Acknowledgments: We are grateful to members of the
                                                        2. J. D. Murray, Mathematical Biology: I. An Introduction            University of Hong Kong Team for The International
implemented by adding a second cI gene, whose                                                                                Genetic Engineering Machine Competition (iGEM) 2008
                                                           (Springer, New York, ed. 3, 2002).
expression is titratable in an AHL-independent          3. C. M. Chuong, M. K. Richardson, Int. J. Dev. Biol. 53, 653        for their contribution to the project. We thank H. Berg,
manner (Fig. 4B). This is implemented in strain            (2009).                                                           A. Courey, A. Danchin, D. Smith, J. Tailleur, and C. Voigt for
CL5 (18). Figure 4C shows that the basal cI             4. E. Ben-Jacob, I. Cohen, H. Levine, Adv. Phys. 49, 395             valuable comments. This project was supported by a Hong
                                                           (2000).                                                           Kong University (HKU) University Development Fund, a
expression level of CL5 cells could indeed be                                                                                Small Project Grant from the HKU Committee on Research
                                                        5. E. H. Davidson, D. H. Erwin, Science 311, 796 (2006).
smoothly tuned by adjusting the dosage of an            6. A. S. Khalil, J. J. Collins, Nat. Rev. Genet. 11, 367             and Conference Grants, and a Collaborative Research Fund
inducer, anhydrotetracycline (aTc), at a fixed cell        (2010).                                                           from the Research Grants Council (RGC) (HKU1/CRF/10) to
density (red symbols). Corresponding suppres-           7. M. Elowitz, W. A. Lim, Nature 468, 889 (2010).                    J.D.H., and a HKU Faculty of Medicine Development Fund to
                                                        8. S. Mukherji, A. van Oudenaarden, Nat. Rev. Genet. 10,             W.H. TH is supported by the NSF through the Center for
sion of cheZ expression at low cell density was
                                                           859 (2009).                                                       Theoretical Biological Physics (grant PHY-0822283) and
also observed (blue symbols). For a fixed aTc           9. S. Basu, Y. Gerchman, C. H. Collins, F. H. Arnold,                additionally acknowledges an HKU Distinguished Visiting
level, the density dependence of cheZ expres-              R. Weiss, Nature 434, 1130 (2005).                                Professorship. P.L. acknowledges support through the LOEWE
sion remained (fig. S8). When strain CL5 was           10. M. Isalan, C. Lemerle, L. Serrano, PLoS Biol. 3, e64              program of the State Hessen. L.H.T. acknowledges support
spotted at the center of semisolid agar, consist-          (2005).                                                           by the RGC of the Hong Kong Special Administrative
                                                       11. P. Rørth, Trends Cell Biol. 17, 575 (2007).                       Region under grant 201606. E. coli strains and plasmids
ent with the model predictions, the number of                                                                                are available under a material transfer agreement with
                                                       12. D. J. Montell, Curr. Opin. Genet. Dev. 16, 374 (2006).
stripes decreased gradually as the aTc concen-         13. A. Aman, T. Piotrowski, Dev. Biol. 341, 20 (2010).                the University of Hong Kong.
tration increased from 0.1 to 3.0 ng ml−1 (Fig.        14. L. E. Weiss et al., Biotechnol. Bioeng. 100, 1251
4D). As a control, the pattern of strain CL3 did           (2008).                                                       Supporting Online Material
not change even when the aTc concentration was         15. C. M. Waters, B. L. Bassler, Annu. Rev. Cell Dev. Biol. 21,
                                                           319 (2005).
increased to 100 ng ml−1 (fig. S2G).                   16. A. J. Wolfe, H. C. Berg, Proc. Natl. Acad. Sci. U.S.A. 86,
                                                                                                                         SOM Text
     Natural occurrences of well-coordinated spa-          6973 (1989).                                                  Figs. S1 to S21
tial and/or temporal patterns are abundant in de-      17. M. G. Sanna, M. I. Simon, J. Bacteriol. 178, 6275             Tables S1 to S3
velopmental systems and are believed to involve            (1996).                                                       References (33–63)
                                                       18. Materials and methods are available as supporting             Movies S1 to S5
elaborate control mechanisms (3, 5). Similarly,            material on Science Online.
stripe formation in various bacterial systems          19. J. S. Parkinson, S. E. Houts, J. Bacteriol. 151, 106          31 May 2011; accepted 2 September 2011
(27–29) has been attributed to complex effects             (1982).                                                       10.1126/science.1209042

                                           SCIENCE           VOL 334          14 OCTOBER 2011                                                                     241
                                                                                                                                    Substantial changes in the 1H-13C HMQC
      NMR Detection of Structures in the                                                                                       spectra of AUG-labeled 5′-L were observed
                                                                                                                               upon incubation at physiological ionic strength
      HIV-1 5′-Leader RNA That Regulate                                                                                        [physiological ion (PI) buffer: 10 mM Tris-HCl,
                                                                                                                               140 mM KCl, 10 mM NaCl, 1 mM MgCl2,

      Genome Packaging                                                                                                         pH 7.0], which correspond to a reversible equi-
                                                                                                                               librium shift from a predominantly monomeric
                                                                                                                               species to a mixture of monomeric and dimeric
      Kun Lu,1*† Xiao Heng,1† Lianko Garyu,1 Sarah Monti,1 Eric L. Garcia,2 Siarhei Kharytonchyk,2                             species [dimer dissociation constant (Kd) = 0.9 T
      Bilguujin Dorjsuren,1 Gowry Kulandaivel,1 Simonne Jones,1 Atheeth Hiremath,1                                             0.1 mM] (Fig. 1, G and H). Signals of the AUG
      Sai Sachin Divakaruni,1 Courtney LaCotti,1 Shawn Barton,1 Daniel Tummillo,1 Azra Hosic,1                                 hairpin in 5′-L exhibited uniformly reduced inten-
      Kedy Edme,1 Sara Albrecht,1 Alice Telesnitsky,2‡ Michael F. Summers1‡                                                    sities as a function of incubation time, and new
                                                                                                                               signals were observed for the 3′-residues of AUG
      The 5′-leader of the HIV-1 genome regulates multiple functions during viral replication via                              (A345 to A356) (Fig. 1H). Similar NMR spectral
      mechanisms that have yet to be established. We developed a nuclear magnetic resonance approach                           changes were observed for an isolated AUG oligo-
      that enabled direct detection of structural elements within the intact leader (712-nucleotide dimer)                     RNA upon titration with an oligo-U5 fragment
      that are critical for genome packaging. Residues spanning the gag start codon (AUG) form a hairpin in                    (Fig. 1I). The NMR chemical shifts and narrow
      the monomeric leader and base pair with residues of the unique-5′ region (U5) in the dimer. U5:AUG                       line widths observed for A345 to A356 indicate
      formation promotes dimerization by displacing and exposing a dimer-promoting hairpin and enhances                        that these residues are unstructured and mobile in
      binding by the nucleocapsid (NC) protein, which is the cognate domain of the viral Gag polyprotein                       the dimeric form of the leader ([5′-L]2). No 1H-13C
      that directs packaging. Our findings support a packaging mechanism in which translation, dimerization,                   HMQC signals were detected for G328 to G344 of
      NC binding, and packaging are regulated by a common RNA structural switch.                                               [5′-L]2 because of severe line broadening (dis-
                                                                                                                               cussed below). These findings are consistent with
              he 5′-leader is the most conserved region               (Fig. 1) (10, 15–23), there is less agreement re-        a phylogenetically derived structural model (10, 20),

      T       of the HIV-1 genome and is responsible
              for regulating multiple activities during
      viral replication, including RNA encapsidation
                                                                      garding the structures that regulate packaging
                                                                      (14). In particular, residues overlapping the gag
                                                                      start codon [G328 to A356 (AUG)] that are crit-
                                                                                                                               in which the 5′-residues of AUG base pair with
                                                                                                                               U5 and the 3′-residues are disordered (Fig. 1C).
                                                                                                                                    To directly probe for U5:AUG base pairing,
      during virus assembly (1, 2). Like all retroviruses,            ical for genome packaging (24) and RNA dimer             we developed an NMR approach that involves
      HIV-1 packages two copies of its genome, enabling               stability (25) have been proposed to form a hair-        replacement of a short stretch of adjacent base
      strand-transfer–mediated recombination during                   pin (Fig. 1B), to base pair with residues of the         pairs by A-U base pairs [long-range probing by
      reverse transcription and promoting genetic evo-                upstream unique-5′ element (U5) (Fig. 1C) (10),          adenosine interaction detection (lr-AID)]. The sub-
      lution under environmental and chemotherapeut-                  or to adopt other conformations (14). (Single-           stituting element (ideally [UiUjAk]:[UlAmAn],
      ic pressure (3). Dimeric genomes are trafficked                 letter abbreviations for the nucleic acids are as        but other sequences can suffice) (Fig. 2A) affords
      from the cytoplasm to plasma membrane assem-                    follows: A, adenosine; G, guanosine; C, cytosine;        an upfield-shifted Am-C2-1H NMR signal (~6.5
      bly sites by a small number of viral Gag proteins,              U, uridine.) In vivo nucleotide reactivity map-          parts per million), enabling direct detection of
      where additional Gag proteins assemble and bud-                 ping has supported multiple AUG models, without          cross-strand Ak-H2 and -H1′ 1H-1H nuclear Over-
      ding occurs (4–8). Dimerization and packaging                   consensus (14, 21, 22). Nuclear magnetic resonance       hauser effects (NOEs), without the need for het-
      are mediated by interactions between the nucleo-                (NMR) is potentially well suited to probe RNA            eronuclear spectral editing. The HIV-1NL4-3 leader
      capsid (NC) domains of the viral Gag polypro-                   structure, but signal degeneracy and relaxation          naturally contains one [UUA:UAA] element in
      teins and RNA elements within the 5′-leader of                  problems have thus far limited applications to rel-      the TAR hairpin ([U12-A14]:[U45-A47]), and,
      the genome, and there is evidence that these ac-                atively small oligonucleotides (typically fewer than     to preclude signal overlap, A46 was mutated to G
      tivities and translational control are mechanisti-              50 residues) (26, 27). Here, we describe an NMR          [a naturally occurring substitution in 7% of the
      cally coupled (7, 9–13).                                        approach that enabled direct detection of structures     reported HIV-1 TAR sequences (31)] (Fig. 2, B
          Understanding the mechanisms that regulate                  formed by AUG and other packaging elements with-         to D). NOE spectroscopy (NOESY) data (28)
      5′-leader activities is limited, in part because of             in the intact, 230-kD dimeric 5′-leader as well as the   obtained for the lr-AID–modified dimeric 5′-
      incomplete knowledge of the leader structure                    identification of conformational changes that reg-       leader ([5′-Llr-AID-U5:AUG]2, C110 to G112 and
      (14). Recombinant leader-containing RNAs have                   ulate dimerization and NC binding and packaging.         G338 to G339 substituted by UUA and AA,
      been probed by nucleotide accessibility mapping,                    A 356-nucleotide HIV-1NL4-3 5′-leader RNA            respectively) (Fig. 2A) exhibited well-resolved
      mutagenesis, and biochemical approaches, and                    that includes the entire 5′-untranslated region (5′-     A338-H2 signals with frequencies similar to
      although there is general consensus that transcrip-             UTR) and the first 21 residues of gag (5′-L) was         those observed for an isolated lr-AID–modified
      tional activation, primer binding, dimerization,                prepared by means of enzymatic ligation of non-          U5:AUG oligoribonucleotide (Fig. 2, E and F).
      and splicing activities are promoted by discrete                labeled 5′-RNA (residues 1 to 327) and 13C-              Other outlier signals in the NOESY spectrum
      hairpin structures [transacting responsive (TAR),               enriched AUG fragments (residues 328 to 356)             were unaffected by the lr-AID substitutions. All
      primer-binding site (PBS), dimer initiation site (DIS),         (Fig. 1) (28). The 5′-leader exists predominantly        of the expected A338-H2 inter-adenosine NOEs
      and splice donor (SD) hairpins, respectively]                   as a monomer at low ionic strength (Fig. 1D),            were observed, including cross-strand NOEs with
                                                                      and under these conditions, the AUG residues of          A112-H1′ and -H2. The A338-H2 T1 and T2 1H
       Howard Hughes Medical Institute (HHMI) and Department of
                                                                      5′-L gave rise to 1H-13C–correlated heteronuclear        NMR relaxation rates (8.7 s and 9.4 ms, re-
      Chemistry and Biochemistry, University of Maryland Baltimore    multiple-quantum coherence (HMQC) (28) NMR               spectively) are consistent with restricted rotation-
      County (UMBC), 1000 Hilltop Circle, Baltimore, MD 21250,        spectra similar to those observed for the isolated       al motion and indicate that U5:AUG resides
      USA. 2Department of Microbiology and Immunology, University     AUG fragment, which is known to form a hairpin           within a globular region of the 5′-L structure.
      of Michigan Medical School, Ann Arbor, MI 48109–0620, USA.      (Fig. 1, E and F) (29). NMR line widths of AUG                The relationship between U5:AUG forma-
      *Present address: RNA Institute, State University of New York   in the intact 5′-leader and the isolated AUG RNA         tion and dimerization was investigated by means
      at Albany, 1400 Washington Avenue, Albany, NY 12222, USA.
      †These authors contributed equally to this work.
                                                                      were also similar, indicating that the hairpin is        of site-directed mutagenesis. Mutations in AUG
      ‡To whom correspondence should be sent. E-mail: ateles@         structurally mobile and does not form A-minor–           designed to stabilize the hairpin and disrupt base (A.T.); (M.F.S.)                like contacts (30).                                      pairing with U5 (5′-LAUG-HP) favored the monomer,

242                                                  14 OCTOBER 2011              VOL 334        SCIENCE
whereas mutations that promote U5:AUG base            attenuated dimerization activity of the AUGHP               To determine whether U5:AUG formation
pairing and destabilize the hairpin ([5′-LU5:AUG]2)   form of 5′-L is not due to refolding of the DIS (32).   influences NC binding, isothermal titration calo-
favored the dimer (fig. S1). Deletion of AUG          Sequence complementarity between U5 and the             rimetry (ITC) experiments were performed with
(5′-LDAUG) also favored the monomer (Fig. 3A),        GC-rich loop of the DIS suggested that dimeriza-        5′-L and mutant 5′-LAUG-HP and 5′-LU5:AUG RNAs.
whereas incubation of 5′-LDAUG with a 17-             tion may instead be inhibited by U5:DIS base            Under conditions of the ITC experiments, 5′-
nucleotide AUG oligoribonucleotide [G328 to           pairing. Consistent with this hypothesis, muta-         LAUG-HP and 5′-LU5:AUG exist predominantly as
G344 (AUG-17)] promoted dimerization (Fig.            tions in 5′-L designed to enhance U5:DIS base           monomers and dimers, respectively, and native
3B). These findings indicate that dimerization        pairing promoted monomer formation (Fig. 3,             5′-L exists as a 70:30 monomer:dimer equilibri-
is suppressed when AUG exists in a hairpin con-       C and D), whereas mutations to disrupt U5:DIS           um mixture (Fig. 4A). All three RNAs gave rise
formation (AUGHP), that U5:AUG formation              base pairing induced dimerization of 5′-LDAUG           to two-component NC binding isotherms that in-
promotes dimerization, and that dimerization is       (fig. S2). In addition, NMR data obtained for an        cluded an initial exothermic event associated
induced by intramolecular U5:AUG base pairing         lr-AID substituted 5′-LDAUG RNAwith 5′-LDAUG–           with high-affinity NC binding and a subsequent
rather than intermolecular tethering.                 like dimerization, and NC binding properties            endothermic event attributed to NC-induced RNA
    NMR chemical shifts and NOEs associated           (fig. S4) exhibited cross-strand NOEs and NMR           unfolding at high NC:RNA ratios (Fig. 4B and
with A268-H2 in 5′-L, 5′-LDAUG, and an isolated       chemical shifts that are consistent with the pre-       fig. S3) (33). Whereas 5′-LAUG-HP binds 7 T 1 NC
DIS oligo-RNA were similar, indicating that the       dicted U5:DIS interface (Fig. 3, E to G).               molecules with high affinity (Kd = 87 T 3 nM),

Fig. 1. Structure of the HIV-1 5′-leader. (A to C) Secondary structure            subsequent dilution (100-fold) in PI buffer. (H) 1H-13C HMQC NMR data
predictions showing AUG (green) in (B) hairpin and (C) U5:AUG conforma-           obtained for 5′-LAUG* in LI buffer (black) and immediately upon dissolution in
tions. (D) 5′-L forms a monomer at low ionic strength [low ion (LI) buffer: 10    PI buffer (red). Spectral changes correlate with an equilibrium shift toward the
mM Tris-HCl, 10 mM NaCl, pH 7.0] ([RNA] = 60 mM). (E) 2D 1H-13C HMQC              dimer. Residues 349, 351, 352, and 355 are unstructured and detectable, and
NMR spectrum obtained for 5′-L containing 13C-labeled AUG (5′-LAUG*; LI           residues 337 and 341 are rotationally restricted and undetectable, in the [5′-L]2
buffer). (F) Overlay of spectra obtained for 5′-LAUG* (gray) and oligo-AUG*       dimer. (I) Similar spectral changes were observed upon titration of an iso-
hairpin (green). (G) 5′-L exists as a reversible monomer-dimer equilibrium in     lated AUG* hairpin (black) with an unlabeled U5 oligonucleotide, except that
PI buffer ([5′-L] = 30 mM, left). The equilibrium shifts to the monomer upon      all U5:AUG signals were detectable (red).

                                        SCIENCE        VOL 334       14 OCTOBER 2011                                                      243
      [5′-LU5:AUG]2 binds 32 T 2 NC molecules with               We have developed an approach that extends        dard two-dimensional (2D) NMR experiments
      high affinity (Kd = 71 T 3 nM) (Fig. 4B). The ITC      the size of RNAs that can be structurally probed      and conservative base-pair mutagenesis. Our
      profile observed for native 5′-L was similar to        by NMR to ~230 kD. The lr-AID method avoids           findings support a translation/packaging RNA
      that obtained for a 70:30 mixture of 5′-LAUG-HP        NMR relaxation problems associated with hetero-       structural switch mechanism, in which the dimer-
      and 5′-LU5:AUG (Fig. 4, A and B), validating the       nuclear editing, probes interactions over a narrow    promoting GC-rich loop of the DIS hairpin is
      use of 5′-LAUG-HP and 5′-LU5:AUG as models for         distance range (≲5 Å), provides both chemical         sequestered by base pairing with U5 in the
      the native monomer and dimer, respectively.            shift and distance information for structural anal-   AUGHP form of the 5′-leader and is displaced
      The NC binding and dimerization properties of          ysis, and is readily implemented by using stan-       by AUG upon U5:AUG formation (Fig. 4D).
      5′-Llr-AID-U5:AUG and 5′-Llr-AID-U5:DIS were sim-
                                                                                                                         Fig. 2. AUG base pairs with U5 in the
      ilar to those of the 5′-LU5:AUG and 5′-LAUG-HP
                                                                                                                         dimer. (A) lr-AID mutations (red) designed
      controls, respectively (fig. S4), further indicating
                                                                                                                         to probe for predicted U5:AUG base pair-
      that the lr-AID substitutions did not alter the                                                                    ing. Black arrows denote 1H-1H NOEs. (B)
      structure of the RNA. Substitution of the dimer-                                                                   Portions of 1D 1H NMR spectra showing
      promoting GC-rich loop of the DIS hairpin by                                                                       (top) the TAR A46-H2 signal of native [5′-L]2;
      a GAGA tetraloop prevented dimerization of                                                                         (middle) A46G substitution ([5′-LA46G]2)
      5′-LU5:AUG but did not affect its NC-binding prop-                                                                 eliminates the TAR A46-H2 signal; and
      erties (fig. S5), indicating that enhanced NC                                                                      (bottom) A338-H2 signal observed for lr-AID
      binding by the U5:AUG form of the 5′-leader is                                                                     substituted [5′-LA46G]2. (C and D) Similar-
      due to intramolecular conformational changes                                                                       ities in NOE spectra observed for A46-H2
      associated with U5:AUG base pairing and not                                                                        of (C) an isolated TAR oligonucleotide and
      to dimerization per se.                                                                                            (D) native [5′-L]2. (E and F) Similar A338-H2
          The above findings suggested that in vivo RNA                                                                  NOEs observed for (E) an isolated lr-AID U5:
      packaging should be dependent on U5:AUG                                                                            AUG oligonucleotide and (F) lr-AID substi-
      formation, and we therefore measured packag-                                                                       tuted [5′-LA46G]2.
      ing efficiencies of vector RNAs 5′-LAUG-HP and
      5′-LU5:AUG mutations in competition experiments.
      An HIV-1NL4-3 helper construct that expressed
      the native 5′-leader and all viral proteins except
      Env was co-expressed with test vector RNAs with
      5′-LAUG-HP or 5′-LU5:AUG mutations (Fig. 4C).
      Consistent with the structural and NC binding
      studies, the 5′-LU5:AUG RNAs were packaged
      nearly as avidly as the native construct (Fig. 4C,
      lane 5), whereas the 5′-LAUG-HP RNAs exhibited
      severe packaging defects (Fig. 4C, lane 6).
          Previously observed packaging defects asso-
      ciated with mutations in AUG (24, 34) can be
      attributed to defects in U5:AUG–dependent ex-
      posure of NC binding sites rather than inhibition                                                                   Fig. 3. U5:AUG formation promotes di-
      of a cis-packaging mechanism (34). The inability                                                                    merization. (A) 5′-LDAUG (1 mM) forms a
      of helper RNAs to rescue packaging of RNAs                                                                          stable monomer. (B) Addition of AUG-17
      with AUG mutations (24) is likely due to a de-                                                                      (10-fold molar excess) promotes dimeri-
      fect in U5:AUG–dependent exposure of the                                                                            zation of 5′-LDAUG. (C and D) Substitutions
      DIS because rescue requires DIS-mediated het-                                                                       that enhance U5:DIS interactions favor
      erodimer formation. The relationship between                                                                        the monomer. (E) Proposed U5:DIS base
                                                                                                                          pairing and lr-AID sequences used for
      U5:AUG formation and dimerization also ex-
                                                                                                                          structural probing. (F) The A259-H2 sig-
      plains why mutations in AUG can lead to the
                                                                                                                          nal is only observed when U5 and DIS
      production of virions containing genomes that                                                                       both contain lr-AID substitutions. (Top) 5′-
      are sensitive to dissociation by mild denaturants                                                                   LDAUG,A46G. (Middle) 5′-LDAUG,A46G contain-
      (25). Hybrid 5′-leader structures containing both                                                                   ing the lr-AID substitution in the DIS loop
      U5:AUG and AUGHP features—predicted from                                                                            only. (Bottom) 5′-LDAUG,A46G containing lr-
      in vivo ribose reactivity measurements (22)—                                                                        AID substitution in both DIS and U5. (G)
      were not observed with NMR but can be ex-                                                                           2D NOE data for A259-H2 in lr-AID–
      plained by the presence of the AUGHP/U5:AUG                                                                         substituted 5′-LDAUG,A46G, assigned from
      equilibrium. Elevated U5 nucleotide reactivities                                                                    the corresponding U5:DIS oligonucleotide
      observed by traditional chemical probing, which                                                                     control spectrum.
      are incompatible with an exclusive U5:AUG
      structure (21), may also be explained by the ob-
      served AUGHP/U5:AUG equilibrium. The find-
      ing that the gag start codon is exposed in a
      mobile hairpin in the monomeric leader and
      sequestered in the dimer is consistent with ob-
      servations that dimerization attenuates both the
      chemical reactivity of the gag start codon and the
      in vitro translational activity of the genome (16).

244                                           14 OCTOBER 2011           VOL 334       SCIENCE

                                                                                                                                                               Fig. 4. U5:AUG forma-
                                                                                                                                                               tion promotes NC bind-
                                                                                                                                                               ing and packaging. (A)
                                                                                                                                                               5′-LAUG-HP and 5′-LU5:AUG
                                                                                                                                                               form predominantly mono-
                                                                                                                                                               mers and dimers, respec-
                                                                                                                                                               tively, and native 5′-L
                                                                                                                                                               adopts a 70:30 monomer:
                                                                                                                                                               dimer equilibrium (PI buf-
                                                                                                                                                               fer; [RNA] = 0.8 mM).
                                                                                                                                                               Mix, 70:30 mixture of
                                                                                                                                                               5′-LAUG-HP and 5′-LU5:AUG.
                                                                                                                                                               (B) ITC NC titration data
                                                                                                   for samples shown in (A). (C) Packaging of native HIV-1NL4-3 5′-L, 5′-LU5:AUG, and
                                                                                                   5′-LAUG-HP RNAs under competition conditions [assayed by means of ribonuclease
                                                                                             protection (28)]. Lanes 2 to 8: 2, undigested probe; 3, RNA size standards; 4, native
                                                                                             HIV-1NL4-3 helper versus native HIV-1NL4-3 test vector RNAs; 5, native HIV-1NL4-3 helper
                                                                                             versus 5′-LU5:AUG test vector; 6, native HIV-1NL4-3 helper versus 5′-LAUG-HP test vector; 7,
                                                                                             HIV-1NL4-3 helper expressed without test RNA; 8, mock transfected-cells. Bands corre-
                                                                                             sponding to native HIV-1NL4-3 helper RNA (helper) and copackaged test RNAs (test) are
                                                                                             labeled. (D) Nucleotide displacement mechanism regulates dimeric genome packaging.

Conformational changes associated with U5:AUG                         11. R. S. Russell, C. Liang, M. A. Wainberg, Retrovirology              30. E. T. Yu, A. Hawkins, J. Eaton, D. Fabris, Proc. Natl. Acad.
base pairing simultaneously sequester the gag                             1, 23 (2004).                                                           Sci. U.S.A. 105, 12248 (2008).
                                                                      12. J. Greatorex, Retrovirology 1, 23 (2004).                           31. The HIV Sequence Database; available at
start codon and expose the DIS and high-affinity                      13. V. D’Souza, M. F. Summers, Nat. Rev. Microbiol. 3, 643 (2005).          gov/content/sequence/HIV/mainpage.html (2011).
NC binding sites, attenuating translation and pro-                    14. K. Lu, X. Heng, M. F. Summers, J. Mol. Biol. 410, 609 (2011).       32. H. Huthoff, B. Berkhout, RNA 7, 143 (2001).
moting the packaging of a dimeric genome.                             15. G. P. Harrison, A. M. L. Lever, J. Virol. 66, 4144 (1992).          33. M. F. Summers et al., Protein Sci. 1, 563 (1992).
                                                                      16. F. Baudin et al., J. Mol. Biol. 229, 382 (1993).                    34. D. T. Poon, E. N. Chertova, D. E. Ott, Virology 293, 368 (2002).
    References and Notes                                              17. J. Clever, C. Sassetti, T. G. Parslow, J. Virol. 69, 2101 (1995).   Acknowledgments: This research was supported by a grant
 1. J. M. Coffin, S. H. Hughes, H. E. Varmus, Retroviruses            18. M. S. McBride, A. T. Panganiban, J. Virol. 70, 2963 (1996).             (R01 GM42561) from the National Institute of General
    (Cold Spring Harbor Laboratory Press, Plainview, NY, 1997).       19. J. L. Clever, D. Miranda Jr., T. G. Parslow, J. Virol. 76,              Medical Sciences (NIGMS). L.G. was supported by a NIGMS
 2. A. M. Lever, Adv. Pharmacol. 55, 1 (2007).                            12381 (2002).                                                           grant for maximizing doctoral diversity (R25 MBRS-IMSD
 3. A. Onafuwa-Nuga, A. Telesnitsky, Microbiol. Mol.                  20. C. K. Damgaard, E. S. Andersen, B. Knudsen, J. Gorodkin,                GM55036). B.D., K.E., S.J., G.K., and S.B. were supported by
    Biol. Rev. 73, 451 (2009).                                            J. Kjems, J. Mol. Biol. 336, 369 (2004).                                a NIGMS grant for enhancing minority access to research
 4. E. Poole, P. Strappe, H. P. Mok, R. Hicks, A. M. Lever,           21. J. C. Paillart et al., J. Biol. Chem. 279, 48397 (2004).                careers (MARC U*STAR 2T34 GM008663). B.D., G.K., and
    Traffic 6, 741 (2005).                                            22. K. A. Wilkinson et al., PLoS Biol. 6, e96 (2008).                       S.B. were supported by a HHMI undergraduate education
 5. M. D. Moore et al., PLoS Pathog. 5, e1000627 (2009).              23. J. M. Watts et al., Nature 460, 711 (2009).                             grant. We thank S. R. King (Michigan) and HHMI staff
 6. S. B. Kutluay, P. D. Bieniasz, PLoS Pathog. 6, e1001200 (2010).   24. O. Nikolaitchik, T. D. Rhodes, D. Ott, W.-S. Hu, J. Virol.              (UMBC) for technical assistance.
 7. J.-C. Paillart, M. Shehu-Xhilaga, R. Marquet, J. Mak,                 80, 4691 (2006).
    Nat. Rev. Microbiol. 2, 461 (2004).                               25. R. Song, J. Kafaie, M. Laughrea, Biochemistry 47, 3283 (2008).      Supporting Online Material
                                                                      26. F. H.-T. Allain, G. Varani, J. Mol. Biol. 267, 338 (1997).
 8. N. Jouvenet, S. M. Simon, P. D. Bieniasz, Proc. Natl.
                                                                      27. B. S. Tolbert et al., J. Biomol. NMR 47, 205 (2010).                Materials and Methods
    Acad. Sci. U.S.A. 106, 19114 (2009).
                                                                      28. Materials and methods are available as supporting                   Figs. S1 to S5
 9. G. Miele, A. Mouland, G. P. Harrison, E. Cohen,
    A. M. Lever, J. Virol. 70, 944 (1996).                                material on Science Online.                                         References (35–48)
10. T. E. M. Abbink, B. Berkhout, J. Biol. Chem. 278,                 29. D. J. Kerwood, M. J. Cavaluzzi, P. N. Borer, Biochemistry           28 June 2011; accepted 9 September 2011
    11601 (2003).                                                         40, 14518 (2001).                                                   10.1126/science.1210460

                                                                                                                                              tection is dependent on the adaptor molecule that
Successful Transmission of a Retrovirus                                                                                                       signals downstream of most TLRs, expressed by
                                                                                                                                              myeloid differentiation primary response gene
Depends on the Commensal Microbiota                                                                                                           88 (MyD88) (1–3). Retroviruses employ various
                                                                                                                                              mechanisms of immune evasion (4, 5), however,
Melissa Kane,1 Laure K. Case,1* Karyl Kopaskie,1 Alena Kozlova,1 Cameron MacDearmid,1                                                         and can destroy the immune system (e.g., immu-
Alexander V. Chervonsky,2† Tatyana V. Golovkina1†‡                                                                                            nodeficiency viruses of various species) or sub-
                                                                                                                                              vert it (5, 6) to enable successful transmission.
To establish chronic infections, viruses must develop strategies to evade the host’s immune responses.                                        Establishment of a state of immunological tol-
Many retroviruses, including mouse mammary tumor virus (MMTV), are transmitted most efficiently                                               erance to viral proteins in infected animals should
through mucosal surfaces rich in microbiota. We found that MMTV, when ingested by newborn mice,                                               also support virus spread. To test whether an-
stimulates a state of unresponsiveness toward viral antigens. This process required the intestinal                                            imals infected with the orally transmitted MMTV
microbiota, as antibiotic-treated mice or germ-free mice did not transmit infectious virus to their                                           were tolerant to MMTV antigens, we immunized
offspring. MMTV-bound bacterial lipopolysaccharide triggered Toll-like receptor 4 and subsequent
interleukin-6 (IL-6)–dependent induction of the inhibitory cytokine IL-10. Thus, MMTV has evolved to rely                                     1
                                                                                                                                               Department of Microbiology, The University of Chicago,
on the interaction with the microbiota to induce an immune evasion pathway. Together, these findings                                          Chicago, IL 60637, USA. 2Department of Pathology, The Uni-
reveal the fundamental importance of commensal microbiota in viral infections.                                                                versity of Chicago, Chicago, IL 60637, USA.
                                                                                                                                              *Present address: The University of Vermont, Given C249,
                                                                                                                                              89 Beaumont Avenue, Burlington, VT 05405, USA.
       uccessful pathogens have developed means                       their own benefit. Retroviruses, including mouse

S      to counteract the immune system or even
       to use established immune mechanisms to
                                                                      mammary tumor virus (MMTV), are detected by
                                                                      at least one Toll-like receptor (TLR7), and de-
                                                                                                                                              †These authors contributed equally to this work.
                                                                                                                                              ‡To whom correspondence should be addressed. E-mail:

                                                           SCIENCE             VOL 334            14 OCTOBER 2011                                                                        245
      them with virion proteins mixed with an adjuvant.     sion and also implicated the TLR4 ligand, LPS,         of M cells, which are critical for MMTV trans-
      Adult animals that have ingested MMTV-laden           as a key player in this process.                       mission (14). Second, MMTV transmission by GF
      milk as neonates failed to produce antibodies             Thus, the incomplete penetrance of the phe-        MyD88KO mice implies that virus clearance in
      (Abs) against viral proteins (Fig. 1A), whereas       notype (loss of MMTV transmission) in the GF           GF wild-type and TLR4LPS-del animals is MyD88
      response to a control antigen was intact (Fig. 1B).   wild-type group could result from the presence of      dependent.
      In contrast, neonates infected through the par-       LPS in the autoclaved animal feed. The limulus             TLR4 signals through either MyD88 or TIR
      enteral route produced virus-specific Abs (Fig.       amebocyte lysate (LAL) test for LPS established        domain–containing adapter inducing interferon-b
      1A). TLR4-dependent production of the immu-           that a GF mouse may consume up to 500 ng of            (TRIF). We found that TRIF was dispensable for
      noregulatory cytokine interleukin-10 (IL-10) is       LPS per day (fig. S2C). This consumption is not        virus-elicited IL-10 production, whereas MyD88
      required for the persistence of MMTV, and de-         likely to be uniform, which explains the variation     was not (fig. S3C). Thus, MyD88 is required for
      ficiency in TLR4 leads to the eventual loss of        in MMTV loss between different families. In            both TLR4- and TLR7-mediated responses to
      the original virus (6). MMTV-infected TLR4-           fact, we found LPS in the virus fraction of the        the virus.
      deficient C57B10/ScNJ (B10.TLRLps-del) mice           milk from one out of five lactating MMTV-                  Our results suggest that intestinal microbiota
      that eliminated the virus responded to MMTV           infected GF females (fig. S2D).                        are important for successful passage of the virus
      proteins after immunization (Fig. 1A). Thus, both         Infected GF C3H/HeN.MyD88KO and C3H/               through the oral route. Furthermore, reconstitu-
      the oral route of infection and TLR4 sufficiency      HeN.MyD88KOTLR4Lps-del mice transmitted                tion of GF wild-type mice with a defined bac-
      were required for induction of immune tolerance       MMTV without exception through multiple gen-           terial community [altered Schaedler’s flora, ASF
      to MMTV.                                              erations (Fig. 2, C and D, and fig. S3A), similarly    (15)] restored their ability to transmit the virus
          TLR4 is a pattern recognition receptor (7)        to SPF MyD88KO mice (fig. S3B). This observa-          (Fig. 2E). The random (from the viral point of
      with primary specificity for bacterial lipopoly-      tion makes two critical points. First, it alleviates   view) composition of ASF suggests that immune
      saccharide (LPS) (8). Other ligands have been         the concern that the lack of MMTV transmission         subversion is not dependent on a particular type
      proposed to activate this receptor (9), however,      in GF mice may be due to the underdeveloped            of bacteria but requires a bacterially derived
      including MMTV proteins (10). Therefore, MMTV         gut-associated lymphoid tissues (13) or a paucity      ligand, presumably LPS.
      could use either its own proteins or the products
      of the host intestinal microbiota to trigger TLR4.
      To discriminate between these possibilities, we
      used two complementary approaches. First, we
      treated naturally infected pregnant C57BL/6J (B6)
      mice with a complex of broad-spectrum antibiotics
      (fig. S1A). When the first pregnancy offspring
      were immunized as adults, they responded to
      MMTV antigens, unlike their antibiotic-free lit-
      termates (Fig. 1C). Moreover, these mice were
      MMTV-free; no integrated proviruses were found
      in the spleens, and there was no deletion of T
      cells responsive to viral superantigen (SAg)—
      both sensitive indicators of MMTV infection
      (11, 12) (Fig. 1D), even though their mothers con-
      tinued to produce MMTV into the milk (fig. S1B).
      Thus, depletion of the commensal microbiota
      by antibiotics has a profound effect on MMTV
          Second, we produced germ-free (GF) C3H/HeN
      (wild-type) mice, as well as GF C3H/HeN mice
      deficient in TLR4 (C3H/HeN.TLR4Lps-del ), MyD88
      (C3H/HeN.MyD88KO ), or both TLR4 and
      MyD88 (C3H/HeN.TLR4Lps-del MyD88KO ). GF
      females of all four strains and control conven-
      tional specific pathogen–free (SPF) mice were
      injected with filter-sterilized MMTV virions iso-
      lated from the milk of infected SPF wild-type
      mice (generation 0, G0) and allowed to breed and
      to produce progeny (G1). As expected, all G0 fe-      Fig. 1. Unresponsiveness to MMTV is dependent on the oral route of infection and on microbiota. (A and
                                                            B) Mice of indicated strains—MMTV-free or infected by fostering on MMTV-infected lactating MMTV(LA)
      males injected with the virus became infected and
                                                            females or infected intraperitoneally (i.p.) at 3 to 5 days of age—were immunized at 8 weeks of age with
      produced similar amounts of virus into the milk
                                                            either MMTV(LA) proteins (A) or ovalbumin (B) in Freund’s complete adjuvant. Production of specific Abs to
      (Fig. 2A and fig. S2A). Whereas all injected SPF      respective antigens was tested by enzyme-linked immunosorbent assay (ELISA). Sera from four to eight
      mice transmitted infectious virus to their off-       mice were used per group. One of three experiments is shown. (C) The progeny from the first pregnancy of
      spring, the majority of wild-type GF mice (five       uninfected and MMTV(LA)-infected B6 females, treated with a mixture of broad-spectrum antibiotics or left
      families out of seven tested) failed to pass the      untreated, were immunized at 6 weeks of age with MMTV(LA) proteins and tested for virus-specific Abs by
      virus to their offspring (Fig. 2, B and C, and fig.   ELISA (detailed experimental scheme is shown in fig. S1A). n, number of mice. All graphs, means T SEM. (D)
      S2B). It is noteworthy that none of the eight in-     MMTV infection of mice used in (C) was probed by polymerase chain reaction (PCR) amplification (under
      dependent C3H/HeN.TLR4Lps-del GF families             saturating conditions) of integrated proviruses in splenic DNA followed by Southern blot with an MMTV
      transmitted infectious MMTV to their young            long terminal repeat (LTR)–specific probe (lanes 1 to 4, offspring of infected antibiotic-treated females;
      (Fig. 2, B and C). These results supported the        lanes 5 to 10, offspring of infected, untreated females) and by evaluating deletion of SAg-cognate CD4+
      importance of microbiota for MMTV transmis-           Vb6+ T cells. Uninfected B6 mice (n = 5) had 8.5 T 0.3% of CD4+Vb6+ T cells among CD4+ T cells.

246                                          14 OCTOBER 2011            VOL 334       SCIENCE
    Either LPS could directly bind to MMTVor it        stricting viral replication) (Fig. 2D) lacked de-     infection (fig. S5), which suggested that the back-
could stimulate a milieu of inhibitory cytokines       tectable LPS and failed to elicit IL-10 secretion     ground levels of IL-10 in the suckling mice were
independently of viral infection. An in vitro sys-     from wild-type C3H/HeN splenocytes (Fig. 3C           microbiota-independent. Thus, binding to and
tem was used in which addition of MMTV vi-             and fig. S4C). MMTV-GF isolates contained equal       concentrating of LPS by MMTV likely occurs
rions to splenocytes induced production of IL-10       or even greater quantities of the virus compared      in a compartmentalized fashion without affect-
in a TLR4-dependent fashion (6). Polymyxin B           with the control MMTV-SPF isolates (fig. S4, B        ing the global levels of IL-10 in the gut. Never-
(PMB), which binds to and neutralizes LPS ac-          and C). Moreover, MMTV-GF was capable of              theless, LPS, in addition to stimulating IL-10
tivity, alleviated MMTV’s ability to induce IL-10      pelleting LPS during gradient centrifugation and      production, may also induce proliferation of the
secretion by wild-type splenocytes (Fig. 3A). PMB      then stimulating IL-10 production in in vitro         virus’ primary targets and thus increase overall
was not directly toxic to responding cells, be-        splenocyte cultures (Fig. 3C). Furthermore, bind-     viral titers to give the virus an advantage over the
cause IL-10 secretion could be restored by ad-         ing of LPS by MMTV strongly enhances its              immune response.
dition of an excess of LPS in the presence of          ability to induce IL-10 production compared               IL-10 is the cytokine likely responsible for
PMB (Fig. 3B). These data suggest that MMTV            with unbound LPS (Fig. 3, B and D). Note that         the induction of tolerance to MMTV. Whereas
might bind LPS and thereby act as a factor that        MMTV-LPS complexes were more sensitive to             TLR4-sufficient myeloid cells were required for
extracts and concentrates LPS from the environ-        PMB treatment compared with free LPS (Fig.            IL-10 production by B cells (6), the nature of
ment. In fact, MMTV isolated from mammary              3B), which suggests a special type of interaction     the IL-10–induced signal produced by myeloid
glands of lactating females did not contain LPS,       between MMTV, LPS, and TLR4, or involve-              cells was not known. To identify this signal, we
whereas MMTV in the stomachs of pups ingest-           ment of a highly PMB-sensitive type of LPS.           tested splenocytes (Fig. 4A) and cells from gut-
ing their milk was laden with LPS (fig. S4A).          Even the minimal amount of LPS found in a few         associated secondary lymphoid organs (fig. S6)
Gradient centrifugation of MMTV-SPF isolates           GF virus isolates was capable of inducing de-         from cytokine-deficient mice for their ability to
demonstrated that all LPS in the viral preparation     tectable levels of IL-10 in normal splenocytes        secrete IL-10 upon exposure to the virus in vitro.
(and its IL-10–eliciting activity) was MMTV-           (Fig. 3D), which supported the conclusion that        Whereas IL-4, IL-12, and IL-1 were dispensable
bound (Fig. 3C and fig. S4B). Furthermore, no          MMTV propagation in some GF mice (Fig. 2C)            for induction of IL-10, IL-6 was found to be
LPS was pelleted in the absence of the virus,          was indeed aided by contaminating LPS and             critical (Fig. 4A). Secretion of IL-6 was depen-
even when 25 mg of LPS (40 mg/ml) was added            the argument that efficient binding of LPS by         dent on TLR4, but not IL-10, which places TLR4
(1000× the dose of LPS found in virus isolates)        MMTV is an evolutionary adaptation. We also           upstream of IL-6, and IL-6 upstream of IL-10.
during gradient centrifugation (Fig. 3C).              found that background IL-10 levels in the small       B6.CD14KO splenocytes also failed to produce
    MMTV-GF produced by the majority of                intestines of 2-week-old SPF and GF mice were         IL-10 when exposed to MMTV (Fig. 4A), which
MyD88KOTLR4Lps-del mice (incapable of re-              similar and were independent of TLR4 or MMTV          suggests that CD14 (a co-receptor for TLR4

Fig. 2. Differential MMTV persistence in GF and SPF mice of distinct genotypes.   assembled from nonconcurrent portions of the same image. (C) Virus fate in
(A) Reverse transcription–PCR detection of MMTV virion RNA in the milk of GF      the offspring of parenterally infected (G0) mice of different genotypes and
C3H/HeN mice infected parenterally with MMTV(C3H). NC, negative control:          maintenance conditions. Summary of data from MMTV-infected mice, showing
RNA from the milk of an uninfected C3H/HeN mouse; PC, positive control: RNA       virus loss or persistence at G1. (D) Transmission of infectious MMTV(C3H) in GF
from the milk of an SPF mouse injected with the same viral isolate. MMTV(C3H)     MyD88KOTLR4Lps-del mice. Detection of integrated proviruses was done as
and MMTV(LA) were used in two independent studies, data with MMTV(C3H) are        in (B). NC, DNA from the spleen of an uninfected C3H/HeN mouse. A single
shown. To ensure sufficient time for virus amplification in injected mice, the    representative family (a) is shown. (E) Transmission of infectious MMTV(C3H)
second litters by these dams were used for breeding and testing. (B) Splenic      in ASF-associated gnotobiotic wild-type C3H/HeN mice as detected by
DNA from the first progeny (G1) of i.p. injected mice (G0) was subjected to       MMTV(C3H)-specific PCR, followed by Southern blot analysis, performed with
MMTV-specific PCR, followed by Southern blot hybridization with an MMTV           splenic DNA of mice from G0 to G2. a to c, different families. Splenic DNA from
LTR–specific probe. Splenic DNA from an SPF G1 progeny of mice injected with      SPF G1 progeny of mice injected with the same virus isolate served as PC, and
the same virus isolate was used as PC, splenic DNA from an MMTV-negative          splenic DNA from an MMTV-negative C3H/HeN mouse was used as NC. For (A),
C3H/HeN mouse was used as NC. a and b, independent families. The figure is        (B), (D), and (E), PCRs were performed under saturating conditions.

                                        SCIENCE       VOL 334       14 OCTOBER 2011                                                     247
      binding of LPS) also participates in the MMTV-            TLR4 → IL-6 → IL-10) is involved in the                 was followed through multiple generations in
      mediated LPS interaction with TLR4.                       MMTV subversion pathway in vivo, we utilized            infected mouse pedigrees. Both IL-6– and IL-10–
         To confirm by a genetic approach that the              B6.IL-6KO, B6.IL-10KO, B6.CD14KO, B10.                  deficient animals eliminated MMTV in succes-
      proposed chain of events (MMTV+LPS →                      TLR4Lps-del, and B6.TLR2KO mice. MMTV fate              sive generations, as did B10.TLR4Lps-del and

      Fig. 3. MMTV-bound LPS induces production of IL-10. (A) Splenocytes from               assay. In parallel, indicated amounts of LPS (from Escherichia coli serotype
      wild-type C3H/HeN (wt) or C3H/HeN TLR4Lps-del (Lps-del) mice were incu-                026:B6) incubated with PBS or MMTV-GF were ultracentrifuged through a
      bated with an MMTV-SPF isolate in the presence or absence of 0.1 mg/ml                 sucrose cushion and similarly tested. Several fractions (samples a to f) were
      PMB. IL-10 was detected in tissue-culture supernatants by ELISA 16 hours               tested for the ability to elicit IL-10 in in vitro cultures of C3H/HeN spleno-
      later. Graph shows means T SEM from four independent experiments. (B)                  cytes (IL-10 and final LPS concentrations in the in vitro cultures are shown
      Splenocytes from wild-type C3H/HeN mice were incubated with different                  in the table). LPS added: total amount of LPS added to MMTV-GF or PBS
      concentrations of LPS with or without 0.1 mg/ml PMB followed by IL-10                  before centrifugation. ND, not determined. Graph shows means T SEM. (D)
      detection as in (A). In parallel, cells were exposed to an MMTV-SPF isolate in         Four independent MMTV-GF isolates (either LPS-free or containing LPS at
      the presence or absence of 0.1 mg/ml PMB. Concentration of LPS in the                  1 pg/ml) alone with unbound LPS were compared for their ability to induce
      cultures containing the MMTV-SPF isolate was ~13 ng/ml as determined by                IL-10 secretion by splenocytes (left). All virus isolates were normalized by
      LAL assay. Graph shows means T SEM. One of three experiments shown. (C)                ELISA (not shown). MMTV-SPF at several dilutions, MMTV-GF mixed with
      MMTV-SPF or MMTV-GF virions were ultracentrifuged through 30% sucrose                  various concentrations of LPS, and the same concentrations of free LPS were
      with a phosphate-buffered saline (PBS) cushion; the pellet, PBS supernatant            compared for their ability to induce IL-10 in splenocyte cultures. One of
      (PBS fraction), and initial “before spin” fraction were assayed for LPS by LAL         three experiments is shown (right). Graphs show means T SEM.

      Fig. 4. Genetic delineation
      of the immune subversion
      pathway induced by MMTV-
      LPS triggering of TLR4. (A)
      Splenocytes from B6 or B10
      wild-type (used interchange-
      ably), or from indicated knock-
      out or mutant mice were
      incubated with an MMTV-
      SPF isolate followed by de-
      tection of IL-6 and IL-10 in
      supernatants by ELISA. Three
      to five mice were used per
      group. Graphs show means T
      SEM. (B) Virus elimination
      in subsequent generations of mice with deficiencies within the immune sub-             the virus was allowed to produce the next generation of mice, which was also
      version pathway. G0 mice were fostered by SPF MMTV(LA)-infected C3H/HeN                tested to confirm virus loss. MMTV(LA)-infected B6 and B10 mice were used
      females. At least three animals per family were analyzed for hallmarks of              as controls. To control for background modifiers, 129/SvJ mice were also in-
      infection: deletion of SAg-cognate T cells, viral RNA in the milk (table S1 and fig.   cluded, as many targeted knockout mice were originally generated on the
      S8A), and integrated proviruses in spleens (fig. S8B). A family that eliminated        129/SvJ background.

248                                             14 OCTOBER 2011              VOL 334         SCIENCE
CD14KO, but not TLR2KO, mice (Fig. 4B and              interactions with viruses. The lack of knowledge                  16. S. Rakoff-Nahoum, J. Paglino, F. Eslami-Varzaneh,
table S1). Thus, our results support a model (fig.     in this area makes it important to expand such                        S. Edberg, R. Medzhitov, Cell 118, 229 (2004).
                                                                                                                         17. S. Rakoff-Nahoum, R. Medzhitov, Mucosal Immunol. 1,
S7) whereby LPS-induced signaling drives a             investigations to other systems. It is not yet clear                  (Suppl 1), S10 (2008).
viral “subversion” pathway via IL-10 production        whether other viruses take advantage of bacte-                    18. L. Wen et al., Nature 455, 1109 (2008).
that promotes viral transmission to successive         rial products, such as LPS, to achieve successful                 19. M. C. Moreau, G. Corthier, Infect. Immun. 56, 2766
generations.                                           transmission. Retroviruses transmitted through                        (1988).
                                                                                                                         20. N. Sudo et al., J. Immunol. 159, 1739 (1997).
    Detailed analysis of the actual viral load in      mucosal surfaces may also use similar strat-                      21. B. Stecher, W. D. Hardt, Trends Microbiol. 16, 107
subsequent generations of infected IL-10KO and         egies. In humans, the highest risk of human                           (2008).
IL-6KO mice revealed that it was reduced grad-         immunodeficiency virus (HIV) transmission oc-                     22. A. Benson, R. Pifer, C. L. Behrendt, L. V. Hooper,
ually and that it took different numbers of pas-       curs across mucosal surfaces among individuals                        F. Yarovinsky, Cell Host Microbe 6, 187 (2009).
sages for various families to completely eliminate     who practice receptive anal intercourse (26), and                 23. M. J. Wargo, D. A. Hogan, Curr. Opin. Microbiol. 9,
                                                                                                                             359 (2006).
the virus (table S1 and fig. S8). Thus, it appears     risk is also high in infants breastfed by HIV-                    24. K. S. Hayes et al., Science 328, 1391 (2010).
that, in early generations, a high viral load can      infected mothers (27, 28). This study sheds light                 25. T. Ichinohe et al., Proc. Natl. Acad. Sci. U.S.A. 108,
compensate for the loss of the TLR4-dependent          on the previously unknown role of commensal                           5354 (2011).
subversion pathway and overpower the adaptive          microbiota in retroviral pathogenesis and suggests                26. R. A. Royce, A. Seña, W. Cates Jr., M. S. Cohen, N. Engl.
immune response. However, the virus is eventu-         new approaches to the prevention of mucosal trans-                    J. Med. 336, 1072 (1997).
                                                                                                                         27. R. Nduati et al., JAMA 283, 1167 (2000).
ally lost in each infected mouse pedigree, as the      mission, viral-specific vaccination, and therapies.
                                                                                                                         28. J. H. Humphrey et al.; ZVITAMBO study group, BMJ 341,
virus load is reduced with each subsequent pas-                                                                              c6580 (2010).
sage (fig. S8).                                                                                                          Acknowledgments: We thank B. Theriault and A. Vest for
                                                           References and Notes
    Commensal microbiota are required for many          1. E. P. Browne, D. R. Littman, PLoS Pathog. 5, e1000298
                                                                                                                             their help in monitoring gnotobiotic animals. This
homeostatic functions of the intestinal mucosa                                                                               work was supported by T32GM007183 to M.K., K.K.,
                                                           (2009).                                                           and C.M.; by T32 AI065382-01 to L.C.; by Juvenile
and other barrier tissues. Microbiota control tis-      2. A. W. Hardy, D. R. Graham, G. M. Shearer,                         Diabetes Research Foundation grants 2005-204 and
sue repair (16), induction of tolerance to self            J. P. Herbeuval, Proc. Natl. Acad. Sci. U.S.A. 104,               2007-353; by the National Institute of Allergy and
                                                           17453 (2007).
[including tolerance to itself (17) and to auto-        3. M. Kane et al., Immunity 35, 135 (2011).
                                                                                                                             Infectious Diseases (NIAID) NIH AI082418 and the
antigens of the host (18)], and oral tolerance of                                                                            National Institute of Diabetes and Digestive and Kidney
                                                        4. M. H. Malim, M. Emerman, Cell Host Microbe 3, 388 (2008).         Diseases, NIH, Digestive Disease Research Core Center
adults to ingested antigens (19, 20). At present, it    5. U. Dittmer et al., Immunity 20, 293 (2004).                       grant DK42086 to A.V.C.; by National Cancer Institute,
is unknown whether neonatal oral tolerance is           6. B. A. Jude et al., Nat. Immunol. 4, 573 (2003).                   NIH, grant CA100383 and NIAID grant AI090084 to
                                                        7. R. Medzhitov, P. Preston-Hurlburt, C. A. Janeway Jr.,
also dependent on microbiota or whether MMTV                                                                                 T.V.G.; and by a grant (P30 CA014599) to The University
                                                           Nature 388, 394 (1997).                                           of Chicago. A material transfer agreement is required
induces neonatal oral tolerance to itself by a          8. A. Poltorak et al., Science 282, 2085 (1998).                     for use of the SPF and GF C3H-based mutant strains of
unique mechanism available to retroviruses. Com-        9. T. Kawai, S. Akira, Nat. Immunol. 11, 373 (2010).                 mice. The data reported in this paper are tabulated in the
mensals interact with various pathogens and pro-       10. J. C. Rassa, J. L. Meyers, Y. Zhang, R. Kudaravalli,              main paper and the supporting online material.
tect the host against infection with pathogenic            S. R. Ross, Proc. Natl. Acad. Sci. U.S.A. 99, 2281 (2002).
                                                       11. T. V. Golovkina, J. P. Dudley, S. R. Ross, J. Immunol. 161,
and opportunistic bacteria (21), protozoa (22),            2375 (1998).                                                  Supporting Online Material
and fungi (23) or facilitate infection with hel-       12. P. Marrack, E. Kushnir, J. Kappler, Nature 349, 524 
                                                                                                                         Materials and Methods
minthes (24). However, the role of commensal               (1991).
                                                       13. D. A. Hill, D. Artis, Annu. Rev. Immunol. 28, 623 (2010).     Figs. S1 to S8
bacteria in viral transmission and/or pathogenesis                                                                       Table S1
                                                       14. T. V. Golovkina, M. Shlomchik, L. Hannum,
is only beginning to unravel. It is highly likely          A. Chervonsky, Science 286, 1965 (1999).                      References (29–48)
that bacterial microbiota can play both protective     15. F. E. Dewhirst et al., Appl. Environ. Microbiol. 65, 3287     5 July 2011; accepted 24 August 2011
(25) and abetting roles (present report) in their          (1999).                                                       10.1126/science.1210718

                                                                                                                         tions (2). Orally acquired poliovirus undergoes a
Intestinal Microbiota Promote                                                                                            primary replication cycle in the gastrointestinal
                                                                                                                         tract before dissemination. Poliovirus occasion-
Enteric Virus Replication and                                                                                            ally disseminates from the intestine to the central
                                                                                                                         nervous system, which results in paralytic polio-

Systemic Pathogenesis                                                                                                    myelitis days to weeks after initial infection in the
                                                                                                                         gastrointestinal tract. A key question is whether
Sharon K. Kuss,1 Gavin T. Best,1 Chris A. Etheredge,1* Andrea J. Pruijssers,2,3                                          microbiota influence viral replication in the gastro-
Johnna M. Frierson,3,4 Lora V. Hooper,1,5,6 Terence S. Dermody,2,3,4 Julie K. Pfeiffer1†                                 intestinal tract to augment systemic dissemination.
                                                                                                                             To investigate the effect of intestinal micro-
Intestinal bacteria aid host health and limit bacterial pathogen colonization. However, the influence                    biota on poliovirus infection, mice susceptible to
of bacteria on enteric viruses is largely unknown. We depleted the intestinal microbiota of mice with
antibiotics before inoculation with poliovirus, an enteric virus. Antibiotic-treated mice were less                      1
                                                                                                                          Department of Microbiology, University of Texas Southwestern
susceptible to poliovirus disease and supported minimal viral replication in the intestine. Exposure to                  Medical Center, Dallas, TX 75390, USA. 2Department of Pe-
bacteria or their N-acetylglucosamine–containing surface polysaccharides, including lipopolysaccharide                   diatrics, Vanderbilt University School of Medicine, Nashville,
                                                                                                                         TN 37240, USA. 3Elizabeth B. Lamb Center for Pediatric
and peptidoglycan, enhanced poliovirus infectivity. We found that poliovirus binds lipopolysaccharide,
                                                                                                                         Research, Vanderbilt University School of Medicine, Nash-
and exposure of poliovirus to bacteria enhanced host cell association and infection. The pathogenesis                    ville, TN 37240, USA. 4Department of Pathology, Micro-
of reovirus, an unrelated enteric virus, also was more severe in the presence of intestinal microbes.                    biology, and Immunology, Vanderbilt University School of
These results suggest that antibiotic-mediated microbiota depletion diminishes enteric virus infection                   Medicine, Nashville, TN 37240, USA. 5Department of Im-
and that enteric viruses exploit intestinal microbes for replication and transmission.                                   munology, University of Texas Southwestern Medical Center,
                                                                                                                         Dallas, TX 75390, USA, 6Howard Hughes Medical Institute.
                                                                                                                         *Present address: Neurosciences Department, Medical Univer-
       nteric viruses encounter up to 1014 bacteria    affect enteric viruses. Poliovirus is an enteric hu-

E      in the mammalian intestine (1). It is un-
       clear whether commensal microorganisms
                                                       man pathogen transmitted by the fecal-oral route
                                                       and serves as a model for enteric virus infec-
                                                                                                                         sity of South Carolina, Charleston, SC 29425, USA.
                                                                                                                         †To whom correspondence should be addressed. E-mail:

                                           SCIENCE           VOL 334          14 OCTOBER 2011                                                                 249
      poliovirus were treated with antibiotics to deplete     over, poliovirus intestinal replication was equiv-       ease. Collectively, these results indicate that the
      microbes, and viral disease was monitored (fig. S1)     alent in Ifnar1M +/+ and Ifnar1−/− mice, which           microbiota enhance gastrointestinal poliovirus
      (3). Murine poliovirus infection requires expres-       suggested that intestinal replication was IFNAR-         replication.
      sion of the human poliovirus receptor, PVR (4–6).       independent. Because poliovirus infection was le-            We gathered several lines of evidence sug-
      PVR-transgenic mice (PVRtg), however, are not           thal for a fraction of antibiotic-treated mice (Fig.     gesting that diminished poliovirus replication and
      susceptible to oral poliovirus infection unless ren-    1B), it is possible that either minimal viral rep-       disease in antibiotic-treated mice is due to mi-
      dered immunodeficient by interferon-a/b recep-          lication was sufficient for lethality or inoculum        crobiota depletion rather than direct effects of
      tor gene inactivation (PVRtg-Ifnar1−/−) (7, 8).         virus breached the epithelium and replicated             antibiotic treatment. We first tested whether an-
      PVRtg-Ifnar1−/− mice were untreated or treated          in extraintestinal sites, occasionally initiating dis-   tibiotics directly affect poliovirus and found that
      orally with four antibiotics before oral inocu-
      lation with poliovirus. Antibiotic treatment re-
      duced culturable intestinal bacteria by a millionfold
      (Fig. 1A). The mortality of untreated mice was
      twice that of antibiotic-treated mice (Fig. 1B).
      Reintroduction of fecal bacteria into antibiotic-
      treated mice enhanced poliovirus disease, which
      suggested that microbiota promote poliovirus
      pathogenesis. However, when the intestinal
      lumen was bypassed by intraperitoneal inocula-
      tion of poliovirus, pathogenesis was microbiota-
      independent (Fig. 1C and fig. S2). Given that
      orally inoculated poliovirus enters the intestine
      and encounters the large number of bacteria that
      reside there, the microbiota-mediated enhance-
      ment of poliovirus pathogenesis in orally inocu-
      lated mice is likely initiated in the intestine.
           To determine whether mice harboring micro-
                                                              Fig. 1. Poliovirus pathogenesis, shedding, and replication in microbiota-depleted mice. (A) Bacterial
      biota support more efficient poliovirus replication
                                                              loads in feces. PVRtg-Ifnar1−/− mice (n = 4 to 7) were untreated, antibiotic-treated (Abx) for 10 days, or
      than mice with depleted microbiota, we quanti-
                                                              antibiotic-treated for 8 days and recolonized for 2 days with fecal bacteria (Abx + recol). Feces were plated
      fied viral titers from fecal samples (Fig. 1D and       and grown anaerobically, yielding colony-forming units (CFU) per milligram of feces. (B) Survival of
      fig. S3A), because poliovirus was undetectable          PVRtg-Ifnar1−/− mice orally inoculated with poliovirus (untreated, n = 30; Abx, n = 26; Abx + recol, n = 8).
      in intestinal tissue (fig. S4), and minimal intesti-    *P = 0.012, log-rank test. (C) Survival of PVRtg-Ifnar1−/− mice intraperitoneally inoculated with poliovirus
      nal pathology was evident (fig. S5). Peak polio-        (n = 10 mice each). (D) Poliovirus shedding from PVRtg-Ifnar1−/− mice. Mice were orally inoculated with
      virus titers in feces from antibiotic-treated animals   poliovirus, feces were collected (n = 2 to 26 per interval), and poliovirus was isolated and quantified by
      were lower than those from untreated mice, but          plaque assay, yielding plaque-forming units (PFU) per milligram of feces. (E and F) Poliovirus replication
      titers from antibiotic-treated mice were higher at      in intestinal tracts of PVRtg-Ifnar1−/− (E) or PVRtg (F) mice orally inoculated with light-sensitive poliovirus
      later times. Prolonged shedding from antibiotic-        (n = 3 to 9 mice per interval). Feces were harvested, and virus was quantified with or without light
      treated mice was due to slower peristalsis, because     exposure to determine percent replication. Symbols represent means + SEM, *P < 0.05, **P < 0.01,
      dye transit also was delayed (fig. S6) (9). We          Student’s t test. n = 2 to 6 experiments for all.
      postulated that increased poliovirus titers from
      antibiotic-treated mice at late times might be due      Fig. 2. The effects of an-
      to extended shedding of unreplicated inoculum           tibiotic treatment on po-
      virus. To differentiate between replicated and in-      liovirus replication and
      oculum virus, we first quantified fecal shedding of     pathogenesis. (A) Polio-
      poliovirus from nonpermissive mice lacking PVR          virus replication kinetics
      and observed elevated late titers in antibiotic-        in MEFs and HeLa cells
      treated mice, which suggested that total viral titers   with or without antibi-
      in feces and replication are not linked (fig. S3B).     otics. (B) Fecal bacterial
      We then quantified viral replication in PVR mice        loads from untreated or
      using light-sensitive poliovirus. Poliovirus propa-     antibiotic-treated mice har-
      gated in the presence of neutral red dye is sensitive   boring antibiotic-resistant
      to light-induced inactivation by RNA cross-linking      (abxR) bacteria. Feces were
      but loses light sensitivity upon replication in the     plated on rich medium
                                                              with or without four an-
      dark inside mice, facilitating assessment of rep-
                                                              tibiotics. (C) Survival of
      lication (10). We orally inoculated untreated or
                                                              PVRtg-Ifnar1−/− mice oral-
      antibiotic-treated mice with light-sensitive po-
                                                              ly inoculated with polio-
      liovirus and collected feces in the dark. Fecal         virus premixed with four
      viruses were light-exposed or unexposed and             antibiotics (Untreated+abx
      quantified to determine replication status (fig.        PV, n = 9 mice) or polio-
      S7). PVRtg-Ifnar1−/− and PVRtg mice harboring           virus alone in antibiotic-treated mice harboring AbxR bacteria (Abx + abxR, n = 8 mice). (Results from
      microbiota supported efficient intestinal polio-        untreated and antibiotic-treated mice are from Fig. 1B.) (D) Replication of light-sensitive poliovirus in
      virus replication, whereas antibiotic-treated mice      untreated mice receiving poliovirus + antibiotics inoculum and antibiotic-treated mice harboring abxR
      did not (Fig. 1, E and F). Therefore, total fecal       bacteria in comparison with antibiotic-treated mice. (Results from antibiotic-treated mice are from
      titers do not reflect viral replication, a fact only    Fig. 1E.) Each symbol represents mean T SEM. (A) and (B), n = 2 to 5 experiments, (C) and (D) are
      revealed by using light-sensitive viruses. More-        from a representative experiment.

250                                            14 OCTOBER 2011            VOL 334        SCIENCE
                                                                                                                   poliovirus replication kinetics were identical in
                                                                                                                   the presence and absence of antibiotics in HeLa
                                                                                                                   cells and PVRtg mouse embryo fibroblasts (MEFs)
                                                                                                                   (Fig. 2A). We next assayed poliovirus replication
                                                                                                                   and pathogenesis in antibiotic-treated mice har-
                                                                                                                   boring antibiotic-resistant bacteria. For these ex-
                                                                                                                   periments, we treated PVRtg-Ifnar1−/− mice with
                                                                  Fig. 3. Reovirus pathogenesis in microbiota-     antibiotics to select antibiotic-resistant micro-
                                                                  depleted mice. (A) PVRtg-Ifnar1−/− mice          biota (fig. S8). After several weeks, fecal bac-
                                                                  were uninfected, untreated (n = 5), or           teria were insensitive to antibiotics in vitro (Fig.
                                                                  antibiotic-treated (n = 5) or infected           2B). The strain resistant to multiple antibiotics
                                                                  perorally with reovirus, untreated (n =          was identified as Ochrobactrum intermedium,
                                                                  13) or antibiotic-treated (n = 15). Feces        a Gram-negative aerobe, by 16S ribosomal DNA
                                                                  were collected 24 hours post inoculation.        sequencing of fecal-derived subclones (fig. S9).
                                                                  (B) Fecal pathology (table S1). (C) Up-          Poliovirus replicated and was pathogenic in
                                                                  per (top) and lower (bottom) small intes-        antibiotic-treated mice harboring O. intermedium
                                                                  tines were harvested from untreated and          (Fig. 2, C and D). Finally, poliovirus mixed with
antibiotic-treated PVRtg-Ifnar1−/− mice on day 4 post infection or from uninfected mice. Arrows indi-              antibiotics before oral inoculation of mice repli-
cate Peyer’s patches. (D) Quantification of Peyer’s patch sizes [from (C)] from uninfected and infected            cated and was pathogenic (Fig. 2, C and D).
mice. (E) Reovirus titers from day 4 post infection PVRtg-Ifnar1−/−mouse tissues. Plaque assays were               Therefore, diminished poliovirus replication and
performed using murine L929 cells, yielding PFU per milligram of tissue. For (B) to (E), n = 4 to 9                pathogenesis in antibiotic-treated mice is not due
untreated mice, n = 2 to 9 antibiotic-treated mice. Each symbol or bar denotes the mean + SEM. *P <                to direct antiviral effects of antibiotics.
0.05, **P < 0.01, Student’s t test. Scale bars in (A) and (C), 5 mm. (A) and (C), representative of 3 to 5             Because all enteric viruses encounter intes-
experiments; n = 2 to 4 experiments for (B), (D), and (E).                                                         tinal bacteria within the host, we examined the
                                                                                                                   specificity of the microbiota effects using reo-
                                                                                                                   virus, an enteric virus that infects most mam-
                                                                                                                   mals (11). Although immunocompetent adult
                                                                                                                   mice do not display overt reovirus disease symp-
                                                                                                                   toms, immunocompromised adult mice devel-
                                                                                                                   op nonfatal disease after oral inoculation with
                                                                                                                   reovirus strain T3SA+. We orally inoculated un-
                                                                                                                   treated or antibiotic-treated immunocompromised
                                                                                                                   PVRtg-Ifnar1−/− mice with reovirus. Feces from
                                                                                                                   untreated mice were yellow, oily, and hardened,
                                                                                                                   typical of biliary obstruction from T3SA+ reo-