CONTENTS Volume 330 Issue 6012

EDITORIAL BOOKS ET AL.

1724 Model Organisms and Human Health 1747 Evolution Since Darwin
Bruce Alberts M. A. Bell et al., Eds., reviewed by D. P. Mindell
>> Perspective p. 1758; Research Articles 1748 Brain
pp. 1775 & 1787 R. DeSalle et al., curators, reviewed by
A. Rabinowitz and C. E. Schoonover
NEWS OF THE WEEK
1730 Polio Outbreak Breaks the Rules POLICY FORUMS
1731 Court to Weigh University’s Decision 1749 The Challenge of Feeding Scientific Advice
Not to Hire Astronomer into Policy-Making
1732 French Nobelist Escapes ‘Intellectual R. Schenkel
Terror’ to Pursue Radical Ideas in China 1752 Boosting CITES
1733 From Science’s Online Daily News Site J. Phelps et al.
page 1736
1734 Discoverer Asks for Time, Patience Over PERSPECTIVES
Arsenic Bacteria Controversy
1754 Ubiquitination Inhibits Neuronal Exit
1735 From the Science Policy Blog C. Métin and C. Luccardini
>> Report p. 1834
NEWS FOCUS
1755 Generating an Atmosphere
1736 What’s Next for Disease Eradication?
D. P. Cruikshank
Scientists’ New Eradication Target:
>> Report p. 1813
A Word in Their Lexicon
1756 Computational Physics in Film
1740 Altering the Past: China’s
R. Bridson and C. Batty
Faked Fossils Problem
1758 Revealing the Dark Matter of the Genome
1742 CIRM: The Good, the Bad, and the Ugly
M. Blaxter
>> Editorial p. 1724; Research Articles
LETTERS
pp. 1775 & 1787
1744 Genetic Future for Florida Panthers
1759 Stretching Dielectric Elastomer
P. Hedrick
Performance
Response F. Carpi et al.
W. E. Johnson et al.
1761 Germ Cell Genes and Cancer
Biodiversity Transcends Services X. Wu and G. Ruvkun
D. P. Faith >> Report p. 1824
Response 1763 Retrospective: Allan Sandage
C. Perrings et al. (1926–2010)
1746 CORRECTIONS AND CLARIFICATIONS D. Lynden-Bell
page 1749
SCIENCE PRIZE ESSAY
1764 Science 101: Building the Foundations
for Real Understanding
A. Thanukos et al.

CONTENTS continued >>

COVER DEPARTMENTS
An adult Caenorhabditis elegans nematode, ~1 millimeter long, 1722 This Week in Science
pictured along with eggs and young worms. C. elegans is the 1725 Editors’ Choice
first multicellular organism to have its genome fully sequenced, 1726 Science Staff
followed by the fruit fly Drosophila melanogaster. High-resolution 1729 Random Samples
genomic analyses presented on pages 1775 and 1787 provide 1766 AAAS News & Notes
new insights into the organization, structure, and function of the 1839 New Products
genomes of these organisms. See the related Editorial on page 1840 Science Careers
1724 and Perspective on page 1758.
Image: Carolina Biological Supply, Co/Visuals Unlimited, Inc.

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1719

Published by AAAS
 CONTENTS
REVIEW
1768 Has the Microbiota Played a Critical Role
in the Evolution of the Adaptive Immune
System?
Y. K. Lee and S. K. Mazmanian
BREVIA
1774 Decreased Clearance of CNS β-Amyloid
in Alzheimer’s Disease
K. G. Mawuenyega et al.
Alzheimer’s disease is associated with reduced
β-amyloid clearance from the brain.
RESEARCH ARTICLES
1775 Integrative Analysis of the
Caenorhabditis elegans Genome
by the modENCODE Project
M. B. Gerstein et al.
Extensive analysis of the Caenorhabditis
elegans genome reveals regions highly
occupied by multiple transcription factors.
pages 1755 & 1813
1787 Identification of Functional Elements
and Regulatory Circuits by Drosophila
modENCODE
The modENCODE Consortium et al.
The Drosophila modENCODE project
demonstrates the functional regulatory
network of flies.
>> Editorial p. 1724; Perspective p. 1758
REPORTS
1797 High-Flux Solar-Driven Thermochemical
Dissociation of CO2 and H2O Using
Nonstoichiometric Ceria
W. C. Chueh et al.
Solar heating of ceric oxide enables a cycle
for conversion of carbon dioxide to carbon
monoxide or water to hydrogen.
1801 Spin Hall Effect Transistor
J. Wunderlich et al.
pages 1761 & 1824 Manipulation of the spin degree of freedom
of electrons is used to build a spin transistor
without magnetic materials.
1804 Brownian Motion of Stiff Filaments
in a Crowded Environment
N. Fakhri et al.
The thermal motion of single-walled
carbon nanotubes is used to track the
dynamic motion of stiff macromolecules.
1807 Tunable Field Control Over the Binding
Energy of Single Dopants by a Charged
Vacancy in GaAs
D. H. Lee and J. A. Gupta
The electrostatic field of manganese atoms
in gallium arsenide depends on its distance
from a nearby arsenic vacancy site.
1720 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org
Published by AAAS
1810 Dynamics of Magnetic Domain Walls
Under Their Own Inertia
L. Thomas et al.
The current-induced motion of magnetic
domain walls is controlled by the length
of the current pulse.
1813 Cassini Finds an Oxygen–Carbon Dioxide
Atmosphere at Saturn’s Icy Moon Rhea
B. D. Teolis et al.
Rhea’s atmosphere is maintained by chemical
decomposition of surface water ice under
irradiation from Saturn’s magnetosphere.
>> Perspective p. 1755
1816 Structures of C3b in Complex with
Factors B and D Give Insight into
Complement Convertase Formation
F. Forneris et al.
A double-safety–catch mechanism controls
amplification of the complement cascade
during immune responses.
1820 Hsp90 and Environmental Stress
Transform the Adaptive Value
of Natural Genetic Variation
D. F. Jarosz and S. Lindquist
A molecular chaperone both buffers and
potentiates the adaptive nature of genetic
variation in yeast.
1824 Ectopic Expression of Germline Genes
Drives Malignant Brain Tumor Growth
in Drosophila
A. Janic et al.
Inactivation of germline genes suppresses
brain tumor growth in Drosophila.
>> Perspective p. 1761
1827 A Pollen Factor Linking Inter- and
Intraspecific Pollen Rejection in Tomato
W. Li and R. T. Chetelat
The inability to cross with distant relatives in
the nightshade family is linked to mechanisms
preventing self-pollination.
1830 The Social Sense: Susceptibility to Others’
Beliefs in Human Infants and Adults
A. M. Kovács et al.
Knowledge of what others believe is present
in 7-month-old infants.
1834 Siah Regulation of Pard3A Controls
Neuronal Cell Adhesion During
Germinal Zone Exit
J. K. Famulski et al.
A ubiquitination cascade regulates
formation of cell adhesions that immature
neurons require in the developing mouse
brain.
>> Perspective p. 1754

SCIENCEONLINE
SCIENCEXPRESS
www.sciencexpress.org
Crystal Structure of the Eukaryotic 40S Ribosomal
Subunit in Complex with Initiation Factor 1
J. Rabl et al.
The structure provides insight into how protein
synthesis is initiated and into the evolution of
the eukaryotic ribosome.
10.1126/science.1198308
Phosphorylation of ULK1 (hATG1) by
AMP-Activated Protein Kinase Connects
Energy Sensing to Mitophagy
D. F. Egan et al.
A protein kinase links energy stores to control
of autophagy.
10.1126/science.1196371
Induction of Colonic Regulatory T Cells
by Indigenous Clostridium Species
K. Atarashi et al.
Bacteria of the genus Clostridium promote the
induction of suppressor T cells in the colons of mice.
10.1126/science.1198469
LysM-Type Mycorrhizal Receptor Recruited for
Rhizobium Symbiosis in Nonlegume Parasponia
R. Op den Camp et al.
Parasponia uses a mycorrhizal signaling receptor
essential for arbuscle formation to control
rhizobium nodule symbiosis.
10.1126/science.1198181
Intramembrane Cleavage of AMA1 Triggers
Toxoplasma to Switch from an Invasive
to a Replicative Mode
J. M. Santos et al.
Membrane proteins govern a change from invasion
to replication of an intracellular parasite.
10.1126/science.1199284
SCIENCENOW
www.sciencenow.org
Highlights From Our Daily News Coverage
What Makes Glaciers Shake?
Geologists probe the cause of icequakes.
Fearless Woman Lacks Key Part of Brain
Study suggests that the amygdala plays a
crucial role in fear, but not other emotions.
A Dog’s Growl Announces Its Size
Canines know how big their foes are
just by listening to them.
SCIENCESIGNALING
www.sciencesignaling.org
The Signal Transduction Knowledge Environment
RESEARCH ARTICLE: The Mammalian
MAPK/ERK Pathway Exhibits Properties
of a Negative Feedback Amplifier
O. E. Sturm et al.
Analysis of ERK pathway circuitry suggests
appropriate targets for inhibition, providing
a guide for drug development.
www.sciencemag.org
RESEARCH ARTICLE: Cyclic GMP and Protein
Kinase G Control a Src-Containing Mechanosome
in Osteoblasts
H. Rangaswami et al.
PERSPECTIVE: Mechanosomes Carry
a Loaded Message
J. P. Bidwell and F. M. Pavalko
Drugs that activate protein kinase G could mimic the
bone-building effects of mechanical stimulation.
RESEARCH RESOURCE: Phosphoproteomic
Analysis Reveals Interconnected System-Wide
Responses to Perturbations of Kinases and
Phosphatases in Yeast
B. Bodenmiller et al.
PODCAST
B. Bodenmiller et al.
Targeted removal of individual enzymes elicits
changes throughout the entire network of kinases
and phosphatases in yeast.
PERSPECTIVE: T Cell Receptor Signaling
Kinetics Takes the Stage
Y. Sykulev
The kinetics of TCR signaling influence the quality
of the T cell response.
SCIENCECAREERS
www.sciencecareers.org/career_magazine
Free Career Resources for Scientists
The Best of Science Careers, 2010
Science Careers Staff
It was a difficult year for careers in science,
but another good year for Science Careers.
SCIENCETRANSLATIONAL MEDICINE
www.sciencetranslationalmedicine.org
Integrating Medicine and Science
EDITORIAL: 2010: Awards Show What
Translation Can Accomplish
S. Desmond-Hellmann
Four of the most coveted awards in science
celebrate truly translational research.
COMMENTARY: Advancing Translational
Research Collaborations
L. M. Portilla et al.
Barriers to collaboration among academia,
government, and industry must be identified
and overcome to maximize the clinical return
on the investment in science.
MEETING REPORT: Training the Translational
Scientist
R. D. Jackson et al.
MEETING REPORT: Collaborations Among
Academia, Government, and Industry in the
Diagnostics Space: Barriers and Some Ideas
for Solutions
G. Evans and F. Austin
MEETING REPORT: Development of the First
Inhaled Antibiotic for the Treatment of Cystic
Fibrosis
L. M. Rose and R. Neale
SCIENCE VOL 330
Published by AAAS
CONTENTS
MEETING REPORT: Working with the CTSA
Consortium: What We Bring to the Table
S. J. Steele
MEETING REPORT: Academic/Industry Challenges
for Medical Device Development
J. H. Linehan and A. Chaney
These five meeting reports from an NIH forum held in
2010 on promoting collaborations among stakeholders
in translational medicine discuss impediments to such
partnerships and ways to overcome them.
RESEARCH ARTICLE: Calreticulin Is the Dominant
Pro-Phagocytic Signal on Multiple Human
Cancers and Is Counterbalanced by CD47
M. P. Chao et al.
Calreticulin-induced phagocytosis of cancer cells can
be counterbalanced by CD47 expression.
REPORT: Short-Term Monotherapy in HIV-Infected
Patients with a Virus Entry Inhibitor Against the
gp41 Fusion Peptide
W.-G. Forssmann et al.
A natural HIV-1 entry inhibitor targeting the
gp41 fusion peptide shows antiviral potency
in a Phase I/II clinical trial.
SCIENCEPODCAST
www.sciencemag.org/multimedia/podcast
Free Weekly Show
Download the 24 December Science Podcast
to hear a wrap-up of some of the favorite
ScienceNOW stories of 2010.
SCIENCEINSIDER
news.sciencemag.org/scienceinsider
Science Policy News and Analysis
SCIENCE (ISSN 0036-8075) is published weekly on Friday, except the last
week in December, by the American Association for the Advancement of
Science, 1200 New York Avenue, NW, Washington, DC 20005. Periodicals Mail
postage (publication No. 484460) paid at Washington, DC, and additional mailing
offices. Copyright © 2010 by the American Association for the Advancement of
Science. The title SCIENCE is a registered trademark of the AAAS. Domestic individual
membership and subscription (51 issues): $146 ($74 allocated to subscription).
Domestic institutional subscription (51 issues): $910; Foreign postage extra: Mexico,
Caribbean (surface mail) $55; other countries (air assist delivery) $85. First class,
airmail, student, and emeritus rates on request. Canadian rates with GST available
upon request, GST #1254 88122. Publications Mail Agreement Number 1069624.
Printed in the U.S.A.
Change of address: Allow 4 weeks, giving old and new addresses and 8-digit account
number. Postmaster: Send change of address to AAAS, P.O. Box 96178, Washington,
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
material for internal or personal use under circumstances not falling within the fair
use provisions of the Copyright Act is granted by AAAS to libraries and other users
registered with the Copyright Clearance Center (CCC) Transactional Reporting Service,
provided that $20.00 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
Reader’s Guide to Periodical Literature and in several specialized indexes.
24 DECEMBER 2010 1721
 EDITED BY STELLA HURTLEY

Exiting the Birthplace >>

In the developing mammalian brain, new neurons are not always born where
they are needed. In order for immature neurons of the mouse cerebellum to
leave their birthplace in the germinal zone and find their functional niche in
the brain, the neurons need to migrate. Famulski et al. (p. 1834, published
online 25 November; see the Perspective by Métin and Luccardini) now show
that ubiquitin-mediated protein degradation regulates development of specific
cell adhesions that the neurons need in order to exit their birthplace en route to
their final functional location.

Downloaded from www.sciencemag.org on December 23, 2010

A Gutsy Analysis
Efforts to sequence the human microbiome—
the genomes of all the microbes that inhabit our
bodies—have demonstrated its enormous diver-
sity. Analyses to probe the various functions of
the microbiota, particularly of those that reside
in the gut, have revealed that our microbiota
has a profound impact on the development
and function of our immune systems. Lee and
Mazmanian (p. 1768) review how the microbi-
ota influences the development of the adaptive
immune system. Specific species and families
of microbiota support the differentiation of
particular populations of T cells, and alterations
in intestinal microbiota affect the development
of inflammation and autoimmunity.
Tuning Semiconductor
Dopants
the Caenorhabditis elegans genome, and The
modENCODE Consortium (p. 1787) summarize
for the Drosophila melanogaster genome, full
transcriptome analyses over developmental
stages, genome-wide identification of transcrip-
tion factor binding sites, and high-resolution
maps of chromatin organization. Both studies
identified regions of the nematode and fly
genomes that show highly occupied targets (or
HOT) regions where DNA was bound by more
than 15 of the transcription factors analyzed
and the expression of related genes were charac-
terized. Overall, the studies provide insights
into the organization, structure, and function of
the two genomes and provide basic information
needed to guide and correlate both focused and
genome-wide studies.
Fuel from Heat
Plants grow by using energy from the Sun to
Movement in a
Tight Squeeze
The motion of flexible polymer chains in a dense
melt or concentrated solution is described by
reptation theory, in which a single chain is con-
sidered to snake back and forth inside a virtual
confining tube formed by all its neighboring
chains. A number of theories have been pro-
posed for stiffer molecules, but it has been hard
to obtain experimental data to determine the
thermal motion of stiff filaments. Fakhri et al.
(p. 1804) visualized carbon nanotubes directly
as a model system for stiff polymers diffusing in
a gel, and found that even a slight increase in
flexibility significantly sped up diffusion of stiff
filaments under confinement. The rotational dif-
fusion constant grew linearly with the flexibility
and, counterintuitively, did not depend on the
degree of crowding.

Dopants in semiconductors can alter their

conductivity or introduce spin centers that alter
their magnetic properties. Generally, the charge
state of a dopant and field it creates are fixed.
convert carbon dioxide into sugar-based poly-
mers and aromatics. These compounds in turn
can be stripped of their oxygen, either through
millennia of underground degradation to yield
Moving Walls
The current-induced movement of mag-
netic domain walls in magnetic nanowires is

CREDITS (TOP TO BOTTOM): N. KEVITIYAGALA, SOURCE: MÉTIN ET AL.; CHUEH ET AL.

Lee and Gupta (p. 1807, published online 9
December) studied Mn dopants in GaAs with a
low-temperature scanning tunneling microscope
(STM). Using the STM to position As vacancies
at different distances from the Mn dopants
revealed that the As vacancy tuned the local
electrostatic field of the dopant.
From Genome to
Regulatory Networks
For biologists, having a genome in hand is only
the beginning—much more investigation is
still needed to characterize how the genome is
used to help to produce a functional organ-
ism (see the Perspective by Blaxter). In this
vein, Gerstein et al. (p. 1775) summarize for
fossil fuels, or through a rather more rapid pro-
cess of dissolution,
fermentation, and
hydrogenation to
yield biofuels. Can
we use sunlight
to turn CO2 into
hydrocarbon fuel
without relying on
the intervening
steps of plant growth and breakdown? Chueh
et al. (p. 1797) demonstrate one possible
approach, in which concentrated sunlight heats
cerium oxide to a sufficiently high temperature
(~1500°C) to liberate some oxygen from its
lattice. The material then readily strips O atoms
from either water or CO2, yielding hydrogen or
CO, which can then be combined to form fuels.
a candidate for a new architecture in logic
processing and memory. Control-
ling the motion and position of
the domain walls as they move
along the wires in excess of 100
meters per second requires an
understanding of the processes
involved. Thomas et al. (p. 1810)
investigated the dynamics of
magnetic domain wall motion,
looking at the acceleration, constant motion,
and deceleration processes in detail. The whole
process could be described in terms of the
inertia of the domain wall. The distance trav-
eled was simply proportional to the length of
the current pulse used to move the wall, which
should simplify implementation in a circuit or
network architecture.

1722 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

Published by AAAS
 This Week in Science

Extraterrestrial Atmosphere
The detection of oxygen in the atmospheres of Jupiter’s icy moons, Europa and Ganymede, and the
presence of this gas as the main constituent of the atmosphere that surrounds Saturn’s rings, has
suggested the possibility of oxygen atmospheres around the icy moons that orbit inside Saturn’s
magnetosphere. Using the Ion Neutral Mass Spectrometer onboard the Cassini spacecraft, Teolis
et al. (p. 1813, published online 25 November; see the Perspective by Cruikshank) report the de-
tection of a very tenuous oxygen and carbon dioxide atmosphere around Saturn’s icy moon Rhea. As
with other icy satellites, this atmosphere is maintained through the dissociation of surface molecules
and ejection into the atmosphere as a result of Saturn’s magnetospheric radiation.

A Safety Catch on Immune Response

The complement system is an integral part of the innate immune system.
When triggered, it initiates a cascade that marks intruders for elimination

Downloaded from www.sciencemag.org on December 23, 2010

and stimulates immune responses. The key amplification step is cleav-
age of a complex comprising the complement fragment C3b and factor
B (C3bB) by factor D (FD). Forneris et al. (p. 1816) now describe
the crystal structure of C3bB and its complex with FD. The structures
support a mechanism in which membrane-bound C3b stabilizes an open
form of factor B (FB) that has its scissile bond accessible. FD binds through
a site distant from its catalytic center to the open form of FB, which activates FD. The two con-
formational equilibria represent a double safety-catch that would allow tight regulation of this
immune response pathway.

Exploiting Variation
Molecular chaperones help newly synthesized proteins fold, protecting the macromolecular machin-
ery of the cell from various stresses; for example, the highly conserved heat shock proteins (hsp)
protect against elevated temperature. Hsp90 has also been suggested both to buffer against and to
potentiate existing genetic variation in a population. To investigate the generality of these claims,
Jarosz and Lindquist (p. 1820) screened 96 Saccharomyces cerevisiae strains from various ecologi- AAAS is here –
cal niches—soil, fruit, sake, beer, and infected humans—as well as assessed their adaptive value
under different growth conditions. Hsp90 determined the adaptive value of ~20% of the genetic increasing diversity in the
variation in baker’s yeast, with half of the traits being buffered, and half potentiated by hsp90. scientific work force.

Reds Versus Greens

AAAS is working to ensure that
Self-incompatibility (SI) allows plants to prevent inbreeding. Crosses with distant relatives (outbreed- every student with an aptitude for
ing) can also be problematic and is prevented by unilateral interspecific incompatability (UI). In the science, technology, engineer-
nightshade family, SI functions within green-fruited species, whereas crosses between green-fruited ing, and mathematics gets an
and red-fruited species (which includes tomato) results in UI. Li and Chetelat (p. 1827) found a opportunity to pursue a chosen
gene, related to known SI genes within this family, that differs in transcript length and function profession, no matter what the
between individuals that are red-fruited and those that are green-fruited. A survey of species shows challenges. For over 30 years
that the green-fruited species have a functional allele of this gene, whereas the transcript of this AAAS’s ENTRY POINT! program
gene in red-fruited species, which are self-compatible, produce a putatively nonfunctional protein. has placed talented, differently
abled students in paid intern-
These findings suggest that cultivated tomato may have lost the ability to pollinate other species
ships with leading scientific
within the same family, owing to the loss of this protein.
employers. As a AAAS member
your dues support these efforts.
Mind Reading If you’re not yet a AAAS member,
One core component of social cognition, especially of the kind practiced by humans, is the capacity join us. Together we can make
to formulate a representation of what someone else believes to be true, even if that belief is not an- a difference.
chored in reality. Holding two such beliefs in mind—one false and one true—is no simple feat, and
up until a few years ago, it was generally accepted that such a capacity did not arise until children To learn more,
CREDIT: FORNERIS ET AL.

visit aaas.org/plusyou/entrypoint
were 3 to 4 years old. Since then, a flurry of studies, using a variety of interrogation measures, has
suggested that much-younger humans might, in fact, possess this capacity, commonly referred to as
a theory of mind. Kovács et al. (p. 1830) devised an ingenious behavioral paradigm and applied it
both to adults and to infants, which suggests that the representations of others’ beliefs are indeed
formed in the same way in adults and in infants.

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010

Published by AAAS
 EDITORIAL

Model Organisms and Human Health

IN THIS ISSUE OF SCIENCE, WE HIGHLIGHT THE IMPRESSIVE EFFORTS TO DESCRIBE AND ANALYZE Bruce Alberts is Editor-
the genomes of the two organisms—the fly Drosophila melanogaster and the nematode in-Chief of Science.
worm Caenorhabditis elegans—that serve as the best models for understanding the biol-
ogy of all animals, including humans. Hundreds of scientists have collaborated in these
two major studies, which have moved us far beyond the complete descriptions of the DNA
molecules that make up the fly and worm genomes published a little more than a decade
ago, an accomplishment that seemed amazing then. As summarized in the Perspective on
p. 1758, the new findings reveal essentially all of the tens of thousands of RNA and protein

Downloaded from www.sciencemag.org on December 23, 2010

molecules that each of these organisms produces, as well as how their genetic informa-
tion is packaged. Extensive Web-based databases built on these data are freely available
to everyone, greatly accelerating future discoveries. Strange as it may seem, this research,
aimed at reaching a deep molecular understanding of how the bod-
ies of a fly and a worm are formed and maintained, will be critical
for improving human health.
Most of the government funding for biomedical research in the
United States is distributed through the National Institutes of Health.
Its budget of $31 billion in 2010 reflects a widespread public appre-
ciation that biomedical research will lead to great improvements in
human health. Despite the many advances in our understanding of
cells and tissues produced by this research, many diseases remain
incurable. The disparity between the enormous amount now known
about the chemistry and molecular biology of cells and our ability
to intervene in human disease may seem incongruous to the public,
but it is not at all surprising to the scientists involved: As we have
learned more about how cells work, we have been surprised to dis-
cover how enormously sophisticated and complex are the processes
that produce a human being.
Consider just one example. Unlike a bacterium that keeps growing and dividing as long
as food is available, each cell in an animal requires a position-detection system that causes it
to proliferate only when more cells of its type are needed at its particular position in a tissue.
An animal cell behaves as though it contains a tiny computer, assessing the many signals that
it receives from its neighborhood and then deciding whether to maintain itself unchanged
(its usual fate), grow and divide, or kill itself for the good of the entire cell collective. Power-
ful techniques such as those used in these two landmark studies can provide us with lists of
all the molecules involved. But the crucial next challenge, thus far out of reach, is to decipher
exactly how the elaborate networks of signaling molecules that exist inside a cell enable it
to make its crucial decisions—a process analogous to cell “thinking.” Once scientists truly
understand such processes, they will be able to create precise tools to correct harmful cell
behaviors, as when cells multiply out of control in cancer or when they die inappropriately
in degenerative conditions such as Alzheimer’s disease.
CREDITS: (TOP) TOM KOCHEL; (LEFT) WIKIMEDIA COMMONS

The effort to use what we are learning about how cells and organisms work at the molecu-
lar level to improve human health is often called “translational medicine.” The ultimate suc-
cess of this important endeavor will depend on gaining much more knowledge to “translate.”
Because of the long evolutionary process that has given rise to the diverse array of animals
that populate Earth, the molecules and mechanisms that produce humans, flies, and nema-
todes are nearly the same. But unlike humans, flies and worms can be experimentally manip-
ulated, and they have short generation times that allow the complex mechanisms that form
them to be deciphered with powerful genetic tools. And thus we find ourselves in a surpris-
ing position: As incredible as it seems, future research on flies and worms will quite often
provide the shortest and most efficient path to curing human disease.
– Bruce Alberts
Published online 22 December 2010;
10.1126/science.1201826

1724 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

Published by AAAS
 EDITORS’CHOICE
EDITED BY KRISTEN MUELLER AND JAKE YESTON

setting in which the aftermath of a breach of trust

can be studied. De Cremer et al. show that when
students in the role of player 1 were treated un-
fairly (for instance, when receiving only €5), they
judged an apology from player 2 as being less
valued and less apt to induce reconciliation when
the apology was actually received in comparison
to a scenario where they only imagined receiv-
ing one. This disconnect also influenced their
behavior; the students who received the apology
were less trusting in a second round of the game,
whereas those who had imagined the receipt of
an apology were more willing to resume a trust-
ing stance. Why? One possibility is that mental

Downloaded from www.sciencemag.org on December 23, 2010

simulation may have enabled a more effective
repair of social status via a public acknowledg-
ment of the transgression. — GJC
Psychol. Sci. 10.1177/0956797610391101 (2010).
PHYSICS
BIOMEDICINE
Medium-Sized Bang
Travel Assistance
Most models that strive to describe the state of our universe after the Big Bang predict
the existence of the quark-gluon plasma microseconds after the beginning of time. High Tumor metastases are a major cause of death from
temperatures are thought to have supported a state wherein the constituents of atomic solid tumors. Evidence from preclinical models
nucleons—quarks and gluons—existed unbound. In an effort to recreate such conditions, suggests that tumor cells do not metastasize alone
researchers earlier collided gold ions using the Relativistic Heavy Ion Collider (RHIC) and but rather are assisted by specific host cells that
found that the state of matter they created behaved much more like a liquid than a gas. modify the microenvironment of the target organ
Now, similar experiments have been carried out using the Large Hadron Collider (LHC) to so that it can support the survival and growth
collide lead ions at even higher energies than those achieved at RHIC. The results were of newly arriving tumor cells. Two independent
studies of lung metastasis in mice converge on
CREDITS (TOP TO BOTTOM): COURTESY THE ALICE EXPERIMENT AT CERN; DUDA ET AL., PROC. NATL. ACAD. SCI. U.S.A. 107, 21677 (2010)

collected and analyzed by two groups. Aamodt et al. (ALICE collaboration) found that at
these higher energies and temperatures, the quark-gluon plasma still behaves like a (nearly this theme. Kowanetz et al. show that tumor cells
perfect) liquid, implying that it is a strongly interacting system. This conclusion was further secrete granulocyte colony-stimulating factor, a
corroborated by Aad et al. (ATLAS collaboration), who determined that jets of particles pro- protein that expands and mobilizes bone marrow
duced by the collisions in the plasma are strongly quenched by their interaction with the cells of a specific type called Ly6G+Ly6C+ granu-
surrounding medium. — JS locytes and facilitates their homing into the lung
Phys. Rev. Lett. 105, 252302; 252303 (2010). before the arrival of tumor cells. Upon accumula-
tion in the lungs, these granulocytes then secrete
proteins that enhance the invasive properties of
tumor cells, including
EVOLUTION atmospheric CO2 began to fall in the early Oligo- matrix metalloproteinases
cene. The authors measured the carbon isotopic and Bv8, a protein that
Dark Phase Dating
composition of morphologically indistinguishable stimulates tumor cell
Molecular phylogenies indicate that grasses that C3 and C4 pollen grains in order to determine to migration. Duda et al.
transform CO2 through the C4 photosynthetic which metabolic group they belonged, taking provide evidence that the
pathway developed around 30 million years ago, advantage of the large difference in the carbon stability of circulating
long after the appearance of the first C3 grasses isotopic signatures that characterize the two metastatic tumor cells is
60 million years ago or earlier. Fossil and isotopic photosynthetic pathways. The results thus indicate enhanced when they “co-
records do not show the presence of C4 grasses that factors other than decreasing atmospheric travel” with stromal cells
until around 20 million years ago, however, CO2 concentrations must have driven the evolu- derived from the primary Stromal cells
leaving their date of origin poorly constrained. It tion of C4 photosynthesis. — HJS tumor, such as fibroblasts. (green) accompany
has been suggested that the C4 pathway evolved Geology 38, 1091 (2010). Once these cellular clumps metastasizing tu-
in response to the rapid decrease of atmospheric reach the lung, the stromal mor cells (red).
CO2 from 1000 ppm to 500 ppm that occurred PSYCHOLOGY cells appear to provide an
between 30 and 25 million years ago, as C4 early growth advantage to the tumor cells. Further
photosynthesis confers a competitive advantage
Making Up Is Hard exploration of the cells and signaling molecules
over C3 photosynthesis in low-CO2 conditions. The trust game—(i) player 1 gives €10 to player identified in these studies could lead to therapies
Urban et al. present evidence that C4 grasses 2; (ii) that amount of money is tripled; and (iii) that prevent or inhibit metastases. — PAK
already were abundant in southwestern Europe player 2 decides how much of the €30 is given Proc. Natl. Acad. Sci. U.S.A. 107, 21248;
34 million years ago, before the concentration of back to player 1—provides an experimental 21677 (2010).

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1725

Published by AAAS
 EXECUTIVE PUBLISHER Alan I. Leshner
www.sciencemag.org EDITOR-IN-CHIEF Bruce Alberts
PUBLISHER Beth Rosner
EXECUTIVE EDITOR NEWS EDITOR
Monica M. Bradford Colin Norman
MANAGING EDITOR, RESEARCH JOURNALS Katrina L. Kelner F ULFILLMENT S YSTEMS AND O PERATIONS (membership@aaas.org);

1200 New York Avenue, NW
Washington, DC 20005
Editorial: 202-326-6550, FAX 202-289-7562
News: 202-326-6581, FAX 202-371-9227
Bateman House, 82-88 Hills Road
Cambridge, UK CB2 1LQ
+44 (0) 1223 326500, FAX +44 (0) 1223 326501
SUBSCRIPTION SERVICES For change of address, missing issues, new
orders and renewals, and payment questions: 866-434-AAAS (2227)
or 202-326-6417, FAX 202-842-1065. Mailing addresses: AAAS, P.O.
Box 96178, Washington, DC 20090-6178 or AAAS Member Services,
1200 New York Avenue, NW, Washington, DC 20005
I NSTITUTIONAL S ITE L ICENSES please call 202-326-6755 for any
questions or information
REPRINTS: Author Inquiries 800-635-7181
Commercial Inquiries 803-359-4578
PERMISSIONS 202-326-7074, FAX 202-682-0816
MEMBER BENEFITS AAAS/Barnes&Noble.com bookstore www.aaas.org/bn;
AAAS Online Store www.apisource.com/aaas/ code MKB6; AAAS
Travels: Betchart Expeditions 800-252-4910; Apple Store www.
apple.com/eppstore/aaas; Bank of America MasterCard 1-800-833-6262
priority code FAA3YU; Cold Spring Harbor Laboratory Press
Publications www.cshlpress.com/affiliates/aaas.htm; GEICO Auto
Insurance www.geico.com/landingpage/go51.htm?logo=17624;
Hertz 800-654-2200 CDP#343457; Office Depot https://bsd.
officedepot.com/portalLogin.do; Seabury & Smith Life Insurance 800-
424-9883; Subaru VIP Program 202-326-6417; VIP Moving Services
www.vipmayflower.com/domestic/index.html; Other Benefits: AAAS
Member Services 202-326-6417 or www.aaasmember.org.
science_editors@aaas.org (for general editorial queries)
science_letters@aaas.org (for queries about letters)
science_reviews@aaas.org (for returning manuscript reviews)
science_bookrevs@aaas.org (for book review queries)
Published by the American Association for the Advancement of Science
(AAAS), Science serves its readers as a forum for the presentation and
discussion of important issues related to the advancement of science,
including the presentation of minority or conflicting points of view,
rather than by publishing only material on which a consensus has been
reached. Accordingly, all articles published in Science—including
editorials, news and comment, and book reviews—are signed and reflect
the individual views of the authors and not official points of view adopted
by AAAS or the institutions with which the authors are affiliated.
DEPUTY EDITORS R. Brooks Hanson, Barbara R. Jasny, Andrew
M. Sugden
EDITORIAL SENIOR EDITORS/COMMENTARY Lisa D. Chong, Brad Wible; SENIOR
EDITORS Gilbert J. Chin, Pamela J. Hines, Paula A. Kiberstis (Boston), Marc
S. Lavine (Toronto), Beverly A. Purnell, L. Bryan Ray, Guy Riddihough,
H. Jesse Smith, Phillip D. Szuromi (Tennessee), Valda Vinson, Jake
S. Yeston; ASSOCIATE EDITORS Kristen L. Mueller, Jelena Stajic, Sacha
Vignieri, Nicholas S. Wigginton, Laura M. Zahn (San Diego); BOOK REVIEW
EDITOR Sherman J. Suter; ASSOCIATE LETTERS EDITOR Jennifer Sills; EDITORIAL
MANAGER Cara Tate; SENIOR COPY EDITORS Jeffrey E. Cook, Cynthia Howe, Harry
Jach, Lauren Kmec, Barbara P. Ordway, Trista Wagoner; COPY EDITOR Chris
Filiatreau; EDITORIAL COORDINATORS Carolyn Kyle, Beverly Shields; PUBLICATIONS
ASSISTANTS Ramatoulaye Diop, Joi S. Granger, Emily Guise, Jeffrey
Hearn, Michael Hicks, Lisa Johnson, Scott Miller, Jerry Richardson,
Jennifer A. Seibert, Brian White, Anita Wynn; EDITORIAL ASSISTANTS Emily
C. Horton, Patricia M. Moore, Miriam Weinberg; EXECUTIVE ASSISTANT
Alison Crawford; ADMINISTRATIVE SUPPORT Maryrose Madrid; EDITORIAL
FELLOW Melissa R. McCartney
E DITORIAL DIRECTOR , WEB AND NEW MEDIA Stewart Wills; SENIOR WEB
EDITOR Tara S. Marathe; WEB EDITOR Robert Frederick; RESEARCH ASSOCIATE
Corinna Cohn; WEB DEVELOPMENT MANAGER Martyn Green; WEB DEVELOPER
Andrew Whitesell; INTERN Sophia Cai
NEWS DEPUTY NEWS EDITORS Robert Coontz, David Grimm (Online), Eliot
Marshall, Jeffrey Mervis, Leslie Roberts; CONTRIBUTING EDITORS Elizabeth
Culotta, Polly Shulman; NEWS WRITERS Yudhijit Bhattacharjee, Adrian Cho,
Jennifer Couzin, Jocelyn Kaiser, Richard A. Kerr, Eli Kintisch, Greg
Miller, Elizabeth Pennisi, Lauren Schenkman, Robert F. Service (Pacific
NW), Erik Stokstad; WEB DEVELOPER Daniel Berger; INTERN Kristen Minogue;
CONTRIBUTING CORRESPONDENTS Jon Cohen (San Diego, CA),
Daniel Ferber, Ann Gibbons, Sam Kean, Andrew Lawler, Mitch
Leslie, Charles C. Mann, Virginia Morell, Gary Taubes; COPY
EDITORS Linda B. Felaco, Melvin Gatling, Melissa Raimondi; ADMINISTRATIVE
SUPPORT Scherraine Mack; BUREAUS San Diego, CA: 760-942-3252,
FAX 760-942-4979; Pacific Northwest: 503-963-1940
PRODUCTION DIRECTOR Wendy K. Shank; ASSISTANT MANAGER Rebecca
Doshi; SENIOR SPECIALISTS Steve Forrester, Chris Redwood, Anthony
Rosen; PREFLIGHT DIRECTOR David M. Tompkins; MANAGER Marcus Spiegler;
SPECIALIST Jason Hillman
ART DIRECTOR Yael Fitzpatrick; ASSOCIATE ART DIRECTOR Laura Creveling;
SENIOR ILLUSTRATORS Chris Bickel, Katharine Sutliff; ILLUSTRATOR Yana
Hammond; SENIOR ART ASSOCIATES Holly Bishop, Preston Huey,
Nayomi Kevitiyagala; ART ASSOCIATES Kay Engman, Matthew Twombly;
PHOTO EDITOR Leslie Blizard
SCIENCE INTERNATIONAL
EUROPE (science@science-int.co.uk) EDITORIAL: INTERNATIONAL MANAGING
EDITOR Andrew M. Sugden; SENIOR EDITOR/COMMENTARY Julia Fahrenkamp-
Uppenbrink; SENIOR EDITORS Caroline Ash, Stella M. Hurtley, Ian S.
Osborne, Peter Stern; ASSOCIATE EDITOR Maria Cruz; LOCUM EDITOR Helen
DIRECTOR Waylon Butler; CUSTOMER SERVICE SUPERVISOR Pat Butler;
SPECIALISTS Latoya Casteel, LaVonda Crawford, Vicki Linton,
April Marshall; DATA ENTRY SUPERVISOR Cynthia Johnson; SPECIALISTS
Shirlene Hall, Tarrika Hill, William Jones
BUSINESS OPERATIONS AND A DMINISTRATION DIRECTOR Deborah Rivera-
Wienhold; BUSINESS SYSTEMS AND FINANCIAL ANALYSIS DIRECTOR Randy Yi; MANAGER,
BUSINESS ANALYSIS Eric Knott; MANAGER, BUSINESS OPERATIONS Jessica Tierney;
FINANCIAL ANALYSTS Priti Pamnani, Celeste Troxler; RIGHTS AND PERMISSIONS:
ADMINISTRATOR Emilie David; ASSOCIATE Elizabeth Sandler; MARKETING DIRECTOR
Ian King; MARKETING MANAGERS Allison Pritchard, Alison Chandler, Julianne
Wielga; MARKETING ASSOCIATES Aimee Aponte, Mary Ellen Crowley, Wendy
Wise; SENIOR MARKETING EXECUTIVE Jennifer Reeves; DIRECTOR, SITE LICENSING
Tom Ryan; DIRECTOR, CORPORATE RELATIONS Eileen Bernadette Moran; PUBLISHER
RELATIONS, eRESOURCES SPECIALIST Kiki Forsythe; SENIOR PUBLISHER RELATIONS
SPECIALIST Catherine Holland; PUBLISHER RELATIONS, EAST COAST Phillip Smith;
PUBLISHER RELATIONS, WEST COAST Philip Tsolakidis; FULFILLMENT SUPERVISOR
Iquo Edim; FULFILLMENT COORDINATOR Carrie MacDonald, Destiny Pinson; MAR-
KETING MANAGER Christina Schlecht; MARKETING ASSOCIATE Laura Tutino; ELECTRONIC
MEDIA: MANAGER Lizabeth Harman; PROJECT MANAGER Trista Snyder;
ASSISTANT MANAGER Lisa Stanford; SENIOR PRODUCTION SPECIALISTS Ryan Atkins,
Christopher Coleman, COMPUTER SPECIALIST Walter Jones, Kai Zhang;
PRODUCTION SPECIALISTS Antoinette Hodal, Nichele Johnston, Kimberly Oster;
DIRECTOR, WEB AND NEW MEDIA Will Collins
ADVERTISING DIRECTOR, WORLDWIDE AD SALES Bill Moran
COMMERCIAL EDITOR Sean Sanders: 202-326-6430
ASSISTANT COMMERCIAL EDITOR Tianna Hicklin 202-326-6463
PROJECT DIRECTOR, OUTREACH Brianna Blaser
P R O D U C T (science_advertising@aaas.org); M I D W E S T Rick
Bongiovanni: 330-405-7080, FAX 330-405-7081; EAST COAST /
E . CANADA Laurie Faraday: 508-747-9395, FAX 617-507-8189;
WEST COAST/W. CANADA Lynne Stickrod: 415-931-9782, FAX 415-520-
6940; UK/EUROPE/ASIA Roger Goncalves: TEL/FAX +41 43 243 1358;
JAPAN ASCA Corporation, Nanako Ide +81 (0) 3 6802 4616, FAX
+81 (0) 3 6802 4615; ads@sciencemag.jp; SENIOR TRAFFIC ASSOCIATE
Deiandra Simms
WORLDWIDE ASSOCIATE DIRECTOR OF SCIENCE CAREERS Tracy Holmes: +44 (0)
1223 326525, FAX +44 (0) 1223 326532
CLASSIFIED (advertise@sciencecareers.org); U.S.: MIDWEST/WEST COAST/
SOUTH CENTRAL/CANADA Tina Burks: 202-326-6577; EAST COAST/INDUSTRY
Elizabeth Early: 202-326-6578; SALES ADMINISTRATOR: Marci Gallun
SALES COORDINATORS Rohan Edmonson, Shirley Young; EUROPE /
ROW SALES : Susanne Kharraz, Dan Pennington, Alex Palmer; SALES
ASSISTANT Lisa Patterson; JAPAN ASCA Corporation, Jie Chin +81 (0) 3
6802 4616, FAX +81 (0) 3 6802 4615; careerads@sciencemag.jp;
ADVERTISING SUPPORT MANAGER Karen Foote: 202-326-6740; ADVERTISING
PRODUCTION OPERATIONS MANAGER Deborah Tompkins; SENIOR PRODUCTION
SPECIALIST / GRAPHIC DESIGNER Amy Hardcastle; PRODUCTION SPECIALIST

AAAS was founded in 1848 and incorporated in 1874. Its mission is to
advance science, engineering, and innovation throughout the world for
the benefit of all people. The goals of the association are to: enhance
communication among scientists, engineers, and the public; promote and
defend the integrity of science and its use; strengthen support for the
science and technology enterprise; provide a voice for science on societal
issues; promote the responsible use of science in public policy; strengthen
and diversify the science and technology workforce; foster education in
science and technology for everyone; increase public engagement with
science and technology; and advance international cooperation in science.
INFORMATION FOR AUTHORS
See pages 352 and 353 of the 15 January 2010 issue or
access www.sciencemag.org/about/authors
Pickersgill; EDITORIAL SUPPORT Samantha Hogg, Alice Whaley; Yuse Lajiminmuhip; SENIOR TRAFFIC ASSOCIATE Christine Hall
ADMINISTRATIVE SUPPORT John Cannell, Janet Clements, Louise AAAS B OARD OF D IRECTORS RETIRING PRESIDENT , CHAIR Peter C. Agre;
Hartwell; NEWS: EUROPE NEWS EDITOR John Travis; DEPUTY NEWS EDITOR Daniel PRESIDENT Alice Huang; PRESIDENT-ELECT Nina Fedoroff; TREASURER David
Clery; CONTRIBUTING CORRESPONDENTS Michael Balter (Paris), John E. Shaw; CHIEF EXECUTIVE OFFICER Alan I. Leshner; BOARD Linda P. B.
Bohannon (Vienna), Martin Enserink (Amsterdam and Paris), Gretchen Katehi, Nancy Knowlton, Stephen Mayo, Cherry A. Murray, Julia M.
Vogel (Berlin); INTERN Jennifer Carpenter Phillips, Sue V. Rosser, David D. Sabatini, Thomas A. Woolsey
LATIN AMERICA CONTRIBUTING CORRESPONDENT Antonio Regalado
ASIA Japan Office: Asca Corporation, Tomoko Furusawa, Rustic Bldg. 7F,
77 Tenjin-cho, Shinjuku-ku, Tokyo 162-0808, Japan; +81 3
6802 4616, FAX +81 3 6802 4615, inquiry@sciencemag.jp;
ASIA NEWS EDITOR Richard Stone (Beijing: rstone@aaas.org); CONTRIBUTING
CORRESPONDENTS Dennis Normile [Japan: +81 (0) 3 3391 0630, FAX
+81 (0) 3 5936 3531; dnormile@gol.com]; Hao Xin [China:
cindyhao@gmail.com]; Pallava Bagla [South Asia: +91 (0) 11 2271
2896; pbagla@vsnl.com]

SENIOR EDITORIAL BOARD F. Fleming Crim, Univ. of Wisconsin

John I. Brauman, Chair, Stanford Univ.
Richard Losick, Harvard Univ.
Linda Partridge, Univ. College London
Michael S. Turner, University of Chicago
BOARD OF REVIEWING EDITORS
Adriano Aguzzi, Univ. Hospital Zürich
Takuzo Aida, Univ. of Tokyo.
Sonia Altizer, Univ. of Georgia
David Altshuler, Broad Institute
Arturo Alvarez-Buylla, Univ. of California, San Francisco
Richard Amasino, Univ. of Wisconsin, Madison
Angelika Amon, MIT
Kathryn Anderson, Memorial Sloan-Kettering Cancer Center
Siv G. E. Andersson, Uppsala Univ.
Peter Andolfatto, Princeton Univ.
Meinrat O. Andreae, Max Planck Inst., Mainz
John A. Bargh, Yale Univ.
Ben Barres, Stanford Medical School
Marisa Bartolomei, Univ. of Penn. School of Med.
Jordi Bascompte, Estación Biológica de Doñana, CSIC
Facundo Batista, London Research Inst.
Ray H. Baughman, Univ. of Texas, Dallas
David Baum, Univ. of Wisconsin
Yasmine Belkaid, NIAID, NIH
Stephen J. Benkovic, Penn State Univ.
Gregory C. Beroza, Stanford Univ.
Ton Bisseling, Wageningen Univ.
Mina Bissell, Lawrence Berkeley National Lab
Peer Bork, EMBL
Robert W. Boyd, Univ. of Rochester
Paul M. Brakefield, Leiden Univ.
Christian Büchel, Universitätsklinikum Hamburg-Eppendorf
Joseph A. Burns, Cornell Univ.
William P. Butz, Population Reference Bureau
Mats Carlsson, Univ. of Oslo
Mildred Cho, Stanford Univ.
David Clapham, Children’s Hospital, Boston
David Clary, Oxford University
J. M. Claverie, CNRS, Marseille
Jonathan D. Cohen, Princeton Univ.
Andrew Cossins, Univ. of Liverpool
Alan Cowman, Walter & Eliza Hall Inst.
Robert H. Crabtree, Yale Univ.
Wolfgang Cramer, Potsdam Inst. for Climate Impact Research
Jeff L. Dangl, Univ. of North Carolina
Tom Daniel, Univ. of Washington
Stanislas Dehaene, Collège de France
Emmanouil T. Dermitzakis, Univ. of Geneva Medical School
Robert Desimone, MIT
Claude Desplan, New York Univ.
Ap Dijksterhuis, Radboud Univ. of Nijmegen
Dennis Discher, Univ. of Pennsylvania
Scott C. Doney, Woods Hole Oceanographic Inst.
Jennifer A. Doudna, Univ. of California, Berkeley
Julian Downward, Cancer Research UK
Bruce Dunn, Univ. of California, Los Angeles
Christopher Dye, WHO
Michael B. Elowitz, Calif. Inst. of Technology
Gerhard Ertl, Fritz-Haber-Institut, Berlin
Barry Everitt, Univ. of Cambridge
Paul G. Falkowski, Rutgers Univ.
Ernst Fehr, Univ. of Zurich
Tom Fenchel, Univ. of Copenhagen
Alain Fischer, INSERM
Wulfram Gerstner, EPFL Lausanne
Karl-Heinz Glassmeier, Inst. for Geophysics &
Extraterrestrial Physics
Charles Godfray, Univ. of Oxford
Diane Griffin, Johns Hopkins Bloomberg School of
Public Health
Christian Haass, Ludwig Maximilians Univ.
Steven Hahn, Fred Hutchinson Cancer Research Center
Gregory J. Hannon, Cold Spring Harbor Lab.
Niels Hansen, Technical Univ. of Denmark
Dennis L. Hartmann, Univ. of Washington
Chris Hawkesworth, Univ. of St Andrews
Martin Heimann, Max Planck Inst., Jena
James A. Hendler, Rensselaer Polytechnic Inst.
Janet G. Hering, Swiss Fed. Inst. of Aquatic
Science & Technology
Ray Hilborn, Univ. of Washington
Michael E. Himmel, National Renewable Energy Lab.
Kei Hirose, Tokyo Inst. of Technology
Ove Hoegh-Guldberg, Univ. of Queensland
David Holden, Imperial College
Lora Hooper, UT Southwestern Medical Ctr at Dallas
Ronald R. Hoy, Cornell Univ.
Jeffrey A. Hubbell, EPFL Lausanne
Steven Jacobsen, Univ. of California, Los Angeles
Peter Jonas, Universität Freiburg
Barbara B. Kahn, Harvard Medical School
Daniel Kahne, Harvard Univ.
Bernhard Keimer, Max Planck Inst., Stuttgart
Robert Kingston, Harvard Medical School
Hanna Kokko, Univ. of Helsinki
Alberto R. Kornblihtt, Univ. of Buenos Aires
Leonid Kruglyak, Princeton Univ.
Lee Kump, Penn State Univ.
Mitchell A. Lazar, Univ. of Pennsylvania
David Lazer, Harvard Univ.
Virginia Lee, Univ. of Pennsylvania
Ottoline Leyser, Univ. of New York
Olle Lindvall, Univ. Hospital, Lund
Marcia C. Linn, Univ. of California, Berkeley
John Lis, Cornell Univ.
Richard Losick, Harvard Univ.
Jonathan Losos, Harvard Univ.
Ke Lu, Chinese Acad. of Sciences
Laura Machesky, CRUK Beatson Inst. for Cancer Research
Andrew P. MacKenzie, Univ. of St Andrews
Anne Magurran, Univ. of St Andrews
Oscar Marin, CSIC & Univ. Miguel Hernández
Charles Marshall, Univ. of California, Berkeley
Martin M. Matzuk, Baylor College of Medicine
Grahma Medley, Univ. of Warwick
Virginia Miller, UNC, Chapel Hill
Yasushi Miyashita, Univ. of Tokyo
Richard Morris, Univ. of Edinburgh
Edvard Moser, Norwegian Univ. of Science and Technology
Sean Munro, MRC Lab. of Molecular Biology
Naoto Nagaosa, Univ. of Tokyo
James Nelson, Stanford Univ. School of Med.
Timothy W. Nilsen, Case Western Reserve Univ.
Pär Nordlund, Karolinska Inst.
Helga Nowotny, European Research Advisory Board
Stuart H. Orkin, Dana-Farber Cancer Inst.
Christine Ortiz, MIT
Elinor Ostrom, Indiana Univ.
Andrew Oswald, Univ. of Warwick
Jonathan T. Overpeck, Univ. of Arizona
P. David Pearson, Univ. of California, Berkeley
John Pendry, Imperial College
Reginald M. Penner, Univ. of California, Irvine
John H. J. Petrini, Memorial Sloan-Kettering Cancer Center
Simon Phillpot, Univ. of Florida
Philippe Poulin, CNRS
Colin Renfrew, Univ. of Cambridge
Trevor Robbins, Univ. of Cambridge
Barbara A. Romanowicz, Univ. of California, Berkeley
Jens Rostrup-Nielsen, Haldor Topsoe
Edward M. Rubin, Lawrence Berkeley National Lab
Shimon Sakaguchi, Kyoto Univ.
Michael J. Sanderson, Univ. of Arizona
Jürgen Sandkühler, Medical Univ. of Vienna
Randy Seeley, Univ. of Cincinnati
Christine Seidman, Harvard Medical School
Vladimir Shalaev, Purdue Univ.
Joseph Silk, Univ. of Oxford
Montgomery Slatkin, Univ. of California, Berkeley
Davor Solter, Inst. of Medical Biology, Singapore
Allan C. Spradling, Carnegie Institution of Washington
Jonathan Sprent, Garvan Inst. of Medical Research
Elsbeth Stern, ETH Zürich
Yoshiko Takahashi, Nara Inst. of Science and Technology
Jurg Tschopp, Univ. of Lausanne
Herbert Virgin, Washington Univ.
Bert Vogelstein, Johns Hopkins Univ.
Cynthia Volkert, Univ. of Gottingen
Bruce D. Walker, Harvard Medical School
Ian Walmsley, Univ. of Oxford
Christopher A. Walsh, Harvard Medical School
David A. Wardle, Swedish Univ. of Agric Sciences
Colin Watts, Univ. of Dundee
Detlef Weigel, Max Planck Inst., Tübingen
Jonathan Weissman, Univ. of California, San Francisco
Sue Wessler, Univ. of Georgia
Ian A. Wilson, The Scripps Res. Inst.
Timothy D. Wilson, Univ. of Virginia
Xiaoliang Sunney Xie, Harvard Univ.
John R. Yates III, The Scripps Res. Inst.
Jan Zaanen, Leiden Univ.
Mayana Zatz, University of Sao Paolo
Jonathan Zehr, Ocean Sciences
Huda Zoghbi, Baylor College of Medicine
Maria Zuber, MIT
BOOK REVIEW BOARD
John Aldrich, Duke Univ.
David Bloom, Harvard Univ.
Angela Creager, Princeton Univ.
Richard Shweder, Univ. of Chicago
Ed Wasserman, DuPont
Lewis Wolpert, Univ. College London

1726 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

Published by AAAS
 RANDOMSAMPLES
EDITED BY LAUREN SCHENKMAN

“
Making music can often
become a battle of egos. Now
a group of musician-scientists
at Rensselaer Polytechnic Insti-
tute (RPI) in Troy, New York,
THEY SAID IT
has a solution: a computer
program that decides how each “Space and dinosaurs are the two

DIGITAL MAESTRO musician will contribute.

Saxophonist-acoustician
things that turn kids on more than
anything else. If we could grow
Jonas Braasch and his col-
leagues wrote the program to aid them with their unusual improvisation sessions, in which groups of dinosaurs on the space station,
five or more musicians in up to four places around the world jam via the Internet. To keep everyone in we’d have this thing nailed.”
sync, they tried having one musician conduct the troupe. But unfortunately, even with video monitors
showing the remote groups, conductors kept favoring performers within their own groups. —Advice on how to get more young
To overcome the human bias, Braasch went digital. With his RPI colleagues, including electronic people into science from Mark Uhran,

Downloaded from www.sciencemag.org on December 23, 2010

composer Pauline Oliveros and research specialist Doug Van Nort, Braasch combined predictive a NASA director. Uhran spoke at a
algorithms similar to those used in speech recognition software with adaptive, genetic algorithms 10 December meeting about the
to create software that understands features such as musical timing and that can experiment with
international space station at NASA
combinations of instruments. The program conducts the group via symbols that the musicians view
headquarters in Washington, D.C.
on computer monitors as they play.
“The conductor makes intelligent decisions and knows which direction you want to take,” says
Braasch. After a performance, the musicians can log what they think of the conductor’s choices. ”That’s
the way it learns,” he says. Once the program gains mastery, it could adopt the arrogance of real con- sor of public policy at the Georgia Institute of
ductors: It’s also designed to give humans feedback on their performances. Technology in Atlanta, compiled a Web data-
base (www.stemcellstates.net) of the nearly
750 grants totaling $1.25 billion that states
disbursed for adult and embryonic stem cells
States Pick Up the 180
NIH grants State grants
from 2005 through 2009.
Slack on Stem Cells Counting only research grants, since 2007
Funding awarded

155
150 144
($ millions)

125 states have spent at least as much as the U.S.

120
As a court battle rages in the United 90 76 76 National Institutes of Health (NIH) on hESCs
States over the legality of using tax- 60 (see graph), Levine’s team reports in a letter
33 31 35
payer dollars for research on human 30 21 published online 7 December in Nature Bio-
1
embryonic stem cells (hESCs), a recent 0 technology. At least two-thirds of the scientists
05

06

07

08

09

analysis has uncovered a surprising had no NIH grants for hESC work before 2007.
20

20

20

20

20

fact: The six U.S. states that fund this Human Embryonic Stem Cell Research Grants That means if funds dry up in states like Cali-
CREDITS (TOP TO BOTTOM): RENSSELAER POLYTECHNIC INSTITUTE; NATURE BIOTECHNOLOGY 28 2010; NASA

area now spend more on it than the fed- fornia, which leads the pack in funding (see
eral government does. research funds after President George W. Bush p. 1742), it could hit some researchers hard,
California, Connecticut, Illinois, Maryland, limited which hESC lines could be studied with Levine says: “I think there’s a risk of some
New Jersey, and New York launched stem cell federal dollars in 2001. Aaron Levine, a profes- upheaval.”

Last week, a solar storm—a violent explosion from the sun’s surface also known as a coronal mass
ejection—struck Earth 12 hours earlier than space scientist Chris Davis and colleagues had pre-
dicted. That would have been bad news for any astronauts relying on the forecast. But Davis,
Storm Chasers
who works at Rutherford Appleton Laboratory near Didcot, U.K., was pleased: The prediction
was based on data analyzed entirely by volunteers, and for the virtual team’s first effort,
“half a day … isn’t bad,” he says.
Davis and the Royal Observatory, Greenwich, launched the Solar Stormwatch project
(http://solarstormwatch.com) on the citizen-science site Zooniverse in February. Since then,
about 10,000 people have identified and tracked features in images captured by the Helio-
spheric Imager instruments on NASA’s twin STEREO spacecraft, which study solar activity.
Davis suspects the time lag occurred because volunteers track the middle of a storm instead
of the harder-to-spot front. A systematic correction, he thinks, can help future predictions rival
those of the National Oceanic and Atmospheric Administration, which watches storms erupt from the
sun’s surface, then calculates their arrival times (spot on for this storm) using a computer model.
Bringing in about 50 images daily, all requiring human analysis, the imagers used to swamp Davis’s
three-person team. Thanks to the Zooniverse volunteers, the researchers can now chronicle the sun’s current
state nearly in real time, he says: “I feel very privileged having something like 10,000 research assistants.”

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1729

Published by AAAS
 NEWS>> Montagnier to Q&A with author
study “magnetic of arsenic
waves” from DNA bacteria paper

1732 1734
Full assault. The Republic of Congo is vaccinating
adults as well as children.

fecal and other samples were not collected

at the outset, which has hindered subsequent
efforts to reconstruct exactly what happened,
he says. Even now, samples are just trickling
in, and many are of very poor quality, he says.
As a result, the polio labs in Kinshasa, Johan-
nesburg, and CDC have virologically con-

Downloaded from www.sciencemag.org on December 23, 2010

firmed only about two dozen of the suspected
cases—and some will never be confirmed,
Pallansch says.
Genetic analysis has determined that the
culprit is a wild poliovirus type 1 that some-
how jumped from northwest Angola to Congo
in a single importation. That virus, in turn,
INFECTIOUS DISEASE
came from India several years ago and has

Polio Outbreak Breaks the Rules

Polio is a horrendous disease, but it is seldom
fatal—except now. An explosive outbreak
in the Republic of Congo is writing another
chapter in the book on how this ancient
scourge behaves.
Polio usually strikes children under age 5,
paralyzing one in 200 of those infected and
killing at most 5%, occasionally up to 10% in
developing countries. The new outbreak tear-
ing through this West African country has so
far killed an estimated 42% of its victims,
who, in another unusual twist, are mostly
males between the ages of 15 and 25. Since
it began in early October, the outbreak has
paralyzed more than 476 people and killed
at least 179, according to World Health
Organization (WHO) estimates from early
December, making this one of the larg-
est and deadliest polio outbreaks in recent
history. And one of the most mystifying,
too, says polio expert Neal Nathanson of the
University of Pennsylvania: “There are too
many things that don’t fit or are unexpected.”
unrulier neighbor to the east—had rid itself of
polio in 2000 through countrywide campaigns
to vaccinate each and every child. Since
then, routine immunization has kept Congo-
Brazzaville polio-free, even when outbreaks
swept neighboring Angola and D.R.C. “It
was not considered at high risk. That is why
we were all surprised,” says Mark Pallansch,
who is leading efforts to analyze the virus at
the U.S. Centers for Disease Control and Pre-
vention (CDC) in Atlanta.
“Because this is not a typical outbreak that
occurs in children, people initially looked
for another cause,” Pallansch says. Unfortu-
nately, that meant appropriate
CONGO
DEMOCRATIC
REPUBLIC
OF THE CONGO
ANGOLA
been circulating in Angola ever since—and
has also recently spread from there to D.R.C.,
where a separate and distinct outbreak is under
way. In Angola and D.R.C., it is behaving like
“garden variety” polio that strikes young chil-
dren, says Nathanson. “If it is the same type 1
from Angola and India, how can it be behav-
ing so differently” in Congo? he asks.
Nathanson wonders whether something
in addition to polio is going on in Congo-
Brazzaville—perhaps a simultaneous out-
break of another deadly virus with another
route of transmission, although searches for
other strange viruses have so far come up
empty. Until more research is complete, “I
reserve judgment as to what is going on,” he
says. CDC and WHO have rushed in teams
of epidemiologists to help country authori-
ties investigate.
Outbreaks among adults are not unheard
of—one in Namibia in 2006 was traced
back to inadequate routine vaccination some
16 years earlier. But there had been no such
breakdown that anyone knew of in Congo.
Looking back, says Pallansch, there was some
political instability in the mid-1990s around
CREDITS (TOP TO BOTTOM): AP; ADAPTED FROM WHO

Wild poliovirus type 1

“We are scratching our heads,” says Bruce
Aylward of WHO in Geneva, who runs the
troubled 20-plus-year, $8 billion global pro-
gram to eradicate polio (see p. 1736).
When cases of acute flaccid paralysis first
cropped up among adults a couple of months
ago in the oil-rich city of Pointe-Noire on the
Atlantic coast, no one suspected polio. The
Republic of Congo—also known as Congo-
Brazzaville to distinguish it from the Dem-
ocratic Republic of Congo, its larger and
Pointe-Noire. “So maybe there was a disrup-
Cases of paralysis
tion in vaccination during that time to explain
the adult cases. Maybe, but it doesn’t line up
CONGO Brazzaville nicely as it did in Namibia,” he adds. And what
explains the concentration of cases in males—
Pointe-Noire Kinshasa more than 67% of all cases? To Aylward, that’s
DEM. REP. OF THE CONGO the most interesting question.
Cabinda
As for why it is so deadly, Pallansch says
ANGOLA there are a couple of possibilities. When polio
does strike adults, it tends to be more severe,
One hop. The virus jumped from Angola to Congo. progressing more often to the bulbar form of

1730 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

Published by AAAS
 Candidates for China’s faked
eradication fossils

the disease, which leads to cardiac or respira-
tory failure. But that may typically happen in
10% of adult cases, or perhaps even 20%, says
Nathanson, but nothing to rival the figure in
Congo. One theory being investigated is that
there is some confounding factor among those
who died versus those who didn’t: “Did they
have an underlying health problem? Were they
all from the same location?” asks Pallansch.
The other, disturbing possibility is that the
1736
prove it? It will be hard retrospectively to
put all the pieces together.”
All agree that the first priority is to snuff
out the outbreak before the virus reinfects
other countries. It is already spreading: Cases
have been confirmed in the capital, Brazza-
ville, some 650 km to the east of Point-Noire,
in Cabinda, a sliver of Angola that juts out into
the Atlantic south of Congo, and in the adja-
cent province of D.R.C., Bas Congo. Mas-
1740
vaccine.” Aylward predicts that the outbreak
will be under control in 3 to 4 months, if there
is enough money—in November, WHO and
partners issued an emergency appeal for
$23 million—and vaccination campaigns con-
tinue to go well. But there are no guarantees.
Even if Aylward is right, the outbreak
has raised a new, disturbing question, says
Pallansch. Is the Congo-Brazzaville epidemic
an anomaly, or does it suggest there are other

Downloaded from www.sciencemag.org on December 23, 2010

Republic of Congo is in the midst of a much,
much bigger polio outbreak and somehow the
milder cases have been missed.
“Everyone’s got an opinion. But there
are few data,” says Aylward. “You can pull
together a story, but will you ever be able to
sive emergency campaigns are under way to
vaccinate the entire population, all ages, in
Congo and in neighboring parts of Angola
and D.R.C. They seem to be bringing the out-
break in check, says Pallansch: “The popu-
lation is scared, so there is demand for the
polio-free parts of Africa with susceptible
adult populations that would also be ripe for
a explosive epidemic? “We don’t know where
the susceptibles are,” says Aylward—“where
the next Congo could be.”
–LESLIE ROBERTS

S C I E N C E E D U C AT I O N

Court to Weigh University’s Decision Not to Hire Astronomer

Is it possible to separate religious and scien-
tific beliefs when it comes to evolution? A
federal court will take up that question early
next year in the case of Martin Gaskell, an
astrophysicist who claims that the University
of Kentucky (UK) denied him a job because
he is an evangelical Christian.
Pro-evolution advocates say the univer-
sity was well within its rights. “It’s an employ-
ment law case,” says Eugenie Scott, executive
director of the National Center for Science
Education, an organization in Oakland, Cali-
fornia, that lobbies to preserve the teaching of
evolution in public schools. “Can an employer
discriminate based on the scientific knowl-
edge of an employee?” she asks. “Well, yeah.”
But the case could be more complicated.
“It’s a rather intriguing case,” says Ehrich
Koch, an attorney in Minneapolis, Minnesota,
who represented a school district whose reas-
signment of a biology teacher who declined
to teach evolution was upheld. “It appears as
though what the court is saying is both sides
have arguments, and they may be able to prove
their case.”
On 23 November, Judge Karl Forester of
the U.S. District Court for the Eastern District
of Kentucky ruled that UK’s motivation for
CREDIT: MARK DAHMKE
a nontenured position at the University of his e-mail, Gaskell says he is not a creation-
Nebraska to a research fellow post at the ist or a subscriber to intelligent design, both
University of Texas’s McDonald Observa- of which, to varying degrees, discount nat-
tory when he applied to head Kentucky’s ural selection. However, his lecture notes
new observatory in 2007. During the search cite work by astronomer Hugh Ross, who
process, a UK committee member discov- embraces an old Earth, as geologists do, but
ered an article on Gaskell’s rejects evolution as the guid-
personal Web site titled ing principle for life.
“Modern Astronomy, the “I had no trouble with the
Bible, and Creation.” natural selection process,”
The article, based on talks Gaskell said in his deposition.
Gaskell had given, “appeared But “when it comes to trying
to blend science and religion,” to explain everything, and
according to a brief filed by particularly the origin of life,
the university. The dean of the … we just don’t have any sat-
college decided that “since isfactory theory.”
Gaskell’s viewpoint was dis- Jennifer Wiseman, an
cussed in a scholarly paper, astrophysicist who has known
the committee should con- Gaskell professionally for
sider whether his statements In court. It’s astrophysicist 20 years, says she doesn’t
were ‘good science,’ ” accord- Martin Gaskell versus the Uni- consider Gaskell a creation-
ing to the brief. versity of Kentucky in February. ist. “He doesn’t discount or
Gaskell, who was one of disbelieve evolution,” says
three finalists, didn’t get the job, and in 2009 Wiseman, who directs the Dialogue on
he sued the university. In its legal filings, UK Science, Ethics, and Religion program at
says that although there were other reasons AAAS (which publishes Science). A reli-
Gaskell wasn’t hired, “his apparent inability to gious scientist who cites ongoing puzzles in
separate his personal or religious beliefs from evolution sets off more alarms than when an

rejecting Gaskell “remains hotly contested”
and needs to be examined by a jury. Gaskell
is seeking damages for lost income and emo-
tional distress.
Gaskell, 57, had recently moved from
his scientific comments … raised concerns.” atheist makes the same point, she believes.
In an e-mail to Science, Gaskell called The trial is scheduled to begin on 8 Feb-
himself an “old earth theistic evolution- ruary. On 1 March, Gaskell begins work as
ist,” a label that deems evolution a tool God a professor at the University of Valparaiso
used to develop life. In his deposition and in Chile. –JENNIFER COUZIN-FRANKEL

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1731

Published by AAAS
 NEWS OF THE WEEK
N E W S M A K E R I N T E R V I E W : L U C M O N TA G N I E R
French Nobelist Escapes ‘Intellectual
Terror’ to Pursue Radical Ideas in China
PARIS—Virologist and Nobel laureate Luc
Montagnier announced earlier this month
that, at age 78, he will take on the leader-
ship of a new research institute at Jiaotong
University in Shanghai. What has shocked
many scientists, however, isn’t Montagnier’s
departure from France but what he plans to
study in China: electromagnetic waves that
Montagnier says emanate from the highly
diluted DNA of various pathogens.
Montagnier, who won a 2008 Nobel
Prize for his discovery of HIV, claims that
those signals—which he described in two
little-noticed papers in 2009—can reveal the
bacterial or viral origins of many conditions,
including autism and Alzheim-
er’s disease. The work could sug-
gest novel therapies, he says.
But Montagnier’s new direc-
tion evokes one of the most
notorious affairs in French sci-
ence: the “water memory”
study by immunologist Jacques
Benveniste. Benveniste, who
died in 2004, claimed in a 1988
Nature paper that IgE antibodies
have an effect on a certain cell
type even after being diluted by
a factor of 10120. His claim was
interpreted by many as evidence
for homeo pathy, which uses
extreme dilutions that most sci-
entists say can’t possibly have a
biological effect. After a weeklong investi-
gation at Benveniste’s lab, Nature called the
paper a “delusion.”
Science talked to Montagnier, who is
founder and president of the World Foun-
dation for AIDS Research and Prevention,
last week. Questions and answers have been
edited for brevity and clarity.
–MARTIN ENSERINK
Q: Why are you going to Shanghai?
L.M.: I have been offered a professorship and
a new institute, which will bear my name, to
work on a new scientific movement at the
crossroads of physics, biology, and medicine.
The main topic will be this phenomenon of
electromagnetic waves produced by DNA in
water. We will study both the theoretical basis
and the possible applications in medicine.
Q: What exactly are these waves?
L.M.: What we have found is that DNA pro-
1732 24 DECEMBER 2010
duces structural changes in water, which
persist at very high dilutions, and which
lead to resonant electromagnetic signals that
we can measure. Not all DNA produces sig-
nals that we can detect with our device. The
high-intensity signals come from bacterial
and viral DNA.
Q: What do you think are the potential
medical applications?
L.M.: I have found these signals coming
from bacterial DNA in the plasma of many
patients with autism, and also in most, if not
all, patients with Alzheimer, Parkinson’s dis-
ease, and multiple sclerosis. It seems that the
bacteria we are detecting are coming from
the gut. So it is quite possible that products
from gut bacteria end up in the plasma and
cause damage to the brain.
The waves give us a biomarker to test for
the presence of these bacteria, even when we
can’t detect them with classical techniques
like PCR. So when we treat these diseases
with antibiotics, our hope is to see the patho-
gen disappearing. One idea is to set up a clini-
cal trial in autism here in France. We will first
show that we can detect bacterial DNA in the
plasma of autistic children and not in a healthy
control group. Then, if we get agreement from
an ethical committee, autistic children can
be treated with antibiotics to see whether
the DNA signal disappears and their clini-
cal condition improves. In the future, we may
use these findings not just for diagnostics but
also for treatment. It’s possible that electro-
magnetic waves at some frequency will kill
the waves produced by bacterial DNA.
VOL 330 SCIENCE www.sciencemag.org
Published by AAAS
Q: Many of your colleagues seem to be
extremely skeptical.
L.M.: Well, I was skeptical myself in the
beginning. But these are facts. The findings
are very reproducible and we are waiting for
confirmation by other labs.
Q: You have called Benveniste a modern
Galileo. Why?
L.M.: Benveniste was rejected by everybody,
because he was too far ahead. He lost every-
thing, his lab, his money. … I think he was
mostly right, but the problem was that his
results weren’t 100% reproducible.
Q: Do you think there’s something to
Downloaded from www.sciencemag.org on December 23, 2010
homeopathy as well?
L.M.: I can’t say that homeopathy is right in
everything. What I can say now is that the
high dilutions are right. High dilutions of
something are not nothing. They are water
structures which mimic the origi-
nal molecules. We find that with
DNA, we cannot work at the
extremely high dilutions used in
homeopathy; we cannot go further
than a 10-18 dilution, or we lose the
signal. But even at 10-18, you can
calculate that there is not a single
molecule of DNA left. And yet we
detect a signal.
Q: Can’t you pursue this research
in France?
L.M.: I don’t have much fund-
ing here. Because of French
retirement laws, I’m no longer
allowed to work at a public insti-
tute. I have applied for funding
from other sources, but I have been turned
down. There is a kind of fear around this
topic in Europe. I am told that some people
have reproduced Benveniste’s results, but
they are afraid to publish it because of the
intellectual terror from people who don’t
understand it.
Q: Are the Chinese more open to it?
L.M.: I think so. I have visited Jiaotong Uni-
versity several times, and they are quite open-
minded. The editor-in-chief of [Interdisciplin-
ary Sciences: Computational Life Sciences,]
the journal in which I have published two
papers on this topic, is based there as well.
Q: Aren’t you worried that your colleagues
will think you have drifted into pseudo-
CREDIT: NEWSCOM
science?
L.M.: No, because it’s not pseudoscience.
It’s not quackery. These are real phenomena
which deserve further study.
 NEWS OF THE WEEK

ScienceNOW
From Science’s Online Daily News Site

Celestial Ornament
This gossamer ring in the sky may look as light and lovely as a soap
bubble, but its appearance belies its unimaginably violent birth: The
shell of reddish gas is actually the remnants of a supernova explo-
sion riding a shock wave and ripping through space at more than
18 million kilometers per hour. The supernova remnant, dubbed SNR
0509, was first spotted by the Hubble Space Telescope in October
2006. This newly released picture combines data from that 4-year-
old image, which was taken only at wavelengths that highlight glow-
ing hydrogen, with a visible-light image snapped just last month.

Downloaded from www.sciencemag.org on December 23, 2010

http://scim.ag/red-glow

vast majority of icequakes are caused by calv-
ing events.
That means remote observations of ice-
quakes can serve as an early warning system
CREDITS (TOP TO BOTTOM): NASA/ESA AND THE HUBBLE HERITAGE TEAM (STSCI/AURA); HEMERA/THINKSTOCK; TAMÁS FARAGÓ/EÖTVÖS LORÁND UNIVERSITY, HUNGARY
out fearful feeling, they say. But to prove it,
say other researchers, the team needs to look
at more than one subject.
http://scim.ag/no-fear

for changes in calving patterns, says Larsen.

And that could eventually tell glaciologists just
how much ice a glacier is losing—a concern
with a warming climate and the threat of sea-
level rise. http://scim.ag/ice-quake

Fearless Woman Lacks Key Part of Brain

What Makes Glaciers Shake?
A new study has uncovered the cause of most
“icequakes,” seismic tremblings of the earth
that can be felt up to 200 kilometers away.
Glaciologist Chris Larsen of the Geophysi-
cal Institute in Fairbanks, Alaska, and col-
league Shad O’Neel of the U.S. Geological
Survey in Anchorage used 21 seismometer
stations on the southeastern coast of Alaska to
observe icequakes in the St. Elias mountains
for a year and a half. As they will report in an
upcoming issue of the Journal of Geophysical
Research, 85% of the quakes originated at the
ends of glaciers.
Some seismologists had predicted that
icequakes begin in the glacier’s interior, as the
massive slab of ice slides across the ground.
But given the location of the icequakes and
the high rate of calving, or glacial ice falling
into the sea, observed on the glaciers with the
most seismic noise, the team suspects that the
“SM” is a bit of an emotional anomaly. The
44-year-old mother, given those initials to
preserve her anonymity, is pretty much
fearless—and now scientists think they’ve
figured out why.
SM has interested researchers because a
rare genetic condition called Urbach-Wiethe
disease destroyed her amygdala, a pair of
almond-shaped clusters of neurons in the
brain that play a role in fear and anxiety.
Justin Feinstein, a graduate student in clinical
psychology at the University of Iowa in Iowa
City, and colleagues wanted to test whether
SM could experience fear. They recorded her
reaction to snakes and spiders, horror films,
and a trip to a notoriously scary haunted
house. They also gave her an electronic diary
that asked her to rate her current emotional
state several times a day for 3 months.
SM didn’t report feeling scared during
her trials. Meanwhile, the emotion that
received the highest average rating in her
diary was “fearless.”
The results suggest that “the amygdala is
a critical brain region for triggering a state of
fear when an individual encounters threaten-
ing stimuli,” Feinstein and his co-authors
write in Current Biology. It’s the first human
study to show that amygdala damage can wipe
A Dog’s Growl Announces Its Size
Dogs can tell another canine’s size simply
by listening to its growl, a new study reveals.
Péter Pongrácz, an ethologist at Eötvös Loránd
University in Hungary, and colleagues showed
24 canine test subjects images of two dogs
projected onto a screen in front of them.
One image showed a small dog less than
52 centimeters tall; the other image was of
the same dog but projected to be 30% larger.
The researchers then played recorded food-
guarding growls—from either a large or a
small dog—on a speaker placed between the
two projected images. The scientists filmed the
dogs, recording where the canines looked as
they listened to the growls.
Twenty of the dogs looked first and for a
longer period of time at the dog whose size
matched the growl, the team reports online in
PLoS ONE. Matching a sound to a photograph
is a complex cognitive talent previously seen
only in primates, the researchers say.
http://scim.ag/growl-size
Read the full postings, comments, and more at
http://news.sciencemag.org/sciencenow.

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1733

Published by AAAS
 NEWS OF THE WEEK
N E W S M A K E R I NT E RV I E W: F E L I SA WO L F E - S I M O N
Discoverer Asks for Time, Patience
Over Arsenic Bacteria Controversy
Three weeks ago, Felisa Wolfe-Simon, 33, a
former performance oboist with a doctorate
in oceanography and a NASA fellowship in
astrobiology, published a paper in Science
about bacteria that can use arsenic instead
of phosphorus in DNA and other biomol-
ecules (http://scim.ag/arsenicpaper). Four
days before the publication, NASA sent out
a media advisory that it would hold a press
briefing “to discuss an astrobiology finding
that will impact the search for evidence of
extraterrestrial life.” That led to wild specu-
lations on the Web about extraterrestrial life,
and when the paper was published, many
headlines made the most of the “alien” nature
of the discovery by Wolfe-Simon and her
colleagues at the U.S. Geological Survey in
Menlo Park, California. Then came a torrent
of criticism by scientists. A highly critical
blog post by Rosie Redfield, a microbiologist
at the University of British Columbia, Van-
couver, quickly drew hundreds of comments,
many also finding fault with the study. Wolfe-
Simon and her co-author Ronald Oremland
then came under attack by journalists when
they declined to respond to media calls for a
response to these comments. On 16 Decem-
ber, the authors posted responses to some of
the issues on http://scim.ag/arsenicresponse,
and Science will publish technical comments
1734 24 DECEMBER 2010
and responses in early 2011. In the meantime,
Wolfe-Simon agreed to share some of her
thoughts in an interview with Science’s news
department last week. The following has been
edited for brevity; a longer version is avail-
able online at http://scim.ag/arsenicqa.
–ELIZABETH PENNISI
Q: How would you characterize your life
since the press conference?
F.W.-S.: Since the press conference, my life
has been really busy and stressful. When the
paper was accepted for publication, we told
the Astrobiology Program and NASA. …
And when they asked me to come in and talk
about the paper, I said, “Sure.” I thought this
would be great; I’ll bring the information to
the public.
Q: NASA approached you about doing a
press conference, and you thought that was
a good idea?
F.W.-S.: I wouldn’t say I thought it was a good
or bad idea. But we weren’t clearly prepared,
in terms of understanding how it might be,
again, with the new types of media that are
really rather amazing, what was exactly
going to happen. We thought that our find-
ings would generate some discussion, but we
didn’t anticipate the reaction we saw.
VOL 330 SCIENCE www.sciencemag.org
Published by AAAS
Under attack. The media and scientists have del-
uged Felisa Wolfe-Simon with questions.
Q: Why do you think you got the reaction
that you did?
F.W.-S.: I think maybe it has something
to do with that there was some hype gen-
erated around it. I was receiving a lot of
inquiries from all sorts of people—science
journalists and scientists and other sorts of
reporters—even before the paper went out
under embargo. On Monday, NASA had sent
out the media advisory, and it seemed to have
people talking. And I thought, “Oh, we’re all
talking about science.” You know, as a science
communicator and a person, what I’d like to
Downloaded from www.sciencemag.org on December 23, 2010
communicate is how passionate I am about
science and understanding these fundamental
properties and principles of nature.
We, as scientists and other science com-
municators moving forward, need to under-
stand how the Internet gives voice to things we
can’t necessarily anticipate, and I think that
that’s something I will think a lot more about.
Q: You answered questions at the press con-
ference, but then after that, when did you
stop talking to the press?
F.W.-S.: For the press conference, I was pre-
pared to talk about our findings reported in
the paper. I did not show any data, nor did
I describe the study as definitive. I was not
giving a scientific talk, so I was really not
prepared to engage in a scientific debate on
that spot.
Q: After Saturday [4 December], when
[Rosie] Redfield’s blog came out, at least
some journalists took a look at the paper
again and wanted to talk to you. If my
information is correct, that’s when you and
NASA declined to talk to reporters anymore
about this. Is that right?
F.W.-S.: There are two issues. One is that,
well, we wanted to be able to have that
discourse in the scientific community, as
a record. That’s the record, the literature
record that we go back to—or has been up
until now. So that was the one issue, and
the other issue was the rapidness. We spent
a lot of time really crafting our paper and
crafting the SOM [supplemental online
material] and crafting all the data, in
terms of trying to show it as clearly as we
thought. We wanted to give voice to that,
in responses to these queries and some of
CREDIT: GETTY IMAGES
the questions and issues brought up in the
press, and we didn’t want to respond to it
in a way that we thought would not give us
the opportunity to think as deeply as we
might need to. I was under a lot of pressure,

and I’ll be honest, I was exhausted. I would
really be lying if I told you that the barrage
of criticism didn’t hurt. It did. I know my
colleagues in the community aren’t thrilled
or happy about this delay, but, again, I’m
really doing my best.
Q: Some researchers have suggested that
it would be very easy to conclusively tell
whether the arsenic is in DNA using differ-
ent techniques. One was the cesium chlo-
ride density gradient ultracentrifugation.
Did you do those tests?
F.W.-S.: We’re aware of all these other tech-
niques you mentioned. In fact, I have done
a cesium chloride gradient experiment, and
it showed what my gel showed: something
unusual that we couldn’t quite explain.
We could have waited until we did a
really exhaustive selection of all of these
alternative techniques, many requiring col-
laborations with groups well outside of our
field, but instead, I and my co-authors, we
wanted to provide a strongly suggestive and
convincing argument to our community to
initiate these new collaborations and really
inspire other people to go out and do this.
Q: Has anybody asked to collaborate or for
your samples?
F.W.-S.: Yes, absolutely. People have asked
for cells, which we had been already work-
ing on getting to make available. That was
the goal: to stimulate the discussion and
[have] somebody say, “Hey, we have this
technique; could we help you?”
Q: But, do you have the capability of sending
stuff out?
F.W.-S.: We’re currently working to submit
the bacteria to two culture collections so that
they are available to all interested scientists.
Our lab is not currently equipped to provide
the cells at the scale needed to ensure fair
access for everyone. We got requests from a
lot of people.
Q: It’s also been suggested that you didn’t
wash the DNA of any arsenic that may be
stuck to the DNA. What’s your comment
on that?
F.W.-S.: First, we take the cells, and we
collect them by centrifugation, and we
wash them very well. We did a standard
DNA-extraction protocol, multiple phenol
chloroform steps to remove all the impuri-
ties, including things like any free arsenic,
which the washing will have removed. The
DNA fraction that was used for further
analyses, including things like PCR, would
require highly purified DNA. So any of
www.sciencemag.org
those contaminants, if indeed they were
there, would have been a problem [for PCR
amplification]. So, we really don’t feel it’s a
valid concern.
Q: Do you think the other researchers are
overstating how easy it is going to be to
resolve this matter?
F.W.-S.: I would immediately say, “If we’re
lucky, it’s going to be very easy,” but I don’t
think so. The cells are not easy to deal with.
They’re kind of soft and fluffy, and they’re
different. And so testing with the alterna-
tive techniques will fill in more pieces to the
puzzle, and, again, no doubt will open up
new questions.
Q: Were you surprised by the personal tone
of some of the criticism?
F.W.-S.: I’ll be honest. Of course I was dis-
turbed by the personal attacks, and I can
only speak for myself right now. I’ve worked
really hard on this project. I’ve solicited the
advice and assistance from the top scientists
across a variety of different fields—not just
my co-authors, but many, many individuals.
They’ve been incredibly generous with their
time and expertise, and I’m deeply grateful
to everyone who’s helped.
Q: Are you going to start taking media
calls again, or are you going to lay low for
a while?
F.W.-S.: That’s a hard question because, defi-
nitely, in talking to my co-authors, we want
to get to work. We’re scientists, and it’s hard
if all your time is taken up talking. I’m happy
to explain the results, but there’s one thing,
I think, to explain the results, and there’s
another thing to be under what feels like an
attack; it’s hard to do that. There’s only so
many hours in a day.
Q: It sounds like what you want to do is
not really spend much time with the media
right now.
F.W.-S.: What we would want to iterate is
that we’re thrilled that the public is talking
about science. I think the media is an impor-
tant part of the process. We absolutely don’t
want to come off as evasive. We wanted the
time to think. I think the physical volume at
which questions and comments were coming
in, I don’t know how others would respond.
I mean, it was so much and so quick. In fact,
during the press conference, I had a couple
hundred, at least, e-mails coming in. I’m
still on stage. I didn’t have my PDA with
me. When I checked my e-mail later, they’re
demanding, “Answer all my questions right
now.” It’s really hard.
SCIENCE VOL 330 24 DECEMBER 2010
Published by AAAS
NEWS OF THE WEEK
From the Science
Policy Blog
The U.S. government should keep a close
eye on the new field of synthetic biology,
says a report by the president’s bioethics
commission, which doesn’t think new regu-
lations are needed.
http://scim.ag/synthetic_bio_report
A plan to use €1.4 billion in unused 2010
budget funds to fill a gap in 2012–13
caused by the ballooning costs of the ITER
fusion reactor project in France has fallen
Downloaded from www.sciencemag.org on December 23, 2010
apart. The ongoing dispute between the
directly elected European Parliament and its
27 member states could raise the ultimate
cost of the massive project and push it fur-
ther behind schedule.
http://scim.ag/ITER_budget_falters
A new science law in Venezuela gives the
government control over a large pool of tax
money previously spent by companies on
internal research projects or in collabora-
tion with universities. Scientists fear that
the changes could hinder progress across
many fields. http://scim.ag/venezuela_law
A South Korean court has knocked 6 months
off of a 2-year suspended sentence to
disgraced scientist Woo Suk Hwang while
upholding his conviction.
http://scim.ag/shorter_sentence
Personnel moves: Harvard Provost Steven
Hyman is stepping down in June to resume
teaching and research. … Former NIH
Director Elias Zerhouni will become head
of research and development at French
pharmaceutical giant Sanofi-Aventis.
http://scim.ag/hyman_stepsdown
http://scim.ag/zerhouni_job
Beefed-up investments in malaria control
are having a major impact, according to a
new report by the World Health Organiza-
tion (WHO), although some hard-hit coun-
tries are losing ground. … WHO needs a
complete overhaul to remain relevant in
the 21st century, says Jack Chow, a former
high-ranking WHO official, in an online
essay in Foreign Policy.
http://scim.ag/antimalaria_gains
http://scim.ag/WHO_overhaul
For more science policy news, visit http://
news.sciencemag.org/scienceinsider.
1735
 NEWSFOCUS
What’s Next
For Disease
Eradication?
Despite major setbacks, the idea of wiping entire
diseases from the face of the planet hasn’t lost its
appeal. But the rules of the game have changed
IT WAS TIME, ONCE AGAIN, TO BASK IN
the glory and share heroic tales. Late in
August, approximately 260 scientists and
public health leaders met in Rio de Janeiro
to commemorate the 30th anniversary of
what is often considered one of the major
human accomplishments of the 20th cen-
tury: the eradication of smallpox. Leaders of
the global effort—many now in their 70s or
80s—reanalyzed the dramatic 2-decade fight
to obliterate a virus that had killed countless
millions of people.
But many of those present in Rio wished
that by now a younger generation of disease
fighters would have similar victories under
their belt and fresh tales to tell. Thirty years
on, smallpox remains the only human disease
to have been eradicated. Its demise inspired
dreams that many pathogens might be wiped
off the planet, and two eradication cam-
paigns were launched in its wake. But nei-
ther has finished, and many are now question-
ing whether such global operations—which
require extraordinary amounts of devotion
and money—make sense.
That’s why, just 2 days after the commem-
orations ended in Rio, 30 scientists and public
health experts from around the world gathered
1736 24 DECEMBER 2010
for a weeklong meeting in the German city of
Frankfurt am Main* to try to chart a new path
for disease eradication in the 21st century.
Their meeting was triggered by sev-
eral developments. Interest in tackling
global health problems has surged the past
Ernst Strüngmann Forum on Disease Eradication in the
*
Context of Global Health in the 21st Century, Frankfurt
am Main, 29 August–3 September.
SMALLPOX
Agents: Variola major and minor
Reported cases (since 1978): 0
Smallpox killed an estimated
2 million people a year—and grossly
disfigured millions more—before 1959, when hen a
worldwide eradication campaign kicked off. Victory
was declared in 1980. Two labs, one in Russia and
one in the United States, are allowed to retain the
virus, but fears of illicit stocks linger.
VOL 330 SCIENCE www.sciencemag.org
Published by AAAS
Downloaded from www.sciencemag.org on December 23, 2010
decade—as has funding—but the two ongo-
ing eradication campaigns have proven far
more difficult than predicted. In 1986, the
World Health Assembly called for the eradi-
cation of the painful and disfiguring guinea
worm disease; 1995 was chosen as the target
date a few years later. In 1988, polio received
a similar death sentence, to be carried out
by 2000. The deadlines came and went, and
although numbers of cases have plummeted,
both pathogens are still with us. Polio is cur-
rently on a demoralizing rampage through
central Africa and has struck anew in Tajiki-
stan, fueling more doubts about its demise.
Meanwhile, a key rationale for past
CREDITS (TOP TO BOTTOM): MARION KAPLAN/ALAMY; CDC
eradication efforts—the promised financial
windfall from stopping all control measures
once a disease is gone—all but disappeared
as a result of 9/11 and the 2001 anthrax let-
ters. Wealthy countries in particular are
determined never to let their guard down
against diseases like smallpox, polio, or
measles. Meanwhile, developing countries
have their own questions: Why should they
keep spending inordinate amounts of time
and money on a disease such as polio—now
down to fewer than 2000 cases a year—
while their health systems are struggling

Forever gone. An Ethiopian child is vaccinated
in 1976 during the final stage of the smallpox-
eradication campaign.
with far more devastating diseases such as
AIDS and TB?
And yet, getting rid of a disease once and
for all will never lose its appeal, says Walter
Dowdle, a consultant for the Task Force for
Global Health in Atlanta. Eradication cam-
paigns offer the inspiring promise of perpet-
ual benefits and the chance to write health
history. The 2007 call by Bill and Melinda
Gates to eradicate malaria, for instance, has
reenergized many of those working on the
disease—even though the couple was criti-
cized for using the term prematurely.
POLIO
Agent: Poliovirus type 1 and 3
Reported cases (2009): 1604
Two decades into the polio-eradication
on
campaign, the annual number of chil-
dren paralyzed by the disease has come
down by more than 99%. But scientific and
political problems make the home stretch
frustratingly difficult, and some have suggested
calling it quits.
So the Frankfurt meeting sought a new
way forward. The participants, many of
them involved in past and current eradica-
tions, believed that eradication campaigns
should continue. But “proceed with caution”
could have been the motto of the 44-page
CREDITS: MICROSCOPIC PHOTOS (TOP TO BOTTOM): MARY NG MAH LEE/CDC; CDC (2)
draft report cobbled together on the 5th and
6th day. Future eradication campaigns “will
be put under the microscope in a way that
smallpox or polio never were,” says Stephen
Cochi of the U.S. Centers for Disease Con-
trol and Prevention (CDC) in Atlanta, who
co-organized the meeting and has been heav-
ily involved in the polio campaign.
New eradication plans must be more
evidence-based than the old ones, par-
ticipants concluded. There should be an
analysis of economic costs and benefits, a
thorough funding plan, and new financial
tricks to prevent perennial budget gaps like
those hampering the polio campaign. Any
new eradication program should also help
poor countries build stronger health systems
along the way, the report said. A smart com-
www.sciencemag.org SCIENCE VOL 330
munication plan that reaches out to marginal-
ized populations is key.
Much of that was never done—or done on
the fly—for polio, says Cochi. “We were well-
meaning but totally naive,” he says. “We built
the boat as we sailed,” adds T. Jacob John, a
member of the India Expert Advisory Group
for Polio Eradication.
Annihilation
The concept of disease eradication has a long
history of high hopes and dashed dreams.
Edward Jenner, the British doctor who pio-
neered the smallpox vaccine, realized the
huge potential of his discovery. “It now
becomes too manifest to admit of contro-
versy,” he wrote in 1801, “that the annihilation
of the Small Pox, the most dreadful scourge
of the human species, must be the result of
this practice.” It would take 180 years—but
at least it came true. In the meantime, several
other eradication efforts died, and their ghosts
haunt the field to this day.
The problem was often that scientists
underestimated the problem. In 1955, the
World Health Assembly endorsed a plan
to eradicate malaria through an aggressive
mosquito-control program that relied on
spraying DDT inside homes. In the 1960s,
it became evident that the strategy wouldn’t
work, in part because the insects began to
develop resistance against DDT, an insecti-
cide that came under fire from environmen-
talists as well. Eventually, political will and
money ran out. The 1952-to-1964 attempt to
eradicate yaws, an easily treatable disease
caused by the spirochete Treponema palli-
dum pertenue that leads to disfiguring skin
lesions, met a similar fate.
The smallpox campaign proved that erad-
ication was feasible given the right tools—
in this case, a very effective single-dose
GUINEA WORM DISEASE
Agent: Dracunculus medinensis
Reported cases (Jan.-Sep. 2010): 1626
Before an eradication campaign
began in 1986, 20 African and Asian
countries had an estimated 3.5 million
cases of this painful infection, also known as
dracunculiasis. Now only a few thousand cases a
year occur in four countries; southern Sudan, home
to more than 95% of them, is set to become the final
battleground.
24 DECEMBER 2010
Published by AAAS
NEWSFOCUS
SARS
Agent: The SARS coronavirus
Reported cases (since 2004): 0
SARS, a highly fatal respiratory infec-
ec-
tion that erupted in China in November er
2002, was mopped up 8 months later thanks
to intense global efforts to isolate patients and
quarantine their contacts. Nobody called it an
eradication campaign at the time—but in hindsight,
some scientists say, why not?
Downloaded from www.sciencemag.org on December 23, 2010
vaccine—and offered enthusiasts a fresh
argument to forge ahead with new pathogens.
The experiences since then with polio and
guinea worm have provided new lessons in
modesty—and new pitfalls to avoid.
With polio, like with malaria, part of the
problem is that the tools aren’t working as
well as scientists believed. The oral polio vac-
cine is much less effective than the small-
pox vaccine, especially in some regions of
the world. And while it’s easy to recognize a
smallpox case—and roll out a mop-up cam-
paign to protect everyone in the area—polio
can cause silent outbreaks; hundreds of chil-
dren can be infected before one develops
paralysis or dies.
On top of that came what smallpox veteran
William Foege calls “people problems.” War
and civil unrest have sidetracked vaccination
campaigns, and in 2004, suspicions about
vaccine safety led to a vaccine boycott in
northern Nigeria that sparked new outbreaks
in 20 polio-free countries (Science, 6 Febru-
ary 2009, p. 702). This year, there have been
458 cases in Tajikistan—the first epidemic in
the region since it was declared polio-free in
2002—and an explosive new outbreak in the
Republic of Congo (see p. 1730). The battle
has been so difficult that some suggested in
2006 that it was time to throw in the towel and
just settle for keeping the disease in check.
The eradication of guinea worm dis-
ease, led by the Carter Center in Atlanta,
has had fewer setbacks; it relies on a sim-
ple change in human behavior, a strategy
that has worked everywhere. Guinea worm
larvae are ingested via contaminated drink-
ing water, and they make a very painful
exit, usually from the lower leg, a year later.
Teaching people to filter their water and pre-
venting those with an exiting worm from
walking in sources of drinking water can
interrupt transmission.
1737
 NEWSFOCUS
Although the guinea worm campaign has
taken longer than anyone expected, today it
is getting tantalizingly close; so far in 2010,
there have been fewer than 1700 cases in
four countries, down from an estimated
3.5 million cases in 20 countries 2 decades
ago. More than 95% of those cases occur in
southern Sudan, however, a region expected
to vote for independence in a January ref-
erendum. If that leads to a new civil war, as
some expect, it could mean new delays, says
Donald Hopkins of the Carter Center, who
has led the campaign since its inception.
LYMPHATIC FILARIASIS
Agents: Wuchereria bancrofti, Brugia
malayi, and Brugia timori
Estimated number of people
infected: 120 million
Three mosquito-borne nematodes cause use
LF, often called elephantiasis because of some
patients’ horribly swollen limbs. A global alliance is
coordinating annual rounds of mass treatment using
drugs donated by GlaxoSmithKline and Merck. The
target: interrupt transmission globally by 2020.
Inflammatory word
One direct effect of these sobering experi-
ences is that some prefer to avoid the word
eradication in order to keep expectations
low. That’s the case for the global campaign
to wipe out lymphatic filariasis (LF), a
mosquito-borne parasitic infection that
afflicts an estimated 120 million people in
more than 80 countries.
LF is also known as elephantiasis for the
grossly swollen limbs and scrotal sacs that
it can cause. Five or six annual rounds of
mass drug administration often stop trans-
mission in affected regions, and that’s the
main strategy a global partnership is using
against the disease. But its stated goal
is “global elimination,” not eradication,
which some dismiss as a semantic differ-
ence (see sidebar). “You don’t want to use
a word that is so inflammatory if you want
your program to run smoothly,” says Eric
Ottesen, who heads the Lymphatic Filariasis
Support Center at the Task for Global
Health in Atlanta.
At the meeting, participants offered
many recommendations to deal with the
tougher climate. Some are already being
1738 24 DECEMBER 2010
applied in the attempt to
put measles on the path
Scientists’ New Eradication Target:
toward eradication, says A Word in Their Lexicon
CDC’s Cochi, who is
involved in those plans. Malaria has been eradicated from Europe. True or false?
Measles is an obvious True—malaria no longer occurs in Europe, except for “imported”
next candidate because cases, of course. Yet most scientists would argue that the statement
it kills hundreds of thou- is wrong because eradication, by definition, means that a disease is
sands of children, and gone from the entire planet. When talking about one or more coun-
there is a very effec- tries, or even a whole continent, you have to use the second E-word:
tive vaccine that works elimination. Smallpox has been eradicated. Malaria has been elimi-
even after a single dose. nated from Europe.
Stepped-up vaccina- At least until now. The semantic distinction, adopted at a 1997
tion efforts have led to a landmark meeting on disease eradication in Berlin, has often
plunge in global mortal- caused confusion, and participants at a recent follow-up meeting in
ity since 2000. But reim- Frankfurt—including some who were present 13 years ago—pro-
Downloaded from www.sciencemag.org on December 23, 2010
portations occur fre- posed doing away with it. Their proposal, already contested, would
quently in measles-free make the assertion at the top of this story perfectly fine.
countries, and this year The debate over what constitutes eradication has been going on
has seen an upsurge in for decades. In trying to draw up a clean set of definitions, the 1997
Africa. forum, one in a series of brainstorming sessions called the Dahlem
At WHO’s request, an Workshops, said that “elimination” was to become the word for any-
expert panel has studied thing less than the global scale, while “eradication” was reserved for
the feasibility of measles pathogens completely gone from the planet—except any remaining
eradication; as part of the stocks in lab freezers, as is the case with the smallpox virus. (Once
analysis, two indepen-
dent teams have done an
extensive economic analysis. Their reports, international agencies, nongovernmental
presented at a July meeting in Washington, organizations, and national governments—
D.C., concluded that eradication would cost should also have a solid business plan, the
an estimated $7 billion to $14 billion but Frankfurt meeting concluded. This “invest-
would be cost-effective. ment case,” as some call it, would outline
Such studies are controversial because the rationale for a disease campaign, along
they rely on mathematical models, which with the risks and obstacles, to policymakers,
can be a poor proxy for the real world. Wars funders, and drug- and vaccinemakers. The
or political problems can add billions to the report also argued for innovative financing
final tab. Still, they’re better than nothing, schemes. Because the benefits of eradication
says Cochi—and they force scientists to accrue over generations to come, government
take into account the new reality in which could issue bonds, for instance, just as they
control measures will continue. After small- do for major capital investments such as
pox was eradicated, vaccination programs roads or railways.
around the world were halted, which saved
billions of dollars. Today, security experts
shudder at the notion of entire populations
MEASLES
vulnerable to fast-spreading diseases like Agent: The measles virus
polio or measles, and the assumption is Estimated annual number of cases:
10 million
that many countries will keep vaccinating The Americas eliminated measles in
against them no matter what. 2002, and cases elsewhere have dropped ed
Both of the measles studies showed that precipitously as a result of vaccination campaigns—
although this year, there’s been a wave of new
eradication would pay off even if current outbreaks. An expert panel advising WHO has
vaccination schemes continue. Based on green-lighted eradication, but many argue it’s
these and other findings, the expert group better to finish polio first.
recommended in August that measles
CREDITS: MICROSCOPIC PHOTOS: CDC
“can and should be eradicated,” and the
World Health Assembly may adopt a resolu-
tion to go ahead at its next meeting, in May
in Geneva.
Eradication bonds
A cost-benefit analysis is one step, but advo-
cates of eradication—often an alliance of
VOL 330 SCIENCE www.sciencemag.org
Published by AAAS
 NEWSFOCUS

e·rad·i·ca·tion (i- răd i-ka-y’ shuhn) noun 1. Permanent reduction to zero

of the worldwide incidence of infection caused by a specific agent as a says. (Even worse, many say, is the ill-defined term “elimi-
result of deliberate efforts; intervention measures are no longer needed. nation as a public health problem.”)
That’s why Dowdle and Stephen Cochi of the U.S. Cen-
ters for Disease Control and Prevention want to drop the
e·rad·i·ca·tion (i- răd i-ka-y’ shuhn) noun 1. The absence of disease in a term “elimination.” Talk about national, regional, and
defined geographical area as a result of deliberate control efforts. global eradication from now on, they suggested in a dis-
cussion paper for the Frankfurt meeting.
The proposal sparked yet more debate. Some called it
those are destroyed as well, the third E-word applies: extinction. But this an inevitable correction of a historic mistake. “Let’s simply eradicate the word
has never happened.) elimination,” quipped Ciro de Quadros, the former head of the Pan American
The distinction created trouble from the start, recalls Walter Dowdle, a Health Organization. Others weren’t so sure. “Why upset a system of defini-
veteran of the polio-eradication effort. At a 1998 meeting in Atlanta, co- tions that has become firmly established in the medical literature?” they
sponsored by the World Health Organization (WHO), polio pioneer Frederick wondered. And how exactly do you define national or regional eradication?
Robbins said elimination was a vague, useless word. “What do you mean, we Those questions weren’t solved, but the proposal to eliminate the word

Downloaded from www.sciencemag.org on December 23, 2010

‘eliminated’ polio from the Americas? We eradicated it!” Dowdle recalls the “elimination” carried the day. In the process, participants agreed to also
Nobel laureate yelling. “Fred’s words are still ringing in my ears.” drop the part of the 1997 definition of eradication that said that “all con-
Although scientists have generally adopted the distinction, confusion trol efforts can be stopped,” as bioterror worries have put an end to that
has persisted. Journalists use the terms interchangeably. WHO often pref- promise (see main text, p. 1736).
aces “eradication” with “global” or “worldwide,” which is clear but techni- Not that this is the final answer. In a paper on measles elimination
cally redundant. At meetings of Rotary International, whose members have published in the the 3 December issue of Weekly Epidemiological Record,
donated more than $900 million for polio eradication, “I never use the for instance, WHO acknowledged the Frankfurt proposal in a footnote but
word elimination,” says Robert Scott, chair of the group’s PolioPlus Com- firmly stuck to existing definitions of eradication and elimination. The con-
mittee. People don’t understand it, and it doesn’t inspire enthusiasm, Scott fusion is unlikely to be eliminated—or eradicated—anytime soon. –M.E.

Eradication programs should not hurt
RIVER BLINDNESS existing health services by siphoning away
Agent: Onchocerca volvulus money and effort from basic health ser-
Estimated number of people vices for an increasingly rare disease, the
infected: 18 million Frankfurt report says—and to the extent
A problem primarily in Africa, river that they can, they should have a broader
blindness, or onchocerciasis, is caused
sed
by roundworms and transmitted by black ck fli
flies. benef icial effect. Current eradication
An international control program relying on mass efforts have tried to do this. Polio vaccina-
treatment has been more successful than expected,
tion has sometimes been combined with
and some suggest going full throttle for eradication
dispensing vitamin A tablets or distributing
later this decade.
bed nets against malaria. But everyone at
the Frankfurt meeting agreed that it’s not
the job of an eradication campaign to fix a
broken health care system. And ultimately,
Dowdle points out, eradication campaigns
need to be focused if they are to have any
chance of success.
on the success—or failure—of the current
campaigns. WHO is reticent to embark on
a measles-eradication campaign as long as
polio isn’t finished, says Cochi; it worries
that two simultaneous campaigns would be
too much.
MALARIA
Agent: Plasmodium spp.
Estimated number of cases in 2009:
225 million
A flopped campaign in the 1950s made ade
malaria eradication a dirty word. Now, Bill
and Melinda Gates have revived the buzz.
And although most scientists say global
eradication is decades away, countries on the
fringes of the malaria map are busy ridding
themselves of the parasites.

Nor can future eradication campaigns
afford to bypass poor countries’ broader
health concerns, like diarrhea or respiratory
disease, which kill far more children, the
group concluded. The relentless focus on one
CREDITS: MICROSCOPIC PHOTOS: CDC (2)
More candidates
There are other candidates for eradica-
tion, despite the more demanding environ-
ment. One is the rubella virus, which causes
severe malformations in newborn babies.
The total burden is considered too low to
warrant a standalone eradication campaign,

disease has fueled resistance to the polio cam-
paign, for instance, in Nigeria and India. Gov-
ernments of developing countries are rightly
wondering whether their sacrifices for a global
public health goal make sense, says Stewart
Tyson, a consultant at Liverpool Associates in
Tropical Health in the United Kingdom—and
they should give any new eradication plan a
“good grilling” before signing on, he says.
but rubella could piggyback on measles
eradication since the vaccines are com-
bined easily, advocates say. Some also see
chances for onchocerciasis, also known as
river blindness, and perhaps schistosomia-
sis. Yaws, which has already been elimi-
nated from India, could once again become
a candidate.
But these ideas’ fortunes depend in part
Such ambivalence is one reason why
many eradication enthusiasts say giving up
on polio is not an option. “I’m very worried
about polio,” says Hopkins. “It must suc-
ceed. If it didn’t, it would be a big setback
for the whole concept of eradication.”
–MARTIN ENSERINK

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1739

Published by AAAS
 NEWSFOCUS
PA L E O N T O L O G Y
Altering the Past: China’s
Faked Fossils Problem
A booming fossils market has resulted in a flood of “improved,” reconfigured, and
composite specimens; many are finding their way into China’s museums
BEIJING—Frozen in time, the 5-meter-long
ichthyosaur embedded in dark limestone
seems to be darting after prey in a turbid Tri-
assic sea. But look more closely at the star-
tlingly lifelike skeleton here in the Geologi-
cal Museum of China, and you will see that
something isn’t quite right. The beast’s lower
jaw and shoulder girdle are visible, which
requires a ventral view—but the lower body
is a lateral-dorsal view. Such an odd juxtapo-
sition can mean only one thing, says Li Chun,
a marine reptile expert here at the Institute of
Vertebrate Paleontology and Paleoanthropol-
ogy (IVPP): The centerpiece of the museum’s
prehistoric life exhibit is a composite of two
individuals, and possibly more.
Specialists and collectors around the
world have long decried the flood of sham
fossils pouring out of China. But Science has
learned that many composites and fakes are
now finding their way into Chinese museums,
especially local museums. “The fake fos-
sil problem has become very, very serious,”
says Peking University paleontologist Jiang
Da-yong. Li estimates that more than 80% of
marine reptile specimens now on display in
Chinese museums have been “altered or artifi-
cially combined to varying degrees.”
Geological Museum officials are not
inclined to remove the fishy ichthyosaur,
1740 24 DECEMBER 2010
from southwestern China. “Farmers prepared
the specimen. They might have made some
mistakes when they put it together, but it is
not a fake. You can call it a kind of model,”
says Lu Liwu, director of prehistoric life
research at the museum.
Nevertheless, Li is concerned about what
he deems a misleading display. “This is a
national museum,” he says. Another Geolog-
ical Museum exhibit, a pair of huge dinosaur
eggs embedded in siltstone, is also a sham:
Large chunks of the shell are not the orig-
inal material, says Li. Lu says the museum
intends to add signage to clear up any mis-
conceptions. Outside of Beijing, curators are
not so conscientious. Chinese and Western
paleontologists concur that many provincial
museums are chock-full of composites, chi-
meras, and other phony fossils. But several
contacted by Science said they are reluctant
to speak out. “We would seriously jeopar-
dize our own opportunities to work with our
Chinese colleagues on very important mate-
rial,” says one Western paleontologist, who
requested anonymity.
One consequence of the fakery is an ero-
sion of trust in museums, which are supposed
to enlighten—not con—the public. Schol-
ars, too, pay a price: They waste time sifting
authentic specimens from counterfeit chaff.
VOL 330 SCIENCE www.sciencemag.org
Published by AAAS
Crack me up. Archaeoraptor turned out to be a
bird-dinosaur chimera, not a missing link.
And a genuine blockbuster fossil can be
destroyed by attempts to enhance its appeal.
“A fake part in a fossil ruins the value of the
entire specimen,” says Ryosuke Motani, a
paleontologist at the University of California,
Davis. “Even though the genuine part of the
same specimen may provide important infor-
mation that is otherwise unknown,” he says,
“skepticism emerges as to whether we can
trust it or not.”
“Normally we know right away if a fossil
is fake, although it can take some time to be
sure,” says IVPP Director Zhou Zhonghe. But
Downloaded from www.sciencemag.org on December 23, 2010
fraudulent specimens can end up in the peer-
reviewed literature. For example, the holo-
type—for which a species is named—of Typi-
cusichthyosaurus tsaihuae, a marine reptile
from southwestern China, is “a forged speci-
men” with carved features, says Motani. (Li
and other paleontologists agree with that anal-
ysis; the team that described the species could
not be reached for comment.) More contro-
versially, an IVPP paleontologist asserts that a
2009 report in the Proceedings of the National
Academy of Sciences (PNAS) describing a
new species of early cheetah is based on a
forged skull; he has demanded a retraction.
The authors insist the skull is authentic and
stand by the report, as has the journal.
Stamping out sham fossils will require a
crackdown on how fossils are collected and
sold in China. A new law that comes into
force next month aims to protect fossils of
high scientific value (Science, 17 September,
p. 1453). But experts doubt that the law will
pose a sufficient deterrent: Forging fossils is
simply too lucrative, they say.
CREDITS (TOP TO BOTTOM): COURTESY OF ZHOU ZHONGHE; R. STONE/SCIENCE
High fidelity. Zhao Lijun’s exhibition in Zhejiang
has won praise for its authenticity.

The art of faking fossils has a long his-
tory. Perhaps the most infamous fraud is Pilt-
down Man, a skull, unveiled in 1912, that was
touted as a missing link between humans and
apes. It was exposed as a hoax in 1923, when
a German anatomist determined that Piltdown
was a chimera: a modern human skull and an
orangutan jawbone. In another notorious case
in the annals of bogus fossils, noted paleontol-
ogist Friedrich von Huene described in 1966
a juvenile Leptopterygius from Germany. Von
Huene, 91 at the time, had not realized that
the ichthyosaur was a total fabrication: Its
“bones” were carved from the substrate.
China, too, has suffered a Piltdown
moment. “Archaeoraptor,” purported to be
a missing link between birds and dinosaurs,
made its debut in the November 1999 issue of
National Geographic. Rumors that the skele-
ton, unearthed in northeastern China’s Liaon-
ing Province, was a fake began to swirl even
before publication, says Zhou. Archaeoraptor
was later thoroughly discredited as a chimera
consisting of the body of Yanornis martini—
a primitive fish-eating bird—and the tail of
Microraptor zhaoianus, a feathered dinosaur.
Now, an IVPP paleontologist fears that a
prestigious peer-reviewed journal has pub-
lished a fake. In a 13 January 2009 report
in PNAS, Per Christiansen of the Zoological
Museum in Copenhagen and Ji H. Mazák
of the Shanghai Science and Technology
Museum presented a nearly complete skull
of a primitive cheetah, which they have said
was unearthed from a fossil layer in Gansu’s
Linxia Basin dated to around 2.2 million
to 2.5 million years ago. Christiansen and
Mazák describe a “unique combination of
primitive and derived traits” that places the
species, which they named Acinonyx kurteni,
as “the most primitive cheetah known to date.”
When IVPP’s Deng Tao saw the PNAS
paper, he says, “I knew immediately the skull
was a fake.” Deng says the published pho-
tos show that several features of the skull
had been concocted from bone or plaster.
For example, he wrote in a 16 January 2009
letter to PNAS, “the parietal area is glued by
some bone pieces to imitate the skull of a
modern cheetah, but the forger did not make
the parietal crests.” With that one slip-up,
Deng says, the forger “gave the game away.”
According to Deng, who has collected fos-
sils in Linxia every year since 1998, it is
common there to encounter dealers peddling
CREDIT: R. STONE/SCIENCE
fake skulls. “Unqualified collectors are often
cheated,” he wrote to PNAS. Because the
paper’s “unfounded” conclusions are “based
on a fossil forgery,” Deng urged the authors
or the journal to retract the paper. IVPP’s Qiu
Zhanxiang, an academician and top specialist
www.sciencemag.org SCIENCE VOL 330
on mammalian fossils, says he concurs with
Deng’s opinion that the skull is a composite
and that the paper should be retracted.
Mazák, whose birth name is Huang Ji,
told Science that the skull is genuine and that
Deng’s concerns amount to a “scientific dis-
pute” because the PNAS paper did not cite
Deng’s 2004 description of a primitive chee-
tah from Linxia, Sivapanthera linxiaensis.
Mazák declined to explain how he obtained
the skull, and Christiansen, now at the Zoo-
logical Garden in Ålborg, Denmark, did not
respond to requests for comment. In a 4 Feb-
Pièce de contrefaçon.
IVPP’s Li Chun with a composite
ichthyosaur at the Geological
Museum of China.
ruary 2009 letter to Deng, PNAS declined to
publish Deng’s letter and stated that his obser-
vations “can be explained by sloppy prepara-
tion, incomplete preservation of the skull, or
as characters that differ from Deng’s expecta-
tions that are based on an a priori hypothesis
of relationship or ancestry.” Deng says he has
not pursued the matter further because Mazák
has declined to give him access to the skull.
The growing problem of faked specimens
stems from China’s fossil economy. Most fos-
sils, including prized specimens, are unearthed
by farmers, who often gussy up specimens to
make them look more complete or unusual
and thus fetch a higher price. Some dealers
are fooled, and some also engage in such chi-
canery, says Zhou. Few buyers are discerning.
“For officials and businessmen, beautiful fos-
sils can upgrade their reputation. For some
researchers, strange fossils mean they may
have a chance to publish in a top journal and
get more funding and a higher position,” says
Jiang. “In this hurried and blundering situa-
tion, anything may happen.”
Exacerbating the problem is a recent boom
in museum building across China that has
24 DECEMBER 2010
Published by AAAS
NEWSFOCUS
resulted in sizable acquisition budgets and
competition for prize specimens. In October,
the Shanghai History Museum invited Zhao
Lijun, curator of paleontology at the Zheji-
ang Museum of Natural History (ZMNH)
in Hangzhou, to examine fossils it intended
to purchase for an exhibition hall to open in
2012. Zhao identified a dozen specimens,
including a 15-meter-long ichthyosaur, that
were “totally fake,” she says. “When I told
them the truth, they were astonished.” To its
credit, Zhao says, the Shanghai museum can-
celed the deal. Many other museums, how-
Downloaded from www.sciencemag.org on December 23, 2010
ever, snap up fossils with inept or derisory
expert advice. As a result, Li says, many
“local museums are full of fakes.”
One remedy is for museums to create
closer ties with academics; few now have
paleontologists on staff. A rare success is “Sea
Monsters,” a yearlong exhibition of marine
reptile fossils that wrapped up last month at
ZMNH. Zhao joined IVPP’s Li and others in
the field for a few summers to collect fossils
for the exhibit; other specimens were on loan
from IVPP. “Without IVPP’s cooperation,
we would not have been able to do this,” says
ZMNH Director Kang Xi Min.
Another boost would be a training pro-
gram for fossil preparators. And preventing
fake fossils from contaminating the scientific
literature, says IVPP’s Xu Xing, “can be eas-
ily avoided by careful and experienced sci-
entists.” But Li and others admit they don’t
have a strategy for combating the root of their
ills: a legion of fakers assiduously despoiling
China’s paleontological riches. “Our fossils
are some of the best in the world,” says Li.
“But they are being destroyed, and there is
little we can do about it.” –RICHARD STONE
1741
 NEWSFOCUS
R E G E N E R AT I V E M E D I C I N E
CIRM: The Good, the Bad,
And the Ugly
Having bolstered basic research, California’s stem cell agency must choose a new
leader and figure out how to develop therapies
December has been an eventful month for the
California Institute for Regenerative Medi-
cine (CIRM), the agency created by Cali-
fornia voters to disburse $3 billion for stem
cell research. Real estate developer Robert
Klein, who spearheaded the 2004 ballot initia-
tive that created CIRM and served as its first
chair, had promised to step down when his
term expired on 17 December. On 15 Decem-
ber, CIRM’s governing board was sched-
uled to elect Klein’s successor. Sounds pretty
straightforward, right?
Not so. Klein initially backed Alan
Bernstein, a Canadian scientist and executive
director of the Global HIV Vaccine Enter-
prise, to succeed him. But Bernstein’s can-
didacy ran aground because of concerns that
state law precludes non-U.S. citizens from
holding the job. (It’s not clear that it actually
does.) The governor nominated Klein for a
second 6-year term. Klein accepted but said
he would serve only up to 6 months more
to help find a replacement. The state con-
troller nominated a different candidate, who
accepted the nomination and then withdrew
8 days later. Eleven prominent California sci-
entists threw their support behind Klein, who
is both widely admired for getting CIRM off
the ground and criticized for micromanag-
ing the agency. But the state controller wrote
a letter that criticized the selection process
as “fundamentally flawed” and urged the
board to postpone the election and start over.
Instead, they reelected Klein.
“It’s the usual CIRM circus,” says Marie
1742 24 DECEMBER 2010
Csete, the institute’s former chief scien-
tific officer, who now oversees research and
development at Organovo, a San Diego bio-
tech company.
Controversy is nothing new for CIRM.
Csete resigned abruptly last year after only
15 months on the job, for reasons she has
declined to explain (Science, 17 July 2009,
p. 249). In 2007, founding president Zach
Hall resigned after a contentious meeting
that exposed a rift between board members
representing research institutions and those
who are patient advocates (Science, 27 April
2007, p. 526). Watchdog groups have blasted
the institute about what they see as exorbi-
tant staff salaries and conflicts of interest on
the board. And patient advocates are tired of
waiting for the stem cell cures they feel they
were promised during the campaign.
Despite all this, many scientists insist
that the institute has been a tremendous suc-
cess. It has so far awarded 385 grants total-
ing more than $1.1 billion, money that has
been used to build new labs, train scientists,
and conduct research throughout the state. “I
see CIRM as a major advance for the entire
world of stem cell research,” says Elaine
Fuchs of Rockefeller University in New
York City. “Its effect has spread way beyond
the state of California.”
Still, even CIRM supporters say the insti-
tute has to improve its relationships with
industry if it hopes to fulfill its mandate:
generating stem cell therapies that help peo-
ple suffering from conditions like diabetes
VOL 330 SCIENCE www.sciencemag.org
Published by AAAS
and Parkinson’s disease. A recent report by
a panel of external scientists convened by
the CIRM board said translating basic sci-
ence into therapies should be a major prior-
ity going forward. “We’ve fallen short on the
clinical translation side,” says Jeff Sheehy, a
CIRM board member and HIV/AIDS activ-
ist. “I think we can do better.”
A solid foundation
A major impetus for CIRM was the restric-
tions the Bush Administration imposed in
2001 on federal funding for research using
human embryonic stem cells (hESCs)
(Science, 17 August 2001, p. 1242). Proposi-
tion 71 authorized California to raise $3 billion
Downloaded from www.sciencemag.org on December 23, 2010
for stem cell research through the sale of state
bonds, offering an alternative source of fund-
ing for the state’s scientists. “The ‘yes’ vote
on Prop. 71 changed the world,” says Sean
Morrison, a stem cell researcher at the Uni-
versity of Michigan, Ann Arbor. “Prior to that,
the conversation in most states was, ‘Should
we allow embryonic stem cell research?’ ”
Morrison says. “But once California put that
stake in the ground, the conversation shifted
to, ‘How do we keep up with California?’ ”
Even with the Obama Administration’s
support for hESC research, some research-
ers see CIRM as a bulwark against political
and economic turbulence. In August, a fed-
eral judge in Washington, D.C., issued a tem-
porary injunction blocking hESC research
on the grounds that it violates a law banning
federal funding for research that destroys
embryos (Science, 3 September, p. 1132).
That injunction has been put on hold, but the
legal battles continue. “If the injunction goes
back into place, that would be a huge blow
to the field,” says Morrison. “People in Cali-
fornia may have a safe harbor, and that’s a
big deal.” (CIRM’s portfolio now includes
research on adult and induced pluripotent
stem cells.) The agency also provides a buf-
fer against increased competition for funds at
the National Institutes of Health (NIH), says
George Daley, a stem cell researcher at Har-
vard Medical School in Boston.
Not surprisingly, CIRM grantees are not
complaining. “I just moved into a spectacu-
lar new building,” says Arnold Kriegstein, the
director of the new Eli and Edythe Broad Cen-
ter of Regeneration Medicine and Stem Cell
Research at the University of California, San
Francisco (UCSF). CIRM kicked in $35 mil-
lion for the new facility, about a third of its
cost. The Broad Center at UCSF is funded by
one of the 12 major facilities grants awarded
CREDIT: CIRM
by CIRM. Seven of these were completed in
2010, and all but one are expected to be up and
running by the end of 2011.
 NEWSFOCUS

TOP 12 CIRM GRANTEES

Institution Total awarded Awards Representation on CIRM board Position
Stanford University $ 175,862,473 50 Philip A. Pizzo Dean of medical school
University of California, Los Angeles $ 135,154,660 40 Eugene Washington Dean of medical school
University of California, San Francisco $ 110,532,518 35 Sam Hawgood Dean of medical school
University of California, San Diego $ 77,177,593 32 David Brenner Dean of medical school
University of Southern California $ 71,933,514 19 Carmen A. Puliafito Dean of medical school
University of California, Irvine $ 71,878,458 27 Susan V. Bryant Vice chancellor for research
University of California, Davis $ 61,187,635 19 Claire Pomeroy Dean of medical school
Sanford Consortium $ 43,000,000 1 John C. Reed Board member (CEO, Burnham Research Institute)
for Regenerative Medicine William R. Brody Board member (President of Salk Institute)
City of Hope National Medical Center $ 41,586,199 9 Michael A. Friedman CEO
Scripps Research Institute $ 37,377,357 14 Floyd E. Bloom Professor emeritus
The Salk Institute for Biological Studies $ 36,818,181 14 William R. Brody President

Downloaded from www.sciencemag.org on December 23, 2010

University of California, Berkeley $ 36,746,646 12 Robert Birgeneau Chancellor

The recent external review gave CIRM includes many people in high positions at the point. “Companies are trying to stabilize
high marks for building research infrastruc- the institutions that receive the most fund- a technology and commercialize it rather than
ture and fostering high-quality research. ing from CIRM (see table). A June 2009 push the bleeding edge,” he says.
The agency has so far committed $108 mil- report from the Little Hoover Commis- Even companies that have succeeded
lion to training more than 600 young scien- sion, an independent state oversight com- say it hasn’t been easy. Earlier this year, San
tists. CIRM says it has contributed funding mittee, found that 80% of funding had gone Francisco–based iPierian won a $6 million
to research published in more than 600 jour- to institutions with representatives on the early translation award and a $1.5 million
nal articles, with roughly a quarter of those in board. The commission recommended that a basic biology award. “We put a ton of effort
high-profile journals. “Progress during this smaller board with more independent voices into understanding what was being asked
first stage of CIRM’s development has been would have more credibility. for,” says CEO Michael Venuti.
remarkable,” the panel wrote Some companies have been discouraged
in its report. (The fact that the Looking forward from applying for CIRM funds by terms
panel was chaired by Bernstein, According to the recent exter- that require paying CIRM back with equity
who became a candidate to chair nal review, CIRM is now enter- in the company or with cash equivalent to
CIRM’s board, caused some to ing a second stage in which it several times the original loan if a project
question its objectivity. “Was should maintain its strength in bears commercially viable fruit, says Hans
he really going to criticize the basic research and extend its Keirstead, a stem cell researcher at UC Irvine.
agency before he came in?” asks reach toward clinical applica- Such arrangements are reasonable in princi-
Sheehy.) tions. To do that, CIRM will ple, but the terms CIRM imposes can be oner-
Even some of CIRM’s fre- have to improve its ties with ous, Keirstead says: “CIRM has to pay a lot of
quent critics give the agency biotechnology companies, says attention now to becoming industry-friendly.”
credit. “If you look realisti- Daley, who served on the panel: CIRM President Alan Trounson says he is
cally at the scientific work that’s Same as the old boss. “We felt that engagement with sensitive to these concerns but doesn’t think
been done, there’s been very Robert Klein says his sec- industry had been underempha- the review process is problematic: “Compa-
important progress,” says John ond term will last no more sized and needs to be encour- nies that have put in well-formed proposals
CREDITS (TOP TO BOTTOM): CIRM; NOAH BERGER/BLOOMBERG VIA GETTY IMAGES

Simpson, stem cell project direc- than 6 more months. aged.” So far, slightly more have done very well.” But he acknowledges
tor for Consumer Watchdog in than 7% of CIRM money has that CIRM has had difficulty attracting pro-
Santa Monica, California. “They did manage gone to companies. posals, particularly from larger companies.
to get some world-class laboratories built, and Several California biotech leaders say He’d like to set up an industry advisory board
they did it in a clever way where they used the they have been frustrated by their interactions to help improve industry relations.
public money as seed money to attract match- with CIRM. “In the past, there’s been a lack In the coming months, he and others will
ing contributions,” Simpson says. of recognition that it takes a company to actu- be waiting anxiously to see who succeeds
At the same time, Simpson criticizes what ally take a treatment forward from the bench Klein as chair. Patient advocates want an
he sees as “outrageously high” salaries for top into the clinic,” says Chris Airriess, chief advocate at the helm. Scientists would prefer
CIRM staff members. The chair and presi- operating officer of California Stem Cell Inc. a scientist. Trounson, who may have to work
dent, for example, can be paid up to $529,000. in Irvine. Airriess says his company has twice most closely with the new boss, says he’s hop-
(Klein has declined a salary for most of his applied, unsuccessfully, for CIRM money. He ing for someone with expertise in the delivery
tenure.) That’s more than double the $199,700 and others place much of the blame on the stage of therapeutic development. “The basic
paid to the director of NIH, Simpson notes— review process, which he says is structured science, as long as we look after it, will take
or the $225,000 paid to the state’s governor: too much like the NIH review process for aca- care of itself,” he says. The real challenge for
“That kind of largesse has often come back to demic research grants. CIRM reviewers criti- CIRM, he says, is getting the science into the
embarrass the agency, and rightly.” cized his company’s applications for the lack clinic. “We need more help on how to make it
Critics have also noted that the board of new science, but Airriess says that misses all happen.” –GREG MILLER

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1743

Published by AAAS
 COMMENTARY
Interactive introduction Interacting with the
to the brain public for science policy

1748
LETTERS I BOOKS I POLICY FORUM I EDUCATION FORUM I PERSPECTIVES
1749

LETTERS
edited by Jennifer Sills

Genetic Future for Florida Panthers 0.20, 0.10, 0.06, 0.04, and 0.01, for a total of 41%. Ordinarily, 50%
percent of the ancestry is from each sex; a contribution of 41% is

Downloaded from www.sciencemag.org on December 23, 2010

W. E. JOHNSON ET AL. (“GENETIC RESTORATION OF THE FLORIDA
panther,” Reports, 24 September, p. 1641) document genetic
changes in the Florida panther population after the 1995 introduc-
tion of eight Texas puma females. This translocation has been a great
success; the population size has increased more than threefold, and
several detrimental traits have
substantially decreased in fre-
quency. However, there are com-
pelling reasons to continue the
close genetic management and
monitoring of the population in
the future.
First, only five of the eight
female Texas pumas had off-
spring. The distribution of off-
spring from these five females
was unequal—one female con-
tributed nearly half of the off-
spring—and the total ancestry
from these five females was very
high. Specifically, the authors
stated that “[t]he estimated rel-
ative genetic contribution[s]
of the [Texas] females to the
descendant population” are
equivalent to saying that about 80% of the female ancestry is from
the five Texas females, nearly the maximum possible. In other
words, the Texas females may have been too successful and man-
agement should evaluate whether to actively preserve the original
Florida panther ancestry.
Second, the success may be threatened by inbreeding and low
effective population size in the current and future generations. For
example, a male offspring of a Texas female and a Florida panther
male mated with three of his daughters and produced seven off-
spring with inbreeding coefficients of 0.25. The effective population
size estimate was based only on the number of breeding males and
breeding females. If the variance in contributions in males is equal
to that found in Yellowstone pumas (1), which resulted in the effec-
tive number of males being only 18.5% of the observed number of
males, and the variance in females reflects the contributions above,
the overall effective size in 2007 is probably only between 10 and 15
animals, rather than the 32.1 estimated.
Overall, swamping of the Florida panther ancestry, inbreeding,
and low effective population size may endanger the gains made from
translocation for genetic restoration (2). PHIL HEDRICK
School of Life Sciences, Arizona State University, Tempe, AZ 85287–4501, USA. E-mail:
philip.hedrick@asu.edu
References
1. M. Culver et al., Anim. Conserv. 11, 1045 (2008).
2. P. W. Hedrick, R. Fredrickson, Conserv. Genet. 11, 615 (2010).

CREDIT: U.S. NATIONAL PARK SERVICE, EVERGLADES NATIONAL PARK/WIKIMEDIA COMMONS

Response
WE AGREE WITH HEDRICK THAT THERE ARE
cogent reasons for continuing to monitor
the surviving Florida panthers in the future.
Inbreeding is by no means solved and may
increase as available habitat is developed. It
is true that the relative genetic contribution of
the Texas pumas was restricted to five of the
eight females released in 1995 and that they
account for about 50% of the genetic heritage
in Florida panthers today. Whether this repre-
sents “swamping” or natural subspecies reas-
sortment in the aftermath of demographic
and genetic perils experienced by canonical
Florida panthers is a matter of opinion. What
is clear is that suitable habitat must be pre-
served and additional populations must be
established for the continued survival of this
critically endangered group.
The lessons learned from the genetic
restoration project highlight the many ben-
efits to the Florida panther population while
also demonstrating that there is no quick or
universally accepted solution to conserving
small, endangered populations. Incorporat-
ing interdisciplinary data and the expertise
of scientists from varied backgrounds can
only improve the development of effective
management regimes to help ensure the
recovery of endangered animals, including
the Florida panther.
WARREN E. JOHNSON,1* DAVID P. ONORATO,2
MELODY E. ROELKE,3 E. DARRELL LAND,2
STEPHEN J. O’BRIEN1
1
Laboratory of Genomic Diversity, National Cancer Insti-
tute, Frederick, MD 21702, USA. 2Florida Fish and Wildlife
Conservation Commission, Naples, FL 34114, USA. 3SAIC-
Frederick, Laboratory of Genomic Diversity, National Cancer
Institute, Frederick, MD 21702, USA.
*To whom correspondence should be addressed. E-mail:
warjohns@mail.nih.gov
Biodiversity Transcends
Services
IN THEIR POLICY FORUM “ECOSYSTEM SERVICES
for 2020” (15 October, p. 323), C. Perrings et
al. discuss possible missing elements in the
Convention on Biological Diversity’s pro-
posed new targets. They suggest that targets
for biodiversity be based directly on ecosys-

1744 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

Published by AAAS

Enhanced
devices
1759
tem services because people will then have a
stake in the program’s success.
This approach undersells both biodiver-
sity and the role of ecosystem services. Bio-
diversity’s value extends beyond current eco-
system services and includes likely future
benefits we cannot anticipate. Recognizing
the benefits of ecosystem services can reduce
the cost of retaining relatively intact areas
of local biodiversity, but we need to plan for
larger-scale conservation. A recognized eco-
system service does more than support some
local elements of biodiversity; it makes a low-
cost contribution toward conserving the bio-
diversity of the larger region. Regionally, eco-
system services may be more important as
indicators of relative cost and intactness than
of biodiversity. When considering regional
trade-offs, we cannot simply target ecosys-
tem services and ignore the elements of bio-
diversity that are not required for the service.
Adopting the ecosystem services option for a
specific locality may not be as good for bal-
anced regional biodiversity conservation as
adopting full conversion of that locality (1).
An example that has been used to illustrate
this point is a locality offering either com-
plete conversion to forestry logging or “sym-
pathetic” logging with partial biodiversity
retention. Adopting the ecosystem service
based on sympathetic logging, while lowering
opportunity costs and maintaining some bio-
diversity in that locality, nevertheless would
mean greater regional biodiversity loss for a
given level of regional forestry production.
As an alternative to targets focused on
current perceptions of important services, it
is time to consider higher-level targets and
goals in an effort to better balance overall
biodiversity conservation, ecosystem ser-
vices, and other needs of society. I propose
Letters to the Editor
Letters (~300 words) discuss material published
in Science in the previous 3 months or issues of
general interest. They can be submitted through
the Web (www.submit2science.org) or by regular
mail (1200 New York Ave., NW, Washington, DC
20005, USA). Letters are not acknowledged upon
receipt, nor are authors generally consulted before
publication. Whether published in full or in part,
letters are subject to editing for clarity and space.
www.sciencemag.org SCIENCE VOL 330
SPORE Prize
Essay
1764
that we implement new systematic conser-
vation planning to more efficiently serve
these different needs (2, 3). Because greater
efficiency can mean more biodiversity pro-
tection for a given rate of land conversion,
higher-level targets could allow us to focus
on reducing the rate of biodiversity loss as
opposed to the more narrow goal of main-
taining ecosystem services.
DANIEL P. FAITH
The Australian Museum, Sydney, NSW 2010, Australia.
E-mail: danfaith8@yahoo.com.au
References
1. D. P. Faith, Biodiversity and Regional Sustainability Analy-
sis, (CSIRO, Canberra, 1995); http://australianmuseum.
net.au/document/Biodiversity-and-regional-sustainability-
analysis/.
2. D. P. Faith, Glob. Environ. Change Soc. Pol. Dimensions
15, 5 (2005).
3. F. Grant, J. Young, P. Bridgewater, A. D. Watt, Eds.,
“Targets for biodiversity beyond 2010: Research support-
ing policy” (Report of e-conference, 2009), p. 44;
www.epbrs.org/PDF/Final%20long%20report.pdf.
Response
FAITH ARGUES THAT OUR APPROACH TO BIO-
diversity conservation, which focuses on
people’s interest in the benefits of ecosystem
services, may deprive us of future, unantici-
pated benefits. This claim is misplaced. In
our Policy Forum, we argue that conservation
goals should reflect the benefits we get from
biodiversity. The argument is not conditional
on the type or timing of benefits delivered. We
agree that it is not just biodiversity’s value in
producing marketed commodities that mat-
ters. Its indirect value in supporting ecosystem
services is often more important (1), and its
potential value to future users (option value)
and to future science (quasi-option value) has
been recognized as the most important of all
for at least three decades (2–4).
Our goal was to clarify the trade-offs
between these benefits, which are inevita-
ble as we strive to meet the basic needs of a
growing world population, alleviate poverty,
and protect those species on which our future
well-being depends (5). Only by being clear
about the benefits put at risk by the loss of
biodiversity now and in the future can we
approach these trade-offs wisely. The ecosys-
tem services approach helps clarify the bene-
fits at risk, whether they are direct, indirect, or
options. Science-based information on what
24 DECEMBER 2010
Published by AAAS
we gain and lose from biodiversity change
can inform decisions by those charged with
representing particular constituencies in
forums such as the Convention on Biological
Diversity (CBD).
Now that the CBD’s 2020 Targets have
been set (6), the problem has shifted from
goal setting to implementation. The best
way, now, to prioritize and assess targets is
to select appropriate indicators. For example,
given that achievement of the goal for sus-
tainable agriculture, target 7, is conditional
on achievement of the goal for agricultural
subsidies, target 3, the indicators for target 7
should include measures of the achievement
of target 3.
To implement the 2020 targets success-
Downloaded from www.sciencemag.org on December 23, 2010
fully, decision-makers need to be convinced
that the costs of biodiversity loss are real.
The ecosystem services approach provides
the evidence base to argue this case. Trum-
peting “intrinsic value” has had little effect
in the past and is likely to have less effect in
the future as other environmental concerns
escalate in policy significance. Using the
resources of science to identify and value the
consequences of biodiversity change is likely
to be the most effective strategy.
C. PERRINGS,1* S. NAEEM,2 F. AHRESTANI,2
D. E. BUNKER,3 P. BURKILL,4 G. CANZIANI,5
T. ELMQVIST,6 R. FERRATI,5 J. FUHRMAN,7
F. JASIC,8 Z. KAWABATA,9 A. KINZIG,1 G. M. MACE,10
F. MILANO,5 H. MOONEY,11 A.-H. PRIEUR RICHARD,12
J. TSCHIRHART,13 W. WEISSER14
1
School of Life Sciences, Arizona State University, Tempe,
AZ 85287, USA. 2Department of Ecology, Evolution, and
Environmental Biology, Columbia University, New York, NY
10027, USA. 3Department of Biological Sciences, New Jer-
sey Institute of Technology, Newark, NJ 07102, USA. 4Sir Ali-
ster Hardy Foundation for Ocean Science, Plymouth PL1 2PB,
UK. 5Instituto Multidisciplinario sobre, Ecosistemas y Desar-
rollo Sustentable, Universidad Nacional del Centro, Argen-
tina. 6The Resilience Centre, Stockholm University, SE-106
91 Stockholm, Sweden. 7Department of Biological Sciences,
University of Southern California, Los Angeles, CA 90089,
USA. 8Centro de Estudios Avanzados en Ecologia y Biodivers-
idad, Pontificia Universidad Católica de Chile, Chile. 9Center
for Ecological Research, Kyoto University, Japan. 10Centre for
Population Biology, Imperial College London, Ascot SL5 7PY,
UK. 11Department of Biology, Stanford University, Stanford,
CA 94305, USA. 12DIVERSITAS, Muséum National d’Histoire
Naturelle, 75231 Paris Cedex 05, France. 13Department of
Economics and Finance, University of Wyoming, Laramie,
WY 82071, USA. 14Institut für Ökologie, Friedrich-Schiller-
Universität, Jena 07743, Germany.
*To whom correspondence should be addressed. E-mail:
charles.perrings@asu.edu
References
1. Millennium Ecosystem Assessment, Ecosystems and
Human Well-Being: General Synthesis (Island Press,
Washington, DC, 2005).
2. K. J. Arrow, A. Fisher, Q. J. Econ. 88, 312 (1974).
3. J. M. Conrad, Q. J. Econ. 94, 813 (1980).
4. A. C. Fisher, W. M. Hanemann, Nat. Res. Model. 1, 111
(1986).
5. J. D. Sachs et al., Science 325, 1502 (2009).
6. Convention on Biological Diversity, COP 10, Nagoya
(2010); www.cbd.int/nagoya/outcomes/.
1745
 LETTERS

CORRECTIONS AND CLARIFICATIONS

News Focus: “What shall we do with the x-ray laser?”
by A. Cho (10 December, p. 1470). The story mistakenly
states that Linda Young, an atomic physicist at Argonne
National Laboratory in Illinois, and colleagues shined
x-rays from the Linac Coherent Light Source onto xenon
gas. The physicists used neon.

News Focus: “Genes link epigenetics and cancer” by

J. Kaiser (29 October, p. 577). The article failed to note
that a Canadian team is among the researchers who have
found cancer genes involved in modifying chromatin
[K. C. Wiegand et al., N. Engl. J. Med. 363, 1532 (2010)].

Research Articles: “Nonlocal transport in the quantum

spin Hall state” by A. Roth et al. (17 July 2009, p. 294).
An unintended duplication of figure elements was intro-
duced during manuscript preparation. Despite their dif-
ferent horizontal scales, the red R14,23 curve in Fig. 1 is

Downloaded from www.sciencemag.org on December 23, 2010

the same as that in Fig. 3A; likewise, the red R14,14 curve
in Fig. 1 is the same as the green curve in Fig. 3A. The
configuration of current contacts and voltage probes
shown in Fig. 3A is fully equivalent to the four- and two-
terminal configurations of a standard Hall bar as shown
in Fig. 1. Therefore, this unintended duplication does not
affect any claims in the paper. A corrected version of Fig.
1, based on data taken from a Hall bar device different
from the one shown in Fig. 3A, is shown here. The origi-
nal caption is correct.

Research Article: “Identifying autism loci and genes by acknowledgment to the contribution of the late Ahmad idea of studying complex disorders in consanguineous
tracing recent shared ancestry” by E. M. Morrow et al. Teebi to the work presented here. He pioneered the study populations. We are indebted to his generous collabora-
(11 July 2008, p. 218). The authors wish to add an of genetic disorders in the Arab world and inspired the tion and dedicate this work to his memory.

Call for
Papers Science Translational Medicine
Integrating Medicine and Science
Science Translational Medicine, from Submit your manuscripts for
AAAS, the publisher of Science, focuses review in the following areas
on the conversion of basic biomedical of translational medicine:
research into practical applications, thus • Cardiovascular Disease
Chief Scientific Adviser bridging the research-to-application gap, • Neuroscience/Neurology/
Elias A. Zerhouni, M.D. linking basic scientists and researchers. Psychiatry
Former Director, • Infectious Diseases
National Institutes of Health For more information see • Cancer
ScienceTranslationalMedicine.org or • Health Policy
contact scitranslmededitors@aaas.org • Bioengineering
• Chemical Genomics/
Submit your research at Drug Discovery
www.submit2scitranslmed.org • Other Interdisciplinary
Approaches to Medicine

ScienceTranslationalMedicine.org

1746 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

Published by AAAS

EVOLUTION
At the Sesquicentennial of Origin
David P. Mindell
M
ore than a few biol-
ogists have fanta-
sized about tak-
ing Charles Darwin out
for a beer, to pick his brain
and hear his reactions to a
whirlwind update on evo-
lutionary science. Old news
items to relate would have to
include the discoveries of the
material of inheritance, continen-
tal drift, and radiometric dating. More
recent topics to consider are equally compel-
ling and a bit kaleidoscopic, ranging from
extreme archaean lifestyles to epigenetics,
macroecology, ribozymes, Hox genes, units
of selection, and xenology in eukaryotes,
among many others. If you could put a single
book-length survey into the pigeon breeder’s
hands to focus the conversation, Evolution
Since Darwin: The First 150 Years, edited
by Michael Bell, Douglas Futuyma, Walter
Eanes, and Jeffrey Levinton, would be an
excellent choice as a comprehensive guide to
current understanding and controversies.
The volume, based on a November 2009
workshop at Stony Brook University, brings
together an accomplished set of authors rep-
resenting diverse subdisciplines of evolution-
ary biology. In 22 chapters and 8 commen-
taries, they trace the history of their subjects
from Darwin’s work, with particular focus on
the past 50 years. Stimulating and insightful
individually, the reviews combine to provide
a picture of a vibrant and expanding field.
Perhaps most impressive in the growth of
evolutionary science has been its success in
explaining change across scales. This includes
extremely different scales of time and geogra-
phy, hierarchical levels of biological organiza-
tion, disparate groups of organisms, and even
different scientific disciplines. Evolution may
be unparalleled among the natural sciences
CREDIT: COURTESY DAVID HILLIS AND SINAUER ASSOCIATES
in integrating so comprehensively, touching
everything from cells to awe.
Much of the growth in explanatory power
for evolution over the past 150 years stems
from growth in the kinds and richness of data.
We now have a wealth of fossils with reli-
able age estimates, from the earliest known
life forms 3.5 billion years ago to evidence
The reviewer is at the California Academy of Sciences, 55
Music Concourse Drive, San Francisco, CA 94118, USA.
E-mail: dmindell@calacademy.org
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
BOOKS ET AL.
viously hidden dimension for phylogeny, as
well as the dynamism of evolutionary con-
cepts (genealogy, heritability, selection) to
span natural hierarchies (species, genes) and
to yield new insights about the processes and
consequences of change over time.
Sketched tree of life. Estimated The new genomics data reveal new lev-
from RNA genes. els of complexity; however, students of orga-
nisms in nature are well aware of the limita-
for 25 different hominid tions of genetics and other molecular data,
species during the past 6 which are not yet well integrated, especially
million years. We have not causally, with the evolution of variable
detailed, long-term stud- traits, populations, species, species commu-
ies from the field and lab nities, and ecosystems functioning.
regarding the evolutionary Much of the work of integration lies ahead
impacts that diverse orga- and will require syntheses of new data and
Downloaded from www.sciencemag.org on December 23, 2010
nisms and species have on one expanded views on evolutionary process. In
another through competition, preda- a volume to mark the centennial of Origins
tion, parasitism, and mutualisms. We have that had objectives similar to those of Bell et
many well-documented instances of natural al. (1), many authors worked within the tenets
selection yielding adaptive change in popu- of the (then) modern synthesis. They empha-
lations and of speciation influenced by both sized gradual change as the dominant tempo,
natural selection and chance events. We have speciation largely by geographic isolation,
extensive molecular databases, including general equivalency between micro- and mac-
sets of homologous gene sequences allowing roevolution, and strictly bifurcating phylog-
phylogenetic estimation for all enies, and they paid relatively
living things. Growth in data little attention to processes of
feeds growth in the analytical Evolution Since Darwin development, epigenetics, and
methods and, subsequently, The First 150 Years ecology. The Bell et al. volume
the depth at which topics can Michael A. Bell, Douglas J. nicely illustrates how the com-
be addressed. Futuyma, Walter F. Eanes, munity of evolutionary biol-
Our time traveler would and Jeffrey S. Levinton, Eds. ogists is changing the view.
certainly have interest in some Sinauer, Sunderland, MA, Contributors elucidate highly
general truths from 50 years 2010. 704 pp. Paper, $69.95. variable rates of evolution for
of evolutionary genetics. For ISBN 9780878934133. different traits and lineages,
example, all life shares one diverse speciation modes, and
genetic code (with minor lateral gene transfer. Chap-
illuminating exceptions) and very similar ters move beyond gene-centrism for organis-
means of DNA processing; a good deal of mal evolution to greater concern and data for
the DNA in genomes has no apparent func- study of epigenetics, multiple levels of selec-
tion; the number of functional genes does not tion, and evolvability.
correlate with organismal complexity; gene The volume reflects the dramatic expan-
duplication is the main source of new genes; sion of the applications of evolutionary sci-
both protein function and gene expression ence in recent years. These now include
are important in phenotypic evolution; both individual, public, and environmental
chance and selection are important in evolu- health as well as forensics, vaccine devel-
tionary change; and genes can move laterally opment, conservation of species, and eco-
between distinct species as well as vertically system sustainability.
from parents to progeny. Back in the pub, Darwin may wonder,
Lateral gene transfer, particularly com- “Does all this activity mean evolution has
mon among Bacteria and Archaea, yields an lost its ability to excite fear and opposition?”
atomization of phylogeny for many lineages, Not yet. As the root for natural explanations
into reticulating networks of genealogy of human origins (quiescent for the current
among genes and suites of genes within and news cycle) and ultimate impetus for human
among organismal lineages. The pleasingly moral behavior and values, evolution remains
disruptive potential to distinguish between the disturbing discovery.
differing organismal and genetic histories
has been claimed by some (falsely I think) to References
1. S. Tax, Ed., Evolution After Darwin: The University of
invalidate Darwin’s view of the tree of life, Chicago Centennial (Univ. Chicago Press, Chicago, 1960).
and it would be fun to ask the man about that.
Lateral gene transfer does demonstrate a pre- 10.1126/science.1199052
1747
Published by AAAS
 BOOKS ET AL.
EXHIBITION: NEUROSCIENCE
A Natural History
of the Brain
Abigail Rabinowitz1 and Carl E. Schoonover2
T
he American Museum of Natural His-
tory’s Brain: The Inside Story does not
open with the customary brain numer-
ology—the billions of neurons and synapses,
the eons of evolution spent packing it all into
1.4 kg of tissue. Instead, you feel your way
down a winding corridor surrounded by 680
kg of tangled electrical wire and optical fiber.
Spanish artist Daniel Canogar’s installation is
lit up with rapid trickles of light and sheets
of shifting color—a cross between forest and
funhouse. There is no better way to see the
organ as many a brain scientist does: a stag-
geringly complex, interconnected tangle in
which countless subtle signals whizz by at
breakneck speed.
Once the challenge of understanding the
brain is made viscerally clear, the exhibition
begins. Five main sections cover topics from
the nervous system’s cellular workings and its
role in sensation to how our brains learn and
change over time. The immensely ambitious
exhibition, based on a knowledge set that is
still patchy, aims to explain the human brain
to an audience of all ages.
If you are new to gray matter, trying to
understand everything from synapses to syn-
esthesia in two hours can feel
demoralizing, like cramming
from an encyclopedia for a mul-
tiple-choice test. But at its best,
the show moves away from text-
heavy placards to displays that
encourage you to understand the
brain intuitively. Some of this
science exhibition’s most effec-
tive teaching tools are works of
art. To explain how we perceive,
Devorah Sperber’s visual puzzle
assembles delicately tinted spools
of thread into an abstract shape
that, once refracted through a
glass orb, resolves into a famous
portrait. Her installation offers a
powerful metaphorical account
of how a nervous system takes
in disjointed bits of information
The reviewers are at 1EastWest Institute,
11 East 26th Street, 20th Floor, New York,
NY 10010, USA, and 2Department of
Neuroscience, Columbia University, New
York, NY 10032, USA. E-mail: ces2001@
columbia.edu
1748
and synthesizes them into a clear, seamless cerebellum is, for instance, doesn’t help you
percept—and does so in a manner that feels understand what it does.
effortless to the perceiver. Just across the way, But in a few brilliant moments, the show
a 1.8-m-tall homunculus with monstrously really succeeds in explaining why brain local-
large lips, hands, and feet
symbolizes the relative size
ization matters. At a hands-
on table, you are invited
Brain
of the somatosensory cor- to trace a star shape with
The Inside Story
tex’s representation of vari- a stylus while looking at
Rob DeSalle, Joy Hirsch,
ous body parts. (Parents, do your hand in a mirror. It’s a
and Margaret Zellner, curators
not fear: one oversized hand clumsy, frustrating task … at
American Museum of Natural
is strategically placed.) least at first. Above the sta-
History, New York, through 14
The curators often August 2011. Guangdong
tion, a panel lucidly teases
employed art effectively, so Science Center, Guangzhou, China, apart why practice makes
it is particularly unfortunate 19 November 2011 to 30 April 2012. perfect: As the procedural
that they didn’t better use Parque de las Ciencias, Granada, memory of the task becomes
visual cues to help unite the Spain, 14 July 2012 to 6 January ingrained over time, differ-
Downloaded from www.sciencemag.org on December 23, 2010
vast amount of information. 2013. Codie Idee per la Cultura, ent brain areas gradually
Moving from section to Torino, Italy, 2 March 2013 to 18 take over its execution. The
section, one finds the same August 2013. www.amnh.org/ more frontal (“planning/
brain regions in panel illus- exhibitions/brain/ thinking”) regions initially
trations colored differently, recruited to solve the prob-
which forces visitors to keep track of confus- lem are replaced by areas that specialize in
ing new neuro-jargon to connect concepts coordinated motion, which explains why,
across the exhibition. Early on, an excel- with repetition, the task requires less and less
lent video linking a Juilliard dance student mental effort. It has become mindless.
practicing her routine to her brain’s activity Some of the show’s most dazzling
(illuminated in a clear model) gives a very moments occur at the end, in a section
immediate and useful overview of the brain’s that explores advances we might look for-
geography. But instead of being referenced ward to thanks to this brave new science.
throughout to anchor the exhibition, it stands In one video, we encounter a scientist who
alone, the point it so vividly conveys forgot- has developed a system for a disabled man
ten. More generally, we felt that the show that, by capturing his neural activity, allows
often devotes too much time to naming brain him to control the motion of a virtual hand
regions without explaining why the informa- (and perhaps one day an actual prosthetic
tion is important. Simply knowing where the one). Another uses functional magnetic
resonance imaging to link musi-
cal and athletic aptitude to brain
activation patterns, exploring the
notion that we, our talents, our
tastes, our very selves, are noth-
ing but the product of our ner-
vous systems.
Despite our reservations about
the exhibition’s lack of unity and
unnecessary difficulty in parts,
we found that Brain: The Inside
Story presents a compelling snap-
shot of brain science and does so
CREDIT: COURTESY AMERICAN MUSEUM OF NATURAL HISTORY
without overselling researchers’
claims. Some of the exhibition’s
most moving and successful dis-
plays are also its most simple. In a
section on aging, two plasticized
human brains are placed side by
side—one plump and healthy and
one whose cortical folds have
been thinned by Alzheimer’s—
requiring no explanation and
serving as a silent reminder of
how far we have yet to go.
Tricky tracing task. 10.1126/science.1200403
24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org
Published by AAAS

SCIENCE POLICY
The Challenge of Feeding Scientific
Advice into Policy-Making
Roland Schenkel
B
oth the United States and the Euro-
pean Union are facing new challenges
in terms of how science is viewed and
used. There continues to be tension between
scientific information and societal and polit-
ical priorities. How can we explain the gap
between science and policy-making while
confronting misperceptions and promoting
positive views of science among the public?
The United States and the European
Union have recently undergone major politi-
cal changes, reflected in a reevaluation of the
role of science in the policy-making appara-
tus. The Obama Administration has focused
on science as a central component of the pol-
icy agenda. Concrete measures include re-
invigorating the President’s Council of Advi-
sors on Science and Technology (PCAST);
increasing investment in research and edu-
cation; and appointing respected scientists,
such as John Holdren and Stephen Chu, to
senior-level positions. Similarly in Europe,
the president of the European Commission,
José Manuel Barroso, has initiated the Europe
2020 agenda, which has research and inno-
vation at its core. Perhaps less trumpeted but
also important has been President Barroso’s
announcement of the creation of a Chief Sci-
entific Adviser position.
However, our governments are currently
faced with a significant financial crisis. Will
they continue to take a long-term view and
invest in the future? Or will they succumb to
pressures and cut funding for science? Like-
wise, will the scientific community be mod-
est enough to accept that science is just one
important consideration on the table when
decision-makers have to make choices?
What follows is a plea for a more frequent
and more issue-driven dialog between
policy-makers, the scientific community, and
all relevant stakeholders, based on observa-
tions that have shaped my career in providing
scientific advice.
Robust Science for Policy
Policy-makers today want to receive evidence-
based information, including a cost-benefits
CREDIT: SKB
Former Director-General, Joint Research Centre, European
Commission, Square de Meeus 8, 1049 Brussels, Belgium.
E-mail: roland.schenkel@t-online.de
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
Nuclear waste. Citizens visiting
the Äspo Hard Rock Laboratory as
part of public consultations about
the Swedish National Nuclear
Repository for spent nuclear fuel.
analysis of every scenario under consider-
ation. They also expect to receive indepen-
dent, unbiased advice. Yet, this is not always
obvious when decisions reflect the need to
balance different opinions, as well as claims
and counterclaims from interest groups,
including scientists.
Three case-studies spring to mind that
illustrate best practices and pitfalls when
feeding scientific advice into policy-making:
the long-standing debate about nuclear waste
management, the more recent global debate
on biofuel production, and Europe’s dis-
jointed response to the Icelandic volcanic ash
crisis. In considering these studies, it would
help to keep in mind five seemingly obvious,
but nevertheless critical, observations:
First, science is at the heart of invention
and the drive to make our lives better in a glo-
balized world. Legislative answers founded
on scientific evidence increasingly shape the
world we live in.
Second, science should not claim to have
“the” answer. Scientists from different dis-
ciplines should not be afraid to engage in
a “contradictory evidence-based mode” of
discussion, challenging each other with con-
flicting facts and uncertainties to arrive at
a better-informed, yet less narrow and more
harmonized view. This mode has advantages
above the classical peer-review used by scien-
tific journals and is better suited for a proper
treatment of multidimensional topics, even if
Published by AAAS
POLICYFORUM
Three case studies illustrate general principles
to guide scientists and policy-makers in
interactions with each other and the public.
Downloaded from www.sciencemag.org on December 23, 2010
it may result in “gray” literature only.
Third, industry and other core interest
groups have natural vested interests in policy-
making. Scientific outcomes are often better if
they participate in the process.
Fourth, public opinion is crucial and pub-
lic debate is instrumental in forming it. Scien-
tists must speak in a language that the public
understands, engaging in real dialog and mov-
ing away from the often arrogant “ex cathe-
dra” presentation style.
Finally, robust scientific advice has to be
multidimensional and inclusive. It must con-
sider economic, social, environmental, ethi-
cal, and scientific aspects, while indicating
how best to deal with uncertainties.
Nuclear Waste
The legacy of public opinion following the
Harrisburg and Chernobyl accidents, when
nobody could be seen to promote nuclear
energy, least of all the nuclear energy indus-
try, was a major impasse in having rational,
evidence-based dialog for many years (1).
Such was the veil of secrecy around these
events and so significant were the short-
comings in communications at that time, it
is easy to understand why decision-makers
were paralyzed.
However, in 2008, 37% of Europeans were
in favor of nuclear energy, rising to 58% if the
waste issue was solved (2). This represents a
sea-change compared with views expressed
1749
 POLICYFORUM

two decades ago. How has it been possible— lar positions have clouded a clear focus on the Volcanic Ash
in Northern Europe—to convince the public scientific evidence. One only has to think of The Icelandic volcanic ash disruption was
of the long-term safety of highly active waste the enormous influence wielded by the agro- the largest disruption to aviation since 9/11.
stored underground for up to hundreds of economic sector or the competition between Thousands of tons of mineral ash were thrown
thousands of years? The history of discussions oil and biofuels in the energy sector. There into the air, creating a plume of fine particles
in Sweden provides a good example of how to are also environmental groups concerned that rose 6 to 10 km (20,000 to 35,000 feet)
make progress. about issues such as monocultures, pesti- into the atmosphere (see the third figure).
Starting in the early 1990s Swedish author- cides, genetically modified (GMO) crops, These abrasive particles had the potential
ities conducted an extensive public engage- and deforestation. The developing-country to erode metal; clog fuel, sensing, and cool-
ment campaign to identify candidate sites for and export perspectives are also important in ing systems; and melt to form glassy depos-
deep geological waste disposal (3). In 2009, terms of interregional and international trade. its in jet engines. What astonished many was
the Forsmark site was In assessing the many that in such a (in principle, predictable) cri-
selected as the location issues involved, one must sis the European Union had little real power
of the Swedish National contend with consider- or competence to act and that there appeared
Nuclear Repository for able scientific uncertain- to be so little scientific evidence available.
highly active wastes (see ties, especially in predict- Nobody knew what kind or amount of volca-

Downloaded from www.sciencemag.org on December 23, 2010

first figure). The safety ing the competitiveness nic ash would be safe for an aircraft. Nobody
assessment prepared by of biofuels, determining had rigorously tested how different engines
the Swedish Nuclear Fuel whether biofuels are sus- cope with different concentrations of ash. The
and Waste Management tainable, and calculating result was that 75% of European airspace was
Company (SKB) is a ref- their overall greenhouse closed from 14 to 21 April 2010. Closures
erence document in terms gas emissions balance were based on a zero-tolerance strategy ema-
of its focus on long-term ( 10). Greenhouse gas nating from the precautionary principle with
safety aspects and cover- release from fertilizers the weight of evidence being given to previ-
age of plausible scenarios (11), the role of refores- ous instances of jet engine failure when flying
(4). Last July, after intense tation from fast grow- through some volcanic ash clouds.
consultation, the Swed- ing trees, and competi- There are several lessons to be learned.
ish government decided Biofuels. Some of the complex issues tion with other bioenergy First, in crisis situations, it is difficult to pull
to proceed with build- around the definition of a EU biofuel policy. sectors are just some of together the expertise of different scientific
ing new nuclear reactors. the issues where more disciplines (in this case, volcanology, atmo-
Sweden currently has 12 operating reactors at research is needed. We still need a robust, spheric science, material sciences, remote
three sites, producing 45% of its gross elec- global map of existing biofuel production and sensing, engineering, security, economics,
tricity generation (5). its impacts. Uncertainties can also be “beyond etc.), with the aim of getting better-informed
What distinguishes the Swedish approach science,” when, for instance, one needs to use answers. Second, there were glaring method-
from previous attempts is that the nuclear assumptions of future oil prices to calculate ological shortcomings. For example, existing
industry placed strong emphasis on socio- competitiveness and land-use changes. scientific models of ash cloud movements
economic studies, interaction with all stake- Yet, there are positive lessons to be taken could not be adapted effectively to the uncer-
holders, fostering a spirit of openness and from the biofuel debate in Europe (see the tainties of actual eruptions. Similarly, the
transparency and above all, careful listening second figure). Early interaction among the technology for satellite observation of ash
and accurate communication (6). Represen- scientific community, interest groups, and and SO2 in the atmosphere existed, but its
tatives literally knocked on every door and policy-makers has helped inform decision- ability to generate quantitative results had not
engaged with their most ardent critics so as making. Swift and intensive scientific con- been validated. Neither was there sufficient
to challenge the “not in my backyard” argu- sultation (12) has resulted in policy-support integration between airborne and ground
ments. The result was strong local support for documents that better address many of the measurements.
candidates in favor of the final repository (7– scientific questions raised, particularly con- We should not point the finger of blame at
9). Perhaps there are lessons to be learned here cerning biofuel sustainability. EU legislation, policy-makers. They reacted correctly in opt-
for future Yucca Mountains. for example, now states that, in order to be ing for a zero-tolerance policy in the absence
sustainable, future energy-related agriculture of clear scientific guidance. Individual gov-
Biofuels must guarantee a greenhouse gas reduction of ernments and various European authori-
There have been major international debates at least 35% with respect to fossil fuel. This ties have been quick to request urgent basic
in recent years over the production of biofuels target rises to 60% by 2018 (13). research. New satellite information and dis-
for transport and energy applications, stimu- Policy-makers worldwide must be cred- persal models should be integrated into a
lated by political targets and fears about the ited for having financed a number of stud- common European Union platform for use by
environment and food price impacts (partic- ies (14) addressing the most uncertain issues, risk managers.
ularly in poorer countries). Growth, harvest- which will serve as a basis for future deci-
ing, and processing release CO2, the amount sion-making and adaptation or revision of the Scientists Wanted
of which depends on a wide range of factors, present legislative texts and targets. However, Public engagement with all stakeholders on
such as plant species, growth conditions, and achieving a level of transparency that can sat- sensitive scientific issues is the key to suc-
CREDIT: EU/JRC

the soil used. isfy all stakeholders, even in large public con- cess. We have to engage more on the oppor-
The vested interests of different lobbying sultation processes, continues to be elusive tunity-rich, but risk-prone, developments in
groups pressing governments to take particu- (15). nanotechnologies, genetic engineering, stem

1750 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

Published by AAAS
 POLICYFORUM

cell, and brain research, to name but a few, to We must engage with the media, explain horizon, be they general trends in nature and
remain at the forefront of research and innova- what we are doing, and speak to them in society or technological developments, sci-
tion. Otherwise, doubters will win out, even if terms that they and their consumers can entists must help policy-makers tackle the
they are a minority. understand. I would like to see more scien- right issues at the right time. At the Euro-
Europe needs to analyze critically the tific organizations offering real support to pean Commission’s Joint Research Centre,
damage being done by an overzealous reli- science journalism and training programs. we are trying to meet this need by setting up
ance on the precautionary principle. It seems Similarly, when organizing scientific con- a scientific foresight and policy anticipation
to me that if there is any doubt, we delay deci- ferences, we might consider greater political capacity. Feeding scientific advice into pol-
sions and others pass us by. and public engagement. Put elected officials icy-making will be challenging. But what
The licensing of GMO technologies is a and third-party groups on your panels, listen is important is that we defend and assert
clear example. It took 13 years for the Amflora to your opponents, confront them (if neces- the inherent integrity of science, demon-
genetically optimized starch potato strate openness, speak in terms the public
to get approval for commercial can understand, and show that we take our
application in Europe. Despite the duty to society seriously (17). If we strive
European Food Safety Author- to achieve this, then evidence-based policy
ity’s having confirmed on several may just win over policy-biased evidence.

Downloaded from www.sciencemag.org on December 23, 2010

occasions that Amflora is safe for
humans, animals, and the environ- References and Notes
1. Nuclear Energy Agency, Public Attitudes to Nuclear
ment, even now, it can only be used Power (OECD, Paris, 2010), p. 45; www.oecdnea.org/ndd/
for industrial starch. This level reports/2010/nea6859-public-attitudes.pdf.
of precaution when the science is 2. European Commission (EC), Special Eurobarometer 227:
clear must be challenged. Radioactive Waste (EC, Brussels, 2005), pp. 26 and
30; http://ec.europa.eu/public_opinion/archives/ebs/
The scientific community must ebs_227_en.pdf.
learn to better defend its basic inter- 3. SKB (A.B. Svensk Kärnbränslehantering), Swedish
ests, such as funding for education, Nuclear Fuel and Waste Management Co., Final Reposi-
tory for Spent Fuel in Forsmark—Basis for Decision and
research, and innovation. In 2000, Reasons for Site Selection (SKB, Stockholm, 2009);
the combined total of GDP for www.skb.se/f0a51b8e-d851-492c-a218-32bd7a857b1b.
research expenditure for EU mem- fodoc.
4. SKB, Long-Term Safety for KBS-3 Repositories at Fors-
bers was 1.85%, and faced with
mark and Laxemar—A First Evaluation: Main Report of
stiff competition from the United the SR-Can Project (TR-06–09;KB, Stockholm, 2006);
States, China, and Japan; in par- www.skb.se/Templates/Standard____17139.aspx.
ticular, we declared our intention Volcanic ash over Europe. Ash from Iceland’s erupting Eyjaf- 5. EC, EU Energy and Transport in Figures (EC, Brussels,
jallajökull Volcano had drifted over northern Europe by April 2010), p.80; http://ec.europa.eu/energy/publications/
to increase this to 3% by 2010. This statistics/doc/2010_energy_transport_figures.pdf.
16, 2010. The brown ash is mixed with clouds in this photo-
“Lisbon Agenda” target became a like image taken by the Moderate Resolution Imaging Spectro- 6. Public consultation reports of SKB, www.skb.se/Tem-
rallying call for all, but has slowly radiometer (MODIS) on NASA’s Terra satellite. The airborne ash plates/Standard____24151.aspx.
7. Eight out of 10 support a final repository in Forsmark
lost out. In 2008, the real figure grounded flights across much of northern and western Europe [press release] (SKB, Stockholm, 2009); www.skb.se/
was still stagnating at 1.9% (16). starting on April 15. As the ash moved south, more countries Templates/Standard____29472.aspx.
Perhaps we are too good at setting began to close their airspace. 8. Annual Report 2009 (SKB, Stockholm, 2009); www.skb.
targets in the EU in the knowledge se/upload/publications/pdf/SKB_Verksamhet_2009_
Engelsk_web.pdf.
that we may never achieve them. sary), but do not be content to exist in a com- 9. Forsmark for Swedish Nuclear Waste, World Nuclear
It is too easy to point the finger of blame fortable, parallel universe. News, 3 June 2009; www.world-nuclear-news.org/WR_
at the politicians. But what has the scientific There are policy-makers who bury their Forsmark_for_Swedish_nuclear_waste_0306091.html.
10. T. Searchinger et al., Science 319, 1238 (2008).
community really done to make the 3% tar- heads in the sand when faced with compel- 11. P. J. Crutzen, A. R. Mosier, K. A. Smith, W. Winiwarter,
get happen? Have we organized ourselves to ling scientific evidence for unpopular policy Atmos. Chem. Phys. 8, 389 (2008).
protect it? Do we leave protests to students changes, believing all too easily that sci- 12. R. Edwards, D. Mulligan, L. Marelli, Indirect Land Use
Change from Increased Biofuels Demand: Comparison
only? Do we speak up to defend the future of ence is an à la carte menu. I would like to see
CREDIT: JEFF SCHMALTZ/MODIS RAPID RESPONSE TEAM AT NASA GSFC

of Models and Results for Marginal Biofuels Production

our countries? Do we think that things will more members of the European Parliament from Different Feedstocks (EC Joint Research Center
get any better during a period of cutbacks recruit qualified scientists into their teams. Scientific and Technical Report, EU, Luxembourg, 2010);
and austerity packages if we leave public The European Commission might consider http://re.jrc.ec.europa.eu/bf-tp/.
13. EU Announces Guidelines for Biofuel Sustainability,
debate to others? a new policy of temporarily placing high- www.biofuelsb2b.com/B2B_news.php.
Scientists should not be afraid to engage level scientists directly around its Commis- 14. Intelligent Energy—Europe programme, http://
in politics, from local to global. This sioners and senior teams, as well as offering ec.europa.eu/energy/intelligent/index_en.html.
15. J. Rankin, European Voice, 3 September 2010;
increases the risk of not being perceived as to do so in the European Parliament, when www.europeanvoice.com/article/imported/biofuel-policy-
independent, but this can be mitigated by working on major dossiers needing daily lacks-transparency-/69027.aspx
always sticking to the facts. We must strive scientific expertise. 16. Gross domestic expenditure on R&D (GERD), http://epp.
to be unbiased, no matter the context. This We scientists need to be a little more mod- eurostat.ec.europa.eu/tgm/table.do?tab=table&init=1&p
lugin=1&language=en&pcode=t2020_20.
implies that we must speak up if researchers est in understanding that decisions taken by 17. Recommendations of the European Commission’s Joint
do not follow the principles of good scientific governments are, of course, ultimately polit- Research Centre and the American Association for the
behavior. At the same time, we must raise our ical and that science is just one of the many Advancement of Science, Euroscience Open Forum, Turin,
Italy, 2 to 7 July 2010.
voices if individuals or groups try to distract elements under consideration. Given their
the public from the evidence. feeling for issues that are appearing on the 10.1126/science.1197503

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1751

Published by AAAS
 POLICYFORUM

CONSERVATION

Boosting CITES
To protect biodiversity, more, improved
biological and trade data and analyses
are needed.
Jacob Phelps,1*† Edward L. Webb,1* David Bickford,1† Vincent Nijman,2 Navjot S. Sodhi1

I
nternational wildlife trade remains a lead-
ing threat to biodiversity conservation (1)
and is a common vector for infectious
diseases (2, 3) and invasive species (4) that
also affect agriculture, livestock, and public
health. With 175 member countries, the Con-
vention on International Trade in Endangered
Species of Wild Flora and Fauna (CITES) is
the most important global initiative to moni-
tor and regulate international trade of plants
ing mechanism for species-level informa-
tion (16). Yet this information is central to
CITES function (9, 15), as exporters must
complete nondetriment f inding (NDF)
reports to prove that international trade is
not harming populations of regulated spe-
cies (17). Such baseline data are also funda-
mental to listing species for CITES protec-
tion; commercially high-value species have
been listed on the basis of robust, empiri-
remains a leading challenge. For example,
more than 50% of documented live-animal
imports into the United States from 2000 to
2006 were identified only by class; only about
14% were identified to species (3). Weak data
sets overlook species introductions, substitu-
tions, and exporter misidentifications [e.g.,
(20)]. Traditional identification protocols and
methods are proving inadequate (3, 15) and
require revision and innovation (19, 21).

Downloaded from www.sciencemag.org on December 23, 2010

and animals (5). CITES regulates trade of
nearly 34,000 species and has reduced threats
associated with overharvest of imperiled spe-
cies for international trade.
Credible biological and trade data are core
to informing decisions and garnering politi-
cal will and consensus among CITES parties
(6). This does not preclude party bargaining,
as occurred during the March 2010 Confer-
ence of Parties (CoP) debate over bluefin
tuna [e.g., (7)]. Nevertheless, CITES deci-
sions are also frequently hindered by a lack of
basic data [e.g., (8–10)]. We highlight CITES
limitations and describe potential solutions
related to systematic data collection, rigorous
data analysis, flexible research methods, and
peer review.
Systematic, Standardized Data Collection
The CITES secretariat, Animals and Plant
Committees (APCs), and external agencies
[e.g., International Union for Conservation of
Nature (IUCN) Specialist Groups] depend on
national agencies to regulate trade. Yet many
CITES parties fail to systematically monitor
and report international wildlife trade [e.g.,
(11–13)]. Some of the largest exporters and
importers of wildlife products are not fully
compliant: Brazil, a significant source coun-
try for illegal fauna (14), lacks a function-
ing central mechanism for reporting wildlife
confiscations (15). The United States, a lead-
ing importer of wildlife, lacks a coordinated
national authority for monitoring wildlife
imports (3).
Many CITES parties fail to collect
domestic population and harvest data, and
CITES lacks a standard international report-
1
Department of Biological Sciences, National University
of Singapore, Singapore, 117543, Singapore. 2School of
Social Sciences and Law, Oxford Brookes University, Oxford,
OX3 0BP, UK.
*These authors contributed equally to this work. †Authors
for correspondence. E-mail: jacob.phelps@gmail.com
(J.P.), dbsbdp@nus.edu.sg (D.B.)
cal population data [e.g., (6, 18)]. However,
most taxa are understudied, and there is a Rigorous Analysis
lack of coordinated, systematic data collec- When data are available, analyses under the
tion within and among parties [supporting Secretariat, APCs, and their collaborators
online material (SOM)]. often remain insufficient to identify species
Data collection at all levels depends on threatened by trade and to detect trade inac-
proper species identification (19), which curacies and loop-holes. For instance, ~20%
of species threatened in four mega-diversity
Genera Count* for countries (Brazil, China, Colombia, and the
identified each genus Philippines) have not been assessed at the
international level (22). Similarly, the IUCN
CITES Trade Ascocentrum 5 holds “no information” about the status of
Database most of the Orchidaceae (23); only three spe-
Dendrobium 5 cies were added to the Red List of Threat-
Rhynchostylis 10 ened Species from 2007 to 2009, although
sufficient information exists to list many
Total count 20
others (24). A handful of studies have high-
lighted the need for enhanced, rigorous anal-
Market
observations Aerides 60 ysis (SOM), yet critical trade linkages often
remain undetected when CITES relies on
Arundina 14 the interest, resources, and often informal
Ascocentrum 7 or irregular input of independent research-
ers and organizations (25). Encouragingly,
Bulbophyllum 50
(including
CITES partners are developing tools to
Cirrhopetalum) enhance analysis capacity, such as the Trade
Data Dashboard (26).
Dendrobium 10
Eria 5 Flexible Methods
Vanda 6
Wildlife trade occurs openly at public bor-
der markets (27) and discrete black markets
Vanilla 16
(28). Trade activity shifts and cycles among
Total count 168 countries as wild populations are depleted
(12, 29), and innovative smuggling tech-
Orchid trade between Lao PDR and Thailand. niques are adopted in response to enforce-
Comparing CITES Trade Database (2000–09) and a ment pressures (28). However, trade data
1-day survey of a single market trader along the are collected using conventional techniques
Mekong River (February 2010). *The CITES count implemented along easily accessed trade
is based on the number reported, method unre- routes (e.g., airports), which cannot cap-
ported. Observed count is based on the number of
ture the true dynamics. For example, CITES
plant bundles (potentially including multiple indi-
viduals) plus the number of individuals (potentially
reports an insignificant fraction of CITES-
regulated wild orchid trade into Thailand
divisions of larger plants), both recorded as single
counts. This is conservative relative to traditional from Lao People’s Democratic Republic (see
customs recording, but not necessarily representa- the chart), Myanmar, Cambodia, and Viet-
tive of the number of genetically distinct individu- nam. A single small-scale trader at an infor-
als. (See SOM for details.) mal border market on the Mekong can sell

1752 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

Published by AAAS

more plants in a single day than reported by
CITES over a 9-year period (SOM, see the
charts on page 1752). Similar trade inaccu-
racies are evident across taxa (bears, edible
tubers, medicinal plants, seahorses, bush-
meat, and frogs) and regions (12, 20, 27, 30–
33). Some efforts have been made to inte-
grate alternative, investigative approaches
into CITES (e.g., the Lusaka Agreement and
CITES-INTERPOL collaborations), but the
overall CITES “airport bias” fails to detect
the majority of illicit trade.
CITES shortcomings may be overlooked
because the convention lacks internal
and external checks and balances. CITES
relies exclusively on country self-report-
ing, although incentives are high for biased
analyses and misreporting (34), and most
CITES-listed species occur in the tropics
where governance is often weak and corrup-
tion high (35). This is especially problematic
when CITES National Management Author-
ities lack independence from their advisory
Scientific Authorities (SOM) and because
parties’ submissions to CITES are not publi-
cally available (36).
Critical, independent peer-review offers
a legitimate means of party validation, par-
ticularly when addressing contentious issues
such as harvest quotas, approvals of NDFs,
proof of captive breeding, and national man-
agement procedures for protected species
(8). These reviews may meet with party
resistance that could hamper future inves-
tigative efforts, especially if they are fol-
lowed by legal action. However, the recent
pilot CITES Policy Review Project in four
exporting countries provides an encourag-
ing precedent for future external reviews
(37) (SOM).
Solutions in Context
CITES credibility, effectiveness, and suc-
cess at catalyzing consensus depend heav-
ily on punctilious data collection, analysis,
and synthesis. Yet the convention is bound
by political and economic realities. General
strategies through which to improve CITES
(table S1) must recognize that some mea-
sures may overlap, prioritization depends on
party needs and resources, and recommen-
dations may vary in their political feasibility.
CITES has improved party compliance
and science-based decision-making despite
political sensitivities, through provision of
technical support; mission visits and rec-
ommendations; simplified reporting pro-
cedures; and legal strategies, such as warn-
ings and threats of trade suspensions (5,
36). Such progress demonstrates CITES
recognition of the importance of enhanced
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
enforcement and data collection. Further
increasing the demands on CITES parties
and secretariat is necessary, but remains
administratively demanding, costly, and
politically challenging.
Some of the most urgent solutions (table
S1) require the greatest coordination among
parties and institutions. For example, col-
lection of baseline biological data on traded
species will require coordinated activities
among diverse stakeholders, ranging from
rural harvesters to multilateral agencies.
CITES has already enhanced data-sharing
and analysis through collaborations with
nongovernmental organizations and part-
nerships, such as the Wildlife Enforcement
Monitoring System. At the March 2010 CoP,
CITES instituted an illegal-trade database
working group to enhance data collection
and analysis (38). The majority of proposed
solutions depends on enhanced active, sus-
tained, and reciprocal engagement of CITES
parties with external partners.
Funding remains a principal limitation to
CITES, especially for on-the-ground execu-
tion of mandates and for proposed enhance-
ments (table S1) (25). The secretariat oper-
ates on meager party donations (25, 36)
of U.S. $5.2M per year for 2009–11 (39).
National-level funding for CITES enforce-
ment is similarly restricted, especially in
many tropical exporting countries. There
is a need for parties, particularly importing
nations, to increase contributions dramati-
cally. CITES costs should also be extended
to participating industries and consumers,
consistent with the “polluter pays” princi-
ple, while doing no harm to poor harvest-
ers (40). This can be accomplished through
trade levies on CITES-listed wildlife (9),
increased infraction penalties ( 19), and
wildlife certification schemes (41). Only
through increased resources can CITES
move toward proactive, real-time monitor-
ing and regulation to strengthen enforce-
ment and data quality.
After 35 years, the CITES framework
remains highly relevant, and the secre-
tariat and CoP should continue to facili-
tate progress among noncompliant coun-
tries and should exercise legal tools to
create consensus. However, current rigors
are inadequate, and meaningful improve-
ments will require greater financial and
political commitments. We propose targeted
CITES negotiations to establish new part-
nerships; to review financial commitments;
and to develop clear rules and progressive
standards for data collection, analysis, and
review. A strengthened convention is essen-
tial to protecting imperiled biodiversity.
Published by AAAS
POLICYFORUM
References and Notes
1. W. Sutherland et al., Conserv. Biol. 23, 557 (2009).
2. P. Daszak, A. A. Cunningham, A. D. Hyatt, Science 287,
443 (2000).
3. K. F. Smith et al., Science 324, 594 (2009).
4. P. M. Vitousek, C. M. D’Antonio, L. L. Loope, R. West-
brooks, Am. Sci. 84, 468 (1996).
5. P. H. Sand, Eur. J. Int. Law 8, 29 (1997).
6. T. Gehring, E. Ruffing, Glob. Environ. Polit. 8, 123 (2008).
7. S. Milius, Sci. News, 26 March 2010, Web edition;
www.sciencenews.org/view/generic/id/57679.
8. V. Nijman, C. R. Shepherd, Wildlife Trade from ASEAN
to the EU: Issues with the Trade in Captive-Bred Reptiles
from Indonesia (TRAFFIC Europe Report for the EC,
Brussels, 2010); www.traffic.org/species-reports/traffic_
species_reptiles26.pdf.
9. V. Nijman, Biodivers. Conserv. 19, 1101 (2010).
10. E. C. M. Parsons, N. A. Rose, T. M. Telecky, Mar. Policy 34,
384 (2010).
11. L. Yi-Ming, L. Dianmo, Biodivers. Conserv. 7, 895 (1998).
12. B. G. Giles, T. S. Ky, D. H. Hoang, C. J. Vincent, Biodivers.
Downloaded from www.sciencemag.org on December 23, 2010
Conserv. 15, 2497 (2006).
13. O. G. Amir, Wildlife Trade in Somalia (Antelope Specialist
Group, Northeast African Subgroup, IUCN Species Sur-
vival Commission, Geneva, 2006).
14. RENCTAS, Primer Relatorio Nacional Sobre o Trafico de
Fauna Silvestre [RENCTAS (Brazilian National Network to
Fight the Trafficking of Wild Animals), Brazília, 2001];
www.renctas.com.br/files/REL_RENCTAS_pt_final.pdf.
15. J. Pistoni, L. F. Toledo, Braz. Soc. Herpetol. 5, 51 (2010).
16. H. Gerson et al., Conserv. Biol. 22, 4 (2008).
17. CITES, “Non-detriment findings”; www.cites.org/eng/
prog/ndf/index.shtml.
18. A. G. Blundell, Oryx 38, 84 (2004).
19. G. E. Rosen, K. F. Smith, EcoHealth 7, 24 (2010).
20. M. Veith, J. Kosuch, R. Feldmann, H. Martens, A. Seitz,
Biodivers. Conserv. 9, 333 (2000).
21. E. P. Green, H. Hendry, Coral Reefs 18, 403 (1999).
22. D. Brito et al., Biol. Conserv. 143, 1154 (2010).
23. United Nations Environment Programme–World Con-
servation Monitoring Centre (UNEP-WCMC), Threatened
Species Database (UNEP-WCMC, Cambridge, 2010);
www.unep-wcmc.org.
24. IUCN Species Survival Commission, Orchid Specialist
Group General Meeting, Quito, Ecuador, 7 February
2009; www.orchidconservation.org/osg/Docs/OSG%
20Quito%20minutes.pdf.
25. L. Wilson-Wilde, Forensic Sci. Med. Pathol. 6, 221 (2010).
26. CITES, Trade Data Dashboards, http://cites-dashboards.
unep-wcmc.org/about.
27. N. Van Song, J. Environ. Dev. 17, 145 (2008).
28. B. Moyle, Global Crime 10, 124 (2009).
29. A. I. Carpenter, J. M. Rowcliffe, A. R. Watkinson, Biol.
Conserv. 120, 291 (2004).
30. C. R. Shepherd, V. Nijman, Biodivers. Conserv. 17, 35
(2008).
31. T. R. B. Davenport, H. J. Ndangalasi, Oryx 37, 55 (2003).
32. C. S. Olsen, N. Bhattarai, Mt. Res. Dev. 25, 37 (2005).
33. M. Mohneke, A. B. Onadeko, M. Hirschfeld, M. O. Rodel,
TRAFFIC Bull. 22, 117 (2010).
34. F. Courchamp et al., PLoS Biol. 4, e415 (2006).
35. N. S. Sodhi, B. W. Brook, C. J. A. Bradshaw, Tropical Con-
servation Biology (Blackwell Publishing, Oxford, 2007).
36. R. Reeve, Int. Aff. 82, 881 (2006).
37. Wildlife Trade Policy Reviews, www.cites.org/eng/prog/
policy/index.shtml.
38. Gathering and analysis of data on illegal trade,
www.cites.org/eng/dec/valid15/15_42-43.shtml.
39. Resolution on financing, www.cites.org/eng/res/
14/14-01.shtml.
40. B. Dickson, Oryx 42, 548 (2008).
41. I. G. Warkentin, D. Bickford, N. S. Sodhi, C. J. Bradshaw,
Conserv. Biol. 23, 1056 (2009).
42. Funding for J.P. provided by the Harry S. Truman Founda-
tion, for E.L.W. from the Singapore Ministry of Education
grant no. R-154-000-400-133.
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1752/DC1
10.1126/science.1195558
1753
 PERSPECTIVES
NEUROSCIENCE

Ubiquitination Inhibits Newborn neurons shut down a protein

destruction mechanism to migrate to their

Neuronal Exit
final destination.

Christine Métin and Camilla Luccardini

I
n the developing brain, neurons increase ing and recycling of adhesion molecules increased neuronal exit. In contrast, overex-
their numbers in an “amplification com- (which enable cells to stick to each other or pressing a functional Siah in the prolifera-
partment,” from which they emigrate to to extracellular products) and regulate actin tive area blocks neuronal exit and/or alters
colonize distant brain structures. How young cytoskeleton dynamics (2, 3). In the cerebral the direction of neuronal movements.
neurons leave these compartments, however, cortex, researchers have suggested that poly- Siah’s target in granule cells is the parti-
has been unclear, although researchers have ubiquitination and degradation arrests the tioning defective-3 (Pard3) protein, which
hypothesized that they turn comprises a “degron,” or bind-

Downloaded from www.sciencemag.org on December 23, 2010

on a biochemical “migration ing motif (sequence) that acts
program” that enables them as a degradation signal (5).
to exit. On page 1834 of this By observing the pattern of
issue, however, Famulski et Siah expression in the devel-
al. (1) show that one common oping cerebellum, Famulski
type of neuron uses an oppo- et al. showed that the radi-
site strategy. To exit, young ally migrating granule cells
cerebellar granule neurons express higher levels of Pard3
shut down a protein ubiqui- protein than progenitor cells.
tination mechanism that inhib- Pard3 associates to partition-
its the formation of a molec- ing defective-6 (Pard6) and
ular complex that controls a protein kinase C (PKC) to
adhesion and allows the neu- form the PAR polarity com-
rons to migrate to a final des- plex, which plays a major
tination. Granule cells, a large role in cell orientation. PAR
population of small neurons is a major regulator of api-
in the central nervous system, cal-basal polarity in epithe-
proliferate at the surface of the lial cells, for instance, and of
developing cerebellum. They front-rear polarity in migrat-
then undergo a transient phase ing cells (6). Solecki et al. had
of “tangential” migration near previously shown that the PAR
the cerebellar surface before complex localizes to the cen-
extending a radial process trosome, and that Pard6 sig-
(sprout-like extension) and naling regulates centrosomal
migrating to deeper layers far- motility by acting on micro-
ther from their birthplace (see tubules and the actomyosin
the figure). The mechanisms cytoskeleton (7, 8). The pres-
that control the onset of radial ent study investigates another
migration are poorly under- cellular mechanism by which
stood. In their study, Famulski No exit. Siah ubiquitin ligase inhibits the exit of developing young neurons (top) by pre- the PAR complex controls the
et al. explore the role of ubiq- venting the formation of PAR3/Jam-C complexes that are required for migration along migration of granule cells.
uitination, a common cellular the radial glia. After their last division, young cerebellar granule cells located near the In particular, the authors
process in which the protein brain surface extend long tangential processes (middle). They are prevented from moving identified the junctional adhe-
ubiquitin bonds to proteins, deeper by Siah activity. If Siah is switched off, the cell extends a third process along the sion molecule C (JAM-C)
marking them for destruction radial glia (bottom), along which the nucleus will move toward the deeper brain structure. as a molecular partner of the
and recycling. Research is PAR complex in granule cells.
revealing that ubiquitination is as important radial migration of neurons (4). JAM-C directly and strongly associates to
as phosphorylation in regulating biological Famulski et al. combined a clever exper- Pard3 by an intracellular PDZ domain and
CREDIT: N. KEVITIYAGALA/SCIENCE

processes. Early analyses of ubiquitination’s imental approach with sophisticated imag- localizes at apical tight junctions together
role in cell migration, for instance, showed ing techniques to show that the ubiquitin with the PAR complex in epithelial cells
that it can activate the intracellular traffick- ligase Seven in Absentia (Siah) regulates (9). They then performed acute gain- and
the movements of neurons in the develop- loss-of-function experiments in developing
ing cerebellum. Silencing Siah, or over- brain tissue to show that Pard3 controls the
Institut du Fer à Moulin, INSERM U839, 75005 Paris,
France. E-mail: christine.metin@inserm.fr; camilla.lucca- expressing a truncated Siah ligase in the radial migration of granule cells by control-
rdini@inserm.fr region where neurons proliferate, results in ling the formation of JAM-C positive cell-

1754 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

Published by AAAS

cell contacts. This work identified a previ-
ously unknown function of the Pard3/JAM-C
complex and shed light on a new mechanism
involved in radial migration.
The radial migration of newborn gran-
ule cells starts with the extension of a pro-
cess along thin radial fibers in the developing
brain. This process emerges in front of the
centrosome, which moves in afterward (10).
Several models suggest that radial migration
requires forward movements of the centro-
some toward the tip of the elongating pro-
cess and coordinated nuclear translocations
(11). Several factors that can affect this pro-
cess in granule cells have been identified. In
particular, previous studies have shown that
the protein Semaphorin-6A and its Plexin A2
receptor control the switch from tangential
to radial migration and regulate the nuclear-
centrosomal coupling (12). Nevertheless,
the cell mechanisms and signaling pathways
involved remain poorly understood.
Famulski et al. showed that Siah regu-
lates the morphogenetic movements of gran-
ule cells through the post-translational regu-
lation of the Pard3/JAM-C adhesive com-
PLANETARY SCIENCE
Generating an Atmosphere
Dale P. Cruikshank
T
he presence of water ice on most of
the large satellites of the outer plan-
ets was established many years ago
through near-infrared (1- to 2.5-µm wave-
length) observations with ground-based
telescopes. Frozen carbon dioxide, sul-
fur dioxide, methane, nitrogen, and other
molecular ices are also found in various
combinations on inner planets such as Mars
to bodies far beyond Pluto. Recent discov-
eries of ice varieties on some asteroids and
sequestered in protected regions on Mercury
and the Moon point to the near-universal
distribution of frozen volatiles throughout
the solar system (1–3). On page 1813 of this
issue, Teolis et al. (4) report the detection
CREDIT: NASA/JPL/SPACE SCIENCE INSTITUTE
of a tenuous (approximately one in five tril-
lionth of Earth’s atmospheric pressure at sea
level) oxygen (O2) and carbon dioxide (CO2)
atmosphere surrounding Saturn’s icy moon
Rhea (diameter of 1529 km) measured as
the Cassini spacecraft passed by only 97 km
Astrophysics Branch, NASA Ames Research Center, Moffett
Field, CA 94035–1000, USA. E-mail: dale.p.cruikshank@
nasa.gov
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
plex. A limitation of their findings, however,
was that they could not directly demonstrate
that the Pard3 protein is both translated and
degraded by Siah-dependent ubiquitination in
immature granule cells that are located near
the surface of the brain structure. Indeed, as
granule cells overexpessing Pard3 are induced
to migrate radially, Siah invalidation that
increases the Pard3 signal may also induce the
exit of transfected cells toward deeper brain
regions. Famulski et al. circumvented this dif-
ficulty by using in vivo imaging to visualize
the extinction of a fluorescent signal linked to
the degron motif of Pard3 in Siah-expressing
cells located at the surface of the cerebellum.
Using the cerebellum as a model system,
Solecki and colleagues shed light on a previ-
ously unknown mechanism to control neu-
ronal migration from proliferative compart-
ments. Interestingly, their results suggest that
increased migration from the compartment
is associated with reduced cell cycle exit.
Repressing neuronal exit by posttranslational
modification of a protein in a complex appears
to be a very efficient means of quickly adapt-
ing both the proliferation and the migration
Rhea close up. Image in visible light
on 17 October 2010 from a distance
of 44,000 km above the moon’s sur-
face. The image was obtained with
the Cassini Imaging Science Subsys-
tem narrow-angle camera, showing
ancient cratered terrain. The frame is
280 km on each side.
above the surface in March 2010
(see the figure).
The Ion Neutral Mass Spec-
trometer onboard Cassini cap-
tured a sample of Rhea’s atmo-
sphere and sorted the molecules
by mass, confirming the pres-
ence of O2 and CO2. When these
measurements are combined
with data acquired by two other
Cassini instruments on more dis-
tant flybys of Rhea in 2005 and
2007, a picture emerges in which O2 formed
by the irradiation of water ice is ejected
from the surface by charged-particle inter-
actions, and the tenuous gas is then swept
away into space. These new results expand
and clarify our understanding of the pro-
Published by AAAS
PERSPECTIVES
of young neurons to environmental changes.
It is now important to determine whether a
similar mechanism controls the migration of
embryonic neurons along radial glia in other
brain regions, particularly in the cerebral cor-
tex, and how it interacts with other guidance
mechanisms. At the cellular level, another
important question is how this new function of
the PAR complex at the periphery of migrat-
ing neurons correlates with the function of the
PAR complex function in the centrosome.
References
1. J. K. Famulski et al., Science 330, 1834 (2010).
2. V. H. Lobert et al., Dev. Cell 19, 148 (2010).
3. Y. Chen et al., Mol. Cell 35, 841 (2009).
4. L. Feng, N. S. Allen, S. Simo, J. A. Cooper, Genes Dev. 21,
Downloaded from www.sciencemag.org on December 23, 2010
2717 (2007).
5. C. M. House et al., Proc. Natl. Acad. Sci. U.S.A. 100,
3101 (2003).
6. S. Etienne-Manneville, Oncogene 27, 6970 (2008).
7. D. J. Solecki et al., Nat. Neurosci. 7, 1195 (2004).
8. D. J. Solecki et al., Neuron 63, 63 (2009).
9. G. Mandicourt et al., J. Biol. Chem. 282, 1830 (2007).
10. A. Chédotal, Trends Neurosci. 33, 163 (2010).
11. C. Métin, R. B. Vallee, P. Rakic, P. G. Bhide, J. Neurosci.
28, 11746 (2008).
12. J. Renaud et al., Nat. Neurosci. 11, 440 (2008).
10.1126/science.1200475
The oxygen and carbon dioxide atmosphere
on Saturn’s moon Rhea is produced by a
photochemical reaction mechanism.
cesses by which new molecules are synthe-
sized, ejected, and then lost.
Solar system ices are not static deposits
that remain undisturbed for all time. Ices in
the interiors of planetary satellites and com-
ets can warm, evaporate, and burst through
1755

cell contacts. This work identified a previ-
ously unknown function of the Pard3/JAM-C
complex and shed light on a new mechanism
involved in radial migration.
The radial migration of newborn gran-
ule cells starts with the extension of a pro-
cess along thin radial fibers in the developing
brain. This process emerges in front of the
centrosome, which moves in afterward (10).
Several models suggest that radial migration
requires forward movements of the centro-
some toward the tip of the elongating pro-
cess and coordinated nuclear translocations
(11). Several factors that can affect this pro-
cess in granule cells have been identified. In
particular, previous studies have shown that
the protein Semaphorin-6A and its Plexin A2
receptor control the switch from tangential
to radial migration and regulate the nuclear-
centrosomal coupling (12). Nevertheless,
the cell mechanisms and signaling pathways
involved remain poorly understood.
Famulski et al. showed that Siah regu-
lates the morphogenetic movements of gran-
ule cells through the post-translational regu-
lation of the Pard3/JAM-C adhesive com-
PLANETARY SCIENCE
Generating an Atmosphere
Dale P. Cruikshank
T
he presence of water ice on most of
the large satellites of the outer plan-
ets was established many years ago
through near-infrared (1- to 2.5-µm wave-
length) observations with ground-based
telescopes. Frozen carbon dioxide, sul-
fur dioxide, methane, nitrogen, and other
molecular ices are also found in various
combinations on inner planets such as Mars
to bodies far beyond Pluto. Recent discov-
eries of ice varieties on some asteroids and
sequestered in protected regions on Mercury
and the Moon point to the near-universal
distribution of frozen volatiles throughout
the solar system (1–3). On page 1813 of this
issue, Teolis et al. (4) report the detection
CREDIT: NASA/JPL/SPACE SCIENCE INSTITUTE
of a tenuous (approximately one in five tril-
lionth of Earth’s atmospheric pressure at sea
level) oxygen (O2) and carbon dioxide (CO2)
atmosphere surrounding Saturn’s icy moon
Rhea (diameter of 1529 km) measured as
the Cassini spacecraft passed by only 97 km
Astrophysics Branch, NASA Ames Research Center, Moffett
Field, CA 94035–1000, USA. E-mail: dale.p.cruikshank@
nasa.gov
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
plex. A limitation of their findings, however,
was that they could not directly demonstrate
that the Pard3 protein is both translated and
degraded by Siah-dependent ubiquitination in
immature granule cells that are located near
the surface of the brain structure. Indeed, as
granule cells overexpessing Pard3 are induced
to migrate radially, Siah invalidation that
increases the Pard3 signal may also induce the
exit of transfected cells toward deeper brain
regions. Famulski et al. circumvented this dif-
ficulty by using in vivo imaging to visualize
the extinction of a fluorescent signal linked to
the degron motif of Pard3 in Siah-expressing
cells located at the surface of the cerebellum.
Using the cerebellum as a model system,
Solecki and colleagues shed light on a previ-
ously unknown mechanism to control neu-
ronal migration from proliferative compart-
ments. Interestingly, their results suggest that
increased migration from the compartment
is associated with reduced cell cycle exit.
Repressing neuronal exit by posttranslational
modification of a protein in a complex appears
to be a very efficient means of quickly adapt-
ing both the proliferation and the migration
Rhea close up. Image in visible light
on 17 October 2010 from a distance
of 44,000 km above the moon’s sur-
face. The image was obtained with
the Cassini Imaging Science Subsys-
tem narrow-angle camera, showing
ancient cratered terrain. The frame is
280 km on each side.
above the surface in March 2010
(see the figure).
The Ion Neutral Mass Spec-
trometer onboard Cassini cap-
tured a sample of Rhea’s atmo-
sphere and sorted the molecules
by mass, confirming the pres-
ence of O2 and CO2. When these
measurements are combined
with data acquired by two other
Cassini instruments on more dis-
tant flybys of Rhea in 2005 and
2007, a picture emerges in which O2 formed
by the irradiation of water ice is ejected
from the surface by charged-particle inter-
actions, and the tenuous gas is then swept
away into space. These new results expand
and clarify our understanding of the pro-
Published by AAAS
PERSPECTIVES
of young neurons to environmental changes.
It is now important to determine whether a
similar mechanism controls the migration of
embryonic neurons along radial glia in other
brain regions, particularly in the cerebral cor-
tex, and how it interacts with other guidance
mechanisms. At the cellular level, another
important question is how this new function of
the PAR complex at the periphery of migrat-
ing neurons correlates with the function of the
PAR complex function in the centrosome.
References
1. J. K. Famulski et al., Science 330, 1834 (2010).
2. V. H. Lobert et al., Dev. Cell 19, 148 (2010).
3. Y. Chen et al., Mol. Cell 35, 841 (2009).
4. L. Feng, N. S. Allen, S. Simo, J. A. Cooper, Genes Dev. 21,
Downloaded from www.sciencemag.org on December 23, 2010
2717 (2007).
5. C. M. House et al., Proc. Natl. Acad. Sci. U.S.A. 100,
3101 (2003).
6. S. Etienne-Manneville, Oncogene 27, 6970 (2008).
7. D. J. Solecki et al., Nat. Neurosci. 7, 1195 (2004).
8. D. J. Solecki et al., Neuron 63, 63 (2009).
9. G. Mandicourt et al., J. Biol. Chem. 282, 1830 (2007).
10. A. Chédotal, Trends Neurosci. 33, 163 (2010).
11. C. Métin, R. B. Vallee, P. Rakic, P. G. Bhide, J. Neurosci.
28, 11746 (2008).
12. J. Renaud et al., Nat. Neurosci. 11, 440 (2008).
10.1126/science.1200475
The oxygen and carbon dioxide atmosphere
on Saturn’s moon Rhea is produced by a
photochemical reaction mechanism.
cesses by which new molecules are synthe-
sized, ejected, and then lost.
Solar system ices are not static deposits
that remain undisturbed for all time. Ices in
the interiors of planetary satellites and com-
ets can warm, evaporate, and burst through
1755
 PERSPECTIVES
the surface, sending jets of gas and dust into
space, as we have seen emanating from Sat-
urn’s moon Enceladus and numerous com-
ets. Ices exposed on the cold surfaces of outer
planetary moons interact with the local space
environment, as the incident solar ultraviolet
light and charged particles from deep space
and trapped in the parent planet’s magne-
tosphere cause chemical changes in the ice
and its evaporation into space by sputtering.
Although these chemical changes occur at the
molecular and even the atomic level, remote-
sensing instruments on Earth and on passing
spacecraft can detect them directly by opti-
cal (ultraviolet through infrared wavelengths)
spectroscopy and by measurements from fly-
bys high above the surface.
The origin of carbon dioxide is less clear,
and requires either that CO2 is native to
Rhea’s icy inventory, or that it forms at the
surface from the released O2 acting on car-
bon-rich grains. Such grains may be native
to Rhea or entrained in its ice, but a more
likely source is the carbonaceous microme-
COMPUTER SCIENCE
Computational Physics in Film
Robert Bridson1,2,3*and Christopher Batty1
C
omputer simulation of solid and fluid
dynamics underlies many visual
effects seen in films produced dur-
ing the past decade. This approach not only is
less expensive than filming live action but also
can avoid putting actors and crews in danger-
ous settings and can allow visualization of the
impossible. Compared with more traditional
animation methods that rely chiefly on artists’
efforts, numerical solutions to the equations
of physics allow computers to calculate realis-
tic motion, such of that of smoke, fire, explo-
sions, water, rubble, clothing, hair, muscles,
and skin. Algorithmic advances now afford
artists a higher-level, more efficient role in
guiding the physics as they produce anima-
tion. We provide an overview here of current
challenges in physics-based animation.
The movement and collisions of rigid
bodies have long been the mainstay of phys-
1
Department of Computer Science, University of Brit-
ish Columbia, 201-2366 Main Mall, Vancouver, BC
V6T 1Z4, Canada. 2Exotic Matter AB, Svardvagen 7,
182 33 Danderyd, Sweden. 3Weta Digital Ltd., 9-11
Manuka Street, Miramar, Wellington, 6022 New Zealand.
*To whom correspondence should be addressed. E-mail:
rbridson@cs.ubc.ca
1756 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org
teorites that continuously dust the solid bod-
ies of the solar system, including Earth.
In the tenuous O2 atmosphere of Rhea,
molecules rarely collide with one another,
such that the rate of escape into space
approximates the rate of ejection from the
surface. Therefore, in the current epoch,
the atmosphere is probably not increasing
appreciably in density and surface pressure.
However, Teolis et al. find that the rate of
O2 generation in the ice exceeds the rate of
ejection from it, leading to the buildup of an
oxygen reservoir. The episodic or long-term
release of this stored oxygen could increase
the total atmospheric density, but it would
still be considered tenuous.
The presence of an oxygen-rich atmo-
sphere of entirely radiolytic (photodriven) ori-
gin raises the question of using the detection
of oxygen on an extrasolar planet as a crite-
rion indicating the occurrence of life. The first
detection of an oxygen-rich atmosphere on an
extrasolar planet is likely to be accomplished
by spectroscopy, which will require a some-
ics-based animation, but modeling and inte-
grating frictional contact remains a seri-
ous challenge. Structured stacks of blocks,
highly nonconvex geometry, and delicate
balances between pressure and friction all
can pose torture tests for numerical methods
that must exactly balance forces to keep these
assemblies stable. Kaufman et al. (1) discuss
new methods that use alternating projections
(a way to calculate where interactions occur)
to solve a constrained optimization formula-
tion of contact.
Some objects, such as hair and cloth-
ing, are naturally deformable, which com-
plicates the collision problem. In hair sim-
ulation, modeling the contacts between
individual hairs creates a problem of compu-
tational scale. Resolving all of the collisions
between the 100,000 hairs on a human head
overwhelms brute-force methods. McAdams
et al. (2) have taken a multiscale approach by
treating hair as a continuum fluid, rather than
discrete strands. This approach resolves the
motion of the hair as a whole by averaging
the motion into a continuous vector field, but
truly accurate vector-field equations have yet
to be derived. Kaldor et al. (3) have taken the
opposite route in clothing simulation. Rather
Published by AAAS
what denser atmosphere than Rhea currently
has. It is notable, however, that emission of
atomic oxygen in the tenuous atmospheres of
Jupiter’s moons Europa and Ganymede was
detected by ultraviolet spectroscopy with the
Hubble Space Telescope (5) and similarly in
an extrasolar planet atmosphere (6). Addi-
tional laboratory and theoretical studies of
O2 production in ice by interaction with the
nearby space environment and the develop-
ment of a dense atmosphere should further
clarify the feasibility of using this particular
criterion, often cited as a hopeful sign of life
in a remote planetary system (7).
References
1. J. B. Dalton et al., Space Sci. Rev. 153, 113 (2010).
Downloaded from www.sciencemag.org on December 23, 2010
2. H. Campins et al., Nature 464, 1320 (2010).
3. A. Colaprete et al., Science 330, 463 (2010).
4. B. D. Teolis et al., Science 330, 1813 (2010); 10.1126/
science.1198366.
5. D. T. Hall, P. D. Feldman, M. A. McGrath, D. F. Strobel,
Astrophys. J. 499, 475 (1998).
6. A. Vidal-Madjar et al., Astrophys. J. 604, L69 (2004).
7. L. Kaltenegger, Astrophys. J. 712, L125 (2010).
10.1126/science.1200473
Numerical modeling of how objects and
fluids move, collide, and break up underlies
spellbinding video animations.
than use models that homogenize the two-
dimensional (2D) surface of clothing, they
perform a full simulation of every loop and
twist in the yarn of knitwear and create subtle
behaviors that simpler methods cannot repro-
duce. However, densely woven fabrics still
require more efficient modeling as isomet-
ric surfaces, ones that bend but do not stretch
or shear. English and Bridson (4) recently
resolved the “locking” problem plaguing ear-
lier efforts in which isometry constraints inad-
vertently prevent the natural bending. Para-
doxically, their solution involves allowing
holes to open up in the cloth between mesh tri-
angles (the numerical regions into which the
surface is decomposed). This finding poses
interesting questions in discontinuous geom-
etry, in that the mapping from surface param-
eters is neither continuous nor differentiable.
Volumetric elasticity—handling fully 3D
deformation—is used in biomechanical mod-
els of the flesh of virtual creatures (5). Stu-
dios are rapidly increasing the anatomical
detail of their models, from the complexities
of muscles and tendons to delicate wrinkles
in the skin. The amount of detail in the surface
as well as the structures underneath the skin
(muscles, tendons, bones, and other organs),
 PERSPECTIVES
the surface, sending jets of gas and dust into
space, as we have seen emanating from Sat-
urn’s moon Enceladus and numerous com-
ets. Ices exposed on the cold surfaces of outer
planetary moons interact with the local space
environment, as the incident solar ultraviolet
light and charged particles from deep space
and trapped in the parent planet’s magne-
tosphere cause chemical changes in the ice
and its evaporation into space by sputtering.
Although these chemical changes occur at the
molecular and even the atomic level, remote-
sensing instruments on Earth and on passing
spacecraft can detect them directly by opti-
cal (ultraviolet through infrared wavelengths)
spectroscopy and by measurements from fly-
bys high above the surface.
The origin of carbon dioxide is less clear,
and requires either that CO2 is native to
Rhea’s icy inventory, or that it forms at the
surface from the released O2 acting on car-
bon-rich grains. Such grains may be native
to Rhea or entrained in its ice, but a more
likely source is the carbonaceous microme-
COMPUTER SCIENCE
Computational Physics in Film
Robert Bridson1,2,3*and Christopher Batty1
C
omputer simulation of solid and fluid
dynamics underlies many visual
effects seen in films produced dur-
ing the past decade. This approach not only is
less expensive than filming live action but also
can avoid putting actors and crews in danger-
ous settings and can allow visualization of the
impossible. Compared with more traditional
animation methods that rely chiefly on artists’
efforts, numerical solutions to the equations
of physics allow computers to calculate realis-
tic motion, such of that of smoke, fire, explo-
sions, water, rubble, clothing, hair, muscles,
and skin. Algorithmic advances now afford
artists a higher-level, more efficient role in
guiding the physics as they produce anima-
tion. We provide an overview here of current
challenges in physics-based animation.
The movement and collisions of rigid
bodies have long been the mainstay of phys-
1
Department of Computer Science, University of Brit-
ish Columbia, 201-2366 Main Mall, Vancouver, BC
V6T 1Z4, Canada. 2Exotic Matter AB, Svardvagen 7,
182 33 Danderyd, Sweden. 3Weta Digital Ltd., 9-11
Manuka Street, Miramar, Wellington, 6022 New Zealand.
*To whom correspondence should be addressed. E-mail:
rbridson@cs.ubc.ca
1756 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org
teorites that continuously dust the solid bod-
ies of the solar system, including Earth.
In the tenuous O2 atmosphere of Rhea,
molecules rarely collide with one another,
such that the rate of escape into space
approximates the rate of ejection from the
surface. Therefore, in the current epoch,
the atmosphere is probably not increasing
appreciably in density and surface pressure.
However, Teolis et al. find that the rate of
O2 generation in the ice exceeds the rate of
ejection from it, leading to the buildup of an
oxygen reservoir. The episodic or long-term
release of this stored oxygen could increase
the total atmospheric density, but it would
still be considered tenuous.
The presence of an oxygen-rich atmo-
sphere of entirely radiolytic (photodriven) ori-
gin raises the question of using the detection
of oxygen on an extrasolar planet as a crite-
rion indicating the occurrence of life. The first
detection of an oxygen-rich atmosphere on an
extrasolar planet is likely to be accomplished
by spectroscopy, which will require a some-
ics-based animation, but modeling and inte-
grating frictional contact remains a seri-
ous challenge. Structured stacks of blocks,
highly nonconvex geometry, and delicate
balances between pressure and friction all
can pose torture tests for numerical methods
that must exactly balance forces to keep these
assemblies stable. Kaufman et al. (1) discuss
new methods that use alternating projections
(a way to calculate where interactions occur)
to solve a constrained optimization formula-
tion of contact.
Some objects, such as hair and cloth-
ing, are naturally deformable, which com-
plicates the collision problem. In hair sim-
ulation, modeling the contacts between
individual hairs creates a problem of compu-
tational scale. Resolving all of the collisions
between the 100,000 hairs on a human head
overwhelms brute-force methods. McAdams
et al. (2) have taken a multiscale approach by
treating hair as a continuum fluid, rather than
discrete strands. This approach resolves the
motion of the hair as a whole by averaging
the motion into a continuous vector field, but
truly accurate vector-field equations have yet
to be derived. Kaldor et al. (3) have taken the
opposite route in clothing simulation. Rather
Published by AAAS
what denser atmosphere than Rhea currently
has. It is notable, however, that emission of
atomic oxygen in the tenuous atmospheres of
Jupiter’s moons Europa and Ganymede was
detected by ultraviolet spectroscopy with the
Hubble Space Telescope (5) and similarly in
an extrasolar planet atmosphere (6). Addi-
tional laboratory and theoretical studies of
O2 production in ice by interaction with the
nearby space environment and the develop-
ment of a dense atmosphere should further
clarify the feasibility of using this particular
criterion, often cited as a hopeful sign of life
in a remote planetary system (7).
References
1. J. B. Dalton et al., Space Sci. Rev. 153, 113 (2010).
Downloaded from www.sciencemag.org on December 23, 2010
2. H. Campins et al., Nature 464, 1320 (2010).
3. A. Colaprete et al., Science 330, 463 (2010).
4. B. D. Teolis et al., Science 330, 1813 (2010); 10.1126/
science.1198366.
5. D. T. Hall, P. D. Feldman, M. A. McGrath, D. F. Strobel,
Astrophys. J. 499, 475 (1998).
6. A. Vidal-Madjar et al., Astrophys. J. 604, L69 (2004).
7. L. Kaltenegger, Astrophys. J. 712, L125 (2010).
10.1126/science.1200473
Numerical modeling of how objects and
fluids move, collide, and break up underlies
spellbinding video animations.
than use models that homogenize the two-
dimensional (2D) surface of clothing, they
perform a full simulation of every loop and
twist in the yarn of knitwear and create subtle
behaviors that simpler methods cannot repro-
duce. However, densely woven fabrics still
require more efficient modeling as isomet-
ric surfaces, ones that bend but do not stretch
or shear. English and Bridson (4) recently
resolved the “locking” problem plaguing ear-
lier efforts in which isometry constraints inad-
vertently prevent the natural bending. Para-
doxically, their solution involves allowing
holes to open up in the cloth between mesh tri-
angles (the numerical regions into which the
surface is decomposed). This finding poses
interesting questions in discontinuous geom-
etry, in that the mapping from surface param-
eters is neither continuous nor differentiable.
Volumetric elasticity—handling fully 3D
deformation—is used in biomechanical mod-
els of the flesh of virtual creatures (5). Stu-
dios are rapidly increasing the anatomical
detail of their models, from the complexities
of muscles and tendons to delicate wrinkles
in the skin. The amount of detail in the surface
as well as the structures underneath the skin
(muscles, tendons, bones, and other organs),

and the precise calculation of force response
for different materials, all contribute to the
ongoing challenge of controlling the motion
of characters while making it appear that sim-
ulated muscles are doing the work.
Objects not only move, they also break.
Research efforts in depicting fracture mechan-
ics began with O’Brien and Hodgins’s work
(6) using remeshing, which improves the geo-
metric fidelity where surfaces break by chang-
ing the mesh in that region. More recent tech-
niques embed crack geometry in finite ele-
ments models of the mechanics (7, 8). These
simulations are still difficult to control and
have yet to truly break into film production.
Some of the most spectacular examples
of physics in film involve fluids, where non-
linearities in the underlying Navier-Stokes
equations that describe fluid motion lead to
accumulation of remarkable geometric com-
plexity. A recent trend for depicting liquids
has been the emergence of mesh-based sur-
face tracking. In a preliminary attempt to fol-
low the details of a water surface as closely
as possible, Brochu et al. (9), inspired by
approaches for cloth collision processing,
developed a method that matches the degrees
of freedom in the simulation to the geometry
of the deforming surface mesh, rather than the
other way around. However, the difficulties
involved in making this approach truly robust
are still daunting; meshes in three dimensions
find endless ways to cause numerical and
combinatorial troubles.
Horvath and Geiger (10) have probably
achieved the greatest level of detail yet in fire
and smoke with a two-level approach. The
hybrid particle-grid method (fluid-implicit
particle, or FLIP) (11) provides a high-qual-
ity (albeit relatively low resolution) 3D sim-
ulation. They use it to guide extremely high-
resolution simulations on 2D slices through
the volume (oriented to the camera), running
in parallel on commodity graphics process-
ing unit (GPU) hardware. This idea of get-
ting the bulk motion from fast, low-resolution
simulations and then adding localized detail
with secondary simulation is being pursued
by many groups, although properly account-
ing for dynamics within a grid element at low
resolutions remains a major hurdle.
CREDIT: IMAGE COURTESY OF EXOTIC MATTER AB
The problem of scale in general looms
large (and small). Consider a ship on a rough
sea; the figure shows fairly convincing detail
that is achievable on a single workstation
today [a full movie clip is available, see (12); a
compressed version is available at (13)]. This
simulation still falls far short of a shot encom-
passing stormy waves to the horizon as well as
small scales down to the tiny droplets break-
ing up on the rigging. Brute-force methods
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
Realistic simulations. A simulated ship upon a simulated ocean. Here, the film industry’s Naiad software is
used to evolve the incompressible Navier-Stokes equations for the water, strongly coupled with the rigid-body
dynamics of the longboat, with additional phenomenological simulation of foam and spray.
of today would require orders of magnitude
more computing resources than are available.
The other performance challenge is
increasing speed at the simulation resolu-
tions currently in use. Studios can cope with
simulations running overnight, but such time
scales do not allow much scope for iterative
refinement. When film-quality simulations
can run at real-time rates, remarkably more
effective artist interaction is possible—not
just more design cycles but also experimen-
tation with continuous feedback. With more
general simulations emerging that fully cou-
ple all of the solid and fluid dynamics men-
tioned above, we are already beginning to see
the advent of “virtual practical effects.” Art-
ists can use their natural intuition about how
physics works to build virtual devices to con-
trol the virtual world of a shot, rather than awk-
wardly manipulating parameters in equations.
The quality of physics-based animation
methods is not simple to evaluate. Ultimately,
success is judged by the director, and the worth
of the underlying algorithms can be judged by
the users based on how well it helps them do
their jobs. However, given the time and effort
involved in using a new method in production,
and the difficulty of creating objective, quan-
titative metrics, researchers need something
better to analyze their efforts.
Classic means of evaluating algorithms,
such as convergence rates for iterative
schemes, have limited use. Qualitatively good
results (ones that convince audiences) are usu-
ally obtained well before a model simulation
converges. Psychophysical results in video
compression, for example, also suggest that
the mathematically convenient norms com-
monly used in numerical analysis are not a
Published by AAAS
PERSPECTIVES
Downloaded from www.sciencemag.org on December 23, 2010
good match for human-oriented evaluation.
Moreover, assessing the errors in underlying
models on which algorithms are based is a
particular challenge for film.
Unlike traditional science, reality does
not necessarily provide a ground truth against
which film models can be compared. Film
production on a set already demands some-
thing enhanced beyond reality. Tackling the
problem of objective and useful evaluation
will likely demand cross-disciplinary efforts
in understanding the human perception of
complex dynamics. In the short term, success
can continue to be judged by increasing calls
for “lights, camera, simulation.”
References and Notes
1. D. M. Kaufman, S. Sueda, D. L. James, D. K. Pai, ACM
Trans. Graph. 27, 10.1145/1457515.1409117 (2008).
2. A. McAdams, A. Selle, K. Ward, E. Sifakis, J. Teran, ACM
Trans. Graph. H 28, 10.1145/1531326.1531368 (2009).
3. J. M. Kaldor, D. L. James, S. Marschner, ACM Trans. Trans.
Graph 29, 10.1145/1778765.1778842 (2010).
4. E. English, R. Bridson, A. C. M. Trans. Graph. 27,
10.1145/1360612.1360665 (2008).
5. S.-H. Lee, E. Sifakis, D. Terzopoulos, ACM Trans. Graph.
28, 10.1145/1559755.1559756 (2009).
6. J. O’Brien, J. Hodgins, in Proc. SIGGRAPH, 18 (ACM, New
York, 1999), pp. 137–146.
7. N. Molino, Z. Bao, R. Fedkiw, ACM Trans. Graph. 23, 385
(2004).
8. P. Kaufmann, S. Martin, M. Botsch, E. Grinspun, M. Gross,
ACM Trans. Graph. 28, 10.1145/1531326.1531356
(2009).
9. T. Brochu, C. Batty, R. Bridson, ACM Trans. Graph. 29,
10.1145/1778765.1778784 (2010).
10. C. Horvath, W. Geiger, ACM Trans. Graph. 28,
10.1145/1531326.1531347 (2009).
11. Y. Zhu, R. Bridson, ACM Trans. Graph. 24, 965 (2005).
12. www.youtube.com/watch?v=3pojNrRfJFM
13. http://people.cs.ubc.ca/~rbridson/movies/NaiadLongboat.
mov
14. Supported in part by a grant from the Natural Sciences
and Engineering Research Council of Canada.
10.1126/science.1198769
1757
 PERSPECTIVES

GENETICS

Revealing the Dark Matter Integrated data sets from two animal model
organisms provide insights into the organization,

of the Genome
structure, and function of their genomes.

Mark Blaxter

A
nimal embryos successfully trans-
form the two-dimensional code
of their genome into multidimen-
sional organisms that are ready to meet
the challenge of natural selection. In addi-
tion to the three dimensions of the body,
by embryos, but two articles in this issue,
by Gerstein et al. on page 1775 (1) and the
modENCODE Consortium on page 1787
(2), bring this goal closer.
The genomes of multicellular animals
are big and complex, but functions have
expression posttranslationally, and marking
of the histone proteins on which the DNA is
wound with chemical tags to define regions
of the genome that are active or silent.
One could analyze this complex regula-
tory landscape one factor or region at a time,

Downloaded from www.sciencemag.org on December 23, 2010

animal genomes inform additional dimen-
sions: of cells coordinating to form tissues,
been defined for only a small proportion
of them. Only 1% of the human genome
but this would miss the big picture. The
ENCODE (Encyclopedia of DNA Elements)
projects are using large-scale,
genome-wide assays to identify the
interactions among transcription
factors, ncRNA, chromatin marks,
and gene expression—seeking
functions for the dark genome. Ini-
tial data from the human ENCODE
project (3) revealed an incredible
density of regulatory marks and
interactions on a small portion
(1%) of the human genome. The
modENCODE (model organism

CREDITS: DNA, Y. HAMMOND/SCIENCE; FRUIT FLY, TOMASZ ZACHARIASZ/ISTOCKPHOTO; C. ELEGANS, BOB GOLDSTEIN/WIKIMEDIA COMMONS
Encyclopedia of DNA Elements)
strand of the project is using the
power of model organism genom-
ics to reveal genome-wide patterns
of regulatory interactions (1, 2).
Model organisms, such as the
fruit fly Drosophila melanogaster
and the nematode Caenorhabditis
elegans, are chosen for many rea-
sons, including ease of cell culture
and amenability to experimenta-
tion. In the era of genome science,
one of their key benefits is their
small genomes: 100 million bases
Computing the organism. Integrated datasets across transcription, epigenome, and protein-DNA interactions describe (Mb) for C. elegans (5) and 180 Mb
the dynamic regulation of gene expression in the nematode and fly model organisms. for D. melanogaster (6) [compared
to the 3000-Mb human genome (7,
tissues functioning together as organs, and is transcribed into protein-coding messen- 8)]—and a much larger proportion of their
organs shaping the body’s systems; and of ger RNA (mRNA) and non–protein-coding genomes shows signatures of evolutionary
individuals responding appropriately to the RNA (ncRNA), and DNA elements that con- constraint. Both models have been examined
varied challenges of life and surviving to trol the expression of genes occupy another with a huge armory of genetic and molecu-
breed. Poisons in food are detoxified, patho- ~0.5%, suggesting that the remaining “dark lar tools, and our understanding of how their
gens are killed, parasites are eliminated, and genome” is nonfunctional padding. How- embryos develop, and how the adult organ-
predators avoided through the deft employ- ever, 5% of the human and other mammalian isms function, exceeds that for any other ani-
ment of responses encoded in the genome. genomes are under evolutionary constraint, mal. Their role in modENCODE is to pilot
It is not currently possible to compute an suggesting biological functions (3, 4). What technologies, especially those of data anal-
organism from its genome, performing are these functions and how are they inte- ysis, and to provide reference points for the
the transformation so efficiently executed grated? Three interacting systems coordi- emerging human ENCODE data. Fruit fly
nate gene expression in space and time: tran- and nematode modENCODE projects have
Institute of Evolutionary Biology, University of Edinburgh, scription factors that bind to DNA in pro- performed hundreds of experiments and pro-
Edinburgh EH9 3JT, UK. E-mail: mark.blaxter@ed.ac.uk moters of genes, ncRNA that modifies gene duced billions of data points to permit the

1758 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

Published by AAAS

building of new models of gene expression
regulation. These can be used to describe the
idiosyncratic development and biology of
each animal, but the excitement lies in the
commonalities in the overall structures of the
regulatory landscape they reveal.
Before modeling gene expression pat-
terns, one first has to know what genes are
present. Despite the deep annotation avail-
able for the fruit fly and nematode genomes,
both projects have identified many new
genes and parts of genes. In C. elegans, Ger-
stein et al. found evidence supporting 95%
of existing protein-coding gene predictions,
but 1650 new genes are now added for a total
of ~22,000. The number of RNA transcripts
identified is now triple the previous estimate
(to about three per gene), and the ncRNA
set is greater by a factor of 20. Evidence
obtained by the modENCODE Consortium
increases the D. melanogaster gene set by a
similar amount, to ~17,000 distinct genes.
Both reports suggest that the gene catalogs
for the two model organisms may now be
essentially complete, with new discoveries
inherently indistinguishable from biological
noise and likely to be unimportant.
The model animals have a similar com-
plexity of patterns of chromatin marks and
their correlations with gene expression. Using
data for 18 different histone modifications
in a D. melanogaster cell line, the modEN-
CODE Consortium identified 30 chromatin
states that are associated with different gene
expression patterns and gene positions. C.
elegans lacks some chromatin states found in
D. melanogaster (such as heterochromatin—
repressed DNA that makes up ~30% of the
D. melanogaster genome), but Gerstein et al.
found that chromatin states are similarly pre-
dictive of nematode gene expression patterns,
including expression of ncRNAs.
A third shared discovery of the modEN-
CODE teams is the very high degree of con-
nectivity, and beguiling simplicity, in the reg-
ulatory systems (9). Regulators (transcription
factors and ncRNAs) function in hierarchies
with few levels, in which master regulators
control many other regulators. These, in turn,
feed back in a set of simple network connec-
tion motifs involving ncRNAs. In D. melano-
gaster each regulator is, on average, only two
(and no more than five) links away from any
other. Predictive models of gene expression
based on their regulatory interactions were
built and tested against observed gene expres-
sion patterns. In developing D. melanogas-
ter embryos, the model predicted 62% of the
expression patterns observed in isolated cell
lines. Given the stochasticity of biological
systems, this is a remarkable achievement.
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
Both Gerstein et al. and the modEN-
CODE Consortium report a curious class of
short (100-base) elements in the genomes
called highly occupied target (HOT) regions
(10, 11). HOT regions were repeatedly iden-
tified as binding many different transcription
factors, but are curiously not enriched in the
known DNA motifs to which these factors
bind, suggesting that the interactions may be
indirect. HOT regions are stable and asso-
ciate with gene transcription start sites, and
in C. elegans they are associated with genes
that are universally expressed through devel-
opment at high levels. In D. melanogaster,
HOT sites are also sites of binding by pro-
teins involved in originating DNA replica-
tion. Both studies identified novel sequence
motifs that are enriched in HOT regions, but
these motifs are not shared between the two
species and most do not match known tran-
scription factor binding sites, suggesting
that the proteins that bind to these motifs are
yet to be identified.
The Large Hadron Collider is the preem-
inent, long-term cooperative enterprise in
the physical sciences, dedicated to gathering
MATERIALS SCIENCE
Stretching Dielectric
Elastomer Performance
Federico Carpi,1 Siegfried Bauer,2 Danilo De Rossi1
Devices using materials that deform in response to electricity are based on a phenomenon that was
observed more than two centuries ago.
T
he idea that a solid material can
deform when stimulated by electric-
ity originated in the late-18th cen-
tury with observations of ruptures in over-
charged Leyden jars, the first electrical
capacitors. In 1776, Italian scientist Ales-
sandro Volta mentioned in a letter that Ital-
ian experimenter Felice Fontana had noted
volume changes in the Leyden jar upon elec-
trification (1), an observation that launched
a new field of investigation—“deformable”
materials affected by electricity. More than
two centuries later, the concept of “electri-
cally stretchable materials” is at the fore-
front of devising bioinspired robots, tactile
1
Interdepartmental Research Centre “E. Piaggio,” School of
Engineering, University of Pisa, Pisa 56100, Italy. 2Depart-
ment of Soft Matter Physics, Johannes Kepler University, Linz
A-4040, Austria. E-mail: f.carpi@centropiaggio.unipi.it
Published by AAAS
PERSPECTIVES
data to fully parameterize the basic physi-
cal constants of the universe and understand
dark matter. In the same way, the modEN-
CODE and ENCODE deep genomics pro-
grams will, in time, deliver the power to
model and predict organism function from
multidimensional data, shine light on the
dark genome, and hopefully allow a bet-
ter understanding of the healthy human and
how to treat human disease.
References
1. M. B. Gerstein et al., Science 330, 1775 (2010).
2. modENCODE Consortium et al., Science 330, 1787 (2010).
3. ENCODE Project Consortium, Nature 447, 799 (2007).
4. L. Eory, D. L. Halligan, P. D. Keightley, Mol. Biol. Evol.
27, 177 (2010).
Downloaded from www.sciencemag.org on December 23, 2010
5. C. elegans Sequencing Consortium, Science 282, 2012
(1998).
6. M. D. Adams et al., Science 287, 2185 (2000).
7. J. C. Venter et al., Science 291, 1304 (2001).
8. International Human Genome Sequencing Consortium,
Nature 409, 860 (2001).
9. U. Alon, Nat. Rev. Genet. 8, 450 (2007).
10. S. MacArthur et al., Genome Biol. 10, R80 (2009).
11. C. Moorman et al., Proc. Natl. Acad. Sci. U.S.A. 103,
12027 (2006).
Published online 22 December 2010;
10.1126/science.1200700
and haptic interfaces, and adaptive optical
systems (2, 3).
This diversity of applications took a
great leap 10 years ago in a landmark study
by Pelrine and colleagues (4). They reported
high-speed, giant-strain, electrically actu-
ated elastomers with unprecedented elec-
tromechanical transduction performance.
These materials were demonstrated for
so-called dielectric elastomer actuators,
deformable capacitors made of a film of
a soft insulator (such as acrylic, polyure-
thane, or silicone elastomer), with compli-
ant electrodes. Upon electrical charging,
purely electrostatic forces caused the elas-
tomer film to undergo substantial thickness
compression and surface expansion (4). The
exceptional performance of these dielectric
elastomer actuators gave rise to a scientific
and technological revolution in the field of
1759

building of new models of gene expression
regulation. These can be used to describe the
idiosyncratic development and biology of
each animal, but the excitement lies in the
commonalities in the overall structures of the
regulatory landscape they reveal.
Before modeling gene expression pat-
terns, one first has to know what genes are
present. Despite the deep annotation avail-
able for the fruit fly and nematode genomes,
both projects have identified many new
genes and parts of genes. In C. elegans, Ger-
stein et al. found evidence supporting 95%
of existing protein-coding gene predictions,
but 1650 new genes are now added for a total
of ~22,000. The number of RNA transcripts
identified is now triple the previous estimate
(to about three per gene), and the ncRNA
set is greater by a factor of 20. Evidence
obtained by the modENCODE Consortium
increases the D. melanogaster gene set by a
similar amount, to ~17,000 distinct genes.
Both reports suggest that the gene catalogs
for the two model organisms may now be
essentially complete, with new discoveries
inherently indistinguishable from biological
noise and likely to be unimportant.
The model animals have a similar com-
plexity of patterns of chromatin marks and
their correlations with gene expression. Using
data for 18 different histone modifications
in a D. melanogaster cell line, the modEN-
CODE Consortium identified 30 chromatin
states that are associated with different gene
expression patterns and gene positions. C.
elegans lacks some chromatin states found in
D. melanogaster (such as heterochromatin—
repressed DNA that makes up ~30% of the
D. melanogaster genome), but Gerstein et al.
found that chromatin states are similarly pre-
dictive of nematode gene expression patterns,
including expression of ncRNAs.
A third shared discovery of the modEN-
CODE teams is the very high degree of con-
nectivity, and beguiling simplicity, in the reg-
ulatory systems (9). Regulators (transcription
factors and ncRNAs) function in hierarchies
with few levels, in which master regulators
control many other regulators. These, in turn,
feed back in a set of simple network connec-
tion motifs involving ncRNAs. In D. melano-
gaster each regulator is, on average, only two
(and no more than five) links away from any
other. Predictive models of gene expression
based on their regulatory interactions were
built and tested against observed gene expres-
sion patterns. In developing D. melanogas-
ter embryos, the model predicted 62% of the
expression patterns observed in isolated cell
lines. Given the stochasticity of biological
systems, this is a remarkable achievement.
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
Both Gerstein et al. and the modEN-
CODE Consortium report a curious class of
short (100-base) elements in the genomes
called highly occupied target (HOT) regions
(10, 11). HOT regions were repeatedly iden-
tified as binding many different transcription
factors, but are curiously not enriched in the
known DNA motifs to which these factors
bind, suggesting that the interactions may be
indirect. HOT regions are stable and asso-
ciate with gene transcription start sites, and
in C. elegans they are associated with genes
that are universally expressed through devel-
opment at high levels. In D. melanogaster,
HOT sites are also sites of binding by pro-
teins involved in originating DNA replica-
tion. Both studies identified novel sequence
motifs that are enriched in HOT regions, but
these motifs are not shared between the two
species and most do not match known tran-
scription factor binding sites, suggesting
that the proteins that bind to these motifs are
yet to be identified.
The Large Hadron Collider is the preem-
inent, long-term cooperative enterprise in
the physical sciences, dedicated to gathering
MATERIALS SCIENCE
Stretching Dielectric
Elastomer Performance
Federico Carpi,1 Siegfried Bauer,2 Danilo De Rossi1
Devices using materials that deform in response to electricity are based on a phenomenon that was
observed more than two centuries ago.
T
he idea that a solid material can
deform when stimulated by electric-
ity originated in the late-18th cen-
tury with observations of ruptures in over-
charged Leyden jars, the first electrical
capacitors. In 1776, Italian scientist Ales-
sandro Volta mentioned in a letter that Ital-
ian experimenter Felice Fontana had noted
volume changes in the Leyden jar upon elec-
trification (1), an observation that launched
a new field of investigation—“deformable”
materials affected by electricity. More than
two centuries later, the concept of “electri-
cally stretchable materials” is at the fore-
front of devising bioinspired robots, tactile
1
Interdepartmental Research Centre “E. Piaggio,” School of
Engineering, University of Pisa, Pisa 56100, Italy. 2Depart-
ment of Soft Matter Physics, Johannes Kepler University, Linz
A-4040, Austria. E-mail: f.carpi@centropiaggio.unipi.it
Published by AAAS
PERSPECTIVES
data to fully parameterize the basic physi-
cal constants of the universe and understand
dark matter. In the same way, the modEN-
CODE and ENCODE deep genomics pro-
grams will, in time, deliver the power to
model and predict organism function from
multidimensional data, shine light on the
dark genome, and hopefully allow a bet-
ter understanding of the healthy human and
how to treat human disease.
References
1. M. B. Gerstein et al., Science 330, 1775 (2010).
2. modENCODE Consortium et al., Science 330, 1787 (2010).
3. ENCODE Project Consortium, Nature 447, 799 (2007).
4. L. Eory, D. L. Halligan, P. D. Keightley, Mol. Biol. Evol.
27, 177 (2010).
Downloaded from www.sciencemag.org on December 23, 2010
5. C. elegans Sequencing Consortium, Science 282, 2012
(1998).
6. M. D. Adams et al., Science 287, 2185 (2000).
7. J. C. Venter et al., Science 291, 1304 (2001).
8. International Human Genome Sequencing Consortium,
Nature 409, 860 (2001).
9. U. Alon, Nat. Rev. Genet. 8, 450 (2007).
10. S. MacArthur et al., Genome Biol. 10, R80 (2009).
11. C. Moorman et al., Proc. Natl. Acad. Sci. U.S.A. 103,
12027 (2006).
Published online 22 December 2010;
10.1126/science.1200700
and haptic interfaces, and adaptive optical
systems (2, 3).
This diversity of applications took a
great leap 10 years ago in a landmark study
by Pelrine and colleagues (4). They reported
high-speed, giant-strain, electrically actu-
ated elastomers with unprecedented elec-
tromechanical transduction performance.
These materials were demonstrated for
so-called dielectric elastomer actuators,
deformable capacitors made of a film of
a soft insulator (such as acrylic, polyure-
thane, or silicone elastomer), with compli-
ant electrodes. Upon electrical charging,
purely electrostatic forces caused the elas-
tomer film to undergo substantial thickness
compression and surface expansion (4). The
exceptional performance of these dielectric
elastomer actuators gave rise to a scientific
and technological revolution in the field of
1759
 PERSPECTIVES
electroactive polymers, materials that can
undergo electrically induced deformations.
Since that milestone study, elastomers with
improved electromechanical properties have
been developed, such as interpenetrating
networks of acrylic polymers (3).
The most widely recognized potential of
dielectric elastomer actuators is for creating
artificial muscles (2, 3). Indeed, electrically
driven elastomers have already exceeded
the performance of natural muscles in
terms of strain (up to 380% in area), stress
(up to 7.2 MPa), and elastic energy density
(up to 3.4 J cm–3) (3). Moreover, they show
fast response and long lifetime; have high
New frontiers. Future applications of electrically charged dielectric elastomers
include replacing traditional Braille printed pages with refreshable Braille dis-
plays, using millimeter-sized actuators that control the tactile pattern of dots.
resilience; are light weight, scalable, shock-
tolerant, and noise- and heat-free; and are
inexpensive (2–5).
Possible future applications of dielectric
elastomer actuators that have been under
development over the past decade, and seem
to be promising, deal with haptics and optics.
For example, they are expected to be used
in electronic smart devices such as mobile
phones to provide users with vibro-tactile
feedback, transmitting clicks and vibra-
tions through the sense of touch. Smooth
touch screens do not provide key sensorial
experience (see the photo, above). Dielec-
tric elastomer actuators may deliver feed-
back that integrates visual, audio, and tactile
responses. Adaptive optical devices, such as
lenses and diffraction gratings, are another
area in an advanced stage of development
(6). Lenses with electrically tunable focal
length are potentially useful in autofocus
cameras. Miniaturization of dielectric elas-
tomer actuators is a new frontier of research
(2, 3, 7) in which millimeter-sized actuators
could, for example, control the tactile pattern
1760 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org
of Braille interfaces, allowing for simple and
compact electrically refreshable tactile dis-
plays (2, 3).
Key early investigations on electri-
cally induced deformations of insulators
include studies in 1775 by the French sci-
entist Nicolas-Phillipe Ledru (also known as
Comus), who observed the rise and fall of
mercury within glass tubes upon electrifica-
tion (8). In 1776, Volta explained Fontana’s
observations on the Leyden jar: “The glass
is strongly compressed…by the two arma-
tures, i.e., exterior metallic leaf, and interior
water, which…armatures weight, I will say
so, one against the other, because they are
Touch screens that are used in smart devices and laptop computers that lack
tactile feedback could have this new dimension. Dielectric elastomer actuators
could produce effects that enhance the user’s experience with such devices.
oppositely electric...So...it behaves alike,
both when it is interiorly charged by excess,
and when it is charged by defect” (1). Volta
was the first to interpret the observations as
electrostatically induced deformations of
the solid dielectric, independent of the volt-
age polarity. Today, this appears as a sort of
unaware anticipation of the physical princi-
ple that underpins dielectric elastomer actu-
ation, introduced two centuries later. More-
over, electromechanical effects experienced
in those early capacitors that were filled with
conductive liquids have recently inspired
contractile “hydrostatically coupled” dielec-
tric elastomer actuators (9), which exactly
follow Volta’s explanation. Here, an incom-
pressible liquid confined between elasto-
mer membranes allows for electrically safe
transmission of forces to loads.
In 1880 German physicist Wilhelm Con-
rad Röntgen reported that a “caoutchouc
stripe...[is] electrified by...an isolated comb
of needles...connected... to...a strong Holtz
influence machine...As...the caoutchouc
becomes electrified...one observes a con-
Published by AAAS
tinuous increase of the length of the band”
(10). Today, this experiment is a milestone
in the historical background of dielectric
elastomer actuation, as the first example of
a charge-controlled electrode-free actua-
tor. Indeed, this result was reproduced in a
study this year (11) using the acrylic elasto-
mer tested by Pelrine et al. A process called
corona charging was used, in which a piece
of elastomer is electrified by ionizing the
surrounding medium with high voltages.
Corona-charged electrified actuators per-
formed better than those controlled directly
by electrodes. This is because the former
overcome the elastomer “pull-in” insta-
Downloaded from www.sciencemag.org on December 23, 2010
bility—that is, the mechanical collapse of
the elastomer when the electrostatic force
exceeds the elastic force.
At present, the greatest challenge for
dielectric elastomer actuation is reduc-
ing the driving voltages (on the order of
1 kV for electrode separations of 10 to 100
um). To this end, developing high–dielectric
constant elastomers and processing them
as thin films is strategic (2, 3). There are
CREDIT: (LEFT) COMSTOCK; (RIGHT) NICOLAS LORAN/ISTOCKPHOTO
promising advances in boosting the dielec-
tric constant of a material while preserving
low mechanical stiffness and high dielectric
strength (12). Reaching voltages compara-
ble to those of piezoelectrics (on the order
of 100 V) may be feasible through thin-film
processing in a few years. This would allow
the dielectric elastomer actuator technology
to permeate an enormous variety of prod-
ucts in haptic, automation, robotic, fluidic,
biomedical, optical, and acoustic systems
(2, 3).
Fault tolerance is increased with “self-
healing” electrodes; on this front, conduc-
tive polyaniline nanofiber or carbon nano-

tube electrodes, wherein dielectric break-
down results in isolated local burns, show
promise (3). Ion-implanted electrodes (7)
are paving the way to dielectric elastomer-
based microelectromechanical systems and
microfluidic devices, such as electrically
controlled valves (3).
The use of dielectric elastomer transduc-
tion technology in a reverse mode to convert
mechanical energy into electrical energy
(especially from renewable sources, such as
offshore ocean waves) is one of the latest
frontiers. The voltage is increased when the
force of a stretched and charged elastomer
is reduced (2). Here, high-voltage operation
is of utmost advantage for energy distribu-
tion. While experimental assessments are in
progress (13), theoretical estimates antici-
pate energy densities around 6 J/g ( 14).
CANCER
Germ Cell Genes and Cancer
Xiaoyun Wu1,2 and Gary Ruvkun1,2
C
ancer cells and germ cells share sev-
eral characteristics. For instance,
both have the ability to rapidly pro-
liferate, typically do not lose the ability to
divide as they age (lack senescence), and
exist in undifferentiated states. Although
some genes involved in cancer may initiate
disease simply by activating the cell divi-
sion cycle, others may spur tumors by acti-
vating early developmental pathways asso-
ciated with programming for multipotency
(the ability to differentiate into different cell
types). On page 1824 of this issue, Janic
et al. (1) show that in fruit flies a number of
genes typically involved in early program-
ming of germline cells are also involved in
the formation of one type of malignant brain
tumor. They also show that inactivating these
germ cell genes—some of which have related
genes shown to be abnormally expressed in
certain human cancers—can suppress tumor
growth, suggesting new avenues for therapy.
Janic et al. studied brain tumors of the fruit
fly Drosophila melanogaster that are caused
by a mutation in the tumor suppressor gene
lethal (3) malignant brain tumor [l(3)mbt]
(see the figure). Gene expression profiles of
tumor and normal brain cells revealed that
1
Department of Molecular Biology, Massachusetts General
Hospital, Boston, MA 02114, USA. 2Department of Genetics,
Harvard Medical School, Boston, MA 02115, USA. E-mail:
wux@molbio.mgh.harvard.edu; ruvkun@molbio.mgh.
harvard.edu
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
Current challenges include the develop-
ment of optimized cycles to maximize elec-
tromechanical efficiency at any strain.
The field of dielectric elastomer trans-
ducers is rapidly maturing and broadening,
and the limits of their applications surely
will be stretched. The question is whether
future applications will be enabled by the
two key factors that have thus far prompted
their vast and diverse impacts: a simple and
reliable physical principle, and the possibil-
ity of effective implementation with inex-
pensive and off-the-shelf materials.
References
1. G. I. Montanari, Ed., Lettere inedite di Alessandro Volta,
(Stamperia Nobili, Pesaro, Italy, 1835), pp. 15–17.
2. F. Carpi, D. De Rossi, R. Kornbluh, R. Pelrine, P. Sommer-
Larsen, Eds, Dielectric Elastomers as Electromechanical
Transducers (Elsevier, Amsterdam, 2008).
3. P. Brochu, Q. Pei, Macromol. Rapid Commun. 31, 10
one-fourth of the genes that had increased
activity in the tumors (up-regulated) have
germline-associated functions. Among these
genes are two (vasa and nanos) involved in a
germ cell organelle called the nuage, or polar
granule, and four (spn-E, piwi, aubergine, and
ago3) that express proteins in the Piwi-inter-
acting RNA (piRNA) pathway.
The germline organelles have been
intensively studied. Drosophila develop-
mental geneticists first discovered their role
as couriers of maternally contributed germ-
line specification and early patterning sig-
Out of place? In normal fruit fly brain cells (left, blue), vasa, a gene normally associated with germ cell
organelle function, is not active. Flies with a mutation in the tumor suppressor gene l(3)mbt develop cancer-
ous brain tumors, and vasa shows activity in tumor cells (right, red). The expression of vasa and other germ
cell genes appear to be essential for tumor growth, and inactivating these genes can suppress tumors.
Published by AAAS
PERSPECTIVES
(2010).
4. R. E. Pelrine, R. D. Kornbluh, Q. Pei, J. P. Joseph, Science
287, 836 (2000).
5. G. Kovacs, L. Düring, S. Michel, G. Terrasi, Sens.
Actuators A Phys. 155, 299 (2009).
6. M. Aschwanden, A. Stemmer, Opt. Lett. 31, 2610
(2006).
7. S. Rosset, M. Niklaus, P. Dubois, H. Shea, Adv. Funct.
Mater. 19, 470 (2009).
8. Scelta di opuscoli interessanti, Nuova edizione, Tomo II
(Stamperia di G. Galeazzi, Milano, 35, 1782), p. 33.
9. F. Carpi, G. Frediani, D. De Rossi, IEEE ASME Trans.
Mech. 15, 308 (2010).
10. W. C. Röntgen, Ann. Phys. Chem. 247, 771 (1880).
11. C. Keplinger, M. Kaltenbrunner, N. Arnold, S. Bauer,
Proc. Natl. Acad. Sci. U.S.A. 107, 4505 (2010).
12. H. Stoyanov, M. Kollosche, D. N. McCarthy, G. Kofod,
J. Mater. Chem. 20, 7558 (2010).
13. S. Chiba, M. Waki, R. Kornbluh, R. Pelrine, Proc. SPIE
6927, 692715 (2008).
Downloaded from www.sciencemag.org on December 23, 2010
14. S. J. A. Koh, X. Zhao, Z. Suo, Appl. Phys. Lett. 94,
262902 (2009).
10.1126/science.1194773
In fruit flies, genes that help to program germ
cells also play a role in a brain cancer.
nals to the developing embryo (2). Then,
over the past decade, small RNA biologists
discovered that some of the same genes
in the polar granule pathway also mediate
small RNA control of transposons (mov-
able segments of DNA) (3). Germline cells
are more active than other cells in small
RNA surveillance, perhaps to protect the
genome from damage by transposons and
viruses, or perhaps also to carry out small
RNA–mediated regulation of maternally
encoded mRNAs and the chromatin-reg-
ulated deshrouding of the zygotic genome
1761

tube electrodes, wherein dielectric break-
down results in isolated local burns, show
promise (3). Ion-implanted electrodes (7)
are paving the way to dielectric elastomer-
based microelectromechanical systems and
microfluidic devices, such as electrically
controlled valves (3).
The use of dielectric elastomer transduc-
tion technology in a reverse mode to convert
mechanical energy into electrical energy
(especially from renewable sources, such as
offshore ocean waves) is one of the latest
frontiers. The voltage is increased when the
force of a stretched and charged elastomer
is reduced (2). Here, high-voltage operation
is of utmost advantage for energy distribu-
tion. While experimental assessments are in
progress (13), theoretical estimates antici-
pate energy densities around 6 J/g ( 14).
CANCER
Germ Cell Genes and Cancer
Xiaoyun Wu1,2 and Gary Ruvkun1,2
C
ancer cells and germ cells share sev-
eral characteristics. For instance,
both have the ability to rapidly pro-
liferate, typically do not lose the ability to
divide as they age (lack senescence), and
exist in undifferentiated states. Although
some genes involved in cancer may initiate
disease simply by activating the cell divi-
sion cycle, others may spur tumors by acti-
vating early developmental pathways asso-
ciated with programming for multipotency
(the ability to differentiate into different cell
types). On page 1824 of this issue, Janic
et al. (1) show that in fruit flies a number of
genes typically involved in early program-
ming of germline cells are also involved in
the formation of one type of malignant brain
tumor. They also show that inactivating these
germ cell genes—some of which have related
genes shown to be abnormally expressed in
certain human cancers—can suppress tumor
growth, suggesting new avenues for therapy.
Janic et al. studied brain tumors of the fruit
fly Drosophila melanogaster that are caused
by a mutation in the tumor suppressor gene
lethal (3) malignant brain tumor [l(3)mbt]
(see the figure). Gene expression profiles of
tumor and normal brain cells revealed that
1
Department of Molecular Biology, Massachusetts General
Hospital, Boston, MA 02114, USA. 2Department of Genetics,
Harvard Medical School, Boston, MA 02115, USA. E-mail:
wux@molbio.mgh.harvard.edu; ruvkun@molbio.mgh.
harvard.edu
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
Current challenges include the develop-
ment of optimized cycles to maximize elec-
tromechanical efficiency at any strain.
The field of dielectric elastomer trans-
ducers is rapidly maturing and broadening,
and the limits of their applications surely
will be stretched. The question is whether
future applications will be enabled by the
two key factors that have thus far prompted
their vast and diverse impacts: a simple and
reliable physical principle, and the possibil-
ity of effective implementation with inex-
pensive and off-the-shelf materials.
References
1. G. I. Montanari, Ed., Lettere inedite di Alessandro Volta,
(Stamperia Nobili, Pesaro, Italy, 1835), pp. 15–17.
2. F. Carpi, D. De Rossi, R. Kornbluh, R. Pelrine, P. Sommer-
Larsen, Eds, Dielectric Elastomers as Electromechanical
Transducers (Elsevier, Amsterdam, 2008).
3. P. Brochu, Q. Pei, Macromol. Rapid Commun. 31, 10
one-fourth of the genes that had increased
activity in the tumors (up-regulated) have
germline-associated functions. Among these
genes are two (vasa and nanos) involved in a
germ cell organelle called the nuage, or polar
granule, and four (spn-E, piwi, aubergine, and
ago3) that express proteins in the Piwi-inter-
acting RNA (piRNA) pathway.
The germline organelles have been
intensively studied. Drosophila develop-
mental geneticists first discovered their role
as couriers of maternally contributed germ-
line specification and early patterning sig-
Out of place? In normal fruit fly brain cells (left, blue), vasa, a gene normally associated with germ cell
organelle function, is not active. Flies with a mutation in the tumor suppressor gene l(3)mbt develop cancer-
ous brain tumors, and vasa shows activity in tumor cells (right, red). The expression of vasa and other germ
cell genes appear to be essential for tumor growth, and inactivating these genes can suppress tumors.
Published by AAAS
PERSPECTIVES
(2010).
4. R. E. Pelrine, R. D. Kornbluh, Q. Pei, J. P. Joseph, Science
287, 836 (2000).
5. G. Kovacs, L. Düring, S. Michel, G. Terrasi, Sens.
Actuators A Phys. 155, 299 (2009).
6. M. Aschwanden, A. Stemmer, Opt. Lett. 31, 2610
(2006).
7. S. Rosset, M. Niklaus, P. Dubois, H. Shea, Adv. Funct.
Mater. 19, 470 (2009).
8. Scelta di opuscoli interessanti, Nuova edizione, Tomo II
(Stamperia di G. Galeazzi, Milano, 35, 1782), p. 33.
9. F. Carpi, G. Frediani, D. De Rossi, IEEE ASME Trans.
Mech. 15, 308 (2010).
10. W. C. Röntgen, Ann. Phys. Chem. 247, 771 (1880).
11. C. Keplinger, M. Kaltenbrunner, N. Arnold, S. Bauer,
Proc. Natl. Acad. Sci. U.S.A. 107, 4505 (2010).
12. H. Stoyanov, M. Kollosche, D. N. McCarthy, G. Kofod,
J. Mater. Chem. 20, 7558 (2010).
13. S. Chiba, M. Waki, R. Kornbluh, R. Pelrine, Proc. SPIE
6927, 692715 (2008).
Downloaded from www.sciencemag.org on December 23, 2010
14. S. J. A. Koh, X. Zhao, Z. Suo, Appl. Phys. Lett. 94,
262902 (2009).
10.1126/science.1194773
In fruit flies, genes that help to program germ
cells also play a role in a brain cancer.
nals to the developing embryo (2). Then,
over the past decade, small RNA biologists
discovered that some of the same genes
in the polar granule pathway also mediate
small RNA control of transposons (mov-
able segments of DNA) (3). Germline cells
are more active than other cells in small
RNA surveillance, perhaps to protect the
genome from damage by transposons and
viruses, or perhaps also to carry out small
RNA–mediated regulation of maternally
encoded mRNAs and the chromatin-reg-
ulated deshrouding of the zygotic genome
1761
 PERSPECTIVES
in the early embryo. Consistent with these
roles, Janic et al. found that many of the
small RNAs (such as micro-RNAs and
piRNAs) that accumulate at high levels
in fruit fly brain tumors are also highly
expressed in normal ovaries. Therefore,
tumor cells deficient in l(3)mbt “reanimate”
multiple germ cell characteristics.
This unusual, “ectopic” expression of
germ cell genes in somatic tumor cells may
constitute a pathway for multipotent cell
specification, perhaps leading to tumor-
ous attributes. Indeed, Janic et al. show that
some of the ectopically expressed genes
are essential for tumor growth. Loss of the
germ plasm component genes vasa and
nanos or of the piRNA pathway genes piwi
and aubergine substantially reduced l(3)
mbt tumor mass or almost completely pre-
vented tumor growth.
What triggers the reanimation of germ
cell characteristics in these brain tumors?
Ectopic expression of germline genes was
not associated with malignant Drosoph-
ila brain tumors caused by other genetic
lesions (1), and thus are unique to the l(3)
mbt-derived tumors. L(3)MBT is an MBT
domain–containing chromatin factor that
represses the transcription of particular suites
of genes, some of which we now know are
germline-specific. It is known to interact with
other transcriptional repressors such as the
Rb-containing dREAM/MMB complex (4),
which contains homologs of the retinoblas-
toma protein (Rb) tumor suppressor pathway
involved in human retinoblastoma, an eye
cancer. Intriguingly, inactivation of compo-
nents of dREAM/MMB in cultured fly cell
lines also led to the up-regulation of germ-
line genes including (but not limited to) vasa,
nanos, spn-E, piwi, and zpg (5). Therefore,
preventing ectopic germline gene expres-
sion in somatic cells requires transcriptional
repression by retinoblastoma protein (Rb)
and its associated chromatin cofactors.
An analogous repression of germline
genes in somatic cells has been observed in
Caenorhabditis elegans. Mutations in genes
encoding homologs of dREAM/MMB com-
plex components and two other functionally
related chromatin repressors (MEP-1 and
Mi2) cause ectopic expression of the germ-
line granule proteins PGL-1, PGL-3, and
GLH-1 (a VASA homolog) (6, 7). These
similarities suggest that the retinoblastoma
pathway genes repress germline expression
in somatic cells in many animal systems,
perhaps including mammals.
There is precedent for exploring the
links between tumors and cellular program-
ming for totipotency (the ability to produce
1762 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org
all kinds of differentiated cells). Years ago,
investigators reported that testis or germ-
line antigens normally expressed only in
germ cells were highly expressed in certain
human solid tumors (8, 9). Higher levels
of these antigens are often correlated with
advanced tumors and are associated with a
poor prognosis, suggesting that these anti-
gens play a role in tumor etiology. Janic et
al. show that interfering with this program
of germline expression can blunt the growth
of the brain tumors induced by the recessive
l(3)mbt oncogene mutation.
The ectopically expressed germ cell
genes that are required for tumor growth are
present in the very earliest stages of germ
cells and germline stem cells. Janic et al.
detected ectopic expression of these genes
in the proliferative neuroblast cell regions
of the tumorous brain. Similarly, the most
convincing germline antigens in mam-
malian tumors are those that are normally
expressed in male germline stem cells (8,
9). Moreover, homologs of vasa along with
nanos or piwi are expressed in somatic stem
cells in other animals such as planaria, poly-
chaetes, and cnidaria (10–12). The expres-
sion of these totipotency factors, normally
limited to early developmental stages, may
inappropriately specify cancer stem cells
that divide and differentiate into heteroge-
neous cell types in tumors. Loss or inhibi-
tion of these stem cell program genes may
prevent the formation of cancer stem cells
or lead to their death, and may thereby pre-
vent tumor formation.
These findings suggest some new paths
for research. If the expression of germline
characteristics is common in tumors, for
instance, it should be seen in gene expres-
sion analyses of human tumors. Indeed, the
Piwil2 protein, a human Piwi family mem-
ber, was found to be widely expressed in a
variety of solid tumors (13). It should be
feasible to examine more systematically the
expression of germ cell genes, including
vasa and nanos, in human tumors by micro-
array or deep RNA sequencing. The retino-
blastoma tumor that triggered the study of
this pathway is a good candidate for studying
germline gene activity in tumorigenesis. In
addition, mutations in the human homologs
of L(3)MBT, Rb, and its chromatin cofactors
may be common in cancer genomes as they
are sequenced (14). A query of the human
homologs of these genes at the Cosmic
web site (www.sanger.ac.uk/genetics/CGP/
cosmic), for instance, revealed somatic
mutations in L(3)MBT, Rb, and CHD3 (an
Mi2 homolog) in a small fraction of tumors.
Because there are so many mutations in
Published by AAAS
these tumors, however, a more sophisticated
statistical analysis is needed.
The up-regulation of germline path-
ways in the l(3)mbt brain tumors and the
required role for some of these genes in
tumor growth also suggest new possibilities
for tumor therapy. Screens in C. elegans Rb
pathway mutants have identified 32 other
chromatin factors that reverse the devel-
opmental defects potentially induced by
somatic misexpression of germline genes
(6, 7, 15, 16). Interestingly, many of these
suppressor genes were also identified by
RNA interference and genetic screens for
defects in small RNA pathways (15, 17–
19), suggesting that the somatic cell speci-
Downloaded from www.sciencemag.org on December 23, 2010
fication defects in l(3)mbt-type tumors may
be mainly small RNA–based. Inactivation
of the Drosophila homologs of these sup-
pressor genes would be predicted to sup-
press tumor formation in the l(3)mbt brain
tumor model. These genes are also con-
served in mammals and could be potential
targets for drugs that treat tumors similar to
those analyzed by Janic et al.
References
1. A. Janic, L. Mendizabal, S. Llamazares, D. Rossell,
C. Gonzalez, Science 330, 1824 (2010).
2. S. Strome, R. Lehmann, Science 316, 392 (2007).
3. M. Ghildiyal, P. D. Zamore, Nat. Rev. Genet. 10, 94
(2009).
4. P. W. Lewis, E. L. Beall, T. C. Fleischer, D. Georlette,
A. J. Link, M. R. Botchan, Genes Dev. 18, 2929
(2004).
5. D. Georlette, S. Ahn, D. M. MacAlpine, E. Cheung,
P. W. Lewis, E. L. Beall, S. P. Bell, T. Speed, J. R. Manak,
M. R. Botchan, Genes Dev. 21, 2880 (2007).
6. Y. Unhavaithaya, T. H Shin, N Miliaras, J. Lee, T. Oyama,
C. C. Mello, Cell 111, 991 (2002).
7. D. Wang, S. Kennedy, D. Conte, Jr., J. K Kim, H. W.
Gabel, R. S. Kamath, C. C. Mello, G. Ruvkun, Nature
436, 593 (2005).
8. O. L. Caballero, Y. T. Chen, Cancer Sci. 100, 2014
(2009).
9. A. J. Simpson, O. L. Caballero, A. Jungbluth, Y. T. Chen,
L. J. Old, Nat. Rev. Cancer 5, 615 (2005).
10. N. Rebscher, F. Zelada-González, T. U. Banisch, F. Raible,
D. Arendt, Dev. Biol. 306, 599 (2007).
11. N. Shibata, L. Rouhana, K. Agata, Dev. Growth Differ.
52, 27 (2010).
12. C. G. Extavour, K. Pang, D. Q. Matus, M. Q. Martindale,
Evol. Dev. 7, 201 (2005).
13. J. H. Lee, D. Schütte, G. Wulf, L. Füzesi, H.-J. Radzun,
S. Schweyer, W. Engel and K. N. Hum. Mol. Genet. 15
(2005).
14. S. van den Heuvel, N. J. Dyson, Nat. Rev. Mol. Cell Biol.
9, 713 (2008).
15. M. Cui, E. B. Kim, M. Han, PLoS Genet. 2, e74 (2006).
16. E. C. Andersen, X. Lu, H. R. Horvitz, Development 133,
2695 (2006).
17. J. K. Kim, H. W. Gabel, R. S. Kamath, M. Tewari, J.-F.
Pasquinelli, Science 308, 1164 (2005).
18. T. Takasaki, Z. Liu, Y. Habara, K. Nishiwaki, J.-I.
Nakayama, K. Inoue, H. Sakamoto, S. Strome, Develop-
ment 134, 757 (2007).
19. N. L. Vastenhouw, K. Brunschwig, K. L. Okihara,
F. Müller, M. Tijsterman, R. H. A. Plasterk, Nature 442,
882 (2006).
10.1126/science.1200772

RETROSPECTIVE
Allan Sandage (1926–2010)
Donald Lynden-Bell
A
llan Sandage spent long nights lead-
ing observational cosmology until
studies of the cosmic microwave
background led to a new dawn. For 60 years,
he surveyed and measured the universe. He
died on 13 November at age 84.
His fascination with stars began as a youth
in Iowa and led to his study of physics at the
University of Illinois and a Ph.D. in astronomy
from the California Institute of Technology
(CalTech) in 1953, where he was a student of
Walter Baade. He joined the Carnegie Obser-
vatories in Pasadena (where he remained),
became Edwin Hubble’s assistant a year
before Hubble’s sudden death, and inher-
ited Hubble’s program to determine the size
and age of the universe. Hubble had used the
magnitudes of the brightest stars in galaxies
to determine their distances. Sandage realized
that many of these were actually tight groups
of bright stars, together brighter than any star
in the Milky Way, and this doubled the dis-
tances estimated to those galaxies. Thus began
his main work improving distance measure-
ments to determine the age of the universe.
Refining the distance estimates to variable
stars became a recurring theme of his work.
These estimates were needed to determine
the size of the first step on the ladder of dis-
tances through which the scale of the universe
was measured. By comparing the brightness
of stars of the same period in different galax-
ies, the ratio of their distances can be found.
However, a star’s color and chemical composi-
tion can influence its intrinsic brightness, and
interstellar dust can change its color. Accu-
racy can only be achieved by learning how
to compensate for such effects. Sandage dis-
covered good secondary distance indicators
in the brightest galaxies of clusters and super-
novae. He attributed this use of supernovae
to the brilliant and opinionated astronomer
Fritz Zwicky, but it was only through detailed
CREDIT: CARNEGIE INSTITUTION OF WASHINGTON
refinements, to which Sandage and his long-
time collaborator Gustav Tammann contrib-
uted greatly, that type IA supernovae became
the standard candles for measuring large cos-
mological distances.
At different times, Sandage’s attention
was dominated by the latest new discoveries:
x-ray sources, the origin of the Galaxy, qua-
Institute of Astronomy, Cambridge University, Cambridge,
CB3 0HA UK. E-mail: dlb@ast.cam.ac.uk
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
sars (quasi-stellar radio sources), active galax-
ies, and the Virgo cluster. He studied the the-
ory of how stars evolve into red giants under
Martin Schwarzschild at Princeton, and in
1963 they were jointly awarded the Edding-
ton medal of the Royal Astronomical Soci-
ety. The ages of the globular star clusters esti-
mated from theory were uncomfortably large
(at times reaching 18 billion years rather than
the true 12 billion years). The time scale found
from the rate of expansion of the universe
was shorter—closer to 10 billion years—but
depended inversely on the size of the Hubble
constant (the measure of how fast the universe
expands per unit distance; the speeds of gal-
axies moving away from us are approximately
proportional to their distances). Sandage was
convinced that nature must be consistent, and
worked unceasingly with Tammann to refine
distances and redetermine the Hubble con-
stant. For years their best estimate was around
52 ± 5 km s−1 Mpc−1, but their paper this year
on the linearity of the Hubble flow, based on
more accurate distance estimators, gives a
value of 62.3 km s−1 Mpc−1. That paper gives
the most accurate relative distances of local
structures in the universe, but the current con-
sensus, which is not always right, is that Hub-
ble’s constant is about 72 km s−1 Mpc−1.
Sandage was elected to the National Acad-
emy of Sciences in 1963 but resigned when
the Academy failed to elect Olin Eggen.
Eggen, Sandage, and I had collaborated on a
seminal paper about the formation of the gal-
axy. This paper gave birth to galactic archaeol-
ogy, in which the ages, chemical constitutions,
and galactic orbits of stars are used to discuss
the formation of the galaxy they are in.
In 1961, Sandage was the first to deter-
mine the spectrum of a quasar called 3C48,
Published by AAAS
PERSPECTIVES
An astronomer launched the field of
observational cosmology and influenced our
view of the universe over the past half-century.
but could not find a reasonable
interpretation of it. He consulted
colleagues Ira Bowen and Jesse
Greenstein about it, and mean-
while showed that the light of
“radio-star” 3C48 varied over a
few months, which demonstrated
that it must be only a few light-
months across, much smaller than
Downloaded from www.sciencemag.org on December 23, 2010
a galaxy. It was the discovery by
Maarten Schmidt of the large
redshift in another radio source
(3C273) two years later that led
to the interpretation of Sandage’s
spectrum. When Schmidt and Greenstein
finally understood and published the spectra
of 3C48, Sandage felt excluded. A few years
later, he discovered that there are many more
radio-quiet quasi-stellar objects than quasars.
This contributed to the current view that most
large galaxies contain giant black holes (dead
quasi-stellar objects) in their nuclei.
On one occasion, Sandage showed me a
photographic plate of a quasar that he was
studying. It lay near a bright galaxy, and on
inspection I saw a faint mark that led from the
quasar to the galaxy. Sandage pointed to the
office of Halton Arp and said with a laugh,
“Perhaps he could be right after all.” That pos-
sibility did not survive. The mark was merely
one side of the residue left by a droplet that
had dried after the plate had been developed.
Arp’s claims of connections between objects
of very different redshifts were never sub-
stantiated, although Fred Hoyle believed this.
Sandage was awarded the Crafoord prize
in 1991, was the joint recipient of the first
Gruber prize for Cosmology in 2000, and
was a Foreign Member of the Royal Soci-
ety. He was also awarded all the major astro-
nomical medals in the United States and the
United Kingdom. His public lectures were
filled with wonder and enthusiasm for astron-
omy, and his atlases of galaxies influenced
young observers and theorists alike. Sandage
was open to new ideas and was generous and
helpful to those who worked with him. He
was a true gentleman and expected equally
high standards of honor and acknowledg-
ment from others. He never stopped work-
ing and has left us with a wider under-
standing of the scale of the universe and a
greater wonder at the remarkable objects in it.
10.1126/science.1201221
1763
 ESSAY
SPORE* SERIES WINNER

Science 101: Building the Two online projects offer one-stop shopping
for teaching evolution, as well as the nature
Foundations for Real Understanding and process of science.

Anastasia Thanukos,1 Judith G. Scotchmoor,1 Roy Caldwell,1,2† David R. Lindberg1,2

I
t’s not just about evolution Since its launch in 2004, Understanding
anymore. Growing anti- Evolution’s impact has grown. The site now
science sentiment in the averages more than a million page accesses
United States now infuses public per month during the academic year, is avail-
discourse on conservation, vac- able in Spanish and Turkish (www.sesbe.org/
cination, distribution of research evosite/evohome.html, www.evrimianlamak.
funds, and climate change (1). org/e/Ana_Sayfa), and has been distributed

Downloaded from www.sciencemag.org on December 23, 2010

Low rates of scientific literacy (2) in Tibetan as part of the Dalai Lama’s Emory-
exacerbate the problem. Although Tibet Science Initiative. Additionally, site
the public recognizes its indebt- resources have been shown to improve teach-
edness to the products of scien- ers’ and students’ understandings of evolu-
tific knowledge, few understand tion and to increase instructors’ confidence
much about the nature of that in their ability to teach this challenging, and
knowledge or the processes that sometimes charged, material (4).
generated it (3). Without a basic As we developed Understanding Evolu-
understanding of how science Teaching the process of science. Mark Stefanski, a teacher ad- tion and noticed similar tensions and misin-
works, the public is vulnerable viser, uses the Science Flowchart with high-school biology students formation arising around topics like climate
to antiscience propaganda, which at Marin Academy, where the science faculty employs the flowchart change, we realized that much of the pub-
engenders distrust of science in lab activities to help students focus on the process of science. lic’s mistrust of evolution stems from more
when it comes to social issues, basic and even more important issues: poor
consumer choices, and policy decisions. pedagogy; review and editing by scientific understanding of how science works to build
The University of California Museum of experts; field testing through teacher advisers; reliable knowledge and confusion about the
Paleontology’s interest in this issue stemmed revision of materials with additional expert strengths and limitations of this process.
from a project on evolution education, which review; and summative evaluation performed Hence, we envisioned a Web site that would
expanded into an effort to support more by the evaluation firm Rockman et al. (4). leverage the format of Understanding Evolu-
effective teaching of the nature and process The result of this process was the Under- tion toward the goal of helping teachers rein-
of science (see the first figure). In 2000, we standing Evolution Web site, which provides force the true nature and process of science
hosted a conference on evolution instruc- educational materials targeting teachers of throughout their teaching.
tion that brought together stakeholders from kindergarten through college, students, and The Understanding Science Web site
education, academia, and the media. Partic- the general public (see the second figure). (www.understandingscience.org) was
ipants identified a critical need for a collec- Teacher advisers requested resources that launched in January 2009. Its development
tion of vetted tools for teaching evolution- engage students with data, explore scientific process followed that for Understanding
ary biology. Understanding Evolution (www. reasoning and science as a human endeavor, Evolution, bringing together scientists, phi-
understandingevolution.org) was built, in and demonstrate the relevance of evolution to losophers of science, teachers, writers, and
collaboration with the National Center for everyday life. Site resources that respond to Web designers to conceive, develop, and vet
Science Education, to meet this need and to these needs include (i) “Evo in the News,” a content. The site is unique in its straightfor-
provide a clear and comprehensive reference monthly feature that reveals the evolutionary ward presentation of science, not as an eso-
for the general public. science behind a current news story and inte- teric collection of vocabulary and facts, but
Understanding Evolution brought together grates data from the primary literature with as an intensely human endeavor—a multi-
an advisory board of scientists, Web design- discussion questions and background read- faceted process that both students and sci-
ers, authors, and master teachers to create a ing; (ii) research profiles and case studies, entists can use to better understand the natu-
vision for the project and develop content. which follow a particular scientist or inves- ral world. Instead of oversimplifying science
Key aspects of this process included teacher tigation and step students through the logic into a five-step recipe, the site emphasizes the
guidance on content types, Web features, and of testing evolutionary hypotheses; and (iii) dynamic and iterative nature of the process,
interactive investigations (e.g., Visualizing as well as the roles of creativity and com-
Life on Earth, http://evolution.berkeley.edu/ munity in scientific progress. Understanding
PHOTO CREDIT: LIZ GOTTLIEB

1
University of California Museum of Paleontology, Univer-
sity of California, Berkeley, Berkeley, CA, 94720–4780, evolibrary/article/ldg_01) that ask students to Science is designed to give students and the
USA. 2Department of Integrative Biology, University of put scientific reasoning into practice. Many general public the tools they need to recog-
California, Berkeley, Berkeley, CA, 94720–4780, USA. such resources will be housed in our Under- nize the relevance of science to their lives and
*SPORE, Science Prize for Online Resources in Education;
graduate Library, an area devoted to college- to keep pace with the ways in which science
www.sciencemag.org/special/spore/. level evolution instruction, which will open informs personal and societal decision-mak-
†Author for correspondence. E-mail: rlcaldwell@berkeley.edu in January 2011. ing. These ideas are communicated through

1764 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

Published by AAAS
 ESSAY

a friendly primer on the nature

of science, as well as through
“Science in Action” features,
which use stories from the his-
tory of science, animations, and
graphics to reinforce basic sci-
entific concepts and show how
science works.
By providing a compre-
hensive, practical resource for
teaching the nature and process
of science, Understanding Science also fills a
major gap in the landscape of science educa-
tion materials. Teaching resources on the site
(second figure) are informed by educational
research showing that instruction in this area
About the Authors
From left to right: Roy Caldwell, Josh
Frankel, David R. Lindberg, Judith G. Scotch-
moor, Anastasia Thanukos, and David Smith.
R. Caldwell and D. R. Lindberg, co-principal
investigators on the project, are curators in
the University of California Museum of Paleon-
tology (UCMP) and professors of Integrative
Biology at the University of California, Berkeley.
J. G. Scotchmoor, project coordinator, is an assistant director at UCMP, in charge of education
and outreach. A. Thanukos, primary author, is principal editor at UCMP. J. Frankel and D. Smith
work in education and outreach at UCMP and direct Web design for the project.

Downloaded from www.sciencemag.org on December 23, 2010

is most effective when it is explicit, reflective,
and reinforced in many contexts (5). Three
tools from the site help teachers put these
guidelines into practice in multiple instruc-
tional settings. The Science Checklist helps
students identify key characteristics of science
in different investigations. The Science Flow-
chart provides a more accurate and appealing
representation of the scientific process than
the rigid Scientific Method. The Science Tool-
kit helps students analyze policies and media
messages to get to the science behind the spin.
Educational resources available through Understanding
Evolution and Understanding Science
Resource
Understanding Evolution
Searchable lesson database
Tips, strategies, and teaching help
Common misconceptions and explanations
Conceptual framework
Image library
Evolution 101 and basic content
Advanced tutorials
Interactive online labs
Research profiles and case studies
Evo in the News articles
Understanding Science
Individual lessons and activities
Searchable lesson database
Tips, strategies, and teaching help
Guidelines for modifying lessons
Tips from the education research literature
Common misconceptions and explanations
Conceptual framework
First-hand instructor reports
PHOTO CREDIT: MOLLY WRIGHT/UCMP
Although less than 2 years old, Under-
standing Science has had far-reaching
impacts and now averages more than 60,000
page accesses per month during the school
year. The project is endorsed by organiza-
tions such as the American Institute of Bio-
logical Sciences, and materials from the site
have been incorporated into middle- and
high-school textbooks from major publish-
ers. Encouragingly, an evaluation of a year-
long in-service training program indicates
that site materials generate a high level of
Target Audience
Undergraduate General
Students K–12 Teachers Instructors Public
teacher buy-in, meaningful increases in stu-
dent understanding, and reports of increased
student motivation (6).
In the current climate of both funding
constraints and concern for the future of sci-
ence education in the United States, we see
opportunities for additional contributions
from these projects, such as new resources
and collaborations with scientists, as well as
challenges, such as maintaining vibrant and
freely accessible teaching materials while
seeking a sustainable funding model. Fortu-
nately, many other initiatives have also set
their sights on improving science literacy
(7), and this complements a growing move-
ment within the scientific community itself
to reach out to students and the broader pub-
lic. We are proud to be a part of this cam-
paign and are committed to working with
scientists, scientific agencies, the media,
and educators to build a more scientifically
literate society.
References and Notes
1. C. Mooney, S. Kirshenbaum, Unscientific America (Basic
Books, New York, 2009).
2. R. Bybee, B. McCrae, R. Laurie, J. Res. Sci. Teach. 46, 865
(2009).
3. National Science Board, Science and Engineering Indicators
2010 (National Science Foundation, Arlington, VA, 2010).
4. Rockman et al., Understanding Evolution Evaluation,
http://evolution.berkeley.edu/evolibrary/evaluation.php.
5. N. G. Lederman, in Handbook of Research on Science
Education, S. K. Abell and N. G. Lederman, Eds. (Law-
rence Erlbaum Associates, Mahwah, NJ, 2007),
pp. 831–880.
6. M. A. M. Stuhlsatz, Final Evaluation Report for the
University of California Museum of Paleontology:
Understanding Science (ER 2010-08, Biological Sciences
Curriculum Study, Colorado Springs, CO, 2010); http://

undsci.berkeley.edu/BERKreport_7_12_10.doc.
Understanding Science 101 and basic content 7. American Association for the Advancement of Science,
Advanced supplementary content Center for Public Engagement with Science and Technol-
Science in Action stories ogy, www.aaas.org/programs/centers/pe/.
8. Supported by the NSF 0096613, EAR-0624436, DUE-
Currently available Launching this academic year 0918741; Howard Hughes Medical Institute 51003439;
and the University of California Museum of Paleontology.
Educational resources. The Understanding Evolution and Understanding Science Web sites offer a wealth
of resources for teaching and learning evolutionary biology and the nature and process of science. K–12, Published online 2 December 2010;
kindergarden to high school. 10.1126/science.1186994

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1765

Published by AAAS
 AAASNEWS&NOTES EDITED BY EDWARD W. LEMPINEN

INTERNATIONAL dous resource that can be tapped for the pur-

poses of development of our continent.”
Seeking Growth, East Africa Science education and training are crucial
to the region’s ability to address its challenges,

Expands Science, Education Ties said Romain Murenzi, formerly a minister

in the Kagame administration overseeing
education, science, and communication, and
KIGALI —High-ranking government lead- Rwanda’s neighbors are embracing a similar now director of the AAAS Center for Sci-
ers from six African nations have pledged approach to strong economic growth. ence, Technology and Sustainable Develop-
to expand their collaboration in science and The conference, convened 8 to 9 Decem- ment. But with less than 1% of East Africa’s
science education to further economic and ber in Rwanda’s capital, reflected the region’s 200 million people holding college degrees,

Downloaded from www.sciencemag.org on December 23, 2010

human development in their resource-rich energy and optimism. It attracted about 60 he said, “regional integration is critical to
but long-impoverished region. influential science leaders—government efficiently use this scarce resource.”
At a landmark meeting organized by the science and education ministers, heads of Before the conference, Jo Ellen Rose-
Rwandan Ministry of Education and AAAS, national science organizations, and univer- man, director of AAAS’s Project 2061 science
leaders from the East African nations of sity rectors, along with top officials from the literacy initiative, gave a 2-hour briefing on the
Burundi, Kenya, Rwanda, Tanzania, program’s extensive science learning
and Uganda, along with the Demo- resources to about 40 administrators
cratic Republic of Congo, agreed to and staff at Rwanda’s National Cur-
establish a forum for their science riculum Development Centre.
ministers. By working together on At the conference, some 40
a range of issues—from education undergraduate and graduate students
and health to energy and the environ- won praise for their research post-
ment—they hope to support sustain- ers, many focused on agriculture,
able economic growth while advanc- hydrology, and health, and most with
ing East Africa’s emerging commit- direct practical applications.
ment to regional integration. Speakers from both Africa and
“The integration spirit in the the United States put particular
region is very high,” said Charles focus on the need to recruit more
Murigande, Rwanda’s minister of women into science and engineer-
education. “The objective of this con- Regional partners. (left to right): Alexis Kanyenye, deputy direc- ing studies. Mabel Imbuga, vice
ference was to bring us together. We tor of cabinet, Ministry of Scientific Research, Democratic Republic of chancellor of the Jomo Kenyatta
did not want to come together once the Congo; Romain Murenzi, director of the AAAS Center for Science, University of Agriculture and Tech-
and end it there, and so I expected … Technology and Sustainable Development; Julien Nimubona, minister nology, described Kenya’s incen-
that we should continue to meet and of higher education and scientific research, Burundi; Ignace Gatare, tives to bring women into the sci-
exchange ideas on how we can pro- minister in the office of the president for Information Communication ences. Uganda and Tanzania are
mote science and technology.” Technologies, Rwanda; Nelson G. Gagawala, minister of state for trade, making similar efforts, speakers
That could be a crucial oppor- Uganda; Alan I. Leshner, CEO of AAAS; Charles Murigande, minister of said. Advocates need to demon-
tunity for harmonizing science and education, Rwanda; Charles Kitwanga, deputy minister of communi- strate more forcefully that bring-
education goals and procedures in cation, science, and technology, Tanzania; and Shaukat A. Abdulrazak, ing more women into science has
the region, said AAAS Chief Execu- executive secretary, Kenya National Council for Science and Technology. a positive economic impact, said
tive Officer Alan I. Leshner, who led Shirley Malcom, head of Education
a delegation to the conference. “This is a very African Development Bank and UNESCO. and Human Resources at AAAS.
important and promising development,” said W. Stuart Symington, the U.S. ambassador to Vaughan Turekian, AAAS’s chief interna-
Leshner, the executive publisher of Science. Rwanda, also spoke at the event. tional officer, was among many speakers who
“If we’re going to be able to work for the bet- Discussions were collegial and candid, urged science leaders to build on momen-
terment of humankind on a global scale, the and they ranged broadly across issues— tum created by the conference. “Cooperating
scientific community has to be able to func- from Internet technology and data collection across borders to improve the lives of people
tion as a global community in and of itself.” to food security and approaches to scien- and drive prosperity represents one of the
Rwanda, devastated by genocide in 1994, tific collaboration. But participants returned greatest goals of our scientific and education
has become an international model for build- repeatedly to the importance of education. enterprise,” he said. “We have to seize on this
ing economic strength with education, sci- “Our continent has reached the milestone opportunity through sustained action.”
ence, and technology; Rwandan President of 1 billion people,” said Boukary Savadogo, The science ministers forum can help
Paul Kagame discussed that strategy and division manager for education, science, and lead that effort, said Ugandan Trade Minister
plans for the future in an hour-long meet- technology at the African Development Bank. Nelson G. Gagawala. “We need action,” he
ing with Leshner and the AAAS delegation. “These 1 billion brains constitute a tremen- said. “We need action in real time.”

1766 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

Published by AAAS
PUBLIC ENGAGEMENT
Science, Law Enforcement Build Biotech Bridges
With scientists working to create new life
forms and amateur biology clubs springing
up nationwide, it stands to reason that the U.S.
security community would be concerned that
one rogue researcher or one innocent error
might create a grave problem.
But before uneasiness could turn to con-
flict, the FBI, working closely with AAAS,
embraced a new strategy. The Bureau held
conferences with university and private sector
researchers, attended synthetic biology sci-
ence fairs, and spent time with do-it-yourself
(DIY) biologists. The message, though tai-
lored for each audience, was consistent.
“We want science and security communi-
ties to come to an understanding to promote
a culture of responsibility,” says Edward H.
You, an experienced researcher and now the
FBI supervisory special agent guiding the
outreach effort. By bringing those communi-
ties together, “we can…identify what some of
the risks and gaps might be, and then come up
with strategies that make sense to both com-
munities to mitigate those risks and gaps.”
A certain amount of uneasiness was inevi-
table after the deadly blitz of anthrax letters
that followed the 9/11 terror attacks and, more
recently, the stunning advances and increasing
accessibility of biotech research. In research
published last May in Science, genomics
ANNUAL MEETING
In Washington, Researchers Seek Science Without “Silos”
Some of the most compelling scientific work of
the 21st century depends on researchers who
seek inspiration and partnerships across disci-
plines and national borders. It’s an approach
that Frances H. Arnold, a plenary speaker at
the 2011 AAAS Annual Meeting, uses when she
combines mechanical engineering, chemistry,
and evolutionary biology to design new enzymes
for medical and energy research. It’s also the
driving force behind the work of Colin Phil-
lips, one of the meeting’s topical speakers, who
employs computer science, anthropology, and
neuroscience in his studies of human grammar.
Under the banner “Science Without Borders,”
Arnold, Phillips, and scientists and engineers
from more than 50 countries will convene from
17 to 21 February at the 177th Annual Meet-
ing in Washington, D.C. Their innovative proj-
ects “cross conventional borders or break out
from silos, especially in groundbreaking areas of
research,” said AAAS President Alice S. Huang.
The program will also highlight the international
nature of scientific collaborations, said Huang,
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
pioneer J. Craig Venter announced develop-
ment of the first cell controlled by a synthetic
genome. That breakthrough underscored that
biotech will likely create unpredictable impli-
cations for science and society.
In a recent appearance at AAAS, bioeth-
icist Thomas H. Murray said synthetic biol-
ogy—fundamentally altering life or creating
new life forms—offers “mind-boggling” pos-
sible benefits, from production of new phar-
maceuticals to cleaning up oil spills. But, he
added, the benefits must be weighed against
bioterrorism and other hard-to-define risks.
“If I didn’t think the potential benefits...
were massive, there would be no point in hav-
ing this conversation,” said Murray, president
and chief executive officer of the Hastings
Center, in the annual AAAS-Hitachi Lecture
on Science & Society on 28 October.
Finding the best balance of benefits
and risks is the rationale for the collabora-
tion between the FBI and AAAS, said AAAS
biosecurity expert Kavita Berger, an associate
program director in the Center for Science,
Technology and Security Policy.
Just a few years ago, Berger contributed to
a survey of researchers that found only a third
were comfortable sharing their research with
agents, and a mere 14% felt comfortable with
the FBI having a role in monitoring research.
who has consulted on sci- i-
ence policy for governmentnt
agencies in China, Taiwan,n,
and Singapore.
The 2011 meeting willl
continue AAAS’s tradition
of boundary-crossing sci-
ence, featuring multi-
disciplinary research on
oceans, human health,
sustainability, and next-generation
generation engi-
neering. Special events include seminars on neu-
roscience and robotics, molecular machines, the
search for Earth-like planets, and a plenary panel
on emerging issues in biosecurity.
For registration and other information about
the 2011 Annual Meeting, see www.aaas.org/
meetings. Information from the D.C. gathering
will also be posted at the Annual Meeting News
site at http://news.aaas.org, at the 2011 AAAS
Annual Meeting and ScienceNOW pages on
Facebook, and on Twitter at #AAASmtg.
—Becky Ham
Published by AAAS
Genspace President Ellen Jorgensen (left) and FBI
Supervisory Special Agent Edward H. You (right).
But if science and security couldn’t build
Downloaded from www.sciencemag.org on December 23, 2010
a working relationship, she thought, then
policy-makers, acting out of mistrust or fear,
might impose rules that impede research
without affecting real security concerns.
Collaboration, she said, is “ultimately
going to be a lot more productive and a lot
more useful in reaching the end goals of
security and science.”
In professional conferences organized by
the FBI and AAAS, You and Berger have had
agents and researchers work through simu-
lated problems related to biotech and biosecu-
rity. In the process, they learned about each
other’s values, perspectives, and practices.
Now the uneasiness is giving way to
closer interaction between researchers and
law enforcement, with major universities
offering to host the conferences. “We’re see-
ing a paradigm shift,” said You, who had
worked in gene therapy and cancer research
before joining the FBI.
AAAS is helping forge a similar relation-
ship with amateur biologists, who number
an estimated 4000 or more nationwide. An
iinformal meeting this fall brought three of
tthem together with You and others from the
F
FBI, along with government and AAAS
oofficials. The DIY speakers described how
a love of science and commitment to public
en
engagement has led them to hold exhibits at
st
street fairs and form community labs.
Ellen Jorgensen, an assistant professor in
pathology at New York Medical College and
president of the Genspace community lab
in New York City, acknowledged that coop-
eration with agents does not come easily for
many in the DIY movement.
But, she said, “I think that the meetings we
have had were very useful in terms of foster-
ing some trust between the FBI and the DIY
biocommunity…To kill a movement that
embodies a reawakened public enthusiasm
about science due to concerns about biosecu-
rity would be a terrible shame.”
—Brian Vastag contributed to this report.
1767
 REVIEW
How Does the Microbiota Shape Host Immune
Development and Function?
Although microbes reside in several anatomical
Has the Microbiota Played a Critical locations including the skin, vagina, and mouth,
the lower gastrointestinal tract of mammals
Role in the Evolution of the Adaptive harbors the greatest density and diversity of com-
mensal microorganisms. These include bacteria,

Immune System? archaea, fungi, viruses, protozoans, and (in some

cases) multicellular helminths; however, bacteria
predominate and reach 100 trillion microbial cells
Yun Kyung Lee and Sarkis K. Mazmanian* in the colon. Recent efforts to sequence the bac-
terial genomes of the microbiota (known as the
Although microbes have been classically viewed as pathogens, it is now well established that microbiome) have begun to reveal its genetic
the majority of host-bacterial interactions are symbiotic. During development and into adulthood, identity (1) and suggest that our microbiome con-
gut bacteria shape the tissues, cells, and molecular profile of our gastrointestinal immune system. tains more than 150 times as many nonredundant
This partnership, forged over many millennia of coevolution, is based on a molecular exchange genes as in the human genome (2). For decades,
involving bacterial signals that are recognized by host receptors to mediate beneficial outcomes microbiological techniques to culture bacteria in

Downloaded from www.sciencemag.org on December 23, 2010

for both microbes and humans. We explore how specific aspects of the adaptive immune system the laboratory have only identified cultivatable
are influenced by intestinal commensal bacteria. Understanding the molecular mechanisms that microorganisms, which represent a minority of
mediate symbiosis between commensal bacteria and humans may redefine how we view the the microbial species of the gut. The aggregate
evolution of adaptive immunity and consequently how we approach the treatment of numerous human microbiota likely contains 1000 to 1150
immunologic disorders. bacterial species (spread among all people sam-
pled), with each person harboring about 160 bac-
e are (fortunately) not alone: Humans cause they share similar molecular patterns that terial species (2). This suggests that an individual’s

W provide residence to numerous micro-

bial communities comprising hundreds
of individual bacterial species. Although teleo-
logical design may predict that the immune sys-
tem evolved to eliminate infectious microbes, we
now know that almost every environmentally
exposed surface of our bodies is teeming with
symbiotic microbes (Fig. 1). These polymicrobial
communities contribute profoundly to the archi-
tecture and function of the tissues they inhabit
and thus play an important role in the balance
between health and disease. The notion that com-
mensal microbes critically affect tissue and cell de-
velopment in humans can be rationalized when this
process is viewed from an evolutionary perspective.
Bacteria populated Earth 2 billion years be-
fore the first signs of eukaryotic life, and they
occupy almost every terrestrial and aquatic niche
on our planet. Mitochondria and chloroplasts of
eukaryotic cells are descended from bacteria,
which suggests that bacteria may have had an
active role in the evolution of higher organisms.
As multicellular metazoans evolved more com-
plex body plans, bacteria acquired the ability to
inhabit new anatomical niches. Animals represent
a stable, nutrient-rich ecosystem for microbes to
thrive; hence, host health is paramount to the mi-
crobiota. In turn, the host benefits from a diverse
commensal microbiota that helps to digest com-
plex carbohydrates and provide essential nutrients
to mammals.
Symbionts are not the only microbes the host
encounters, however. An important challenge faced
by the host immune system is to distinguish be-
tween beneficial and pathogenic microbes, be-
Division of Biology, California Institute of Technology,
Pasadena, CA 91125, USA.
*To whom correspondence should be addressed. E-mail:
sarkis@caltech.edu
are recognized by the innate immune system
(such as lipopolysaccharide, peptidogycan, lipo-
proteins, and flagellin). Discrimination between
specific microbes may be a feature of the adapt-
ive immune system, which can recognize discrete
molecular sequences and mount both pro- and
anti-inflammatory responses depending on the
nature of the antigen. In particular, CD4+ T cells
are quite plastic and differentiate into numerous
subsets after development in the thymus and
thus are capable of sensing environmental cues
from the microbiota. As adaptive immunity
evolved in higher vertebrates, the ability of this
system to recognize and respond to specific mi-
croorganisms may have been driven by evolu-
tionary forces provided by the microbiota itself,
resulting in immune functions beyond simply clear-
ing microbial pathogens (which in theory also
helps the microbiota by improving host health).
Recent evidence shows that the commensal mi-
crobiota “programs” many aspects of T cell dif-
ferentiation, thus augmenting the developmental
instructions of the host genome to engender the
full function of the adaptive immune system.
Here, we review concepts derived from gnoto-
biology (Greek for “known life”) to unravel how
commensal bacteria promote the development
and function of adaptive immunity. In particular,
we explore how CD4+ T helper cell subsets with-
in the gastrointestinal and systemic immune sys-
tem are shaped (perhaps even controlled) by our
microbiota and theorize how and why gut bac-
teria evolved to so profoundly influence immuno-
logic well-being. Understanding human coevolution
with our microbiota may lead to a philosophical
and conceptual redefinition of the microbial world
and may yield clinical advances toward the treat-
ment of autoimmunity and inflammatory dis-
eases by harnessing the immunomodulatory
properties of human commensal bacteria.
microbiome is relatively distinct in composition
and is adaptable to environmental changes and/or
host genetics.
Germ-free animals (born and raised in the
absence of all microbes) provide important in-
sights into how the microbiota affects the host
immune system. The development of gut-associated
lymphoid tissue (GALT), the first line of defense
for the intestinal mucosa, is defective in germ-
free mice. Germ-free mice display fewer and
smaller Peyer’s patches, smaller and less cellular
mesenteric lymph nodes, and less cellular lamina
propria of the small intestine relative to animals
with a microbiota (3–7). Besides developmental
defects in tissue formation, the cellular and mo-
lecular profile of the intestinal immune system is
also compromised in the absence of symbiotic
bacteria. In germ-free mice, intestinal epithelial
cells (IECs), which line the gut and form a
physical barrier between luminal contents and the
immune system, exhibit reduced expression of
Toll-like receptors (TLRs) and class II major
histocompatibility complex (MHC II) molecules
(8, 9), which are involved in pathogen sensing
and antigen presentation, respectively. Inter-
spersed between epithelial cells is a specialized
population of T cells known as intraepithelial
lymphocytes (IELs). IELs from germ-free mice
are reduced in number, and their cytotoxicity is
compromised (10, 11). Microbial colonization
expands specific subsets of intestinal gd T cells
(12). Germ-free mice also have reduced num-
bers of CD4+ T cells in the lamina propria (13).
The development of isolated lymphoid follicles,
specialized intestinal structures made of mostly
dendritic cells and B cell aggregates, is also de-
pendent on the microbiota (14). Therefore, mul-
tiple populations of intestinal immune cells
require the microbiota for their development
and function.

1768 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


The absence of a microbiota also leads to
several extra-intestinal defects, including reduced
numbers of CD4+ T cells in the spleen, fewer and
smaller germinal centers within the spleen, and
reduced systemic antibody levels, which suggests
that the microbiota is capable of shaping sys-
temic immunity (15–17). Beyond development,
the microbiota also influences functional aspects
of intestinal and systemic immunity, including
pathogen clearance. Germ-free mice are more
susceptible to infectious agents such as Shigella
flexneri, Bacillus anthracis, and Leishmania (18).
Peptidoglycan from the microbiota enhances
neutrophil cytotoxicity after systemic infections
by Streptococcus pneumoniae and Staphylococcus
aureus (19). During challenge with Listeria
monocytogenes, sterile mice harbored an increased
bacterial burden in the liver, spleen, and peritoneal
to the peritoneal cavity in response to Listeria
infection is impaired in germ-free mice (20).
The contributions of the microbiota to the
development and function of the immune system
appear to be fundamental. A more robust im-
mune system, equipped with a diverse arsenal of
cells and molecules, is better able to combat
microbial pathogens and ultimately provides a
healthier residence for commensal bacteria. This
view implies that host mechanisms and the mi-
crobiota may have evolved to collaborate against
infectious agents. Indeed, several reports show an
antagonistic relationship between overt pathogens
and the microbiota. For example, Salmonella
triggers intestinal inflammation, which reduces
the numbers and diversity of the microbiota—a
process that promotes bacterial infection (21).
Depletion of the microbiota diminishes intestinal
REVIEW
infections by Citrobacter rodentium and Cam-
pylobacter jejuni (22). Given the role of the mi-
crobiota in immune system function, harnessing
the immunomodulatory capabilities of the micro-
biota may offer novel avenues for the development
of antimicrobial therapies for infectious disease.
How Does the Microbiota Provide Signals
to Instruct Peripheral Regulatory T Cell
Differentiation?
Although many cell types are influenced by the
microbiota, we focus here on the emerging role
of the microbiota on effector CD4+ T cell differ-
entiation. After lineage commitment in the thy-
mus, naïve CD4+ T cells enter the periphery, where
they sense environmental signals that further in-
struct their maturation and function. During an
infection, microbial and host signals provide cues

Downloaded from www.sciencemag.org on December 23, 2010

cavity (20). Moreover, trafficking of T lymphocytes immune responses that help to control enteric to naïve CD4+ T cells to induce their differentia-

Airways
Mouth

Skin

Intestines

Vagina

Fig. 1. The microbiome of various anatomical locations of the human body. regions of the body share similarities, they each have a unique site-specific “fingerprint”
Numerous bacterial species colonize the mouth, upper airways, skin, vagina, and made of many distinct microbes. Each site has a very high level of diversity, as shown by
intestinal tract of humans. The phylogenetic trees show the speciation of bacterial clades the individual lines on the dendrograms. Data are from the NIH-funded Human
from common ancestors at each anatomical site. Although the communities in different Microbiome Project; circles represent bacterial species whose sequences are known.

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1769

 REVIEW
tion into various pro- and anti-inflammatory
subsets. For instance, infection by intracellular
pathogens drives the development of T helper
1 (TH1) cells, whereas extracellular pathogens
induce the differentiation of TH2 and TH17
subsets (23). These proinflammatory cells coor-
dinate many aspects of the innate and adaptive
immune response to clear microbial invaders.
CD4+ T cells can also adopt an anti-inflammatory
phenotype. Regulatory T cells (Tregs) control un-
wanted immune system activation and dampen
inflammation after microbial infection. Expres-
sion of the Treg cell–specific transcription factor
Foxp3 (forkhead box P3) induces regulatory
phenotypes and functions by CD4+ T cells (24).
Foxp3+ T cells develop in the thymus shortly
after birth, and deletion or depletion of Foxp3+ T
cells leads to severe multi-organ lymphoproliferative
disease and autoimmunity (24). Besides the
thymus-derived CD4+Foxp3+ T cells (“natural”
Tregs), various subsets of Tregs can be generated
in the gut from naïve T cells (“inducible” Tregs),
some of which produce the anti-inflammatory
cytokine interleukin-10 (IL-10) (25–28). More-
over, intestinal bacteria may be critically involved
in the differentiation of some gut Treg subsets
(29–31). Accordingly, several commensal bacteria
(e.g., Bifidobacteria infantis, Faecalibacterium
prausnitzii) have been shown to induce Foxp3+
Tregs and IL-10 production in the gut (32, 33).
Members of the genus Bacteroides are prom-
inent in the mammalian gastrointestinal tract and
are also potent stimulators of the mucosal im-
mune system of mammals (34). The gut micro-
organism Bacteroides fragilis has emerged as a
model system for the study of immune-bacterial
symbiosis. During colonization of mice with B.
fragilis, the bacterial molecule polysaccharide A
(PSA) directs the cellular and physical develop-
ment of the immune system (16). Moreover, B.
fragilis is able to prevent intestinal pathology in
two independent models of experimental colitis
in a PSA-dependent manner (35). Furthermore, in
mouse models of experimental colitis, oral treat-
ment of mice with purified PSA protects against
weight loss, decreases proinflammatory cyto-
kine expression in the gut, and inhibits lympho-
cyte infiltration that is associated with disease
(35). The protective effects of PSA were likely
mediated by CD4+ T cell production of IL-10,
because CD4+ T lymphocytes from mesenteric
lymph nodes of PSA-treated mice produced ele-
vated amounts of IL-10. IL-10–deficient CD4+ T
cells abolished the protective effects of PSA in
colitis models. These studies identify PSA as a
beneficial microbial molecule that suppresses
inflammation-driven host pathology.
No consensus has been reached about wheth-
er Foxp3+ Treg cells in the intestinal tissues of
germ-free mice are defective (36–40); however,
production of IL-10 is reduced within the GALT
of germ-free animals (13, 36, 41). Foxp3+ Treg
cells in the colon of germ-free mice exhibit
reduced IL-10 expression, and monocolonization
with PSA-producing bacteria (but not PSA-
1770 24 DECEMBER 2010
deficient B. fragilis) restores IL-10 expression
(42). PSA increases Foxp3 expression by Treg
cells, and colonization of germ-free animals with
B. fragilis augments the in vitro suppressive
activity of Tregs in a PSA-dependent manner
(42). PSA protects and cures animals from
experimental colitis by inducing Foxp3+ Treg
cells and IL-10 production (42). Recently, it was
shown that a defined set of Clostridium strains
induce Foxp3+ Tregs that produce IL-10 in the
colon and protect animals from colitis (43). These
findings imply that optimal Foxp3+ Treg cell
differentiation in the colon requires signals from
the microbiota and the host genome. They also
suggest that specific commensal bacteria may have
evolved to promote Treg cell differentiation in the
gut to actively engender mucosal tolerance. If
validated in human disease, these findings may
lead to probiotic therapies for colitis based on
microbial-driven Treg induction.
How Does the Microbiota Instruct T Helper
Cell Differentiation?
Although the microbiota has been shown to af-
fect the TH1-TH2 balance in systemic immune
compartments (44), studies have not yet observed
symbiotic microbial effects on TH1 or TH2 cells
at mucosal surfaces. In contrast, TH17 cell de-
velopment in the gut is specifically affected by
commensal bacteria (45). Germ-free mice are
deficient in the production of IL-17 from CD4+ T
cells (the hallmark cytokine of TH17 cells) of the
small intestinal lamina propria (39). Only a minor
defect was noted for gd T cells, which suggests
that the lack of TH17 cells was not due to an
overall deficiency in immune activation and that
specific features of the immune response are sen-
sitive to the microbiota. One mechanism of in-
testinal TH17 cell differentiation may be production
of adenosine 5′-triphosphate (ATP) in the lamina
propria by commensal bacteria, which drives the
production of TH17-inducing cytokines by resi-
dent lamina propria cells (46). Germ-free animals
display a reduction in fecal ATP amounts, and
treatment of mice with a nonhydrolyzable ATP an-
alog increased the number of gut TH17 cells (46).
Not all bacterial species of the microbiota are
similar in their ability to promote nonpathogenic
T cell responses during normal colonization of
animals. Of the numerous bacterial phylotypes
that constitute the normal microbiota of mice,
only segmented filamentous bacteria (SFBs) have
been shown to direct intestinal T helper cell de-
velopment. A role for SFBs was identified by
reconstituting germ-free mice with various sub-
sets of bacterial consortia and measuring cytokine
production in gut mucosal tissues (41). SFBs,
which are known to tightly adhere to the
intestinal mucosa (and to Peyer’s patches of the
ileum), induced the development of T helper cells
in the lamina propria and in cell aggregates of
Peyer’s patches. This activity was greatly re-
duced even when very complex groups of bacteria
were tested if they were missing SFBs. In a
contemporary report, a comparison of the micro-
VOL 330 SCIENCE www.sciencemag.org
biota of mice that contained TH17 cells with mice
deficient in these cells identified SFBs as being
sufficient to restore TH17 cells to germ-free mice
and conventionally raised mice that lack TH17
cells (47). Gene expression analysis showed that
SFBs induce a spectrum of intestinal immune
responses including production of cytokines and
chemokines, antimicrobial peptides, and serum
amyloid A (SAA), which was shown in vitro to
support TH17 cell differentiation (47). SFB coloni-
zation protected animals from intestinal infection
with C. rodentium, a bacterial pathogen of animals
that causes acute intestinal inflammation similar to
enteropathogenic Escherichia coli in humans (47).
Thus, commensal SFBs induce a tonic (or
controlled) inflammatory response in the gut
through TH17 cell development that does not
cause pathology and is protective against infec-
Downloaded from www.sciencemag.org on December 23, 2010
tion with pathogenic bacteria. These new studies
build on research done several decades ago, which
showed that SFBs promote germinal center de-
velopment, mucosal immunoglobulin A responses,
and recruitment of intraepithelial lymphocytes
(48–50). Collectively, it appears that only a par-
ticular subset of bacteria from the gut microbiota
directly influences TH17 immune responses
during steady-state colonization.
Are Noninfectious Human Diseases
Influenced by the Microbiota?
Numerous autoimmune diseases result from dys-
regulation of the adaptive immune system. The
incidences of autoimmune diseases such as mul-
tiple sclerosis (MS), type 1 diabetes (T1D), and
rheumatoid arthritis (RA) are rapidly increasing
in Western societies, suggesting alterations in en-
vironment factors that regulate the adaptive
immune system. As appreciation for the immu-
nomodulatory potential of commensal bacteria
has increased, we and others have proposed that
lifestyle changes have caused a fundamental al-
teration in our association with the microbial world
(51, 52). Altered diets, widespread antibiotic use,
and other societal factors in developed countries
may result in an unnatural shift in the community
composition of a “healthy” microbiota, leading to
altered microbial colonization known as dysbio-
sis. Whether dysbiosis causes any human disease
is yet unproven (insights may come from micro-
biome sequencing projects); however, evidence
in mice suggests that dysbiosis may affect auto-
immunity by altering the balance between tole-
ragenic and inflammatory members of the microbiota
(Fig. 2). PSA from B. fragilis, previously shown
to treat experimental colitis in the gut, is also able
to prevent and cure experimental autoimmune
encephalomyelitis (EAE, an animal model for
multiple sclerosis) (53). Oral treatment of animals
with PSA reduced TH17 cell development and
increased Treg numbers in the central nervous
system (CNS). Furthermore, germ-free animals
display reduced TH17 cell numbers in the spleen
and spinal cords, and do not develop RA or EAE
(inflammation in joints and in the CNS, respec-
tively) (54, 55). The inflammatory responses in

both RA and EAE are promoted by TH17 cells
and prevented by Tregs, which suggests that the
effects of gut bacteria on the adaptive immune
system likely extend beyond the gastrointestinal
tract to influence autoimmune diseases that are
seemingly unrelated to microbial infections.
Why only specific commensal bacteria in-
duce TH17 cell differentiation remains unclear.
TH17 responses are critical at mucosal surfaces to
control infections by extracellular pathogens. IL-
17 production recruits neutrophils to the site of
infection and induces antimicrobial peptide ex-
pression and other mediators of immunity. If there
is an evolutionary rationale for the ability of
SFBs to induce TH17 cell differentiation, one in-
terpretation is that they mediate a state of “controlled
inflammation” that prepares the gastrointestinal
tract for an invading pathogen. SFBs are not
overt pathogens and colonize animals as sym-
bionts, and thus TH17 induction may lead to
more enhanced immune responses that protect
against acute infectious agents (such as C.
rodentium). Besides this beneficial outcome, it
appears that SFB colonization also leads to ad-
verse host effects. TH1 and TH17 cells of the adapt-
ive immune system promote autoimmunity. As a
result, microbes that stimulate T helper cell de-
velopment may (inadvertently) also increase the
inherent immune reactivity of the host, potential-
ly leading to host-destructive pathologies medi-
ated by the adaptive immune system. This notion
is supported by a role for SFBs in promoting RA
and EAE during induced animal models, both of
which involve TH17 cell inflammation (54, 55).
REVIEW
The enhancement of RA and EAE by SFBs
establishes that the microbiota can adversely
influence autoimmune disease outside the gut.
Therefore, SFBs can colonize healthy animals
without causing illness; however, when the host
is immunocompromised or under inflammatory
conditions, SFBs can be detrimental.
We propose that certain microbes, such as
SFBs, that can peacefully coexist with a healthy
host but still retain pathogenic potential be termed
“pathobionts” to distinguish them from opportun-
istic pathogens that are acquired from the envi-
ronment and cause acute infections (56). Pathobionts
may represent microorganisms on the evolution-
ary continuum between acute pathogens and com-
mensal microbes, whose sustained relationships
with the host induce the development of additional

Downloaded from www.sciencemag.org on December 23, 2010

Human microbiome Human genome TH17/ Treg profile
SFB
B. fragilis
A
Treg TH 17
Healthy
microbiome + = Health

B
Treg
TH 17
Dysbiosis
(increased pro-
inflammatory
+ = Disease
bacteria)

C
Treg TH 17
Dysbiosis
(decreased anti- + = Disease
inflammatory
bacteria)

B cell TH 17 cell
Treg cell IL-10+ Treg cell

Fig. 2. How the microbiome and the human genome contribute to
inflammatory disease. In a simplified model, the community composition of
the human microbiome helps to shape the balance between immune
regulatory (Treg) and proinflammatory (TH17) T cells. The molecules produced
by a given microbiome network work with the molecules produced by the
human genome to determine this equilibrium. (A) In a healthy microbiome,
there is an optimal proportion of both pro- and anti-inflammatory organisms
(represented here by SFBs and B. fragilis), which provide signals to the
developing immune system (controlled by the host genome), leading to a
balance of Treg and TH17 cell activities. In this scenario, the host genome can
contain “autoimmune-specific” mutations (represented by the stars), but
disease does not develop. (B and C) The genomes of patients with multiple
sclerosis, type I diabetes, rheumatoid arthritis, and Crohn’s disease contain a
spectrum of variants that are linked to disease by genome-wide association
studies [reviewed in (63)]. Environmental influences, however, are risk factors
in all of these diseases. Altered community composition of the microbiome
due to life-style, known as dysbiosis, may represent this disease-modifying
component. An increase in proinflammatory microbes (for example, SFBs in
animal models) may promote TH17 cell activity to increase and thus predispose
genetically susceptible people to TH17-mediated autoimmunity (B). Alterna-
tively, a decrease or absence in anti-inflammatory microbes—for example, B.
fragilis in animal models—may lead to an underdevelopment of Treg cell
subsets (C). The imbalance between TH17 cells and Tregs ultimately leads to
autoimmunity.

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1771

 REVIEW

B Expansion of Treg subsets

by commensals

D Modern adaptive
immune system
A Primordial adaptive
immune system

Downloaded from www.sciencemag.org on December 23, 2010

C Development of TH 17 cells
by pathobionts

B. fragilis Treg cell TGF␤ TH 17 cell

SFB

TH 17
B cell TH 17 cell inducing cytokines

Treg cell IL-10+ Treg cell

Fig. 3. A model for the coevolution of adaptive immunity with the microbiota.
(A) The adaptive immune system develops under the control of the vertebrate
genome to produce various cell types. The evolutionarily ancient molecule TGFb
directs the differentiation of Foxp3+ Treg cells. Although the earliest mammals
contained a gut microbiota, bacteria may or may not have influenced features
of the primordial adaptive immune system. (B) Over millennia of coevolution,
commensal microbes (B. fragilis used as an example here) produced molecules
that networked with the primordial immune system to help expand various Treg
cell subsets (for example, IL-10–producing Foxp3+ Treg cells). This process may
have evolved to allow these microorganisms to colonize the gut by inducing
antigen-specific tolerance to the microbiota. (C) Proinflammatory pathobionts
(such as SFBs) may have induced TH17 cell differentiation to increase mucosal
defenses against enteric pathogens. (D) The modern adaptive immune system
may have arisen from two distinct events: Tregs and TH17 cell types evolved
independently [(A) to (B) and (A) to (C)] or through the sequential development
of TH17 cells from Treg cell precursors [(A) to (B) to (C) to (D)]. This may have
been achieved by a combinatorial signal of TGFb, augmented by the addition of
IL-6 to promote TH17 cell evolution over time (inset). Together, the modulation
of Tregs and TH17 cells by commensal microorganisms and pathobionts, re-
spectively, appears to shape the immune status of the host and thus represents a
possible risk factor for autoimmune diseases that appear to depend on balanced
Treg-TH17 proportions.

layers of mucosal defense while promoting the
unwanted side effect of autoimmune disease. The
importance of TH17 cell–inducing microorganisms
(such as SFBs) to animal models of autoimmunity
remains to be further established; caveats exist, such
as the fact that animals from colonies devoid of
SFBs can develop autoimmune disease. Also, it
remains to be determined how the microbiota may
contribute to human autoimmunity. SFB coloniza-
tion of animals, however, does provide a model
system for testing concepts linking specific gut
bacteria to nonintestinal immune disorders. The
identification of bacterial molecules required for
SFBs to induce TH17 cell responses may reveal
why this particular microorganism is capable of
promoting the development of proinflammatory T
cells. Furthermore, studies that delineate the gene
regulatory networks induced by SFB colonization
may enhance our understanding of the evolutionary
forces that resulted in TH17 lineage development.
Autoimmune diseases such as MS, T1D, and
RA are associated with a spectrum of genetic poly-
morphisms, as shown by recent genome-wide
association studies. Given that concordance rates
for disease among monozygotic twins are 20 to
40% on average, environmental factors are cru-
cial for the manifestations of symptoms (57). We
predict that autoimmunity can result from the com-
bination of an altered human genome and an
altered microbiome (Fig. 2). Patients with auto-
immunity likely have a genetic landscape that
predisposes them to self-reactivity, and in some
cases, certain gut bacteria may promote disease
by activating the adaptive immune system. Po-
tential future treatments for autoimmunity may
include treatment of dysbiosis, because whereas
the human genome is static and intransigent to
manipulation, the microbiome is conceivably
more amenable to therapeutic alterations. Under-
standing the molecular mechanisms of how sym-
biotic microbes affect immune reactions to self
antigens may provide insight into the causes, and
potential cures, for autoimmune diseases.
Did the Microbiota Influence the Evolution
of Adaptive Immunity?
The adaptive immune system distinguishes be-
tween self and foreign antigens and mounts an
appropriate response to clear invading pathogens
by recognizing non-self molecules. The micro-
biota presents a challenge to the adaptive im-
mune system because it contains an enormous
foreign antigenic burden, which must be either
ignored or tolerated to maintain health. One hy-
pothesis for how this occurs is “immunologic
ignorance,” whereby spatial separation of bacteria
from the immune system or down-modulation of
innate immunity prevents overt inflammation
(58). This notion rests on the inability of the in-
nate immune system to distinguish pathogens
from symbionts because they share similar mo-
lecular patterns (such as TLR ligands). Rather
than ignorance, tolerance could also be induced
by the microbiota, given the capacity of gut bac-
teria to induce Treg lineage differentiation. Mol-
ecules produced by our microbiome may be
considered “self,” because inflammatory bowel
disease is thought, in part, to involve a loss of
tolerance to antigens of the microbiota. Therefore,
it appears that we may tolerate the microbiota in
the same way that we tolerate antigens encoded
by our own genome. This then raises the question

1772 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


of whether symbiotic bacteria evolved mecha-
nisms to suppress unwanted inflammation to-
ward the microbiota by actively inducing mucosal
tolerance. Several studies now suggest this to be
the case (32, 33, 42). The necessity for the mi-
crobiota to induce tolerance as a requirement for
colonization, if true, provides a rationale for why
symbiotic bacteria may have influenced critical
aspects of the adaptive immune system through-
out mammalian evolution.
Although T cells can adopt numerous effector
cell fates (such as TH1, TH2, TH3, TH9, etc.),
there are common mechanistic foundations to
TH17 and Treg cell development. The differen-
tiation of both lineages is promoted by trans-
forming growth factor b (TGFb); Tregs require
TGFb (and retinoic acid), whereas TGFb and IL-
6 promote TH17 development (23). The central
transcription factors for Treg cells and TH17 cells
(Foxp3 and RORgt, respectively) are coexpressed
in naïve and effector CD4+ T cells, physically
interact with each other, and differentially
respond to cytokine stimulation to help determine
lineage commitment between Foxp3+ Treg and
TH17 differentiation (59). Furthermore, TH17
cells can develop from Foxp3+ Treg cell pre-
cursors (60). As mentioned above, germ-free
animals show decreased TH17 cell development
in several anatomic locations (39, 46, 54, 55),
and recolonization with a microbiota (containing
SFBs) promotes TH17 cells (47, 54, 55).
On the basis of this knowledge, we propose a
hypothetical model for how specific commensal
bacteria network with an evolving adaptive im-
mune system (Fig. 3). Thymically derived Foxp3+
Treg cells developed under the control of the
evolutionarily ancient molecule TGFb. It is
tempting to speculate that commensal microbes
(for example, B. fragilis) may have “learned” to
augment this process by further promoting the
differentiation of existing Treg cells into
expanded subsets, such as those producing IL-
10 at mucosal surfaces. This expanded Treg cell
repertoire could have provided the host with a
mechanism to tolerate foreign antigens of the
microbiota. Pathobionts such as SFBs may have
further modified the development of adaptive
immunity by promoting the differentiation of
TH17 cells, in part from Foxp3+ precursors. In
support of this notion, IL-17 family members
have been found only in vertebrates (61). Other
proinflammatory cytokines (IL-6, IL-21, and IL-1)
along with TGFb are required to induce IL-17
production, perhaps suggesting that TH17 cells
might be a more recent invention than Treg cells.
Perhaps the evolution of specific immune
responses was not mainly driven by pathogens
(as is popular assumption), but instead by
organisms that developed more sustained rela-
tionships with the host such as commensals and
pathobionts. As these new symbiotic relationships
were forged through host-microbial coevolution,
novel additions to the immune system were
introduced over millennia. Treg and TH17 cells
provide a powerful means by which mucosal
www.sciencemag.org
surfaces can be protected from unwanted inflam-
matory responses to the microbiota (by Treg cells)
while still being capable of potently responding to
microbial infections (with TH17 cells). B. fragilis
and SFBs represent model organisms that have
been experimentally validated to provide these
functions; other microbes may possess similar
activities. Although this concept requires experi-
mental validation, the coordination of Treg and
TH17 responses appears ideally suited to benefit
both the microbiota and the host, and may rep-
resent an important evolutionary partnership for
human health.
Co-opting an antigen-specific adaptive im-
mune system by the microbiota may extend be-
yond simply a host-derived process for controlling
microbial infections. During cohabitation with
the microbiota, evolution of the vertebrate ge-
nome occurred under the influence of signals
from symbiotic bacteria. In fact, the evolutionary
forces that contributed to immune system devel-
opment during lifelong microbial associations
may be dominant relative to those of transient
encounters by microbial pathogens (which are
rare and opportunistic) (62). Thus, symbiotic mi-
crobes may have influenced features of adaptive
immune system evolution and function more pro-
foundly than pathogens, possibly to protect both
host and microbiota from invading infections.
As an increasing body of knowledge links the
microbiota to Treg and TH17 phenotypes that
mediate autoimmunity, it is imperative to deter-
mine how signals from the microbiota shape gene
regulatory networks within CD4+ T cells after
their thymic development. If these co-
evolutionary interactions are relatively recent
inventions, is autoimmunity an unwanted side
effect from fine-tuning of the peripheral adaptive
immune system by the microbiota? As the
influences of the microbiota on autoimmune
diseases are unraveled, it can be envisioned that
harnessing the ability of the microbiota to induce
tolerance through Treg cells may provide novel
treatments for autoimmunity by correcting im-
munologic imbalances found in an evolving adapt-
ive immune system. Finally, because we harbor
10 times as many bacterial cells as human cells,
explorations into how the microbiota may have
influenced the evolution of adaptive immunity
might redefine how we view our “microbial selves.”
References and Notes
1. P. J. Turnbaugh et al., Nature 449, 804 (2007).
2. J. Qin et al., Nature 464, 59 (2010).
3. P. G. Falk, L. V. Hooper, T. Midtvedt, J. I. Gordon,
Microbiol. Mol. Biol. Rev. 62, 1157 (1998).
4. A. J. Macpherson, N. L. Harris, Nat. Rev. Immunol. 4,
478 (2004).
5. M. Pollard, N. Sharon, Infect. Immun. 2, 96 (1970).
6. H. Hoshi et al., Tohoku J. Exp. Med. 166, 297 (1992).
7. J. R. Glaister, Int. Arch. Allergy Appl. Immunol. 45,
719 (1973).
8. A. Lundin et al., Cell. Microbiol. 10, 1093 (2008).
9. S. Matsumoto, H. Setoyama, Y. Umesaki,
Gastroenterology 103, 1777 (1992).
10. A. Imaoka et al., Eur. J. Immunol. 26, 945 (1996).
11. Y. Umesaki, H. Setoyama, S. Matsumoto, Y. Okada,
Immunology 79, 32 (1993).
SCIENCE VOL 330 24 DECEMBER 2010
REVIEW
12. J. Duan, H. Chung, E. Troy, D. L. Kasper,
Cell Host Microbe 7, 140 (2010).
13. J. H. Niess, F. Leithäuser, G. Adler, J. Reimann,
J. Immunol. 180, 559 (2008).
14. D. Bouskra et al., Nature 456, 507 (2008).
15. M. C. Noverr, G. B. Huffnagle, Trends Microbiol. 12,
562 (2004).
16. S. K. Mazmanian, C. H. Liu, A. O. Tzianabos, D. L. Kasper,
Cell 122, 107 (2005).
17. H. Bauer et al., Am. J. Pathol. 42, 471 (1963).
18. K. Smith, K. D. McCoy, A. J. Macpherson,
Semin. Immunol. 19, 59 (2007).
19. T. B. Clarke et al., Nat. Med. 16, 228 (2010).
20. H. Inagaki, T. Suzuki, K. Nomoto, Y. Yoshikai,
Infect. Immun. 64, 3280 (1996).
21. B. Stecher et al., PLoS Biol. 5, e244 (2007).
22. C. Lupp et al., Cell Host Microbe 2, 119 (2007).
23. E. Bettelli et al., Nature 441, 235 (2006).
24. J. D. Fontenot et al., Immunity 22, 329 (2005).
25. A. Foussat et al., J. Immunol. 171, 5018 (2003).
26. M. Battaglia, S. Gregori, R. Bacchetta, M. G. Roncarolo,
Semin. Immunol. 18, 120 (2006).
Downloaded from www.sciencemag.org on December 23, 2010
27. S. Sakaguchi et al., Cell 133, 775 (2008).
28. Y. P. Rubtsov et al., Immunity 28, 546 (2008).
29. J. L. Coombes et al., J. Exp. Med. 204, 1757 (2007).
30. M. Boirivant, W. Strober, Curr. Opin. Gastroenterol. 23,
679 (2007).
31. J. L. Coombes, K. J. Maloy, Semin. Immunol. 19, 116 (2007).
32. C. O’Mahony et al., PLoS Pathog. 4, e1000112 (2008).
33. H. Sokol et al., Proc. Natl. Acad. Sci. U.S.A. 105,
16731 (2008).
34. S. K. Mazmanian, D. L. Kasper, Nat. Rev. Immunol. 6,
849 (2006).
35. S. K. Mazmanian, J. L. Round, D. L. Kasper, Nature 453,
620 (2008).
36. U. G. Strauch et al., Gut 54, 1546 (2005).
37. S. Östman, C. Rask, A. E. Wold, S. Hultkrantz, E. Telemo,
Eur. J. Immunol. 36, 2336 (2006).
38. B. Min et al., Eur. J. Immunol. 37, 1916 (2007).
39. I. I. Ivanov et al., Cell Host Microbe 4, 337 (2008).
40. C. Zaph et al., J. Exp. Med. 205, 2191 (2008).
41. V. Gaboriau-Routhiau et al., Immunity 31, 677 (2009).
42. J. L. Round, S. K. Mazmanian, Proc. Natl. Acad.
Sci. U.S.A. 107, 12204 (2010).
43. K. Atarashi et al., Science 10.1126/science.1198469
(2010).
44. R. Dobber, A. Hertogh-Huijbregts, J. Rozing, K. Bottomly,
L. Nagelkerken, Dev. Immunol. 2, 141 (1992).
45. I. I. Ivanov et al., Cell 126, 1121 (2006).
46. K. Atarashi et al., Nature 455, 808 (2008).
47. I. I. Ivanov et al., Cell 139, 485 (2009).
48. J. Snel et al., Can. J. Microbiol. 44, 1177 (1998).
49. G. L. Talham, H. Q. Jiang, N. A. Bos, J. J. Cebra,
Infect. Immun. 67, 1992 (1999).
50. Y. Umesaki, Y. Okada, S. Matsumoto, A. Imaoka,
H. Setoyama, Microbiol. Immunol. 39, 555 (1995).
51. M. C. Noverr, G. B. Huffnagle, Clin. Exp. Allergy 35,
1511 (2005).
52. J. L. Round, S. K. Mazmanian, Nat. Rev. Immunol. 9,
313 (2009).
53. J. Ochoa-Repáraz et al., Mucosal Immunol. 3, 487 (2010).
54. H. J. Wu et al., Immunity 32, 815 (2010).
55. Y. K. Lee et al., Proc. Natl. Acad. Sci. U.S.A. 10.1073/
pnas.1000082107 (2010).
56. J. Chow, S. K. Mazmanian, Cell Host Microbe 7, 265 (2010).
57. S. E. Baranzini et al., Nature 464, 1351 (2010).
58. L. V. Hooper, Nat. Rev. Microbiol. 7, 367 (2009).
59. L. Zhou et al., Nature 453, 236 (2008).
60. X. O. Yang et al., Immunity 29, 44 (2008).
61. C. T. Weaver, R. D. Hatton, P. R. Mangan,
L. E. Harrington, Annu. Rev. Immunol. 25, 821 (2007).
62. M. McFall-Ngai, Nature 445, 153 (2007).
63. S. E. Baranzini, Curr. Opin. Immunol. 21, 596 (2009).
64. We thank members of the Mazmanian laboratory for their
critical review of the manuscript. Supported by NIH grants
DK078938, DK083633, and AI088626; the Damon Runyon
Cancer Research Foundation; and the Crohn’s and Colitis
Foundation of America (S.K.M.).
10.1126/science.1195568
1773
 BREVIA
ance mechanisms may be critically important in the
development of AD (6). Estimates based on a 30%
Decreased Clearance of CNS decrease in Ab clearance rates suggest that brain Ab

b-Amyloid in Alzheimer’s Disease

accumulates over about 10 years in AD. The im-
paired clearance of both Ab40 and Ab42 is con-
sistent with prior findings of deposition of Ab40
Kwasi G. Mawuenyega,1 Wendy Sigurdson,1,2 Vitaliy Ovod,1 Ling Munsell,1 Tom Kasten,1 and Ab42 in parenchymal amyloid plaques and the
John C. Morris,1,2,3 Kevin E. Yarasheski,4 Randall J. Bateman1,2,5* substantial deposition of Ab40 in cerebral amyloid
angiopathy in about 80% of cases of AD (7).
lzheimer’s disease (AD) is characterized P = 0.98). The average clearance rate of Ab42 was Limitations of this study include the relatively

A by increased amounts of soluble and in- slower for AD individuals compared with that for
soluble b-amyloid (Ab), predominantly in cognitively normal controls (5.3%/hour versus
the form of Ab42 in amyloid plaques and Ab40 in 7.6%/hour, P = 0.03), as was the average clear-
small numbers of participants (12 in each group)
and the inability to prove causality of impaired Ab
clearance for AD. In addition to decreased CNS Ab
amyloid angiopathy. The amyloid hypothesis pro- ance rate of Ab40 (5.2%/hour for AD individuals clearance, CSF Ab42 concentrations are decreased
poses that AD is caused by an imbalance between versus 7.0%/hour for controls; P = 0.01). in AD compared with those in controls (fig. S3).
Ab production and clearance (1), resulting in in- To determine the balance of Ab production to Taken together, these may be consistent with de-
creased amounts of Ab in various forms such as clearance rates in AD versus controls, we mea- creased Ab42 clearance (efflux) from the brain to

Downloaded from www.sciencemag.org on December 23, 2010

monomers, oligomers, insoluble fibrils, and pla- sured the ratios of production to clearance (fig. S2). the CSF. However, the relationship between de-
ques in the central nervous system (CNS). High The ratio of Ab42 production to clearance rates creased concentrations of CSF Ab42 and decreased
levels of Ab then initiate a cascade of events cul- was balanced for cognitively normal participants CNS clearance of labeled Ab (fig. S4) is not fully
minating in neuronal damage and death manifesting (0.95); however, because of decreased clearance in understood. Additional possibilities include more
as progressive clinical dementia of the Alzheimer’s the AD participants, there was an imbalance in the than one pool of Ab in CSF, undetected pools of Ab
type (2). Ab42 production to clearance ratio (1.35). Simi- in CSF by enzyme-linked immunosorbent assay
In rare cases of AD, genetic alterations increase larly, we observed an imbalance in the AD Ab40 (e.g., oligomers), or a combined increase in Ab pro-
the production of Ab (3). However, Ab dysregula- production to clearance ratio (1.37) compared with duction with impaired efflux from parenchyma to
tion in the far more common late-onset “sporadic” the ratio in cognitively normal participants (0.99). CSF. Overall, these results suggest impaired metab-
AD is less well understood. Possible mechanisms The technique of measuring Ab production and olism of Ab in AD compared with that in controls.
of increased Ab production for late-onset AD in- clearance has been used to measure effects of drugs
References and Notes
clude alterations in gamma or beta secretase activ- that target Ab generation, demonstrating decreases 1. J. Hardy, D. J. Selkoe, Science 297, 353 (2002).
ity. Alternatively, impaired clearance of Ab may in production (5). We found that late-onset AD is 2. J. L. Cummings, N. Engl. J. Med. 351, 56 (2004).
also cause late-onset AD through interactions with associated with a 30% impairment in the clearance 3. D. Scheuner et al., Nat. Med. 2, 864 (1996).
4. R. J. Bateman et al., Nat. Med. 12, 856 (2006).
ApoE4, decreased catabolism of Ab via reduced of both Ab42 and Ab40, indicating that Ab clear- 5. R. J. Bateman et al., Ann. Neurol. 66, 48 (2009).
proteolysis, impaired transport 6. R. B. DeMattos et al., Neuron 41, 193 (2004).
A B
across the blood-brain barrier, 7. R. J. Ellis et al., Neurology 46, 1592 (1996).
or impaired cerebrospinal flu- 8. We are grateful to the participants for their time and effort.
The authors thank J. X. Wang for gas chromatography–mass
id (CSF) transport. spectrometry (MS) analysis and R. Connors and R. Potter
To measure the production for immunoprecipitation-MS analysis. Special thanks to
and clearance of Ab in AD, we D. M. Holtzman for mentoring, support, and review of the
developed a method to measure manuscript. This work was supported by grants from NIH
(nos. K08 AG027091, K23 AG030946, R01 NS065667, P50
human CNS Ab production and C D AG05681, P01 AG03991, UL1 RR024992, P41 RR000954,
clearance (fig. S1) (4) and com- P60 DK020579, and P30 DK056341), an anonymous
pared Ab42 and Ab40 pro- foundation, Eli Lilly research, the Knight Initiative for
duction and clearance rates in Alzheimer’s Research, and the James and Elizabeth
McDonnell Fund for Alzheimer’s Research. R.J.B. and
individuals with symptomatic
D. M. Holtzman are cofounders of a company (C2N
AD and in cognitively normal Diagnostics) that has licensed a pending Washington
persons to determine whether University patent on the technology described in this article.
either or both are altered in AD.
We plotted the average time E Supporting Online Material
F www.sciencemag.org/cgi/content/full/science.1197623/DC1
course results of labeled Ab42 Materials and Methods
and Ab40 for the production Figs. S1 to S4
phase (hours 5 to 14) and the References
clearance phase (hours 24 to 10 September 2010; accepted 11 November 2010
36) (Fig. 1). The production and Published online 9 December 2010;
clearance rates were calculated 10.1126/science.1197623
for each participant and com- 1
Department of Neurology, Washington University School of
pared by group status (AD ver- Medicine, St. Louis, MO 63110, USA. 2Alzheimer’s Disease
sus control). The average Ab42 Research Center, Washington University School of Medicine,
production rate did not differ Fig. 1. Ab kinetics in the CNS of 12 AD participants (red triangles) and 12 St. Louis, MO 63110, USA. 3Department of Pathology and Im-
between the control (6.7%/hour) controls (blue circles). The amount of labeled Ab42 and Ab40 was mea- munology, Washington University School of Medicine, St. Louis,
and AD (6.6%/hour) groups sured and compared between groups to measure production and clearance MO 63110, USA. 4Department of Medicine, Washington Uni-
versity School of Medicine, St. Louis, MO 63110, USA. 5Hope
(P = 0.96), nor did Ab40 pro- rates of both Ab species. Error bars indicate SEM. (A) Normalized labeled Center for Neurological Disorders, Washington University School
duction rate differ between Ab42 production phase. (B) Ab42 clearance phase. (C) Normalized labeled of Medicine, St. Louis, MO 63110, USA.
groups (6.8%/hour for controls Ab40 production phase. (D) Ab40 clearance phase. (E) Fractional synthesis *To whom correspondence should be addressed. E-mail:
and 6.8%/hour for the AD group; rates of Ab42 and Ab40. (F) Fractional clearance rates of Ab42 and Ab40. batemanr@wustl.edu

1774 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

 RESEARCH ARTICLES
genes as a human and all of the information
necessary to specify the major tissues and cell
types of metazoans.
Integrative Analysis of the From the project start in 2007 (2), the C.
elegans modENCODE groups had by February
Caenorhabditis elegans Genome 2010 collected 237 genome-wide data sets (table
S1) bearing on gene structure, RNA expression

by the modENCODE Project profiling, chromatin structure and regulation, and

evolutionary conservation. To ensure the complete-
ness and standardization of modENCODE data,
Mark B. Gerstein,1,2,3*† Zhi John Lu,1,2* Eric L. Van Nostrand,4* Chao Cheng,1,2* all data sets were submitted to the modENCODE
Bradley I. Arshinoff,5,6* Tao Liu,7,8* Kevin Y. Yip,1,2* Rebecca Robilotto,1* Andreas Rechtsteiner,9* Data Coordinating Center; hand curated with ex-
Kohta Ikegami,10* Pedro Alves,1* Aurelien Chateigner,11* Marc Perry,5* Mitzi Morris,12* tensive, structured metadata; validated for com-
Raymond K. Auerbach,1* Xin Feng,5,22* Jing Leng,1* Anne Vielle,13* Wei Niu,14,15* pleteness; and checked for consistency before
Kahn Rhrissorrakrai,12* Ashish Agarwal,2,3 Roger P. Alexander,1,2 Galt Barber,16 Cathleen M. Brdlik,4 release at www.modencode.org.
Jennifer Brennan,10 Jeremy Jean Brouillet,4 Adrian Carr,11 Ming-Sin Cheung,13 Hiram Clawson,16 Analyses of these data reveal (i) directly sup-
Sergio Contrino,11 Luke O. Dannenberg,17 Abby F. Dernburg,18 Arshad Desai,19 Lindsay Dick,38 ported protein-coding genes containing 5′ and 3′

Downloaded from www.sciencemag.org on December 23, 2010

Andréa C. Dosé,18 Jiang Du,3 Thea Egelhofer,9 Sevinc Ercan,10 Ghia Euskirchen,14 Brent Ewing,20 ends and alternative splice junctions; (ii) sets of
Elise A. Feingold,21 Reto Gassmann,19 Peter J. Good,21 Phil Green,20 Francois Gullier,11 noncoding RNAs, including RNAs belonging to
Michelle Gutwein,12 Mark S. Guyer,21 Lukas Habegger,1 Ting Han,23 Jorja G. Henikoff,24 known classes and previously unknown types; (iii)
Stefan R. Henz,29 Angie Hinrichs,16 Heather Holster,17 Tony Hyman,26 A. Leo Iniguez,17 gene expression and transcription factor (TF)–
Judith Janette,15 Morten Jensen,10 Masaomi Kato,28 W. James Kent,16 Ellen Kephart,5 binding profiles across developmental stages; (iv)
Vishal Khivansara,23 Ekta Khurana,1,2 John K. Kim,23 Paulina Kolasinska-Zwierz,13 Eric C. Lai,30 genomic locations bound by many of the TFs
Isabel Latorre,13 Amber Leahey,20 Suzanna Lewis,31 Paul Lloyd,5 Lucas Lochovsky,1 analyzed, designated as HOT (high-occupancy
Rebecca F. Lowdon,21 Yaniv Lubling,32 Rachel Lyne,11 Michael MacCoss,20 Sebastian D. Mackowiak,33 target) regions; (v) a hierarchy of candidate regu-
Marco Mangone,12 Sheldon McKay,34 Desirea Mecenas,12 Gennifer Merrihew,20 latory interactions among TFs and its relationship
David M. Miller III,27 Andrew Muroyama,19 John I. Murray,20 Siew-Loon Ooi,24 Hoang Pham,18 to the network of microRNAs (miRNAs) and their
Taryn Phippen,9 Elicia A. Preston,20 Nikolaus Rajewsky,33 Gunnar Rätsch,25 Heidi Rosenbaum,17 targets; (vi) differences in histone modifications
Joel Rozowsky,1,2 Kim Rutherford,11 Peter Ruzanov,5 Mihail Sarov,26 Rajkumar Sasidharan,2 and nuclear-envelope interactions between the
Andrea Sboner,1,2 Paul Scheid,12 Eran Segal,32 Hyunjin Shin,7,8 Chong Shou,1 Frank J. Slack,28 centers and arms of autosomes and between auto-
Cindie Slightam,35 Richard Smith,11 William C. Spencer,27 E. O. Stinson,31 Scott Taing,7 somes and the X chromosome; (vii) evidence
Teruaki Takasaki,9 Dionne Vafeados,20 Ksenia Voronina,19 Guilin Wang,15 Nicole L. Washington,31 for chromatin-mediated epigenetic transmission
Christina M. Whittle,10 Beijing Wu,35 Koon-Kiu Yan,1,2 Georg Zeller,25,36 Zheng Zha,5 Mei Zhong,14 of the memory of gene expression from adult
Xingliang Zhou,10 modENCODE Consortium,‡ Julie Ahringer,13† Susan Strome,9† germ cells to embryos; and (viii) predictive mod-
Kristin C. Gunsalus,12,37† Gos Micklem,11† X. Shirley Liu,7,8† Valerie Reinke,15† Stuart K. Kim,4,35† els that relate chromatin state to TF-binding sites
LaDeana W. Hillier,20† Steven Henikoff,24† Fabio Piano,12,37† Michael Snyder,4,14† and to expression levels of protein- and miRNA-
Lincoln Stein,5,6,34† Jason D. Lieb,10† Robert H. Waterston20† encoding genes.
The summation of features annotated through
these functional data sets provides a potential
We systematically generated large-scale data sets to improve genome annotation for the nematode explanation for most of the conserved sequences
Caenorhabditis elegans, a key model organism. These data sets include transcriptome profiling in the C. elegans genome and lays the foundation
across a developmental time course, genome-wide identification of transcription factor–binding for further study of how the genome of a multi-
sites, and maps of chromatin organization. From this, we created more complete and accurate cellular organism accurately directs development
gene models, including alternative splice forms and candidate noncoding RNAs. We constructed and maintains homeostasis.
hierarchical networks of transcription factor–binding and microRNA interactions and discovered
chromosomal locations bound by an unusually large number of transcription factors. Different The Transcriptome
patterns of chromatin composition and histone modification were revealed between chromosome Accurate and comprehensive annotation of all
arms and centers, with similarly prominent differences between autosomes and the X chromosome. RNA transcripts (the transcriptome) provides a
Integrating data types, we built statistical models relating chromatin, transcription factor binding, and framework for interpreting other genomic features,
gene expression. Overall, our analyses ascribed putative functions to most of the conserved genome. such as TF-binding sites and chromatin marks.
At the project’s inception [WS170; WormBase
omplete genome sequences provide a ila melanogaster so as to systematically annotate versions used for specific analyses can be found

C view of the full instruction set of an or-

ganism. However, understanding the
functional content of a genome requires more
than DNA sequence. To address this need, in
2003 the U.S. National Human Genome Re-
search Institute (NHGRI) initiated the Encyclo-
pedia of DNA Elements (ENCODE) project in
order to study the human genome in greater depth
(1). Recognizing the importance of well-annotated
model genomes, in 2007 the NHGRI initiated
the model organism ENCODE (modENCODE)
project on Caenorhabditis elegans and Drosoph-
the functional genomic elements in these orga-
nisms (2).
Given its intermediate complexity between
single-celled eukaryotes and mammals, C. elegans
offers an outstanding system for studies of ge-
nome organization and function. C. elegans was
the first multicellular organism with a fully de-
fined cell lineage, a nervous system reconstructed
through serial-section electron microscopy, and
a sequenced genome (3–5). Its 100.3-Mb genome
is only about eight times larger than that of S.
cerevisiae, and yet it contains almost as many
in (6)], the C. elegans genome lacked direct ex-
perimental support for about one third of pre-
dicted splice junctions, and some of these
predictions were erroneous (7, 8). Many genes
lacked transcript start sites and polyadenylate
[poly(A)] addition sites; exons and even whole
genes were missing. To address these deficiencies,
cDNA-based evidence was obtained through high-
throughput sequencing (RNA-seq), reverse tran-
scription polymerase chain reaction (RT-PCR)/
rapid amplification of cDNA ends (RACE), and
tiling arrays from a variety of stages, conditions,

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1775

 RESEARCH ARTICLES
and tissues (tables S1, S3, and S4). Analysis of
the data yielded previously unrecognized protein-
coding genes, refined the structure of known
protein-coding genes, revealed the dynamics of
expression and alternative splicing, provided evi-
dence of pseudogene transcription, and suggested
previously unknown noncoding RNAs (ncRNAs).
Through mass spectrometry, we verified pre-
dicted proteins and distinguished short single-
exon protein-coding transcripts from ncRNAs.
Protein-coding genes. We used RNA-seq to
generate more than 1 billion uniquely aligned
short sequence reads from 19 different nematode
populations, including all major developmental
stages (embryonic, larval, dauer, and adult), em-
bryonic and late L4 males, animals exposed to
pathogens, and selected mutants (fig. S3) (9, 10).
Data sets targeting the 3′ ends of poly(A)-plus
transcripts were also collected, and additional
sequence tags representing polyadenylated 3′
ends that were acquired by using 3P-Seq [poly(A)-
position profiling by sequencing] were made
available to the consortium (11, 12).
RNA-seq reads were mapped exhaustively
and, together with the 3P-Seq data, allowed us
to detect with nucleotide resolution features of
protein-coding genes independently of previous
WormBase models (fig. S7). The number of con-
firmed splice junctions increased from 70,028 at
project start to 111,786, with 8174 of these not
previously represented in WormBase (Fig. 1A
and fig. S8). The number of genes with a trans-
spliced leader (either Spliced Leader 1 or 2) at
the 5′ end increased from 6012 to 12,413,
covering 20,515 different trans-spliced transcript
start sites (TSSs), and the number of poly(A)
sites associated with genes increased from 1330
to 28,199 (table S2A) (13). RT-PCR/RACE and
mass spectrometry provided direct support for
40,114 splice junctions (6). About 95% of these
1
Program in Computational Biology and Bioinformatics, Yale
University, Bass 432, 266 Whitney Avenue, New Haven, CT
06520, USA. 2Department of Molecular Biophysics and Bio-
chemistry, Yale University, Bass 432, 266 Whitney Avenue,
New Haven, CT 06520, USA. 3Department of Computer
Science, Yale University, 51 Prospect Street, New Haven, CT
06511, USA. 4Department of Genetics, Stanford University
Medical Center, Stanford, CA 94305, USA. 5Ontario Institute
for Cancer Research, 101 College Street, Suite 800, Toronto,
Ontario M5G 0A3, Canada. 6Department of Molecular Ge-
netics, University of Toronto, 27 King’s College Circle, Toronto,
Ontario M5S 1A1, Canada. 7Department of Biostatistics and
Computational Biology, Dana-Farber Cancer Institute, 44
Binney Street, Boston, MA 02115, USA. 8Department of
Biostatistics, Harvard School of Public Health, 677 Huntington
Avenue, Boston, MA 02115, USA. 9Molecular, Cell, and De-
velopmental Biology, University of California, Santa Cruz, Santa
Cruz, CA 95064, USA. 10Department of Biology and Carolina
Center for Genome Sciences, University of North Carolina at
Chapel Hill, Chapel Hill, NC 27599, USA. 11Department of
Genetics, University of Cambridge, Cambridge CB2 3EH, UK,
and Cambridge Systems Biology Centre, Tennis Court Road,
Cambridge CB2 1QR, UK. 12Center for Genomics and Systems
Biology, Department of Biology, New York University, 1009
Silver Center, 100 Washington Square East, New York, NY
10003–6688, USA. 13Wellcome Trust/Cancer Research UK
Gurdon Institute, University of Cambridge, Tennis Court Road,
Cambridge CB2 1QN, UK. 14Department of Molecular, Cellu-
1776 24 DECEMBER 2010
overlapped with those detected with RNA-seq,
providing independent support for 37,830 of
these features (fig. S9). In addition, mass spec-
trometry proved that of 359 tested, 73 single-
exon genes produced protein.
We used several avenues to estimate how
many features of protein-coding genes remain
to be supported in C. elegans. Of predicted
WormBase transcripts, only 1108 (5%) do not
have support through RNA-seq (table S2B). Of
these, 369 are members of rapidly evolving gene
families implicated in environmental response
and may be nonfunctional or only expressed un-
der specific conditions. The yield of new features
discovered with additional RNA-seq samples is
clearly diminishing, and features such as newly
discovered exons are approaching saturation
(fig. S10). Intersection of the data sets produced
here with previous evidence from WormBase
suggests as few as 2000 to 3000 exons (2 to 3%)
remain undetected (fig. S10). However, we con-
tinue to detect rare splice-junction and spliced-
leader events, particularly those associated with
more abundantly expressed genes. These could
be biologically important but might also result
from RNA-processing errors.
Gene models. We built probable gene models
from the results of transcript sequencing, allow-
ing for multiple transcripts (isoforms) from a
given region (10). These models, called genelets
because they could be fragments of full genes,
were initiated with the most highly represented
splice junction in a region and extended in each
direction so as to incorporate regions covered
by above-threshold sequence reads and splice
junctions (6). The model was terminated when
either a transcript start or stop signal was encoun-
tered or when coverage was interrupted (fig. S5).
By iterating the process, we generated alternative
isoforms. We used the longest open reading frame
lar, and Developmental Biology, Yale University, New Haven,
CT 06824, USA. 15Department of Genetics, Yale University
School of Medicine, New Haven, CT 06520–8005, USA.
16
Department of Biomolecular Engineering, University of
California, Santa Cruz, Santa Cruz, CA 95064 USA. 17Roche
NimbleGen, 500 South Rosa Road, Madison, WI 53719, USA.
18
Howard Hughes Medical Institute, Department of Molecular
and Cell Biology, University of California, Berkeley, Berkeley,
CA 94720, USA, and Life Sciences Division, Lawrence Berkeley
National Laboratory, Berkeley, CA 94720, USA. 19Ludwig In-
stitute Cancer Research/Department of Cellular and Molecular
Medicine, University of California, San Diego, 9500 Gilman
Drive, La Jolla, CA 92093–0653, USA. 20Department of Ge-
nome Sciences, University of Washington School of Medicine,
William H. Foege Building S350D, 1705 NE Pacific Street, Post
Office Box 355065, Seattle, WA 98195–5065, USA. 21Division
of Extramural Research, National Human Genome Research
Institute, National Institutes of Health, 5635 Fishers Lane, Suite
4076, Bethesda, MD 20892–9305, USA. 22Department of
Biomedical Engineering, State University of New York at
Stonybrook, Stonybrook, NY 11794, USA. 23Life Sciences Institute,
Department of Human Genetics, University of Michigan, 210
Washtenaw Avenue, Ann Arbor, MI 48109–2216, USA. 24Basic
Sciences Division, Fred Hutchinson Cancer Research Center, 1100
Fairview Avenue North, Seattle, WA 98109, USA. 25Friedrich
Miescher Laboratory of the Max Planck Society, Spemannstrasse
39, 72076 Tübingen, Germany. 26Max Planck Institute of Mo-
lecular Cell Biology and Genetics, Pfotenhauerstrasse 108, 01307
VOL 330 SCIENCE www.sciencemag.org
to annotate protein-coding sequences (CDSs) and
5′ and 3′ untranslated regions (UTRs).
For each of the 19 stages and conditions, we
built transcript sets purely on the basis of RNA-seq
data from a given stage (stage-specific RNA-seq–
only genelets), along with three aggregate sets:
(i) aggregate RNA-seq–only genelets; (ii) aggre-
gate integrated genelets, which combined RNA-
seq data with available ESTs (expressed sequence
tags), cDNAs, and OSTs (open reading-frame
sequence tags) (7, 8, 11), as well as the RT-PCR/
RACE and mass spectrometry data produced in
the project; and (iii) aggregate integrated tran-
scripts, which incorporates all evidence from
“(ii)” above and allows WormBase predictions
to fill small coverage gaps within exons. The last
set incorporates all of the splice junctions and
spliced-leader sites, as well as multiple poly(A)
Downloaded from www.sciencemag.org on December 23, 2010
addition sites, and thus often contains multiple
isoforms. Altogether, we generated 64,824 tran-
scripts from 21,733 genes, as compared with
23,710 transcripts from 20,082 genes in WormBase
at the project start. Our gene models, which
come from direct experimental evidence, exactly
match the internal splice junction pattern for
10,123 WormBase transcripts, but we provide
revised 5′- or 3′UTRs for many of these. For 6418
models, the internal gene structure was unchanged
from WormBase, but new 5′ or 3′ exons and as-
sociated splice junctions were added. The re-
maining fall into three categories: Our models
overlap WormBase transcripts but differ in splice
junctions (3292); they fail to cover all of the splice
junctions (2235); or they are not represented in
WormBase at all (1952).
Expression dynamics. To determine the dy-
namics of gene expression during development
and in specific cell types, we analyzed tiling ar-
ray data from 42 biological samples, comprising
17 different growth stages and conditions from
Dresden, Germany. 27Department of Cell and Developmental
Biology, Vanderbilt University, 465 21st Avenue South, Nash-
ville, TN 37232–8240, USA. 28Department of Molecular,
Cellular and Developmental Biology, Post Office Box 208103,
Yale University, New Haven, CT 06520, USA. 29Max Planck
Institute for Developmental Biology, Spemannstrasse 37-39,
72076 Tübingen, Germany. 30Sloan-Kettering Institute, 1275
York Avenue, Post Office Box 252, New York, NY 10065, USA.
31
Genomics Division, Lawrence Berkeley National Laboratory, 1
Cyclotron Road, Mailstop 64-121, Berkeley, CA 94720 USA.
32
Department of Computer Science and Applied Mathematics,
Weizmann Institute of Science, Rehovot, 76100, Israel. 33Max-
Delbrück-Centrum für Molekulare Medizin, Division of Systems
Biology, Robert-Rössle-Strasse 10, D-13125 Berlin-Buch, Ger-
many. 34Cold Spring Harbor Laboratory, 1 Bungtown Road, Cold
Spring Harbor, NY 11542 USA. 35Department of Developmental
Biology, Stanford University Medical Center, 279 Campus Drive,
Stanford, CA 94305–5329, USA. 36European Molecular Biology
Laboratory, 69117 Heidelberg, Germany. 37New York University,
Abu Dhabi, United Arab Emirates. 38David Rockefeller Graduate
Program, Rockefeller University, 1230 York Avenue New York,
NY 10065, USA.
*These authors contributed equally to this work.
†To whom correspondence should be addressed. E-mail:
modencode.worm.pi@gersteinlab.org
‡The modENCODE Consortium is a group of NHGRI-funded
investigators defining genomic elements in C. elegans and
D. melanogaster.
 RESEARCH ARTICLES
whole animals, and 25 samples from different files at a given stage with all other stages. For simple alternative exons to more complicated pat-
isolated cell and tissue types (table S3) (6). For simplicity, we focused on a set of 8428 genes with terns, such as splicing or retention of an entire series
almost all whole-animal samples, RNA-seq data non-overlapping transcripts and found that pro- of introns in different stages (Fig. 1C and fig. S6).
were also obtained from the same or similarly files over the time course cluster into distinct em- We also developed algorithms that infer quan-
prepared samples. Calibration and processing bryo and larval phases (Fig. 2A) (6). This division titative transcript-level expression by distribut-
were done to facilitate the integration of se- was consistent with a principal-components anal- ing sequence reads among alternative isoforms
quencing and arrays for both RNA-seq and for ysis on the tiling-array data from matched tis- in a probabilistic manner (6). Pairwise compari-
chromatin immunoprecipitation (ChIP) followed sues from embryo and L2 (Fig. 2C) (6). The RNA sons of staged samples showed that overall, iso-
by high-throughput sequencing (ChIP-seq), al- for the embryos and larvae was isolated through form usage does not change dramatically between
lowing them to be used for a merged data set different procedures, but on the basis of a num- stages: Of 12,875 genes with multiple isoforms,
(figs. S1, S2, and S4) (6, 14). Overall, we found ber of controls and comparisons these differ- 280 on average switch isoform usage between
that only a small number of genes (~100 per ences are unlikely to confound the analysis (6). any two stages, totaling 1324 genes with switch-
stage) showed strong stage-specific expression in Alternative splicing. Alternative mRNA pro- ing (Fig. 1B and fig. S14) (6). Using a different
the whole-animal samples, but fewer than half cessing, including selection of alternative splice approach, we grouped transcript-level expression
of the genes were detectably expressed in all junctions, promoters, or poly(A) addition signals, profiles across many stages into 48 distinct clus-
stages by means of RNA-seq, and tiling arrays provides another mechanism for differential tran- ters (figs. S15 and S16). We identified 1320 genes
suggest that >75% of genes show a greater than script generation. To discover prominent stage- for which one isoform fell into a separate cluster
twofold range of expression across all the tis- specific alternative isoforms among the aggregate from all the others and then classified these

Downloaded from www.sciencemag.org on December 23, 2010

sues (figs. S11 and S12) (15). integrated transcript models, we identified genes according to the type of processing events that
To investigate the relationship between gene with two or more isoforms whose abundance distinguish them (figs. S17 and S18) (6). These
expression and developmental stages in greater changed more than fivefold during development; analyses illustrate the range of alternative mRNA
detail, we correlated the RNA-seq expression pro- differential splice junction usage ranged from processing that takes place during development.

115000
A novel C let-2 Aggregate transcripts
105000 3'
internal 200 bases
5' EE
95000
Confirmed splice junctions

Hillier et al. '09

LE
Project start
85000
L1

75000 L2

L3
65000
L4
55000
YA

45000
D Parent Gene Pseudogene
35000
Y-Scale: [0.0, 3.6] DCPM Y-Scale: [0.0, 8.6] DCPM
Project start
Hillier et al. '09
Agg RNA-seq
Agg total

EE
LE
L1
L2
L3
dauer entry
dauer
dauer exit
L4
YA

L4 male
L4 soma
L1 (lin-35)
aged adult

Hs
Hs ctrl

Sm ctrl
MxE

Sm

L1

L1 (lin-35)
Time course Males & other Pathogens
L2
B
10000

L3

dauer entry (daf-2)

1000

dauer (daf-2)
Number of Genes

280
dauer exit (daf-2)
100

L4

L4 male
10

YA
1

5’ 3’ 5’ 3’
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 9 10
0.0 0.2 0.4 0.6 0.8 1.0 ncRNA
Fractional Differences in Isoform Composition T01B11.7.1 PP00501
Between Stages chrIV: 8,464,749 - 8,472,587 chrIV: 8,461,023 - 8,463,781

Fig. 1. Transcriptome features and alternative splicing. (A) Bar graphs indicate comparisons across seven developmental stages. A fractional difference close to
the number of confirmed splice junctions categorized by type. The leftmost bars 1 indicates large differences in the relative composition. (C) Representative
show the progression from project start (6) to the aggregate integrated tran- example (F01G12.5; let-2), illustrating alternative exon usage across stages.
script set. The three other groups provide data for the various developmental (D) Example of a differentially transcribed pseudogene creating a ncRNA. Rows
stages, males, mutants, and populations exposed to pathogens. Specific sample are normalized signal tracks for the various developmental stages, showing the
names are described in table S3. (B) Histogram of fractional differences in expression pattern of the parent gene (T01B11.7.1; orange) and an associated
isoform composition for 12,875 genes with multiple isoforms in 21 pair-wise duplicated pseudogene (PP00501, green).

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1777

 RESEARCH ARTICLES
Pseudogenes. Several gene models derived
from RNA-seq fell in regions previously anno-
tated as pseudogenes. Pseudogenes are DNA se-
quences similar to protein-coding genes that are
generally thought not to produce functioning pro-
teins (16). However, some pseudogenes are tran-
scribed and may potentially act as endo-siRNA
(endogenous small-interfering RNA) regulators
of their parent genes (17). Using computational
methods, we identified 1293 probable pseudo-
genes in the C. elegans genome, adding 173 to
and removing 213 from the previous annotation
set (WS170), and established the probable source
(parent) gene for 1198 of them (fig. S19) (6).
Using RNA-seq data, we found evidence of tran-
scription for 323 pseudogenes (6). For 191 of
the 323, we determined that the transcription was
clearly independent of the parent gene, ruling out
potential mismapping artifacts. Of these 191, 104
had a discordant expression pattern across stages
relative to the parent (Fig. 1D), and 87 were greater
than two times more expressed than the parent (6).
Intriguingly, 17 of the transcribed pseudogenes
have a unique peptide match through mass spec-
trometry, suggesting that they are translated and
may create novel short peptides.
ncRNAs. The genome produces a variety of
transcripts that do not code for proteins but in-
stead function directly as noncoding RNA
(ncRNA). At the start of the project, there were
1061 known ncRNAs in C. elegans (table S5).
These include small nucleolar RNAs (snoRNAs),
RNAs involved in mRNA translation and splicing
[such as ribosomal RNAs (rRNAs) and tRNAs],
miRNAs, piwi-associated RNAs (piRNAs, called
21U-RNAs in C. elegans), and multiple classes
of endo-siRNAs (18).
To provide a more comprehensive annotation
of small ncRNAs, we profiled small-RNA gene
expression using RNA-seq on size-fractionated
total RNA. In particular, using 81 million aligned
1778 24 DECEMBER 2010
reads from 11 different stages enabled us to iden-
tify 154 out of 174 previously annotated miRNA
genes (19, 20). Most of these are products of the
canonical Drosha-Dicer cleavage pathway. How-
ever, four are mirtrons—miRNAs for which the
precursor hairpins are generated directly by intron
splicing (21). Our computational and experimen-
tal analysis validated 13 previously unidentified
mirtrons (6, 22). Small-RNA data also defined
102 additional candidate canonical miRNAs and
thousands of 21U-RNAs, although these latter
were from previously identified loci (6, 19, 23).
To identify other candidate ncRNAs, partic-
ularly ones longer than those discussed above,
we combined all the transcriptome data sets to
integrate both tiling-array and RNA-seq data.
We found that in comparison to other genomic
“elements” (such as well-curated CDSs, UTRs,
or intergenic regions), the known ncRNAs tend
to have a higher small RNA-seq signal and very
little poly(A)-plus RNA-seq signal. However, no
single transcriptome feature was able to reliably
distinguish them (fig. S21A) (24). Therefore, we
developed a multivariate machine-learning mod-
el combining all the transcriptome data sets and
found support for 21,521 previously unknown
ncRNAs (4.3 Mb in total), which we call the 21k-
set of ncRNAs (tables S6 to S8 and fig. S20) (6).
Because identifying ncRNAs by using tiling
arrays can be problematic (14), we added con-
servation and RNA secondary structure to our
model. However, doing so restricted the predic-
tions of this second model to only the ~15% of
the C. elegans genome that was readily alignable
to C. briggsae. Overall, the second model pre-
dicted 7237 previously unidentified ncRNA can-
didates (the 7k-set, comprising 1.0 Mb), with an
estimated positive-predictive value of 91% (from
testing against an independent validation set of
known ncRNAs) (24). Of these, 1678 ncRNA
candidates (181 kb) fell in intergenic regions,
Pol II–binding in embryonic stages shows low correlation with gene expression in larval and
young adult stages, expression in the embryo correlates moderately well with RNA Pol II–
binding later. (C) Principal components analysis (PCA) of six matched tissue samples from mixed
embryo (MxE) and L2 (7). GABA, g-aminobutyric acid.
VOL 330 SCIENCE www.sciencemag.org
with the remainder in introns, pseudogenes, or
regions antisense to exons (fig. S21B). We tested
a number of these intergenic candidates to vali-
date expression: RT-PCR detected RNA products
for 14 of 15, and Northern blots detected expres-
sion for three of five (24).
The 7k-set contains many RNA structural mo-
tifs, including some not found in known RNA
secondary structure families (24). Additionally,
these ncRNA candidates tend to be differen-
tially expressed across development (24), with
many preferentially expressed in the embryo.
Comparing the expression profiles of the 7k-set
with those of well-characterized genes allowed
us to identify putative functions for some can-
didate ncRNAs (table S9) (6). Lastly, in com-
paring the 7k and 21k sets of ncRNAs the
overlap was small, with just 1259 overlaps. Thus,
Downloaded from www.sciencemag.org on December 23, 2010
when conservation and structure were consi-
dered we detected candidate ncRNAs not found
from the expression data alone; conversely, many
previously uncharacterized transcripts in C. elegans
may occur in nonconserved parts of the genome.
Thus, the 7k and 21k sets provide comple-
mentary types of ncRNA candidates for further
study.
In summary, the improved annotation of tran-
scribed portions of the genome from these data
sets provides the community with new sub-
strates for further experimentation. However,
gaps remain in some transcript models, some
protein-coding genes remain to be discovered,
and direct evidence is needed to support the can-
didate ncRNAs.
Regulatory Sites and Interactions
Accurate annotation of sites bound by TFs is
central to understanding the regulatory networks
underlying development and homeostasis. How-
ever, at the start of the project very few TF-binding
sites had been annotated in the nematode ge-
Fig. 2. Expression and
binding dynamics. (A)
Spearman correlations of
gene expression and RNA
Pol II binding across sev-
en stages. Expression-
level correlations are shown
above the diagonal; RNA
Pol II–binding correla-
tions appear below. For
both expression and bind-
ing, there is a notable
transition between em-
bryonic and larval stages.
(B) Correlation of RNA Pol
II–binding levels with gene
expression. Although RNA

nome, in part because of a lack of suitable meth-
ods with which to assay binding sites in whole
animals (25). We developed these methods
and have applied them to map the binding sites
for 23 green fluorescent protein (GFP)–tagged
fusion proteins and RNA polymerase II (RNA
Pol II) using ChIP-seq (table S10) (6, 26). Most
factors were assayed at their stage of highest ex-
pression, but both PHA-4 (a well-studied factor
required for pharyngeal development) and RNA
Pol II were analyzed at six developmental stages.
Some of the factors were expressed in as few
as 10% of the cells in the whole animal.
TF-binding sites, motifs, and targets. Bind-
ing sites were identified by first finding relative-
ly broad regions of enrichment and then, for some
analyses, refining these to narrow [≤200 base
pairs (bp)] peak summits (figs. S24 and S46).
basis of their proximity to the TSS, we were able
to assign most sites to specific protein-coding or
known ncRNA genes, creating a set of candi-
date targets for each TF (6); however, some
sites were ambiguously located and remain un-
assigned. Although most factors target both
protein-coding and known ncRNA genes, GEI-
11 preferentially targets ncRNAs (Fig. 3D and
fig. S22, A and B). Analysis of TF-binding sites
adjacent to ncRNA candidates from the 7k-set
showed that 59% are potential targets of the 22
TFs examined, which is significantly more than
would be expected by chance (P < 0.001, derived
from a z score assuming a normal distribution
of random sequences) (6, 24). Pairwise correla-
tion of target genes revealed that factors with
related functions often show substantial over-
lap in their protein-coding gene targets (fig. S23A).
RESEARCH ARTICLES
tend to cluster together more by stage than by
factor type (fig. S23B), which is consistent with
observations that expression of miRNAs tends
to show strong stage-specific enrichment (19).
To further characterize TF-binding sites, we
searched for 8- to 12-bp cis-regulatory motifs
within the ChIP-seq peaks (6) and found strong
motifs for eight TFs (BLMP-1, CEH-14, CEH-30,
EGL-5, HLH-1, LIN-39, NHR-6, and PHA-4)
(fig. S35). Two of these are similar to previously
described motifs (PHA-4 and HLH-1).
The binding sites (defined from narrow peaks)
cover a total of 5,165,949 bp (5.2% of the ge-
nome) and target 8859 protein-coding genes, as
well as 652 known ncRNAs, indicating that each
gene may have sites for many factors.
Clustered binding in HOT regions. We iden-
tified 304 short binding regions (average length

Downloaded from www.sciencemag.org on December 23, 2010

Most TF-binding sites defined by means of ChIP-
seq peaks for protein-coding genes lie within
500 bp upstream of transcript start sites. Binding
sites assigned to known ncRNAs are even closer
to the 5′ end of the transcript (fig. S22C). On the
Three homeobox (HOX) genes involved in es-
tablishing the body plan provide particularly
striking examples (mab-5, lin-39, and egl-5)
(26). In contrast, pairwise correlation of targeted
miRNAs shows that the factors bound to them
of ~ 400 bp) that were significantly enriched (q
value < 1e-5) in most TF ChIP-seq experiments
despite the fact that the 22 analyzed factors have
diverse functions and expression patterns. These
regions, which we term HOT regions, were bound

Fig. 3. Integrated miRNA-TF regulatory network. (A) TFs are organized hierar-
chically, and those miRNAs either regulating or being regulated by the TFs are
shown. (TF names are in fig S36.) All larval TF-TF interactions in HOT regions
were removed. Tissue specificity and number of protein-protein interactions are
shown for each of the hierarchical levels (6). (B) TF network after filtering out
edges that do not show a significant correlation in their expression patterns.
Also shown is a schematic representation of the target genes of the 18 larval
TFs. (C) One of the three significantly enriched network motifs (other two are in
fig. S37). (D) Enrichment of binding targets and signal of TFs in noncoding
versus coding genes. Max signal equals the ratio of maximum binding signal of
a TF at noncoding versus coding genes. Target fraction represents the ratio of
target percentage in noncoding genes to that in coding genes (fig. S22A).

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1779

 RESEARCH ARTICLES
by 15 or more factors (Fig. 4, A and B, and fig.
S25A) (6). Control experiments revealed that
these regions are not enriched in input DNA, nor
do they appear in control ChIPs from strains lack-
ing GFP-tagged TFs (fig. S26) (6). The number of
factors bound to HOT regions was relatively in-
sensitive to the width of the peaks used to identify
them because peak summits occur within 100 bp
for over 80% of HOT regions (fig. S25B) (6).
In addition to the HOT regions, most TFs
also cross-link to “factor-specific” DNA regions
(bound by one to four total factors) (Fig. 4A).
Using HLH-1, a typical factor with both known
tissue specificity and a known binding motif, we
compared these two different classes of sites
(HOT and factor-specific) for functional dif-
ferences. HLH-1 drives muscle development in
C. elegans (27) and is associated with 598 factor-
(fig. S27, A and C) (29). Similar enrichment for
motifs and tissue-specific expression of targets
was also observed for other TFs when factor-
specific sites were compared with HOT regions
(fig. S27B) (6), suggesting that factor-specific
and HOT regions are functionally distinct.
Genes associated with HOT regions are dis-
tinguished by several other measures. HOT-region
genes assayed for expression at the individual-
cell level in L1 larvae are expressed in most or
all cell types, whereas other genes mostly showed
tissue-specific expression (Fig. 4C and fig. S29)
(30). Genes associated with HOT regions were
also expressed at higher levels in whole-animal
and tissue-enriched measurements and were less
likely to be stage-specific (fig. S28) (6). Com-
pared with 3% of genes associated with factor-
specific regions, 21% of the HOT region–
enrichment, P < 1e-12) (table S11). In compar-
ison, GO analysis of the remaining (non-HOT)
targeted genes identified functional terms that
are consistent with the known tissue specificity
and function of the individual TFs (26).
Extensive overlap in binding sites between
TFs with disparate functions has previously been
observed in both limited (33) as well as whole-
genome ChIP-chip experiments (34, 35). Using
ChIP-seq data, we have shown that hundreds of
regions in C. elegans are bound by the majority
of TFs within a 100-bp window. Our results sug-
gest that many TFs that are cross-linked to HOT
regions are not directly associated with DNA via
specific binding, which is consistent with findings
for highly occupied regions in Drosophila (34).
Rather, they suggest that association with HOT
regions may be driven by protein-protein interac-

Downloaded from www.sciencemag.org on December 23, 2010

specific and 165 HOT regions. Relative to HOT
regions, factor-specific HLH-1 ChIP-seq regions
were over twofold enriched for the HLH-1–
binding motif (Fisher’s exact test, P < 0.0001)
(28), and genes associated with these regions
were more than ninefold enriched for muscle-
specific expression (Fisher’s exact test, P < 0.01)
associated genes are essential (P < 1e-40; c2 test)
(fig. S27C) (6, 31). Gene Ontology (GO) (32)
analysis revealed a variety of biological pro-
cesses highly represented in HOT-associated
genes, including growth, reproduction, and larval
and embryonic development (each P < 1e-15), as
well as 19 ribosomal protein genes (>12×
tions to a currently unknown set of HOT region–
associated DNA-binding factors. We searched for
sequence motifs that might be broadly associated
with HOT regions and found a few that were
significantly enriched (fig. S35), but the protein
factors that bind directly to these motifs are
currently unknown.

Fig. 4. HOT regions. (A) A HOT Factor-specific B Chromosome

304 HOT regions 50 random
TF-binding peaks at a HOT region regions
regions
region and two “factor- PES-1
# of HOT
regions bound I II III IV V X I II III IV V X
specific regions” on chro- GEI-11 304/304 EOR-1 L3
PQM-1 302 MDL-1 L1
mosome III: 7,206,000 MAB-5 301 SKN-1 L1
to 7,220,000. Top tracks LIN-39 299 PES-1 L4
EGL-5 298 LIN-15B L3
show read density (scaled LIN-15B 294 MEP-1 MxE
293 PHA-4 L1
based on the total mapped EOR-1
291 LIN-39 L3
LIN-11
reads) from 22 ChIP-seq CEH-14
284
268
LIN-13 MxE
CEH-30 LE
experiments. Bottom tracks ALR-1 267 MAB-5 L3
UNC-130 221 EGL-5 L3
show ChIP-seq controls, SKN-1 220 ELT-3 L1
RNA-seq expression lev- MDL-1 220
212
LIN-11 L2
BLMP-1 L1
ELT-3
els, and ChIP-chip signals EGL-27
209 EGL-27 L1
185 CEH-14 L2
for two histone modifica- BLMP-1 175 ALR-1 L2
tions. (B) 304 HOT regions CEH-30 165 HLH-1 MxE
MEP-1 164 PQM-1 L3
bound by 15 or more LIN-13 133 UNC-130 L1
PHA-4 62 GEI-11 L4
factors and 50 randomly 29 DPY-27 MxE
HLH-1
chosen TF-bound regions. EGL-27 IgG Peak significance
Each row represents a TF, Input -log10(q-value)
0 5 10 20 30 40 50 60 70 80 90 100
H3K27ac
and each region is colored H3K4me C
by enrichment q value (6). Pol II Expression in 363 L1 cells
(C) Genes associated with RNA-seq Promoter neu.
Refseq(+)
int. b.w.m. hyp. blast
Contains:
HOT regions are broadly rpl-6 R151.2 A1
expressed. Single-cell gene HOT region(s)
expression measurements of 93 mCherry reporters (30) are shown separated by A4
whether the promoter contains a HOT region, contains a region bound by 10 to 14 B1
Region(s)
factors, or contains only regions bound by 0 to 9 factors (gene names are in fig. S29). bound by
The x axis represents 363 specific cells present in L1-stage animals. 10-14 factors
B7
C1

Only regions
bound by
0-9 factors

C82

log (gene expression)

2

1780 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


Building a TF hierarchy. Following the
assignment of binding sites to target genes,
we investigated the resulting “binding network,”
as had previously been done in yeast and
Escherichia coli (36). The network for 18 fac-
tors assayed in larval stages (Fig. 3, A and B,
and fig. S36) is relatively dense, with each TF
bound to an average of 828 genes, including
TFs and other gene targets. We pruned the
network to the strongest interactions, using the
fact that the expression profile of a TF tends to
be more strongly correlated over the time course
with that of its targets than nontargets, being
positive for activators and negative for repres-
sors (table S12) (6). The pruned network shows
a high level of autoregulation among the factors.
Within the network, we organized TFs hier-
archically according to the degree to which they
target other TFs (top of the hierarchy) or are
themselves targets for other TFs (bottom) (37).
We observed clear differences between the TFs
at each level (Fig. 3, A and B). TFs at the lower
levels tended to be more uniformly expressed
across multiple tissues (P = 0.07, Student’s t
test) (6). Consistent with this, TFs at the bottom
level were essential more often than those at
the top. In contrast, members of the Hox family
were more often at the top of the hierarchy—
among the six Hox TFs examined, four were at
the top layer of nine TFs—perhaps reflecting
their role in modulating specific developmental
processes across multiple tissues. Lastly, TFs
showed connectivity in the existing C. elegans
protein-protein interaction network so that those
at the hierarchy top tended to have significantly
fewer protein-protein interactions than those be-
low (P = 0.002, Student’s t test) (38). This sug-
gests that TFs in the middle and bottom layers act
as “mediators” or “effectors,” more likely to ex-
change information with other proteins. Although
the predicted larval-stage TF network here is
small and one cannot make strong statistical state-
ments, these conclusions follow a pattern that is
consistent with regulatory hierarchies in yeast and
E. coli, in which essential and highly connected
“workhorse” regulators tend to occupy lower levels
whereas overall modulators are on the top (37).
An integrated miRNA-TF network and its mo-
tifs. Next, we added miRNAs to our TF hierarchy
in order to enable us to explore the interplay
between transcriptional and posttranscriptional
regulation. In particular, we identified the targets
of miRNAs on the basis of annotated 3′UTRs
and sequence conservation (table S13) (6). We
then constructed an integrated network between
miRNAs expressed during larval stages and the
above 18 TFs (all assayed in the same stages).
For simplicity in this network, we describe con-
nections between two entities as “A regulates
B”—though more properly, we should describe
them as “A is predicted to bind near B and reg-
ulate it.” In the integrated network, the level of a
miRNA was assigned according to the highest-
level TF it regulates or, if it does not regulate a
TF, the lowest-level TF that regulates it. The
www.sciencemag.org
miRNAs fall into distinct levels, paralleling the
arrangement of TFs (Fig. 3A). Moreover, the net-
work reveals two different classes of miRNAs:
those that are more strongly regulated by TFs
versus those that predominantly regulate TFs (Fig.
3A, bottom right versus top left, respectively).
We can further analyze our integrated net-
work in terms of motifs, which is a common ap-
proach used to decompose a complex network
into simple building blocks (36). Many different
types of network motifs exist; as a simple ex-
ample, we observed miRNA-TF loops in our in-
tegrated network, in which a miRNA regulates
a TF and the same TF regulates the miRNA
(39). Of particular interest are patterns that are
overrepresented as compared with randomized,
rewired null models (6). We observed three over-
represented motifs in the integrated miRNA-TF
network (fig. S37) (6). One example is a miRNA-
mediated feed-forward loop, in which a TF reg-
ulates a miRNA and, together with the miRNA,
regulates a target coding gene (Fig. 3C). This
particular motif structure is potentially responsi-
ble for buffering noise and maintaining target
protein homeostasis (40).
RNA Pol II binding and expression. We pro-
filed RNA Pol II and the specific factor PHA-4
in each of the main stages of C. elegans devel-
opment and compared their binding profiles
with the corresponding RNA-seq data. Similar
to the above approach for gene-expression dy-
namics, for RNA Pol II we focused on a set of
8428 genes with non-overlapping transcripts
and used the binding profiles at promoters to
generate correlation matrices between each
stage. We found a similar differential clustering
of the embryonic and larval stages (Fig. 2A).
This embryonic-larval division was also observed
for PHA-4 binding across stages (fig. S30) and
presumably reflects the different transcriptional
programs between embryos and larvae.
Next, we correlated the RNA Pol II–binding
profiles with expression profiles across all the
stages. As expected, the same-stage correlation
was fairly high (0.64 to 0.70) (Fig. 2B) but was
notably lower for embryonic stages than for larval
ones, perhaps reflecting the presence of maternal
transcripts in embryos (6, 41, 42). Unexpected-
ly, we found expression at earlier developmental
stages more tightly correlated with binding at
later stages, rather than RNA Pol II–binding an-
ticipating RNA production (Fig. 2B). Specifi-
cally, the correlation is low initially, reaches a
maximum at the matching stage, and then re-
mains high for later stages. This can be inter-
preted as RNA Pol II binding to genes at the
same developmental stage at which they are ini-
tially expressed, and Pol II then remaining bound
in later stages, even if expression drops. The ini-
tial round of transcription may affect the accessi-
bility of the promoter, which may then remain
unaltered in later stages for nondividing cells.
Alternatively, this result may reflect paused RNA
Pol II at genes with reduced expression at later
stages. We have found several examples of genes
SCIENCE VOL 330 24 DECEMBER 2010
RESEARCH ARTICLES
in which RNA Pol II binding remains high in
later stages but gene expression is low [such as
isl-1 and pgp-2 (fig. S31)], which is consistent
with RNA Pol II stalling.
Overall, we have shown how the analysis of
relatively few TFs allows the construction of a
fairly elaborate network. To improve these net-
works in the future, we will need to identify the
precise cells and stages in which the TFs and
miRNAs are expressed.
Chromatin Organization and Its Implications
One modENCODE goal is to identify elements
that control chromosome behavior and regulate
the function of DNA elements. C. elegans chro-
mosomes have several distinctive features. In-
stead of having centromeres embedded in highly
repeated sequences, its chromosomes are holo-
Downloaded from www.sciencemag.org on December 23, 2010
centric, with microtubule attachment sites dis-
tributed along their length. In hermaphrodites
(XX), gene expression from both X chromo-
somes is down-regulated in somatic cells by a
dosage compensation mechanism and so better
match expression in males, which have one X
chromosome (XO) (43). Furthermore, the entire
X chromosome is under-expressed relative to
the autosomes in the germline cells of both her-
maphrodites and males (44). C. elegans auto-
somes have distinct domains—a central region
flanked by two distal “arms” that together com-
prise more than half of the chromosome. Com-
pared with the centers, the arms have higher
meiotic recombination rates, lower gene density,
and higher repeat content (5, 45, 46). Arms are
not as sharply defined on the X chromosome.
Chromosome-scale domains of histone mod-
ification. The distribution of 19 histone modifica-
tions and two key histone variants (C. elegans
homologs of H2A.z and H3.3) revealed striking,
broad domains of histone modification states on
the autosomes, with relatively sharp boundaries
between the central region of each autosome
and the arms (Fig. 5, A to C) (47–49). Mod-
ifications traditionally associated with gene ac-
tivity and euchromatin such as acetylation and
H3K4 and H3K36 methylation are enriched in
the central regions of the chromosomes. In con-
trast, H3K9 mono-, di-, and trimethylation marks
associated with transcriptional repression and
heterochromatin formation are relatively depleted
from the central regions and enriched on the arms
of the autosomes (Fig. 5A). These megabase-
scale chromosomal domains are not homoge-
neous; there are small zones of repressive marks
within the generally active central regions and
active marks within the generally repressed arms.
The chromosome-scale domains of histone mod-
ification do not vary substantially in composi-
tion or position between embryos and L3 larvae.
Despite the biased distribution of repressive
marks, the arms of the chromosomes do not
appear heterochromatic through 4´,6´-diamidino-
2-phenylindole (DAPI) staining or classical band-
ing techniques (50). Although our samples did
not include appreciable meiotic tissue, the broad
1781
 RESEARCH ARTICLES
domains of histone modifications correspond
to regions defined by differences in recombi-
nation rate, with the boundaries located at the
recombination rate inflection points (Fig. 5A)
(5, 46). On each chromosome, one arm contains
a meiotic pairing center that mediates homolo-
gous pairing and synapsis (50, 51). As previ-
ously reported, H3K9me3 is more highly enriched
on that arm (Fig. 5A) (52). However, methylation
is not particularly enriched within the pairing
center regions themselves (53). H3K9me3 is also
highly enriched on silent genes on arms, and all
forms of H3K9 methylation are enriched in re-
petitive elements, which are more prevalent on
chromosome arms (fig. S32).
The X chromosome. Gene density, recombi-
nation rates, and repeat content are more uni-
formly distributed along the X chromosome than
autosomes (5). Consistent with this, chromatin
marks on the X are more uniformly distributed.
A high density of repressive marks, similar to
that seen throughout the autosome arms, is as-
sociated with only two narrow ~300-kb regions
at the left end of the X that flank the meiotic
pairing center (Fig. 5B). The genomic distribu-
tion of DPY-26, DPY-27, DPY-28, and SDC-3,
proteins mediating dosage compensation, is
highly enriched on the X chromosome (Fig.
5B) (25, 54, 55). H4K20me1, a modification
linked in mammals to chromosome maturation
and X-chromosome inactivation (56), is also
enriched on the X. This X-enrichment is detec-
table in early embryo populations, when some
embryos have initiated dosage compensation, and
becomes more pronounced in L3 animals, when
dosage compensation is fully established.
Chromosomes and nuclear envelope interac-
tions. Interactions between the genome and the
nuclear envelope were determined by means
of ChIP of LEM-2, a transmembrane protein
associated with the nuclear lamina (57). In em-
bryos, LEM-2 interacts with the repeat-rich,
H3K9-methylated arms of the autosomes but not
with the autosome centers (Fig. 5, A and D).
Similar to H3K9 methylation, the transition be-

A Chromosome III B Chromosome X

Downloaded from www.sciencemag.org on December 23, 2010

Gene Density
Repeat Density
EE HCP-3
EE H3K27me3
L3 H3K27me3
20

MxE LEM-2
L3 H3K9me3
EE H3K9me3
Group 1 EE H3K9me1
L3 H3K9me1
L3 H3K9me2
10

EE H3K9me2
EE H4K20me1
L3 H4K20me1
Group 2 MxE SDC-3
MxE MIX-1
cM

MxE DPY-27
0

L3 RNAseq
EE RNAseq
EE H3K9ac
EE H3K4me2
EE H3K4me3
-10

EE H3K27ac
L3 H3K4me2
L3 H3K4me3
L3 H3K27ac
Group 3 EE H4tetra-ac
EE RNA Pol II
-20

EE H3K79me1
EE H3K79me2
EE H3K79me3
EE MES-4
EE H3K36me2
EE H3K36me3
L3 H3K36me2
L3 H3K36me3

0 2 0 5 Color Key
4 6 Mb 8 10 12 Mb 10 15
−1.5 0 1.5
Value
C 3Mb 4Mb 9Mb 10Mb D
RefSeq Genes H3K9 methylation regions Repeat regions

2_ LEM-2
EE H3K9me1
-2 _
2_ H4K20me1
EE H3K9me2 Chromosome III
-2 _
2_ Chromosome III
EE H3K9me3 Chromosome X
-2 _
2_ LEM-2
MxE LEM-2
-2 _
2_
L3 H3K9me1
-2 _
2_
L3 H3K9me2
-2 _
2_ Transcribed genes with
L3 H3K9me3 RNA Pol II and active histone marks
-2 _

Fig. 5. Chromosome-scale domains of chromatin organization. (A and B)
Whole-genome ChIP-chip data for various histone modifications and chromatin-
associated proteins, along with relevant genome annotations, were normalized,
placed into 10-kb bins, and displayed as a heat map. Red indicates a stronger
signal, and blue indicates a weaker signal. The continuous black line plots the
relationship between physical (x axis) and genetic (y axis) distance. Three major
groups were identified by hierarchical clustering. Group 1 contains H3K9 meth-
ylation marks and LEM-2, which tend to be enriched at distal autosomal
regions, and correlate with repetitive DNA and a high recombination rate.
Group 2 contains dosage compensation complex members and H4K20me1,
which are highly enriched on X. Group 3 contains marks associated with active
chromatin. Generally, signals for active marks are weaker on the X chromo-
some than the autosomes. This megabase-scale chromatin organization persists
through all stages examined. (A) Chromosome III is representative of autosomes.
(B) X has a distinct chromatin configuration. (C) H3K9me1, - 2, and -3 signals
decrease gradually at the boundaries between the central and distal domains,
whereas the boundaries defined by LEM-2 are relatively sharp. (D) A schematic
representation of key findings.

1782 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


tween LEM-2–enriched arms and the central
chromosomal regions is relatively sharp, coincid-
ing with the transition between regions of high
and low meiotic recombination rate (Fig. 5B).
Within the arm regions, LEM-2 enrichment ex-
hibits a complex underlying subdomain structure
(57). On the X chromosome, LEM-2 interacts
with only the small regions on the left end that
harbor repressive chromatin marks (Fig. 5B).
This suggests a particular organization for the X
chromosome within the nucleus (Fig. 5D).
Histone mono-methylation. We plotted the
distribution of each chromatin mark relative to
transcript starts and ends and further subdivided
these plots by the expression level of the asso-
ciated gene on autosomes versus the X chromo-
some (Fig. 6 and fig. S34). Overall, the results
are consistent with the known distributions and
fined sharply to gene bodies on X, in contrast to
broader enrichment that spans promoters and 3′
UTRs on autosomal genes. Conversely, H3K36me3
and H3K36me2 are more associated with auto-
somal genes than with X-linked ones (Fig. 6 and
fig. S34). Differences in several marks are ob-
served between early embryogenesis and more
differentiated L3 animals—most notably a redis-
tribution of H3K27me1 and H3K27me3 (Fig. 6
and fig. S34, bottom row).
Nucleosome organization. Consistent with mi-
crococcal nuclease (MNase) nucleosome-mapping
experiments (52, 59, 60), both X and autosomal
genes exhibit a typical nucleosome-depleted re-
gion upstream of TSSs, a well-positioned +1 nu-
cleosome, and nucleosome depletion at the 3′
ends. However, we observed that the average nu-
cleosome occupancy immediately upstream of
RESEARCH ARTICLES
(61–63). We observed a similar difference be-
tween X and autosomal promoters when naked
DNA was digested with MNase, although this
result was expected because the known DNA se-
quence preferences of MNase are similar to the
sequence preferences of linker DNA (64, 65).
DNA sequences associated with nucleosome oc-
cupancy evolve according to expression require-
ments (66, 67), suggesting that the higher GC
content on X promoters may relate to mechanisms
of X-specific gene regulation in the soma and
germline.
Epigenetic transmission of chromatin state to
progeny. The activity of the C. elegans protein
MES-4—a histone H3K36 methyltransferase
required for the survival of nascent germ cells
in developing animals—mediates the transmis-
sion of information about the pattern of germ-

Downloaded from www.sciencemag.org on December 23, 2010

functions of chromatin marks in other eukary-
otes (58). However, the distribution of several
mono-methyl marks—including H4K20me1,
H3K9me1, and H3K27me1—are associated more
with the bodies of highly transcribed genes on
the X chromosome than with similarly expressed
genes on autosomes. Further, H3K36me1 is con-
H3K4me2
2
the +1 nucleosome on the X chromosome was
1.6-fold higher than that of genes on autosomes
(at –300 to +200 bp relative to the TSS; P <
2.2e−16, Wilcoxon rank-sum test) (61). Relative
to autosomal genes, promoters of X-linked genes
have higher GC content, which is predictive of
high nucleosome occupancy in vitro (fig. S33)
H3K4me3 H4 tetra ac
line gene expression from mother to progeny.
Similar to other H3K36 methyltransferases,
MES-4 is associated with gene bodies. However,
in contrast to previously studied H3K36 methyl-
transferases (68) MES-4 is able to associate with
genes in an RNA Pol II–independent manner (69).
In the embryo, MES-4 is preferentially bound to
genes that were highly expressed in the maternal
germline but may no longer be expressed in em-
bryos (69). Conversely, MES-4 is not associated

with genes expressed specifically in early em-

1

bryos, despite recruitment of RNA Pol II to those

genes (69). Therefore, RNA Pol II association
Active marks in EE
0

with genes is neither necessary nor sufficient to

recruit MES-4 in embryos (69). These findings
H3K36me1 H3K36me3 H3K79me3 suggest that MES-4, which is required for fer-
2

tility, functions as a maintenance histone methyl-

transferase and propagates the memory of gene
1

expression from the maternal germline to the

0

cells of the next generation (69).

Models relating chromatin to TF binding.
−1

H3K9me1 H3K9me2 H3K9me3 To integrate chromatin with other types of

2

modENCODE data, we sought to relate the pat-

terns of histone marks with the observed TF-
1
Repressive marks in EE

binding sites. Across the whole genome, we

observed only weak direct correlations between
0

the two (fig. S38A). However, the relationship

between chromatin and TFs may involve com-
H3K27me1 H3K27me3 H4K20me1
plex, nonlinear relationships. To probe these, we
2

built machine-learning models to identify TF-

binding peaks from chromatin features (fig. S39).
1

Investigating the association of individual his-

0

tone marks with TF-binding sites, we found some

−1

that discriminate TF-binding sites from the ge-

H3K27me1 H3K27me3 H4K20me1 Transcript Level
nomic background with reasonable accuracy (Fig.
7A). Often, this is connected with their actual
2

Autosomal
top 20% presence at binding sites; for example, when
1

bottom 20%
L3

comparing the background to binding peaks, on

X-linked
0

average, some marks have stronger signals, where-

top 20%
as others have weaker ones [such as H3K4me3
−1

bottom 20%
TSS TES TSS TES TSS TES versus H3K9me3 (fig. S41)]. Individual chroma-
-1k +1.5k -1.5k +1k -1k +1.5k -1.5k +1k -1k +1.5k -1.5k +1k tin marks and RNA Pol II–binding signals could
Fig. 6. Chromatin patterns around genes. Average gene profiles around the TSS and TTS of various also distinguish HOT regions from the genomic
histone marks displayed for the (red) X chromosome and (blue) autosomes. Genes were further stratified background, highlighting the association with ac-
according to their expression level, with the top 20% of expressed genes shown in darker shade and the tive transcription in these regions.
bottom 20% of expressed genes shown in lighter color. Marks typically associated with active or repressed Because chromatin features work in combi-
transcription are labeled on the left. nation to influence binding-site selection (70),

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1783

 RESEARCH ARTICLES
we combined all the histone marks together in a
classifier. The resulting models could identify bind-
ing sites better than those based on any individual
mark (Fig. 7A and figs. S38B and S40A).
We further observed that chromatin features
are particularly good at identifying the bind-
ing peaks of some specific TFs. For example,
H3K4me2 and H3K4me3, which are usually
enriched in promoters, identified the binding
peaks of a group of five factors (CEH-14, CEH-
30, LIN-13, LIN-15B, and MEP-1) better than
the other TFs. This association is specifically
due to a relative enrichment of these H3K4me2
and H3K4me3 at the binding peaks of this
group of five TFs (fig. S41). It further suggests
that the chromatin features can be useful in
discriminating not only binding sites from the
genomic background but also the sites of
To provide additional predictive power, we By combining all features at each of the 160
incorporated into our models the information bins, we built a model for gene expression, pre-
from the specific sequence motif recognized by dicting the quantitative expression levels of tran-
a TF, summarized by a position-weight matrix. scripts with support vector regression (SVR) (6).
The combined models with both chromatin and Predicted expression levels were highly correlated
sequence information were more accurate than with measured ones [correlation coefficient (r) =
were models involving either type of information 0.75, cross-validated]. As an overall benchmark,
alone (Fig. 7B and fig. S43). Thus, chromatin fea- we compared our chromatin model with one
tures enable one to predict TF-accessible regions based on the level of RNA Pol II–binding alone
and broad classes of binding sites, and motifs pro- (r = 0.37); our model achieves better prediction
vide additional information on the exact sites bound accuracy for expression levels.
by particular factors, chosen from these broad classes. To find the relative importance for gene ex-
Models relating chromatin to gene expression. pression of the 160 possible bin locations, we
Next, we developed a model to relate chromatin divided genes into highly and lowly expressed
marks to gene expression levels. We divided the classes and predicted the class of each gene from
regions around each TSS and transcript termi- each bin. The best predictions were obtained from
nation site (TTS) into small (100 bp) bins and bins immediately after the TSS and just before
calculated the average signal of each chromatin the TTS. With increasing distance upstream of

Downloaded from www.sciencemag.org on December 23, 2010

specific TFs in comparison with other TFs.
Indeed, we were able to build integrated models
to do this with reasonable accuracy (fig. S40B).
The same approach was also successful in dis-
criminating HOT regions from all TF-binding re-
gions (fig. S40B). Our models perform best when
chromatin features are measured at the same
stage as the TFs, suggesting a dynamic relation-
ship between chromatin and binding sites across
developmental stages (fig. S42).
feature and RNA Pol II (13 features in total) in a the TSS, predictive power decreased smoothly.
set of 160 bins up to 4 kb upstream and Intriguingly, the predictive capability of chroma-
downstream of these two anchors (to include tin features extended as much as 4 kb upstream
even long-range effects). Then at each bin, we of the TSS and 4 kb downstream of the TTS,
correlated the chromatin signals with the stage- even when we restricted the analysis to widely
matched gene expression value (Fig. 7C). There separated genes with distant neighbors. This
is clear variation across the bins in this correla- may indicate a long-range influence of chroma-
tion, with the effect of making activating marks tin on gene expression.
more sensitive than are repressive ones to their In contrast to protein-coding genes, the asso-
exact positioning relative to the TSS or TTS. ciation between histone modifications and miRNA

Fig. 7. Statistical models pre- A HLH-1

dicting TF-binding and gene Histone modifications RNA Pol II Integrated B 70
Predictors
expression from chromatin fea-
EE

EE

EE

H3K79me1 EE

EE

H3K79me3 EE

H3K79me2 EE

EE

EE

EE

YA
LE
L3

L3

L3

L3

H3K27me3 L3

L3

L1
L2
L3
L4

Histone marks

All RNA Pol II

Positive predictive value (PPV)

60
tures. (A) Modeling TF-binding
H3K36me3

H3K36me2

All features
H3K9me2

H3K9me3

H3K4me2
H3K4me3

PWM + Chromatin model

sites with chromatin features. Binding 50
experiments
The color of each cell represents AUROC

Chromatin model only

PQM1
1 40
the accuracy of a statistical model ALR1

in which a chromatin feature or a

GEI11
D
Observed Expression ( )
BLMP1 30

PWM only
( )
EGL27
set of features acts as predictor HLH1 * R=0.37
for TF binding or HOT regions. (B) LIN39
20
*
PHA4
An example of combining chro- ELT3 10
MAB5
matin and sequence features. PES1
0.5 0
Potential binding sites of HLH-1 SKN1
UNC130
were predicted by using only se- EOR1 Pol II
quence motifs, only chromatin EGL5
D *
)
2

MDL1
Observed Expression ( ()

features, or both. (C) Correlation LIN11 * R=0.75

1

LIN15B
pattern for a number of chro- CEH14
LIN13
-3 -2 -1 0

matin features in 100-bp bins MEP1

around the TSS (T 4 kb) and CEH30

TTS (T 4 kb) of transcripts at the HOT

early embryo (EE) stage. The

Spearman correlation coeffi- 0.8
cient of each chromatin feature C H3K4me2 -2 -1 0 1
with gene-expression levels 0.6
H3K4me3 Predicted expression from
was calculated for each bin. (D) Pol II+13HMs (SVR model)
Obs. miRNA Expression ( )

H3K9me2
-1 0 1 2 3 4 5 6

Chromatin features can predict 0.4 * R=0.60

expression levels for both protein- H3K9me3
coding genes and miRNAs. (Top) 0.2

A model involving all chromatin H3K36me2

features. (Bottom) The model for 0 H3K36me3
protein-coding genes can also be
used to predict accurately miRNA -0.2
H3K79me1
expression levels. H3K79me2 -6 -4 -2 0 2
-0.4
H3K79me3 Predicted miRNA expression
(SVR model above)
-4kb TSS 4kb -4kb TTS 4kb
* = LOG10RPKM

1784 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


expression has not been explored in detail. Be-
cause protein-coding and miRNA genes are both
transcribed by RNA Pol II, we applied the above
chromatin model, derived from protein-coding
genes, to the regions around candidate pre-
miRNAs. We then predicted expression levels
for 162 microRNAs, for which genomic loca-
tions are provided by miRBase (71), and com-
pared these predictions to the measurements in
the modENCODE small RNA-seq data set. We
found a correlation of 0.60 (r = 0.62 for just
miRNAs far from known genes) (Fig. 7D). That
expression of miRNAs can be predicted ac-
curately by using a chromatin model trained on
protein-coding genes is consistent with miRNAs
and protein-coding gene regulation sharing sim-
ilar mechanistic connections to histone marks.
Conservation Analysis
Because mutations are constantly accumulating
over evolutionary time, purifying selection slows
the rate of divergence of functional relative to
nonfunctional sequences (72). For this reason,
evolutionarily constrained regions can assist in
identifying functional elements (73). Although
some functional sequences may not be conserved,
are conserved in a way that we are unable to
detect, or are under positive selection (resulting
in accelerated divergence), the coverage of con-
strained bases by identified functional elements
is a valuable measure of the completeness of our
understanding of the genome. We characterized
regions of the C. elegans genome under evolu-
tionary constraint by constructing a multiple
alignment among the nematodes C. elegans,
C. remanei, C. briggsae, C. brenneri, C. japonica,
and Pristionchus pacificus using methods pre-
viously developed (1). We then calculated con-
servation scores with PhastCons (6, 74). These
Fig. 8. Relative proportion of annota-
tions among constrained sequences. (A)
Relative proportion of constrained and
unconstrained bases in the C. elegans
genome. Within the constrained re-
gion, the stacked bar chart shows the
cumulative proportion covered by var-
ious classes of annotated genomic ele-
ments. (B) Fraction of element classes
covering (red) constrained and (gray) un-
constrained bases. The error bars show
the 95% confidence interval for ran-
dom placement of elements calculated
with GSC. If the ends of the columns
are outside the confidence interval,
then it is unlikely that the fraction of
the element class overlapping con-
strained and/or unconstrained bases
could have occurred by chance. (C) Con-
straint profiles of broad categories
of elements. The x axis indicates the
PhastCons score of bases covered by
the element ranging from 0 (no con-
servation) to 1.0 (perfect conservation). The y axis indicates the log ratio of the number of bases with the given score covered, relative to what would be
expected by random element placement (dotted line) (fig. S45 shows more detail).
www.sciencemag.org
procedures identified 59,504 constrained blocks
that cover 29.6% of the C. elegans genome as a
whole and range from 27.4% of chromosome
IV to 31.9% of chromosome X. The single largest
constrained block was 3558 bp on chromosome
V, but conserved blocks were typically much
smaller (mean 49 T 58.6 bp).
These conserved regions are highly corre-
lated with functional elements. We first exam-
ined the proportion of evolutionarily constrained
regions that overlap experimentally annotated
portions of the genome (Fig. 8A and fig. S44).
In the last WormBase freeze before the incorpo-
ration of modENCODE data (6), 50.8% of the
constrained regions were covered by annotations
supported by direct experimental evidence. Ad-
ding modENCODE protein-coding gene evi-
dence increased the coverage of constrained bases
to 58.3%. Other modENCODE increases came
from the 7k-set of ncRNAs (1.9%), TF-binding
sites, (5.9%), dosage compensation (9.3%), and
other chromatin-associated factors (2.8%). Thus,
modENCODE explains an additional 27.4%
(8.1 Mb) of the constrained portion of the ge-
nome; together with remaining unconfirmed
WormBase gene predictions (0.7%) and pseudo-
genes (0.6%), coverage now totals 79.5% of
constrained bases.
We then estimated the extent of constraint on
different functional elements by plotting the dis-
tribution of the PhastCons conservation scores
for each type of element (Fig. 8, B and C, and
fig. S45). The most constrained elements were
ncRNAs (both known and the 7k-set), presumably
reflecting the fact that conservation was a criterion
used to identify them. Next came protein-coding
elements, followed by miRNAs, TF-binding sites,
and other chromatin factor–binding sites. Pseudo-
genes, introns, and regions of the genome not
SCIENCE VOL 330 24 DECEMBER 2010
RESEARCH ARTICLES
covered by modENCODE data sets all have low
levels of conservation. We then used the genome
structure correction (GSC) statistic (1, 75) to cal-
culate confidence intervals on the degree of over-
lap between evolutionarily constrained bases and
functional elements defined by modENCODE and
other sources. This demonstrated that coding re-
gions, ncRNAs, TF-binding sites, and other chro-
matin factor–binding sites are significantly more
constrained than would be expected by chance,
whereas regions covered by pseudogenes, introns,
and unannotated regions are significantly depleted
in constrained regions relative to chance.
Roughly 20.5% of the constrained genome
remains uncovered by known functional ele-
ments, but a portion of this sequence directly
abuts known functional elements. If the borders
of transcribed regions and chromatin-associated
Downloaded from www.sciencemag.org on December 23, 2010
protein-binding sites are extended across all con-
strained blocks that neighbor them, ~ 4.1 Mb
(14%) in isolated constrained blocks remains.
These residual constrained bases are highly en-
riched in introns and intragenic regions (table
S14), are moderately enriched in the 1-kb regions
upstream of TSSs, and are depleted in the 1-kb
regions downstream of TTSs. One potential ex-
planation for the residual constrained bases is that
they correspond to the binding sites of untested
TFs. Indeed, a plot of coverage of constrained se-
quence against numbers of TF experiments shows
that the relatively small numbers of TFs studied
here are far from saturating constrained bases (fig.
S47), implying that additional TFs may explain
part of the remaining constrained bases in these
regions. Other explanations for the residual con-
strained regions include other intronic regulatory
sites, transcribed regions that are expressed only
under rare circumstances, and possibly as-yet
unknown classes of functional elements.
1785
 RESEARCH ARTICLES
Discussion
Our analysis illustrates patterns at multiple ge-
nomic scales: individual gene, chromosomal
domain, and whole-chromosome. At the first
scale, in addition to improving annotation of
protein-coding genes, we identified transcribed
pseudogenes and many previously unidentified
ncRNAs, mapped binding sites of TFs, built
regulatory networks, and constructed models
predicting binding location and expression
levels from chromatin marks. On a larger scale,
we found chromosomal domains—characterized
by repressive marks and interactions with the
nuclear envelope on the autosome arms—and
noted how the boundaries in these domains
align with changes in recombination frequency.
We also identified additional properties of the
entire X chromosome, including the preferential
The modENCODE data sets are intended as 30.
an enduring resource for the genomics com- 31.
32.
munity. All raw and analyzed data, metadata, 33.
and interpreted results are available at www. 34.
modencode.org, where they can be searched,
displayed, and downloaded. Raw sequencing 35.
reads and microarray data are archived at the 36.
Short-read Archive and the Gene Expression 37.
Omnibus, and higher-order results are being
incorporated into WormBase (77). In addition, 38.
39.
we have assembled a compact guide to the data 40.
sets used (at www.modencode.org/publications/
integrative_worm_2010) (table S1) (6) and have 41.
populated a community cloud-computing re-
42.
source with the data and analysis tools to
43.
facilitate further investigation by interested
researchers (6). We expect that analyses of these
data sets in the coming years will provide 44.
X. Liu et al., Cell, 139, 623 (2009).
R. S. Kamath et al., Nature 421, 231 (2003).
M. Ashburner et al., Nat. Genet. 25, 25 (2000).
A. Carr, M. D. Biggin, EMBO J. 18, 1598 (1999).
C. Moorman et al., Proc. Natl. Acad. Sci. U.S.A. 103,
12027 (2006).
S. MacArthur et al., Genome Biol. 10, R80
(2009).
U. Alon, Nat. Rev. Genet. 8, 450 (2007).
H. Y. Yu, M. Gerstein, Proc. Natl. Acad. Sci. U.S.A. 103,
14724 (2006).
N. Simonis et al., Nat. Methods 6, 47 (2009).
N. J. Martinez et al., Genes Dev. 22, 2535 (2008).
E. Hornstein, N. Shomron, Nat. Genet. 38 (suppl.),
S20 (2006).
L. G. Edgar, N. Wolf, W. B. Wood, Development 120,
443 (1994).
G. Seydoux, A. Fire, Development 120, 2823 (1994).
B. J. Meyer, “X-Chromosome Dosage Compensation,”
WormBook, The C. elegans Research Community, Eds.
(WormBook, 2005), 10.1895/wormbook.1.8.1.
W. G. Kelly et al., Development 129, 479 (2002).

Downloaded from www.sciencemag.org on December 23, 2010

accumulation of multiple mono-methylated his-
tone marks. Our large-scale approach also
discovered unexpected biological phenomena
that would be difficult to uncover in conven-
tional studies. In particular, upon profiling the
binding sites of 23 factors we identified regions
of clustered binding (HOT regions).
One limitation of the modENCODE strategy
is that we cannot readily distinguish low levels
of biochemical noise, such as a rare nonfunc-
tional transcription splice form, from biological-
ly important phenomena. The presence of such
noise may be an unavoidable part of the cell
regulatory machinery (76) and will only be dis-
tinguished from biologically important signals
through careful follow-up experimentation. An-
other limitation is that almost all experiments
were performed in populations of whole animals
composed of multiple tissues. Future studies will
increase the tissue-specific resolution of the data.
Model organisms such as C. elegans have
long served as key experimental systems for
developing technology and providing funda-
mental insights into human biology. Comparing
our modENCODE results with the ENCODE
pilot, which assessed functional elements in 1%
of the human genome, we can already begin to
see commonalities (6). For instance, for some ag-
gregated binding signals (such as for RNA Pol
II) the overall shape of the signal distributions
relative to the TSS are quite similar between hu-
man and C. elegans. Likewise, the overall amount
(per base pair) of transcription and binding by
TFs is comparable (fig. S49 and tables S15 and
S16). However, there are differences in the shape
of the aggregated signal distributions for a few
matched histone modifications (Fig. 6 versus
fig. S50). Moreover, the relative proportion of
constrained genome covered by experimental an-
notation is quite different in human and nema-
tode, perhaps reflecting evolutionary pressures
for a compact genome in the latter (fig. S48). A
more comprehensive comparison, including the
Drosophila genome data presented in the ac-
companying article, must await genome-wide
analysis of human cells—an effort currently
underway in the ENCODE project.
additional insights into general principles of 45. T. M. Barnes, Y. Kohara, A. Coulson, S. Hekimi, Genetics
141, 159 (1995).
genome organization and function, which will 46. M. V. Rockman, L. Kruglyak, M. Przeworski, PLoS Genet.
ultimately aid in annotating and deciphering the 5, e1000419 (2009).
human genome. 47. T. Liu et al., Genome Res., 10.1101/gr.115519.110.
48. S. L. Ooi, J. G. Henikoff, S. Henikoff, Nucleic Acids Res.
38, e26 (2010).
49. T. A. Egelhofer et al., Nat. Struct. Mol. Biol.,
References and Notes
1. E. Birney et al., ENCODE Project Consortium; NISC 10.1038/nsmb.1972.
Comparative Sequencing Program; Baylor College of 50. D. G. Albertson, A. M. Rose, A. M. Villeneuve, in
Medicine Human Genome Sequencing Center; C. elegans II, D. L. Riddle, T. Blumenthal, B. J. Meyer,
Washington University Genome Sequencing Center; J. R. Preiss, Eds. (Cold Spring Harbor Laboratory Press,
Cold Spring Harbor, NY, 1997), pp. 47–78.
Broad Institute; Children’s Hospital Oakland Research
Institute, Nature 447, 799 (2007). 51. A. J. MacQueen et al., Cell 123, 1037 (2005).
2. S. E. Celniker et al., Nature 459, 927 (2009). 52. S. G. Gu, A. Fire, Chromosoma 119, 73 (2010).
3. J. E. Sulston, E. Schierenberg, J. G. White, J. N. Thomson, 53. C. M. Phillips et al., Nat. Cell Biol. 11, 934
Dev. Biol. 100, 64 (1983). (2009).
4. J. G. White, E. Southgate, J. N. Thomson, S. Brenner, 54. S. Ercan, L. L. Dick, J. D. Lieb, Curr. Biol. 19, 1777
Philos. Trans. R. Soc. London B Biol. Sci. 314, 1 (2009).
(1986). 55. J. Jans et al., Genes Dev. 23, 602 (2009).
5. C. elegans Sequencing Consortium, Science 282, 2012 56. A. Kohlmaier et al., PLoS Biol. 2, E171 (2004).
(1998). 57. K. Ikegami, T. A. Egelhofer, S. Strome, J. D. Lieb,
6. Materials and methods are available as supporting Genome Res., 10.1186/gb-2010-11-12-r120.
58. T. Kouzarides, Cell 128, 693 (2007).
material on Science Online.
7. J. Reboul et al., Nat. Genet. 34, 35 (2003). 59. S. M. Johnson, F. J. Tan, H. L. McCullough, D. P. Riordan,
8. P. Lamesch et al., Genome Res. 14, (10B), 2064 A. Z. Fire, Genome Res. 16, 1505 (2006).
(2004). 60. A. Valouev et al., Genome Res. 18, 1051
9. A. Mortazavi, B. A. Williams, K. McCue, L. Schaeffer, (2008).
61. S. Ercan, Y. Lubling, E. Segal, J. D. Lieb, Genome Res.,
B. Wold, Nat. Methods 5, 621 (2008).
10. L. W. Hillier et al., Genome Res. 19, 657 (2009). 10.1101/gr.115931.110.
11. M. Mangone et al., Science 329, 432 (2010). 62. N. Kaplan et al., Nature 458, 362 (2009).
12. C. H. Jan, R. C. Friedman, J. G. Ruby, C. B. Burge, 63. D. Tillo et al., PLoS ONE 5, e9129 (2010).
D. P. Bartel, Nature, published online 17 November 2010 64. W. Hörz, W. Altenburger, Nucleic Acids Res. 9,
2643 (1981).
(10.1038/nature09616).
13. M. A. Allen, L. W. Hillier, R. H. Waterston, T. Blumenthal, 65. E. Segal, J. Widom, Curr. Opin. Struct. Biol. 19, 65
Genome Res., 10.1101/gr.113811.110. (2009).
14. A. Agarwal et al., BMC Genomics 11, 383 (2010). 66. Y. Field et al., PLoS Comput. Biol. 4, e1000216
15. W. C. Spence et al., Genome Res., 10.1101/gr.114595.110 (2008).
16. P. M. Harrison, M. Gerstein, J. Mol. Biol. 318, 1155 67. A. M. Tsankov et al., PLoS Biol. 8, e1000414
(2002). (2010).
17. R. Sasidharan, M. Gerstein, Nature 453, 729 68. K. O. Kizer et al., Mol. Cell. Biol. 25, 3305
(2008). (2005).
18. J. S. Mattick, Science 309, 1527 (2005). 69. A. Rechtsteiner et al., PLoS Genet. 6, e1001091
19. M. Kato, A. de Lencastre, Z. Pincus, F. J. Slack, (2010).
Genome Biol. 10, R54 (2009). 70. S. L. Berger, Nature 447, 407 (2007).
20. M. Stoeckius et al., Nat. Methods 6, 745 (2009). 71. S. Griffiths-Jones, H. K. Saini, S. van Dongen,
21. J. G. Ruby, C. H. Jan, D. P. Bartel, Nature 448, 83 A. J. Enright, Nucleic Acids Res. 36 (Database issue),
(2007). D154 (2008).
22. W. Chung et al., Genome Res., 10.1101/gr.113050.110. 72. G. M. Cooper et al., Genome Res. 14, 539 (2004).
73. A. M. Moses et al., PLOS Comput. Biol. 2, e130
23. J. G. Ruby et al., Cell 127, 1193 (2006).
24. Z. J. Lu et al., Genome Res., 10.1101/gr.110189.110. (2006).
25. S. Ercan et al., Nat. Genet. 39, 403 (2007). 74. A. Siepel et al., Genome Res. 15, 1034 (2005).
26. W. Niu et al., Genome Res., 10.1101/gr.114587.110. 75. P. J. Bickel, N. Boley, J. B. Brown, H. Huang, N. Zhang,
27. T. Fukushige, M. Krause, Development 132, 1795 Annals Appl. Stat. 0, 1 (2010).
(2005). 76. A. Eldar, M. B. Elowitz, Nature 467, 167
28. C. A. Grove et al., Cell 138, 314 (2009). (2010).
29. P. J. Roy, J. M. Stuart, J. Lund, S. K. Kim, Nature 418, 77. T. W. Harris et al., Nucleic Acids Res. 38 (Database issue),
975 (2002). D463 (2010).

1786 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


78. Funding for this work came from the NHGRI of the
NIH as part of the modENCODE project, NIH (grant
R01GM088565), Muscular Dystrophy Association, and the
Pew Charitable Trusts (J.K.K.); the Helmholtz-Alliance on
Systems Biology (Max Delbrück Centrum Systems
Biology Network) (S.D.M.); the Wellcome Trust (J.A.);
the William H. Gates III Endowed Chair of Biomedical
Sciences (R.H.W.); and the A. L. Williams Professorship
(M.B.G.). M. Snyder has an advisory role with DNANexus,
a DNA sequence storage and analysis company. Transfer
of GFP-tagged fosmids requires a Materials Transfer
Agreement with the Max Planck Institute of Molecular Cell
RESEARCH ARTICLES
Biology and Genetics. Raw microarray data are available fellow curator P. Davis for reviewing and hand-checking
from the Gene Expression Omnibus archive, and raw the list of pseudogenes.
sequencing data are available from the SRA archive
Supporting Online Material
(accessions are in table S18). We appreciate help from
www.sciencemag.org/cgi/content/science.1196914/DC1
S. Anthony, K. Bell, C. Davis, C. Dieterich, Y. Field,
Materials and Methods
A. S. Hammonds, J. Jo, N. Kaplan, A. Manrai, B. Mathey-Prevot,
Figs. S1 to S50
R. McWhirter, S. Mohr, S. Von Stetina, J. Watson,
Tables S1 to S18
K. Watkins, C. Xue, and Y. Zhang, and B. Carpenter. We
References
thank C. Jan and D. Bartel for sharing data on poly(A)
sites before publication, WormBase curator G. Williams 24 August 2010; accepted 18 November 2010
for assistance in quality checking and preparing the Published online 22 December 2010;
transcriptomics data sets for publication, as well as his 10.1126/science.1196914

biology. The functions of ~40% of the protein-

Identification of Functional Elements and nonprotein-coding genes [FlyBase 5.12 (4)]
have been determined from cDNA collections
and Regulatory Circuits by (5, 6), manual curation of gene models (7), gene
mutations and comprehensive genome-wide
Drosophila modENCODE RNA interference screens (8–10), and compara-

Downloaded from www.sciencemag.org on December 23, 2010

tive genomic analyses (11, 12).
The Drosophila modENCODE project has
The modENCODE Consortium,* Sushmita Roy,1,2† Jason Ernst,1,2† Peter V. Kharchenko,3† generated more than 700 data sets that profile
Pouya Kheradpour,1,2† Nicolas Negre,4† Matthew L. Eaton,5† Jane M. Landolin,6† transcripts, histone modifications and physical
Christopher A. Bristow,1,2† Lijia Ma,4† Michael F. Lin,1,2† Stefan Washietl,1† nucleosome properties, general and specific tran-
Bradley I. Arshinoff,7,18† Ferhat Ay,1,33† Patrick E. Meyer,1,30† Nicolas Robine,8† scription factors (TFs), and replication programs
Nicole L. Washington,9† Luisa Di Stefano,1,31† Eugene Berezikov,23‡ Christopher D. Brown,4‡ in cell lines, isolated tissues, and whole orga-
Rogerio Candeias,1‡ Joseph W. Carlson,6‡ Adrian Carr,10‡ Irwin Jungreis,1,2‡ nisms across several developmental stages (Fig. 1).
Daniel Marbach,1,2‡ Rachel Sealfon,1,2‡ Michael Y. Tolstorukov,3‡ Sebastian Will,1‡ Here, we computationally integrate these data
Artyom A. Alekseyenko,11 Carlo Artieri,12 Benjamin W. Booth,6 Angela N. Brooks,28 Qi Dai,8 sets and report (i) improved and additional ge-
Carrie A. Davis,13 Michael O. Duff,14 Xin Feng,13,18,35 Andrey A. Gorchakov,11 Tingting Gu,15 nome annotations, including full-length protein-
Jorja G. Henikoff,8 Philipp Kapranov,16 Renhua Li,17 Heather K. MacAlpine,5 John Malone,12 coding genes and peptides as short as 21 amino
Aki Minoda,6 Jared Nordman,22 Katsutomo Okamura,8 Marc Perry,18 Sara K. Powell,5 acids; (ii) noncoding transcripts, including 132
Nicole C. Riddle,15 Akiko Sakai,29 Anastasia Samsonova,19 Jeremy E. Sandler,6 Yuri B. Schwartz,3 candidate structural RNAs and 1608 nonstruc-
Noa Sher,22 Rebecca Spokony,4 David Sturgill,12 Marijke van Baren,20 Kenneth H. Wan,6 tural transcripts; (iii) additional Argonaute (Ago)–
Li Yang,14 Charles Yu,6 Elise Feingold,17 Peter Good,17 Mark Guyer,17 Rebecca Lowdon,17 associated small RNA genes and pathways,
Kami Ahmad,29 Justen Andrews,21 Bonnie Berger,1,2 Steven E. Brenner,28,32 Michael R. Brent,20 including new microRNAs (miRNAs) encoded
Lucy Cherbas,21,24 Sarah C. R. Elgin,15 Thomas R. Gingeras,13,16 Robert Grossman,4 within protein-coding exons and endogenous small
Roger A. Hoskins,6 Thomas C. Kaufman,21 William Kent,34 Mitzi I. Kuroda,11 Terry Orr-Weaver,22 interfering RNAs (siRNAs) from 3′ untranslated
Norbert Perrimon,19 Vincenzo Pirrotta,27 James W. Posakony,26 Bing Ren,26 Steven Russell,10 regions; (iv) chromatin “states” defined by com-
Peter Cherbas,21,24 Brenton R. Graveley,14 Suzanna Lewis,9 Gos Micklem,10 Brian Oliver,12 binatorial patterns of 18 chromatin marks that are
Peter J. Park,3 Susan E. Celniker,6§|| Steven Henikoff,25§|| Gary H. Karpen,6,28§|| Eric C. Lai,8§|| associated with distinct functions and properties;
David M. MacAlpine,5§|| Lincoln D. Stein,18§|| Kevin P. White,4§|| Manolis Kellis1,2|| (v) regions of high TF occupancy and replication
activity with likely epigenetic regulation; (vi) mixed
To gain insight into how genomic information is translated into cellular and developmental TF and miRNA regulatory networks with hierar-
programs, the Drosophila model organism Encyclopedia of DNA Elements (modENCODE) project chical structure and enriched feed-forward loops;
is comprehensively mapping transcripts, histone modifications, chromosomal proteins, transcription (vii) coexpression- and co-regulation–based func-
factors, replication proteins and intermediates, and nucleosome properties across a developmental tional annotations for nearly 3000 genes; (viii)
time course and in multiple cell lines. We have generated more than 700 data sets and discovered stage- and tissue-specific regulators; and (ix)
protein-coding, noncoding, RNA regulatory, replication, and chromatin elements, more than predictive models of gene expression levels and
tripling the annotated portion of the Drosophila genome. Correlated activity patterns of these regulator function.
elements reveal a functional regulatory network, which predicts putative new functions for genes, Overview of data sets. Our data sets provide
reveals stage- and tissue-specific regulators, and enables gene-expression prediction. Our results an extensive description of the transcriptional, epi-
provide a foundation for directed experimental and computational studies in Drosophila and genetic, replication, and regulatory landscapes of
related species and also a model for systematic data integration toward comprehensive genomic the Drosophila genome (table S1). Experimental
and functional annotation. assays include high-throughput RNA sequencing
(RNA-seq), capturing-small and large RNAs and
everal years after the complete genetic se- through systematic generation and computational splice variants; chromatin immunoprecipitation

S quencing of many species, it is still unclear

how to translate genomic information into
a functional map of cellular and developmental
programs. The Encyclopedia of DNA Elements
(ENCODE) (1) and model organism ENCODE
(modENCODE) (2) projects use diverse genomic
assays to comprehensively annotate the Homo
sapiens (human), Drosophila melanogaster (fruit
fly), and Caenorhabditis elegans (worm) genomes,
integration of functional genomic data sets.
Previous genomic studies in flies have made
seminal contributions to our understanding of
basic biological mechanisms and genome func-
tions, facilitated by genetic, experimental, compu-
tational, and manual annotation of the euchromatic
and heterochromatic genome (3), small genome
size, short life cycle, and a deep knowledge of
development, gene function, and chromosome
(ChIP)–chip and ChIP followed by high-throughput
sequencing (ChIP-seq), profiling chromosomal
and RNA binding or processing proteins; tiling-
arrays, identifying and measuring replication pat-
terns, nucleosome solubility, and turnover; and
genomic DNA sequencing, measuring copy-
number variation. We conducted most assays in
the sequenced strain y; cn bw sp (13), with mul-
tiple developmental samples (30 for RNA expres-

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1787


78. Funding for this work came from the NHGRI of the
NIH as part of the modENCODE project, NIH (grant
R01GM088565), Muscular Dystrophy Association, and the
Pew Charitable Trusts (J.K.K.); the Helmholtz-Alliance on
Systems Biology (Max Delbrück Centrum Systems
Biology Network) (S.D.M.); the Wellcome Trust (J.A.);
the William H. Gates III Endowed Chair of Biomedical
Sciences (R.H.W.); and the A. L. Williams Professorship
(M.B.G.). M. Snyder has an advisory role with DNANexus,
a DNA sequence storage and analysis company. Transfer
of GFP-tagged fosmids requires a Materials Transfer
Agreement with the Max Planck Institute of Molecular Cell
RESEARCH ARTICLES
Biology and Genetics. Raw microarray data are available fellow curator P. Davis for reviewing and hand-checking
from the Gene Expression Omnibus archive, and raw the list of pseudogenes.
sequencing data are available from the SRA archive
Supporting Online Material
(accessions are in table S18). We appreciate help from
www.sciencemag.org/cgi/content/science.1196914/DC1
S. Anthony, K. Bell, C. Davis, C. Dieterich, Y. Field,
Materials and Methods
A. S. Hammonds, J. Jo, N. Kaplan, A. Manrai, B. Mathey-Prevot,
Figs. S1 to S50
R. McWhirter, S. Mohr, S. Von Stetina, J. Watson,
Tables S1 to S18
K. Watkins, C. Xue, and Y. Zhang, and B. Carpenter. We
References
thank C. Jan and D. Bartel for sharing data on poly(A)
sites before publication, WormBase curator G. Williams 24 August 2010; accepted 18 November 2010
for assistance in quality checking and preparing the Published online 22 December 2010;
transcriptomics data sets for publication, as well as his 10.1126/science.1196914

biology. The functions of ~40% of the protein-

Identification of Functional Elements and nonprotein-coding genes [FlyBase 5.12 (4)]
have been determined from cDNA collections
and Regulatory Circuits by (5, 6), manual curation of gene models (7), gene
mutations and comprehensive genome-wide
Drosophila modENCODE RNA interference screens (8–10), and compara-

Downloaded from www.sciencemag.org on December 23, 2010

tive genomic analyses (11, 12).
The Drosophila modENCODE project has
The modENCODE Consortium,* Sushmita Roy,1,2† Jason Ernst,1,2† Peter V. Kharchenko,3† generated more than 700 data sets that profile
Pouya Kheradpour,1,2† Nicolas Negre,4† Matthew L. Eaton,5† Jane M. Landolin,6† transcripts, histone modifications and physical
Christopher A. Bristow,1,2† Lijia Ma,4† Michael F. Lin,1,2† Stefan Washietl,1† nucleosome properties, general and specific tran-
Bradley I. Arshinoff,7,18† Ferhat Ay,1,33† Patrick E. Meyer,1,30† Nicolas Robine,8† scription factors (TFs), and replication programs
Nicole L. Washington,9† Luisa Di Stefano,1,31† Eugene Berezikov,23‡ Christopher D. Brown,4‡ in cell lines, isolated tissues, and whole orga-
Rogerio Candeias,1‡ Joseph W. Carlson,6‡ Adrian Carr,10‡ Irwin Jungreis,1,2‡ nisms across several developmental stages (Fig. 1).
Daniel Marbach,1,2‡ Rachel Sealfon,1,2‡ Michael Y. Tolstorukov,3‡ Sebastian Will,1‡ Here, we computationally integrate these data
Artyom A. Alekseyenko,11 Carlo Artieri,12 Benjamin W. Booth,6 Angela N. Brooks,28 Qi Dai,8 sets and report (i) improved and additional ge-
Carrie A. Davis,13 Michael O. Duff,14 Xin Feng,13,18,35 Andrey A. Gorchakov,11 Tingting Gu,15 nome annotations, including full-length protein-
Jorja G. Henikoff,8 Philipp Kapranov,16 Renhua Li,17 Heather K. MacAlpine,5 John Malone,12 coding genes and peptides as short as 21 amino
Aki Minoda,6 Jared Nordman,22 Katsutomo Okamura,8 Marc Perry,18 Sara K. Powell,5 acids; (ii) noncoding transcripts, including 132
Nicole C. Riddle,15 Akiko Sakai,29 Anastasia Samsonova,19 Jeremy E. Sandler,6 Yuri B. Schwartz,3 candidate structural RNAs and 1608 nonstruc-
Noa Sher,22 Rebecca Spokony,4 David Sturgill,12 Marijke van Baren,20 Kenneth H. Wan,6 tural transcripts; (iii) additional Argonaute (Ago)–
Li Yang,14 Charles Yu,6 Elise Feingold,17 Peter Good,17 Mark Guyer,17 Rebecca Lowdon,17 associated small RNA genes and pathways,
Kami Ahmad,29 Justen Andrews,21 Bonnie Berger,1,2 Steven E. Brenner,28,32 Michael R. Brent,20 including new microRNAs (miRNAs) encoded
Lucy Cherbas,21,24 Sarah C. R. Elgin,15 Thomas R. Gingeras,13,16 Robert Grossman,4 within protein-coding exons and endogenous small
Roger A. Hoskins,6 Thomas C. Kaufman,21 William Kent,34 Mitzi I. Kuroda,11 Terry Orr-Weaver,22 interfering RNAs (siRNAs) from 3′ untranslated
Norbert Perrimon,19 Vincenzo Pirrotta,27 James W. Posakony,26 Bing Ren,26 Steven Russell,10 regions; (iv) chromatin “states” defined by com-
Peter Cherbas,21,24 Brenton R. Graveley,14 Suzanna Lewis,9 Gos Micklem,10 Brian Oliver,12 binatorial patterns of 18 chromatin marks that are
Peter J. Park,3 Susan E. Celniker,6§|| Steven Henikoff,25§|| Gary H. Karpen,6,28§|| Eric C. Lai,8§|| associated with distinct functions and properties;
David M. MacAlpine,5§|| Lincoln D. Stein,18§|| Kevin P. White,4§|| Manolis Kellis1,2|| (v) regions of high TF occupancy and replication
activity with likely epigenetic regulation; (vi) mixed
To gain insight into how genomic information is translated into cellular and developmental TF and miRNA regulatory networks with hierar-
programs, the Drosophila model organism Encyclopedia of DNA Elements (modENCODE) project chical structure and enriched feed-forward loops;
is comprehensively mapping transcripts, histone modifications, chromosomal proteins, transcription (vii) coexpression- and co-regulation–based func-
factors, replication proteins and intermediates, and nucleosome properties across a developmental tional annotations for nearly 3000 genes; (viii)
time course and in multiple cell lines. We have generated more than 700 data sets and discovered stage- and tissue-specific regulators; and (ix)
protein-coding, noncoding, RNA regulatory, replication, and chromatin elements, more than predictive models of gene expression levels and
tripling the annotated portion of the Drosophila genome. Correlated activity patterns of these regulator function.
elements reveal a functional regulatory network, which predicts putative new functions for genes, Overview of data sets. Our data sets provide
reveals stage- and tissue-specific regulators, and enables gene-expression prediction. Our results an extensive description of the transcriptional, epi-
provide a foundation for directed experimental and computational studies in Drosophila and genetic, replication, and regulatory landscapes of
related species and also a model for systematic data integration toward comprehensive genomic the Drosophila genome (table S1). Experimental
and functional annotation. assays include high-throughput RNA sequencing
(RNA-seq), capturing-small and large RNAs and
everal years after the complete genetic se- through systematic generation and computational splice variants; chromatin immunoprecipitation

S quencing of many species, it is still unclear

how to translate genomic information into
a functional map of cellular and developmental
programs. The Encyclopedia of DNA Elements
(ENCODE) (1) and model organism ENCODE
(modENCODE) (2) projects use diverse genomic
assays to comprehensively annotate the Homo
sapiens (human), Drosophila melanogaster (fruit
fly), and Caenorhabditis elegans (worm) genomes,
integration of functional genomic data sets.
Previous genomic studies in flies have made
seminal contributions to our understanding of
basic biological mechanisms and genome func-
tions, facilitated by genetic, experimental, compu-
tational, and manual annotation of the euchromatic
and heterochromatic genome (3), small genome
size, short life cycle, and a deep knowledge of
development, gene function, and chromosome
(ChIP)–chip and ChIP followed by high-throughput
sequencing (ChIP-seq), profiling chromosomal
and RNA binding or processing proteins; tiling-
arrays, identifying and measuring replication pat-
terns, nucleosome solubility, and turnover; and
genomic DNA sequencing, measuring copy-
number variation. We conducted most assays in
the sequenced strain y; cn bw sp (13), with mul-
tiple developmental samples (30 for RNA expres-

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1787

 RESEARCH ARTICLES
sion and 12 for TF and histone studies), and in
cultured cells, predominantly with four lines (S2,
BG3, Kc, and Cl.8; table S2).
Annotation of gene transcripts and their pro-
moter regions. To comprehensively characterize
transcribed sequences, we performed RNA-seq
using poly(A)+ and total RNA, cap analysis of
gene expression, rapid amplification of cDNA ends,
and produced expressed sequence tags (table S1)
(14–16) and cDNAs. These data support more than
90% of annotated genes, exons, and splice
junctions and provide experimental evidence for
a total of 17,000 protein-coding and noncoding
genes, of which 1938 are previously unannotated.
In addition to genes, we discovered 52,914
previously undescribed or modified exons (65%
supported by cDNAs) and 22,965 new splice
Developmental
Stages Embryo
RNAPolymerase
Transcription
Spliceosome
DNA splicing
Intron
Extract RNA
RNA
Long RNA Short RNA
mRNA miRNA
hnRNA piRNA
ncRNA siRNA
etc etc
Transcription/Splicing
Fig. 1. Overview of Drosophila modENCODE data sets. Range of genomic elements and trans factors studied, with relevant techniques and resulting genome
annotations. hnRNA, heterogeneous nuclear RNA.
1
Computer Science and Artificial Intelligence Laboratory,
Massachusetts Institute of Technology (MIT), Cambridge, MA
02139, USA. 2Broad Institute of MIT and Harvard, Cambridge,
MA 02140, USA. 3Center for Biomedical Informatics, Harvard
Medical School, 10 Shattuck Street, Boston, MA 02115, USA.
4
Institute for Genomics and Systems Biology, Department of
Human Genetics, The University of Chicago, 900 East 57th
Street, Chicago, IL 60637, USA. 5Department of Pharmacology
and Cancer Biology, Duke University Medical Center, Durham,
NC 27710, USA. 6Department of Genome Dynamics, Lawrence
Berkeley National Laboratory (LBNL), 1 Cyclotron Road,
Berkeley, CA 94720 USA. 7Department of Molecular Genetics,
University of Toronto, 27 King’s College Circle, Toronto, Ontario
M5S 1A1, Canada. 8Sloan-Kettering Institute, 1275 York Av-
enue, Box 252, New York, NY 10065, USA. 9Genome Sciences
Division, LBNL, 1 Cyclotron Road, Berkeley, CA 94720, USA.
10
Department of Genetics and Cambridge Systems Biology
Centre, University of Cambridge, Downing Street, Cambridge,
CB2 3EH, UK. 11Department of Medicine and Department of
Genetics, Brigham and Women’s Hospital, Harvard Medical
School, 77 Avenue Louis Pasteur, Boston, MA 02115, USA.
12
Section of Developmental Genomics, Laboratory of Cellular
and Developmental Biology, National Institute of Diabetes
and Digestive and Kidney Diseases (NIDDK), National In-
stitutes of Health (NIH), Bethesda, MD 20892, USA. 13Cold
Spring Harbor Laboratory, Cold Spring Harbor, NY 11724,
USA. 14Department of Genetics and Developmental Biol-
ogy, University of Connecticut Stem Cell Institute, 263 Farmington,
1788 24 DECEMBER 2010
junctions in 14,016 distinct alternative transcripts
[35% supported by cDNAs, reverse transcription
polymerase chain reaction products, and long
poly(A)+ RNA-seq (14)]. Overall, 74% of annotated
genes show at least one previously undescribed or
modified exon or alternative splice form, despite
extensive previous annotation efforts, illustrating
the importance of probing additional cell types. Of
the 21,071 newly predicted exons expressed in S2
cells, 89% are associated with chromatin signatures
characteristic of transcribed regions (17).
We also characterized the shapes and tran-
scription start site (TSS) distributions for 56% of
annotated genes (70% of embryonically expressed
genes). We discovered and validated 2075 al-
ternative promoters for known genes. Of 427
discovered alternative promoters adjacent to
Larva Pupae
Transcription
Start Site Chromatin &
Histone Binding
Replication Proteins
Nucleosome
Origins
physical
properties
Pre-Replicative
Complex
Histone
Histone
tails
Modifications
Origin Mapping & Variants
Timing
Salt Chromatin
Differential
Fractionation ImmunoPrecipitation
Replication
(ChIP)
Microarray or
Sequence
Replication Epigenetics
CT 06030–6403, USA. 15Department of Biology CB-1137,
Washington University, Saint Louis, MO 63130, USA.
16
Affymetrix, Santa Clara, CA 95051, USA. 17Division of Extra-
mural Research, National Human Genome Research Institute,
NIH, 5635 Fishers Lane, Suite 4076, Bethesda, MD 20892–
9305, USA. 18Ontario Institute for Cancer Research, 101
College Street, Suite 800, Toronto, Ontario M5G 0A3, Canada.
19
Department of Genetics and Drosophila RNAi Screening Cen-
ter, Harvard Medical School, 77 Avenue Louis Pasteur, Boston,
MA 02115, USA. 20Center for Genome Sciences, Washington
University, 4444 Forest Park Boulevard, Saint Louis, MO
63108, USA. 21Department of Biology, Indiana University, 1001
East 3rd Street, Bloomington, IN 47405–7005, USA. 22White-
head Institute, Cambridge, MA 02142, USA. 23Hubrecht In-
stitute, Royal Netherlands Academy of Arts and Sciences and
University Medical Center Utrecht, Utrecht, Netherlands. 24Cen-
ter for Genomics and Bioinformatics, Indiana University, 1001
East 3rd Street, Bloomington, IN 47405–7005, USA. 25Basic
Sciences Division, Fred Hutchinson Cancer Research Center,
1100 Fairview Avenue North, Seattle, WA 98109, USA. 26Divi-
sion of Biological Sciences, Section of Cell and Developmental
Biology, University of California San Diego, 9500 Gilman Drive,
La Jolla, CA 92093, USA. 27Department of Molecular Biology
and Biochemistry, Rutgers University, Piscataway, NJ 08854,
USA. 28Department of Molecular and Cell Biology, University of
California, Berkeley, CA 94720, USA. 29Department of Biolog-
ical Chemistry and Molecular Pharmacology, Harvard Medical
School, 240 Longwood Avenue, Boston, MA 02115, USA.
VOL 330 SCIENCE www.sciencemag.org
active S2 cell transcripts, 72.5% are supported
by promoter-associated chromatin marks in that
cell type (18), confirming predictions and suggest-
ing that these regions contain regulatory ele-
ments. Similarly, comparison to chromatin marks
in whole animals yielded 1117 additional vali-
dated promoters (19).
We detect all but 1498 (9.9%) of previously
annotated D. melanogaster genes (4) in either the
poly(A)+ or total RNA-seq samples. Undetected
genes include members of multicopy gene families
[e.g., ribosomal RNAs, paralogs, small nucleolar
RNAs (snoRNAs), tRNAs] and those with known
low or constrained expression. We discovered new
snoRNAs, scaRNAs, and pri-miRNA transcripts
in the total embryonic RNA-seq data alone, even
without including larval, pupal, or adult samples.
Downloaded from www.sciencemag.org on December 23, 2010
OR
Adult Cell lines
Nucleus
Transcription
Factors
Chromosomes
Nucleosomes
Chromatin
Generate Antibodies
or Tagged Lines
Transcription Regulation
30
Machine Learning Group, Université Libre de Bruxelles,
CP212, Brussels 1050, Belgium. 31Massachusetts General
Hospital Cancer Center, Harvard Medical School, Charlestown,
MA 02129, USA. 32Department of Plant and Microbial Biology,
University of California, Berkeley, CA 94720, USA. 33Computer
and Information Science and Engineering, University of
Florida, Gainesville, FL 32611, USA. 34Center for Biomolecular
Science and Engineering, School of Engineering and Howard
Hughes Medical Institute (HHMI), University of California Santa
Cruz, Santa Cruz, CA 95064, USA. 35Department of Biomedical
Engineering, Stony Brook University, Stony Brook, NY 11794,
USA.
*The complete list of authors appears at the end of the
paper.
†These authors contributed equally to this work.
‡These authors contributed equally to this work (listed
alphabetically).
§These authors contributed equally to this work (listed
alphabetically).
||To whom correspondence should be addressed. E-mail:
manoli@mit.edu (M.K.) (integrative analysis); celniker@fruitfly.
org (S.E.C.) (transcripts); karpen@fruitfly.org (G.H.K.) (chro-
matin); kpwhite@uchicago.edu (K.P.W.) (transcription fac-
tors); david.macalpine@duke.edu (D.M.M.) (replication);
laie@mskcc.org (E.C.L.) (small RNAs); steveh@fhcrc.org
(S.H.) (nucleosomes); lincoln.stein@gmail.com (L.D.S.)
(data availability)

Protein-coding, structural, and noncoding
transcripts. We searched for evolutionary sig-
natures of conserved protein-coding DNA se-
quences in alignments of 12 Drosophila genomes
(12, 20) and for similarity to known proteins.
Only 57 of 1938 previously undescribed gene
models (17) contain a complete, conserved open
reading frame (ORF) likely to represent uniden-
tified protein-coding genes (Fig. 2A). An addi-
tional 81 gene models are likely to be incompletely
reconstructed coding genes, because they contain
at least one protein-coding exon but lack clearly
identifiable translation start or stop sites (17).
These 138 genes show nearly sixfold lower aver-
age expression than known protein-coding genes
[fragments per kilobase of transcript per million
fragments sequenced (FPKM) of 6.7 versus 34.8],
and 40% have expression restricted to late larvae,
predicted transcripts (Fig. 2B), as well as pre-
viously unidentified structural elements in well-
studied ncRNAs, including sex-chromosome
dosage compensation regulator roX2 and heat-
shock regulator HSRw (fig. S1) (17). However,
the lack of highly structured regions in the vast
majority of ncRNAs suggests functions indepen-
dent of secondary structure.
Argonaute-associated small regulatory RNAs.
Our analysis of deeply sequenced ~18- to 28-
nucleotide (nt) RNAs dramatically extended the
catalog of Ago-dependent small regulatory RNAs
(22), including miRNAs, siRNAs, and piwi-
associated RNAs (piRNAs). In the canonical
miRNA pathway, ~21- to 24-nt RNAs are
cleaved from hairpin precursors by Drosha and
Dicer-1 ribonuclease (RNase) III enzymes and
loaded into AGO1 effector complexes to repress
RESEARCH ARTICLES
We and others recognized several classes of
endogenous siRNAs (endo-siRNAs), 21-nt RNAs
that are processed by Dicer-2 RNase III enzyme
and preferentially loaded into AGO2 (26–31).
Endo-siRNAs derive from three distinct sources:
(i) diverse transposable elements (TEs), whose
activity they restrict; (ii) seven genomic regions
encoding long inverted-repeat transcripts, which
direct the cleavage of specific mRNA targets; and
(iii) bi-directionally transcribed regions. This last
class mostly comprises convergent transcripts that
overlap in their 3′ untranslated regions (3′ UTRs),
termed 3′ cis-natural antisense transcripts (3′ cis-
NATs). Our current analysis doubled the number of
3′ cis-NAT–siRNA regions to 237, including near-
ly one-quarter of overlapping 3′ UTRs (table S4).
Lastly, piRNAs are ~24- to 30-nt RNAs
bound by the largely gonadal Piwi-class Argo-

Downloaded from www.sciencemag.org on December 23, 2010

pupae, and adult males, providing a potential
explanation for why they were missed in previous
annotations. For the remaining 1800 gene models,
we find no evidence of protein-coding selection
using PhyloCSF and no similarity to known pro-
tein sequences using blastx, suggesting that they
are unlikely to represent protein-coding genes (20).
We looked for properties of noncoding RNAs
(ncRNAs) among the 1740 transcripts (excluding
60 snoRNA and miRNA transcripts) detected by
RNA sequencing that do not appear to encode
proteins. We examined folding thermodynamics
and comparative evidence of local secondary
structures in the predicted ncRNAs and in 140
ncRNAs listed in FlyBase (4) that do not belong
to major classes of structural RNAs, such as
miRNAs and snoRNAs. We predicted high-
confidence structures for 132 transcripts (7.6%)
using the RNAz program (21), suggesting con-
served function as structural RNAs, similar to the
fraction (7.8%) of transcripts with predicted
structure observed in FlyBase ncRNAs (4). We
revealed candidate structural RNAs in the newly
mRNA targets. We annotated 61 additional ca-
nonical miRNAs, 12 of which are derived from
the antisense strands of known miRNA loci (23),
which may provide an efficient route for the
evolution of new miRNA activities. We unex-
pectedly detected miRNAs that overlap mRNAs,
including nine cases where conserved protein-
coding regions harbor RNA hairpins cleaved into
duplexes of miRNA and partner strand miRNA*
species, many of which are found in AGO1 com-
plexes (e.g., Fig. 2C). It remains to be seen
whether these mRNA-resident miRNAs have de-
tectable trans-regulatory activities, affect their host
transcripts in the cis configuration, or are simply
neutral substrates. We identified 15 additional
mirtrons that generate miRNAs by splicing of
short hairpin introns (24), doubling the number
of known cases from 14 to 29. We defined up to
seven hybrid mirtrons bearing 3′ tails, which
appear to require processing by the exosome
before dicing (25). In total, we recognize at least
three miRNA biogenesis strategies, producing
miRNAs from at least 240 genomic loci.
nautes, Piwi, Aubergine (Aub), and AGO3. The
majority of piRNAs match TEs in sense or an-
tisense orientation and are essential to repress
their activity (32). Though many Drosophila
piRNAs map uniquely to tens of master loci that
serve as genetic repositories for TE defense (32),
we found that the 3′ UTRs of hundreds of cellular
transcripts also generate abundant Piwi-loaded
primary piRNAs in somatic ovarian follicle cells
(33–35). This suggests that beyond transposon
control, the piRNA pathway may play a more
general role in cellular gene regulation.
Large-scale organization of the chromatin
landscape. Eukaryotic genomes are organized
into large domains (~10 kb to megabases) that
exhibit distinct chromatin properties, such as het-
erochromatic regions that cover one-third of the
genome and are typically known for transcrip-
tional silencing (36). Our analyses show that the
chromatin composition, organization, and bound-
aries of heterochromatin display surprising com-
plexity and plasticity among cell types (37). We
find surprisingly active heterochromatic regions,

Fig. 2. Coding and noncoding genes and structures. (A) Extended region of
male-specific expression in chromosome 2R including new protein-coding and
noncoding transcripts. MIP03715 contains two short ORFs of 23 and 21 codons,
respectively. ORF multispecies alignments (color coded) show abundant synony-
mous (bright green) and conservative (dark green) substitutions and a depletion
of nonsynonymous substitutions (red), indicative of protein-coding selection
[ratio of nonsynonymous to synonymous substitutions (dN/dS) < 1 for both, P <
10−7 and P < 10−11, respectively, likelihood ratio test]. Surrounding regions
show abundant stop codons (blue, magenta, yellow) and frame-shifted positions
(orange). (B) A transcribed region in chromosome 3R (26,572,290 to 26,573,456),
identified by RNA-seq and supported by promoter-specific and transcription-
associated chromatin marks, shows RNA secondary-structure conservation in eight
Drosophila species. (C) Example of a new miRNA derived from a protein-coding exon
of CG6700, with 21- to 23-nt RNAs indicative of Drosha/Dicer-1 processing and also
recovered in AGO1-immunoprecipitate libraries from S2 cells and adult heads
indicative of Argonaute loading. Evolutionary evidence suggests protein-coding
constraint, no conservation for the mature arm, and conservation of the star arm. Red
boxes indicate 8-mer “seed” sequence potentially mediating 3′ UTR targeting.

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1789

 RESEARCH ARTICLES
with expression of 45% of pericentric hetero-
chromatin genes (compared with 50% for eu-
chromatic genes), and enrichment for both active
and silent marks in active heterochromatic genes.
Conversely, we find that domains enriched for
heterochromatic marks (e.g., H3K9me2) cover a
surprisingly large proportion of euchromatic
sequences (12% in BG3 cells and 6% in S2) (37).
We identified large domains with similar rep-
lication patterns by characterizing the Drosophila
DNA replication program in cell lines, and we
observed that the temporal replication program
is determined by local chromatin environment
(18, 38) and the density of replication initiation
factors (39). We also found that specific euchro-
matic regions up to 300 kb were under-replicated
in a tissue-specific manner in the polytene sali-
vary glands, larval midgut, and fat bodies (40),
insulator-binding proteins, and TFs to character-
ize chromatin signatures of each type of element
(Fig. 3A). TSS-proximal regions were marked by
H3K4me3 enrichment (45), depletion of nucleo-
some density, increased nucleosome turnover,
and enrichment in the pellet chromatin fraction
(43, 44). Gene bodies showed H2B ubiquitination
covering the entire transcribed region and a 3′-
biased enrichment of H3K36me3 and K3K79me1
marks. Moreover, large introns are enriched for
H3K36me1, H3K18ac, and H3K27ac; specific
chromatin remodelers; high nucleosome turn-
over; the H3.3 histone variant; and DNase I
hypersensitive sites, all suggestive of regulatory
functions (18). These features are generally absent
from short genes and from genes with a low
fraction of intronic sequence. Most transcription-
ally silent genes lack pronounced chromatin sig-
structure (43), consistent with the biological func-
tions associated with the corresponding marks.
We mapped origins of replication activated
early in the S phase of the cell cycle and binding
sites of the origin recognition complex (ORC), a
conserved replication initiation factor that ex-
hibits little, if any, sequence specificity in vitro
(46, 47). ORC-associated sequences are often found
at TSSs and depleted for bulk nucleosomes, but
are enriched for the variant histone H3.3 (39) and
undergo active nucleosome turnover (43). These
findings suggest that local nucleosome occupan-
cy and organization are determinants of ORC
binding in Drosophila, as in yeast (48, 49). By
subdividing the ORC sites into TSS-proximal
and -distal sites, we found that local enrichment
for GAGA factor (GAF), and H4Ac tetra,
H3K27Ac, H4K8Ac, and H3K18Ac are com-

Downloaded from www.sciencemag.org on December 23, 2010

which suggests that copy-number variation may
help regulate gene expression levels.
Chromatin signatures characteristic of func-
tional elements. Many genomic regulatory regions
are difficult to identify because of a lack of char-
acteristic sequence signatures, but they are often
marked by specific histone modifications, var-
iants, and other epigenetic factors (41, 42). To
identify such signatures, we assayed 18 histone
modifications and variants by ChIP-chip in mul-
tiple cell lines (18) and developmental stages (19),
and we defined the physical properties of nu-
cleosomes (43, 44). We correlated this informa-
tion with gene annotations, transcriptome data
sets, binding site profiles for replication factors,
natures, except when positioned within Pc domains
(H3K27me3) or heterochromatin (H3K9me2/3,
HP1a, H3K23ac depletion) (37).
Positional correlation analysis identified rela-
tionships between histone marks and nucleosome
physical properties. Active marks [e.g., H3K27Ac,
RNA polymerase II (RNA Pol II), H3K4me3]
correlate with high chromatin solubility and high
nucleosome-turnover rates, whereas marks asso-
ciated with silent chromatin (e.g., H3K27me3,
H1, H3K9me2/3) show the opposite, correlating
with increased nucleosome density (fig. S2).
High chromatin solubility indicates less stable nu-
cleosomes (44), and high levels of nucleosome
turnover are indicative of a dynamic chromatin
mon to both, whereas H3K36me1 appears to be
specific for TSS-distal ORC sites (Fig. 3A). ORC
marks sites of cohesin complex loading in
Drosophila (38); H3K36me1, which is also
enriched at cohesin sites (18), may be required
in the absence of TSS-associated marks to
promote ORC binding and subsequent cohesin
loading (50, 51).
Insulator elements and proteins (e.g., CP190,
CTCF, SUHW, and BEAF) block enhancer-
promoter interactions and restrict the spread of
histone modifications (52). Analysis of the ge-
nomic distributions of insulator proteins showed
that BEAF32, CP190, and ZW5 preferentially
bind upstream of TSSs, whereas SUHW binds

Fig. 3. Chromatin-based annotation of functional elements. (A) Average
enrichment profiles of histone marks, chromosomal proteins, and physical
chromatin properties at genes, origins of replications, insulator proteins, and
TF binding positions. Each panel shows 4 kb centered at a specified location,
either proximal to TSS (prox.) or distal (dist.). (B) Example of a transcript
predicted by chromatin signatures associated with promoter (red trace) and
gene bodies (blue box) and supported by cDNA evidence. Strong RNA Pol II
and H3K4me3 peaks in the promoter region and strong H2B ubiquitination
extending toward the previously annotated luna gene are confirmed by RNA-
seq junction reads that were not used in the prediction. (C) Intergenic
H3K36me1 chromatin signatures predict replication activity. Enrichment of
multiple chromatin marks were used to identify putative large (>10 kbp)
intergenic H3K36me1/H3K18ac domains located outside of annotated genes.
Although these marks generally correspond to long introns within transcripts,
their intergenic domains were enriched for replication activity (fig. S5). In this
example from BG3 cells, such a domain was found upstream of the bi locus
and is associated with early replication, contains an early origin, is enriched
for ORC binding, and is further supported by NippedB binding.

1790 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


almost exclusively distal to TSSs, with CTCF bind-
ing both equally (53). Insulator regions displayed
distinct chromatin signatures (Fig. 3A), but most
of the variation is explained by the differences
between TSS-proximal and -distal chromatin con-
texts, suggesting that specific marks are not re-
quired for insulator binding or function. However,
nucleosome depletion is a common feature of both
TSS-proximal and -distal insulator binding sites,
as in mammals (54), a property that may facilitate
insulator binding or reflect the ability of insulator
proteins to displace nucleosomes.
Chromatin-based annotation of functional
elements. Chromatin signatures associated with
TSSs and transcribed regions (45) identified
genes and promoters missed by transcript-based
annotation. We developed a predictive model for
active promoters in cell lines using positional
of the intergenic H3K36me1 domains overlapped
with cohesin (18) and early origin activity, re-
spectively, as observed for a 20-kb region upstream
of the bi gene (Fig. 3C and fig. S5). Although
only 15% of early replication origins appear to be
defined by intergenic H3K36me1 domains, the
overlap with cohesion enrichment (18) suggests a
shared mechanism to ensure faithful chromo-
some inheritance.
De novo discovery of combinatorial chroma-
tin states. Multiple histone modifications act in
concert to determine genome functions pro-
ducing combinatorial chromatin states (55). We
used two unsupervised, multivariate hidden Markov
models to segment the genome on the basis of the
combinatorial patterns of 18 histone marks in S2
and BG3 cells (Fig. 4 and fig. S6) (18). We did
not seek a true number of distinct chromatin
RESEARCH ARTICLES
d26, d28, and d29 (c7 and c8) in heterochromatin
(characterized by H3K9me2/3 enrichment and
H3K23ac depletion). These states lack enrich-
ments for other mapped factors [e.g., insulators,
histone deacetylases (HDACs), TFs] and exhibit
low levels of chromatin solubility and nucleo-
some turnover.
In contrast, expressed genes display numer-
ous and complex enrichments for several factors
and chromatin properties. Most active TSSs were
associated with state c1, defined by known promoter-
associated marks H3K4me3 and H3K9ac (45).
Other active TSSs were additionally enriched for
H3K36me1 and multiple acetylations (d13). Even
within c1, some TSSs showed higher association
with nucleosome turnover, group 1 insulator pro-
teins and HDACs (d1, d3), whereas others were
associated with heterochromatic genes of medium

Downloaded from www.sciencemag.org on December 23, 2010

enrichments of 18 histone marks, ORC complex
localization, and nucleosome stability and
turnover in the 1-kb regions surrounding vali-
dated active promoters. Our logistic regression
classifier achieved 93.7% sensitivity at a 21.5%
false discovery rate (FDR) (fig. S4) and predicted
2203 additional promoter positions at least 500
base pairs (bp) away from annotated TSSs (17).
These included promoters for 10 primary miRNA
transcripts, of which 7 were also identified by
RNA-seq (14). We also used H3K36me3/H2B-
ubiquitination signatures (fig. S3) to identify 53
transcribed gene bodies outside annotated genes,
11 of which are additionally supported by promoter
predictions (e.g., Fig. 3B). These included four
primary miRNA transcripts, of which three are also
supported by RNA-seq (14) and one is also sup-
ported by our promoter predictions (for mir-317).
Chromatin signatures also identify functional
elements involved in other chromosomal pro-
cesses such as duplication and segregation. We
identified 133 sites in BG3 and 78 sites in S2
cells that contained large (>10-kbp) intergenic do-
mains of H3K36me1. In BG3 cells, 90 and 68%
states; instead, we sought to identify models that
balance resolution and interpretability given the
available chromatin marks, as more states led to
increased enrichment for specific genomic features
but captured progressively smaller fractions of
each type of feature (fig. S7).
From these considerations, we focused on a
9-state, intensity-based model reflecting broad
classes of chromatin function (continuous model
states c1 to c9) and a 30-state model that iden-
tifies combinatorial patterns at a finer resolution
(discrete model states d1 to d30) (Fig. 4, left panel)
(17). These showed distinct functional and ge-
nomic enrichments (Fig. 4, right panel) associated
with different chromosomes (chromosome 4, male
X), regulatory elements (promoters, enhancers),
gene length and exonic structure (e.g., long first
introns), gene function (e.g., developmental regu-
lators), and gene expression levels (high or
medium, low, or silent).
Intergenic regions and silent genes are as-
sociated with state d30 (c9) in euchromatin (cov-
ering 51% of the genome and lacking enrichments
for any of the marks examined) and with states
(d5) or low expression (d6).
The state analysis also captured the correla-
tion between ORC binding and TSSs for both
euchromatin and heterochromatin, as well as the
correlation between early origins and open chro-
matin in euchromatic regions. However, ORC
binding is largely limited to a subset of TSS-
associated states (d1, d5, d6, d13, d17, and not d3
or d24), and some states enriched for ORC bind-
ing are not found at TSSs (d11, d14, d21). Early
origins are primarily associated with states c3
(active intron, enhancer) and c4 (open chromatin)
and often display distinct state enrichments from
ORC binding in accord with the broad domains
they cover, compared with the near nucleotide
resolution of the ORC binding data.
Our states showed some similarities with the
recently published five “colors” of chromatin from
DNA adenine methyltransferase identification–
mapped chromosomal proteins in Kc cells (56), but
even highly specific states were sometimes split
across multiple colors (fig. S8). This suggests a more
complex picture with many highly specific chro-
matin states with specific functional enrichments.

Fig. 4. Discovery and characterization of chromatin states and their
functional enrichments. Combinatorial patterns of chromatin marks in S2
and BG3 cells reveal chromatin states associated with different classes of
functional elements. A discrete model (states d1 to d30) captures the
presence/absence information, and a continuous model (states c1 to c9) also
incorporates mark intensity information (22). States were learned solely
from mapped locations of marks (left) and were associated with
modENCODE-defined elements (right) with most pronounced patterns in
euchromatin (green) and heterochromatin (blue) shown here (additional
variations shown in fig. S6).

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1791

 RESEARCH ARTICLES
Chromatin and motif properties of high-
occupancy TF binding sites. Extensive overlap in
the binding profiles of multiple TFs has revealed
highly occupied target (HOT) regions or hotspots
(19, 57–61). Using the binding profiles of 41 TFs
in early embryo development, we assigned a TF
complexity score to each of 38,562 distinct TF
binding sites corresponding to the number of
distinct TFs bound (from 1 to ~21), resulting in
1962 hotspots with TF complexity of eight or
greater, corresponding to ~10 overlapping factors
bound (19). We correlated these regions with our
and other data sets to gain insight into the possible
mechanisms of HOT region establishment and
how they may impact or be affected by chromatin
properties.
We studied the enrichment of regulatory mo-
tifs for 32 TFs for which we have both genome-
varying complexity levels. In all eight cases, motif
matches were preferentially found in regions of
lower complexity, which is suggestive of non-
specific binding. For an additional 9 TFs, bound
regions were enriched in the known motif, but no
bias for lower-complexity regions was found; for
another 10 factors, the known motif did not show
a substantial enrichment in bound regions, sug-
gesting that either the motif is incorrect, or a larger
fraction of TFs than previously expected binds in
non–sequence-specific ways.
We found a strong correlation between HOT
spots of increasing TF complexity and decreased
nucleosome density (fig. S9A) (19), increased
nucleosome turnover (fig. S9B), and histone
variant H3.3, which is associated with nucleo-
some displacement (fig. S9C), but a surprising
depletion in previously annotated enhancers (19),
across all complexity ranges. We also found
concordance between HOT regions and ORC
binding sites (Fig. 5B), with the likelihood of
ORC binding increasing monotonically with the
complexity of the TF-bound regions. Coupled
with the lack of a detectable specific sequence for
ORC binding in Drosophila (39), this suggests
hotspots as an alternative mechanism for ORC
localization via nonspecific binding in high-
accessibility regions, as well as widespread inter-
play between chromatin regulation, TF binding,
and DNA replication. Given the high agreement
between embryo and cell-line data sets, we pro-
pose that hotspots are stable genomic regions,
kept open via recruitment of specific chromatin
marks or remodelers, that facilitate binding of
additional TFs at their motifs or nonspecifically.
We looked for potential “driver” motifs that

Downloaded from www.sciencemag.org on December 23, 2010

wide bound regions and well-established regulatory suggesting potentially distinct roles for these may be recognized by TFs potentially involved
motifs (Fig. 5A). We sorted each TF on the basis elements. We observed enrichment for HOT in establishing HOT regions (Fig. 5C). Applying
of its average complexity [the average number of regions across a wide range of complexity values our motif-discovery pipelines (19) within bound
TFs that co-bind (19)], which ranges from 10.8 for several chromatin states associated with TSS regions of varying complexity resulted in seven
for KNI to 1.3 for FTZ-F1. We studied the rel- and open chromatin regions (d1, d5, d6, d13, distinct motifs associated with hotspots of dif-
ative enrichment of each factor’s known motif in d14, d21), whereas some states (d3 and d24) ferent complexities. Motifs M2 and M3 were
bound regions and found eight factors (KNI, were enriched only at lower complexity (fig. similar to the BEAF-32 and Trl/GAF insulator
DLL, GT, PRD, KR, SNA, DA, and TWI) with S9D). In contrast, transcriptional elongation (d7 motifs, suggesting interplay between hotspots and
average complexity greater than four that showed to d9), intergenic (d30), and heterochromatic insulator proteins. Motif M1 differed in only one
significant differences in motif enrichment at states (d26, d27, d29) were strongly depleted position from the known Sna motif and was strong-
ly enriched for high-complexity regions (Fig. 5C),
whereas the Sna motif was depleted in Sna-bound
regions of higher complexity (Fig. 5A), suggest-
ing that the single-nucleotide difference may be
important for recognition. The other four motifs
did not match any known TFs, suggesting that
yet-uncharacterized potential sequence-specific
regulators may be involved in the establishment
of hotspots.
Fraction of the genome assigned to can-
didate functions. We assigned candidate func-
tions to the fraction of the nonrepetitive genome
covered by the data sets, excluding large blocks
of repeats and low-complexity sequences (Fig. 6A).
Protein-coding exons cover 21% of the genome,
and adding Argonaute-associated small regula-
tory RNAs, UTRs, other ncRNAs, bases covered
by Pol II, the binding sites of TFs, and other
chromatin-interacting factors brings the total ge-
nome coverage to 73%. Inclusion of Pc and ORC
binding sites, and derived chromatin states, brings
the total genome coverage to 81.5%, and the ad-
dition of transcribed intronic positions raises the
total coverage to more than 89% (Fig. 6A). Com-
pared with previous annotations [FlyBase (4)],
we have increased coverage of the Drosophila ge-
nome with putative associated functions by 26.3%
Fig. 5. High-occupancy TF binding regions and their relation to motifs, ORC, and chromatin. (A) En- (47 Mb). Euchromatic regions had much higher
richment of known motifs for regions bound by corresponding TF, sorted by average complexity, denoting coverage than heterochromatic regions (90.6
the number of distinct TFs bound in the same region. For eight TFs, motifs are depleted (blue) for higher- versus 69.5%) in a comparison of the respective
complexity regions, suggesting non–sequence-specific recruitment. In seven of eight cases, known motifs nonrepetitive portions.
were enriched in bound regions (Enrich), suggesting sequence-specific recruitment in lower-complexity We next determined the overlap between our
regions. For each factor, binding sites were highly reproducible between replicates (Reprod). (B) ORC predicted functional elements and PhastCons evo-
versus TF complexity. The relation between HOT spot complexity (x axis) and enrichment in ORC binding lutionarily conserved elements across 12 Dro-
(y axis). (C) Discovered motifs in high- or low-complexity regions (boxed range) and their enrichment in sophila species, mosquitoes, honeybees, and beetles
regions of higher (red) or lower (blue) complexity. M1 to M5 are candidate “drivers” of HOT region (62). These elements cover 38% of the D. melano-
establishment. gaster genome in 1.2 million blocks, over which

1792 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


we repeated our previous individual and cumu-
lative calculations. Thirty-two percent of con-
strained bases are covered by protein-coding exons
alone, increasing to a cumulative total of 80% for
transcribed and regulatory elements and 91.8%
after inclusion of specific chromatin states (Fig. 6A).
Nearly all modENCODE-defined functional ele-
ments were more likely to cover constrained bases
than is expected by chance, providing additional
independent evidence for the predicted elements
(fig. S10). The only exceptions were some less
active chromatin states, as expected, and introns,
UTRs, and ncRNAs (63) providing additional in-
dependent evidence for the predicted elements.
Overlap among the annotations produced by
different types of elements resulted in dense mul-
tiple coverage (Fig. 6B), even for regions that
previously lacked any annotation (Fig. 6C). Even
though the genome coverage average is 2.8 data
sets, 10.8% of the genome is covered by 15 or
more data sets, and coverage peaks at 103 data
sets overlapping a single region on chromosome
3R. We found strong positive correlations be-
tween bound regulators and transcribed element
densities, as well as regulators and chromatin el-
ement densities (fig. S11). In the case of chro-
matin data sets, additional chromatin marks
resulted in higher accuracy in chromatin-state re-
covery (fig. S12), and we expect similar addi-
tional data sets to have an effect on other classes
of functional elements.
TF targets and physical regulatory network
inference. We examined the network of regu-
latory relationships between TFs, miRNAs, and
RESEARCH ARTICLES
their target genes. In these networks, “nodes”
represent the transcriptional and posttranscrip-
tional regulators and target genes, and “edges” or
“connections” represent their directed regulatory
relationships. We inferred a physical regulatory
network of TF binding and miRNA targeting,
where connections represent physical contact be-
tween regulators and genomic regions of their
target genes.
The structural properties of the physical
regulatory network were inferred from the ex-
perimentally derived binding profiles of 76 TFs
(table S5) and genome-wide occurrences of 77
distinct evolutionarily conserved miRNA seed
motifs for 105 miRNAs (17). The structure of the
resulting network shows high connectivity and
rapid spread of regulatory information, requiring
traversal of only ~two regulatory connections, on

Downloaded from www.sciencemag.org on December 23, 2010

average, between any two genes and no more
than five connections between any pair of genes.
Target genes are regulated by ~12 TFs, on av-
erage, and can have up to 54 regulatory TFs (17).
The most heavily targeted genes are associated
with increased pleiotropy, as measured by the
number of distinct functional processes and tis-
sues with which they are associated (17).
The physical regulatory network includes both
pre- and posttranscriptional regulators, identify-
ing the interplay between these two types of reg-
ulation. We organized the TFs of the physical
regulatory network into five levels (Fig. 7A and
fig. S13) on the basis of the relative proportion of
TF targets versus TF regulators for each TF (64),
and we augmented this network with the miRNA
regulators most closely interacting with each lev-
el. The presumed “master regulator” TFs at the
top level targeted almost all of the other TFs in
the network, whereas only 8% of lower-level edges
pointed upward to higher levels, supporting a
hierarchical nature and suggesting little direct feed-
back control of master regulators among the TFs
surveyed. We also observed that even though the
number of TF targets decreases for TFs at lower
levels of the hierarchy, the number of their miRNA
targets increases (0.58 miRNA targets per TF for
the two topmost levels versus 1.55 for the two
lowest levels, fold enrichment of 2.66). This sug-
gests that at least some feedback from the lower
levels to the master regulators may occur in-
directly through miRNA regulators.
We next searched for significantly overrep-
resented network connectivity patterns, or “net-
work motifs” (Fig. 7B), likely to represent building
blocks of gene regulation (65). We found eight
network motifs in the physical regulatory net-
work (66), five of which correspond to TF co-
operation (motifs 1, 2, 4, 7, and 8), confirming
observations of cobinding and cotargeting (57–61).
Fig. 6. Genome coverage by modENCODE data sets. (A) Unique (bars) and cumulative (lines) coverage of In all five motifs, at least two TFs bind each
nonrepetitive (blue line) and conserved (red line) genomes. (B) Multiple coverage for data sets grouped other’s promoter regions, suggesting extensive
into transcribed elements (red), bound regulators (blue), and chromatin domains (green) (17). Across all positive and negative feedback. Two other motifs
three classes (black), 10.8% of the genome is covered 15 or more times, and 69.5% is covered at least correspond to mixed feed-forward loops involv-
twice. (C) Increased coverage in a Chr2R region with no prior annotation (left half), now showing multiple ing cooperation of TFs and miRNAs (motifs 3
overlapping data sets. Coverage by different tracks is highly clustered (fig. S11), with some regions and 6), which can lead to different delay proper-
showing little coverage and others densely covered by many types of data. ties in the expression of target genes depending

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1793

 RESEARCH ARTICLES
on the activating or repressive action of the TF.
Lastly, one motif (motif 5) corresponds to a
feedback loop of a downstream TF targeting an
upstream TF through a miRNA, which is also
observed as a means for feedback in the hie-
rarchical network layout (17).
Data set integration predicts a functional
regulatory network. We integrated the physical
network with patterns of coordinated activity of
regulators and targets to derive a functional reg-
ulatory network (fig. S14A). Although TF bind-
ing is strongly associated with the true regulatory
targets, binding alone can occur without a sequence-
specific TF-motif interaction and does not always
result in changes in gene expression (60). Thus, a
functional regulatory network should consider both
binding and its functional consequences, such as
changes in expression or chromatin, which are cor-
curated functional network (17). The functional
network included a similar number of target genes
as both the binding and motif physical networks
(~10,000 targets each), but more regulators over-
all (576 versus 104 and 76, respectively) and
more regulators per target (24 versus 7 and 13,
respectively) (fig. S14B). The functional network
showed similarity to both the motif and binding
networks, which were both used as input evi-
dence; connections of the functional network
showed more than fourfold enrichment in both
networks, even though the two only showed a
1.6-fold enrichment to each other’s connections
(fig. S14C). Compared with either the motif or
the binding network, the functional network
showed the strongest connectivity similarity to
the REDfly network, even though it was not
specifically trained to match known edges.
(fig. S15) (17) and predicted likely functional
GO terms for every gene with a guilt-by-
association approach that uses GO terms of anno-
tated genes to predict likely functions of unannotated
genes, allowing for multiple annotation predic-
tions for each gene (17). This resulted in a higher
predictive power than the use of expression or
regulators alone (Fig. 8). At FDR < 0.25, we
predicted GO terms for 1286 previously unan-
notated genes and additional terms for 1586 pre-
viously annotated genes (fig. S17, table S6, data
set S15). In general, tissue-specific enrichments of
new GO predictions matched those of known
genes in the same GO terms (fig. S18), providing
an independent validation of our approach.
Predicting stage-specific regulators of gene
expression. We predicted stage-specific regula-
tors of gene expression on the basis of tran-

Downloaded from www.sciencemag.org on December 23, 2010

related with gene function (fig. S15). Neither net-
work is a strict subset of the other, as some physical
connections may not lead to functional changes,
and functional connections may be indirect or
simply missing in the physical regulatory map.
We integrated multiple types of evidence in-
cluding conserved sequence motifs of 104 TFs in
promoter regions across the genome (table S5),
ChIP-based TF binding for 76 factors, and the
correlation between chromatin marks and gene
expression patterns of regulators and their target
genes (fig. S16). We combined these lines of
evidence with unsupervised machine learning to
infer the confidence of each regulatory edge be-
tween 707 proteins classified as TFs (17) and
14,444 targets for which at least one line of
evidence was available (17).
We compared the resulting functional net-
work to the physical network inferred from TF
binding, a predicted physical network constructed
from motif occurrences, and the REDfly literature-
The functional regulatory network showed
increased biological relevance compared with
both the motif and binding networks, including
increased functional similarity, increased expres-
sion correlation, and increased protein-protein in-
teractions of cotargeted genes (fig. S14D) (17).
The REDfly network slightly outperformed the
functional network, confirming the relevance of
the metrics. However, the functional network con-
tains 100 times more targets (9436 versus 88) and
1000 times more connections (231,181 versus
233) than the REDfly network, suggesting it will
be more valuable for predicting gene function and
gene expression at the genome scale.
Predicting gene function from the functional
regulatory network. We provided candidate
functional annotations for genes that lack Gene
Ontology (GO) terms on the basis that targets of
similar regulators and with similar expression are
likely to share similar functions. We probabilis-
tically assigned genes to 34 expression clusters
scriptional changes during development. With
the Dynamic Regulatory Events Miner (DREM)
(67), we searched for splits (a point at which pre-
viously coexpressed genes begin to exhibit diver-
gence into two or three distinct expression
patterns) among a set of more than 6000 genes
with the largest expression changes occurring
during the developmental time course (Fig. 9A
and fig. S19). We mined the physical and
functional regulatory networks to predict stage-
specific regulators from the over-representation
of regulator targets along specific trajectories or
“paths” from each split (17). Several predictions
agreed with literature support. For example, TIN,
a known regulator of organ development (68),
was a predicted regulator of genes with an early
increase in expression and enriched for organ de-
velopment (P < 10–53), and E2F2, a known cell-
cycle regulator (69), was a predicted regulator of
genes with an early decrease in expression and
enriched for cell-cycle function (P < 10–100).

Fig. 7. Properties of the physical regulatory network. (A) Hierarchical view of
mixed ChIP-based/miRNA physical regulatory network that combines transcrip-
tional regulation by 76 TFs (green) from ChIP experiments and posttranscriptional
regulation by 52 miRNAs (red). TFs are organized in a five-level hierarchy on the
basis of their relative proportion of TF targets versus TF regulators. miRNAs are
separated into two groups: the ones that are regulated by TFs (left) and the ones
that only regulate TFs (right). The horizontal position of the TFs in each level shows
whether they regulate miRNAs (left), have no regulation to or from miRNAs
(middle), or do not regulate but are targeted by miRNAs (right). Different shades
of green and red represent the total number of target genes for TFs and miRNAs,
respectively (darker nodes indicate more targets). Ninety-two percent of TF reg-
ulatory connections are downstream connections from higher levels to lower levels
(green), and only 8% are upstream (blue). miRNA regulatory connections are red.
(B) Highly enriched network motifs in a mixed physical regulatory network in-
cluding TFs (green), miRNAs (red), and target genes (black). For each motif, five
examples are shown. Known activators, blue; known repressors, red; other TFs, black.

1794 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


To provide additional support for regulator
predictions made using the physical network, we
examined the time-course expression profiles of
the regulators, which were not directly used in the
prediction scheme. Even though several caveats
could hinder this analysis, the time-course ex-
pression of the regulators was often consistent
with DREM’s predictions. For example, a sharp
decline in SU(HW) expression coincides with sharp
expression increase of its targets (Fig. 9A), con-
sistent with a repressive role (70). We generally
observed a notable correspondence among the
Fig. 8. Gene function prediction from
coexpression and co-regulation patterns. Re-
ceiver operator characteristic curves for GO
stage-specific expression changes of predicted
regulators at developmental stages that correspond
with concomitant expression changes in their tar-
get genes. Regulators predicted to be associated
with a split had, on average, a significantly great-
er absolute expression change than those not
associated with a split (P < 10−10) (fig. S19) (17).
Predicting cell type–specific regulators of
chromatin activity. We computed enrichments
of conserved regulatory motif instances in cell
type–specific annotations for 22 chromatin fac-
tors in both S2 and BG3 cells. We defined signa-
RESEARCH ARTICLES
tures of cell-type–specific activators and repressors
probably involved in establishing the chromatin
differences between S2 and BG3 cells (Fig. 9B)
by comparing these enrichments to the expres-
sion patterns of the TFs that recognize these mo-
tifs in the same cell types (17). Activators were
defined as TFs whose cell type–specific expres-
sion coincided with activation of their predicted
targets, and repressors were defined as TFs whose
cell type–specific expression was correlated
with repression of their predicted targets. This
resulted in one to eight predicted regulators for
each cell, including, for example, CREBA as a
predicted S2 activator, H as a predicted BG3
repressor, and factors with the stereotypical homeo-
box binding motif (HOX-like) as a predicted BG3
activator.
For most regulatory motifs, enrichment in ac-

Downloaded from www.sciencemag.org on December 23, 2010

terms with predicted new members and
tivating chromatin marks was coupled with
area-under-the-curve statistics. False neg-
depletion in repressive chromatin marks. This
atives for each GO term are predictions for
genes previously annotated for “incompatible” coupling leads to more robust predictions of ac-
GO terms, defined as pairs of GO terms that tivators and repressors and also enables a high-
have less than 10% common genes relative level distinction between active and repressive
to the union of their gene sets. chromatin marks that agrees with previous studies
and with our chromatin-state analysis (Fig. 4)
(18, 19). For a small number of motifs, however,
the chromatin enrichments did not show a con-
sistent picture of opposite enrichments in activat-
ing versus repressive marks. These could be false
positives and not actually associated with chro-
matin regulation, or they could be active in other
cell types and not relevant to the distinction be-
tween S2 and BG3 chromatin marks.

Fig. 9. Predictive models of regulator, region, and gene activity. (A) Dynamic
regulatory map produced by DREM predicts stage-specific regulators
associated with expression changes (y axis, log space relative to first time
point) across developmental stages (x axis) (17). Each path (colored lines)
indicates the average expression of a group of genes (solid circles) and its
standard deviation (size of circle). Predicted bifurcation events, or splits, (open
circles) are numbered 1 through 19. The colored insets show the expression
level of each individual gene going through the split and ranked regulators
from the physical (black) or functional (blue) regulatory network associated
with the higher (H), lower (L), or middle (M) path. The uncolored inset shows
the expression of repressor SU(HW), whose expression decrease coincides with
an expression increase of its targets (red asterisk). (B) Predicted S2 activators
(top group) or repressors (bottom group), based on the coherence between
relative expression of the TF in S2 (yellow) versus BG3 (green) and the relative
motif enrichment (red) or depletion (blue) in S2 versus BG3 for activating (left
columns) or repressive marks (right columns). (C) True (top of shaded area)
and predicted (dotted blue line) expression levels for target genes, from the
expression levels of inferred activators (red) and repressors (green). Only the
top five positive and negative regulators are shown, ranked by their
contribution to the expression prediction (weight of linear-regression model).
Examples are shown from 8 of 1487 predictable genes, ranked by prediction
quality scores (rank in upper right corner), evaluated as the averaged squared
error between predicted and true expression levels across the time course. An
expanded set of examples is shown in fig S23.

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1795

 RESEARCH ARTICLES
Predicting target gene expression from
regulator expression. Developmental regulatory
programs are defined by multiple interacting reg-
ulators contributing to observed changes in gene
or region activity (71). We sought to predict the
specific expression levels of target genes across
numerous stages and cell lines on the basis of the
expression levels of their regulators. With the 30
distinct measurements of expression levels ob-
tained by RNA-seq across development (14), we
represented the expression level of each target
gene as a linear combination of its regulators, as
defined by the functional regulatory network (Fig.
9C). We split the time course into 10 intervals of
three samples each and learned stable coefficients
for linear combinations of TFs across 9 intervals
to predict expression in the tenth (17).
We predicted the expression levels of 1991
genes better than random control networks (23.6%
of genes), a 2.5-fold enrichment (control net-
works perform better on 9.5% of genes) (figs.
S20 and S21). In contrast, physical networks showed
almost no predictive value over the randomized
networks (table S7), suggesting that they are best
used when combined with additional information
for inferring functional regulatory networks.
Genes whose expression levels are predicta-
ble from the expression levels of their regulators
(those with consistently lower errors than ran-
dom) may be more precisely regulated and, thus,
associated with less noisy expression patterns.
Indeed, the expression correlation between the
30–time-point data set used for expression pre-
diction (14) and an independently generated 12–
time-point data set sampled at longer intervals
(19) was significantly higher for predictable genes
compared with unpredictable genes (Kolmogorov-
Smirnov test P value < 1E–7) (fig. S22). These
results validate our methodology for gene ex-
pression prediction and suggest that unpredict-
able genes may be due to intrinsic variability in
gene expression levels.
We also tested whether the regulatory models
obtained with whole-embryo time-course data
sets can predict gene expression under novel con-
ditions: specifically the Cl.8+, Kc167, BG3, and
S2-DRSC cell lines. For each “predictable” gene,
the expression levels of its regulators were
combined, as dictated by the weights learned in
the time-course experiment, and used to predict
target gene expression. The expression of 932
predictable genes also showed better-than-random
predictions (compared with 296 genes for the
binding network and 214 genes for the motif
network). Overall, 62% of embryo-defined pre-
dictable genes were also predictable in cell lines,
compared with only 10 to 15% for embryo-based
unpredictable genes, providing further validation
of our methodology.
Our results suggest that the primary data sets
are highly relevant for inferring functional reg-
ulatory relations that are predictive of expression
(Fig. 9C and figs. S20 and S23). However, genome-
scale gene expression prediction remains an enor-
mously difficult problem, as only one-quarter of
1796 24 DECEMBER 2010
all genes was predictable, a fraction that we ex-
pect to improve with additional data sets gen-
erated from more and more genome-scale projects.
Discussion. This first phase of the mod-
ENCODE project has provided the foundation
for integrative studies of metazoan biology, en-
hancing existing genome annotations; broadening
the number and diversity of small RNA genes and
pathways; revealing chromatin domains and sig-
natures; and elucidating the interplay between
replication, chromatin, and TF binding in high-
occupancy regions. Together, our resulting anno-
tations cover 82% of the genome, a nearly fourfold
increase compared with previously annotated protein-
coding exons, and have important implications
for interpreting the molecular basis of genetically
linked phenotypes.
Our integrative analysis revealed connections
between elements in physical and functional reg-
ulatory networks, enabling the prediction of gene
function, tissue- and stage-specific regulators, and
gene expression levels. Though our initial results
are promising, only one-quarter of all genes showed
predictable expression, suggesting the need for
continued mapping of regulatory interconnec-
tions and functional data sets, as well as new
predictive models.
It remains to be seen how the general reg-
ulatory principles elucidated here will be con-
served across the animal kingdom and especially
in humans, through comparison across the
ENCODE and modENCODE projects. Toward
this end, we are expanding our exploration of
functional elements, cell types, and developmen-
tal stages and prioritizing orthologous assays and
conditions across species. Given the extensive
conservation of biological molecules and pro-
cesses between flies and vertebrates (72), these
will not only improve our understanding of fly
biology, but can also serve as a template for
understanding of human biology and disease.
References and Notes
1. www.genome.gov/10005107
2. S. E. Celniker et al., Nature 459, 927 (2009).
3. R. A. Hoskins et al., Science 316, 1625 (2007).
4. Compared to FlyBase release 5.12 (October 2008),
available at http://fb2008_09.flybase.org/
5. M. Stapleton et al., Genome Biol. 3, RESEARCH0080
(2002).
6. K. H. Wan et al., Nat. Protoc. 1, 624 (2006).
7. R. Drysdale, FlyBase Consortium, Methods Mol. Biol. 420,
45 (2008).
8. G. Dietzl et al., Nature 448, 151 (2007).
9. S. Mohr, C. Bakal, N. Perrimon, Annu. Rev. Biochem. 79,
37 (2010).
10. H. J. Bellen et al., Genetics 167, 761 (2004).
11. Drosophila 12 Genomes Consortium, Nature 450,
203 (2007).
12. A. Stark et al., Nature 450, 219 (2007).
13. M. D. Adams et al., Science 287, 2185 (2000).
14. B. Graveley, Nature, 10.1038/nature09715.
15. L. Cherbas et al., Genome Res., 10.1101/gr.112961.110.
16. R. A. Hoskins et al., Genome Res., 10.1101/gr.112466.110.
17. Supplemental text and materials and methods are
available as supporting material on Science Online.
18. P. V. Kharchenko et al., Nature, 10.1038/nature09725.
19. TF binding, hotspots, TF motif instances, promoter and
enhancer validations, 12-point expression, and chromatin
time course are available at www.cistrack.org.
VOL 330 SCIENCE www.sciencemag.org
20. M. F. Lin et al., Genome Res. 17, 1823 (2007).
21. S. Washietl, I. L. Hofacker, P. F. Stadler, Proc. Natl. Acad.
Sci. U.S.A. 102, 2454 (2005).
22. V. N. Kim, J. Han, M. C. Siomi, Nat. Rev. Mol. Cell Biol.
10, 126 (2009).
23. E. Berezikov et al., Genome Res., 10.1101/gr.116657.110.
24. W.-J. Chung et al., Genome Res., 10.1101/gr.113050.110.
25. A. S. Flynt, J. C. Greimann, W. J. Chung, C. D. Lima,
E. C. Lai, Mol. Cell 38, 900 (2010).
26. W. J. Chung, K. Okamura, R. Martin, E. C. Lai, Curr. Biol.
18, 795 (2008).
27. Y. Kawamura et al., Nature 453, 793 (2008).
28. B. Czech et al., Nature 453, 798 (2008).
29. M. Ghildiyal et al., Science 320, 1077 (2008); 10.1126/
science.1157396.
30. K. Okamura, J. W. Hagen, H. Duan, D. M. Tyler, E. C. Lai,
Cell 130, 89 (2007).
31. K. Okamura et al., Nature 453, 803 (2008).
32. J. Brennecke et al., Cell 128, 1089 (2007).
33. N. C. Lau et al., Genome Res. 19, 1776 (2009).
34. N. Robine et al., Curr. Biol. 19, 2066 (2009).
35. K. Saito et al., Nature 461, 1296 (2009).
Downloaded from www.sciencemag.org on December 23, 2010
36. J. C. Eissenberg, G. Reuter, Int Rev. Cell. Mol. Biol. 273,
1 (2009).
37. N. C. Riddle et al., Genome Res., 10.1101/gr.110098.110
38. M. L. Eaton et al., Genome Res., 10.1101/gr.116038.110.
39. H. K. MacAlpine, R. Gordân, S. K. Powell, A. J. Hartemink,
D. M. MacAlpine, Genome Res. 20, 201 (2010).
40. J. Nordman, S. Li, T. Eng, D. MacAlpine, T. L. Orr-Weaver,
Genome Res., 10.1101/gr.114611.110
41. G. C. Hon, R. D. Hawkins, B. Ren, Hum. Mol. Genet. 18,
R195 (2009).
42. T. Kouzarides, Cell 128, 693 (2007).
43. R. B. Deal, J. G. Henikoff, S. Henikoff, Science 328,
1161 (2010).
44. S. Henikoff, J. G. Henikoff, A. Sakai, G. B. Loeb,
K. Ahmad, Genome Res. 19, 460 (2008).
45. N. D. Heintzman et al., Nat. Genet. 39, 311 (2007).
46. S. Vashee et al., Genes Dev. 17, 1894 (2003).
47. D. Remus et al., Cell 139, 719 (2009).
48. M. L. Eaton, K. Galani, S. Kang, S. P. Bell,
D. M. MacAlpine, Genes Dev. 24, 748 (2010).
49. N. M. Berbenetz, C. Nislow, G. W. Brown, PLoS Genet. 6,
e1001092 (2010).
50. P. J. Gillespie, T. Hirano, Curr. Biol. 14, 1598 (2004).
51. T. S. Takahashi, P. Yiu, M. F. Chou, S. Gygi, J. C. Walter,
Nat. Cell Biol. 6, 991 (2004).
52. K. S. Scott, P. K. Geyer, EMBO J. 14, 6258 (1995).
53. N. Nègre et al., PLoS Genet. 6, e1000814 (2010).
54. C. Jin et al., Nat. Genet. 41, 941 (2009).
55. J. Ernst, M. Kellis, Nat. Biotechnol. 28, 817 (2010).
56. G. J. Filion et al., Cell 143, 212 (2010).
57. S. MacArthur et al., Genome Biol. 10, R80 (2009).
58. A. Carr, M. D. Biggin, EMBO J. 18, 1598 (1999).
59. C. Moorman et al., Proc. Natl. Acad. Sci. U.S.A. 103,
12027 (2006).
60. X. Y. Li et al., PLoS Biol. 6, e27 (2008).
61. R. P. Zinzen, C. Girardot, J. Gagneur, M. Braun,
E. E. Furlong, Nature 462, 65 (2009).
62. A. Siepel et al., Genome Res. 15, 1034 (2005).
63. S. Meader, C. P. Ponting, G. Lunter, Genome Res. 20,
1335 (2010).
64. H. Yu, M. Gerstein, Proc. Natl. Acad. Sci. U.S.A. 103,
14724 (2006).
65. U. Alon, Nat. Rev. Genet. 8, 450 (2007).
66. S. Wernicke, F. Rasche, Bioinformatics 22, 1152 (2006).
67. J. Ernst, O. Vainas, C. T. Harbison, I. Simon,
Z. Bar-Joseph, Mol. Syst. Biol. 3, 74 (2007).
68. E. E. M. Furlong, E. C. Andersen, B. Null, K. P. White, M. P. Scott,
Science 293, 1629 (2001); 10.1126/science.1062660.
69. L. A. Lee, T. L. Orr-Weaver, Annu. Rev. Genet. 37,
545 (2003).
70. D. A. Harrison, D. A. Gdula, R. S. Coyne, V. G. Corces,
Genes Dev. 7, 1966 (1993).
71. E. H. Davidson et al., Science 295, 1669 (2002).
72. A. C. Spradling, Genetics 174, 1 (2006).
73. This work was supported by the National Human Genome
Research Institute as part of the modENCODE project
under RC2HG005639 (M.K.), U01HG004271 (S.E.C.),
U01HG004258 (G.H.K.), U01HG004264 (K.P.W.),
 U01HG004279 (D.M.M.), U01HG004261 (E.L.),
U01HG004274 (S.H.), and U41HG004269 (L.S.). Awards
to S.E.C. and G.H.K. were carried out at LBNL under contract
no. DE-AC02-05CH11231. Additional support was provided
by the NSF under grant 0937060 to the Computing Research
Association for the CIFellows Project (S.R.) and under award
no. 0905968 (J.E.), a Natural Sciences and Engineering
Research Council of Canada (NSERC) fellowship (B.A.),
T. Kahveci (F.A.), the Japan Society for the Promotion of
Science (K.O.), the Swedish Research Council (Q.D.), a NIH
National Research Service Award postdoctoral fellowship
(C.A.B.), a National Defense Science and Engineering
Graduate Fellowship (R.S.), an Erwin Schrödinger Fellowship
of the Austrian Fonds zur Förderung der wissenschaftlichen
Forschung (S.W.), a Leukemia and Lymphoma Society
fellowship (S.W.), a Lilly-Life Sciences Research Foundation
fellowship (C.D.B.), a NSERC postdoctoral fellowship (C.G.A.),
Affymetrix (T.G.R.), a fellowship from the Swiss National
Science Foundation (D.M.), a German Research Foundation
grant WI 3628/1-1 (S.W.), a HHMI Damon Runyon Cancer
Research fellowship (J.T.N.), the Indiana Genomics Initiative
(T.C.K.), H. Smith and the NIDDK genomics core laboratory
Benjamin Booth, Angela N. Brooks, Carrie A. Davis, Michael O. Duff,
Philipp Kapranov, Anastasia A. Samsonova, Jeremy E. Sandler,
Marijke J. van Baren, Kenneth H. Wan, Li Yang, Charles Yu,
Justen Andrews, Steven E. Brenner, Michael R. Brent, Lucy Cherbas,
Thomas R. Gingeras, Roger A. Hoskins, Thomas C. Kaufman,
Norbert Perrimon, Peter Cherbas, Brenton R. Graveley, Susan E.
Celniker, Charles L. G. Comstock, Alex Dobin, Jorg Drenkow,
Sandrine Dudoit, Jacqueline Dumais, Delphine Fagegaltier,
Srinka Ghosh, Kasper D. Hansen, Sonali Jha, Laura Langton,
Wei Lin, David Miller, Aaron E. Tenney, Huaien Wang, Aarron
T. Willingham, Chris Zaleski, Dayu Zhang
Karpen (chromatin): Peter V. Kharchenko, Michael Y. Tolstorukov,
Artyom A. Alekseyenko, Andrey A. Gorchakov, Tingting Gu,
Aki Minoda, Nicole C. Riddle, Yuri B. Schwartz, Sarah C. R. Elgin,
Mitzi I. Kuroda, Vincenzo Pirrotta, Peter J. Park, Gary H. Karpen,
David Acevedo, Eric P. Bishop, Sarah E. Gadel, Youngsook L.
Jung, Cameron D. Kennedy, Ok-Kyung Lee, Daniela Linder-Basso,
Sarah E. Marchetti, Gregory Shanower
White (transcription factors): Nicolas Nègre, Lijia Ma, Christopher
D. Brown, Rebecca Spokony, Robert L. Grossman, James W.
Posakony, Bing Ren, Steven Russell, Kevin P. White, Richard
Auburn, Hugo J. Bellen, Jia Chen, Marc H. Domanus, David
Lai (small RNAs): Nicolas Robine, Eugene Berezikov, Qi Dai,
Katsutomo Okamura, Eric C. Lai, Qi Dai, Gregory J. Hannon,
Martin Hirst, Marco Marra, Michelle Rooks, Yongjun Zhao
Henikoff (nucleosomes): Jorja G. Henikoff, Akiko Sakai, Kami
Ahmad, Steven Henikoff, Terri D. Bryson
Stein (data coordination center): Bradley I. Arshinoff, Nicole
L. Washington, Adrian Carr, Xin Feng, Marc D. Perry, William
J. Kent, Suzanna E. Lewis, Gos Micklem, Lincoln D. Stein, Galt
Barber, Aurelien Chateigner, Hiram Clawson, Sergio Contrino,
Francois Guillier, Angie S. Hinrichs, Ellen T. Kephart, Paul
Lloyd, Rachel Lyne, Sheldon McKay, Richard A. Moore, Chris
Mungall, Kim M. Rutherford, Peter Ruzanov, Richard Smith,
E. O. Stinson, Zheng Zha
Oliver (comparative transcription): Carlo G. Artieri, Renhua
Li, John H. Malone, David Sturgill, Brian Oliver, Lichun Jiang,
Nicolas Mattiuzzo
RNA structure: Sebastian Will, Bonnie Berger
Program management: Elise A. Feingold, Peter J. Good, Mark
S. Guyer, Rebecca F. Lowdon
Supporting Online Material
www.sciencemag.org/cgi/content/full/science.1198374/DC1

Downloaded from www.sciencemag.org on December 23, 2010

(B.O.), NIH R01HG004037, NSF CAREER award 0644282,
and the Sloan Foundation (M.K.). A full list of author
contributions is available in the SOM.
Complete Author List
Kellis (integration): Sushmita Roy, Jason Ernst, Pouya Kheradpour,
Christopher A. Bristow, Michael F. Lin, Stefan Washietl, Ferhat Ay,
Patrick E. Meyer, Luisa Di Stefano, Rogerio Candeias, Irwin
Jungreis, Daniel Marbach, Rachel Sealfon, Manolis Kellis
Celniker (transcription): Jane M. Landolin, Joseph W. Carlson,
Hanley, Elizabeth Heinz, Zirong Li, Folker Meyer, Steven W. Materials and Methods
Miller, Carolyn A. Morrison, Douglas A. Scheftner, Lionel SOM Text
Senderowicz, Parantu K. Shah, Sarah Suchy, Feng Tian, Koen Figs. S1 to S23
J. T. Venken, Robert White, Jared Wilkening, Jennifer Zieba Tables S1 to S7
MacAlpine (replication): Matthew L. Eaton, Heather K. Data Sets S1 to S17 (available at www.modencode.org/
MacAlpine, Jared T. Nordman, Sara K. Powell, Noa Sher, publications/integrative_fly_2010/)
Terry L. Orr-Weaver, David M. MacAlpine, Leyna C. DeNapoli, 28 September 2010; accepted 30 November 2010
Queying Ding, Thomas Eng, Helena Kashevsky, Sharon Li, Published online 22 December 2010;
Joseph A. Prinz 10.1126/science.1198374

REPORTS
mochemical cycles using metal oxide redox
High-Flux Solar-Driven Thermochemical reactions further bypass the CO-O2 or H2-O2
separation problem (11). Among candidate redox
Dissociation of CO2 and H2O Using materials, ferrite-based oxides exhibit relatively
slow reaction rates, degradation in rates because
Nonstoichiometric Ceria of sintering, and losses because of uncontrolled
volatilization, whereas ZnO, SnO2, and analo-
gous volatile oxides that sublime during de-
William C. Chueh,1 Christoph Falter,2 Mandy Abbott,1 Danien Scipio,1 Philipp Furler,2 composition require rapid quenching of gaseous
Sossina M. Haile,1* Aldo Steinfeld2,3* products to avoid recombination (10–18). Ceri-
um oxide (ceria) has emerged as a highly
Because solar energy is available in large excess relative to current rates of energy consumption, attractive redox active material choice for two-
effective conversion of this renewable yet intermittent resource into a transportable and step thermochemical cycling because it displays
dispatchable chemical fuel may ensure the goal of a sustainable energy future. However, low rapid fuel production kinetics and high selectiv-
conversion efficiencies, particularly with CO2 reduction, as well as utilization of precious ity (17, 19–24), where such features result, in
materials have limited the practical generation of solar fuels. By using a solar cavity-receiver part, from the absence of distinct oxidized and
reactor, we combined the oxygen uptake and release capacity of cerium oxide and facile catalysis reduced phases. However, ceria-based thermo-
at elevated temperatures to thermochemically dissociate CO2 and H2O, yielding CO and H2, chemical studies to date have largely been limited
respectively. Stable and rapid generation of fuel was demonstrated over 500 cycles. Solar-to-fuel to bench-top demonstrations of components or
efficiencies of 0.7 to 0.8% were achieved and shown to be largely limited by the system scale individual steps of the solar fuel production cy-
and design rather than by chemistry. cle; assessment of cyclability has been limited,
and the energy conversion efficiency has re-
mained uncertain because of the relatively low
ong-term storage and long-range transport Chemical fuels, derived from CO2 and/or H2O, gravimetric fuel productivity inherent to the

L of the vast, yet intermittent and unevenly

distributed, solar energy resource is essen-
tial for a transition away from fossil energy (1).
1
Materials Science, California Institute of Technology, MC 309-81,
Pasadena, CA 91125, USA. 2Department of Mechanical and
Process Engineering, Eidgenössische Technische Hochschule (ETH)
Zürich, 8092 Zürich, Switzerland. 3Solar Technology Laboratory,
Paul Scherrer Institute, 5232 Villigen PSI, Switzerland.
*To whom correspondence should be addressed. E-mail:
smhaile@caltech.edu (S.M.H.); aldo.steinfeld@ethz.ch (A.S.)
offer exceptional energy density and convenience
for transportation, but their production using solar
energy input has remained a grand challenge (2–9).
Solar-driven thermochemical approaches to CO2
and H2O dissociation inherently operate at high
temperatures and use the entire solar spectrum; as
such, they provide an attractive path to solar fuel
production at high rates and efficiencies in the ab-
sence of precious metal catalysts (10). In contrast to
direct thermolysis of CO2 and H2O, two-step ther-
nonstoichiometric process. Here, we demonstrate
high-rate solar fuel production from both CO2
and H2O using a solar reactor subjected directly
to concentrated radiation under realistic operating
conditions relevant to large-scale industrial im-
plementation, without the need for complex ma-
terial microstructures and/or system design (e.g.,
additional quench or separation steps). The re-
sults provide compelling evidence for the viabil-
ity of thermochemical approaches to solar fuel

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1797

 U01HG004279 (D.M.M.), U01HG004261 (E.L.),
U01HG004274 (S.H.), and U41HG004269 (L.S.). Awards
to S.E.C. and G.H.K. were carried out at LBNL under contract
no. DE-AC02-05CH11231. Additional support was provided
by the NSF under grant 0937060 to the Computing Research
Association for the CIFellows Project (S.R.) and under award
no. 0905968 (J.E.), a Natural Sciences and Engineering
Research Council of Canada (NSERC) fellowship (B.A.),
T. Kahveci (F.A.), the Japan Society for the Promotion of
Science (K.O.), the Swedish Research Council (Q.D.), a NIH
National Research Service Award postdoctoral fellowship
(C.A.B.), a National Defense Science and Engineering
Graduate Fellowship (R.S.), an Erwin Schrödinger Fellowship
of the Austrian Fonds zur Förderung der wissenschaftlichen
Forschung (S.W.), a Leukemia and Lymphoma Society
fellowship (S.W.), a Lilly-Life Sciences Research Foundation
fellowship (C.D.B.), a NSERC postdoctoral fellowship (C.G.A.),
Affymetrix (T.G.R.), a fellowship from the Swiss National
Science Foundation (D.M.), a German Research Foundation
grant WI 3628/1-1 (S.W.), a HHMI Damon Runyon Cancer
Research fellowship (J.T.N.), the Indiana Genomics Initiative
(T.C.K.), H. Smith and the NIDDK genomics core laboratory
Benjamin Booth, Angela N. Brooks, Carrie A. Davis, Michael O. Duff,
Philipp Kapranov, Anastasia A. Samsonova, Jeremy E. Sandler,
Marijke J. van Baren, Kenneth H. Wan, Li Yang, Charles Yu,
Justen Andrews, Steven E. Brenner, Michael R. Brent, Lucy Cherbas,
Thomas R. Gingeras, Roger A. Hoskins, Thomas C. Kaufman,
Norbert Perrimon, Peter Cherbas, Brenton R. Graveley, Susan E.
Celniker, Charles L. G. Comstock, Alex Dobin, Jorg Drenkow,
Sandrine Dudoit, Jacqueline Dumais, Delphine Fagegaltier,
Srinka Ghosh, Kasper D. Hansen, Sonali Jha, Laura Langton,
Wei Lin, David Miller, Aaron E. Tenney, Huaien Wang, Aarron
T. Willingham, Chris Zaleski, Dayu Zhang
Karpen (chromatin): Peter V. Kharchenko, Michael Y. Tolstorukov,
Artyom A. Alekseyenko, Andrey A. Gorchakov, Tingting Gu,
Aki Minoda, Nicole C. Riddle, Yuri B. Schwartz, Sarah C. R. Elgin,
Mitzi I. Kuroda, Vincenzo Pirrotta, Peter J. Park, Gary H. Karpen,
David Acevedo, Eric P. Bishop, Sarah E. Gadel, Youngsook L.
Jung, Cameron D. Kennedy, Ok-Kyung Lee, Daniela Linder-Basso,
Sarah E. Marchetti, Gregory Shanower
White (transcription factors): Nicolas Nègre, Lijia Ma, Christopher
D. Brown, Rebecca Spokony, Robert L. Grossman, James W.
Posakony, Bing Ren, Steven Russell, Kevin P. White, Richard
Auburn, Hugo J. Bellen, Jia Chen, Marc H. Domanus, David
Lai (small RNAs): Nicolas Robine, Eugene Berezikov, Qi Dai,
Katsutomo Okamura, Eric C. Lai, Qi Dai, Gregory J. Hannon,
Martin Hirst, Marco Marra, Michelle Rooks, Yongjun Zhao
Henikoff (nucleosomes): Jorja G. Henikoff, Akiko Sakai, Kami
Ahmad, Steven Henikoff, Terri D. Bryson
Stein (data coordination center): Bradley I. Arshinoff, Nicole
L. Washington, Adrian Carr, Xin Feng, Marc D. Perry, William
J. Kent, Suzanna E. Lewis, Gos Micklem, Lincoln D. Stein, Galt
Barber, Aurelien Chateigner, Hiram Clawson, Sergio Contrino,
Francois Guillier, Angie S. Hinrichs, Ellen T. Kephart, Paul
Lloyd, Rachel Lyne, Sheldon McKay, Richard A. Moore, Chris
Mungall, Kim M. Rutherford, Peter Ruzanov, Richard Smith,
E. O. Stinson, Zheng Zha
Oliver (comparative transcription): Carlo G. Artieri, Renhua
Li, John H. Malone, David Sturgill, Brian Oliver, Lichun Jiang,
Nicolas Mattiuzzo
RNA structure: Sebastian Will, Bonnie Berger
Program management: Elise A. Feingold, Peter J. Good, Mark
S. Guyer, Rebecca F. Lowdon
Supporting Online Material
www.sciencemag.org/cgi/content/full/science.1198374/DC1

Downloaded from www.sciencemag.org on December 23, 2010

(B.O.), NIH R01HG004037, NSF CAREER award 0644282,
and the Sloan Foundation (M.K.). A full list of author
contributions is available in the SOM.
Complete Author List
Kellis (integration): Sushmita Roy, Jason Ernst, Pouya Kheradpour,
Christopher A. Bristow, Michael F. Lin, Stefan Washietl, Ferhat Ay,
Patrick E. Meyer, Luisa Di Stefano, Rogerio Candeias, Irwin
Jungreis, Daniel Marbach, Rachel Sealfon, Manolis Kellis
Celniker (transcription): Jane M. Landolin, Joseph W. Carlson,
Hanley, Elizabeth Heinz, Zirong Li, Folker Meyer, Steven W. Materials and Methods
Miller, Carolyn A. Morrison, Douglas A. Scheftner, Lionel SOM Text
Senderowicz, Parantu K. Shah, Sarah Suchy, Feng Tian, Koen Figs. S1 to S23
J. T. Venken, Robert White, Jared Wilkening, Jennifer Zieba Tables S1 to S7
MacAlpine (replication): Matthew L. Eaton, Heather K. Data Sets S1 to S17 (available at www.modencode.org/
MacAlpine, Jared T. Nordman, Sara K. Powell, Noa Sher, publications/integrative_fly_2010/)
Terry L. Orr-Weaver, David M. MacAlpine, Leyna C. DeNapoli, 28 September 2010; accepted 30 November 2010
Queying Ding, Thomas Eng, Helena Kashevsky, Sharon Li, Published online 22 December 2010;
Joseph A. Prinz 10.1126/science.1198374

REPORTS
mochemical cycles using metal oxide redox
High-Flux Solar-Driven Thermochemical reactions further bypass the CO-O2 or H2-O2
separation problem (11). Among candidate redox
Dissociation of CO2 and H2O Using materials, ferrite-based oxides exhibit relatively
slow reaction rates, degradation in rates because
Nonstoichiometric Ceria of sintering, and losses because of uncontrolled
volatilization, whereas ZnO, SnO2, and analo-
gous volatile oxides that sublime during de-
William C. Chueh,1 Christoph Falter,2 Mandy Abbott,1 Danien Scipio,1 Philipp Furler,2 composition require rapid quenching of gaseous
Sossina M. Haile,1* Aldo Steinfeld2,3* products to avoid recombination (10–18). Ceri-
um oxide (ceria) has emerged as a highly
Because solar energy is available in large excess relative to current rates of energy consumption, attractive redox active material choice for two-
effective conversion of this renewable yet intermittent resource into a transportable and step thermochemical cycling because it displays
dispatchable chemical fuel may ensure the goal of a sustainable energy future. However, low rapid fuel production kinetics and high selectiv-
conversion efficiencies, particularly with CO2 reduction, as well as utilization of precious ity (17, 19–24), where such features result, in
materials have limited the practical generation of solar fuels. By using a solar cavity-receiver part, from the absence of distinct oxidized and
reactor, we combined the oxygen uptake and release capacity of cerium oxide and facile catalysis reduced phases. However, ceria-based thermo-
at elevated temperatures to thermochemically dissociate CO2 and H2O, yielding CO and H2, chemical studies to date have largely been limited
respectively. Stable and rapid generation of fuel was demonstrated over 500 cycles. Solar-to-fuel to bench-top demonstrations of components or
efficiencies of 0.7 to 0.8% were achieved and shown to be largely limited by the system scale individual steps of the solar fuel production cy-
and design rather than by chemistry. cle; assessment of cyclability has been limited,
and the energy conversion efficiency has re-
mained uncertain because of the relatively low
ong-term storage and long-range transport Chemical fuels, derived from CO2 and/or H2O, gravimetric fuel productivity inherent to the

L of the vast, yet intermittent and unevenly

distributed, solar energy resource is essen-
tial for a transition away from fossil energy (1).
1
Materials Science, California Institute of Technology, MC 309-81,
Pasadena, CA 91125, USA. 2Department of Mechanical and
Process Engineering, Eidgenössische Technische Hochschule (ETH)
Zürich, 8092 Zürich, Switzerland. 3Solar Technology Laboratory,
Paul Scherrer Institute, 5232 Villigen PSI, Switzerland.
*To whom correspondence should be addressed. E-mail:
smhaile@caltech.edu (S.M.H.); aldo.steinfeld@ethz.ch (A.S.)
offer exceptional energy density and convenience
for transportation, but their production using solar
energy input has remained a grand challenge (2–9).
Solar-driven thermochemical approaches to CO2
and H2O dissociation inherently operate at high
temperatures and use the entire solar spectrum; as
such, they provide an attractive path to solar fuel
production at high rates and efficiencies in the ab-
sence of precious metal catalysts (10). In contrast to
direct thermolysis of CO2 and H2O, two-step ther-
nonstoichiometric process. Here, we demonstrate
high-rate solar fuel production from both CO2
and H2O using a solar reactor subjected directly
to concentrated radiation under realistic operating
conditions relevant to large-scale industrial im-
plementation, without the need for complex ma-
terial microstructures and/or system design (e.g.,
additional quench or separation steps). The re-
sults provide compelling evidence for the viabil-
ity of thermochemical approaches to solar fuel

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1797

 REPORTS
generation while clarifying the efforts required to Thermochemical H2O-CO2-splitting cycles Higher temperature, TH
transform the concept into a central technology in over a nonstoichiometric oxide are described by  
d MO2 → d MO2
d þ =2 O2 ðgÞ ð1Þ
1 1 1
a sustainable energy future. the following pairs of reactions:
Lower temperature, TL
Concentrated Solar Radiation 
H2 OðgÞ þ 1 d MO2
d → 1 =d MO2 þ H2 ðgÞ
ð2aÞ
Lower temperature, TL

CO2 ðgÞ þ 1 d MO2
d → 1 =d MO2 þ COðgÞ
ð2bÞ

Quartz Window Net H2O dissociation

H2 OðgÞ → 1 =2 O2 ðgÞ þ H2 ðgÞ ð3aÞ
Net CO2 dissociation
CO2 ðgÞ → 1 =2 O2 ðgÞ þ COðgÞ ð3bÞ
CPC

Downloaded from www.sciencemag.org on December 23, 2010

s
Purge Ga

Inlet H2O, CO2 where M in the present case is Ce or a com-

bination of Ce and a dopant element. At the rel-
atively high temperatures of the present study
Inlet Alumina Insulation (>800°C), equilibria of reaction (2b) involving
carbonaceous species can be neglected (21).
Porous Ceria The solar reactor constructed for the purposes
Outlet

of demonstrating these cycles is schematically

shown in Fig. 1. It consists of a cavity receiver
with a windowed aperture through which con-
centrated solar radiation enters. The selected di-
Oxygen Evolution Half-Cycle mensions ensure multiple internal reflections and
Fuel Production Half-Cycle efficient capture of incoming solar energy; the
50 µm apparent absorptivity exceeds 0.94, approaching
Purge Gas, O2 H2, CO the ideal blackbody limit (25). Porous, monolithic
Fig. 1. Schematic of the solar reactor for the two-step, solar-driven thermochemical production of ceria, assembled from quarter-circular-arc pieces
fuels. It consists of a thermally insulated cavity receiver containing a porous monolithic ceria in the form of a cylinder (325 g in mass, 35 mm
cylinder. Concentrated solar radiation enters through a windowed aperture and impinges on the ceria in outer diameter, 102 mm in height, and 80% in
inner walls. Reacting gases flow radially across the porous ceria toward the cavity inside, whereas porosity as fabricated), is placed inside the cavity
product gases exit the cavity through an axial outlet port at the bottom. (Inset) The scanning electron and subjected to multiple heat-cool cycles under
micrograph of the porous ceria tube after 23 cycles. Blue arrows indicate ceria reduction (Eq. 1); red appropriate gases to induce fuel production (Fig.
arrows indicate oxidation (Eqs. 2a and 2b). 2). No additional support or absorber was used.

40
A 2000 40
B 1000

800
30 1500 30
600
20 1000 20
400
10 500 10
200

0 0 0 0
0 100 200 300 400 0 100 200 300

1500 1500

1000 1000

500 500
0 100 200 300 400 0 100 200 300

Fig. 2. Thermochemical cycling of ceria (325 g) using the solar reactor with
(A) CO2 and (B) H2O as oxidant. The oxygen and fuel evolution rates as well
as the total volume of gas evolved are shown. Temperatures were measured
at three positions along the height of the ceria tube. Maximum temperatures
(Tmax) attained in the reactor are shown. Conditions for (A) were as follows:
Ar sweep gas at a flow rate of 0.0062 liter min−1 g−1 of ceria during the
ceria reduction half-cycle, and CO2/Ar at pCO2 = 0.78 atm and a flow rate
of 0.035 liter min−1 g−1 of ceria during the ceria oxidation half-cycle.
Conditions for (B) were as follows: Ar sweep gas at a flow rate of 0.023 liter
min−1 g−1 of ceria during the ceria reduction half-cycle, and H2O/Ar at
pH2O = 0.78 atm and a flow rate of 0.035 liter min−1 g−1 of ceria during
the ceria oxidation half-cycle.

1798 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


With this arrangement, the porous ceria cylinder
is directly exposed to concentrated solar radiation
impinging on its inner walls. An annular gap be-
tween the ceria cylinder and the alumina in-
sulation tiles suppresses undesired reactions
between the two components. Reacting gases
are injected into this annular gap and directed to
flow radially across the porous ceria cylinder
toward the cavity inside, whereas product gases
exit the cavity through an axial outlet port at the
bottom. Complete experimental details, includ-
ing procedures for estimating uncertainty, are
given in (25). Post situ x-ray diffraction showed
that, with the exception of the portion in direct
contact with the insulation material, the ceria
remained phase-pure and free of detectable
alumina incorporation (fig. S1).
To drive oxygen evolution (Eq. 1), we purged
the solar reactor with flowing argon [partial pressure
of O2 (pO2) = 10–5 atm] and ramped the incident
radiation power to about 1.9 kWat a mean solar flux
intensity over the aperture of 1500 suns (1 sun =
1 kW m−2), typical of a commercial solar dish or a
tower concentration system. The temperature of
the ceria tube rose to values between 1420° and
1640°C, with the exact temperature dependent on
the position within the reactor and on the cycle
(Fig. 2A). The rise in temperature was rapid below
1250°C, averaging 140°C min−1, but slowed to
about 8°C min−1 as the temperature approached a
steady-state value because of increasing heat dis-
sipation by re-radiation through the aperture and
conduction heat transfer through the insulation.
Oxygen evolution from ceria was observed at an
onset temperature of about 900°C, consistent with
equilibrium thermogravimetry measurements (fig.
S2). The rate of evolution increased with temper-
ature, reaching a peak value of 34 T 2 ml min−1
(STP, standard temperature and pressure) and an
average value of 16 T 1 ml min−1 (averaged over
Fig. 3. Comparison of (A) O2, (B) A B
H2, and (C) CO evolution between
12
the solar reactor (dotted lines) and
the differential reactor (solid lines).
Experimental conditions for the
solar reactor are the same as those 8 0.10
for Fig. 2. In the differential reactor,
0.429 g of CeO2 was cycled be-
tween 1500°C (Ar sweep gas flow 4 0.05
rate = 2.3 liter min−1 g−1 of ceria,
20 min) and 800°C (for CO2 split-
ting, pCO2 = 0.50 atm, flow rate =
1.2 liter min−1 g−1 of ceria, 10 min;
0 0.00
for H2O splitting, pH2O = 0.44 to 1500
0.52 atm, flow rate = 2.1 to 2.5
liter min−1 g−1 of ceria, 10 min).
1000
500
0 20
www.sciencemag.org
the time required to reach 90% of the extent of
reaction) on the first cycle (Fig. 2A). When the rate
dropped to 20% of the peak value, the evolution
reaction was terminated by decreasing the intensity
of the incident radiation flux. Upon cooling to
~900°C, CO2 was injected into the solar reactor.
Production of CO was immediately observed,
reaching a remarkable peak rate of 1.5 × 103 T
0.1 × 103 ml min−1 (STP) and an average rate of
5.9 × 102 T 0.4 × 102 ml min−1 (STP) (Fig. 2A).
Consistent with the fact that there was no water
present in the reactant stream, no gas-phase C1, C2,
or C3 hydrocarbons were detected by the gas
chromatograph. Carbon-neutral balance (<3% C
unaccounted for, well within error) was achieved
by summing the flow rates of CO2 in the reactant
stream and CO2 and CO in the product steam. As
verification, temperature-programmed oxidation
was performed on the ceria after the CO2 dis-
sociation reaction by flowing oxygen into the solar
reactor while ramping the temperature to 1000°C
(26). Both CO and CO2 levels were below the
detection limit, confirming that no appreciable
amount of carbonaceous species was deposited onto
ceria during CO2 dissociation and that a 100%
selectivity toward CO production was achieved.
Upon the termination of CO production, the
radiation flux was increased and the entire cycle
repeated. A slight decline over the four cycles in
the nominal reactor temperature attained during
the O2 release step (from 1624° to 1581°C) is
responsible for the observed slight decline in O2
release and CO yield. An analogous set of ex-
periments was also performed for H2O dissoci-
ation, with H2 production rate reaching a peak
value of 7.6 × 102 T 0.8 × 102 ml min−1 (STP) and
a maximum average value of 3.1 × 102 T 0.3 ×
102 ml min−1 (STP) (Fig. 2B). A total of 23 cy-
cles were performed. An experimental run per-
formed without the ceria confirmed the absence
O2 Evolution H2 Evolution
0.15 12
Differential
Reactor
8 8
Solar
Reactor
4 4
0
1500
1000
500
40 60 0 5
Time (min) Time (min)
SCIENCE VOL 330 24 DECEMBER 2010
REPORTS
of O2, CO, or H2 evolution under corresponding
reaction conditions.
The characteristics of solar-thermochemical
fuel production from ceria reveal several important
features of the cycling process. Although the
behavior is generally reproducible between cycles,
some run-to-run variations are evident. The oxygen
evolution reaches a peak value between 17 and 34
ml min−1, whereas the total amount evolved ranges
from 0.54 to 0.94 liter for 325 g of ceria. As noted,
these differences are correlated with the peak
reactor temperature obtained (Fig. 2), variations
in which are attributed to unsteady heat transfer.
Taking this temperature variability into account,
the total oxygen evolution is found to be general-
ly consistent with thermodynamic expectations
(fig. S2). Mass balance considerations further
dictate a 2-to-1 molar (and hence volumetric)
Downloaded from www.sciencemag.org on December 23, 2010
ratio of fuel produced to oxygen released for full
utilization of the ceria nonstoichiometry. For CO2
dissociation, the CO:O2 ratio ranged from 1.6 T
0.2:1 to 2.0 T 0.2:1, whereas for H2O dissocia-
tion, the H2:O2 ratio was 1.6 T 0.2:1. The slight
deviation from the ideal value is attributed to
small leaks in the system and to the accuracies of
the electronic mass flow controller and measured
gas composition.
Perhaps the most obvious feature of the cycling
behavior in Fig. 2 is the much faster rate of fuel
production than that of O2 release. It was observed
that lowering the purge gas flow rate during ceria
reduction by a factor of 4 had negligible impact
on the oxygen evolution rate (Fig. 2), indicating
that the convective transport of oxygen gas in the
reactor is not likely the rate-limiting step. Further-
more, the substantial difference in the O2 evolution
and CO2 dissociation rate suggests that gas-phase
transport through the pores of ceria is probably not
rate-limiting. We considered, as an alternative, that
the oxygen evolution kinetics in the solar reactor are
C CO Evolution
12
0
1500
1000
500
10 0 5 10
Time (min)
1799
 REPORTS
Fig. 4. O2 (black) and
H2 (red) evolution rates
for 500 water-splitting cy-
cles. CeO2 was cycled be-
H2 Rate (mL min g ceria)
tween 1500°C (pO2 =10−5
-1
atm, flow rate = 3.2 liter
-1
min−1 g−1 of ceria, 10 min,
ramp rate = 100°C min−1)
and 800°C (pH2O = 0.13
to 0.15 atm, flow rate =
0.75 to 0.76 liter min−1 g−1
of ceria, 10 min). The gas
evolution rate is calcu-
lated by averaging the in-
stantaneous rate over the
time required to reach
90% of the gas produced.
limited by the heating rate, a factor that does not
affect fuel production because this step occurs iso-
thermally. If the heating rate is slow relative to the
surface reaction and solid-state diffusion steps in-
volved in oxygen release, we can express the oxygen
evolution rate as dd=dt ≈ ðdd=dT ÞT ¼TðtÞ ðdT =dtÞ,
where d is ceria oxygen nonstoichiometry (Eq. 1),
T is temperature, and t is time. By using the spa-
tially averaged temperature profile in the first cycle
in Fig. 2A, we compute a maximum oxygen
evolution rate of 65 ml min−1, which, given the
approximate nature of the calculation, is compa-
rable to the observed rate.
To eliminate the effect of gas-phase mass and
heat transfer, we carried out thermochemical cycling
of ceria by using identically prepared monolithic
porous samples (annealed for 50 hours to simulate
cycling conditions) in a smaller-scale infrared
imaging furnace that uses only a 0.4-g sample.
Such an experimental setup permitted thermo-
chemical cycling under high-flow, differential re-
actor conditions, in which the sample temperature
could be changed rapidly (average ramp rate of
1700°C min−1) and the gas composition approached
uniformity. Under these ideal conditions in which
only surface chemical reactions and solid-state
oxygen diffusion in ceria limit the overall reaction
rate, oxygen evolution (Fig. 3) attained a peak
instantaneous rate ~80 times faster than in the solar
reactor and an associated average rate of 2.2 T
0.2 ml min−1 g−1 of ceria; CO2 and H2O dissocia-
tion reactions were about two and four times faster
(although the reactant partial pressures, temper-
atures, and initial extents of reduction differ slight-
ly between the differential and solar reactor), with
rates of 5.1 T 0.4 and 5.3 T 0.4 ml min−1 g−1 of ceria,
respectively. These rates support the proposition
that oxygen evolution kinetics in the solar reactor
are limited predominantly by the heating rate.
The solar-to-fuel energy conversion efficien-
cy is defined as
rfuel DHfuel
h¼
Psolar þ rinert Einert
where rfuel is the molar fuel production rate,
DHfuel is the higher heating value of the fuel,
1800
3 0.3
O2 Rate (mL min g ceria)
-1
-1
2 0.2
1 0.1
0 0.0
0 100 200 300 400 500
Cycle
Psolar is the incident solar radiation power, rinert is
the flow rate of the inert gas during oxygen evo-
lution, and Einert is the energy required to separate
the inert sweep gas from air (usually N2; Ar was
used in this work entirely for reasons of exper-
imental convenience). Based on the experimental
data, the peak instantaneous efficiencies for CO2
and H2O dissociations reached 0.8% and 0.7%,
respectively [for detailed calculation procedures,
see (25)]. No heat recuperation strategy was used.
Upon averaging the efficiencies over the time
required to produce 80% of the fuel, the effi-
ciencies become 0.4% (for both CO2 and H2O
dissociations). These experimentally measured effi-
ciencies reflect the cycle irreversibilities resulting
from intrinsic material properties as well as solar
reactor design and operation. An energy-balance
analysis (25) reveals that 50% of the energy loss
resulted from heat conduction through the reactor
wall and 41% resulted from reradiation through
the aperture. The former energy penalty can be
dramatically reduced by improving thermal in-
sulation and by scaling up to increase the volume-to-
area ratio. The latter can be minimized by
augmenting the solar flux such that the aperture
size can be reduced. Decreasing heat loss also has
the added benefit of increasing the temperature
ramp rate. As shown in the earlier comparison of
the oxygen evolution kinetics under slow and
rapid heating rates, the oxygen evolution kinetics
and conversion efficiency are closely coupled to
the rates at which the active ceria materials can be
heated and cooled.
Beyond efficiency, material stability is an es-
sential criterion for a viable thermochemical process.
With use of the differential reactor system, which
enables rapid access to multiple cycles, 500 cycles
of water dissociation were performed without
interruption. The results (Fig. 4), reported in part
in an earlier work (21), indicate that, after an initial
stabilization period of ~100 cycles, both the
oxygen and hydrogen evolution rates remain
ð4Þ essentially constant for a subsequent 400 cycles.
Scanning electron microscopy examination of
samples of porous ceria that underwent heat treat-
ment under similar conditions (fig. S3) revealed
24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org
that the decrease in reaction rate is accompanied by
an increase in particle size. The morphology
stabilized after 24 hours of heat treatment at
1500°C, much as the fuel production rate stabilized
after an initial period. That ceria can be cycled
between two oxidation states without substantial
loss of activity can be attributed to its sufficiently
large change in oxygen nonstoichiometry at mod-
erate homologous temperatures (<0:6  Tm ).
In sum, the feasibility of a solar-driven thermo-
chemical cycle for dissociating H2O and CO2
using nonstoichiometric ceria has been demon-
strated in terms of materials, reaction rates, cyclabil-
ity, reactor technology, and energy conversion
efficiency. Essential to this demonstration is a
simple and scalable reactor design using porous
ceria directly exposed to concentrated solar
radiation that enables high-temperature heat transfer
Downloaded from www.sciencemag.org on December 23, 2010
to the reaction sites, as required for performing both
steps of the cycle. The solar-to-fuel energy
conversion efficiency obtained in this work for
CO2 dissociation is about two orders of magnitude
greater than that observed with state-of-the-art
photocatalytic approaches (3, 9).The gravimetric
hydrogen production rate exceeds that of other
solar-driven thermochemical processes by more
than an order of magnitude (27, 28). Both the
efficiency and the cycling rates in the reactor were
limited largely by thermal losses, resulting from
conductive and radiative heat transfer. A thermody-
namic analysis of efficiency based solely on the
material properties of CeO2 indicates that values in
the range of 16 to 19% are attainable, even in the
absence of sensible heat recovery (21). Thus, with
reactor optimization and system integration sub-
stantial increases in both efficiency and fuel
production rates are anticipated. The material
stability, showing stable fuel production over 500
thermochemical cycles, is already suitable for
realistic applications. Furthermore, the abundance
of cerium, which is comparable to that of copper
(29), is such that the approach is applicable at scales
relevant to global energy consumption (21).
References and Notes
1. N. S. Lewis, D. G. Nocera, Proc. Natl. Acad. Sci. U.S.A.
103, 15729 (2006).
2. H. Takeda, K. Koike, H. Inoue, O. Ishitani, J. Am. Chem.
Soc. 130, 2023 (2008).
3. S. C. Roy, O. K. Varghese, M. Paulose, C. A. Grimes,
ACS Nano 4, 1259 (2010).
4. Y. Hori, K. Kikuchi, S. Suzuki, Chem. Lett. 14, 1695
(1985).
5. V. P. Indrakanti, J. D. Kubicki, H. H. Schobert,
Energy Environ. Sci. 2, 745 (2009).
6. T. Inoue, A. Fujishima, S. Konishi, K. Honda, Nature 277,
637 (1979).
7. O. Khaselev, J. A. Turner, Science 280, 425 (1998).
8. K. R. Thampi, J. Kiwi, M. Gratzel, Nature 327, 506 (1987).
9. O. K. Varghese, M. Paulose, T. J. Latempa, C. A. Grimes,
Nano Lett. 9, 731 (2009).
10. E. A. Fletcher, J. Sol. Energy Trans. Am. Soc. Mech. Eng.
123, 63 (2001).
11. A. Steinfeld, Sol. Energy 78, 603 (2005).
12. L. O. Schunk, A. Steinfeld, AIChE J. 55, 1497 (2009).
13. L. O. Schunk, W. Lipinski, A. Steinfeld, Chem. Eng. J.
150, 502 (2009).
14. T. Kodama, N. Gokon, Chem. Rev. 107, 4048 (2007).
15. M. Chambon, S. Abanades, G. Flamant, Chem. Eng. Sci.
65, 3671 (2010).

16. D. Gstoehl, A. Brambilla, L. O. Schunk, A. Steinfeld,
J. Mater. Sci. 43, 4729 (2008).
17. J. E. Miller et al., J. Mater. Sci. 43, 4714 (2008).
18. C. Perkins, A. W. Weimer, AIChE J. 55, 286 (2009).
19. P. Singh, M. S. Hegde, Chem. Mater. 22, 762 (2010).
20. W. C. Chueh, S. M. Haile, ChemSusChem 2, 735 (2009).
21. W. C. Chueh, S. M. Haile, Philos. Trans. R. Soc. London
Ser. A 368, 3269 (2010).
22. S. M. Haile, W. C. Chueh, U.S. Patent application
20,090,107,044 (2009).
23. S. Abanades et al., J. Mater. Sci. 45, 4163 (2010).
24. H. Kaneko et al., Energy 32, 656 (2007).
25. Materials and methods are detailed in supporting
material at Science Online.
26. C. Li, L. Minh, T. C. Brown, J. Catal. 178, 275 (1998).
27. M. Roeb et al., Int. J. Hydrogen Energy 34, 4537 (2009).
28. Fuel production rate is compared with other
thermochemical processes by normalizing the average
fuel production rate (as defined in the text) by the
mass of the material, including inactive support used
to prevent material sintering. The rate for oxygen
evolution is not included.
29. G. B. Haxel, J. B. Hedrick, G. J. Orris, “Rare earth
elements—Critical resources for high technology”
(U.S. Geological Survey Fact Sheet 087-02, Reston, VA,
2002).
30. This work was funded in part by NSF (CBET-0829114),
the Initiative for Renewable Energy and the Environment
(under subcontract from the University of Minnesota),
and the Swiss National Science Foundation (contract no.
200021-126512). Additional travel support was provided
by the International Materials Institutes program of NSF
REPORTS
under award no. DMR 08-43934. We thank the technical
staff of the Solar Technology Laboratory of the Paul
Scherrer Institute for supporting the experimental
activities at the High-Flux Solar Simulator. W.C.C.
designed the experiments, and C.F. designed the solar
reactor. Samples were prepared by M.A. and D.S. W.C.C,
C.F., P.F., and D.S. executed the experiments. S.M.H. and
A.S. supervised the project.
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1797/DC1
Materials and Methods
Figs. S1 to S3
15 September 2010; accepted 23 November 2010
10.1126/science.1197834

triguing quantum-relativistic physics (16–19) has

Spin Hall Effect Transistor been established before the first experimental
observations (20, 21), but the field is still striving

Downloaded from www.sciencemag.org on December 23, 2010

Jörg Wunderlich,1,2*† Byong-Guk Park,1* Andrew C. Irvine,3* Liviu P. Zârbo,2 Eva Rozkotová,4 to turn the phenomenon into a concrete device
Petr Nemec,4 Vít Novák,2 Jairo Sinova,5,2 Tomás Jungwirth2,6 functionality. We demonstrate the applicability of
the spin Hall effect in a new type of spin
The field of semiconductor spintronics explores spin-related quantum relativistic phenomena in transistor.
solid-state systems. Spin transistors and spin Hall effects have been two separate leading directions The active semiconductor channel in our de-
of research in this field. We have combined the two directions by realizing an all-semiconductor vices is a two-dimensional electron gas (2DEG)
spin Hall effect transistor. The device uses diffusive transport and operates without electrical in which the spin-orbit coupling induced spin
current in the active part of the transistor. We demonstrate a spin AND logic function in a precession is controlled by external gate electrodes
semiconductor channel with two gates. Our study shows the utility of the spin Hall effect in a and detection is provided by transverse spin Hall
microelectronic device geometry, realizes the spin transistor with electrical detection directly
1
along the gated semiconductor channel, and provides an experimental tool for exploring spin Hitachi Cambridge Laboratory, Cambridge CB3 0HE, UK.
2
Institute of Physics ASCR, v.v.i., Cukrovarnická 10, 162 53
Hall and spin precession phenomena in an electrically tunable semiconductor layer. Praha 6, Czech Republic. 3Microelectronics Research Centre,
Cavendish Laboratory, University of Cambridge, CB3 0HE, UK.
4
wo major themes in semiconductor spin- lished from the outset by Datta and Das (3). The

T
tronics research, the spin transistors and
the spin Hall effects, have followed dis-
tinct and independent scientific paths (1, 2). In
the transistor case, the target device concept of
a ferromagnetic spin injector and detector con-
nected by a semiconductor channel was estab-
ensuing research has focused on the fundamental
physical problems related to the resistance mis-
match between the transistor’s components and
to the spin manipulation in the semiconductor via
spin-orbit coupling effects (4–15). By contrast, in
the spin Hall effect case, much of the related in-
Faculty of Mathematics and Physics, Charles University in
Prague, Ke Karlovu 3, 121 16 Prague 2, Czech Republic. 5De-
partment of Physics, Texas A&M University, College Station,
TX 77843–4242, USA. 6School of Physics and Astronomy, Uni-
versity of Nottingham, Nottingham NG7 2RD, UK.
*These authors contributed equally to this work.
†To whom correspondence should be addressed. E-mail:
jw526@cam.ac.uk

Fig. 1. (A) Schematics of the measurement setup

with optically injected spin-polarized electrical x 10 150
A
current propagating through the Hall bar and 5
corresponding experimental Hall effect signals at RH2 RH1 100
VB 0
crosses H1 and H2. The Hall resistances, RH = IPH
VH/IPH, for the two opposite helicities of the inci- -5
RH1 σ +-
50
dent light are plotted as a function of the focused σ
-10 IPH σ+- σ +- 0

IPH [ nA ]
RH2
RH [ Ω ]

(∼1 mm) light spot position, i.e., of the position of σ σ

the injection point. Increasing x corresponds to
shifting the spot further away from the Hall de- x
20 150
tectors. (The focused laser beam is indicated by B
RH2 RH1 10
the yellow cylinder in the schematics.) The optical VB 100
current IPH is independent of the helicity of the IPH 0
incident light and varies only weakly with the 50
-10
light spot position. The applied bias voltage VB =
−15 V, the laser intensity is 1000 W/cm2, and the -20
0
laser wavelength is 870 nm. (B) Same as (A) for
measurement geometry in which electrical cur- C 1.0
rent is closed before the first detecting Hall cross
H1. (C) Schematics of the diffusive transport of
<Sz>

0.5
injected spin-polarized electrons and Monte-Carlo
simulations of the out-of-plane component of the 0.0
spin of injected electrons averaged over the 1-mm
bar cross section assuming Rashba field a = 5.5 -0.5
meV Å, Dresselhaus field b = −24 meV Å, and 0.0 0.5 1.0 1.5 2.0 2.5
different values of the mean free path l. x µm

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1801


16. D. Gstoehl, A. Brambilla, L. O. Schunk, A. Steinfeld,
J. Mater. Sci. 43, 4729 (2008).
17. J. E. Miller et al., J. Mater. Sci. 43, 4714 (2008).
18. C. Perkins, A. W. Weimer, AIChE J. 55, 286 (2009).
19. P. Singh, M. S. Hegde, Chem. Mater. 22, 762 (2010).
20. W. C. Chueh, S. M. Haile, ChemSusChem 2, 735 (2009).
21. W. C. Chueh, S. M. Haile, Philos. Trans. R. Soc. London
Ser. A 368, 3269 (2010).
22. S. M. Haile, W. C. Chueh, U.S. Patent application
20,090,107,044 (2009).
23. S. Abanades et al., J. Mater. Sci. 45, 4163 (2010).
24. H. Kaneko et al., Energy 32, 656 (2007).
25. Materials and methods are detailed in supporting
material at Science Online.
26. C. Li, L. Minh, T. C. Brown, J. Catal. 178, 275 (1998).
27. M. Roeb et al., Int. J. Hydrogen Energy 34, 4537 (2009).
28. Fuel production rate is compared with other
thermochemical processes by normalizing the average
fuel production rate (as defined in the text) by the
mass of the material, including inactive support used
to prevent material sintering. The rate for oxygen
evolution is not included.
29. G. B. Haxel, J. B. Hedrick, G. J. Orris, “Rare earth
elements—Critical resources for high technology”
(U.S. Geological Survey Fact Sheet 087-02, Reston, VA,
2002).
30. This work was funded in part by NSF (CBET-0829114),
the Initiative for Renewable Energy and the Environment
(under subcontract from the University of Minnesota),
and the Swiss National Science Foundation (contract no.
200021-126512). Additional travel support was provided
by the International Materials Institutes program of NSF
REPORTS
under award no. DMR 08-43934. We thank the technical
staff of the Solar Technology Laboratory of the Paul
Scherrer Institute for supporting the experimental
activities at the High-Flux Solar Simulator. W.C.C.
designed the experiments, and C.F. designed the solar
reactor. Samples were prepared by M.A. and D.S. W.C.C,
C.F., P.F., and D.S. executed the experiments. S.M.H. and
A.S. supervised the project.
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1797/DC1
Materials and Methods
Figs. S1 to S3
15 September 2010; accepted 23 November 2010
10.1126/science.1197834

triguing quantum-relativistic physics (16–19) has

Spin Hall Effect Transistor been established before the first experimental
observations (20, 21), but the field is still striving

Downloaded from www.sciencemag.org on December 23, 2010

Jörg Wunderlich,1,2*† Byong-Guk Park,1* Andrew C. Irvine,3* Liviu P. Zârbo,2 Eva Rozkotová,4 to turn the phenomenon into a concrete device
Petr Nemec,4 Vít Novák,2 Jairo Sinova,5,2 Tomás Jungwirth2,6 functionality. We demonstrate the applicability of
the spin Hall effect in a new type of spin
The field of semiconductor spintronics explores spin-related quantum relativistic phenomena in transistor.
solid-state systems. Spin transistors and spin Hall effects have been two separate leading directions The active semiconductor channel in our de-
of research in this field. We have combined the two directions by realizing an all-semiconductor vices is a two-dimensional electron gas (2DEG)
spin Hall effect transistor. The device uses diffusive transport and operates without electrical in which the spin-orbit coupling induced spin
current in the active part of the transistor. We demonstrate a spin AND logic function in a precession is controlled by external gate electrodes
semiconductor channel with two gates. Our study shows the utility of the spin Hall effect in a and detection is provided by transverse spin Hall
microelectronic device geometry, realizes the spin transistor with electrical detection directly
1
along the gated semiconductor channel, and provides an experimental tool for exploring spin Hitachi Cambridge Laboratory, Cambridge CB3 0HE, UK.
2
Institute of Physics ASCR, v.v.i., Cukrovarnická 10, 162 53
Hall and spin precession phenomena in an electrically tunable semiconductor layer. Praha 6, Czech Republic. 3Microelectronics Research Centre,
Cavendish Laboratory, University of Cambridge, CB3 0HE, UK.
4
wo major themes in semiconductor spin- lished from the outset by Datta and Das (3). The

T
tronics research, the spin transistors and
the spin Hall effects, have followed dis-
tinct and independent scientific paths (1, 2). In
the transistor case, the target device concept of
a ferromagnetic spin injector and detector con-
nected by a semiconductor channel was estab-
ensuing research has focused on the fundamental
physical problems related to the resistance mis-
match between the transistor’s components and
to the spin manipulation in the semiconductor via
spin-orbit coupling effects (4–15). By contrast, in
the spin Hall effect case, much of the related in-
Faculty of Mathematics and Physics, Charles University in
Prague, Ke Karlovu 3, 121 16 Prague 2, Czech Republic. 5De-
partment of Physics, Texas A&M University, College Station,
TX 77843–4242, USA. 6School of Physics and Astronomy, Uni-
versity of Nottingham, Nottingham NG7 2RD, UK.
*These authors contributed equally to this work.
†To whom correspondence should be addressed. E-mail:
jw526@cam.ac.uk

Fig. 1. (A) Schematics of the measurement setup

with optically injected spin-polarized electrical x 10 150
A
current propagating through the Hall bar and 5
corresponding experimental Hall effect signals at RH2 RH1 100
VB 0
crosses H1 and H2. The Hall resistances, RH = IPH
VH/IPH, for the two opposite helicities of the inci- -5
RH1 σ +-
50
dent light are plotted as a function of the focused σ
-10 IPH σ+- σ +- 0

IPH [ nA ]
RH2
RH [ Ω ]

(∼1 mm) light spot position, i.e., of the position of σ σ

the injection point. Increasing x corresponds to
shifting the spot further away from the Hall de- x
20 150
tectors. (The focused laser beam is indicated by B
RH2 RH1 10
the yellow cylinder in the schematics.) The optical VB 100
current IPH is independent of the helicity of the IPH 0
incident light and varies only weakly with the 50
-10
light spot position. The applied bias voltage VB =
−15 V, the laser intensity is 1000 W/cm2, and the -20
0
laser wavelength is 870 nm. (B) Same as (A) for
measurement geometry in which electrical cur- C 1.0
rent is closed before the first detecting Hall cross
H1. (C) Schematics of the diffusive transport of
<Sz>

0.5
injected spin-polarized electrons and Monte-Carlo
simulations of the out-of-plane component of the 0.0
spin of injected electrons averaged over the 1-mm
bar cross section assuming Rashba field a = 5.5 -0.5
meV Å, Dresselhaus field b = −24 meV Å, and 0.0 0.5 1.0 1.5 2.0 2.5
different values of the mean free path l. x µm

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1801

 REPORTS
effect voltages (22) measured along the 2DEG
Hall bar. For spin injection, we use an optical
method described in (22) that permits all three
components of the spin transistor to be realized
within an all-semiconductor structure. The op-
tical injection method is less scalable than elec-
trical injection from ferromagnetic contacts, yet
it does not require any magnetic elements or ex-
ternal magnetic fields for the operation of the
device. Because of the nondestructive nature of
the spin Hall effect detection, one semiconductor
channel can accommodate multiple gates and
Hall cross detectors and is therefore directly suit-
able for realizing spin logic operations.
Semiconductor heterostructures used in our
experiments, described in detail in (23), com-
prise a modulation p-doped AlGaAs/GaAs hetero-
junction on top of the structure, 90 nm of intrinsic
nel, resulting in a spatially varying magnitude
and sign of the Hall signals on several succes-
sive Hall crosses. Because of the limited number
of discrete detection points, these experiments did
not provide a detailed picture of the spin pre-
cession of injected electrons. To better visualize
the effect, we use here the optical activity of the
device presented in Fig. 1, which extends over a
several-micrometer range from the lateral p-n
junction into the unetched p-type side of the epi-
layer. By shifting the focused laser spot, we can
smoothly change the position of the spin injec-
tion point with respect to the detection Hall crosses.
This results in damped oscillatory Hall resistance,
RH = VH/IPH, measured at each of the two suc-
cessive Hall crosses labeled as H1 and H2 in
Fig. 1, placed 6 and 8 mm from the lateral p-n
junction. (VH is the Hall voltage and IPH is the
detection Hall cross H1. In this case, only pure
spin current (25–27) reaches crosses H1 and H2.
The experiments in Fig. 1 demonstrate that in our
2DEG microchannel, we can realize the Hall
effect detection of injected spin-polarized elec-
trical currents, as well as pure spin currents. [For
additional measurements of ungated devices,
see (23).]
The conventional field-effect transistor func-
tionality in our 2DEG channel achieved by the
p-layer top gate is demonstrated in Fig. 2A, where
we show the gate voltage dependence of the
channel current and mobility underneath the gate.
At zero gate voltage, we detect only a small re-
sidual channel current consistent with the partial
depletion of the 2DEG in the unetched part of
the heterostructure. By applying forward or re-
verse voltages of an amplitude less than 1 V, we

Downloaded from www.sciencemag.org on December 23, 2010

GaAs, and an n-doped AlGaAs/GaAs heterojunc-
tion underneath. In the unetched part of the wafer,
the top heterojunction is populated by holes,
whereas the 2DEG at the bottom heterojunction
is partly depleted. The n-side of the coplanar p-n
junction is formed by removing the p-doped sur-
face layer from a part of the wafer, thereby pop-
ulating the 2DEG. At zero or reverse bias, the
device is not conductive in the dark due to charge
depletion at the lateral p-n junction. Counterprop-
agating electron and hole currents can be gener-
ated by illumination at subgap wavelengths (22).
Because of the optical selection rules, the out-of-
plane spin polarization of injected electrons is de-
termined by the sense and degree of the circular
polarization of vertically incident light.
The n-region is patterned by electron-beam
lithography into a 1-mm-wide Hall bar along the
½110
crystallographic axis. The effective width
photocurrent.) The oscillations at each Hall cross
and the phase shift between signals at the two
Hall crosses are consistent with a micrometer-
scale spin precession period and with a spin-
diffusion length that extends over more than one
precession period.
Experiments in Fig. 1 are performed in two
distinct electrical measurement configurations. In
Fig. 1A, we show data obtained with the source
and drain electrodes at the far ends of the p- and
n-type sides of the lateral junction, respectively.
In this geometry, spin-polarized electrical currents
reach the detection Hall crosses, similar to exper-
iments performed in (22). In Fig. 1B, the elec-
trical current is drained 4 mm before the first
A VG VB
1000
µ [ cm /Vs ]
can open or close the 2DEG channel, respectively,
at negligible gate-channel leakage current. With-
in the range of measurable signals, we detect gate
voltage induced changes of the channel current
by five orders of magnitude while the mobility
changes by two orders of magnitude. The main
effect of the gate voltage on the channel current
is therefore via direct charge depletion or accu-
mulation of the 2DEG, but mobility changes are
also important. With increasing reverse gate volt-
age, the mobility decreases because the 2DEG is
shifted closer to the ionized donors on the other
side of the AlGaAs/GaAs heterojunction and be-
cause screening of the donor impurity potential
by the 2DEG decreases with depletion.
1
0.1

of individual Hall contacts for local spin detection 100

I [ µA ]
RH 0.01
is 50 to 100 nm, and separation between neigh-
I
2

boring Hall crosses is 2 mm. Electrical gates con- 10 1E-3

trolling the spin currents are placed between one 1E-4
or more pairs of the Hall crosses. The gates are 1
realized by the p-type surface layer areas of the 1E-5
heterostructure, which were locally masked and -0.2 0.0 0 .2 0.4 0 .6 0.8
remained unetched during the fabrication of the VG [ V ]
n-channel Hall bar (24). The laser beam is focused 0.8
to a ∼1- to 2-mm spot at the lateral p-n junction or B
near the junction on the p-side of the epilayer.
x 1
σ+
VG 0.6
For further details on the fabrication of the series 0
VH [ µV ]

RH VB
IPH [µA]
of devices used in our study, employed experi-
IPH -1
∆ x=1µm 0.4
mental techniques, and the theory of the measured
spin-dependent Hall signals, see (23). All exper-
-2
imental data presented below were measured at 0.2
4 K. As illustrated in (22, 23), our ungated and -3
gated devices operate also at high temperatures. 0.0
In Fig. 1, we show experimental results on a -0.5 0.0 0.5 1.0
control device in which we did not pattern the VG [V]
gate electrodes. These measurements extend pre- Fig. 2. (A) Schematics of the measurement setup corresponding to the conventional field-effect tran-
vious demonstration of the spin injection Hall sistor and experimental dependence of the electrical current (blue) through the channel and mobility
effect in similar ungated structures (22). In the (black) underneath the gate on the gate voltage. (B) Schematics of the setup of the spin Hall transistor
previous work, we observed that injected spin- and experimental Hall signals as a function of the gate voltage at a Hall cross placed behind the gate
polarized electrical currents produce Hall effect electrode for two light spot positions with a relative shift of 1 mm and the dashed black curve corre-
signals that are proportional to the out-of-plane sponding to the spot shifted further away from the detection Hall cross. The applied bias voltage VB =
component of the local spin polarization. We also −10 V, the laser intensity is 700 W/cm2, and the laser wavelength is 870 nm. The data demonstrate the
demonstrated that spins precess along the chan- realization of the spin Hall effect transistor.

1802 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


The main result of our work (Fig. 2B) is the The initial increase of the detected output
sensitivity of the measured Hall signal at the signal upon opening the gate is followed by a
cross placed behind the gate on the voltage ap- non-monotonic gate voltage dependence of the
plied to the gate electrode. To exclude any po- Hall voltage (Fig. 2B). This is in marked contrast
tential gate voltage dependence of spin-injection to the monotonic increase of the normal electrical
conditions in our device, we performed the ex- current in the channel observed in the conven-
periments with the electrical current drained be- tional field-effect transistor measurement in Fig.
fore the gated part of the channel (Fig. 2B). The 2A. Apart from blocking the spin current at large
data show two regimes of operation of our spin reverse gate voltages, the intermediate gate elec-
transistor. At large reverse voltages, the Hall sig- tric fields are modifying spin precession of the
nals disappear as the diffusion of spin-polarized injected electrons and therefore the local spin
electrons from the injection region toward the polarization at the detecting Hall cross when the
detecting Hall cross is blocked by the repulsive channel is open. This is the spin manipulation
potential of the intervening gate electrode. Upon regime analogous to the original Datta-Das pro-
opening the gate, the Hall signal first increases, posal of a spin transistor. We further demon-
in analogy to the operation of the conventional strated the presence of this regime in our device
field-effect transistor. We emphasize, however, by comparing two measurements shown in Fig.
that while the optically generated current IPH is 2B: one where the laser spot is aligned close to
kept constant, the electrical current in our the lateral p-n junction on the p-side (red solid
experiments in the manipulation and detection line), and the other with the spot shifted by ~1 mm
parts of the transistor channel remains zero at all in the direction away from the detecting Hall
gate voltages. The onset of the output transverse crosses (black dashed line). The reverse voltage
electrical signal upon opening the gate is a result at which the Hall signals disappear is the same
of a pure spin current. The mechanism by which in the two measurements. For gate voltages at
the spin current generates the output signal can- which the channel is open, the signals are shifted
not be ascribed to a normal charge Hall effect with respect to each other in the two measure-
because of the absence of magnetic field and ments and have opposite sign at certain gate volt-
charge current underneath the cross. ages, and the overall magnitude of the signal is
A B
− transistor structures.
3
VG2 [ V ] σ In Fig. 3D we demonstrate a simple AND
VH1 [ µV ]
2 -0.1
0.0
1
0.1
0
-1.0 -0.5
0.25 σ+
Hall bar 2
0 -1 V
σ+ Hall bar 1 VG1 [ V ]
D 0
H2 VH1 C put 0, the electrical signal at H2 disappears, i.e.,
-0.25 [µV] [µV]
VG1 [ V ] σ is contained in the polarization dependence of
VG2 [V]
-0.5 0 0 -0.9
-3 -3-1 0.8
VH2 [ µV ]
-7
-0.75 -0.1
0.25 00.0 -2
0 1. 0.0
0 +1 0.4
0 6.
3.0 0.1
+0.9 +311
12
-0.25
0 0
-0.5
-0.75 σ- 0.0
-.5 -.25 0.25 -.75 -.5 -.25 0.25 -1.0 -0.5
VG1 [V] VG1 [V] VG2 [ V ]
Fig. 3. (A) Scanning electron micrograph and schematics of the device with two detecting Hall crosses H1
and H2 and one gate placed before cross H1 and the second gate placed behind cross H1 and before
cross H2. Gates and p-side of the lateral p-n junction are highlighted in red. The focused laser beam is
indicated by the yellow spot. (B) Hall signals at cross H1 measured as a function of the first gate voltage.
These gating characteristics are similar to those of the single-gate device in Fig. 2B and have much weaker
dependence on the second gate voltage. (C) Hall signals at cross H2 measured as a function of the second
gate voltage. The curves show strong dependence on the voltages on both gates. (D) Demonstration of the
spin AND logic function by operating both gates (input signals) and measuring the response at Hall cross
H2 (output signal). Measured data at cross H1 are also shown for completeness. The applied bias voltage
VB = −10 V, the laser intensity is 400 W/cm2, and the laser wavelength is 870 nm.
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
REPORTS
larger for smaller separation between injection
and detection points, all confirming the spin pre-
cession origin of the observed effect. [For addi-
tional measurements of gated devices see (23).]
One of the important attributes of our non-
destructive spin detection method integrated, to-
gether with the electrical spin manipulation, along
the semiconductor channel is the possibility of
fabricating devices with a series of Hall cross
detectors and also with a series of gates. In Fig. 3,
we demonstrate the feasibility of this concept
and of the ensuing logic functionality on a spin
Hall effect transistor structure with two gates, the
first placed before cross H1 and the second be-
fore H2. The scanning electron micrograph of
the device is shown in Fig. 3A. The measured
data plotted in Fig. 3B demonstrate that Hall
cross H1 responds strongly to the electric field
Downloaded from www.sciencemag.org on December 23, 2010
on the first gate, with gate voltage characteristics
similar to those observed in the single-gate de-
vice in Fig. 2. As expected for the relative posi-
tions of the injection point, of Hall cross H1, and
of the two gates in the device, the dependence
of the signal at cross H1 on the second gate is
much weaker. By contrast, Hall cross H2 re-
sponds strongly to both gates (Fig. 3C). Before
the spin can reach the detecting Hall cross H2, it
is manipulated by two external parameters. This
is analogous to the measurement in Fig. 2B in
which the position of the injection point played
the role of the second parameter. The analogy be-
tween results in Figs. 2B and 3C further demon-
strates the spin origin of the functionality of our
logic functionality by operating both gates and
by measuring the Hall electrical signal at cross
H2. Intermediate gate voltages on both gates rep-
resent the input value 1 and give the largest elec-
trical signal at H2 (positive for s− helicity of the
incident light), representing the output value 1.
0.0 When we apply to any of the two gates a large
reverse (negative) gate voltage, representing in-
− the output is 0. Note that additional information
the detected Hall signals, as illustrated in Fig. 3D.
Proceeding to the theoretical analysis of the
measured data, we first characterize the transport
regime in which our devices operate. The 2DEG
mobilities in the etched, n-type part of the wafer
and underneath the p-layer gates are ≲3 × 103
cm2/Vs, corresponding to a mean-free path ≲102
0.0
nm. This is much smaller than the precession
length and the length of our 2DEG channel, i.e.,
the experiments are done in the diffusive,
strong-disorder weak spin-orbit coupling regime.
As explained in (22), the Hall effect and the spin-
precession effect can be decoupled in this regime.
The Hall effect measures the local out-of-plane
component of the spin polarization of carriers
and originates from the spin-orbit coupling in-
duced skew scattering. [See (23) for quantitative
estimates of the Hall signals that are consistent
with experiment.] In the following analysis, we
1803
 REPORTS
focus on the spin-precession and spin-diffusion
lengths. The possibility of observing and using
spin precession of an ensemble of electrons in
the diffusive regime is demonstrated by our nu-
merical Monte Carlo simulations (22, 28) shown
in Fig. 1C.
The numerically obtained spin-precession
period is well described by an analytical formula
derived from the dynamics of the spin-density
matrix (28), LSO = pħ2/m*(|a| + |b|); m* = 0.067
is the electron effective mass in GaAs. There are
two regimes in which spin precession can be ob-
served in the diffusive transport regime. In one
regime, the width of the channel is not relevant
and a spin-diffusion length larger than the pre-
cession length occurs as a result of the single-
particle transport analog of the spin helix state
(9) realized at 2DEG Rashba and Dresselhaus
diffusive regime and for a ≠ −b, and the spin-
precession and spin-diffusion lengths in this re-
gime are independent of the mean-free-path, i.e.,
of the mobility of the 2DEG channel (28).
The strength of the confining electric field of
the 2DEG underneath the gate changes by up to
a factor of ∼2 in the range of applied gate volt-
ages in our experiments. This result implies (22)
comparably large changes in the strength of the
internal spin-orbit fields in the 2DEG channel.
The dependence on the spin-orbit field strength
shown in the above equation and confirmed by
Monte Carlo simulations (28) (and the indepen-
dence on the momentum of injected electrons)
implies also comparably large changes in the
spin-precession length. These estimates corrob-
orate the observed spin manipulation in our spin
Hall effect transistors by external electric fields
13. X. Lou et al., Nat. Phys. 3, 197 (2007).
14. B. Huang, D. J. Monsma, I. Appelbaum, Phys. Rev. Lett.
99, 177209 (2007).
15. H. C. Koo et al., Science 325, 1515 (2009).
16. M. I. Dyakonov, V. I. Perel, Phys. Lett. A 35, 459 (1971).
17. J. E. Hirsch, Phys. Rev. Lett. 83, 1834 (1999).
18. S. Murakami, N. Nagaosa, S.-C. Zhang, Science 301,
1348 (2003).
19. J. Sinova et al., Phys. Rev. Lett. 92, 126603 (2004).
20. Y. K. Kato, R. C. Myers, A. C. Gossard, D. D. Awschalom,
Science 306, 1910 (2004).
21. J. Wunderlich, B. Kaestner, J. Sinova, T. Jungwirth,
Phys. Rev. Lett. 94, 047204 (2005).
22. J. Wunderlich et al., Nat. Phys. 5, 675 (2009).
23. Materials and methods are available on Science online.
24. B. Kaestner, J. Wunderlich, T. J. B. M. Janssen, J. Mod. Opt.
54, 431 (2007).
25. S. O. Valenzuela, M. Tinkham, Nature 442, 176 (2006).
26. C. Brüne et al., Nat. Phys. 6, 448 (2010).
27. E. S. Garlid, Q. O. Hu, M. K. Chan, C. J. Palmstrøm,
P. A. Crowell, Phys. Rev. Lett. 105, 156602 (2010).
28. L. P. Zârbo, J. Sinova, I. Knezevic, J. Wunderlich,

Downloaded from www.sciencemag.org on December 23, 2010

spin-orbit fields of equal or similar strengths,
a ≈ −b for our bar orientation. When the two
spin-orbit fields are not tuned to similar strengths,
the spin-diffusion length is approximately given
by ∼L2SO =w and spin precession is therefore ob-
servable only when the width w of the channel
is comparable to or smaller than the precession
length (28–30).
The complex design of our semiconductor
heterostructure provides simultaneously the means
for spin injection, electrical gating, and detec-
tion, so we did not rely on further fine tuning of
the internal spin-orbit fields to realize the spin
helix state condition. Instead, we fabricated nar-
row Hall bars whose width is smaller than the
precession length and used a strongly focused
light spot for spin injection. As shown in Fig.
1C, several precessions are readily observable in
this quasi one-dimensional geometry even in the
applied to the gates.
References and Notes
1. I. Žutić, J. Fabian, S. Das Sarma, Rev. Mod. Phys. 76, 323
(2004).
2. T. Dietl, D. D. Awschalom, M. Kaminska, H. Ohno, Eds.,
Spintronics, vol. 82 of Semiconductors and Semimetals
(Elsevier, Amsterdam, 2008).
3. S. Datta, B. Das, Appl. Phys. Lett. 56, 665 (1990).
4. J. M. Kikkawa, D. D. Awschalom, Nature 397, 139
(1999).
5. H. J. Zhu et al., Phys. Rev. Lett. 87, 016601 (2001).
6. P. R. Hammar, M. Johnson, Phys. Rev. Lett. 88, 066806
(2002).
7. G. Schmidt, L. W. Molenkamp, Semicond. Sci. Technol.
17, 310 (2002).
8. J. Schliemann, J. C. Egues, D. Loss, Phys. Rev. Lett. 90,
146801 (2003).
9. B. A. Bernevig, J. Orenstein, S.-C. Zhang, Phys. Rev. Lett.
97, 236601 (2006).
10. X. Jiang et al., Phys. Rev. Lett. 94, 056601 (2005).
11. S. A. Crooker et al., Science 309, 2191 (2005).
12. C. P. Weber et al., Phys. Rev. Lett. 98, 076604 (2007).
T. Jungwirth, Phys. Rev. B 82, 205320 (2010).
29. A. A. Kiselev, K. W. Kim, Phys. Rev. B 61, 13115 (2000).
30. S. Kettemann, Phys. Rev. Lett. 98, 176808 (2007).
31. We acknowledge support from European Union grant
FP7-215368 SemiSpinNet; Czech Republic grants
AV0Z10100521, MSM0021620834, KAN400100652,
LC510, and Preamium Academiae; and U.S. grants
NSF-MRSEC DMR-0820414, ONR-N000140610122,
DMR-0547875, and SWAN-NRI. J.S. is a Cottrell Scholar
of Research Corporation. In connection with this work,
we have two pending patent applications with the
European Patent Office, patent numbers EP 2 224 500
A2 and EP 2 190 022 A.
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1801/DC1
Materials and Methods
SOM Text
Figs. S1 to S10
Table S1
References
29 July 2010; accepted 22 November 2010
10.1126/science.1195816

though direct experimental evidence validating

Brownian Motion of Stiff Filaments in this powerful theoretical intuition came over two
decades later, when reptation of flexible and semi-
a Crowded Environment flexible filaments was observed directly by imag-
ing fluorescently labeled DNA (5) and actin (6).
Nikta Fakhri,1 Frederick C. MacKintosh,2 Brahim Lounis,3 Laurent Cognet,3 Matteo Pasquali1* In contrast, little is known about the thermal
motion of stiff filaments such as carbon nano-
The thermal motion of stiff filaments in a crowded environment is highly constrained and tubes, biopolymers, and stiff fibers in a network.
anisotropic; it underlies the behavior of such disparate systems as polymer materials, In particular, the role of the bending stiffness of
nanocomposites, and the cell cytoskeleton. Despite decades of theoretical study, the fundamental such inclusions remains controversial, with long-
dynamics of such systems remains a mystery. Using near-infrared video microscopy, we studied standing conflicting theoretical predictions (7–11).
the thermal diffusion of individual single-walled carbon nanotubes (SWNTs) confined in porous Doi predicted that rotational diffusion is inde-
agarose networks. We found that even a small bending flexibility of SWNTs strongly enhances pendent of stiffness (7), whereas Odijk concluded
their motion: The rotational diffusion constant is proportional to the filament-bending compliance that such diffusion should be enhanced by flex-
and is independent of the network pore size. The interplay between crowding and thermal ibility (9) and Sato concluded the opposite (11).
bending implies that the notion of a filament’s stiffness depends on its confinement. Moreover, the Bulk experiments by means of birefringence and
mobility of SWNTs and other inclusions can be controlled by tailoring their stiffness. dichroism (12–14) have also given conflicting

rowding greatly constrains the transversal Gennes, Doi, and Edwards (1–3) modeled the ef-

C
mobility of a filament and causes aniso-
tropic diffusion, which is limited to the
filament axial direction. In the case of polymer
solutions or melts, understanding the motion of a
single polymer chain confined by the meshwork
of its neighbors was key to a number of advances
in polymer science. In their seminal work, de
fect of crowding on polymer dynamics by intro-
ducing the concept of a confining tube, together
with preferential motion along the polymer’s axis,
known as reptation because of its resemblance to
the slithering of a snake (Fig. 1A, inset). This
model captured many bulk dynamical properties
of flexible polymer melts and solutions (4), al-
1
Department of Chemical and Biomolecular Engineering,
Department of Chemistry, Smalley Institute for Nanoscale Science
and Technology, Rice University, Houston, TX 77005, USA.
2
Department of Physics and Astronomy, Vrije Universiteit, 1081
HV Amsterdam, Netherlands. 3Centre de Physique Moléculaire
Optique et Hertzienne, Université de Bordeaux CNRS, Talence
F-33405, France.
*To whom correspondence should be addressed. E-mail:
mp@rice.edu

1804 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

 REPORTS
focus on the spin-precession and spin-diffusion
lengths. The possibility of observing and using
spin precession of an ensemble of electrons in
the diffusive regime is demonstrated by our nu-
merical Monte Carlo simulations (22, 28) shown
in Fig. 1C.
The numerically obtained spin-precession
period is well described by an analytical formula
derived from the dynamics of the spin-density
matrix (28), LSO = pħ2/m*(|a| + |b|); m* = 0.067
is the electron effective mass in GaAs. There are
two regimes in which spin precession can be ob-
served in the diffusive transport regime. In one
regime, the width of the channel is not relevant
and a spin-diffusion length larger than the pre-
cession length occurs as a result of the single-
particle transport analog of the spin helix state
(9) realized at 2DEG Rashba and Dresselhaus
diffusive regime and for a ≠ −b, and the spin-
precession and spin-diffusion lengths in this re-
gime are independent of the mean-free-path, i.e.,
of the mobility of the 2DEG channel (28).
The strength of the confining electric field of
the 2DEG underneath the gate changes by up to
a factor of ∼2 in the range of applied gate volt-
ages in our experiments. This result implies (22)
comparably large changes in the strength of the
internal spin-orbit fields in the 2DEG channel.
The dependence on the spin-orbit field strength
shown in the above equation and confirmed by
Monte Carlo simulations (28) (and the indepen-
dence on the momentum of injected electrons)
implies also comparably large changes in the
spin-precession length. These estimates corrob-
orate the observed spin manipulation in our spin
Hall effect transistors by external electric fields
13. X. Lou et al., Nat. Phys. 3, 197 (2007).
14. B. Huang, D. J. Monsma, I. Appelbaum, Phys. Rev. Lett.
99, 177209 (2007).
15. H. C. Koo et al., Science 325, 1515 (2009).
16. M. I. Dyakonov, V. I. Perel, Phys. Lett. A 35, 459 (1971).
17. J. E. Hirsch, Phys. Rev. Lett. 83, 1834 (1999).
18. S. Murakami, N. Nagaosa, S.-C. Zhang, Science 301,
1348 (2003).
19. J. Sinova et al., Phys. Rev. Lett. 92, 126603 (2004).
20. Y. K. Kato, R. C. Myers, A. C. Gossard, D. D. Awschalom,
Science 306, 1910 (2004).
21. J. Wunderlich, B. Kaestner, J. Sinova, T. Jungwirth,
Phys. Rev. Lett. 94, 047204 (2005).
22. J. Wunderlich et al., Nat. Phys. 5, 675 (2009).
23. Materials and methods are available on Science online.
24. B. Kaestner, J. Wunderlich, T. J. B. M. Janssen, J. Mod. Opt.
54, 431 (2007).
25. S. O. Valenzuela, M. Tinkham, Nature 442, 176 (2006).
26. C. Brüne et al., Nat. Phys. 6, 448 (2010).
27. E. S. Garlid, Q. O. Hu, M. K. Chan, C. J. Palmstrøm,
P. A. Crowell, Phys. Rev. Lett. 105, 156602 (2010).
28. L. P. Zârbo, J. Sinova, I. Knezevic, J. Wunderlich,

Downloaded from www.sciencemag.org on December 23, 2010

spin-orbit fields of equal or similar strengths,
a ≈ −b for our bar orientation. When the two
spin-orbit fields are not tuned to similar strengths,
the spin-diffusion length is approximately given
by ∼L2SO =w and spin precession is therefore ob-
servable only when the width w of the channel
is comparable to or smaller than the precession
length (28–30).
The complex design of our semiconductor
heterostructure provides simultaneously the means
for spin injection, electrical gating, and detec-
tion, so we did not rely on further fine tuning of
the internal spin-orbit fields to realize the spin
helix state condition. Instead, we fabricated nar-
row Hall bars whose width is smaller than the
precession length and used a strongly focused
light spot for spin injection. As shown in Fig.
1C, several precessions are readily observable in
this quasi one-dimensional geometry even in the
applied to the gates.
References and Notes
1. I. Žutić, J. Fabian, S. Das Sarma, Rev. Mod. Phys. 76, 323
(2004).
2. T. Dietl, D. D. Awschalom, M. Kaminska, H. Ohno, Eds.,
Spintronics, vol. 82 of Semiconductors and Semimetals
(Elsevier, Amsterdam, 2008).
3. S. Datta, B. Das, Appl. Phys. Lett. 56, 665 (1990).
4. J. M. Kikkawa, D. D. Awschalom, Nature 397, 139
(1999).
5. H. J. Zhu et al., Phys. Rev. Lett. 87, 016601 (2001).
6. P. R. Hammar, M. Johnson, Phys. Rev. Lett. 88, 066806
(2002).
7. G. Schmidt, L. W. Molenkamp, Semicond. Sci. Technol.
17, 310 (2002).
8. J. Schliemann, J. C. Egues, D. Loss, Phys. Rev. Lett. 90,
146801 (2003).
9. B. A. Bernevig, J. Orenstein, S.-C. Zhang, Phys. Rev. Lett.
97, 236601 (2006).
10. X. Jiang et al., Phys. Rev. Lett. 94, 056601 (2005).
11. S. A. Crooker et al., Science 309, 2191 (2005).
12. C. P. Weber et al., Phys. Rev. Lett. 98, 076604 (2007).
T. Jungwirth, Phys. Rev. B 82, 205320 (2010).
29. A. A. Kiselev, K. W. Kim, Phys. Rev. B 61, 13115 (2000).
30. S. Kettemann, Phys. Rev. Lett. 98, 176808 (2007).
31. We acknowledge support from European Union grant
FP7-215368 SemiSpinNet; Czech Republic grants
AV0Z10100521, MSM0021620834, KAN400100652,
LC510, and Preamium Academiae; and U.S. grants
NSF-MRSEC DMR-0820414, ONR-N000140610122,
DMR-0547875, and SWAN-NRI. J.S. is a Cottrell Scholar
of Research Corporation. In connection with this work,
we have two pending patent applications with the
European Patent Office, patent numbers EP 2 224 500
A2 and EP 2 190 022 A.
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1801/DC1
Materials and Methods
SOM Text
Figs. S1 to S10
Table S1
References
29 July 2010; accepted 22 November 2010
10.1126/science.1195816

though direct experimental evidence validating

Brownian Motion of Stiff Filaments in this powerful theoretical intuition came over two
decades later, when reptation of flexible and semi-
a Crowded Environment flexible filaments was observed directly by imag-
ing fluorescently labeled DNA (5) and actin (6).
Nikta Fakhri,1 Frederick C. MacKintosh,2 Brahim Lounis,3 Laurent Cognet,3 Matteo Pasquali1* In contrast, little is known about the thermal
motion of stiff filaments such as carbon nano-
The thermal motion of stiff filaments in a crowded environment is highly constrained and tubes, biopolymers, and stiff fibers in a network.
anisotropic; it underlies the behavior of such disparate systems as polymer materials, In particular, the role of the bending stiffness of
nanocomposites, and the cell cytoskeleton. Despite decades of theoretical study, the fundamental such inclusions remains controversial, with long-
dynamics of such systems remains a mystery. Using near-infrared video microscopy, we studied standing conflicting theoretical predictions (7–11).
the thermal diffusion of individual single-walled carbon nanotubes (SWNTs) confined in porous Doi predicted that rotational diffusion is inde-
agarose networks. We found that even a small bending flexibility of SWNTs strongly enhances pendent of stiffness (7), whereas Odijk concluded
their motion: The rotational diffusion constant is proportional to the filament-bending compliance that such diffusion should be enhanced by flex-
and is independent of the network pore size. The interplay between crowding and thermal ibility (9) and Sato concluded the opposite (11).
bending implies that the notion of a filament’s stiffness depends on its confinement. Moreover, the Bulk experiments by means of birefringence and
mobility of SWNTs and other inclusions can be controlled by tailoring their stiffness. dichroism (12–14) have also given conflicting

rowding greatly constrains the transversal Gennes, Doi, and Edwards (1–3) modeled the ef-

C
mobility of a filament and causes aniso-
tropic diffusion, which is limited to the
filament axial direction. In the case of polymer
solutions or melts, understanding the motion of a
single polymer chain confined by the meshwork
of its neighbors was key to a number of advances
in polymer science. In their seminal work, de
fect of crowding on polymer dynamics by intro-
ducing the concept of a confining tube, together
with preferential motion along the polymer’s axis,
known as reptation because of its resemblance to
the slithering of a snake (Fig. 1A, inset). This
model captured many bulk dynamical properties
of flexible polymer melts and solutions (4), al-
1
Department of Chemical and Biomolecular Engineering,
Department of Chemistry, Smalley Institute for Nanoscale Science
and Technology, Rice University, Houston, TX 77005, USA.
2
Department of Physics and Astronomy, Vrije Universiteit, 1081
HV Amsterdam, Netherlands. 3Centre de Physique Moléculaire
Optique et Hertzienne, Université de Bordeaux CNRS, Talence
F-33405, France.
*To whom correspondence should be addressed. E-mail:
mp@rice.edu

1804 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


results, mainly because polydispersity, aggrega-
tion, attractive forces, and strong coupling be-
tween translational and rotational diffusivities of
the filaments all complicate the interpretation of
the results. We directly visualized single-walled
carbon nanotubes (SWNTs) reptating in a gel and
established that flexibility substantially speeds up
diffusion of stiff filaments under confinement,
which is in accord with Odijk’s theory (9). We
found that the rotational diffusion constant grows
linearly with the bending flexibility and, counter-
intuitively, is independent of degree of crowding.
A natural measure of the stiffness of a fila-
ment is its persistence length, Lp = k/kBT, which
measures its thermal bending by Brownian forces.
Here, k is the bending stiffness, T is the temper-
ature, and kB is Boltzmann’s constant. Doi (7)
postulated that as long as the rods are stiff (L < Lp),
flexibility does not affect diffusion, and that such a
stiff filament of length L confined in a tube of
diameter x << L would explore an angle q ≈ x/L in
the reptation time trep = L2/D‖ needed to diffuse a
length L. This yields a rotational diffusivity DDoi
r =
q2/trep = kBTx2/hL5, where D‖ ~ kBT/hL is the
translational diffusivity of an isolated filament in a
solvent of viscosity h. In contrast, Odijk (9) argued
that whenever the amplitude of the thermal un-
dulations u = (L3/Lp)1/2 exceeds the pore diameter,
confinement results in the independent deflection of
segments of length l = (Lpx2)1/3, of which there are
REPORTS
L/l (15). As the filament reptates a distance l, the
ends of the filament reorient by an angle dq ~ x/l.
After a reptation time, this results in a mean-square
angular deflection of the filament of q2 ~ x2L/l3
and an angular diffusivity of DOdijk
r = kBT/hL2Lp.
Doi’s theory is recovered for filaments shorter
than l, a length typically much shorter than the
persistence length, at which the flexibility be-
comes irrelevant. Otherwise, Odijk theory pre-
dicts that rotational diffusion speeds up by a factor
3
DOdijk
r /DDoi
r = (L/l) , for example, by three orders of
magnitude for a 10–mm-long, stiff (Lp = 100 mm)
filament moving through 100-nm pores. (9)
SWNTs are the ideal system to study confined
dynamics of stiff filaments. SWNTs are slender
(typical diameters of d ≈ 0.7 to 1.2 nm), sufficiently
long to be visualized through optical microscopy
(L ≈ 3 to 15 mm), and share many dynamical

Downloaded from www.sciencemag.org on December 23, 2010

characteristics with polymers (17, 18). SWNTs are
considered stiff because their persistence length
ranges from 20 to 150 mm and scales with their
diameter cubed, Lp ~ d 3, similar to the bending
stiffness of a macroscopic hollow pipe (19).
Individual semiconducting SWNTs can be visual-
ized directly because of their bright near-infrared
(NIR) luminescence, and their diameter can be
determined simultaneously spectroscopically (20).
We image the quasi-two-dimensional dynamics of
these individual SWNTs in agarose gel (21), a
permanent network with pores x ≈ 0.1 to 1 mm
[depending on agarose concentration (22, 23)],
which mimics the reptation ansatz of a filament
moving in a fixed network of frozen obstacles
(1–3). The diameter (hence the persistence length)
of each SWNT was determined from its emission
spectrum (19, 20). By means of image analysis,
Fig. 1. (A) (x, y) center-of-mass trajectories of a SWNT reptating in 1.5 w/w % agarose gel and we extracted frame-by-frame each SWNT’s
representative NIR images of the SWNT, illustrating the effect of flexibility on reorientation of SWNT in center-of-mass position ri = [xi, yi] in the lab
different pores (scale bar, 5 mm). (Inset) Schematic of a stiff filament in a fixed network: L is the length of the coordinates and its orientation qi relative to the x
filament, l is the deflection length, and x is the pore size of the network. (B) Individual SWNT emission axis (i represents the frame number spaced by 30
spectrum with peak at 985 nm, implying a (6, 5) structure with a diameter of 0.76 nm and persistence length ms acquisition time). Figure 1A depicts the center-
of 26 mm. (C) Angular MSD showing the subdiffusive-to-diffusive behavior which occurs at disengagement of-mass trajectory of a 4.5-mm-long (6,5) SWNT
time td. The line is the best fit to the data. Dr is calculated from the long time (diffusive behavior). (Inset) [deduced from its emission spectrum (Fig. 1B)],
Representative image of a SWNT in x-y lab frame; the orientation angle q is the angle between the x axis and with a 0.76 nm diameter and Lp = 26 mm (19) in a
the major axis of the best-fitted ellipse to the shape of the SWNT. 1.5% w/w agarose gel (x ≈ 0.2 mm); this figure and
the accompanying video (24) show unequivocally
Fig. 2. Normalized rota- snake-like motion. NIR fluorescence snapshots
tional diffusivity of 35 demonstrate that flexibility substantially affects
SWNTs with different length reorientation of the SWNT in a new confining
and persistence length tube. At first, the SWNT slides back and forth
[denoted by different partially out of the confining tube. By bending
symbols (table S1)], reptat- slightly, the end of the SWNT has more freedom to
ing in different concentra- explore various paths while translating along its
tions of agarose gel versus length, even though most of the SWNT is still caged
normalized length by de- and thus restricted to a certain orientation. Eventu-
flection length. Doi’s theory ally, the SWNT completely slides out of the original
is shown by a dashed line confining tube and reorients in another tube.
and predicts a power law We quantify rotational motion by the statistics of
with scaling exponent –3 the angle qi. A typical time-evolution of the mean-
across the whole range of
square angular displacement (MSAD), 〈Dq2 〉, is
normalized length. Odijk’s
shown in Fig. 1C. At short times, the SWNT’s
theory is denoted by a
solid line and predicts a angular diffusion is subdiffusive (〈Dq2 〉º tn ,
plateau at ~1 for L > l. n << 1), reflecting the confinement in the initial
tube. At longer times, the SWNT diffuses out of
the initial tube, and the mean angular displacement

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1805

 REPORTS
Fig. 3. (A) The disengagement time normalized by deflection length l2 scales linearly with length L. (B) Time
domains in translational MSDs parallel (Ds2) and perpendicular (Dn2) to the time-averaged reptation path
(corresponding to the SWNT in Fig. 1A). t < td is the dynamics inside the tube. td < t < tr is the crossover
region between anisotropic and isotropic diffusion. t > tr is the isotropic diffusive long-time dynamics.
behaves diffusively 〈Dq2 〉 ¼ 2Dr t (25), yielding
the value of the rotational diffusivity Dr.
We measured the rotational diffusivity of 35
SWNTs with different lengths (2 to 10 mm) and
persistence lengths (26 to 60 mm), reptating in
agarose gels of several concentrations (hence
pore sizes). We collapsed the rotational diffusiv-
ity on a master curve (Fig. 2) by plotting the
normalized rotational diffusivities Dr / DOdijk
r =
DrhL2Lp/kBT versus normalized length L/l. In
such a plot, Doi’s theory predicts a power law
with scaling exponent –3 [(L/l)–3] across the
whole range of normalized length (Fig. 2, dashed
line), whereas Odijk’s theory predicts a plateau at
~1 for L > l (Fig. 2, solid line). The data show
that when L ≥ l, flexibility does not affect
mobility (which is in agreement with both Doi
and Odijk), whereas for L > l flexibility clearly
speeds up long-time diffusion, which follows
Odijk’s scaling. Therefore, the effective rigidity
of a filament depends on its degree of confine-
ment. Whereas in the absence of confinement
Brownian filaments can be considered essentially
as rigid when L < Lp, confined filaments (x < L)
behave as rigid when L > l.
We next turned to the short-time subdiffusive
dynamics of the MSAD (Fig. 1C). To cross over
from short time subdiffusive behavior to long-
time diffusive motion, a filament must diffuse by
a length l out of its initial confining tube. This
occurs on a time scale known as the disengagement
time td, which is the time scale a SWNT needs
to reptate by a deflection length and is deter-
mined from the free parallel diffusion constant
of the center of the mass, td = l2/D‖ ~ hl2L/kBT
(9). At times shorter than td, the SWNT wiggles
“freely” inside its initial confining tube, with min-
imal angular reorientation (q < x/L, hence the
subdiffusive behavior of MSAD in Fig. 1C). At
times longer than td, the SWNT slides out of
the initial confining tube and starts exploring the
other accessible tubes. Show in Fig. 3 are the
disengagement times normalized to l2 obtained
for 11 (6,5) SWNTs by fitting the MSAD with
〈Dq2 〉 ¼ q20 þ 2Dr t and settingtd ¼ q20 =2Dr . We
1806 24 DECEMBER 2010
found that the measured td normalized by de-
flection length l2 scales linearly with length L,
confirming Odijk’s prediction (9) for short-time
translational diffusion (26) and showing that
flexibility speeds up disengagement.
Because SWNTs explore orientation space by
reptating in and out of pores, rotational and
translational diffusion should be strongly coupled
at time scales below the rotational diffusion time
tr = 1/2Dr . Such coupling occurs even in the much
simpler case of two-dimensional Brownian mo-
tion of an unconstrained ellipsoid and is well
described in terms of Perrin-Smoluchowski the-
ory (27). Theoretical calculations and simulations
have recently shown that this same theory can
capture such coupling in the motion of confined
rigid rods (infinite Lp) (28). To investigate this
coupling experimentally, we measured the time
evolution of the center-of-mass mean square dis-
placements (MSDs) parallel (Ds2) and perpendic-
ular (Dn2) to the orientation of the reptation tube,
averaged over the same time window (23). The
parallel and perpendicular MSDs versus time
(normalized by the rotational diffusion time) are
shown in Fig. 3B.
At short times (t < td), SWNT diffusion is
anisotropic—Ds2 >> Dn2; SWNTs diffuse much
faster parallel than perpendicular to the tube axis.
In this time regime, the dynamics of center of
mass is dominated by the relaxation of thermally
excited elastic bending modes of the SWNT, with
relaxation times t nr e hln4 =k, where ln is the mode
wavelength (29). For a given time t, long-wave
modes ðtnr > tÞ are effectively “frozen,” whereas
short-wave modes ðtnr < tÞ evolve and contrib-
ute to the amplitude of thermal undulations. At
time t, the longest (dominant) bending mode has
a wavelength of l(t) ~ (kt/h)1/4. The mean square
amplitude of the transverse fluctuations (Du2) of
this mode dominate the transverse diffusion of
the center of mass and evolves with time (29–31)
as Du2 ≈ Dn2 ~ l(t)3/Lp ~ t3/4, which is indeed the
subdiffusive power law t3/4 we measured (Fig.
3B). The same time dependence, t3/4, is also
expected for the mean square amplitude of the
VOL 330 SCIENCE www.sciencemag.org
longitudinal fluctuations of the SWNT (29, 31);
these dominate the mean square longitudinal
displacement of the center of mass Ds2, as shown
in Fig. 3B. At longer times, Ds2 crosses over to a
linear diffusion regime, indicating that the SWNT
has fully reptated along its length (Ds2 ~ t). In this
crossover regime, the transverse MSD Dn2 grows
super-linearly with time because reptation occurs
along a curved path—a motion that couples ro-
tation and translation (Dn2 ~ D‖tDq2 ~ D‖Drt2)
(10, 28). Thus on intermediate time scales between
disengagement and rotational diffusion times (td
and tr), translational diffusion perpendicular to the
filament is also enhanced by flexibility. At times
longer than rotational diffusion time tr , the SWNT
loses memory of its initial orientation, and its
diffusion becomes isotropic. On these time scales,
translational diffusion is weakly reduced by
Downloaded from www.sciencemag.org on December 23, 2010
flexibility (32).
By varying SWNT surface modifications (33),
we can selectively tune the sensitivity of the carbon
nanotubes to the different physical properties of the
porous media for transport and sensing applica-
tions (such as a cellular crowded environment).
The orientation dynamic behavior of SWNTs in a
fixed network is a starting point to study the dy-
namics of concentrated solutions of SWNTs as
well as SWNT composite materials. Our results
indicate that the SWNT shapes are altered by the
presence of the pores and that bent shapes can be
very long lived. Rotational diffusion and coupling
between translational and rotational motion of
SWNTs can provide a useful counterpart to
translational diffusion approaches in microrheol-
ogy techniques and render the ability to probe
different viscoelastic modes or local heterogeneity
in complex fluids and biological media.
References and Notes
1. P. G. de Gennes, J. Chem. Phys. 55, 572 (1971).
2. M. Doi, S. F. Edwards, J. Chem. Soc., Faraday Trans. II 74,
1789 (1978).
3. S. F. Edwards, Proc. Phys. Soc. Lond. 92, 9 (1967).
4. M. Doi, S. F. Edwards, The Theory of Polymer Dynamics
(Oxford Univ. Press, Oxford, 1986).
5. T. T. Perkins, D. E. Smith, S. Chu, Science 264, 819 (1994).
6. J. Käs, H. Strey, E. Sackmann, Nature 368, 226 (1994).
7. M. Doi, J. Phys. 36, 607 (1975).
8. F. Hofling, T. Munk, E. Frey, T. Franosch, Phys. Rev. E
Stat. Nonlin. Soft Matter Phys. 77, 060904R (2008).
9. T. Odijk, Macromolecules 16, 1340 (1983).
10. S. Ramanathan, D. C. Morse, Phys. Rev. E Stat. Nonlin.
Soft Matter Phys. 76, 010501 (2007).
11. T. Sato, Y. Takada, A. Teramoto, Macromolecules 24,
6220 (1991).
12. M. Tracy, R. Pecora, Annu. Rev. Phys. Chem. 43, 525 (1992).
13. S. S. Wijmenga, A. Maxwell, Biopolymers 25, 2173
(1986).
14. K. M. Zero, R. Pecora, Macromolecules 15, 87 (1982).
15. This characteristic length scale has been estimated from
thermal fluctuations of flexible filaments in an ordered
polymer background (16).
16. Z. Dogic et al., Phys. Rev. Lett. 92, 125503 (2004).
17. M. J. Green, N. Behabtu, M. Pasquali, W. W. Adams,
Polymer (Guildf.) 50, 4979 (2009).
18. R. Duggal, M. Pasquali, Phys. Rev. Lett. 96, 246104 (2006).
19. N. Fakhri, D. A. Tsyboulski, L. Cognet, R. B. Weisman,
M. Pasquali, Proc. Natl. Acad. Sci. U.S.A. 106, 14219 (2009).
20. D. A. Tsyboulski, S. M. Bachilo, R. B. Weisman, Nano Lett.
5, 975 (2005).
21. L. Cognet et al., Science 316, 1465 (2007).

22. T. K. Attwood, B. J. Nelmes, D. B. Sellen, Biopolymers 27,
201 (1988).
23. N. Pernodet, M. Maaloum, B. Tinland, Electrophoresis 18,
55 (1997).
24. Materials and methods are available as supporting
material on Science Online.
25. Rotational diffusion is characterized in two dimensions by
a single diffusion coefficient, Dr, and associated diffusion
time, tr = 1/2Dr.
26. The errors introduced by the limited angular resolution in
our measurements can affect the interpretation of the
short time dynamics. The microscope angular resolution
is ≈a/L, where a is the pixel size. Therefore, resolution
limits our experiments below a resolution time of
tresolution ≅ a2Lpph/2kBT. In the experimental conditions
of Fig. 3A, tresolution/td = [(l/L)(a/x)2]/4 ranges from
0.01 to 0.26; therefore, tresolution << td and resolution
Tunable Field Control Over
the Binding Energy of Single Dopants
by a Charged Vacancy in GaAs
D. H. Lee and J. A. Gupta*
Local manipulation of electric fields at the atomic scale may enable new methods for quantum
transport and creates new opportunities for field control of ferromagnetism and spin-based quantum
information processing in semiconductors. We used a scanning tunneling microscope to position
charged arsenic (As) vacancies in the gallium arsenide 110 [GaAs(110)] surface with atomic precision,
thereby tuning the local electrostatic field experienced by single manganese (Mn) acceptors. The effects
of this field are quantified by measuring the shift of an acceptor state within the band gap of GaAs.
Experiments with varying tip-induced band-bending conditions suggest a large binding energy for
surface-layer Mn, which is reduced by direct Coulomb repulsion when the As vacancy is moved nearby.
he scaling of electronic devices such as
T field-effect transistors to nanometer dimen-
sions requires more precise control of in-
dividual dopants in semiconductor nanostructures,
because statistical fluctuations in dopant distribu-
tions are beginning to affect the performance and
functionality of current devices (1–4). Proposals
for next-generation quantum- and spin-based elec-
tronics also rely on the tuning of the charge, spin,
and interactions of dopant atoms with local elec-
tric fields [e.g., P in Si (5) or Mn in GaAs (6)]. On
Si surfaces, the scanning tunneling microscope
(STM) has been used to probe the influence of
charged dangling bonds on molecular conductance
(7). In III-V semiconductors, recent STM studies
of single Si, Zn, and Mn dopants have shown that
the electronic and magnetic properties of such im-
purities depend on proximity to the surface (8–14),
other impurities (15), interactions with the STM tip
(9, 14, 16, 17), and local strain fields (18).
Here, we demonstrate control of single-dopant
properties by using the local electrostatic field
emanating from a charged vacancy. Using an
STM, we can position this vacancy with atomic
precision or reversibly switch it to a neutral state
to tune the binding energy of holes to individual
Department of Physics, Ohio State University, Columbus, OH
43210, USA.
*To whom correspondence should be addressed. E-mail:
gupta.208@osu.edu
www.sciencemag.org
REPORTS
does not substantially affect the measurement of the EEC-0647452), the Welch Foundation (grant C-1668),
subdiffisive regime due to the short-time dynamics in the Advanced Energy Consortium (www.beg.utexas.edu/
the system. aec), the Région Aquitaine, the Agence Nationale pour la
27. Y. Han et al., Science 314, 626 (2006). Recherche (ANR PNANO), the European Research Council
28. T. Munk, F. Hofling, E. Frey, T. Franosch, Europhys. Lett. (grant n 232942), and the Foundation for Fundamental
85, 30003 (2009). Research on Matter (FOM), which is part of the
29. F. Gittes, F. C. MacKintosh, Phys. Rev. E Stat. Phys. Netherlands Organisation for Scientific Research (NWO).
Plasmas Fluids Relat. Interdiscip. Topics 58, R1241
(1998).
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1804/DC1
30. E. Farge, A. C. Maggs, Macromolecules 26, 5041
(1993). Materials and Methods
Fig. S1
31. R. Granek, J. Phys. II 7, 1761 (1997).
32. M. Doi, J. Polymer Sci. Polymer Symp. 73, 93 (1985). Table S1
33. J. G. Duque et al., J. Am. Chem. Soc. 130, 2626 References
Movie S1
(2008).
34. This work was supported by the NSF Center for Biological 9 September 2010; accepted 16 November 2010
and Environmental Nanotechnology (EEC-0118007 and 10.1126/science.1197321
depressions at negative sample voltage (e.g., Fig.
1A) and could be positioned on the surface by
applying a positive voltage pulse (~+1.7 V) (fig.
Downloaded from www.sciencemag.org on December 23, 2010
S1). Fig. 1, A to C, shows STM images with VAs
at three different positions. The average distance
(~8 Å) and direction of the motion could not
always be controlled and changed with different
atomic-scale terminations of the STM tip. Thermal-
and electro-migration of VAs have been studied
previously by STM (20, 21).
The nanometer-scale depression around VAs
reflects downward band bending associated with
its +1e charge in p-GaAs (20). However, we could
reversibly switch VAs to a neutral state by apply-
ing a smaller voltage pulse (~+1.3 V). The neutral
state of VAs in Fig. 1D appears as a protrusion, in
contrast to the apparent depression in Fig. 1, A to
C. We suggest that this state results from the cap-
ture of a tunneling electron by VAs+. The neutral
Mn acceptors in GaAs. Experiments were per- state is stable indefinitely at 7 K but is readily
formed with a custom-built ultra-high vacuum
(UHV) STM operated at 7.3 K (19). The semi-
conductor sample is a commercial p-GaAs wafer
switched back to VAs+ by the STM tip under
typical imaging conditions.
We studied the influence of the electrostatic
doped with 2 × 1018 cm−3 Zn atoms for nonzero field provided by VAs on single Mn acceptors,
conductivity at low temperature. which were formed using an STM-based substi-
Arsenic vacancies, VAs, in the GaAs(110) sur- tution technique (15). Mn adatoms adsorbed onto
face formed during cleavage appeared as dark the surface at 7 K were exchanged with Ga atoms
A B E 0.8 1.42nm
2.47nm
A/V)
5.48nm
dI/dV (nA
0.4 t l
neutral
no vacancy
[-110]
0.0
[001]
0.2 0.4 0.6 0.8 1.0 1.2 1.4
C D Sample Voltage (V)
Fig. 1. Shift of Mn acceptor resonance due to VAs .
(A to D) STM topographic images of a Mn acceptor
and As vacancy in the (110) surface layer of p-GaAs
(V = –1.3 V, I = 0.5 nA). Scale bar, 1 nm. Under these
imaging conditions, the bright dumbbell-like shape
of the Mn acceptor reflects the influence on neigh-
boring As atoms (15). The Mn atomic position is
indicated with a circle in (A). Positively charged VAs appears as a dark depression. (A to C) Manipulation of VAs+
to three positions (1.42 nm, 2.47 nm, and 5.48 nm from Mn). (D) A voltage pulse switches the vacancy to a
neutral state, which is imaged as a protrusion. (E) Corresponding differential conductance (dI/dV) spectra taken
on the Mn acceptor. The in-gap resonance associated with the Mn acceptor shifts toward lower voltage as VAs+ is
moved closer. The peak shifts back to its unperturbed position when VAs is switched to the neutral state.
SCIENCE VOL 330 24 DECEMBER 2010 1807

22. T. K. Attwood, B. J. Nelmes, D. B. Sellen, Biopolymers 27,
201 (1988).
23. N. Pernodet, M. Maaloum, B. Tinland, Electrophoresis 18,
55 (1997).
24. Materials and methods are available as supporting
material on Science Online.
25. Rotational diffusion is characterized in two dimensions by
a single diffusion coefficient, Dr, and associated diffusion
time, tr = 1/2Dr.
26. The errors introduced by the limited angular resolution in
our measurements can affect the interpretation of the
short time dynamics. The microscope angular resolution
is ≈a/L, where a is the pixel size. Therefore, resolution
limits our experiments below a resolution time of
tresolution ≅ a2Lpph/2kBT. In the experimental conditions
of Fig. 3A, tresolution/td = [(l/L)(a/x)2]/4 ranges from
0.01 to 0.26; therefore, tresolution << td and resolution
Tunable Field Control Over
the Binding Energy of Single Dopants
by a Charged Vacancy in GaAs
D. H. Lee and J. A. Gupta*
Local manipulation of electric fields at the atomic scale may enable new methods for quantum
transport and creates new opportunities for field control of ferromagnetism and spin-based quantum
information processing in semiconductors. We used a scanning tunneling microscope to position
charged arsenic (As) vacancies in the gallium arsenide 110 [GaAs(110)] surface with atomic precision,
thereby tuning the local electrostatic field experienced by single manganese (Mn) acceptors. The effects
of this field are quantified by measuring the shift of an acceptor state within the band gap of GaAs.
Experiments with varying tip-induced band-bending conditions suggest a large binding energy for
surface-layer Mn, which is reduced by direct Coulomb repulsion when the As vacancy is moved nearby.
he scaling of electronic devices such as
T field-effect transistors to nanometer dimen-
sions requires more precise control of in-
dividual dopants in semiconductor nanostructures,
because statistical fluctuations in dopant distribu-
tions are beginning to affect the performance and
functionality of current devices (1–4). Proposals
for next-generation quantum- and spin-based elec-
tronics also rely on the tuning of the charge, spin,
and interactions of dopant atoms with local elec-
tric fields [e.g., P in Si (5) or Mn in GaAs (6)]. On
Si surfaces, the scanning tunneling microscope
(STM) has been used to probe the influence of
charged dangling bonds on molecular conductance
(7). In III-V semiconductors, recent STM studies
of single Si, Zn, and Mn dopants have shown that
the electronic and magnetic properties of such im-
purities depend on proximity to the surface (8–14),
other impurities (15), interactions with the STM tip
(9, 14, 16, 17), and local strain fields (18).
Here, we demonstrate control of single-dopant
properties by using the local electrostatic field
emanating from a charged vacancy. Using an
STM, we can position this vacancy with atomic
precision or reversibly switch it to a neutral state
to tune the binding energy of holes to individual
Department of Physics, Ohio State University, Columbus, OH
43210, USA.
*To whom correspondence should be addressed. E-mail:
gupta.208@osu.edu
www.sciencemag.org
REPORTS
does not substantially affect the measurement of the EEC-0647452), the Welch Foundation (grant C-1668),
subdiffisive regime due to the short-time dynamics in the Advanced Energy Consortium (www.beg.utexas.edu/
the system. aec), the Région Aquitaine, the Agence Nationale pour la
27. Y. Han et al., Science 314, 626 (2006). Recherche (ANR PNANO), the European Research Council
28. T. Munk, F. Hofling, E. Frey, T. Franosch, Europhys. Lett. (grant n 232942), and the Foundation for Fundamental
85, 30003 (2009). Research on Matter (FOM), which is part of the
29. F. Gittes, F. C. MacKintosh, Phys. Rev. E Stat. Phys. Netherlands Organisation for Scientific Research (NWO).
Plasmas Fluids Relat. Interdiscip. Topics 58, R1241
(1998).
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1804/DC1
30. E. Farge, A. C. Maggs, Macromolecules 26, 5041
(1993). Materials and Methods
Fig. S1
31. R. Granek, J. Phys. II 7, 1761 (1997).
32. M. Doi, J. Polymer Sci. Polymer Symp. 73, 93 (1985). Table S1
33. J. G. Duque et al., J. Am. Chem. Soc. 130, 2626 References
Movie S1
(2008).
34. This work was supported by the NSF Center for Biological 9 September 2010; accepted 16 November 2010
and Environmental Nanotechnology (EEC-0118007 and 10.1126/science.1197321
depressions at negative sample voltage (e.g., Fig.
1A) and could be positioned on the surface by
applying a positive voltage pulse (~+1.7 V) (fig.
Downloaded from www.sciencemag.org on December 23, 2010
S1). Fig. 1, A to C, shows STM images with VAs
at three different positions. The average distance
(~8 Å) and direction of the motion could not
always be controlled and changed with different
atomic-scale terminations of the STM tip. Thermal-
and electro-migration of VAs have been studied
previously by STM (20, 21).
The nanometer-scale depression around VAs
reflects downward band bending associated with
its +1e charge in p-GaAs (20). However, we could
reversibly switch VAs to a neutral state by apply-
ing a smaller voltage pulse (~+1.3 V). The neutral
state of VAs in Fig. 1D appears as a protrusion, in
contrast to the apparent depression in Fig. 1, A to
C. We suggest that this state results from the cap-
ture of a tunneling electron by VAs+. The neutral
Mn acceptors in GaAs. Experiments were per- state is stable indefinitely at 7 K but is readily
formed with a custom-built ultra-high vacuum
(UHV) STM operated at 7.3 K (19). The semi-
conductor sample is a commercial p-GaAs wafer
switched back to VAs+ by the STM tip under
typical imaging conditions.
We studied the influence of the electrostatic
doped with 2 × 1018 cm−3 Zn atoms for nonzero field provided by VAs on single Mn acceptors,
conductivity at low temperature. which were formed using an STM-based substi-
Arsenic vacancies, VAs, in the GaAs(110) sur- tution technique (15). Mn adatoms adsorbed onto
face formed during cleavage appeared as dark the surface at 7 K were exchanged with Ga atoms
A B E 0.8 1.42nm
2.47nm
A/V)
5.48nm
dI/dV (nA
0.4 t l
neutral
no vacancy
[-110]
0.0
[001]
0.2 0.4 0.6 0.8 1.0 1.2 1.4
C D Sample Voltage (V)
Fig. 1. Shift of Mn acceptor resonance due to VAs .
(A to D) STM topographic images of a Mn acceptor
and As vacancy in the (110) surface layer of p-GaAs
(V = –1.3 V, I = 0.5 nA). Scale bar, 1 nm. Under these
imaging conditions, the bright dumbbell-like shape
of the Mn acceptor reflects the influence on neigh-
boring As atoms (15). The Mn atomic position is
indicated with a circle in (A). Positively charged VAs appears as a dark depression. (A to C) Manipulation of VAs+
to three positions (1.42 nm, 2.47 nm, and 5.48 nm from Mn). (D) A voltage pulse switches the vacancy to a
neutral state, which is imaged as a protrusion. (E) Corresponding differential conductance (dI/dV) spectra taken
on the Mn acceptor. The in-gap resonance associated with the Mn acceptor shifts toward lower voltage as VAs+ is
moved closer. The peak shifts back to its unperturbed position when VAs is switched to the neutral state.
SCIENCE VOL 330 24 DECEMBER 2010 1807
 REPORTS
in the surface layer by applying a positive voltage Ag W* Ni Ir
pulse with the STM tip (~+1.3 V). Figure 1A A B
1nA

dI/dV (nA/V)
shows a Mn acceptor formed in this way; under 0.8 surface
data 1 0.5nA
these conditions, the acceptor is imaged as a r=100nm, θ =160
o Mn 0.1nA
r=1nm, θ =20
o ×3

Peak Position (V)

bright, dumbbell-like feature. Tunneling spectra 0.4
0
taken with the tip positioned over Mn (Fig. 1E)
show a resonance peak within the GaAs band 0.0 0.5nA
sub-surface

dI/dV (nA/V)
0.3nA
gap, which is associated with the Mn acceptor- θ 2 0.25nA Zn
hole complex (15). The peak systematically shifts -0.4 0.1nA
r 0.05nA
toward lower voltage as VAs is moved closer and
returns to its unperturbed position when VAs is -0.8
z 0
switched to the neutral state (Fig. 1E, red). 0.6 0.8 1.0 1.2

To understand this peak shift, we must consider 4.0 4.5 5.0 5.5 Sample Voltage (V)
two possible influences of VAs on acceptor-hole Tip Work function (eV)
complexes: band bending and direct Coulomb re-
pulsion. Vacancy-induced band bending can be Fig. 2. Insensitivity of the Mn peak to varying TIBB. (A) Black dots indicate measured in-gap peak positions
understood in analogy to tip-induced band bend- of surface-layer Mn with four different tip materials [*, W-tip data from (15)]. Colored points represent peak
ing (TIBB). When brought into tunneling range positions predicted from the rigid band-bending model, based on simulated TIBB with varying work

Downloaded from www.sciencemag.org on December 23, 2010

less than 1 nanometer from the surface, the STM
tip affects the local carrier distribution and locally
bends the semiconductor bands by an amount that
depends on the voltage, tip work function, and ge-
ometric factors such as tip radius and shank angle
(22). TIBB has been invoked in previous STM
studies to explain why the in-gap resonances of
function, tip apex radius r, and shank angle q. The shaded region indicates the broad range of predicted
peak positions expected in the experiments. (B) Comparison of tunneling spectra with varying set current
for surface-layer Mn and subsurface Zn. The resonance peak of Zn shifts according to the rigid band-
bending model, whereas no shift is observed for Mn. The dashed green lines are guides to the eye.
A B
data
0.0 Coulomb fit
n resonance peak (V)

subsurface impurities such as Mn and Zn in III-V 0.8

semiconductors (8, 13, 23–25) occur at voltages
not expected from the bulk binding energies. The 25

∆V (V)
acceptor state is thought to rigidly follow the va- -0.1 20
Tip

Counts
lence band, whose bending changes as the voltage 0.7 15
is varied in tunneling spectroscopy. A resonance 10
Mn + VAs 5
peak is produced when the acceptor state crosses - - -0.2 0
Mn

the sample Fermi energy. The peak’s position does
not necessarily locate the acceptor energy level
and is sensitive to TIBB conditions. For example,
tunneling spectroscopy of subsurface Mn in InAs
indicates a peak at ~0.8 V, even though the ac-
ceptor level in the bulk lies only ~28 meV above
the valence band (8, 16). This peak shifts by ~0.1 V
with varying TIBB conditions (16). We will refer
to this interpretation as a “rigid band-bending
model” of the impurity resonances (fig. S2A).
In our experiments, TIBB is negative (i.e.,
downward band bending) for all voltages less than
the flat-band voltage (fig. S4A). Similar to TIBB,
VAs+ causes an additional downward band bend-
ing (20). As a result, the crossing of acceptor level
and sample Fermi energy should occur at a larger
positive voltage. The rigid band-bending model
therefore predicts that the Mn resonance should
shift toward higher positive voltage as VAs is moved
closer, in contrast to the shift toward lower volt-
age shown in Fig. 1E. This contradiction was
initially surprising and led us to hypothesize that
the rigid band-bending model holds for subsur-
face impurities (e.g., native Zn acceptors) (fig. S3)
but does not hold for surface-layer Mn in GaAs.
This hypothesis is consistent with our obser-
vation that the resonance peak for surface-layer
Mn does not shift with varying TIBB conditions
[e.g., tip material, tip termination, and tip-sample
distance (26)]. Figure 2A shows the Mn peak
positions as measured with four different tip
materials whose bulk work function ranges from
4.2 eV (Ag) to 5.2 eV (Ir). Within experimental
Zn - GaAs 0.75 0.80 0.85
Peak position (eV)
0.6
0 2 4 6 8 0 2 4 6 8 10
Distance (nm) Distance (nm)
Fig. 3. Peak shift versus distance to VAs. (A) The peak positions of five different Mn acceptors are plotted as a
function of distance d to VAs. The solid lines are Coulomb fits with the same coefficient but with different
offsets [i.e., V(∞)]. This variation reflects the different local electrostatic environments due to –1e charged Zn
neighbors (inset). (B) Data from 22 Mn acceptors collapse to a single curve after the offset corrections. The
data are well fit with a Coulomb-like law, DV = A/d, where the fit parameter A = –0.22 V nm. The inset shows
statistics of 106 Mn acceptors isolated from VAs. From the 1.65 s point of the normal distribution, we chose a
value, V0 = 0.74 V, as the binding energy of an isolated Mn-hole complex (green arrow).
error (attributed to variations in local environ- from surface-layer Mn and subsurface Zn over a
ment, as discussed further below), the Mn peak range in set current. Consistent with the rigid band-
appeared at ~0.8 V regardless of tip material. bending model and previous studies (23, 25), sub-
For comparison, we performed three-dimensional surface Zn exhibits resonance peaks that shifted
simulations of TIBB using Poisson’s equation toward higher voltage (by ~ 0.13 V) as current was
(22), which then allowed us to calculate the peak increased. Surface-layer Mn, however, showed
positions predicted by the rigid band-bending little if any shift (0.008 T 0.008 V), again suggest-
model. In this model, the predicted peak position ing that such impurities do not respond to TIBB.
corresponds to the bias voltage at which the Mn A similar insensitivity to varying TIBB con-
bulk acceptor level [0.11 eV (27)] crosses the ditions was observed for Fe adatoms on InAs
Fermi energy of the sample (fig. S4A). We used (30), presumably reflecting the stronger localiza-
bulk work function values of the tip materials tion of adatom states compared with subsurface
(28, 29) and considered both sharp and blunt tip impurities that hybridize with the host lattice. Our
terminations. The yellow region in Fig. 2A indi- studies suggest that surface-layer Mn exhibits in-
cates that a broad range of peak positions is ex- termediate characteristics between bulk acceptor
pected, in contrast to our observations. and adatom. Whereas STM images reveal a hole-
To further explore this issue, we systematically acceptor complex whose asymmetry reflects the
varied TIBB by varying the tunneling set current GaAs crystal structure (15, 31, 32), the acceptor en-
for spectroscopy. Higher set current corresponds ergy levels are decoupled from the valence band
to a smaller tip-sample distance and, thus, larger edge, a characteristic of so-called “deep” impurities.
TIBB. Figure 2B compares tunneling spectra taken Recent tight-binding model calculations support

1808 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


A B
D 1.0
three vacancies
two vacancies
one vacancy
dI/dV (nA/V)
no vacancy
0.5
0.0
0.6 0.8 1.0 1.2
Sample Voltage (V)
Fig. 4. Multiple vacancies and a charged adatom. (A to C) STM topographic images of a Mn acceptor with
multiple As vacancies: (A) three, (B) two, and (C) one VAs. (D) Corresponding dI/dV spectra taken on the Mn
acceptor show that the electrostatic potential can be enhanced with multiple VAs. (E to G) STM images and dI/dV
spectra of a Mn acceptor with a Ga adatom. The peak shift of the Mn acceptor suggests that the Ga adatom is
positively charged (V = –1.3 V, I = 0.5 nA). Scale bar, 1 nm.
this suggestion that surface-layer Mn is a deep
acceptor with a more localized wave function
compared with the bulk (32).
Given that the acceptor level appears to be
independent of TIBB, we reconsidered the phys-
ical interpretation of the 0.8 V peak for surface-
layer Mn. In the absence of any evidence for
Fermi-level pinning (15, 33), the peak’s separation
from the valence band edge at ~0 V directly corre-
sponds to the acceptor-hole binding energy. This
value, ~0.8 eV, is much greater than the bulk value
of 0.11 eV but is consistent with theoretical predic-
tions for surface-layer dopant impurities such as
Mn (32) and Si (34). For example, Strandberg et al.
estimate that the binding energy for surface-layer
Mn acceptors is an order of magnitude greater than
the bulk (32). These predictions are consistent with
recent STM experiments in GaAs, which show an
increase in the binding energy of subsurface Mn
(13) and Si (14) with proximity to the surface.
Underlying mechanisms for an increased binding
energy at or near the surface include surface
effects such as strain fields induced by relaxation
(9, 12, 14), a reduction in the effective dielectric
constant at the vacuum interface (14), or dangling-
bond states (33, 34).
Because the peak shift in Fig. 1E could not be
caused by vacancy-induced band bending, its di-
rection suggests that VAs reduces the binding energy
of surface Mn by the direct Coulomb repulsion
between the positively charged hole and VAs+. The
change in binding energy, DE, can be expressed
within an effective model Hamiltonian, which in-
cludes a Coulomb potential from VAs
e2
H ¼ H0 þ
4pe0 eeff d
ð1Þ
e2
DE ¼ 〈yMn j jy 〉
4pe0 eeff d Mn
where H0 is the unperturbed Hamiltonian, d is the
distance between Mn and VAs, eeff is an effective
www.sciencemag.org
C E F 4, E to G, shows the peak shift produced by a
Ga
G 1.2nm
1.5 no Ga adatom
dI/dV (nA/V)
1.0
0.5
d
00
0.0
0.4 0.6 0.8 1.0
Sample Voltage (V)
dielectric constant at the surface, and yMn is
Mn wave function. A simple hydrogenic model
calculation with 1s wave functions predicts a re-
duction of the binding energy with a Coulomb-
like 1/d dependence when a positive point charge
is moved nearby (35).
To test this prediction, Fig. 3A shows the peak
positions of five different Mn acceptors as a
function of distance. These data are well fit with
the same Coulomb curve but are offset from each
other by tens of meV. These offsets result from
variations in the local electrostatic environment
created by the native Zn acceptors that are nega-
tively charged under our tunneling conditions
(Fig. 3A, inset). We verified that the resonance
peak shifts toward higher voltage for Mn accep-
tors near Zn acceptors, opposite to the shift due to
VAs+. Thus, the immobile Zn acceptors contrib-
uted a finite positive offset to the peak position,
whose value was different for each Mn acceptor.
We define the peak shift as DV(d) = V(d ) –
V0, where V0 is the binding energy of a Mn-hole
complex (isolated from both Zn and VAs) and
V(d) is the resonance peak position, corrected for
variations in local environment. From statistics of
106 Mn acceptors, we chose a value V0 = 0.74 V,
which represents the 1.65 s point of the normal
distribution of resonance peak positions (Fig. 3B,
inset). By measuring V(~ ∞), we then calculated
the offset for each Mn. After this correction, data
from 22 different Mn were fit using a single
Coulomb-like curve (Fig. 3B, red line).
These experiments can be extended in sev-
eral ways to characterize and engineer local
electrostatic fields at the nanometer scale. For
example, larger effects can be achieved by in-
troducing additional vacancies (Fig. 4, A to D).
We find that the electrostatic potential experi-
enced by a Mn acceptor is the superposition of
the individual potentials from each VAs. We have
also achieved similar field control using other
charged species such as adatoms (Ga, Mn, or
SCIENCE VOL 330 24 DECEMBER 2010
REPORTS
Ag) or native dopants (Zn). For example, Fig.
positively charged Ga adatom, which was posi-
tioned 1.2 nm from a Mn acceptor using atom
manipulation.
These data suggest a new and direct method
for quantifying the charge of adsorbates (e.g.,
adatoms or molecules) as well as defects (e.g.,
vacancies, antisites, and interstitials) at semicon-
ductor surfaces. We also anticipate that manip-
ulation of charged species can be used to tune
the interactions between pairs of dopants, which
may provide insight into the mechanisms for
ferromagnetism in semiconductors (15, 36), and
new methods for quantum information process-
1.2 ing in solids (5, 6). Although the experiments
demonstrated here required an STM tip to po-
sition the As vacancies, a similar control may also
Downloaded from www.sciencemag.org on December 23, 2010
be achieved in future devices using conventional
electrodes (4).
References and Notes
1. P. S. Peercy, Nature 406, 1023 (2000).
2. T. Shinada, S. Okamoto, T. Kobayashi, I. Ohdomari,
Nature 437, 1128 (2005).
3. G. P. Lansbergen et al., Nat. Phys. 4, 656 (2008).
4. J. J. Yang et al., Nat. Nanotechnol. 3, 429 (2008).
5. B. E. Kane, Nature 393, 133 (1998).
6. J. M. Tang, J. Levy, M. E. Flatté, Phys. Rev. Lett. 97,
106803 (2006).
7. P. G. Piva et al., Nature 435, 658 (2005).
8. F. Marczinowski et al., Phys. Rev. Lett. 99, 157202
(2007).
9. S. Loth, M. Wenderoth, R. G. Ulbrich, Phys. Rev. B 77,
115344 (2008).
10. J. K. Garleff et al., Phys. Rev. B 78, 075313 (2008).
11. J. M. Jancu et al., Phys. Rev. Lett. 101, 196801 (2008).
12. C. C. Celebi et al., Phys. Rev. Lett. 104, 086404 (2010).
13. J. K. Garleff et al., Phys. Rev. B 82, 035303 (2010).
14. A. P. Wijnheijmer et al., Phys. Rev. Lett. 102, 166101
(2009).
15. D. Kitchen, A. Richardella, J. M. Tang, M. E. Flatté,
A. Yazdani, Nature 442, 436 (2006).
16. F. Marczinowski, J. Wiebe, F. Meier, K. Hashimoto,
R. Wiesendanger, Phys. Rev. B 77, 115318 (2008).
17. K. Teichmann et al., Phys. Rev. Lett. 101, 076103 (2008).
18. A. M. Yakunin et al., Nat. Mater. 6, 512 (2007).
19. Materials and methods are available as supporting
material on Science Online.
20. P. Ebert, Surf. Sci. Rep. 33, 121 (1999).
21. G. Lengel, J. Harper, M. Weimer, Phys. Rev. Lett. 76,
4725 (1996).
22. R. M. Feenstra, S. Gaan, G. Meyer, K. H. Rieder,
Phys. Rev. B 71, 125316 (2005).
23. S. Loth, M. Wenderoth, R. G. Ulbrich, S. Malzer,
G. H. Dohler, Phys. Rev. B 76, 235318 (2007).
24. A. M. Yakunin et al., Phys. Rev. Lett. 92, 216806 (2004).
25. S. Loth et al., Phys. Rev. Lett. 96, 066403 (2006).
26. D. Kitchen et al., J. Appl. Phys. 101, 09G515 (2007).
27. M. Linnarsson, E. Janzen, B. Monemar, M. Kleverman,
A. Thilderkvist, Phys. Rev. B 55, 6938 (1997).
28. H. B. Michaelson, J. Appl. Phys. 48, 4729 (1977).
29. Because the work function varies with different crystal
orientations, the work function of a sharp-ended tip may
differ from the bulk value.
30. T. Matsui, C. Meyer, L. Sacharow, J. Wiebe,
R. Wiesendanger, Phys. Rev. B 75, 165405 (2007).
31. J. M. Tang, M. E. Flatté, Phys. Rev. Lett. 92, 047201
(2004).
32. T. O. Strandberg, C. M. Canali, A. H. MacDonald,
Phys. Rev. B 80, 024425 (2009).
33. A. Richardella, D. Kitchen, A. Yazdani, Phys. Rev. B 80,
045318 (2009).
34. J. Wang, T. A. Arias, J. D. Joannopoulos, G. W. Turner,
O. L. Alerhand, Phys. Rev. B 47, 10326 (1993).
1809
 REPORTS
35. P. D. Robinson, Proc. Phys. Soc. 71, 828 (1958). Mabel Beckman Foundation and the Center for Emergent SOM Text
36. J. M. D. Coey, S. A. Chambers, MRS Bull. 33, 1053 Materials at Ohio State University, an NSF Materials Figs. S1 to S4
(2008). Research Science and Engineering Center (DMR-0820414). References
37. We thank D. R. Daughton and X. H. Qiu for help with
the STM construction and A. J. Heinrich, S. Loth, Supporting Online Material 6 September 2010; accepted 29 November 2010
M. E. Flatté, and A. H. MacDonald for helpful discussions. www.sciencemag.org/cgi/content/full/science.1197434/DC1 Published online 9 December 2010;
We are grateful for support from the Arnold and Materials and Methods 10.1126/science.1197434

DWs should exhibit inertial effects when driven

Dynamics of Magnetic Domain Walls by current (14–18). Inertial effects have indeed
been reported, but only when DWs are excited
Under Their Own Inertia while confined at a trapping site (3, 20, 25–28).
In contrast, recent reports of current-driven prop-
Luc Thomas,* Rai Moriya, Charles Rettner, Stuart S. P. Parkin* agation of DWs over long distances of several
micrometers indicate that the distance traveled by
The motion of magnetic domain walls induced by spin-polarized current has considerable potential DWs in response to current pulses varies linearly
for use in magnetic memory and logic devices. Key to the success of these devices is the precise with the length of the pulse (5, 13). This would
positioning of individual domain walls along magnetic nanowires, using current pulses. We show that seemingly indicate that the DWs move steadily at
domain walls move surprisingly long distances of several micrometers and relax over several tens a fixed velocity without any inertia.

Downloaded from www.sciencemag.org on December 23, 2010

of nanoseconds, under their own inertia, when the current stimulus is removed. We also show that the Our experiments were carried out using 20-
net distance traveled by the domain wall is exactly proportional to the current pulse length because nm-thick permalloy (Ni81Fe19) nanowires with
of the lag derived from its acceleration at the onset of the pulse. Thus, independent of its inertia, widths of 200 nm and lengths between 6 and 15 mm.
a domain wall can be accurately positioned using properly timed current pulses. The devices were composed of the magnetic
nanowire and two electrical contact lines, which
lectrical current passing through a magnetic retical (14–18) studies. However, many aspects were used to write, shift, and detect DWs (Fig.

E material becomes spin-polarized along the

local magnetization direction. When the cur-
rent traverses a magnetic domain wall (DW), spin
angular momentum is transferred from the current
to the magnetization, thereby inducing a torque on
the DW and leading to DW motion (1, 2). Such
spin-transfer torque (STT)–driven DW motion has
distinct characteristics that make it very useful for
magnetic memory-storage devices (3). In particular,
two or more adjacent DWs can be moved in the
same direction, contrary to the case when DWs are
driven by a magnetic field. Advances in our under-
standing of current-driven DW dynamics have
resulted from various experimental (4–13) and theo-
IBM Almaden Research Center, 650 Harry Road, San Jose,
CA, USA.
*To whom correspondence should be addressed. E-mail:
lucthom@us.ibm.com (L.T.); parkin@almaden.ibm.com (S.S.P.P.)
of the underlying physical mechanisms remain
unclear. An important question, from both fun-
damental and technological standpoints, is whether
DWs, driven solely by current, exhibit inertial ef-
fects similar to those observed when they are driven
by a magnetic field (19).
Two contributions to STT have been identi-
fied: the adiabatic and nonadiabatic (field-like)
contributions (14–18). The inertial response of the
DW depends on the relative magnitude of these
two terms. Their relative contributions can be
quantified by the ratio b/a, where b and a are
dimensionless constants that reflect the strengths
of the nonadiabatic STT and the Gilbert damping,
respectively. Although there is still considerable
debate as to the precise origin and value of b/a,
many experimental studies have concluded that
b/a > 1 for various magnetic materials (10, 20–24).
Under these circumstances, theory predicts that
1A) (29). A DW was first written into the nanowire
by applying a burst of current pulses through the
injection line (injection pulse), and the dc resist-
ance of the nanowire was then measured. A cur-
rent pulse [shift pulse (30)] was then applied
through the nanowire to move the DW along the
nanowire, and the resistance was measured for a
second time. This procedure was repeated 100
times under identical conditions, except that the
sign of the injection current was reversed in suc-
cessive experiments so as to write, alternately,
head-to-head (HH) and tail-to-tail (TT) DWs. The
shift pulse polarity was maintained unchanged,
so that electrons always flowed in the same di-
rection along the nanowire (from right to left in
Fig. 1A), and, through STT, drove the DWs, wheth-
er HH or TT, in the same direction as the electron
flow (5, 13). The shift pulse voltage was chosen
to give a current density in the nanowire of ~1.2 ×

Fig. 1. (A) Scanning electron micrograph image of a 12-mm-long A C

permalloy nanowire with electrical contact lines at the left and right
ends of the nanowire, shown in beige. Schematics of the electrical
connections and the injection (red) and shift (blue) current pulses
are shown. The blue arrow indicates the direction of the electron
flow. The shapes of the ends of the nanowire, which are hidden
under the contact lines, are shown schematically in the insets at top
left and bottom right. The dc resistance is measured through the
low-frequency port of a bias tee (bottom left contact), and the shift
pulse is applied through the high-frequency port of the bias tee. (B)
Probability Pout of a DW exiting a 6-mm-long nanowire after a shift
pulse is applied. (C) Shift pulse length tout required to move a DW B D
out of the nanowire with a 50% probability, as a function of the
nanowire length. Error bars corresponding to 20/80% probabilities
are shown. (they are about the same size as the symbols). (D) tout as
a function of the inverse number of shift current pulses Np, for two
nanowires with lengths of 6 and 12 mm (open and solid symbols,
respectively). Error bars show 20/80% probabilities.

1810 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

 REPORTS
35. P. D. Robinson, Proc. Phys. Soc. 71, 828 (1958). Mabel Beckman Foundation and the Center for Emergent SOM Text
36. J. M. D. Coey, S. A. Chambers, MRS Bull. 33, 1053 Materials at Ohio State University, an NSF Materials Figs. S1 to S4
(2008). Research Science and Engineering Center (DMR-0820414). References
37. We thank D. R. Daughton and X. H. Qiu for help with
the STM construction and A. J. Heinrich, S. Loth, Supporting Online Material 6 September 2010; accepted 29 November 2010
M. E. Flatté, and A. H. MacDonald for helpful discussions. www.sciencemag.org/cgi/content/full/science.1197434/DC1 Published online 9 December 2010;
We are grateful for support from the Arnold and Materials and Methods 10.1126/science.1197434

DWs should exhibit inertial effects when driven

Dynamics of Magnetic Domain Walls by current (14–18). Inertial effects have indeed
been reported, but only when DWs are excited
Under Their Own Inertia while confined at a trapping site (3, 20, 25–28).
In contrast, recent reports of current-driven prop-
Luc Thomas,* Rai Moriya, Charles Rettner, Stuart S. P. Parkin* agation of DWs over long distances of several
micrometers indicate that the distance traveled by
The motion of magnetic domain walls induced by spin-polarized current has considerable potential DWs in response to current pulses varies linearly
for use in magnetic memory and logic devices. Key to the success of these devices is the precise with the length of the pulse (5, 13). This would
positioning of individual domain walls along magnetic nanowires, using current pulses. We show that seemingly indicate that the DWs move steadily at
domain walls move surprisingly long distances of several micrometers and relax over several tens a fixed velocity without any inertia.

Downloaded from www.sciencemag.org on December 23, 2010

of nanoseconds, under their own inertia, when the current stimulus is removed. We also show that the Our experiments were carried out using 20-
net distance traveled by the domain wall is exactly proportional to the current pulse length because nm-thick permalloy (Ni81Fe19) nanowires with
of the lag derived from its acceleration at the onset of the pulse. Thus, independent of its inertia, widths of 200 nm and lengths between 6 and 15 mm.
a domain wall can be accurately positioned using properly timed current pulses. The devices were composed of the magnetic
nanowire and two electrical contact lines, which
lectrical current passing through a magnetic retical (14–18) studies. However, many aspects were used to write, shift, and detect DWs (Fig.

E material becomes spin-polarized along the

local magnetization direction. When the cur-
rent traverses a magnetic domain wall (DW), spin
angular momentum is transferred from the current
to the magnetization, thereby inducing a torque on
the DW and leading to DW motion (1, 2). Such
spin-transfer torque (STT)–driven DW motion has
distinct characteristics that make it very useful for
magnetic memory-storage devices (3). In particular,
two or more adjacent DWs can be moved in the
same direction, contrary to the case when DWs are
driven by a magnetic field. Advances in our under-
standing of current-driven DW dynamics have
resulted from various experimental (4–13) and theo-
IBM Almaden Research Center, 650 Harry Road, San Jose,
CA, USA.
*To whom correspondence should be addressed. E-mail:
lucthom@us.ibm.com (L.T.); parkin@almaden.ibm.com (S.S.P.P.)
of the underlying physical mechanisms remain
unclear. An important question, from both fun-
damental and technological standpoints, is whether
DWs, driven solely by current, exhibit inertial ef-
fects similar to those observed when they are driven
by a magnetic field (19).
Two contributions to STT have been identi-
fied: the adiabatic and nonadiabatic (field-like)
contributions (14–18). The inertial response of the
DW depends on the relative magnitude of these
two terms. Their relative contributions can be
quantified by the ratio b/a, where b and a are
dimensionless constants that reflect the strengths
of the nonadiabatic STT and the Gilbert damping,
respectively. Although there is still considerable
debate as to the precise origin and value of b/a,
many experimental studies have concluded that
b/a > 1 for various magnetic materials (10, 20–24).
Under these circumstances, theory predicts that
1A) (29). A DW was first written into the nanowire
by applying a burst of current pulses through the
injection line (injection pulse), and the dc resist-
ance of the nanowire was then measured. A cur-
rent pulse [shift pulse (30)] was then applied
through the nanowire to move the DW along the
nanowire, and the resistance was measured for a
second time. This procedure was repeated 100
times under identical conditions, except that the
sign of the injection current was reversed in suc-
cessive experiments so as to write, alternately,
head-to-head (HH) and tail-to-tail (TT) DWs. The
shift pulse polarity was maintained unchanged,
so that electrons always flowed in the same di-
rection along the nanowire (from right to left in
Fig. 1A), and, through STT, drove the DWs, wheth-
er HH or TT, in the same direction as the electron
flow (5, 13). The shift pulse voltage was chosen
to give a current density in the nanowire of ~1.2 ×

Fig. 1. (A) Scanning electron micrograph image of a 12-mm-long A C

permalloy nanowire with electrical contact lines at the left and right
ends of the nanowire, shown in beige. Schematics of the electrical
connections and the injection (red) and shift (blue) current pulses
are shown. The blue arrow indicates the direction of the electron
flow. The shapes of the ends of the nanowire, which are hidden
under the contact lines, are shown schematically in the insets at top
left and bottom right. The dc resistance is measured through the
low-frequency port of a bias tee (bottom left contact), and the shift
pulse is applied through the high-frequency port of the bias tee. (B)
Probability Pout of a DW exiting a 6-mm-long nanowire after a shift
pulse is applied. (C) Shift pulse length tout required to move a DW B D
out of the nanowire with a 50% probability, as a function of the
nanowire length. Error bars corresponding to 20/80% probabilities
are shown. (they are about the same size as the symbols). (D) tout as
a function of the inverse number of shift current pulses Np, for two
nanowires with lengths of 6 and 12 mm (open and solid symbols,
respectively). Error bars show 20/80% probabilities.

1810 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


108 A/cm2. The presence of one or more DWs in
the nanowire between the contacts was inferred
from the dc resistance of the nanowire. Owing to
the anisotropic magnetoresistance of permalloy,
the nanowire resistance was reduced by a fixed
amount of ~180 milliohm for each additional DW
located in the nanowire (31). In these experiments,
we focused on DWs with a vortex structure (29).
The probability that a single DW, located in
the nanowire after the injection pulse, exits the
nanowire after a shift pulse of length tsh exhibits a
clear threshold in tsh above which almost all the
DWs exit (Fig. 1B). We define tout as the value of
tsh corresponding to an exit probability of 50%.
tout increases linearly with the nanowire’s length,
with a slope corresponding to a velocity of ~138 m/s
(Fig. 1C). The small offset when tout = 0 is due to
the distance traveled by the DW along the nano-
wire from its injection point during its creation (13).
This first set of experiments indicates that
the DW propagates at constant velocity along the
nanowire. One possibility, however, is that the
current pulses used are long compared to the time
scale of any inertial effects, so in a second set of
experiments we explored much shorter shift pulse
lengths. We used a train of up to eight shift pulses
to move the DWs, instead of one long pulse. The
interval between these pulses was set to be ~6
times their length. Figure 1D shows that tout varies
as the inverse number of pulses 1/Np for 6-mm-
A B
Fig. 2. (A) Schematic of the shift pulse sequence
applied to the nanowire. Two pulses of opposite
polarities are applied sequentially, separated by a
waiting time twait during which the current is zero.
Cartoons show the DW motion during this pulse
sequence. Different colors are used to highlight the
motion of the DW during the two current pulses (blue
and red) and under its own inertia (green). (B) Prob-
ability of DW exiting the nanowire as a function of the
shift pulse length tsh, for various waiting times twait
(from left to right, twait = 25, 12, 7, and 0 ns), for a
6-mm-long nanowire. (C) The shift pulse length tout for
which Pout = 50% versus twait for 6-mm-long (green diamonds), 12-mm-long (red circles), and 15-mm-long
(blue squares) nanowires. Error bars show 20/80% probabilities. Solid lines show fits to an exponential decay.
www.sciencemag.org
long (open symbols) and 12-mm-long (solid sym-
bols) nanowires (here tout is the length of just one
of the shift pulses). This means that the overall
time that the current needs to be applied for the
DW to traverse the nanowire remains the same,
irrespective of the length of the individual pulses.
This again shows that the distance traveled by the
DW during a single pulse is directly proportional
to the pulse length, even for current pulses as
short as a few nanoseconds.
Τhe value of b/a can be directly derived from
the DW’s steady-state velocity, which is given by
v = (b/a)u, where u has the dimension of a ve-
locity and is given by u = (mB/eMs)PJ, where mB
is the Bohr magneton, e is the electron charge, Ms
is the saturation magnetization (~800 electro-
magnetic units/cm3 for permalloy), P is the spin
polarization of the current, and J is the current
density (16, 17). For a DW velocity of ~138 m/s
for J ~ 1.2 × 108 A/cm2, and assuming P = 0.5, as
reported recently in permalloy wires of similar
thicknesses (32), we find that b/a ~ 3.2. Thus,
because b/a is much larger than 1, the DWs should
theoretically display inertial behavior.
Because of the very small resistance of the
DW, the dynamics of DWs in the nanowires were
probed by their presence or absence using quasi-
static resistance measurements, which take much
longer than the typical time scale of the DW dy-
namics. Thus, to detect any possible inertial mo-
1 tance is calculated from the extra time needed to
0 location along the nanowire, using a current pulse
25 35 45 55
C 120
100
80
60
40
20
0 20 40 60 80 100
SCIENCE VOL 330 24 DECEMBER 2010
REPORTS
tion of the DWs shortly after the end of the current
pulse, we used a second shift current pulse. This
pulse, with opposite polarity from the first, was
applied after a waiting time twait, which was varied
on a nanosecond time scale (Fig. 2A). If the DW
had already exited the nanowire when the second
pulse was applied, the DW exit probability Pout
was not affected. However, if the DW was still
located within the nanowire at the onset of the
second pulse (Fig. 2A), because the current flow
was now reversed, the DW would be pushed
back into the nanowire, thereby modifying Pout.
Figure 2B shows the dependence of Pout on the
shift pulse length tsh for several values of twait for
a 6-mm-long nanowire. As twait was decreased
from 25 to 0 ns, the length of the shift pulse re-
quired for the DW to exit the nanowire signif-
icantly increased, from ~32 T 2 to 42 T 2 ns. For
Downloaded from www.sciencemag.org on December 23, 2010
three nanowires with different lengths, tout de-
creased as twait was increased, until twait ~ 25 T
5 ns, above which tout was approximately constant
(Fig. 2C). Thus, if no time was allowed for the
DW to move after the end of the shift pulse, a
longer shift pulse was needed to drive the DW
out of the nanowire, indicating that the DW kept
on moving after the end of the shift pulse while
decelerating to zero velocity. The dependence of
tout on twait can be fitted to an exponential form
exp(–twait/t), giving a deceleration time of t
~11.5 T 2 ns, which is independent of nanowire
length. During this deceleration period, we esti-
mate that the DW moved ~1.4 T 0.6 mm. This dis-
move the DW along the length of the nanowire
when twait = 0 (Dtout ~10 T 4 ns), during which the
DW is moving at its terminal velocity (v = 138 m/s).
To quantify the distance over which the DW
accelerated, we created a virtual nanowire of var-
iable length by pre-positioning the DW at a given
of length tpos (Fig. 3). The longer the preposition-
ing time is, the shorter is the distance dx between
the DW’s new position and the exit point (at the
left contact). Then, after waiting ~100 ms to ensure
that the DW was completely relaxed, a first current
pulse was applied, followed by a second current
pulse of the opposite polarity, separated by a
waiting time. Here we consider just the two cases
where twait = 0 and 100 ns, which correspond to
either zero or a full contribution of the DW’s
motion during its after-pulse deceleration. When
twait = 0, we find that tout initially increases rap-
idly as dx is increased from zero before reaching
a linear regime. We attribute this dependence to
the initial acceleration of the DW toward its
terminal velocity, given by the slope in the linear
regime. By fitting the data with an exponential
function (29), we find an acceleration time tacc =
13.3 T 4 ns, which is comparable to the de-
celeration time found above. In contrast, when
twait = 100 ns, tout varies linearly with dx for all dx,
with the same slope as for twait = 0 for large dx. In
this case, where both acceleration and decelera-
tion contribute, we find that the effective DW
velocity is independent of distance traveled, which
1811
 REPORTS
is consistent with the first set of results shown in turned off exactly makes up for the distance distance over which the DW accelerates to 90%
Fig. 1. Thus, the additional distance that the DW lagged during acceleration to its terminal veloc- of its terminal velocity is ~2.5 mm, whereas the
travels during deceleration after the current is ity. From the data in Fig. 3, we estimate that the distance moved by the DW during deceleration is
~1.45 mm (shown by the horizontal dashed line in
Fig. 3. Shift pulse length Fig. 3). The latter is in close agreement with the
tout for which Pout = 50% data described earlier.
versus the length of a To support our experimental findings, we an-
current pulse tpos used alyzed the response of a DW to a current pulse
to position the DW at a using the well-known one-dimensional (1D)
distance dx from the exit model of DW dynamics (15, 17, 19, 20, 29).
point of the 12-mm-long
Figure 4A and B show the calculated position
nanowire. Open and sol-
and instantaneous velocity of a DW in response
id symbols correspond to
twait = 0 and 100 ns, re- to a 100-ns-long current pulse, for two different
spectively. Solid lines show values of b/a (1 and 3.2) but the same terminal
fits to the data as dis- velocity (b/a)u = 138 m/s. Clearly, the DW’s
cussed in the text. Error inertial response to the current pulse is strongly
bars show 20/80% prob- influenced by the value of b/a, even though the
total distance traveled by the DW is the same in

Downloaded from www.sciencemag.org on December 23, 2010

abilities. The inset shows
a diagram of the nano- both cases. The DW’s relaxation after the pulse
wire with the DW posi- exactly offsets the acceleration at the beginning
tioned at a distance dx from the exit. The offset in dx between the two curves (dotted line) represents the of the pulse (15), in agreement with our expe-
distance moved by the DW during its deceleration after the end of the current pulse. riments. In particular, the effective DW velocity
(the total displacement divided by the pulse length)
B is equal to the DW’s terminal velocity in both
A
cases. Deceleration also offsets acceleration even
if the current pulses are shorter than the DW’s
acceleration time, as in the experiments of Fig.
1D. The effect of a bipolar pulse is shown in Fig.
4, C and D, for 100-ns-long pulses and different
waiting times twait between the two pulses. We
find that in all cases, the DW comes back to its
starting position, but as shown in Fig. 4D, its
excursion is greater when the waiting time
C D between the pulses is long. Analytical expres-
sions of the time-dependent DW velocity can be
derived within a linear approximation that is val-
id for small currents, allowing us to calculate the
distance lagged due to acceleration (which is the
same as the deceleration distance). This distance
is ~1.1 mm, which is slightly smaller than the
value found experimentally.
The 1D model does not take into account the
structure of the vortex DW used in our experiments.
E F To confirm our findings and enable a more quan-
titative description, we used micromagnetic sim-
ulations to calculate the response of a vortex DW
to a current step (Fig. 4E). Snapshots of the DW
structure during motion show that the acceleration
is associated with the lateral motion of the vortex
core (Fig. 4F). The DW velocity versus time is very
well described by the analytical model. The Gilbert
damping a was varied to match the experimental
value of t (29). Best agreement was found for a =
Fig. 4. (A) DW position and (B) velocity versus time in response to a 100-ns-long current pulse, calculated 0.008 T 0.002, which leads to b ~ 0.026. The dis-
using the 1D model, for the same steady-state velocity, v = (b/a)u = 138 m/s, but two different b/a ratios:
tance lagged due to acceleration was ~1.0 mm, which
b/a = 1 and u = 138 m/s (blue line), and b/a = 3.2 and u = 43.1 m/s (red line). (C) DW position versus
is in close agreement with the analytical model.
time for a shift current pulse sequence of the same form as shown in Fig. 2A with amplitude u = 43.1 m/s,
b/a = 3.2, and three waiting times twait. Each pulse is 100 ns long. Details of the curves highlighting Notwithstanding our findings that a DW, ir-
inertia-driven motion at the end of the first shift current pulse are shown in (D). The vertical dashed line at respective of its inertia, can be precisely moved a
100 ns indicates the end of this first pulse. The solid diamonds at 101 and 110 ns show the onset of the distance proportional to the temporal length of a
second shift pulse. (E) Velocity (red) and position (black) of a vortex DW versus time in response to a current pulse, the DW’s inertial response means
stepwise change in current calculated by micromagnetic simulations in a 20-nm-thick, 200-nm-wide that its position at a given point in time is not
nanowire, for the same set of parameters as found experimentally (b/a ~3.2, J = 1.2 × 108 A/cm2, P = simply linearly related to the time elapsed since
0.5). The blue solid line is a fit to the data as described in the text. (F) Snapshots of the vortex DW structure at the beginning of the current pulse. This must be
zero time and when moving at terminal velocity (60 ns). The snapshots are centered on the DW. The direction taken into account when devising clocking schemes
of magnetization is color-coded, as indicated by the black arrows. The wall is moving from right to left. for memory or logic devices.

1812 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


References and Notes
1. L. Berger, J. Appl. Phys. 55, 1954 (1984).
2. L. Berger, Phys. Rev. B 33, 1572 (1986).
3. S. S. P. Parkin, M. Hayashi, L. Thomas, Science 320, 190
(2008).
4. J. Grollier et al., Appl. Phys. Lett. 83, 509 (2003).
5. A. Yamaguchi et al., Phys. Rev. Lett. 92, 077205 (2004).
6. M. Yamanouchi, D. Chiba, F. Matsukura, H. Ohno,
Nature 428, 539 (2004).
7. N. Vernier, D. A. Allwood, D. Atkinson, M. D. Cooke,
R. P. Cowburn, Europhys. Lett. 65, 526 (2004).
8. M. Kläui et al., Phys. Rev. Lett. 95, 026601 (2005).
9. D. Ravelosona, D. Lacour, J. A. Katine, B. D. Terris,
C. Chappert, Phys. Rev. Lett. 95, 117203 (2005).
10. M. Hayashi et al., Phys. Rev. Lett. 98, 037204 (2007).
11. S. Yang, J. L. Erskine, Phys. Rev. B 75, 220403 (2007).
12. G. Meier et al., Phys. Rev. Lett. 98, 187202 (2007).
13. M. Hayashi, L. Thomas, R. Moriya, C. Rettner,
S. S. P. Parkin, Science 320, 209 (2008).
REPORTS
14. G. Tatara, H. Kohno, Phys. Rev. Lett. 92, 086601 (2004). 28. L. Bocklage et al., Phys. Rev. B 78, 180405 (2008).
15. Z. Li, S. Zhang, Phys. Rev. Lett. 92, 207203 (2004). 29. Materials and methods can found as supporting material
16. S. Zhang, Z. Li, Phys. Rev. Lett. 93, 127204 (2004). on Science Online.
17. A. Thiaville, Y. Nakatani, J. Miltat, Y. Suzuki, 30. The rise and fall times of the shift current pulses,
Europhys. Lett. 69, 990 (2005). measured in a transmission geometry through the devices
18. S. E. Barnes, S. Maekawa, Phys. Rev. Lett. 95, 107204 using a real-time oscilloscope and corresponding to a
(2005). variation of 20/80% of the pulse amplitude, were ~0.5 ns.
19. A. P. Malozemoff, J. C. Slonczewski, Magnetic 31. M. Hayashi et al., Phys. Rev. Lett. 97, 207205 (2006).
Domain Walls in Bubble Materials (Academic Press, 32. V. Vlaminck, M. Bailleul, Science 322, 410 (2008).
New York, 1979). 33. We thank S.-H. Yang, X. Jiang, and B. Hughes for useful
20. L. Thomas et al., Nature 443, 197 (2006). discussions and help with sample fabrication.
21. R. Moriya et al., Nat. Phys. 4, 368 (2008).
22. I. M. Miron et al., Phys. Rev. Lett. 102, 137202 (2009). Supporting Online Material
23. T. A. Moore et al., Phys. Rev. B 80, 132403 (2009). www.sciencemag.org/cgi/content/full/330/6012/1810/DC1
24. M. Eltschka et al., Phys. Rev. Lett. 105, 056601 (2010). Materials and Methods
25. E. Saitoh, H. Miyajima, T. Yamaoka, G. Tatara, Nature Figs. S1 to S3
432, 203 (2004). References
26. D. Bedau et al., Phys. Rev. Lett. 99, 146601 (2007). 7 September 2010; accepted 16 November 2010
27. L. Thomas et al., Science 315, 1553 (2007). 10.1126/science.1197468

Cassini Finds an Oxygen–Carbon loosely bound electrons from the negative ions by

Downloaded from www.sciencemag.org on December 23, 2010

photo or electron impact ionization as the ions
move away from Rhea.
Dioxide Atmosphere at Saturn’s The in situ detection of O2 and CO2 at Rhea is
consistent with remote observations of Jupiter’s
Icy Moon Rhea icy moons, where the Galileo spacecraft’s Near-
Infrared Mapping Spectrometer observed resonantly
B. D. Teolis,1* G. H. Jones,2,3 P. F. Miles,1 R. L. Tokar,4 B. A. Magee,1 J. H. Waite,1 E. Roussos,5 scattered 4.26-mm infrared emission from atmo-
D. T. Young,1 F. J. Crary,1 A. J. Coates,2,3 R. E. Johnson,6 W.-L. Tseng,6 R. A. Baragiola6 spheric CO2 at Callisto (7), and the Hubble Space
Telescope measured 1304 and 1356 Ǻ ultraviolet
The flyby measurements of the Cassini spacecraft at Saturn’s moon Rhea reveal a tenuous oxygen fluorescence from electron-impact dissociatively ex-
(O2)–carbon dioxide (CO2) atmosphere. The atmosphere appears to be sustained by chemical cited atmospheric O2 at Europa and Ganymede (8).
decomposition of the surface water ice under irradiation from Saturn’s magnetospheric plasma. This Oxygen at Europa and Ganymede is generated by
in situ detection of an oxidizing atmosphere is consistent with remote observations of other icy bodies, radiation chemistry and sputtered from the surface
such as Jupiter’s moons Europa and Ganymede, and suggestive of a reservoir of radiolytic O2 locked ice into the atmosphere by bombarding ions and
within Rhea’s ice. The presence of CO2 suggests radiolysis reactions between surface oxidants and electrons from Jupiter’s magnetosphere (8). The
organics or sputtering and/or outgassing of CO2 endogenic to Rhea’s ice. Observations of outflowing Jupiter findings, and the detection by Cassini of
positive and negative ions give evidence for pickup ionization as a major atmospheric loss mechanism. O2 from ultraviolet (UV) photodecomposition of
ice in Saturn’s rings (9), have long suggested the
n 2 March 2010, the Cassini spacecraft atmosphere was observed, with the CO2 seen possibility of oxygen atmospheres around the sa-

O executed a flyby of Saturn’s icy moon

Rhea, with a trajectory inbound toward
Saturn passing 97 km over the surface at 81° north
latitude. The Ion Neutral Mass Spectrometer
(INMS)—a quadrupole mass analyzer equipped
with an antechamber and electron-impact ionizer
for in situ collection and detection of neutral gas
(1) —was operated during the flyby with the an-
techamber inlet pointed favorably at an angle of
44° to Cassini’s trajectory, enabling the measure-
ment of neutral species. INMS detected a tenuous
atmosphere of oxygen and carbon dioxide in mass
channels 32 and 44 daltons, reaching peak densi-
ties along the trajectory of 5 and 2 T 1 × 1010 mol-
ecules per m3, respectively. A highly non-uniform
1
Southwest Research Institute, Space Science and Engineering
Division, 6220 Culebra Road, San Antonio, TX 78238, USA.
2
Mullard Space Science Laboratory, Department of Space and
Climate Physics, University College London (UCL), Holmbury St.
Mary, Dorking, Surrey RH5 6NT, UK. 3The Centre for Planetary
Sciences at UCL/Birkbeck, Gower Street, London WC1E 6BT, UK.
4
Los Alamos National Laboratory, Space Science and Applica-
tions, Los Alamos, NM 87545, USA. 5Max-Planck-Institut für
Sonnensystemforschung, Max-Planck-Strasse 2, 37191 Katlenburg-
Lindau, Germany. 6University of Virginia, Department of Materials
Science and Engineering, 116 Engineer’s Way, Charlottesville, VA
22903, USA.
*To whom correspondence should be addressed. E-mail:
ben.teolis@swri.org
almost exclusively on the outbound portion of the
trajectory over the day-lit hemisphere (Fig. 1). In
contrast, the O2 profile is more symmetrical about
the point of closest approach, but it is nevertheless
shifted slightly outbound to the day side (Fig. 1).
Spectra from the Cassini Plasma Spectrometer
(CAPS) (2), acquired during the more distant 502-
and 5736-km flybys on 26 November 2005 and
30 August 2007, also show clear signatures (Fig. 2)
symptomatic (3) of outflowing streams of positive
and negative ions, which are produced by ioniza-
tion of the atmosphere and electron capture, respec-
tively. These ions are subsequently swept up into
Saturn’s rotating magnetosphere (4). The timing
of the positive and negative ion signatures inbound
and outbound from →
Rhea
→
(Fig. 2) is consistent
with
→
the expected
→
E × B cycloidal trajectories (where
E and B are the electric and magnetic fields, re-
spectively) of pickup ions in the mass ranges of 26
to 56 daltons (possibly O2+ or CO2+ ) and 13 to
26 daltons, respectively; thus, we tentatively iden-
tify the negative species as O–. The mass uncer-
tainty results from the CAPS energy and angular
resolution (2), as well as the still-uncertain corota-
tion electric field and corotation speed in Rhea’s
plasma wake (5). Unlike the 2005 encounter, only
positive ions were detected during the 11 times more
distant 2007 flyby, suggesting rapid (6) removal of
turnian icy satellites (10), which orbit inside Saturn’s
magnetosphere. Ganymede’s ice (11) and that of
Europa and Callisto (12) also exhibit the weak
5770 and 6275 Ǻ optical absorption signatures of
trapped radiolytic O2 (13), which has been shown
in laboratory experiments to lead to ozone as a by-
product (14), along with eventual O2 ejection from
the surface through sputtering (15). Rhea and
Saturn’s icy moon Dione are especially interesting
because O3 is present in their surface ices (16), a
trait that they share with Ganymede (17). Together
with the existence of ozone in Rhea’s ice, the de-
tection of an O2 atmosphere is consistent with sur-
face radiolysis, as seen at other icy satellites, and
indicative of O2 trapped in the surface ice.
On the basis of CAPS and Magnetospheric
Imaging Instrument (MIMI) measurements of the
saturnian ion and electron plasma, as well as up-
dated laboratory estimates of O2 production and
desorption from ice irradiated with different pro-
jectiles and energies, we have modeled the ex-
pected production of O2 from different radiation
sources (18). The principal oxygen source in the
model is bombardment by water group ions (W+)
from Saturn’s corotating plasma (Table 1), which
sweep past Rhea along its orbit while preferen-
tially bombarding its trailing hemisphere. The ox-
ygen is, therefore, produced preferentially on the

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1813


References and Notes
1. L. Berger, J. Appl. Phys. 55, 1954 (1984).
2. L. Berger, Phys. Rev. B 33, 1572 (1986).
3. S. S. P. Parkin, M. Hayashi, L. Thomas, Science 320, 190
(2008).
4. J. Grollier et al., Appl. Phys. Lett. 83, 509 (2003).
5. A. Yamaguchi et al., Phys. Rev. Lett. 92, 077205 (2004).
6. M. Yamanouchi, D. Chiba, F. Matsukura, H. Ohno,
Nature 428, 539 (2004).
7. N. Vernier, D. A. Allwood, D. Atkinson, M. D. Cooke,
R. P. Cowburn, Europhys. Lett. 65, 526 (2004).
8. M. Kläui et al., Phys. Rev. Lett. 95, 026601 (2005).
9. D. Ravelosona, D. Lacour, J. A. Katine, B. D. Terris,
C. Chappert, Phys. Rev. Lett. 95, 117203 (2005).
10. M. Hayashi et al., Phys. Rev. Lett. 98, 037204 (2007).
11. S. Yang, J. L. Erskine, Phys. Rev. B 75, 220403 (2007).
12. G. Meier et al., Phys. Rev. Lett. 98, 187202 (2007).
13. M. Hayashi, L. Thomas, R. Moriya, C. Rettner,
S. S. P. Parkin, Science 320, 209 (2008).
REPORTS
14. G. Tatara, H. Kohno, Phys. Rev. Lett. 92, 086601 (2004). 28. L. Bocklage et al., Phys. Rev. B 78, 180405 (2008).
15. Z. Li, S. Zhang, Phys. Rev. Lett. 92, 207203 (2004). 29. Materials and methods can found as supporting material
16. S. Zhang, Z. Li, Phys. Rev. Lett. 93, 127204 (2004). on Science Online.
17. A. Thiaville, Y. Nakatani, J. Miltat, Y. Suzuki, 30. The rise and fall times of the shift current pulses,
Europhys. Lett. 69, 990 (2005). measured in a transmission geometry through the devices
18. S. E. Barnes, S. Maekawa, Phys. Rev. Lett. 95, 107204 using a real-time oscilloscope and corresponding to a
(2005). variation of 20/80% of the pulse amplitude, were ~0.5 ns.
19. A. P. Malozemoff, J. C. Slonczewski, Magnetic 31. M. Hayashi et al., Phys. Rev. Lett. 97, 207205 (2006).
Domain Walls in Bubble Materials (Academic Press, 32. V. Vlaminck, M. Bailleul, Science 322, 410 (2008).
New York, 1979). 33. We thank S.-H. Yang, X. Jiang, and B. Hughes for useful
20. L. Thomas et al., Nature 443, 197 (2006). discussions and help with sample fabrication.
21. R. Moriya et al., Nat. Phys. 4, 368 (2008).
22. I. M. Miron et al., Phys. Rev. Lett. 102, 137202 (2009). Supporting Online Material
23. T. A. Moore et al., Phys. Rev. B 80, 132403 (2009). www.sciencemag.org/cgi/content/full/330/6012/1810/DC1
24. M. Eltschka et al., Phys. Rev. Lett. 105, 056601 (2010). Materials and Methods
25. E. Saitoh, H. Miyajima, T. Yamaoka, G. Tatara, Nature Figs. S1 to S3
432, 203 (2004). References
26. D. Bedau et al., Phys. Rev. Lett. 99, 146601 (2007). 7 September 2010; accepted 16 November 2010
27. L. Thomas et al., Science 315, 1553 (2007). 10.1126/science.1197468

Cassini Finds an Oxygen–Carbon loosely bound electrons from the negative ions by

Downloaded from www.sciencemag.org on December 23, 2010

photo or electron impact ionization as the ions
move away from Rhea.
Dioxide Atmosphere at Saturn’s The in situ detection of O2 and CO2 at Rhea is
consistent with remote observations of Jupiter’s
Icy Moon Rhea icy moons, where the Galileo spacecraft’s Near-
Infrared Mapping Spectrometer observed resonantly
B. D. Teolis,1* G. H. Jones,2,3 P. F. Miles,1 R. L. Tokar,4 B. A. Magee,1 J. H. Waite,1 E. Roussos,5 scattered 4.26-mm infrared emission from atmo-
D. T. Young,1 F. J. Crary,1 A. J. Coates,2,3 R. E. Johnson,6 W.-L. Tseng,6 R. A. Baragiola6 spheric CO2 at Callisto (7), and the Hubble Space
Telescope measured 1304 and 1356 Ǻ ultraviolet
The flyby measurements of the Cassini spacecraft at Saturn’s moon Rhea reveal a tenuous oxygen fluorescence from electron-impact dissociatively ex-
(O2)–carbon dioxide (CO2) atmosphere. The atmosphere appears to be sustained by chemical cited atmospheric O2 at Europa and Ganymede (8).
decomposition of the surface water ice under irradiation from Saturn’s magnetospheric plasma. This Oxygen at Europa and Ganymede is generated by
in situ detection of an oxidizing atmosphere is consistent with remote observations of other icy bodies, radiation chemistry and sputtered from the surface
such as Jupiter’s moons Europa and Ganymede, and suggestive of a reservoir of radiolytic O2 locked ice into the atmosphere by bombarding ions and
within Rhea’s ice. The presence of CO2 suggests radiolysis reactions between surface oxidants and electrons from Jupiter’s magnetosphere (8). The
organics or sputtering and/or outgassing of CO2 endogenic to Rhea’s ice. Observations of outflowing Jupiter findings, and the detection by Cassini of
positive and negative ions give evidence for pickup ionization as a major atmospheric loss mechanism. O2 from ultraviolet (UV) photodecomposition of
ice in Saturn’s rings (9), have long suggested the
n 2 March 2010, the Cassini spacecraft atmosphere was observed, with the CO2 seen possibility of oxygen atmospheres around the sa-

O executed a flyby of Saturn’s icy moon

Rhea, with a trajectory inbound toward
Saturn passing 97 km over the surface at 81° north
latitude. The Ion Neutral Mass Spectrometer
(INMS)—a quadrupole mass analyzer equipped
with an antechamber and electron-impact ionizer
for in situ collection and detection of neutral gas
(1) —was operated during the flyby with the an-
techamber inlet pointed favorably at an angle of
44° to Cassini’s trajectory, enabling the measure-
ment of neutral species. INMS detected a tenuous
atmosphere of oxygen and carbon dioxide in mass
channels 32 and 44 daltons, reaching peak densi-
ties along the trajectory of 5 and 2 T 1 × 1010 mol-
ecules per m3, respectively. A highly non-uniform
1
Southwest Research Institute, Space Science and Engineering
Division, 6220 Culebra Road, San Antonio, TX 78238, USA.
2
Mullard Space Science Laboratory, Department of Space and
Climate Physics, University College London (UCL), Holmbury St.
Mary, Dorking, Surrey RH5 6NT, UK. 3The Centre for Planetary
Sciences at UCL/Birkbeck, Gower Street, London WC1E 6BT, UK.
4
Los Alamos National Laboratory, Space Science and Applica-
tions, Los Alamos, NM 87545, USA. 5Max-Planck-Institut für
Sonnensystemforschung, Max-Planck-Strasse 2, 37191 Katlenburg-
Lindau, Germany. 6University of Virginia, Department of Materials
Science and Engineering, 116 Engineer’s Way, Charlottesville, VA
22903, USA.
*To whom correspondence should be addressed. E-mail:
ben.teolis@swri.org
almost exclusively on the outbound portion of the
trajectory over the day-lit hemisphere (Fig. 1). In
contrast, the O2 profile is more symmetrical about
the point of closest approach, but it is nevertheless
shifted slightly outbound to the day side (Fig. 1).
Spectra from the Cassini Plasma Spectrometer
(CAPS) (2), acquired during the more distant 502-
and 5736-km flybys on 26 November 2005 and
30 August 2007, also show clear signatures (Fig. 2)
symptomatic (3) of outflowing streams of positive
and negative ions, which are produced by ioniza-
tion of the atmosphere and electron capture, respec-
tively. These ions are subsequently swept up into
Saturn’s rotating magnetosphere (4). The timing
of the positive and negative ion signatures inbound
and outbound from →
Rhea
→
(Fig. 2) is consistent
with
→
the expected
→
E × B cycloidal trajectories (where
E and B are the electric and magnetic fields, re-
spectively) of pickup ions in the mass ranges of 26
to 56 daltons (possibly O2+ or CO2+ ) and 13 to
26 daltons, respectively; thus, we tentatively iden-
tify the negative species as O–. The mass uncer-
tainty results from the CAPS energy and angular
resolution (2), as well as the still-uncertain corota-
tion electric field and corotation speed in Rhea’s
plasma wake (5). Unlike the 2005 encounter, only
positive ions were detected during the 11 times more
distant 2007 flyby, suggesting rapid (6) removal of
turnian icy satellites (10), which orbit inside Saturn’s
magnetosphere. Ganymede’s ice (11) and that of
Europa and Callisto (12) also exhibit the weak
5770 and 6275 Ǻ optical absorption signatures of
trapped radiolytic O2 (13), which has been shown
in laboratory experiments to lead to ozone as a by-
product (14), along with eventual O2 ejection from
the surface through sputtering (15). Rhea and
Saturn’s icy moon Dione are especially interesting
because O3 is present in their surface ices (16), a
trait that they share with Ganymede (17). Together
with the existence of ozone in Rhea’s ice, the de-
tection of an O2 atmosphere is consistent with sur-
face radiolysis, as seen at other icy satellites, and
indicative of O2 trapped in the surface ice.
On the basis of CAPS and Magnetospheric
Imaging Instrument (MIMI) measurements of the
saturnian ion and electron plasma, as well as up-
dated laboratory estimates of O2 production and
desorption from ice irradiated with different pro-
jectiles and energies, we have modeled the ex-
pected production of O2 from different radiation
sources (18). The principal oxygen source in the
model is bombardment by water group ions (W+)
from Saturn’s corotating plasma (Table 1), which
sweep past Rhea along its orbit while preferen-
tially bombarding its trailing hemisphere. The ox-
ygen is, therefore, produced preferentially on the

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1813

 REPORTS
Fig. 1. (A) INMS 32-dalton
measurement (32) of the
O2 density along Cassini’s
trajectory versus time dur-
ing the 2 March 2010
Rhea encounter. The black
vertical dotted line indi-
cates that the closest ap-
proach (CA) was 17:40:39
UT at 96.8-km altitude
and nearly simultaneous
(later by ~0.05 s) with so-
lar terminator traversal to
Rhea’s day side. The blue
dashed curve denotes
along-track density pre-
dicted by a Monte Carlo
simulation of the O2 at-
trailing hemisphere. Relative to the amount of en-
ergy deposited by the different radiation sources,
the W+ ions are also the most efficient O2 produc-
ers (Table 1). Because of their mass, these ions are
A the most effectively stopped on impact within the
ice; that is, they deposit the most energy close to
the material surface where, according to experi-
ments, O2 synthesis is favored most (18). The
predicted total global production rate of ~2.2 ×
1024 O2 molecules per second is within the ~0.4
to 4 × 1024 s–1 range implied by the elevated O2+
densities seen by CAPS near Rhea’s orbit (19).
In contrast to O2, knowledge of Rhea’s CO2
B source is much less well constrained. Atmospher-
ic CO2 might result from sputtering of primordial
CO2 in Rhea’s ice or from radiolysis reactions
between surface water molecules, radiolytic ox-
ygen, and carbonaceous minerals or organics pos-

Downloaded from www.sciencemag.org on December 23, 2010

mosphere that assumes sibly present in the surface ice (13, 20, 21) and/or
100/40 K day/night sur- deposited by micrometeorite bombardment (22).
face temperatures, respec- The trailing hemispheres of Rhea and Dione both
tively (25). (B) Same as Cassini (8.6 km/s) 1012 show a darkening in their visible-infrared reflec-
(A) for CO2 in the 44-
C
17:36 17:46
tance spectrum, which is indicative of such non-
dalton mass channel. (C)

O2 per m3
E ice material. At Dione, the Cassini Visible and
Diagrammatic equatori- 1011
Infrared Mapping Spectrometer (VIMS) detected
al view of Rhea looking B
Saturn
perpendicular to the Cas- the 4.26-mm absorption of CO2 in the dark re-
26 Nov 2005 1010 gions (22). However, the Rhea measurements are
sini 2010 trajectory (red Trajectory
line in Rhea’s reference 1000 km inconclusive: A possible detection of the CO2 ab-
frame) on the same time 109 sorption in Rhea’s global spectrum (22) was not
scale as in (A) and (B). The confirmed in subsequent VIMS mapping measure-
vantage point at 81.8° longitude and 8.9° north latitude is near the apex (90° longitude) of Rhea’s leading ments (23) of the surface. A completely endogenic
hemisphere. Cassini’s motion toward Saturn at 8.6 km/s was nearly perpendicular (at 88.8°) to the day- CO2 source is also possible: for instance, outgassing
night terminator (shown at the time of CA), with CA at 81.1° north latitude, 263.4° longitude. Rhea’s orbit of primordial CO2 or of CO2 produced by aqueous
and Saturn’s corotation direction point out of the page and perpendicular to the magnetic and corotation chemistry from Rhea’s interior, similar to scenarios
→ →
electric fields B and E . Also shown is the O2 density cross section predicted by the Monte Carlo model. suggested at Enceladus (24) and Callisto (7).

Fig. 2. (A) Diagrammatic Rhea north polar view with the 26 November 2005
1000 km Corotating
Cassini flyby trajectory (black line in Rhea’s reference frame) during which A
Plasma 1012
CAPS detected pickup ions. The time scale is matched to that of (B) and (C). E B
The day and night hemispheres are shown during CA at 22:37:39 UT. The

O2 per m3
trajectory traversed Rhea’s plasma wake, with CA at 502-km altitude, 226 km CO2+ 2 March 2010 1011
south (Fig. 1) of the equator. Our model prediction of the O2 density (226-km Trajectory

south cross section) is also shown. The O2+ and O– (orange) and CO2+ trajectories O2+ Saturn
1010
(blue) are those required to enter anodes 4 and 3 (33) of the CAPS Electron FOV O-
Spectrometer (ELS) and Ion Mass Spectrometer (IMS) at the time and energy
of the ion signatures. The trajectories assume (in Saturn’s reference frame) a FOV
Cassini 109
→ →
B of 26 nT (34) and a corotation electric field E [within uncertainty (5)] of (7.3 km/s)
+ – + Rhea
1.77 (O2 , O ) or 1.51 (CO2 ) V/km. Before ionization, most atmospheric neu-
→ orbit
trals have thermal speeds less than 1 km/s, so jE j is optimized such that ions
backtracked from Cassini come nearly to rest (the trajectory starting point). 104
ELS Anode 4
FOV, field of view. (B) ELS negative particle flux spectrogram from anode 4
3
(20° FOV), which had optimal pointing. Negative pickup ions are indicated by 10 - ions
the sharp feature near 22:41 UT (T0.35 min) and 1.14 (T0.15) keV over the B
Particle Flux (m -2 ster -1 s -1)

1012
2
electron background. (C) Positive ions from IMS anode 3: Pickup ions produce 10
Energy (eV)

the sharp 22:32 UT (T0.5 min), 2.06 (T0.2)–keV signature over the 1011
101
background of (mostly) corotating H+/W+ (31).
104 1010
IMS Anode 3
+ ions
103 109

102 C 108

101

22:32 22:34 22:36 22:38 22:40 22:42 22:44

Time UT (26 November 2005)

1814 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


Table 1. Estimated O2 production from differ-
ent radiation sources.
Radiation Energy Estimated O2
source deposition production
(× 1026 eV/s) (× 1022 O2/s)
W+ 14.8 170
H+ 9.5 7.4
Electrons 73 38
Solar UV 8.1 4.2
Total 105 220
The surface source processes compete with
atmospheric loss mechanisms to determine the
atmospheric O2 and CO2 abundances. The loss
mechanisms are Jeans escape and atmosphere-
plasma interactions—that is, ionization, dissocia-
tion, charge exchange, and electron capture. The
plasma-interaction channels result in fast neutral
and ionized species that, depending on their point
of origin (Fig. 2), either (i) collide with Rhea’s
surface (and implant into the ice or adsorb or react
on the surface)→ or →(ii) escape into space directly or
(for ions) by E × B pickup, as seen by CAPS.
We used a Monte Carlo approach to model the
atmosphere by initializing O2 molecules according
to the expected surface position–dependent produc-
tion (18) while allowing the molecules to execute
random ballistic trajectories between surface impacts.
The simulation assumed no surface adsorption, ther-
mally equilibrated molecules with the surface on
impact by reinitializing the speed with a Maxwell-
Boltzmann distribution at the local surface temper-
ature (25), and destroyed the molecules in mid-flight,
according to the loss rate from plasma interactions
or on leaving the Hill sphere (Jeans escape).
Our model predicts a day-side bulge due to
the higher (25) temperatures (and, therefore, in-
creased scale height) that is well matched by the
outbound O2 tail seen by INMS (Fig. 1); i.e., the
warmer day-side temperatures expand the atmo-
spheric gas to the altitude of Cassini’s trajectory
in this hemisphere. The bulge is also consistent
with the predicted origin of the positive pickup
ions at high altitudes during the 2005 flyby (Fig.
2). Non-negligible night-side O2 adsorption could
account (26) for the model’s slight overestimate
of the INMS inbound measurements (Fig. 1). The
mean free path is ~6 to 30 × 103 km at the pre-
dicted day-night surface densities of ~9 to 40 ×
1010 O2 m–3 (26); therefore, the atmosphere can
be considered as collisionless. The estimated Jeans
escape of 6(T1) × 1022 O2 s–1 is ~2.7% of the
estimated total O2 produced (~2.2 × 1024 O2 s–1),
with the remainder lost because of plasma
interactions and pickup. The total atmospheric
O2 abundance is estimated to be 2.5(T0.5) × 1029
molecules, corresponding to an average atmospher-
ic molecule lifetime of ~105 seconds, or ~1 day.
Whereas for O2 the sticking times are short,
CO2 is much less volatile (27); thus, the night side
could act as a much more effective cold trap for
CO2 condensation, possibly explaining the almost
nonexistent CO2 signal inbound on the night side
www.sciencemag.org
(Fig. 1). Locally condensed CO2 would be re-
released by solar heating as the dawn terminator
advances across the surface, although some CO2
might be trapped for longer periods in shadowed
polar regions [analogous to lunar (28) volatiles].
Although surface CO2 on the night side would be
undetectable by Cassini VIMS (which measures
reflected sunlight), CO2 near the faintly illuminated
poles or dawn terminator might be observable.
The estimated mean atmospheric O2 column
density of 3.4(T0.7) × 1016 m–2 over Rhea’s sur-
face is two orders of magnitude below the 2.4 to
14 and 1 to 10 × 1018 m–2 abundances at Europa
and Ganymede (8), respectively, a difference like-
ly attributable to the greater O2 desorption flux
from the warmer and more intensely irradiated
Galilean satellites (29). Rhea’s atmospheric abun-
dance is also well below the 1018 m–2 detection
limit of MIMI and the Cassini Ultraviolet Imaging
Spectrograph, explaining why earlier attempts by
these instruments to detect an atmosphere remote-
ly were unsuccessful (30). In comparison, labo-
ratory measurements on irradiated ice imply that
1019 to 1020 O2 m–2 (14, 15) are synthesized by
penetrating ions as trapped molecules inside the
bulk H2O solid, from which diffusive loss is
expected to be slow; thus, it is likely that a large
fraction of Rhea’s oxygen is actually locked in-
side the moon’s ice. The laboratory column den-
sities correspond to ~0.4 to 4 × 104 metric tons of
trapped O2 globally on Rhea, but these are a lower
limit because diffusion and micrometeorite
gardening can disperse O2 into the subsurface ice.
References and Notes
1. J. H. Waite Jr. et al., Space Sci. Rev. 114, 113 (2004).
2. D. T. Young et al., Space Sci. Rev. 114, 1 (2004).
3. A. J. Coates et al., Icarus 206, 618 (2010).
4. CAPS did not detect pickup ions during the 2010 flyby
because Cassini’s path north of Rhea did not intersect the
allowable ion trajectories (Figs. 1 and 2).
5. Field strengths of 1.77 and 1.51 V/km, consistent with
O2+ and CO2+ (Fig. 2), yield bulk plasma speeds |E|/|B| =
68 and 58 km/s, respectively, both of which are
compatible with published estimates [(31) and
supplemental reference 44 (S44)].
6. Assuming a 49-km/s bulk plasma speed with respect to
Rhea (18, 31), pickup ions reaching Cassini’s 2007
position are approximately 2 min old, compared with
ages of 10 (T5) s in 2005.
7. R. W. Carlson, Science 283, 820 (1999).
8. D. T. Hall, P. D. Feldman, M. A. McGrath, D. F. Strobel,
Astrophys. J. 499, 475 (1998).
9. W.-L. Tseng, W.-H. Ip, R. E. Johnson, T. A. Cassidy,
M. K. Elrod, Icarus 206, 382 (2010).
10. J. Saur, D. F. Strobel, Astrophys. J. 620, L115 (2005).
11. J. R. Spencer, W. M. Calvin, M. J. Person, J. Geophys. Res.
100, 19049 (1995).
12. J. R. Spencer, W. M. Calvin, Astron. J. 124, 3400 (2002).
13. J. Spencer, Icarus 136, 349 (1998).
14. B. D. Teolis, M. J. Loeffler, U. Raut, M. Fama,
R. A. Baragiola, Astrophys. J. 644, L141 (2006).
15. B. D. Teolis, J. Shi, R. A. Baragiola, J. Chem. Phys. 130,
134704 (2009).
16. K. S. Noll, T. L. Roush, D. P. Cruikshank, R. E. Johnson,
Y. J. Pendleton, Nature 388, 45 (1997).
17. K. S. Noll, R. E. Johnson, A. L. Lane, D. L. Domingue,
H. A. Weaver, Science 273, 341 (1996).
18. Materials, methods, and further discussion are available
as supporting material on Science Online.
19. CAPS measurements throughout the Saturn system imply an
average relative O2+/W+ density of 0.31 T 0.01% at 4.5 to
SCIENCE VOL 330 24 DECEMBER 2010
REPORTS
7.5Rs (where Rs is the radius of Saturn) from photo-
ionization of ring-atmosphere O2 and a statistically greater
0.41 T 0.02% at 7.5 to 10.5Rs (S45). Rhea orbits at 8.75Rs.
20. D. P. Cruikshank et al., Icarus 206, 561 (2010).
21. D. P. Cruikshank et al., Icarus 175, 268 (2005).
22. R. N. Clark et al., Icarus 193, 372 (2008).
23. K. Stephan et al., in 40th Lunar and Planetary Science
Conference, Abstract 1377 (Lunar and Planetary Institute,
The Woodlands, TX, 23 to 27 March 2009).
24. C. R. Glein, M. Y. Zolotov, E. L. Shock, Icarus 197, 157
(2008).
25. Considering the day/night hemisphere orientation at the
times of the 2 March 2010 and 26 November 2005
flybys, we reconstructed a temperature map based on
Cassini Composite Infrared Spectrometer (CIRS)
measurements (S16) showing 100/40 K max/min day/night
temperatures, respectively. Because the CIRS coverage did
not include the poles, we assume a constant 35/75 K
(26 Nov 2005, Saturn winter) and 35/35 K (2 March 2010,
near Saturn equinox) beyond 75° north/south latitude.
26. The predicted day-night surface densities correspond to
surface pressures of 1.2 and 2.2 × 10−12 mbar at 100
Downloaded from www.sciencemag.org on December 23, 2010
and 40 K, respectively. The values are 100 × 1010 O2/m3,
4.8 × 10−12 mbar, and 3 × 103 km mean free path for
35 K minimum at the poles. Any night-side or polar O2
adsorption [e.g., into the porous surface regolith (S46)]
would lower the overlying atmospheric density (Fig. 1);
hence, the night-side and polar values are upper limits.
27. The O2 and CO2 sublimation energies are 0.095 eV (S47)
and 0.271 eV (S48) from the pure solid. Extrapolation from
available equilibrium vapor pressure data (S49, S50) yields
O2 and CO2 values of 5.69 × 10−5 and 3.15 × 10−32 mbar
at 35 K, as well as 2547 and 2.11 × 10−4 mbar at 100 K.
28. J. M. Sunshine et al., Science 326, 565 (2009);
10.1126/science.1179788.
29. Compared with Rhea [40 to 100 K (S16)], the warmer
Europa/Ganymede ice [80 to 132 K/90 to 152 K (S51, S52)]
is expected to exhibit O2 yields roughly one order of
magnitude greater (15). For instance, Europa receives a
mean of ~1.2 and 3.6 × 1017 eV/m2/s (S53) from ions
(H+, W+, S+) and electrons, respectively, and mean O2
sources are ~1013 to 1015 O2/m2/s for both moons
(S24–S28), which exceed the Rhea values: ~3.3 (ions) and
9.9 (electrons) × 1014 eV/m2/s and 3.0 × 1011 O2/m2/s.
30. G. H. Jones et al., Science 319, 1380 (2008).
31. R. J. Wilson, R. L. Tokar, W. S. Kurth, A. M. Persoon,
J. Geophys. Res. 115, A05201 (2010).
32. Mass channels are sampled one at a time by integrating
detector counts during 0.031-s intervals. During the
2010 flyby, CO2 and O2 were sampled every 2.3 and
6.9 s, respectively.
33. We assumed ions entering the center of the field of view,
which is 20°.
34. K. K. Khurana, C. T. Russell, M. K. Dougherty, Icarus 193,
465 (2008).
35. We thank E. Roussos, D. Mitchell, and the Cassini MIMI
team for providing the MIMI data used in figs. S1 and
S4 and N. G. Petrik, A. G. Kavetsky, and G. A. Kimmel
at Pacific Northwest National Laboratory, Richland,
Washington, USA, for contributing the O2 electron-
stimulated desorption measurements in fig. S9. The INMS
and CAPS teams acknowledge support from NASA and the
Jet Propulsion Laboratory under SwRI subcontracts
1283095 and 1356497, respectively. G.H.J. and A.J.C.
acknowledge support for CAPS-ELS operations and
analysis by the United Kingdom Space Agency, and G.H.J.
was supported by a UK Science and Technology Facilities
Council Advanced Fellowship. B.D.T. and R.A.B.
acknowledge support by the NSF Astronomy and
Astrophysics Program through grant AST0807830.
Supporting Online Material
www.sciencemag.org/cgi/content/full/science.1198366/DC1
Materials and Methods
SOM Text
Figs. S1 to S10
References and Notes
28 September 2010; accepted 12 November 2010
Published online 25 November 2010;
10.1126/science.1198366
1815
 REPORTS
we solved the structures of two FD mutants, S183A
Structures of C3b in Complex with (S195A) and R202A (R218A) in their unbound
form. Refinement statistics and the quality of the
Factors B and D Give Insight into final models are described in table S1 and fig. S1.
C3b in both C3bB and C3bBD* resembles free

Complement Convertase Formation C3b (fig. S2) (30, 31). Two polypeptide chains,
termed b (residues 1 to 645) and a′ (residues 727
to 1641) chains, together form a core of eight
Federico Forneris,1 Daniel Ricklin,2 Jin Wu,1 Apostolia Tzekou,2 Rachel S. Wallace,1 macroglobulin (MG1 to MG8) domains with a
John D. Lambris,2* Piet Gros1* linker domain (LNK) inserted into domain MG6,
a “complement C1r/C1s, Uegf, Bmp1” (CUB) and
Activation of the complement cascade induces inflammatory responses and marks cells for immune a thioester-containing domain (TED) in between
clearance. In the central complement-amplification step, a complex consisting of surface-bound MG7 and MG8, and a C-terminal C345C domain
C3b and factor B is cleaved by factor D to generate active convertases on targeted surfaces. We attached to MG8 (Fig. 1). Most of the C3b
present crystal structures of the pro-convertase C3bB at 4 angstrom resolution and its complex with domains (MG1 to MG8, LNK, and CUB) show
factor D at 3.5 angstrom resolution. Our data show how factor B binding to C3b forms an open only minor rearrangements (with rotations up to
“activation” state of C3bB. Factor D specifically binds the open conformation of factor B through a 7°) compared to free C3b; however, we observed
site distant from the catalytic center and is activated by the substrate, which displaces factor D’s larger rotations for the peripheral domains TED

Downloaded from www.sciencemag.org on December 23, 2010

self-inhibitory loop. This concerted proteolytic mechanism, which is cofactor-dependent and and C345C (up to 32° and 15° rotations for C3bB
substrate-induced, restricts complement amplification to C3b-tagged target cells. and C3bBD*, respectively) (fig. S2). FB consists
of five domains grouped into two segments: three
he human complement system contributes cobra-venom factor (CVF) indicated how the pro- N-terminal CCP domains that together with the

T to a variety of processes, ranging from

immunosurveillance to maintenance of cell
homeostasis, which depend on a balance of com-
plement activation and regulation (1). Besides a
low level of steady activation through hydrolysis
of the central complement component, C3, the
cascade can be specifically triggered by various
structures on the surfaces of foreign and apoptotic
cells (1, 2). These events initiate distinct pathways
that converge into a central amplification loop,
which relies on an efficient enzyme complex that
allows rapid turnover of C3 (3). This C3 convertase
complex is generated when factor B (FB) interacts
with surface-bound complement fragment C3b and
forms the pro-convertase C3bB, which is cleaved
by factor D (FD) to yield the active yet unstable
C3 convertase C3bBb (4, 5). The C3 convertase
cleaves C3 into anaphylatoxin C3a and opsonin
C3b (2, 6–8). Newly generated C3b molecules
become covalently attached to cell surfaces proxi-
mate to sites of activation and form additional
convertases, thereby rapidly amplifying com-
plement response on targeted cells (9–11). In order
to prevent excessive activation and attack of host
tissue, convertase activity has to be tightly con-
trolled (12). Although recent studies provided
insight into structure, function, stability, and reg-
ulation of the assembled C3 convertase (13, 14),
molecular details about the convertase formation
and activation step remained elusive.
The published crystal structure of free FB
showed a “locked” conformation in which the
scissile bond (Arg234–Lys235) is protected from
being cleaved by FD (15). Moreover, a recent
structure of FB in complex with the C3b homolog
1
Crystal and Structural Chemistry, Bijvoet Center for Biomole-
cular Research, Department of Chemistry, Faculty of Science,
Utrecht University, Padualaan 8, 3584 CH Utrecht, Netherlands.
2
Department of Pathology and Laboratory Medicine, University
of Pennsylvania, 401 Stellar Chance, Philadelphia, PA 19104,
USA.
*To whom correspondence should be addressed. E-mail:
p.gros@uu.nl (P.G.); lambris@upenn.edu (J.D.L.)
peptide segment of FB, Ba (residues 1 to 234), me-
diates the Mg2+-dependent binding of the protease
segment Bb (residues 235 to 739) to the C terminus
of CVF through the metal-ion-dependent adhe-
sion site (MIDAS) of the Von Willebrand factor
type-A (VWA) domain in Bb (16). Unexpectedly,
the structure of CVFB did not show conforma-
tional changes in FB that lead to exposure of its
scissile bond. Single-particle reconstruction of
negatively stained electron microscopy (EM) data
for the C3bB complex, however, indicated the
coexistence of two conformations: one corre-
sponding to a closed form as observed in CVFB
and one to an open form that putatively exposes
the scissile bond in FB (17). Allosteric roles have
been proposed for the regulatory MIDAS and
C-terminal helix a7 of the VWA domain of Bb
[structural elements that play a key role in signal-
ing through the homologous I domains in integrins
(18, 19)] and the three complement-control protein
(CCP) domains and C-terminal helical linker aL of
the Ba segment (15, 20, 21). Whereas it is likely
that binding of FB to C3b induces conformational
changes that expose the scissile bond (Arg234–
Lys235) for cleavage by FD, the mechanisms in-
volved in this process have not yet been elucidated.
FD is a single-domain serine protease, which
circulates in plasma in a mature but self-inhibited
form (22, 23). Even when the inhibition was pre-
vented by mutagenesis, the activity of FD against
peptides remained poor when compared with other
serine proteases like trypsin or thrombin (24–26).
Therefore, it is likely that proteolytic activity and
specificity are controlled during the assembly and
activation of the central enzyme complex.
We determined crystal structures of the pro-
convertase C3bB stabilized by Ni2+ (17, 27) at
4.0-Å resolution and of the pro-convertase C3bB
in complex with a catalytically inactive FD mutant
Ser183→Ala183 [abbreviated as S183A (28)] (S195A,
chymotrypsin numbering is given in parentheses)
in the presence of Mg2+ at 3.5-Å resolution, which
we refer to as C3bBD* (Fig. 1A) (29). In addition,
helical linker aL form the Ba segment and a central
VWA domain and a C-terminal trypsinlike serine
protease (SP) domain that form the protease seg-
ment Bb (Fig. 1). The CCP1-3 and VWA domains
of FB bind C3b in an identical orientation as in
CVFB (16) (Fig. 1B and fig. S2D). In contrast, the
SP domain of FB in C3bB is rotated by 84°, which
effectively shifts the catalytic serine residue by 65
Å and creates additional contacts between FB and
C3b (Fig. 1, A and B). Furthermore, the structure
of C3bBD* shows that FD binds this new con-
formation of FB without making any contacts to
C3b (Fig. 1C), which is confirmed by surface plas-
mon resonance (SPR) measurements showing that
C3b interacts with FB but not with FD (fig. S3).
In the presence of the physiologically relevant
ion Mg2+, two conformations of C3bB have been
observed in EM experiments at a ratio of 35%:65%,
which were referred to as closed and open forms,
respectively (17). When using Ni2+, however, the
equilibrium was shifted to a 2%:98% ratio (17).
We used Ni2+ to crystallize the open form of C3bB.
The crystal structure of C3bB (Ni2+) (Fig. 2A)
fits well, with a correlation coefficient of 0.87 (fig.
S4), to the shape of the open form of the pro-
convertase seen in the EM experiments (17, 32).
In contrast, the crystal structure of CVFB, which
was confirmed by small-angle x-ray scattering
solution studies (16), correlates with the closed
form observed in EM. The closed conformation
of CVFB possibly explains why it is processed
~100-fold slower than C3bB. However, despite
the good agreement between the C3bB crystal
structure and the EM reconstruction of the open
form, our findings do not agree with the proposed
structural model of domain rearrangements that
was derived from the EM data (32). In particular,
the x-ray data show a significant rotation of the
SP domain rather than the CCP1-3 domains, as
suggested by the EM model. The highly similar
conformation of FB in the structure of C3bB in
complex with FD (that is, C3bBD* crystallized
in the presence of Mg2+) confirms that C3bB (Ni2+)

1816 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


adopts a physiological open conformation that is
cleavable by FD. In all three structures (i.e., CVFB,
C3bB, and C3bBD*), the C terminus of CVF or
C3b is chelated to the metal ion in the regulatory
MIDAS of the VWA domain in FB. The electron
densities at 4 Å and 3.5 Å resolution for C3bB and
C3bBD*, respectively, are consistent with high-
affinity configurations for the MIDAS in both
structures, as was observed for the structure of
CVFB determined at 2.2 Å resolution (16). How-
ever, the limited resolution of the data precludes a
detailed interpretation of the conformational changes
that may be induced when Mg2+ is substituted by
Ni2+ (fig. S5). The open form of C3bB is charac-
terized by new interactions between the SP domain
of FB and the MG2 and CUB domains of C3b,
which result from the large rotation of the SP do-
main in FB. Mutations in FB at this new contact
bridges with Glu207 of helix aL and Glu446 of
helix a7 (15, 16) (Fig. 2C). However, in the open
C3bB structure concerted changes at the VWA-SP
interface result in exposure of FB’s scissile loop
(Fig. 2C). Rotation of the SP domain repositions
the VWA-SP linker (residues 447 to 453) and
elongates helix aL by two helical turns. This elon-
gation is stabilized by residues Ile217 and Val220,
which are docked into hydrophobic pockets of the
VWA domain that were formerly occupied by
residues Ile236 and Leu238 of the scissile loop (fig.
S7A). A new interface is formed between four
glutamates (residues 215 and 218 of aL and 422
and 424) of VWA and a cluster of four arginines
(residues 580, 583, 683, and 685) in the SP do-
main (fig. S7B). In combination, these changes
release Arg234 from its stabilizing interactions and
expose the scissile loop (residues 224 to 239),
REPORTS
end of helix aL, and to a lesser extent helix a7 and
the interactions with the succeeding SP domain.
FD binds FB in the open C3bB configuration
at the VWA-SP interface through a binding site
25 Å away from the catalytic center (Fig. 3A).
This interaction of FD induces a downward and
sideways shift (of ~4 Å) of the VWA helix a7 in
FB and unwinds the C-terminal part of the helix
(residues 442 to 446) (Fig. 2C). FD binds the
unwound helix a7 and its surroundings, yielding
a total interface area of ~700 Å2 dominated by 10
hydrogen bonds and nine salt bridges. In agree-
ment with the structural findings, mutations at this
site of FD [i.e., His133 (His146), Val203 (Val219), and
Arg157 (Arg170)] abolish cleavage of FB (Fig. 3, B
and C, and fig. S8) and render FD unable to amplify
complement response (Fig. 3, D to F). Similarly,
biotinylation of FD at residue Lys208 (Lys223) causes

Downloaded from www.sciencemag.org on December 23, 2010

site reduced the C3bB cleavage rate (Fig. 2B and
fig. S6), which shows that the interface observed
in the crystal structure of C3bB in the open
conformation is important for activity.
In both free FB and the closed CVFB com-
plex, the P1 residue of the scissile bond (Arg234)
is buried inside the VWA domain and forms salt
which appears disordered in the structure of C3bB
(fig. S7C). This mechanism differs markedly
from that observed for the I domains of integrins,
where a coupling of the MIDAS to helix a7 is
implicated in the allosteric activation (18, 33). In
the case of FB, however, the conformational
changes involve predominantly the C-terminal
loss of activity (26), likely because biotin blocks
the binding of FD to FB. In addition, Glu446 of
FB, which plays an important role in binding the
P1-Arg234 in the closed form, contacts His133
(His146) of FD in the C3bBD* complex, and mu-
tation of this residue (i.e., FB E446A) strongly
reduces convertase formation (fig. S9). This contact

Fig. 1. Structures of C3bB and C3bBD*. (A) Overall structure of C3bB (left)
and C3bBD* (right). C3b is shown as gray transparent surface; FB is shown as
orange (Ba), green (VWA), and blue (SP) cartoons; and FD is represented as a
magenta cartoon. The black spheres highlight the metal ions (Ni2+ for C3bB,
Mg2+ for C3bBD*) at the MIDAS site. (B) Opened view of the footprint of the
C3b-FB interaction, highlighting the domains of FB on the C3b surface (top)
and the domains of C3b on FB (bottom). (C) Opened view of the footprint of the
FB-FD interaction, highlighting the domains of FB on the FD surface (left) and
the single domain of FD on FB (right). The scheme indicates the domain
compositions and color codes of C3b, FB, and FD used in (B) and (C).

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1817

 REPORTS
site in FD is formed by four loops: residues 132
to 135 (145 to 149), 155 to 159 (169 to 171), 173
to 176 (185 to 188), and 203 to 209 (220 to 224).
In coagulation factor VIIa, loop 155 to 159 (169
to 171) is part of the zymogen activation domain
(i.e., tissue-factor binding site). Loops 132 to 135
(145 to 149), 173 to 176 (185 to 188), and 203 to
209 (220 to 224) and the self-inhibitory loop of
FD overlap with the Na+ ion–binding region in
thrombin (34, 35). In FD, this exosite is critical for
binding of the open form of FB in C3bB.
In the C3bBD* structure, the self-inhibitory
loop of FD is rearranged, resulting in an active
configuration of the catalytic triad in FD (Fig. 4A
and fig. S10, A and B). In contrast, the structure of
the unbound form of the corresponding FD mutant
S183A (S195A), determined at 1.2 Å resolution
(29), shows a self-inhibited state identical to the
either structurally or dynamically, of FD through
binding of FB at the exosite is not supported by the
present data. Instead, the rearrangements in FD are
restricted to the self-inhibitory loop [residues 196
to 202 (212 to 218)]. In the self-inhibited state,
Arg202 (Arg218) of FD forms a salt bridge with
Asp177 (Asp189) at the bottom of the S1 pocket,
whereas the same residue points outward in C3bBD*
(Fig. 4A). Mutation R202A (R218A) in FD in-
creased the activity against peptides threefold
(fig. S11A) because of rearrangements in the self-
inhibitory loop that induce the active conforma-
tion of the catalytic triad (fig. S10A) (25). In contrast,
the same R202A (R218A) mutation reduced
cleavage of FB to ~20% and thus reduced the
convertase formation rate (fig. S11, B to D), indicat-
ing a role for Arg202 (Arg218) in substrate specificity,
as predicted earlier (23, 26, 36–38). However,
apparent poor binding of this loop to FD is con-
sistent with the observed low activity of FD toward
short peptides even when the self-inhibitory loop
is displaced (25). Starting from residues 223 and
240 that are visible in the electron density map,
we could model the scissile loop of FB (224 to
239) in an extended fashion, placing residues 233
to 235 into the S2, S1, and S1′ binding pockets of
FD. This modeling places the P5-residue Glu230
in the vicinity of the exposed Arg202 (Arg218) of
FD (Fig. 4B). Mutating Glu230 in FB (E230A)
reduced the cleavage rate, indicating a putative
interaction between Arg202 (Arg218) in FD and Glu230
in FB. (fig. S11B). Overall, we interpreted the
specificity and activity of FD for cleaving FB in
C3bB as primarily determined by its exosite-
mediated binding to the open form of FB (with
nanomolar affinity, fig. S3C), which allows the

Downloaded from www.sciencemag.org on December 23, 2010

wild-type enzyme (22, 36) (fig. S10, A and B). We
observe no induced structural changes at the exosite
(fig. S10C). The C3bBD* structure and all eight
available structures of free FD and mutants thereof
(which occur in three different crystal forms) show
well-defined and virtually identical conformations
of the exosite loops. Thus, an induced activation,
published data showed complete inactivity for FD
mutant R202D-V203G (R218D-V219G) toward
both peptides and FB (37). We mimicked this by
using a R202G (R218G) mutant of FD, which
yielded almost no activity (fig. S11B).
Unexpectedly, we observed no electron den-
sity for the scissile loop of FB in C3bBD*. The
comparatively weaker interacting scissile loop of
FB to induce activation of FD.
Lastly, cleavage of FB releases the pro-
peptide segment Ba from the complex, inducing
additional structural changes in the residual Bb
fragment that remains bound to C3b (13). In this
step, the helix aL of the Ba fragment is removed

Fig. 2. Comparison between the closed and open states in the pro-convertase
C3bB. (A) Surface representation of the closed (CVFB, left) (16) and open (C3bB,
right) conformations of the pro-convertase. Colors are the same as in Fig. 1A. Red
triangles indicate the position of the catalytic site of FB during the conforma-
tional changes. (B) FD-mediated cleavage of C3bB performed by using SDS–
polyacrylamide gel electrophoresis (PAGE) shows that mutations (28) in FB located
at the interface between the SP and CUB domains in the open C3bB pro-
convertase lower convertase formation rates. The histogram shows the relative pro-
convertase activation rates compared with those of wild-type FB (see also fig. S6).
Error bars represent deviations from the mean observed in multiple experiments
(n > 3). (C) Cartoon diagram of the VWA domain of FB, highlighting conformational
changes in the transition from the closed (left) to the open state of the pro-
convertase in absence (center) or in presence of FD (right). The aL helix is colored in
orange and the a7 helix in green. The putative orientation of the loop containing
the scissile bond of FB is shown with a dashed line. The positions of the C-a atoms
located at the N-termini of aL and a7 helix are shown as spheres.

1818 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


from its groove in the VWA domain, and helix
a7 takes its place (fig. S7A). This results in a
canonical arrangement of the VWA domain with
the MIDAS-a7 presumably arranged in a high-
affinity ligand binding configuration. Similar to the
SP rearrangement in the pro-convertase, rearrange-
ment of the a7 helix affects the VWA-SP linker,
resulting in a rotation of ~140° for the SP domain
REPORTS
(fig. S12). As derived from the crystal structure of
the SCIN-inhibited convertase C3bBb (13), this
arrangement allows cleavage of the substrate C3
thereby amplifying the complement response.

Downloaded from www.sciencemag.org on December 23, 2010

Fig. 3. Analysis of FD exosite. (A) Surface diagram of FD highlighting its exosite in
blue and its catalytic site in red (28). Exosite mutations are shown in yellow, whereas
the biotinylation site associated with FD inactivation (26) is shown in green. (B) FD-
mediated cleavage of C3bB pro-convertases monitored by SPR; FD mutants were
injected onto surface-bound C3bB, and the cleavage activity was compared with that
of wild type (wt). A drop in SPR response upon FD injection corresponds to the removal
of the Ba fragment. Wild-type FD was injected at the end of each experiment to ensure
cleavage sensitivity. (C) FD-mediated cleavage of C3bB performed by using SDS-PAGE
with FD exosite mutants. The histogram shows the relative pro-convertase activation
rates compared with those of wild-type FD. Error bars represent deviations from the
mean observed in multiple experiments (n > 3). (D) Effect of FD exosite mutations on
the formation and decay of the C3bBb convertase monitored by SPR; wild-type FB was
premixed with various FD mutants and injected over immobilized C3b. RU, resonance
units. (E) Reconstitution of complement activity in FD-depleted plasma using FD
mutants determined as C3b deposition on the enzyme-linked immunosorbent assay
(ELISA) plate. O.D., optical density. (F) Hemolytic activity assays using FD-depleted
plasma reconstituted with different mutants of FD: Lysis of rabbit erythrocytes was
monitored by colorimetry and compared with 100% haemolysis in water.

Fig. 4. Analysis of FD catalytic site. (A) Conformational changes observed in FD side chain of His41 (His57) to the catalytic conformation. (B) Zoomed view of FD
catalytic site in C3bBD* structure. Superposition of the structure of FD S183A catalytic site with modeled FB scissile bond loop bound (dark gray). The model
(S195A) from C3bBD* (magenta) with wild-type free FD (green, PDB ID 1DSU) highlights the putative interaction between Glu230 of FB and Arg202 (Arg218) of
(22, 28) showing the displacement of the self-inhibitory loop and flipping of the FD and the P1 residue Arg234 making a salt bridge with Asp177 (Asp189).

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1819

 REPORTS
Our data reveal a highly concerted and specific
activation mechanism based on cofactor-dependent
and substrate-induced proteolysis, which provides
an important “double-safety” catch to restrict com-
plement amplification to C3b-tagged target cells.
These detailed mechanistic insights into the central
and transient, and therefore low-abundance, C3bBD
complex of complement activation offer oppor-
tunities for the development of new therapeutics
to fight complement-mediated diseases (39) such
as age-related macular degeneration, atypical he-
molytic uremic syndrome, membranoprolifera-
tive glomerulonephritis, and chronic inflammations
that are associated with excessive or poorly con-
trolled activation of complement.
References and Notes
1. D. Ricklin, G. Hajishengallis, K. Yang, J. D. Lambris,
14. J. Wu et al., Nat. Immunol. 10, 728 (2009). 38. H. Jing et al., EMBO J. 18, 804 (1999).
15. F. J. Milder et al., Nat. Struct. Mol. Biol. 14, 224 (2007). 39. D. Ricklin, J. D. Lambris, Nat. Biotechnol. 25, 1265
16. B. J. Janssen et al., EMBO J. 28, 2469 (2009). (2007).
17. E. Torreira, A. Tortajada, T. Montes, S. Rodríguez de Córdoba, 40. Coordinates and structure factors of C3bB, C3bBD*,
O. Llorca, J. Immunol. 183, 7347 (2009). FD (S183A), and FD (R202A) have been deposited in the
18. T. A. Springer, Structure 14, 1611 (2006). Protein Data Bank (PDB) with accession numbers 2XWJ,
19. T. Xiao, J. Takagi, B. S. Coller, J. H. Wang, T. A. Springer, 2XWB, 2XW9, and 2XWA, respectively. We gratefully
Nature 432, 59 (2004). thank the European Synchrotron Radiation Facility (ESRF)
and the Swiss Light source (SLS) for the provision of
20. A. A. Bhattacharya, M. L. Lupher Jr., D. E. Staunton,
R. C. Liddington, Structure 12, 371 (2004). synchrotron radiation facilities and beamline scientists of
21. J. Hinshelwood, S. J. Perkins, J. Mol. Biol. 298, 135 (2000). the SLS, ESRF, and the European Molecular Biology
22. S. V. Narayana et al., J. Mol. Biol. 235, 695 (1994). Laboratory for assistance. This work was supported by a
23. H. Jing et al., J. Mol. Biol. 282, 1061 (1998). “Top” grant (700.54.304 to P.G.) by the Council for
Chemical Sciences of the Netherlands Organization for
24. C. M. Kam et al., J. Biol. Chem. 262, 3444 (1987).
Scientific Research (NWO-CW) and NIH grants
25. S. Kim, S. V. Narayana, J. E. Volanakis, J. Biol. Chem.
270, 24399 (1995). (AI030040, AI068730, AI072106, and GM062134 to
26. F. R. Taylor et al., Biochemistry 38, 2849 (1999). J.D.L.). Author contributions: F.F. expressed and purified
all FB and FD mutants, purified C3b, generated protein
27. Z. Fishelson, H. J. Müller-Eberhard, J. Immunol. 129,
2603 (1982). complexes, performed crystallization experiments,
28. Single-letter abbreviations for the amino acid residues collected diffraction data, and solved the structures;
are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; F.F. and P.G. analyzed the structures; J.W. and R.S.W.
helped with cloning and optimization of protein

Downloaded from www.sciencemag.org on December 23, 2010

Nat. Immunol. 11, 785 (2010).
2. M. J. Walport, N. Engl. J. Med. 344, 1058 (2001).
3. M. K. Pangburn, H. J. Müller-Eberhard, Biochem. J. 235,
723 (1986).
4. P. Gros, F. J. Milder, B. J. Janssen, Nat. Rev. Immunol. 8, 48
(2008).
5. Y. Xu, S. V. Narayana, J. E. Volanakis, Immunol. Rev.
180, 123 (2001).
6. J. R. Dunkelberger, W. C. Song, Cell Res. 20, 34 (2010).
7. S. K. Law, A. W. Dodds, Protein Sci. 6, 263 (1997).
8. P. J. Haas, J. van Strijp, Immunol. Res. 37, 161 (2007).
9. M. C. Carroll, Mol. Immunol. 41, 141 (2004).
10. H. J. Müller-Eberhard, Annu. Rev. Immunol. 4, 503
(1986).
11. M. van Lookeren Campagne, C. Wiesmann, E. J. Brown,
Cell. Microbiol. 9, 2095 (2007).
12. A. P. Sjöberg, L. A. Trouw, A. M. Blom, Trends Immunol.
30, 83 (2009).
13. S. H. Rooijakkers et al., Nat. Immunol. 10, 721 (2009).
G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro;
Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.
29. Materials and methods are available as supporting
material on Science Online.
30. B. J. Janssen, A. Christodoulidou, A. McCarthy,
J. D. Lambris, P. Gros, Nature 444, 213 (2006).
31. C. Wiesmann et al., Nature 444, 217 (2006).
32. E. Torreira, A. Tortajada, T. Montes, S. Rodríguez de Córdoba,
O. Llorca, Proc. Natl. Acad. Sci. U.S.A. 106, 882 (2009).
33. M. Shimaoka et al., Cell 112, 99 (2003).
34. P. E. Bock, P. Panizzi, I. M. Verhamme, J. Thromb.
Haemost. 5 (suppl. 1), 81 (2007).
35. B. C. Lechtenberg, D. J. Johnson, S. M. Freund,
J. A. Huntington, Proc. Natl. Acad. Sci. U.S.A. 107,
14087 (2010).
36. S. Kim, S. V. Narayana, J. E. Volanakis, J. Immunol. 154,
6073 (1995).
37. S. Kim, S. V. Narayana, J. E. Volanakis, Biochemistry 33,
14393 (1994).
expression and purification; F.F. performed kinetic
studies; D.R. and A.T. performed the SPR binding studies
and hemolytic assays; F.F., D.R., J.D.L., and P.G.
conceived the experiments; F.F. prepared the figures;
and F.F., D.R., and P.G. wrote the manuscript. P.G. and
J.D.L. co-supervised this work. Competing financial
interests: P.G., F.F., D.R., and J.D.L. are co-inventors of a
patent application titled “Structure of C3bB-factor D
complex and use for rational drug design.”
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1816/DC1
Materials and Methods
Figs. S1 to S12
Table S1
References
29 July 2010; accepted 4 November 2010
10.1126/science.1195821

chaperone system alter relationships between

Hsp90 and Environmental Stress genotypes and phenotypes under conditions of
environmental stress (5–8) and, in so doing, pro-
Transform the Adaptive Value vide at least two routes to the rapid evolution of
new traits: (i) Acting as a potentiator, Hsp90’s

of Natural Genetic Variation folding reservoir allows individual genetic var-

iants to immediately create new phenotypes; when
the reservoir is compromised, the traits previous-
Daniel F. Jarosz1 and Susan Lindquist1,2* ly created by potentiated variants disappear. (ii)
Acting as a capacitor, Hsp90’s excess chaperone
How can species remain unaltered for long periods yet also undergo rapid diversification? By linking capacity buffers the effects of other variants,
genetic variation to phenotypic variation via environmental stress, the Hsp90 protein-folding storing them in a phenotypically silent form; when
reservoir might promote both stasis and change. However, the nature and adaptive value of the Hsp90 reservoir is compromised, the effects
Hsp90-contingent traits remain uncertain. In ecologically and genetically diverse yeasts, we find such of these variants are released, allowing them to
traits to be both common and frequently adaptive. Most are based on preexisting variation, with create new traits (5). To date, however, only two
causative polymorphisms occurring in coding and regulatory sequences alike. A common temperature types of potentiated variants have been defined
stress alters phenotypes similarly. Both selective inhibition of Hsp90 and temperature stress increase (2, 6), and the nature of buffered variants remains
correlations between genotype and phenotype. This system broadly determines the adaptive value of completely enigmatic. Some buffered traits map
standing genetic variation and, in so doing, has influenced the evolution of current genomes. to specific chromosomal regions, suggesting a de-
pendence on preexisting genetic variation. But sim-
any vital proteins have difficulty reach- Hsp90 is thereby a “hub of hubs” in regulatory ilar phenotypes can be produced by epigenetic

M ing their final folds or are inherently

unstable when they do. To contend
with such problems, organisms employ protein-
remodeling factors and chaperones, including a
subset known as heat-shock proteins (Hsps) (1).
Unlike more general chaperones, Hsp90 special-
izes in folding metastable signal transducers (2)
and key components of multiprotein complexes.
These are hubs in interaction networks (3), and
circuits. Also unlike most chaperones, Hsp90 is
constitutively expressed at much higher levels
than required to fulfill its normal functions. The
Hsp90 chaperone system, then, constitutes a large
but highly specific protein-folding reservoir (4).
Environmental stresses can destabilize Hsp90
clients and produce additional unfolded substrates,
straining the capacity of this buffer. We have sug-
gested that these unusual features of the Hsp90
variation (9, 10) and transposon activation (11),
providing alternative explanations for their ap-
1
Whitehead Institute for Biomedical Research, 9 Cambridge
Center, Cambridge, MA 02142, USA. 2Department of Biology
and Howard Hughes Medical Institute, Massachusetts Institute
of Technology, 77 Massachusetts Avenue, Cambridge, MA
02139, USA.
*To whom correspondence should be addressed. E-mail:
lindquist_admin@wi.mit.edu

1820 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

 REPORTS
Our data reveal a highly concerted and specific
activation mechanism based on cofactor-dependent
and substrate-induced proteolysis, which provides
an important “double-safety” catch to restrict com-
plement amplification to C3b-tagged target cells.
These detailed mechanistic insights into the central
and transient, and therefore low-abundance, C3bBD
complex of complement activation offer oppor-
tunities for the development of new therapeutics
to fight complement-mediated diseases (39) such
as age-related macular degeneration, atypical he-
molytic uremic syndrome, membranoprolifera-
tive glomerulonephritis, and chronic inflammations
that are associated with excessive or poorly con-
trolled activation of complement.
References and Notes
1. D. Ricklin, G. Hajishengallis, K. Yang, J. D. Lambris,
14. J. Wu et al., Nat. Immunol. 10, 728 (2009). 38. H. Jing et al., EMBO J. 18, 804 (1999).
15. F. J. Milder et al., Nat. Struct. Mol. Biol. 14, 224 (2007). 39. D. Ricklin, J. D. Lambris, Nat. Biotechnol. 25, 1265
16. B. J. Janssen et al., EMBO J. 28, 2469 (2009). (2007).
17. E. Torreira, A. Tortajada, T. Montes, S. Rodríguez de Córdoba, 40. Coordinates and structure factors of C3bB, C3bBD*,
O. Llorca, J. Immunol. 183, 7347 (2009). FD (S183A), and FD (R202A) have been deposited in the
18. T. A. Springer, Structure 14, 1611 (2006). Protein Data Bank (PDB) with accession numbers 2XWJ,
19. T. Xiao, J. Takagi, B. S. Coller, J. H. Wang, T. A. Springer, 2XWB, 2XW9, and 2XWA, respectively. We gratefully
Nature 432, 59 (2004). thank the European Synchrotron Radiation Facility (ESRF)
and the Swiss Light source (SLS) for the provision of
20. A. A. Bhattacharya, M. L. Lupher Jr., D. E. Staunton,
R. C. Liddington, Structure 12, 371 (2004). synchrotron radiation facilities and beamline scientists of
21. J. Hinshelwood, S. J. Perkins, J. Mol. Biol. 298, 135 (2000). the SLS, ESRF, and the European Molecular Biology
22. S. V. Narayana et al., J. Mol. Biol. 235, 695 (1994). Laboratory for assistance. This work was supported by a
23. H. Jing et al., J. Mol. Biol. 282, 1061 (1998). “Top” grant (700.54.304 to P.G.) by the Council for
Chemical Sciences of the Netherlands Organization for
24. C. M. Kam et al., J. Biol. Chem. 262, 3444 (1987).
Scientific Research (NWO-CW) and NIH grants
25. S. Kim, S. V. Narayana, J. E. Volanakis, J. Biol. Chem.
270, 24399 (1995). (AI030040, AI068730, AI072106, and GM062134 to
26. F. R. Taylor et al., Biochemistry 38, 2849 (1999). J.D.L.). Author contributions: F.F. expressed and purified
all FB and FD mutants, purified C3b, generated protein
27. Z. Fishelson, H. J. Müller-Eberhard, J. Immunol. 129,
2603 (1982). complexes, performed crystallization experiments,
28. Single-letter abbreviations for the amino acid residues collected diffraction data, and solved the structures;
are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; F.F. and P.G. analyzed the structures; J.W. and R.S.W.
helped with cloning and optimization of protein

Downloaded from www.sciencemag.org on December 23, 2010

Nat. Immunol. 11, 785 (2010).
2. M. J. Walport, N. Engl. J. Med. 344, 1058 (2001).
3. M. K. Pangburn, H. J. Müller-Eberhard, Biochem. J. 235,
723 (1986).
4. P. Gros, F. J. Milder, B. J. Janssen, Nat. Rev. Immunol. 8, 48
(2008).
5. Y. Xu, S. V. Narayana, J. E. Volanakis, Immunol. Rev.
180, 123 (2001).
6. J. R. Dunkelberger, W. C. Song, Cell Res. 20, 34 (2010).
7. S. K. Law, A. W. Dodds, Protein Sci. 6, 263 (1997).
8. P. J. Haas, J. van Strijp, Immunol. Res. 37, 161 (2007).
9. M. C. Carroll, Mol. Immunol. 41, 141 (2004).
10. H. J. Müller-Eberhard, Annu. Rev. Immunol. 4, 503
(1986).
11. M. van Lookeren Campagne, C. Wiesmann, E. J. Brown,
Cell. Microbiol. 9, 2095 (2007).
12. A. P. Sjöberg, L. A. Trouw, A. M. Blom, Trends Immunol.
30, 83 (2009).
13. S. H. Rooijakkers et al., Nat. Immunol. 10, 721 (2009).
G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro;
Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.
29. Materials and methods are available as supporting
material on Science Online.
30. B. J. Janssen, A. Christodoulidou, A. McCarthy,
J. D. Lambris, P. Gros, Nature 444, 213 (2006).
31. C. Wiesmann et al., Nature 444, 217 (2006).
32. E. Torreira, A. Tortajada, T. Montes, S. Rodríguez de Córdoba,
O. Llorca, Proc. Natl. Acad. Sci. U.S.A. 106, 882 (2009).
33. M. Shimaoka et al., Cell 112, 99 (2003).
34. P. E. Bock, P. Panizzi, I. M. Verhamme, J. Thromb.
Haemost. 5 (suppl. 1), 81 (2007).
35. B. C. Lechtenberg, D. J. Johnson, S. M. Freund,
J. A. Huntington, Proc. Natl. Acad. Sci. U.S.A. 107,
14087 (2010).
36. S. Kim, S. V. Narayana, J. E. Volanakis, J. Immunol. 154,
6073 (1995).
37. S. Kim, S. V. Narayana, J. E. Volanakis, Biochemistry 33,
14393 (1994).
expression and purification; F.F. performed kinetic
studies; D.R. and A.T. performed the SPR binding studies
and hemolytic assays; F.F., D.R., J.D.L., and P.G.
conceived the experiments; F.F. prepared the figures;
and F.F., D.R., and P.G. wrote the manuscript. P.G. and
J.D.L. co-supervised this work. Competing financial
interests: P.G., F.F., D.R., and J.D.L. are co-inventors of a
patent application titled “Structure of C3bB-factor D
complex and use for rational drug design.”
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1816/DC1
Materials and Methods
Figs. S1 to S12
Table S1
References
29 July 2010; accepted 4 November 2010
10.1126/science.1195821

chaperone system alter relationships between

Hsp90 and Environmental Stress genotypes and phenotypes under conditions of
environmental stress (5–8) and, in so doing, pro-
Transform the Adaptive Value vide at least two routes to the rapid evolution of
new traits: (i) Acting as a potentiator, Hsp90’s

of Natural Genetic Variation folding reservoir allows individual genetic var-

iants to immediately create new phenotypes; when
the reservoir is compromised, the traits previous-
Daniel F. Jarosz1 and Susan Lindquist1,2* ly created by potentiated variants disappear. (ii)
Acting as a capacitor, Hsp90’s excess chaperone
How can species remain unaltered for long periods yet also undergo rapid diversification? By linking capacity buffers the effects of other variants,
genetic variation to phenotypic variation via environmental stress, the Hsp90 protein-folding storing them in a phenotypically silent form; when
reservoir might promote both stasis and change. However, the nature and adaptive value of the Hsp90 reservoir is compromised, the effects
Hsp90-contingent traits remain uncertain. In ecologically and genetically diverse yeasts, we find such of these variants are released, allowing them to
traits to be both common and frequently adaptive. Most are based on preexisting variation, with create new traits (5). To date, however, only two
causative polymorphisms occurring in coding and regulatory sequences alike. A common temperature types of potentiated variants have been defined
stress alters phenotypes similarly. Both selective inhibition of Hsp90 and temperature stress increase (2, 6), and the nature of buffered variants remains
correlations between genotype and phenotype. This system broadly determines the adaptive value of completely enigmatic. Some buffered traits map
standing genetic variation and, in so doing, has influenced the evolution of current genomes. to specific chromosomal regions, suggesting a de-
pendence on preexisting genetic variation. But sim-
any vital proteins have difficulty reach- Hsp90 is thereby a “hub of hubs” in regulatory ilar phenotypes can be produced by epigenetic

M ing their final folds or are inherently

unstable when they do. To contend
with such problems, organisms employ protein-
remodeling factors and chaperones, including a
subset known as heat-shock proteins (Hsps) (1).
Unlike more general chaperones, Hsp90 special-
izes in folding metastable signal transducers (2)
and key components of multiprotein complexes.
These are hubs in interaction networks (3), and
circuits. Also unlike most chaperones, Hsp90 is
constitutively expressed at much higher levels
than required to fulfill its normal functions. The
Hsp90 chaperone system, then, constitutes a large
but highly specific protein-folding reservoir (4).
Environmental stresses can destabilize Hsp90
clients and produce additional unfolded substrates,
straining the capacity of this buffer. We have sug-
gested that these unusual features of the Hsp90
variation (9, 10) and transposon activation (11),
providing alternative explanations for their ap-
1
Whitehead Institute for Biomedical Research, 9 Cambridge
Center, Cambridge, MA 02142, USA. 2Department of Biology
and Howard Hughes Medical Institute, Massachusetts Institute
of Technology, 77 Massachusetts Avenue, Cambridge, MA
02139, USA.
*To whom correspondence should be addressed. E-mail:
lindquist_admin@wi.mit.edu

1820 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


pearance. Further, the adaptive value of buffered Rad and GdA produced very similar changes
traits remains untested. (table S3), confirming that they derived from
To broadly determine the adaptive value of Hsp90 inhibition, not off-target drug effects.
Hsp90’s effects on the relationship between geno- The reproducible character of the traits
type and phenotype, we examined 102 genetically suggests that they arose from preexisting varia-
diverse strains of Saccharomyces cerevisiae, from tion rather than de novo mutation. To investigate
soil, fruit, wine, sake, beer, infected human pa- quantitative trait loci (QTLs) that might underlie
tients, etc. (table S1). We measured growth in 100 such traits, we analyzed 104 densely mapped
conditions, including alternative carbon sources, haploid progeny from a cross between BY4716
oxidative stressors, antifungals, DNA-damaging (BY), a common laboratory strain originally ob-
agents, osmotic stressors, and small molecules tained from a rotten fig, and RM11-1a (RM), a
that perturb varied cellular processes (fig. S1 and vineyard isolate (14). The coding sequences of
table S2). Using two chemically unrelated Hsp90 these strains diverge by approximately 0.5%. Most
inhibitors, radicicol (Rad) and geldanamycin A genes were polymorphic in this cross but, im-
(GdA) (12, 13), we then determined the effects portantly, Hsp90 was not. Our results agreed well
of reducing the Hsp90 reservoir. We used con- with earlier studies under 23 overlapping conditions
centrations that did not induce a general stress (15), demonstrating the robust consequences of
response (fig. S2) or inhibit the growth of any the variation segregating in this cross.
strain in standard medium. However, every strain Again, we reduced the Hsp90 reservoir, and
exhibited substantial changes in growth under again, this strongly affected growth in specific strains
specific conditions. These varied widely and were in different conditions [Fig. 1B, fig. S3, and support-
sometimes positive, sometimes negative (Fig. 1A). ing online material (SOM)]. With the genetic diver-
A been expected by chance. As previously sug-
Brewing Clinical Fruit Soil Wine
1 23 3
A
Log2 growth change
B
C 2
Conditions
D
E and 15% involved three or more.
F 1
G
H As expected in such analyses, the QTLs en-
I 0
J
K
L
M
-1
N
O -2
-3
B BY x RM meiotic segregants
Conditions
C D
1 ing frame (ORF) of NFS1 (Fig. 2A).
1
2 Nfs1 is a cysteine desulfurase that acts as a
2 3
4 4
3
0.5 µM Rapamycin (metabolic perturbation) 1 mM Deoxycholate (membrane stress)
E confer rapamycin sensitivity. Furthermore, Nfs1
2 F
3
1
4 4
12 3
25 mM Hydroxyurea (replication stress) 5 mM CDNB (oxidative stress)
Fig. 1. Reducing the Hsp90 reservoir creates diverse phenotypes. Representative growth changes elicited by
Hsp90 inhibition. The scale bar indicates log2 of the ratio of growth, with and without 5 mM Rad, in each
condition for (A) wild strains and (B) BY × RM progeny. (C to F) Examples of rank-ordered growth distributions
after 64 hours of growth of BY × RM progeny with (orange bars) and without (gray bars) 5 mM Rad.
www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010
REPORTS
sity present in just these two parents, the patterns
of growth changes in their progeny were highly
diverse (Fig. 1C-F), establishing the independent
heritable segregation of many Hsp90-contingent
alleles. No QTLs influenced growth in 5 mM Rad
or GdA alone: Variations in drug pumps or detox-
ification pathways did not confound the analysis.
In three conditions (trichostatin A, iodo-
acetate, and chlorhexidine), we could not map
Hsp90-contingent QTLs. This might reflect ep-
igenetic effects or the segregation of too many
underlying variants (SOM). The vast majority of
Hsp90-contingent phenotypes, however, could
be mapped (table S5 and fig. S4). The QTLs re-
sponsible derived equally from BY and RM: Re-
laxed pressures of laboratory cultivation did not
skew the nature of the accumulated variation.
Hsp90 acted as a potentiator of variation al-
Downloaded from www.sciencemag.org on December 23, 2010
most as frequently as a capacitor: 44 previously
apparent QTLs disappeared when the reservoir
was reduced, and 63 previously silent QTLs be-
came apparent (tables S4 and S5). With a false
discovery rate of 0.05, at most three would have
gested (7, 8), reducing the Hsp90 reservoir pro-
duced genetically complex traits in a single step:
Fully one-third of the traits involved two QTLs,
compassed many polymorphic alleles. To identify
causative variants, we dissected four. In each case,
we used three otherwise diverse progeny that
carried BY sequence in the relevant region and
three that carried RM sequence. In these four sets
of six strains, we substituted every candidate gene,
one by one, with the allele of the opposite parent.
For the first QTL, Hsp90 acted as a capacitor
for rapamycin resistance latent in the RM ge-
nome. Segregants with both RM and BY se-
quence were sensitive to the compound. When
the Hsp90 reservoir was reduced, those with RM
sequence gained the ability to grow. The QTL
spanned eight genes, but all Hsp90-dependent
growth effects were conferred by the open read-
sulfur donor in tRNA thiolation (16). Rapamycin
targets the highly conserved TOR proteins, which
regulate growth in all eukaryotes, primarily via
the protein synthesis machinery (17). Other muta-
tions in this same tRNA modification pathway
function is known to depend on Hsp90 (18). Thus,
changes in the Hsp90 reservoir are logically linked
to polymorphisms in this region.
For the second QTL, Hsp90 acted as a poten-
tiator for deoxycholate (DOC) resistance con-
ferred by standing variation in the RM genome.
Segregants carrying RM sequence were DOC-
resistant, whereas those carrying BY sequence
were sensitive. When the Hsp90 reservoir was
reduced, strains carrying RM sequence lost re-
sistance. Allele replacements demonstrated that
this resistance arose entirely from the RM PDR8
ORF (Fig. 2B).
1821
 REPORTS

Downloaded from www.sciencemag.org on December 23, 2010

Fig. 2. Genetic dissection of Hsp90-contingent alleles. The growth of allele-
replacement strains with (solid bars) and without (open bars) 5 mM Rad is
normalized to that of the BY allele–replacement strain in each condition
without Rad. (A) QTLs conferring Hsp90-buffered rapamycin resistance, due to
the RM NFS1 allele (44 hours). (B) QTLs conferring Hsp90-potentiated DOC
resistance, due to the RM PDR8 allele. (C). QTLs conferring Hsp90-buffered
HU resistance, due to the BY MEC1 allele (25 hours). Hsp90-potentiated
resistance to UV irradiation was due to the same allele (20 J/m2; 25 hours after
irradiation). (D) QTLs conferring CDNB resistance, due to polymorphisms in
the 3′ untranslated region (UTR) of RM NDI1 (44 hours). Overexpression of BY
NDI1 rescues CDNB toxicity. Error bars in the entire figure represent the
standard deviation of three biological replicates.

Fig. 3. Environmental stress recapitulates phenotypic effects of Hsp90 PDR8 (1 mM DOC; 80 hours). (C) MEC1 (25 mM HU; 25 hours) (D) RM
inhibition. Calculations and symbols are as in Fig. 2. Growth of allele- intergenic region between NDI1 and GTR1 (5 mM CDNB; 44 hours). Because
replacement strains at 23°C, 39°C, or after a deletion of one of the Hsp90 the HSP82 deletion reduces Hsp90 function more than does 5 mM Rad, it often
genes, HSP82, at 23°C, is shown. (A) NFS1 (0.5 mM rapamycin; 44 hours). (B) creates stronger phenotypes.

DOC facilitates fat emulsification in the in-
testine and acts as an antimicrobial agent (19).
PDR8 encodes a transcription factor not known
to depend on Hsp90. To determine whether RM
polymorphisms caused Pdr8 to become an Hsp90
client, we examined other Pdr8-dependent pheno-
types: growth in NaCl, hygromycin B, and LiCl
(20). Reducing Hsp90 did not affect any of these
(fig. S6), suggesting that RM Prd8 does not require
Hsp90 for function. More likely, RM polymor-
phisms exert their effects via Hsp90’s interaction
with another, DOC-specific element of Pdr8’s
circuitry.
For the third QTL (Fig. 2C), Hsp90 acted as a
capacitor for hydroxyurea (HU) resistance latent
in the BY genome. Segregants carrying BY se-
quence were initially more sensitive than those
carrying RM sequence. Reducing the Hsp90 re-
servoir increased the resistance of segregants
carrying BY sequence but not RM sequence. This
trait proved to be conferred by MEC1.
HU reduces intracellular deoxynucleotide tri-
phosphate concentrations, eliciting replication stress
(21). Mec1 coordinates multicomponent repair
and checkpoint pathways that differ for different
damage responses (22). A major QTL that con-
ferred resistance to ultraviolet (UV) radiation also
proved to map to MEC1. In this case, however,
Hsp90 acted as a potentiator. The UV resistance
of strains carrying the BY allele was lost when
the Hsp90 reservoir was reduced. Because Hsp90
inhibition affected two Mec1 functions in differ-
ent ways, these results suggest that Mec1 is an
Hsp90 client, whose partitioning between di-
verse complexes is affected by Hsp90-contingent
polymorphisms.
For the fourth QTL, Hsp90 acted as a ca-
pacitor for CDNB (1-chloro-2,4-nitrobenzene)
resistance latent in the RM genome. Segre-
gants with either RM or BY sequence were sen-
sitive to this oxidative stressor. When the Hsp90
reservoir was compromised, those with RM se-
quence gained the ability to grow. Allele replace-
ments established RM NDI1 as the causative
variant, but here, the polymorphisms resided in

1822 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

 REPORTS
YJM326 (clinical) YJM326 (clinical)
YJM436 (clinical)
YJM428 (clinical)
YJM653 (clinical) YJM428 (clinical)
YJM436 (clinical)
YJM653 (clinical)
YJM421 (clinical) I14 (wine) YJM421 (clinical)
SK1 (soil) SK1 (soil) I14 (wine)
Y9 (sake) T73 (wine) Y9 (sake)
Y12 (sake) Y12 (sake) T73 (wine)
WE372 (wine) WE372 (wine)
Clustering by genotype Clustering by phenotype with
reduced Hsp90 reservoir
I14 (wine)

Y12 (Sake) Fig. 4. Hsp90 inhibition and environmental stress improve the
corrrelations between genotype and phenotype. Phylogenetic cluster-
Y9 (sake)
ing is derived from (27) and (25). Phenotypic clustering is described in
SK1 (soil) the SOM.

Downloaded from www.sciencemag.org on December 23, 2010

YJM428 (clinical) YJM326 (clinical)

YJM653 (clinical) YJM436 (clinical)

YJM421 (clinical)
T73 (wine)
WE372 (wine)
Clustering by phenotype

the 3′ untranslated region (Fig. 2D) rather than
in the ORF.
NDI1 encodes an oxidoreductase that defends
against oxidative stresses (23). We found that
CDNB normally had little effect on NDI1 mRNA
levels. But when the Hsp90 reservoir was re-
duced, transcripts produced by NDI1 in response
to CDNB stress increased by ~100-fold in seg-
regants with RM relative to those with BY se-
quence. Increased NDI1 transcripts fully explained
the phenotype: Forced overexpression of the nor-
mally ineffective BY allele (using a Gal1 pro-
moter) was sufficient to confer CDNB resistance
(Fig. 2D).
How might changes in Hsp90 affect the ex-
pression of genetic variation in nature? Hsp90 is
induced by environmental stress (4). We have
postulated that this increase is sometimes insuf-
ficient to maintain the folding reservoir, changing
the manifestation of genetic variation (24). In-
deed, all four alleles analyzed above were affected
by a simple temperature stress (growth at 39°C) in
the same manner as by Hsp90 inhibition (Fig. 3A-
D). Moreover, the same phenotypes were elicited
by genetic deletion of one of two Hsp90 alleles,
confirming that they are due to changes in Hsp90
function. Far more broadly, we find that even with
the abundant genetic diversity present in the wild
strains, the effects of temperature on phenotypic
transitions were similar to those of Rad and GdA
(Pearson correlation ~0.61 and ~0.56, respectively,
and see SOM).
We took advantage of the fact that 48 of these
strains have been sequenced to ask whether their
genomes carry an impress of Hsp90’s selective
forces. As previously reported in other strains and
circumstances (25), across the ~100,000 poly-
morphisms present here with the 100 growth con-
ditions we used, the correlation between genoptye
and phenotype was relatively weak (Spearman cor-
relation ~0.35) in the absence of Hsp90 inhibition.
Similar strains often had divergent phenotypes,
and divergent genotypes often produced similar
phenotypes.
The correlation between genotype and phe-
notype became much stronger when the Hsp90
reservoir was reduced (Spearman correlation
~0.54; Fig. 4). Ten million random data permu-
tations did not produce a single increase of such
magnitude (P < 10−7). This transition was evident
across diverse ecological niches. A simple in-
crease in growth temperature had a similar effect
(Spearman correlation ~0.48). It is difficult to
imagine how environmental stress in general, and
Hsp90 in particular, could have such a strong
impact on genotype-phenotype correlations un-
less it had acted though the evolutionary history
of these strains to influence the retention of a
broad swath of genetic variation.
Our hypothesis that Hsp90 plays a role in
evolutionary processes remains controversial
because of a paucity of hard evidence (26). Here
we establish that Hsp90 operates on roughly 20%
of the preexisting genetic variation in S. cerevisiae
to both preserve phenotypic robustness and pro-
vide a broad conduit to diversification. Further,
environmental stress creates a dynamic interface
for transitioning between these effects in a manner
that has left an impress on current genomes. Half
of the traits buffered by Hsp90 and half poten-
tiated by it had beneficial effects on growth; the
other half were detrimental. What might maintain
such contrasting adaptive effects? Many proteins
in regulatory hubs are metastable, essential for life,
and constitutively dependent on Hsp90. The need
to preserve these functions during environmental
stress might provide all the selective pressure
needed to maintain this protein-folding reservoir.
The accumulation of new Hsp90-contingent al-
leles might simply be an inevitable consequence
of its existence. Once established, however, the
capacity of the reservoir to facilitate the appear-
ance of new traits—evolvability—might have pro-
vided an additional selective advantage. Theory
holds that natural selection is unable to sustain
mechanisms for evolvability because genetic re-
combination would inevitably separate evolvabil-
ity genes from the alleles on which they act (5, 24).
Negating this objection, Hsp90-contingent poly-
morphisms are dispersed throughout the genome;
loss of some through genetic reassortment would
be balanced by the gain of others.
A particular advantage of the Hsp90 system is
that it provides a route to genetically complex
traits in a single step, via combinatorial gain and
loss of phenotypic variation in response to envi-
ronmental stress. Under selective pressure, mul-
tiple mechanisms could lead to the fixation of
such traits (5, 24). In Drosophila, at least, Hsp90
can also create new traits by affecting epigenetic
variation (10) and transposon-mediated mutagen-
esis (11), and it probably affects genome stability
by other mechanisms as well (5). The strength of
the Hsp90 buffer and the wealth of mechanisms
by which it creates heritable new traits in re-
sponse to environmental change may help to ex-
plain two long-puzzling features of evolution: the
stability of phenotypes over long periods despite
the accumulation of genetic variation and their
rapid appearance of heritable new phenotypes in
response to changing environments.
References and Notes
1. S. Lindquist, E. A. Craig, Annu. Rev. Genet. 22, 631 (1988).
2. L. Neckers, J. Biosci. 32, 517 (2007).

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1823

 REPORTS
3. Y. Gong et al., Mol. Syst. Biol. 5, 275 (2009).
4. K. A. Borkovich, F. W. Farrelly, D. B. Finkelstein,
J. Taulien, S. Lindquist, Mol. Cell. Biol. 9, 3919 (1989).
5. D. F. Jarosz, M. Taipale, S. Lindquist, Annu. Rev. Genet.
44, 189 (2010).
6. L. E. Cowen, S. L. Lindquist, Science 309, 2185 (2005).
7. C. Queitsch, T. A. Sangster, S. Lindquist, Nature 417, 618
(2002).
8. S. L. Rutherford, S. Lindquist, Nature 396, 336 (1998).
9. M. Tariq, U. Nussbaumer, Y. Chen, C. Beisel, R. Paro,
Proc. Natl. Acad. Sci. U.S.A. 106, 1157 (2009).
10. V. Sollars et al., Nat. Genet. 33, 70 (2002).
11. V. Specchia et al., Nature 463, 662 (2010).
12. S. V. Sharma, T. Agatsuma, H. Nakano, Oncogene 16,
2639 (1998).
13. L. Whitesell, E. G. Mimnaugh, B. De Costa, C. E. Myers,
L. M. Neckers, Proc. Natl. Acad. Sci. U.S.A. 91, 8324 (1994).
14. R. B. Brem, G. Yvert, R. Clinton, L. Kruglyak, Science 296,
752 (2002).
15. E. O. Perlstein, D. M. Ruderfer, D. C. Roberts,
S. L. Schreiber, L. Kruglyak, Nat. Genet. 39, 496 (2007).
16. S. Leidel et al., Nature 458, 228 (2009).
17. A. Ramanathan, S. L. Schreiber, J. Biol. 6, 3 (2007).
18. A. J. McClellan et al., Cell 131, 121 (2007).
19. T. Inagaki et al., Proc. Natl. Acad. Sci. U.S.A. 103,
3920 (2006).
20. I. Hikkel et al., J. Biol. Chem. 278, 11427 (2003).
21. E. Vitols, V. A. Bauer, E. C. Stanbrough,
Biochem. Biophys. Res. Commun. 41, 71 (1970).
22. A. M. Friedel, B. L. Pike, S. M. Gasser, Curr. Opin.
Cell Biol. 21, 237 (2009).
23. B. B. Seo, M. Marella, T. Yagi, A. Matsuno-Yagi, FEBS Lett.
580, 6105 (2006).
24. T. A. Sangster, S. L. Lindquist, C. Queitsch, Bioessays 26,
348 (2004).
25. G. Liti et al., Nature 458, 337 (2009).
26. T. Mitchell-Olds, C. A. Knight, Science 296, 2348
(2002).
27. J. Schacherer, J. A. Shapiro, D. M. Ruderfer, L. Kruglyak,
Nature 458, 342 (2009).
28. We are grateful to L. Kruglyak, E. Perlstein, L. Cowen,
S. Schreiber, A. Regev, I. Barassa, E. Louis, S. Dietzmann,
and F. Dietrich for materials, discussion, and help. S.L. is a
Howard Hughes Medical Institute (HHMI) investigator, and
D.F.J. is an HHMI fellow of the Damon Runyon Cancer
Research Foundation (DRG-1964-08). This work was
supported by grants from the Broad Institute, the G. Harold
and Leila Y. Mathers Foundation, and HHMI.
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1820/DC1
Materials and Methods
Figs. S1 to S6
Tables S1 to S5
References
22 July 2010; accepted 18 November 2010
10.1126/science.1195487

Downloaded from www.sciencemag.org on December 23, 2010

Hierarchical clustering plots of these data (table
Ectopic Expression of Germline S1) reveal three distinct clusters that include control
larval brains, mbt larval brain tumors, and cultured
Genes Drives Malignant Brain l(3)mbt ts1 tumors, respectively (fig. S1). From these
data, we identified 151 genes that were either over-
Tumor Growth in Drosophila expressed (n = 125) or underexpressed (n = 26) in
all three larval mbt tumor types compared to all
three controls (table S2). From this list, we removed
Ana Janic,1 Leire Mendizabal,1* Salud Llamazares,1 David Rossell,2 Cayetano Gonzalez1,3† those genes that were also up- (n = 23) or down-
regulated (n = 14) in larval brat neoplasms and,
Model organisms such as the fruit fly Drosophila melanogaster can help to elucidate the molecular hence, likely to encode functions generally required
basis of complex diseases such as cancer. Mutations in the Drosophila gene lethal (3) malignant for larval brain tumor growth. We refer to the ex-
brain tumor cause malignant growth in the larval brain. Here we show that l(3)mbt tumors pression levels of the remaining 102 up-regulated
exhibited a soma-to-germline transformation through the ectopic expression of genes normally genes as the mbt signature (MBTS) (table S3).
required for germline stemness, fitness, or longevity. Orthologs of some of these genes were also MBTS is notably enhanced in cultured mbt tumors
expressed in human somatic tumors. In addition, inactivation of any of the germline genes nanos, and can be used unequivocally to distinguish mbt
vasa, piwi, or aubergine suppressed l(3)mbt malignant growth. Our results demonstrate that tumors from other cultured malignant brain neo-
germline traits are necessary for tumor growth in this Drosophila model and suggest that plasms like lgl, mira, pros, pins, or brat (Fig. 1A
inactivation of germline genes might have tumor-suppressing effects in other species. and table S3). Individual MBTS genes, however,
are also up-regulated in some of these tumors.
he Drosophila tumor-suppressor gene le- the two Drosophila Retinoblastoma-family pro- The function of most MBTS genes remains

T thal (3) malignant brain tumor [l(3)mbt]

was identified as a temperature-sensitive
mutation that caused malignant growth in the
larval brain (1). Other l(3)mbt mutant alleles ob-
tained later show the same temperature-sensitive
phenotype (2). L(3)mbt’s closest homologs, Dro-
sophila Scm (Sex comb on midleg) and Sfmbt
(Scm-related gene containing four mbt domains),
encode Polycomb Group proteins (3). L3MBTL1,
the human homolog of Drosophila L(3)MBT (3),
is a transcriptional repressor (4) that is found in
a complex with core histones, heterochromatin
protein 1g (HP1g), and RB (Retinoblastoma pro-
tein) and can compact nucleosomes (5). Drosoph-
ila L(3)MBT is a substoichiometric component
of the dREAM-MMB complex, which includes
1 tively). Brains from homozygous white1118 (w1118),
Cell Division Group, Institute for Research in Biomedicine (IRB-
Barcelona), PCB, c/Baldiri Reixac 10-12, Barcelona, Spain.
2
Bioinformatics and Biostatistics Unit, IRB-Barcelona, PCB, c/Baldiri
Reixac 10-12, Barcelona, Spain. 3Institució Catalana de Recerca i
Estudis Avançats (ICREA), Passeig Lluís Companys 23, Barcelona,
Spain.
*Present address: Genomics Core Facility, Vall d'Hebrón Institute
of Oncology, Passeig Vall d'Hebron 119, 08035 Barcelona, Spain.
†To whom correspondence should be addressed. E-mail:
gonzalez@irbbarcelona.org
teins and the Myb-MuvB (MMB) complex (6).
Depletion of components of the dREAM/MMB
complex in Drosophila Kc cells by RNA inter-
ference results in genome-wide changes in gene
expression (7). These data strongly suggest that
l(3)mbt function might contribute to establishing
and maintaining certain differentiated states through
the stable silencing of specific genes (3, 7).
To identify the genes whose misexpression
might account for the growth of l(3)mbt tumors
(henceforth referred to as mbt tumors), we carried
out genome-wide gene expression profiling of
l(3)mbt E2 and l(3)mbt ts1 homozygous and trans-
heterozygous larval brains raised at restrictive
temperature (29°C). We also analyzed l(3)mbt ts1
tumors at the 1st, 5th, and 10th rounds of allograft
culture in adult flies (T1, T5, and T10, respec-
l(3)mbt E2, or l(3)mbt ts1 larvae raised at permis-
sive temperature (17°C) were used as controls.
For comparison, we also profiled larval brain ma-
lignant neoplasms caused by mutation in brain
tumor (brat) as well as allograft cultures at T1,T5,
and T10 of tumors caused by mutants in brat,
lethal giant larvae (lgl), miranda (mira), prospero
(pros), and partner of inscuteable (pins) (8).
unknown. However, a quarter of them (26 of 102)
are genes required in the germ line (Fig. 1B and
table S4A). For instance, nanos (nos), female
sterile(1)Yb ( fs(1)Yb), and zero population growth
(zpg) function in the establishment of the pole
plasm in the egg and cystoblasts differentiation
(9). The gonad-specific thioredoxins ThioredoxinT
(TrxT) and deadhead (dhd), giant nuclei (gnu),
corona (cona), hold'em (hdm), matotopetli (topi),
and the female germline-specific gTUB37C iso-
form function during oocyte differentiation, meio-
sis, and syncytial embryo development (10–15).
Also piwi, aubergine (aub), krimper (krimp), and
tejas (tej) are involved in the biogenesis of Piwi-
interacting RNAs (piRNAs) that protect germline
cells against transposable elements and viruses
(16, 17). Some of these genes also have functions
that are not germline related. For instance, some piwi
alleles display synthetic lethality (18), and nos is
required during nervous system development (19).
Driven by the high percentage of MBTS genes
that have germline functions, we looked for other
germline-related genes that do not meet the stringent
criteria applied to select the 102 MBTS genes,
but are overexpressed in mbt tumors (table S4B).
Among these we found the genes that encode the

1824 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

 REPORTS
3. Y. Gong et al., Mol. Syst. Biol. 5, 275 (2009).
4. K. A. Borkovich, F. W. Farrelly, D. B. Finkelstein,
J. Taulien, S. Lindquist, Mol. Cell. Biol. 9, 3919 (1989).
5. D. F. Jarosz, M. Taipale, S. Lindquist, Annu. Rev. Genet.
44, 189 (2010).
6. L. E. Cowen, S. L. Lindquist, Science 309, 2185 (2005).
7. C. Queitsch, T. A. Sangster, S. Lindquist, Nature 417, 618
(2002).
8. S. L. Rutherford, S. Lindquist, Nature 396, 336 (1998).
9. M. Tariq, U. Nussbaumer, Y. Chen, C. Beisel, R. Paro,
Proc. Natl. Acad. Sci. U.S.A. 106, 1157 (2009).
10. V. Sollars et al., Nat. Genet. 33, 70 (2002).
11. V. Specchia et al., Nature 463, 662 (2010).
12. S. V. Sharma, T. Agatsuma, H. Nakano, Oncogene 16,
2639 (1998).
13. L. Whitesell, E. G. Mimnaugh, B. De Costa, C. E. Myers,
L. M. Neckers, Proc. Natl. Acad. Sci. U.S.A. 91, 8324 (1994).
14. R. B. Brem, G. Yvert, R. Clinton, L. Kruglyak, Science 296,
752 (2002).
15. E. O. Perlstein, D. M. Ruderfer, D. C. Roberts,
S. L. Schreiber, L. Kruglyak, Nat. Genet. 39, 496 (2007).
16. S. Leidel et al., Nature 458, 228 (2009).
17. A. Ramanathan, S. L. Schreiber, J. Biol. 6, 3 (2007).
18. A. J. McClellan et al., Cell 131, 121 (2007).
19. T. Inagaki et al., Proc. Natl. Acad. Sci. U.S.A. 103,
3920 (2006).
20. I. Hikkel et al., J. Biol. Chem. 278, 11427 (2003).
21. E. Vitols, V. A. Bauer, E. C. Stanbrough,
Biochem. Biophys. Res. Commun. 41, 71 (1970).
22. A. M. Friedel, B. L. Pike, S. M. Gasser, Curr. Opin.
Cell Biol. 21, 237 (2009).
23. B. B. Seo, M. Marella, T. Yagi, A. Matsuno-Yagi, FEBS Lett.
580, 6105 (2006).
24. T. A. Sangster, S. L. Lindquist, C. Queitsch, Bioessays 26,
348 (2004).
25. G. Liti et al., Nature 458, 337 (2009).
26. T. Mitchell-Olds, C. A. Knight, Science 296, 2348
(2002).
27. J. Schacherer, J. A. Shapiro, D. M. Ruderfer, L. Kruglyak,
Nature 458, 342 (2009).
28. We are grateful to L. Kruglyak, E. Perlstein, L. Cowen,
S. Schreiber, A. Regev, I. Barassa, E. Louis, S. Dietzmann,
and F. Dietrich for materials, discussion, and help. S.L. is a
Howard Hughes Medical Institute (HHMI) investigator, and
D.F.J. is an HHMI fellow of the Damon Runyon Cancer
Research Foundation (DRG-1964-08). This work was
supported by grants from the Broad Institute, the G. Harold
and Leila Y. Mathers Foundation, and HHMI.
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1820/DC1
Materials and Methods
Figs. S1 to S6
Tables S1 to S5
References
22 July 2010; accepted 18 November 2010
10.1126/science.1195487

Downloaded from www.sciencemag.org on December 23, 2010

Hierarchical clustering plots of these data (table
Ectopic Expression of Germline S1) reveal three distinct clusters that include control
larval brains, mbt larval brain tumors, and cultured
Genes Drives Malignant Brain l(3)mbt ts1 tumors, respectively (fig. S1). From these
data, we identified 151 genes that were either over-
Tumor Growth in Drosophila expressed (n = 125) or underexpressed (n = 26) in
all three larval mbt tumor types compared to all
three controls (table S2). From this list, we removed
Ana Janic,1 Leire Mendizabal,1* Salud Llamazares,1 David Rossell,2 Cayetano Gonzalez1,3† those genes that were also up- (n = 23) or down-
regulated (n = 14) in larval brat neoplasms and,
Model organisms such as the fruit fly Drosophila melanogaster can help to elucidate the molecular hence, likely to encode functions generally required
basis of complex diseases such as cancer. Mutations in the Drosophila gene lethal (3) malignant for larval brain tumor growth. We refer to the ex-
brain tumor cause malignant growth in the larval brain. Here we show that l(3)mbt tumors pression levels of the remaining 102 up-regulated
exhibited a soma-to-germline transformation through the ectopic expression of genes normally genes as the mbt signature (MBTS) (table S3).
required for germline stemness, fitness, or longevity. Orthologs of some of these genes were also MBTS is notably enhanced in cultured mbt tumors
expressed in human somatic tumors. In addition, inactivation of any of the germline genes nanos, and can be used unequivocally to distinguish mbt
vasa, piwi, or aubergine suppressed l(3)mbt malignant growth. Our results demonstrate that tumors from other cultured malignant brain neo-
germline traits are necessary for tumor growth in this Drosophila model and suggest that plasms like lgl, mira, pros, pins, or brat (Fig. 1A
inactivation of germline genes might have tumor-suppressing effects in other species. and table S3). Individual MBTS genes, however,
are also up-regulated in some of these tumors.
he Drosophila tumor-suppressor gene le- the two Drosophila Retinoblastoma-family pro- The function of most MBTS genes remains

T thal (3) malignant brain tumor [l(3)mbt]

was identified as a temperature-sensitive
mutation that caused malignant growth in the
larval brain (1). Other l(3)mbt mutant alleles ob-
tained later show the same temperature-sensitive
phenotype (2). L(3)mbt’s closest homologs, Dro-
sophila Scm (Sex comb on midleg) and Sfmbt
(Scm-related gene containing four mbt domains),
encode Polycomb Group proteins (3). L3MBTL1,
the human homolog of Drosophila L(3)MBT (3),
is a transcriptional repressor (4) that is found in
a complex with core histones, heterochromatin
protein 1g (HP1g), and RB (Retinoblastoma pro-
tein) and can compact nucleosomes (5). Drosoph-
ila L(3)MBT is a substoichiometric component
of the dREAM-MMB complex, which includes
1 tively). Brains from homozygous white1118 (w1118),
Cell Division Group, Institute for Research in Biomedicine (IRB-
Barcelona), PCB, c/Baldiri Reixac 10-12, Barcelona, Spain.
2
Bioinformatics and Biostatistics Unit, IRB-Barcelona, PCB, c/Baldiri
Reixac 10-12, Barcelona, Spain. 3Institució Catalana de Recerca i
Estudis Avançats (ICREA), Passeig Lluís Companys 23, Barcelona,
Spain.
*Present address: Genomics Core Facility, Vall d'Hebrón Institute
of Oncology, Passeig Vall d'Hebron 119, 08035 Barcelona, Spain.
†To whom correspondence should be addressed. E-mail:
gonzalez@irbbarcelona.org
teins and the Myb-MuvB (MMB) complex (6).
Depletion of components of the dREAM/MMB
complex in Drosophila Kc cells by RNA inter-
ference results in genome-wide changes in gene
expression (7). These data strongly suggest that
l(3)mbt function might contribute to establishing
and maintaining certain differentiated states through
the stable silencing of specific genes (3, 7).
To identify the genes whose misexpression
might account for the growth of l(3)mbt tumors
(henceforth referred to as mbt tumors), we carried
out genome-wide gene expression profiling of
l(3)mbt E2 and l(3)mbt ts1 homozygous and trans-
heterozygous larval brains raised at restrictive
temperature (29°C). We also analyzed l(3)mbt ts1
tumors at the 1st, 5th, and 10th rounds of allograft
culture in adult flies (T1, T5, and T10, respec-
l(3)mbt E2, or l(3)mbt ts1 larvae raised at permis-
sive temperature (17°C) were used as controls.
For comparison, we also profiled larval brain ma-
lignant neoplasms caused by mutation in brain
tumor (brat) as well as allograft cultures at T1,T5,
and T10 of tumors caused by mutants in brat,
lethal giant larvae (lgl), miranda (mira), prospero
(pros), and partner of inscuteable (pins) (8).
unknown. However, a quarter of them (26 of 102)
are genes required in the germ line (Fig. 1B and
table S4A). For instance, nanos (nos), female
sterile(1)Yb ( fs(1)Yb), and zero population growth
(zpg) function in the establishment of the pole
plasm in the egg and cystoblasts differentiation
(9). The gonad-specific thioredoxins ThioredoxinT
(TrxT) and deadhead (dhd), giant nuclei (gnu),
corona (cona), hold'em (hdm), matotopetli (topi),
and the female germline-specific gTUB37C iso-
form function during oocyte differentiation, meio-
sis, and syncytial embryo development (10–15).
Also piwi, aubergine (aub), krimper (krimp), and
tejas (tej) are involved in the biogenesis of Piwi-
interacting RNAs (piRNAs) that protect germline
cells against transposable elements and viruses
(16, 17). Some of these genes also have functions
that are not germline related. For instance, some piwi
alleles display synthetic lethality (18), and nos is
required during nervous system development (19).
Driven by the high percentage of MBTS genes
that have germline functions, we looked for other
germline-related genes that do not meet the stringent
criteria applied to select the 102 MBTS genes,
but are overexpressed in mbt tumors (table S4B).
Among these we found the genes that encode the

1824 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


synaptonemal complex protein Crossover sup-
pressor on 3 of Gowen [C(3)G] and the cell cycle
kinase Pan gu (PNG), which interact with the
proteins encoded by the MBTS genes cona and
gnu, respectively (11, 13). The same applies to
Squash (SQU), Spindle-E (SPN-E), Maelstrom
(MAEL), and AGO3, components of the piRNA
machinery, which colocalize with other MBTS
proteins in nuage (16, 17).
To determine whether the mRNAs that we
found ectopically expressed in mbt tumors are
translated, we checked for protein expression with
a selected number of currently available antibodies.
Given the key role of VASA in the assembly of the
pole plasm and germline development (20), we
included it in this study, even though vasa mRNA
levels are not significantly increased in mbt tumors.
By Western blot, we confirmed that PIWI, AUB,
and VASA are ectopically expressed in mbt tumors
(Fig. 2A). Immunofluorescence studies also re-
vealed the ectopic expression in l(3)mbt ts1 brains
raised at 29°C of C(3)G, SQU, and VASA (Fig.
2B). These results show that some of the germline
genes ectopically expressed in mbt tumors are
translated. However, we have not been able to
confirm the expression of other proteins, including
MAEL, ORB, BAM, GNU, and TOPI, which sug-
Fig. 1. Gene expression profile of mbt tumors. (A) Heatmap of expression levels of the genes that are most
significantly up-regulated in larval brain mbt tumors (mbt tumor signature, MBTS). Samples include wild-type
larval brains, larval mbt tumors, and different types of larval brain malignant neoplasms in the 1st, 5th, and
10th rounds of allograft culture (T1, T5, and T10, respectively). (B) MBTS genes with known germline functions.
www.sciencemag.org
gests that, possible technical problems aside, either
the corresponding mRNAs are not translated or
these proteins might be unstable in such an ectopic
environment. The expression of VASA, by con-
trast, suggests that other mRNAs whose levels are
not appreciably increased in mbt tumors might
actually be ectopically translated.
Prompted by the expression in l(3)mbt ts1
brains of several genes involved in the biogenesis
and regulation of piRNAs, we sequenced 23- to
30-nucleotide RNAs from l(3)mbt ts1 larval brain
tumors and from wild-type brains and ovaries.
We found 117 known piRNAs and microRNAs
(miRNAs) in l(3)mbt ts1 larval brain tumor samples
(table S5). Of these, 31 are either not expressed in
wild-type brains or are expressed there at less than
10% their level in larval brain tumors. Most of
them are highly expressed in wild-type ovaries,
thus substantiating further the ectopic acquisition
of germline traits that characterizes mbt tumors.
We do not know which, if any, of the germ-
line genes that are up-regulated in mbt tumors are
direct targets of l(3)mbt or if their ectopic expres-
sion is a downstream consequence of intermediate
events. The putative direct targets of l(3)mbt are
many. The dREAM-MMB complex, of which
L(3)MBT is a substoichiometric component (6),
SCIENCE VOL 330 24 DECEMBER 2010
REPORTS
has been found to be promoter-proximal to 32% of
Drosophila genes, and MMB factors are known to
regulate transcription of a wide range of genes in
Drosophila Kc cells (7). In addition, we do not
have an estimate for the number of proteins like
VASA that, despite their low mRNA expression
levels, might be up-regulated in mbt tumors. In-
deed, many of these genes, as well as the piRNAs
and miRNAs expressed in mbt tumors, might
themselves regulate the basal transcription and
translation machineries, adding a further layer of
gene expression modulation (21–23).
We then determined the extent to which ec-
topic expression of germline genes contributes to
mbt tumor growth. To this end, we first quan-
tified larval brain growth in individuals that were
mutant for l(3)mbt ts1 alone, or double mutant for
l(3)mbt ts1 and one of several of the germline genes
Downloaded from www.sciencemag.org on December 23, 2010
that are ectopically expressed in mbt tumors (Fig.
3). Measured as the total amount of protein, the
average brain size in l(3)mbt ts1 (21 T 6 mg of pro-
tein per brain, n = 5) is about seven times as large
(P < 0.0001) as that in control w1118 larvae, a
difference that is not significantly reduced by the
additional loss of zpg, Pxt, or AGO3. However,
brain overgrowth is reduced to a size similar to
that of the control in l(3)mbt ts1 larvae that are also
mutant for either piwi (P < 0.0001), vasa (P <
0.0001), aub (P = 0.0003), or nos (P = 0.001)
(Fig. 3). The loss of piwi does not prevent brain
overgrowth in brat k06028 mutant larvae (P = 0.72).
We then quantified tumor growth after allograft
in adult flies (Fig. 3). The frequency with which
l(3)mbt ts1 homozygous larval brain tissue de-
velops tumors in this assay (70%, n = 67) is not
significantly reduced by the additional loss of zpg
or AGO3 and is only moderately reduced by the
loss of Pxt (P = 0.03), but it is markedly reduced
by the additional loss of piwi (P < 0.0001), vasa
(P < 0.0001), aub (P = 0.0002), or nos (P < 0.0001).
The frequency of brat k06028 tumor formation (80%,
n = 10) is not affected by the loss of piwi (73%, n =
15) or nos (86%, n = 7, P = 1). These results
demonstrate that the ectopic expression of germ-
line genes, particularly piwi, vasa, nos, and aub,
significantly contributes to mbt tumor growth.
A closely reminiscent soma-to-germline trans-
formation observed in mutants in the Caenorhab-
ditis elegans Rb homolog LIN-35, as well as in
long-lived C. elegans strains (20, 24, 25), has led
some to propose that the acquisition of germline
characteristics by somatic cells might contribute to
increased fitness and survival, a mechanism that
could contribute to the transformation of mam-
malian cells (24, 25). Also in humans, some genes
that are predominantly expressed in germline cells
and have little or no expression in somatic adult
tissues become aberrantly activated in various ma-
lignancies, including melanoma and several types
of carcinomas (26, 27). These are known as cancer-
testis (CT) genes or cancer-germline (CG) genes
(28). A subset of these CG genes encode antigens
that are immunogenic in cancer patients and are
being pursued as biomarkers and as targets for
therapeutic cancer vaccines (29, 30).
1825
 REPORTS
Fig. 2. Ectopic expression of
germline proteins in l(3)mbt ts1
larval brain tumors. (A) West-
ern blot. PIWI, AUB, and VASA
are ectopically expressed in
l(3)mbt ts1 brain tumors. aTUB
is used as a loading control.
(B) Immunofluorescence.
VASA, SQU, and C(3)G are
overexpressed in l(3)mbt ts1
brains raised at 29°C. Low-
magnification views (left)
reveal VASA staining concen-
trated in the outer prolifer-
ative center (OPC) and in
undifferentiated cells of the
central brain (CB) (scale bar,
50 mm). High-magnification

Downloaded from www.sciencemag.org on December 23, 2010

views (middle and right) show
that SQU and C(3)G localize
in the cytoplasm and on con-
densed chromatin, respectively (scale bar, 10 mm). Brains were counterstained with DAPI (4´,6´-diamidino-2-phenylindole) (DNA) and antibodies against the
neuroblast marker MIRA.

Fig. 3. The role of ectop-

ically expressed germ-
line genes in mbt tumor
growth. Brain micro-
graphs were taken from
larvae of the correspond-
ing genotypes raised at
29°C (scale bars, 100 mm).
Larval brain size is shown
as mean T SD (in micro-
grams) of protein per brain
(n = number of brains).
Adult fly micrographs
were taken 10 days after
implantation of green flu-
orescent protein (GFP)–
labeled larval brain tissue.
In the absence of tumor
growth, GFP signal is
either undetectable or is
localized to a very small
piece of green tissue that
is about the size of the
implant (arrows). Tumor
growth was quantified as
the percentage (%) of n
hosts in which the im-
planted tissue (green) grew
over the entire abdomen
of the host. P-values refer
to the difference between
each double-mutant combination and l(3)mbt ts1, or between piwi1 brat K06028 and brat k06028.

Human CG genes are suspected to contribute
to oncogenesis germline traits like immortality,
invasiveness, and hypomethylation (28), but their
actual role in cancer remains unknown. Our re-
sults demonstrate that ectopic germline traits are
necessary for tumor growth in Drosophila mbt
tumors, suggesting that their inactivation might
have tumor-suppressing effects in other species.
Some germline genes up-regulated in mbt tumors
are orthologs of human CG genes like PIWIL1/piwi
(31, 32), NANOS1/nanos (33), and SYCP1 /c(3)G
(34). The list of genes up-regulated in mbt tumors
includes many other germline genes that might
also be relevant in human cancer.
References and Notes
1. E. Gateff, T. Löffler, J. Wismar, Mech. Dev. 41, 15 (1993).
2. C. B. Yohn, L. Pusateri, V. Barbosa, R. Lehmann, Genetics
165, 1889 (2003).
3. R. Bonasio, E. Lecona, D. Reinberg, Semin. Cell Dev. Biol.
21, 221 (2010).
4. P. Boccuni, D. MacGrogan, J. M. Scandura, S. D. Nimer,
J. Biol. Chem. 278, 15412 (2003).
5. P. Trojer et al., Cell 129, 915 (2007).
6. P. W. Lewis et al., Genes Dev. 18, 2929 (2004).
7. D. Georlette et al., Genes Dev. 21, 2880 (2007).
8. C. Gonzalez, Nat. Rev. Genet. 8, 462 (2007).
9. M. D. Wong, Z. Jin, T. Xie, Annu. Rev. Genet. 39, 173
(2005).
10. M. J. Svensson, J. D. Chen, V. Pirrotta, J. Larsson,
Chromosoma 112, 133 (2003).
11. L. A. Lee, D. Van Hoewyk, T. L. Orr-Weaver, Genes Dev.
17, 2979 (2003).

1826 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


12. E. F. Joyce, S. N. Tanneti, K. S. McKim, Genetics 181, 335
(2009).
13. S. L. Page et al., PLoS Genet. 4, e1000194 (2008).
14. L. Perezgasga et al., Development 131, 1691 (2004).
15. G. Tavosanis, S. Llamazares, G. Goulielmos, C. Gonzalez,
EMBO J. 16, 1809 (1997).
16. C. Klattenhoff, W. Theurkauf, Development 135, 3 (2008).
17. V. S. Patil, T. Kai, Curr. Biol. 20, 724 (2010).
18. T. K. Smulders-Srinivasan, H. Lin, Genetics 165, 1971 (2003).
19. B. Ye et al., Curr. Biol. 14, 314 (2004).
20. S. Strome, R. Lehmann, Science 316, 392 (2007).
21. T. Kai, D. Williams, A. C. Spradling, Dev. Biol. 283, 486
(2005).
22. N. Bushati, S. M. Cohen, Annu. Rev. Cell Dev. Biol. 23,
175 (2007).
23. H. Lin, Science 316, 397 (2007).
24. D. Wang et al., Nature 436, 593 (2005).
25. S. P. Curran, X. Wu, C. G. Riedel, G. Ruvkun, Nature 459,
1079 (2009).
26. www.cancerimmunity.org/CTdatabase/
27. www.cta.lncc.br/
28. A. J. Simpson, O. L. Caballero, A. Jungbluth, Y. T. Chen,
L. J. Old, Nat. Rev. Cancer 5, 615 (2005).
29. L. J. Old, Cancer Immun. 8 (suppl. 1), 1 (2008).
30. E. Jäger, D. Jäger, A. Knuth, Curr. Opin. Immunol. 14,
178 (2002).
31. H. Taubert et al., Oncogene 26, 1098 (2007).
32. L. F. Grochola et al., Br. J. Cancer 99, 1083 (2008).
33. A. Bonnomet et al., Oncogene 27, 3692 (2008).
34. O. Türeci et al., Proc. Natl. Acad. Sci. U.S.A. 95, 5211 (1998).
35. We thank E. Gateff, R. Lehmann, P. Zamore, A. Spradling,
F. Azorin, P. Lasko, S. Hawley, M. Siomi, T. Kai,
T. Orr-Weaver, H. White-Cooper, D. McKearin,
T. Schupbach, Hybridoma Bank, and Bloomington and
Tübingen Drosophila Stock centers for antibodies and
fly stocks; H. Auer, the IRB Functional Genomics
Facility, and the European Molecular Biology Laboratory
Genomics Core Facility for invaluable technical guidance;
REPORTS
J. Januschke and members of our laboratory for discussions;
and M. Llamazares for proofreading. Work in our
laboratory is supported by ONCASYM-037398, BFU2006-
05813, BFU2009-07975, SGR2005, SRG200, CEN-
20091016, and Consolider-Ingenio CSD2006-23.
A.J. is a recipient of a Ministerio de Ciencia e Innovacion
Formacion de Personal Investigador fellowship. Array
data sets are deposited at Gene Expression Omnibus
(accession no. GSE24917).
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1824/DC1
Material and Methods
Fig. S1
Tables S1 to S5
References
22 July 2010; accepted 9 November 2010
10.1126/science.1195481

near the S locus in Solanum (5, 6). Expression of

A Pollen Factor Linking Inter- and

Downloaded from www.sciencemag.org on December 23, 2010

S-RNase transgenes in Nicotiana established their
function in rejecting pollen from related SC species
Intraspecific Pollen Rejection in Tomato but also uncovered S-RNase–independent pollen
rejection pathways (7). However, the differential
Wentao Li and Roger T. Chetelat* timing of rejection of self versus interspecific
pollen tubes, as well as exceptions to the SI × SC
Self-incompatibility (SI)—intraspecific pollen recognition systems that allow plants to avoid rule—including rejection of interspecific pollen
inbreeding—in the Solanaceae (the nightshade family) is controlled by a polymorphic S locus where by accessions that lack S-RNase—demonstrate
“self” pollen is rejected on pistils with matching S alleles. In contrast, unilateral interspecific that there are also differences between SI and UI
incompatibility (UI) prevents hybridization between related species, most commonly when the pollen (8–11).
donor is self-compatible (SC) and the recipient is SI. We observed that in Solanum, a pollen-expressed Tomato and its wild relatives (Solanum sect.
Cullin1 gene with high similarity to Petunia SI factors interacts genetically with a gene at or near the S Lycopersicon and allied species) exhibit a wide
locus to control UI. Cultivated tomato and related red- or orange-fruited species (all SC) exhibit the same range of mating systems, from autogamy to strict
loss-of-function mutation in this gene, whereas the green-fruited species (mostly SI) contain a functional allogamy, with variation between and within species
allele; hence, similar biochemical mechanisms underlie the rejection of both “self” and interspecific pollen. (12). Cultivated tomato (S. lycopersicum) and the
other red- or orange-fruited species are all SC,
arwin noted that many flowering plants unilateral interspecific incompatibility (UI). In the and their pollen is rejected by UI on pistils of all

D reject both self pollen (too similar) and

pollen from foreign species (too dissim-
ilar), but allow fertilization between individuals
of the same species (1). Although the molecular
mechanisms underlying self-incompatibility (SI)
have been well studied, much less is known about
Department of Plant Sciences, University of California, Davis,
CA 95616, USA.
*To whom correspondence should be addressed. E-mail:
trchetelat@ucdavis.edu
73H07-11
nightshade family (Solanaceae), SI specificity is de-
termined by an S locus, encoding S-ribonucleases
(S-RNases) expressed in the pistil (2) and S-locus
F-box (SLF) proteins expressed in pollen (3).
Most cases of UI occur in crosses between SI and
self-compatible (SC) species, when the pollen
source is the SC species (referred to as the SI ×
SC rule) (4). The general validity of the SI × SC
rule in the Solanaceae suggests the two systems
are related. Indeed, quantitative trait loci for both
pistil and pollen-side UI have been detected at or
the green-fruited taxa, which can be either SI or
SC (13). Besides impeding the transfer of cyto-
plasmic traits (maternally inherited in Solanum),
UI may also contribute to reproductive isolation,
because SI and SC tomato species are often
sympatric in their native regions.
Our goal was to isolate the male gametophytic
factors underlying pollen rejection by UI. As se-
lective pistils, we used allotriploid S. lycopersicum ×
S. lycopersicoides hybrids, composed of two ge-
nomes from cultivated tomato and one from the

19E05-16

19E05-29
19E05-17

19E05-36

19E05-15
19E05-21

A C GT AG
SpCUL1 6 7 8
Mi

SL1.50sc04013
SlCUL1 6 7 8
1463

1478
1465
1453

1456
1457

1517

1614

Position/kb
TAG
Recombinants 19-89
44-32
I
C
Fig. 1. Map of the ui6.1 region and analysis of candidate genes. (A) Physical map
50-58 I showing the genotypes of four recombinants and two candidate genes. Open bars,
71-61
22 kb
I
homozygous for S. lycopersicum allele; hatched, heterozygous; solid, homo-
Candidate genes WD-40 Cullin1
zygous S. pennellii. The pollen phenotypes of recombinants on pistils of the
allotriploid (I, incompatible; C, compatible) place ui6.1 between markers 19E05-
16 and 19E05-21, a region that contains a WD-40 domain gene and a Cullin1
B Leaf Pollen Bud Pistil Stem
(CUL1) gene. (B) Expression of CUL1 and WD-40 in pollen of S. pennellii by reverse
Cullin1
transcription PCR (RT-PCR). (C) Gene structure of SlCUL1 (S. lycopersicum allele)
WD-40 and SpCUL1 (S. pennellii allele). Both alleles include 19 introns and 20 exons, but
SlCUL1 contains a 436-bp deletion in the seventh intron, which shifts 46 bp of
Actin
intron sequence (solid segment) to the seventh exon, introducing a premature stop
codon (TAG).

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1827


12. E. F. Joyce, S. N. Tanneti, K. S. McKim, Genetics 181, 335
(2009).
13. S. L. Page et al., PLoS Genet. 4, e1000194 (2008).
14. L. Perezgasga et al., Development 131, 1691 (2004).
15. G. Tavosanis, S. Llamazares, G. Goulielmos, C. Gonzalez,
EMBO J. 16, 1809 (1997).
16. C. Klattenhoff, W. Theurkauf, Development 135, 3 (2008).
17. V. S. Patil, T. Kai, Curr. Biol. 20, 724 (2010).
18. T. K. Smulders-Srinivasan, H. Lin, Genetics 165, 1971 (2003).
19. B. Ye et al., Curr. Biol. 14, 314 (2004).
20. S. Strome, R. Lehmann, Science 316, 392 (2007).
21. T. Kai, D. Williams, A. C. Spradling, Dev. Biol. 283, 486
(2005).
22. N. Bushati, S. M. Cohen, Annu. Rev. Cell Dev. Biol. 23,
175 (2007).
23. H. Lin, Science 316, 397 (2007).
24. D. Wang et al., Nature 436, 593 (2005).
25. S. P. Curran, X. Wu, C. G. Riedel, G. Ruvkun, Nature 459,
1079 (2009).
26. www.cancerimmunity.org/CTdatabase/
27. www.cta.lncc.br/
28. A. J. Simpson, O. L. Caballero, A. Jungbluth, Y. T. Chen,
L. J. Old, Nat. Rev. Cancer 5, 615 (2005).
29. L. J. Old, Cancer Immun. 8 (suppl. 1), 1 (2008).
30. E. Jäger, D. Jäger, A. Knuth, Curr. Opin. Immunol. 14,
178 (2002).
31. H. Taubert et al., Oncogene 26, 1098 (2007).
32. L. F. Grochola et al., Br. J. Cancer 99, 1083 (2008).
33. A. Bonnomet et al., Oncogene 27, 3692 (2008).
34. O. Türeci et al., Proc. Natl. Acad. Sci. U.S.A. 95, 5211 (1998).
35. We thank E. Gateff, R. Lehmann, P. Zamore, A. Spradling,
F. Azorin, P. Lasko, S. Hawley, M. Siomi, T. Kai,
T. Orr-Weaver, H. White-Cooper, D. McKearin,
T. Schupbach, Hybridoma Bank, and Bloomington and
Tübingen Drosophila Stock centers for antibodies and
fly stocks; H. Auer, the IRB Functional Genomics
Facility, and the European Molecular Biology Laboratory
Genomics Core Facility for invaluable technical guidance;
REPORTS
J. Januschke and members of our laboratory for discussions;
and M. Llamazares for proofreading. Work in our
laboratory is supported by ONCASYM-037398, BFU2006-
05813, BFU2009-07975, SGR2005, SRG200, CEN-
20091016, and Consolider-Ingenio CSD2006-23.
A.J. is a recipient of a Ministerio de Ciencia e Innovacion
Formacion de Personal Investigador fellowship. Array
data sets are deposited at Gene Expression Omnibus
(accession no. GSE24917).
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1824/DC1
Material and Methods
Fig. S1
Tables S1 to S5
References
22 July 2010; accepted 9 November 2010
10.1126/science.1195481

near the S locus in Solanum (5, 6). Expression of

A Pollen Factor Linking Inter- and

Downloaded from www.sciencemag.org on December 23, 2010

S-RNase transgenes in Nicotiana established their
function in rejecting pollen from related SC species
Intraspecific Pollen Rejection in Tomato but also uncovered S-RNase–independent pollen
rejection pathways (7). However, the differential
Wentao Li and Roger T. Chetelat* timing of rejection of self versus interspecific
pollen tubes, as well as exceptions to the SI × SC
Self-incompatibility (SI)—intraspecific pollen recognition systems that allow plants to avoid rule—including rejection of interspecific pollen
inbreeding—in the Solanaceae (the nightshade family) is controlled by a polymorphic S locus where by accessions that lack S-RNase—demonstrate
“self” pollen is rejected on pistils with matching S alleles. In contrast, unilateral interspecific that there are also differences between SI and UI
incompatibility (UI) prevents hybridization between related species, most commonly when the pollen (8–11).
donor is self-compatible (SC) and the recipient is SI. We observed that in Solanum, a pollen-expressed Tomato and its wild relatives (Solanum sect.
Cullin1 gene with high similarity to Petunia SI factors interacts genetically with a gene at or near the S Lycopersicon and allied species) exhibit a wide
locus to control UI. Cultivated tomato and related red- or orange-fruited species (all SC) exhibit the same range of mating systems, from autogamy to strict
loss-of-function mutation in this gene, whereas the green-fruited species (mostly SI) contain a functional allogamy, with variation between and within species
allele; hence, similar biochemical mechanisms underlie the rejection of both “self” and interspecific pollen. (12). Cultivated tomato (S. lycopersicum) and the
other red- or orange-fruited species are all SC,
arwin noted that many flowering plants unilateral interspecific incompatibility (UI). In the and their pollen is rejected by UI on pistils of all

D reject both self pollen (too similar) and

pollen from foreign species (too dissim-
ilar), but allow fertilization between individuals
of the same species (1). Although the molecular
mechanisms underlying self-incompatibility (SI)
have been well studied, much less is known about
Department of Plant Sciences, University of California, Davis,
CA 95616, USA.
*To whom correspondence should be addressed. E-mail:
trchetelat@ucdavis.edu
73H07-11
nightshade family (Solanaceae), SI specificity is de-
termined by an S locus, encoding S-ribonucleases
(S-RNases) expressed in the pistil (2) and S-locus
F-box (SLF) proteins expressed in pollen (3).
Most cases of UI occur in crosses between SI and
self-compatible (SC) species, when the pollen
source is the SC species (referred to as the SI ×
SC rule) (4). The general validity of the SI × SC
rule in the Solanaceae suggests the two systems
are related. Indeed, quantitative trait loci for both
pistil and pollen-side UI have been detected at or
the green-fruited taxa, which can be either SI or
SC (13). Besides impeding the transfer of cyto-
plasmic traits (maternally inherited in Solanum),
UI may also contribute to reproductive isolation,
because SI and SC tomato species are often
sympatric in their native regions.
Our goal was to isolate the male gametophytic
factors underlying pollen rejection by UI. As se-
lective pistils, we used allotriploid S. lycopersicum ×
S. lycopersicoides hybrids, composed of two ge-
nomes from cultivated tomato and one from the

19E05-16

19E05-29
19E05-17

19E05-36

19E05-15
19E05-21

A C GT AG
SpCUL1 6 7 8
Mi

SL1.50sc04013
SlCUL1 6 7 8
1463

1478
1465
1453

1456
1457

1517

1614

Position/kb
TAG
Recombinants 19-89
44-32
I
C
Fig. 1. Map of the ui6.1 region and analysis of candidate genes. (A) Physical map
50-58 I showing the genotypes of four recombinants and two candidate genes. Open bars,
71-61
22 kb
I
homozygous for S. lycopersicum allele; hatched, heterozygous; solid, homo-
Candidate genes WD-40 Cullin1
zygous S. pennellii. The pollen phenotypes of recombinants on pistils of the
allotriploid (I, incompatible; C, compatible) place ui6.1 between markers 19E05-
16 and 19E05-21, a region that contains a WD-40 domain gene and a Cullin1
B Leaf Pollen Bud Pistil Stem
(CUL1) gene. (B) Expression of CUL1 and WD-40 in pollen of S. pennellii by reverse
Cullin1
transcription PCR (RT-PCR). (C) Gene structure of SlCUL1 (S. lycopersicum allele)
WD-40 and SpCUL1 (S. pennellii allele). Both alleles include 19 introns and 20 exons, but
SlCUL1 contains a 436-bp deletion in the seventh intron, which shifts 46 bp of
Actin
intron sequence (solid segment) to the seventh exon, introducing a premature stop
codon (TAG).

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1827

 REPORTS
wild parent, which is SI (14). UI in pistils of the (fig. S1). A 436–base pair (bp) deletion was ob- T1BC1 plants lacking the transgene were incom-
allotriploid hybrid is weakened relative to pure S. served in SlCUL1 relative to SpCUL1 (fig. S1). A patible with the allotriploid (Fig. 2, B and C),
lycopersicoides: Arrest of incompatible pollen combination of rapid amplification of cDNA consistent with our earlier findings (5, 15). In
tubes occurs lower in the style, and fewer pollen ends (RACE) and the polymerase chain reaction contrast, pollen of transgenic plants containing both
factors are required to overcome the incompat- (PCR) determined that the SpCUL1 cDNA com- the SpCUL1 transgene and ui1.1 from S. pennellii
ibility, thus presenting a simplified barrier to in- prised 2223 bp and SlCUL1 2269 bp (Fig. 1C and were compatible on allotriploid styles (Fig. 2, B
terspecific pollen (5, 15). On the pollen side, S. fig. S2). Alignment of the genomic and cDNA and C), confirming that CUL1 underlies ui6.1.
lycopersicum stocks containing two gametophyt- sequences showed both alleles are comprised of Because the SlCUL1 gene of cultivated to-
ic factors (ui1.1 and ui6.1) introgressed from an 20 exons and 19 introns, and differ only in the mato contains a loss-of-function mutation abolish-
SC biotype of S. pennellii (most are SI) are com- seventh exon and intron (Fig. 1C and fig. S1), ing pollen compatibility on pistils of SI species
patible on the allotriploid (5, 15). where the 436-bp deletion in SlCUL1 is located. (or their hybrids), we examined other SC red- or
We previously mapped the ui6.1 locus to a The deleted fragment contains the splice recog- orange-fruited tomato species (i.e., S. pimpinelli-
160-kb region on the short arm of chromosome nition sites; this suggests that a small intron may folium, S. cheesmaniae, and S. galapagense) whose
6, between markers 73H07-11 and Mi (15). We have been lost during evolution in SlCUL1. The pollen is also rejected on styles of the SI (and
refined the candidate region to a ~22-kb interval deletion shifts 46 bp of intron to the seventh exon green-fruited SC) taxa. We determined that the
(Fig. 1A and table S1) containing two open read- (Fig. 1C), resulting in a stop codon that is trans- CUL1 deletion was also present in these species,
ing frames: a WD-40 repeat domain gene and a lated into a truncated protein (252 rather than 740 based on the length of the seventh intron (Fig. 3A
Cullin1 gene. The Cullin1 (hereafter CUL1) pro- amino acids; fig. S3). and table S2). In contrast, the green-fruited species

Downloaded from www.sciencemag.org on December 23, 2010

tein was homologous (75% sequence identity) to To test for function in pollen-side UI, we (mostly SI, some SC) all contained a full length
PiCul1-G from Petunia inflata (16) and PhCullin1 introduced a construct containing the full-length intron in CUL1. Both mutant and full-length al-
from P. hybrida (17), which interact with the SI- SpCUL1 cDNA driven by the LAT52 pollen- leles were detected in S. pimpinellifolium, a species
related factors, SBP1 (S-RNase binding protein) specific promoter (19) into S. lycopersicum via that includes facultative outcrossing populations and
and SSK1 (SLF-interacting Skp1-like1), respec- Agrobacterium-mediated transformation. Because is more diverse than the other red- or orange-fruited
tively. These proteins form either a canonical both S. pennellii factors, ui1.1 and ui6.1, are re- species (20); this finding suggests that the intron
SCF (Skp1-Cullin1–F box) complex or a SLF- quired for pollen compatibility on pistils of the deletion may have originated in S. pimpinellifolium.
SBP1-CUL1 complex, which most likely func- allotriploid hybrid, T0 plants were crossed to a S. We found that all green-fruited species
tions as an E3 ubiquitin ligase (SCFSLF) to target lycopersicum stock homozygous for ui1.1 intro- produce a normal length (i.e., shorter) mRNA,
a pistil SI factor for degradation. WD-40 repeat gressed from S. pennellii, to produce “T1BC1” whereas all the accessions with the CUL1 de-
domains function in protein-protein interactions plants. Transformant 1 exhibited T-DNA inheri- letion produce the longer mRNA seen in S.
and contribute to the substrate specificity of F-box tance consistent with a single insertion, whereas lycopersicum (Fig. 3B and table S2). These ob-
proteins (18). Because the WD-40 gene was ex- transformant 2 most likely had two independent servations confirm that variation in length of the
pressed at low levels in pollen of S. pennellii, insertions. seventh intron is responsible for differences in
whereas CUL1 mRNA was more abundant in The SpCUL1 transgene was expressed at high mRNA sizes, consistent with a single mutation
pollen (Fig. 1B), we hypothesized that CUL1 was levels in anthers of transgenic plants (Fig. 2A). event in the red/orange clade. In addition, we
a more likely candidate for ui6.1. The function of SpCUL1 was tested by placing tested the function of CUL1 alleles from three
The S. lycopersicum (SlCUL1) and S. pennellii transgenic pollen on pistils of the allotriploid. diverse SI green-fruited species (S. peruvianum,
(SpCUL1) alleles were obtained by sequencing Pollen tubes of T0 plants (which lack ui1.1) and S. habrochaites, and S. lycopersicoides) using in-

A B T0 T1BC1(-) T1BC1(+)
VF36 LA0716 T0 T1BC1(-) T1BC1(+)
pollen pollen anthers anthers anthers

Cullin1 Stigma

Actin

C Control Style
Transformant 1
Transformant 2
Pollen tube length (% of allotriploid style)

100

80 Ovaries

60

Fig. 2. Expression and functional analysis of S. lycopersicum plants transformed with

40 LAT52-SpCUL1. Primary transformants (T0) were crossed to a ui1.1 homozygote to produce
T1BC1 plants with and without the transgene [T1BC1(+) and T1BC1(–), respectively]. (A)
Analysis of SpCUL1 expression by RT-PCR. SpCUL1 mRNA was abundant in anthers of
20
transgenic plants. (B) Pollen tube phenotypes of transgenic plants on the allotriploid
hybrid stigma, style and ovaries. T0 plants, which lack ui1.1 (left column), and T1BC1
0
0 ui1 ui6 ui1 T T T T T T
plants without the transgene [T1BC1 (–), center column] show an incompatible
.1 .1 .1+ 0
ui6
1 BC
1 (-)
1 BC
1 (+
0 1 BC
1 (-)
1 BC
1 (+
phenotype, whereas plants with the transgene and ui1.1 [T1BC1 (+), right column] are
.1 ) )
compatible. (C) Pollen tube growth on pistils of the allotriploid hybrid. Controls, ui1.1
Genotype of pollen parents
and/or ui6.1 introgressions from S. pennellii [data from (15)], and multiple
transformants of T0, T1BC1(–), and T1BC1(+) are shown. Values are averages T SE.

1828 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


trogression lines containing the ui6.1 region bred
into the genetic background of cultivated tomato
(Fig. 3C). When combined with the ui1.1 allele
introgressed from S. pennellii, pollen bearing any
of the three independent ui6.1 introgressions was
compatible on pistils of the allotriploid hybrid.
This demonstrates that CUL1 genes from SI
species function in interspecific incompatibility,
and that they interact genetically with the ui1.1
gene from SC S. pennellii to change the pollen
phenotype.
REPORTS
Cullin1 proteins function as scaffolds in re- pollen tube, with multiple SLF proteins recogniz-
cruiting other components, including Skp1, Rbx- ing different S-RNase alleles (23). Our results
1, and F-box proteins, to assemble an SCF-type suggest that UI is regulated by a similar bio-
E3 ubiquitin ligase complex that targets proteins chemical mechanism. Although we have not yet
recognized by the F-box for degradation by the identified the product of ui1.1, its map location at
proteasome (21). Biochemical evidence suggests or near the S locus, its gametophytic gene action,
that the ubiquitin-proteasome pathway is involved and its genetic interaction with CUL1 in control-
in the degradation of a pistil-expressed factor re- ling pollen-side UI are consistent with a model in
quired for S-RNase–based SI (16, 17, 22). The which ui1.1 encodes an SLF protein.
putative SCFSLF complex has been proposed to Other genetic evidence suggests that CUL1
target nonself S-RNases for degradation in the functions in an S-RNase–dependent pathway of

S. lycopersicum S. galapagense S. cheesmaniae S. pimpinellifolium S. neorickii S. chmielewskii

Downloaded from www.sciencemag.org on December 23, 2010

S. chilense S. arcanum S. corneliomulleri S. huaylasense S. peruvianum S. habrochaites

S. pennellii S. lycopersicoides S. sitiens S. ochranthum S. juglandifolium

lium

lium

ides
lleri

B
m

s
ii
e

e
ania

num

haite
ewsk
ellifo

ellifo
rsicu

rsico
gens

sens
omu
m
ii

lii
e
eesm

orick

ruvia
ilens

broc
mpin

mpin

canu

rneli

nnel
cope

lapa

miel

ayla

cope

tiens
16
LA07

S. hu
VF36

S. ga

S. ch

S. ne

S. ch

S. ch

S. co

S. pe

S. ha

S. pe
S. ar
S. pi

S. pi
S. ly

S. ly

S. si
Fig. 3. Analysis of CUL1 alleles in wild tomato and allied Solanum species. (A) Distribution of a C ui6.1 IL ui6.1 IL ui1.1p/p
CUL1 intron deletion. Five representative accessions (table S2) from each species were compared 100
to S. lycopersicum cv. VF36 (436-bp deletion in the seventh intron) and S. pennellii LA0716
Pollen tube length (% of allotriploid style)

(intact intron). All green-fruited species examined contained only the full-length intron, whereas
the red- or orange-fruited taxa all contained the intron deletion (both alleles were present in
80
S. pimpinellifolium). (B) Analysis of the seventh exon of CUL1 in accessions of each species. (C)
Lines of S. lycopersicum containing the ui6.1 (CUL1) region introgressed from S. peruvianum
(S. peruv.), S. habrochaites (S. habro.), or S. lycopersicoides (S. lycds.) (28–30) were crossed to
a stock homozygous for ui1.1 from S. pennellii (ui1.1p/p) to produce doubly heterozygous lines. 60
Pollen from the ui6.1 single introgressions was incompatible on pistils of the allotriploid,
whereas the corresponding ui1.1 + ui6.1 double introgressions were compatible, indicating
functional CUL1 alleles. 40

20

0
0 S. peruv. S. peruv. S. habro. S. lycds.
Source of ui6.1

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1829

 REPORTS
pollen rejection. In progeny of some interspecific
tomato hybrids, distorted segregation ratios for
CUL1-linked markers are consistent with the se-
lective elimination of pollen bearing the nonfunc-
tional SlCUL1 allele (15). Distorted transmission
of CUL1 is only observed in progeny of crosses
between SC cultivated tomato and SI wild species,
and only when the F1 hybrid is used as pistillate
parent. Moreover, two SC accessions of mostly
SI species that either accumulate no S-RNase in
the pistil [S. pennellii LA0716 (11)] or express a
mutant protein lacking RNase activity [S. arcanum
LA2157 (24)] exhibit normal F2 segregation for
CUL1-linked markers in hybrids with cultivated
tomato (25, 26). These observations suggest that
selection against CUL1-deficient pollen requires
S-RNase activity in the pistil. If so, loss of CUL1
function in the red- or orange-fruited species was
References and Notes 23. K. Kubo et al., Science 330, 796 (2010).
1. B. A. McClure, J. Exp. Bot. 60, 1069 (2009). 24. J. Royo et al., Proc. Natl. Acad. Sci. U.S.A. 91, 6511
2. B. A. McClure et al., Nature 342, 955 (1989). (1994).
3. P. Sijacic et al., Nature 429, 302 (2004). 25. A. W. van Heusden et al., Theor. Appl. Genet. 99,
4. D. Lewis, L. K. Crowe, Heredity 12, 233 (1958). 1068 (1999).
5. R. T. Chetelat, J. W. DeVerna, Theor. Appl. Genet. 82, 26. T. M. Fulton, R. Van der Hoeven, N. T. Eannetta,
704 (1991). S. D. Tanksley, Plant Cell 14, 1457 (2002).
6. D. Bernacchi, S. D. Tanksley, Genetics 147, 861 27. C. M. Rick, in Solanaceae Biology and Systematics,
W. G. D’Arcy, Ed. (Columbia Univ. Press, New York,
(1997).
1986), pp. 475–495.
7. J. Murfett et al., Plant Cell 8, 943 (1996).
8. N. G. Hogenboom, Euphytica 22, 219 (1973).
28. J. Y. Ho et al., Plant J. 2, 971 (1992).
29. A. J. Monforte, S. D. Tanksley, Genome 43, 803
9. J. J. Hardon, Genetics 57, 795 (1967).
(2000).
10. B. E. Liedl, S. McCormick, M. A. Mutschler, Sex. Plant
30. M. A. Canady, V. Meglic, R. T. Chetelat, Genome 48,
Reprod. 9, 299 (1996).
685 (2005).
11. P. A. Covey et al., Plant J. 64, 367 (2010). 31. We thank J. DeVerna, P. March, K. Smith, and the
12. C. M. Rick, in Plant Evolutionary Biology, L. D. Gottlieb, C. M. Rick Tomato Genetics Resource Center (TGRC) staff
S. K. Jain, Eds. (Chapman & Hall, London, 1988), for providing seed or cuttings of key genotypes; P. March
pp. 133–147. for photos of fruit; B. McClure and P. Bedinger for
13. M. A. Mutschler, B. E. Liedl, in Genetic Control of comments on the manuscript; and S. Tanksley for
Self-Incompatibility and Reproductive Development in providing seed of S. habrochaites introgression lines.
Flowering Plants, E. G. Williams, A. C. Clarke, R. B. Knox, Supported by NSF grant DBI 0605200. Sequence data

Downloaded from www.sciencemag.org on December 23, 2010

likely preceded by a loss of S-RNase expression 14.
[styles of S. lycopersicum do not accumulate
S-RNase (11)]. On the other hand, pistils of LA0716 15.
and LA2157 reject pollen from cultivated tomato 16.
(9, 27) without expressing functional S-RNases, 17.
which suggests that pollen rejection by UI can 18.
19.
also be mechanistically distinct from SI. Although
our results are from an analysis of interspecific
20.
Solanum hybrids, they may be relevant to UI in
other solanaceous plants, and possibly to other 21.
families that use the S-RNase–based SI system. 22.
Eds. (Kluwer, Dordrecht, Netherlands, 1994), pp. 164–188.
C. M. Rick, J. W. De Verna, R. T. Chetelat, M. A. Stevens,
Proc. Natl. Acad. Sci. U.S.A. 83, 3580 (1986).
W. Li, S. Royer, R. T. Chetelat, Genetics 185, 1069
(2010).
Z. Hua, T. H. Kao, Plant Cell 18, 2531 (2006).
L. Zhao et al., Plant J. 62, 52 (2010).
C. Cenciarelli et al., Curr. Biol. 9, 1177 (1999).
D. Twell, J. Yamaguchi, S. McCormick, Development 109,
705 (1990).
C. M. Rick, J. F. Fobes, M. Holle, Plant Syst. Evol. 127,
139 (1977).
N. Zheng et al., Nature 416, 703 (2002).
H. Qiao et al., Plant Cell 16, 582 (2004).
have been deposited in GenBank under accession
numbers HQ610200 and HQ610201. Seed requests
submitted to the TGRC are subject to a material transfer
agreement (http://tgrc.ucdavis.edu/MTA/TGRC-MTA.pdf).
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1827/DC1
Materials and Methods
Tables S1 and S2
Figs. S1 to S3
References
17 September 2010; accepted 18 November 2010
10.1126/science.1197908

their functional characteristics are, is still the topic

The Social Sense: Susceptibility to of important debates.
Decades of research seem to suggest that
Others’ Beliefs in Human Infants ToM emerges after the age of four. Developmen-
tal transitions in ToM have often been assessed
using so-called “false-belief tasks” (3). In such
and Adults tasks, children typically have to predict a person’s
behavior based on the person’s false belief while
Ágnes Melinda Kovács,1,2,3* Ernő Téglás,1,2,3 Ansgar Denis Endress3,4 ignoring their own true belief. For example, chil-
dren are presented with a situation in which an-
Human social interactions crucially depend on the ability to represent other agents’ beliefs other child (say, John) places a toy in a cupboard
even when these contradict our own beliefs, leading to the potentially complex problem of and leaves the scene. In his absence, the toy is
simultaneously holding two conflicting representations in mind. Here, we show that adults and moved to a different location (say, a basket). Three-
7-month-olds automatically encode others’ beliefs, and that, surprisingly, others’ beliefs have year-olds typically predict that, upon his return,
similar effects as the participants’ own beliefs. In a visual object detection task, participants’ beliefs John will search in the basket rather than in the
and the beliefs of an agent (whose beliefs were irrelevant to performing the task) both modulated cupboard, because they themselves know that the
adults’ reaction times and infants’ looking times. Moreover, the agent’s beliefs influenced toy is in the basket. That is, at least in their overt
participants’ behavior even after the agent had left the scene, suggesting that participants computed responses, 3-year olds fail to take into account
the agent’s beliefs online and sustained them, possibly for future predictions about the agent’s that John cannot know that the toy is in the basket
behavior. Hence, the mere presence of an agent automatically triggers powerful processes of belief and must, therefore, (falsely) believe the toy to be
computation that may be part of a “social sense” crucial to human societies. in the cupboard. In contrast, older children (and
adults) take into account John’s false belief and
umans are guided by internal states such be hardly imaginable. Social interactions, from predict that he will search in the cupboard. Based

H as goals and beliefs. Without an ability to

infer others’ mental states, society would
1
Institute for Psychology, Hungarian Academy of Sciences,
H-1132 Budapest, Hungary. 2Cognitive Development Centre,
Central European University, H-1015, Budapest, Hungary.
3
Cognitive Neuroscience Sector, International School for
Advanced Studies (SISSA), I-34014 Trieste, Italy. 4Department
of Brain and Cognitive Sciences, Massachusetts Institute of
Technology, Cambridge MA 02139, USA.
*To whom correspondence should be addressed. E-mail:
agneskovacs@mtapi.hu
collective hunting to playing soccer to criminal
justice, critically depend on the ability to infer
others’ intentions and beliefs. Such abilities are
also at the foundation of major evolutionary
conundra. For example, the human aptitude at
inferring mental states might be one of the crucial
preconditions for the evolution of the cooperative
social structure in human societies (1). However,
despite their importance for all aspects of social
life, the question of how such “theory of mind”
(ToM) abilities (2) emerge and develop, and what
on such findings, it has been argued that ToM
requires complex computations and emerges rel-
atively late in development (4).
However, such data are not necessarily in-
consistent with the possibility that ToM is auto-
matic and innate (5–8). Children might possess
ToM abilities early on; however, these might be
masked by the slower development of other abil-
ities involved in such tasks, such as inhibition
and selection (6) or problem solving (8). That is,
young children might well be able to represent

1830 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

 REPORTS
pollen rejection. In progeny of some interspecific
tomato hybrids, distorted segregation ratios for
CUL1-linked markers are consistent with the se-
lective elimination of pollen bearing the nonfunc-
tional SlCUL1 allele (15). Distorted transmission
of CUL1 is only observed in progeny of crosses
between SC cultivated tomato and SI wild species,
and only when the F1 hybrid is used as pistillate
parent. Moreover, two SC accessions of mostly
SI species that either accumulate no S-RNase in
the pistil [S. pennellii LA0716 (11)] or express a
mutant protein lacking RNase activity [S. arcanum
LA2157 (24)] exhibit normal F2 segregation for
CUL1-linked markers in hybrids with cultivated
tomato (25, 26). These observations suggest that
selection against CUL1-deficient pollen requires
S-RNase activity in the pistil. If so, loss of CUL1
function in the red- or orange-fruited species was
References and Notes 23. K. Kubo et al., Science 330, 796 (2010).
1. B. A. McClure, J. Exp. Bot. 60, 1069 (2009). 24. J. Royo et al., Proc. Natl. Acad. Sci. U.S.A. 91, 6511
2. B. A. McClure et al., Nature 342, 955 (1989). (1994).
3. P. Sijacic et al., Nature 429, 302 (2004). 25. A. W. van Heusden et al., Theor. Appl. Genet. 99,
4. D. Lewis, L. K. Crowe, Heredity 12, 233 (1958). 1068 (1999).
5. R. T. Chetelat, J. W. DeVerna, Theor. Appl. Genet. 82, 26. T. M. Fulton, R. Van der Hoeven, N. T. Eannetta,
704 (1991). S. D. Tanksley, Plant Cell 14, 1457 (2002).
6. D. Bernacchi, S. D. Tanksley, Genetics 147, 861 27. C. M. Rick, in Solanaceae Biology and Systematics,
W. G. D’Arcy, Ed. (Columbia Univ. Press, New York,
(1997).
1986), pp. 475–495.
7. J. Murfett et al., Plant Cell 8, 943 (1996).
8. N. G. Hogenboom, Euphytica 22, 219 (1973).
28. J. Y. Ho et al., Plant J. 2, 971 (1992).
29. A. J. Monforte, S. D. Tanksley, Genome 43, 803
9. J. J. Hardon, Genetics 57, 795 (1967).
(2000).
10. B. E. Liedl, S. McCormick, M. A. Mutschler, Sex. Plant
30. M. A. Canady, V. Meglic, R. T. Chetelat, Genome 48,
Reprod. 9, 299 (1996).
685 (2005).
11. P. A. Covey et al., Plant J. 64, 367 (2010). 31. We thank J. DeVerna, P. March, K. Smith, and the
12. C. M. Rick, in Plant Evolutionary Biology, L. D. Gottlieb, C. M. Rick Tomato Genetics Resource Center (TGRC) staff
S. K. Jain, Eds. (Chapman & Hall, London, 1988), for providing seed or cuttings of key genotypes; P. March
pp. 133–147. for photos of fruit; B. McClure and P. Bedinger for
13. M. A. Mutschler, B. E. Liedl, in Genetic Control of comments on the manuscript; and S. Tanksley for
Self-Incompatibility and Reproductive Development in providing seed of S. habrochaites introgression lines.
Flowering Plants, E. G. Williams, A. C. Clarke, R. B. Knox, Supported by NSF grant DBI 0605200. Sequence data

Downloaded from www.sciencemag.org on December 23, 2010

likely preceded by a loss of S-RNase expression 14.
[styles of S. lycopersicum do not accumulate
S-RNase (11)]. On the other hand, pistils of LA0716 15.
and LA2157 reject pollen from cultivated tomato 16.
(9, 27) without expressing functional S-RNases, 17.
which suggests that pollen rejection by UI can 18.
19.
also be mechanistically distinct from SI. Although
our results are from an analysis of interspecific
20.
Solanum hybrids, they may be relevant to UI in
other solanaceous plants, and possibly to other 21.
families that use the S-RNase–based SI system. 22.
Eds. (Kluwer, Dordrecht, Netherlands, 1994), pp. 164–188.
C. M. Rick, J. W. De Verna, R. T. Chetelat, M. A. Stevens,
Proc. Natl. Acad. Sci. U.S.A. 83, 3580 (1986).
W. Li, S. Royer, R. T. Chetelat, Genetics 185, 1069
(2010).
Z. Hua, T. H. Kao, Plant Cell 18, 2531 (2006).
L. Zhao et al., Plant J. 62, 52 (2010).
C. Cenciarelli et al., Curr. Biol. 9, 1177 (1999).
D. Twell, J. Yamaguchi, S. McCormick, Development 109,
705 (1990).
C. M. Rick, J. F. Fobes, M. Holle, Plant Syst. Evol. 127,
139 (1977).
N. Zheng et al., Nature 416, 703 (2002).
H. Qiao et al., Plant Cell 16, 582 (2004).
have been deposited in GenBank under accession
numbers HQ610200 and HQ610201. Seed requests
submitted to the TGRC are subject to a material transfer
agreement (http://tgrc.ucdavis.edu/MTA/TGRC-MTA.pdf).
Supporting Online Material
www.sciencemag.org/cgi/content/full/330/6012/1827/DC1
Materials and Methods
Tables S1 and S2
Figs. S1 to S3
References
17 September 2010; accepted 18 November 2010
10.1126/science.1197908

their functional characteristics are, is still the topic

The Social Sense: Susceptibility to of important debates.
Decades of research seem to suggest that
Others’ Beliefs in Human Infants ToM emerges after the age of four. Developmen-
tal transitions in ToM have often been assessed
using so-called “false-belief tasks” (3). In such
and Adults tasks, children typically have to predict a person’s
behavior based on the person’s false belief while
Ágnes Melinda Kovács,1,2,3* Ernő Téglás,1,2,3 Ansgar Denis Endress3,4 ignoring their own true belief. For example, chil-
dren are presented with a situation in which an-
Human social interactions crucially depend on the ability to represent other agents’ beliefs other child (say, John) places a toy in a cupboard
even when these contradict our own beliefs, leading to the potentially complex problem of and leaves the scene. In his absence, the toy is
simultaneously holding two conflicting representations in mind. Here, we show that adults and moved to a different location (say, a basket). Three-
7-month-olds automatically encode others’ beliefs, and that, surprisingly, others’ beliefs have year-olds typically predict that, upon his return,
similar effects as the participants’ own beliefs. In a visual object detection task, participants’ beliefs John will search in the basket rather than in the
and the beliefs of an agent (whose beliefs were irrelevant to performing the task) both modulated cupboard, because they themselves know that the
adults’ reaction times and infants’ looking times. Moreover, the agent’s beliefs influenced toy is in the basket. That is, at least in their overt
participants’ behavior even after the agent had left the scene, suggesting that participants computed responses, 3-year olds fail to take into account
the agent’s beliefs online and sustained them, possibly for future predictions about the agent’s that John cannot know that the toy is in the basket
behavior. Hence, the mere presence of an agent automatically triggers powerful processes of belief and must, therefore, (falsely) believe the toy to be
computation that may be part of a “social sense” crucial to human societies. in the cupboard. In contrast, older children (and
adults) take into account John’s false belief and
umans are guided by internal states such be hardly imaginable. Social interactions, from predict that he will search in the cupboard. Based

H as goals and beliefs. Without an ability to

infer others’ mental states, society would
1
Institute for Psychology, Hungarian Academy of Sciences,
H-1132 Budapest, Hungary. 2Cognitive Development Centre,
Central European University, H-1015, Budapest, Hungary.
3
Cognitive Neuroscience Sector, International School for
Advanced Studies (SISSA), I-34014 Trieste, Italy. 4Department
of Brain and Cognitive Sciences, Massachusetts Institute of
Technology, Cambridge MA 02139, USA.
*To whom correspondence should be addressed. E-mail:
agneskovacs@mtapi.hu
collective hunting to playing soccer to criminal
justice, critically depend on the ability to infer
others’ intentions and beliefs. Such abilities are
also at the foundation of major evolutionary
conundra. For example, the human aptitude at
inferring mental states might be one of the crucial
preconditions for the evolution of the cooperative
social structure in human societies (1). However,
despite their importance for all aspects of social
life, the question of how such “theory of mind”
(ToM) abilities (2) emerge and develop, and what
on such findings, it has been argued that ToM
requires complex computations and emerges rel-
atively late in development (4).
However, such data are not necessarily in-
consistent with the possibility that ToM is auto-
matic and innate (5–8). Children might possess
ToM abilities early on; however, these might be
masked by the slower development of other abil-
ities involved in such tasks, such as inhibition
and selection (6) or problem solving (8). That is,
young children might well be able to represent

1830 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


others’ beliefs, but, to correctly predict that, in the
example above, John will search for the toy where
he (falsely) believes it to be, children also need
efficient inhibitory abilities. Specifically, they need
to deal with two conflicting representations, name-
ly John’s (false) belief that the ball is in the cup-
board and their own (true) belief that the ball is in
the basket, and to inhibit their own belief when
they predict John’s behavior. In line with such
views, newer research suggests that ToM abilities
are present in young children, for instance in 3-
year-olds when testing populations with enhanced
inhibitory abilities (9), or even in infants in their
second year when using simpler testing proce-
dures (10–12).
Although these data suggest that ToM abil-
ities may emerge well before the age of four,
another question has received little empirical at-
puted online and effortlessly, just as we compute
representations of what we perceive in the en-
vironment. The experiments below will address
this issue.
Our experiments will also address a second
issue that, to our knowledge, has not been inves-
tigated directly. Representations about the envi-
ronment and representations about others’ beliefs
can differ in many aspects. Most important, un-
like most aspects of the environment, others’
mental states are not directly observable and are
sometimes inconsistent with the true state of af-
fairs. We can represent that “Mary thinks that
John is at home” even if we know that “John is
not at home.” Thus, our knowledge of the envi-
ronment does not reliably predict the contents of
others’ beliefs, nor do others’ beliefs reliably pre-
dict the state of the environment. Accordingly,
REPORTS
such representations are not referenced to the
environment, they should not affect how we in-
teract with the environment either.
Although this conjecture seems consistent with
the proposal that representations about others’ be-
liefs are computed by specialized mechanisms
(5), there is a more parsimonious hypothesis. We
clearly have mechanisms in place to compute
representations about what we experience in the
environment. Perhaps we use very similar mech-
anisms to compute others’ beliefs based on what
they experience. If so, representations about others’
beliefs should have fundamentally similar proper-
ties as representations about the environment. This
hypothesis predicts that representations of others’
beliefs should be referenced to the environment
just as our own beliefs, and thus can affect our be-
havior. We test this idea in the experiments below.

Downloaded from www.sciencemag.org on December 23, 2010

tention. If such abilities are essentially an innate representations of others’ beliefs might be stored Here, we develop a method for investigating
“social sense,” they should be spontaneous and in a way reflecting that their content is not ref- ToM mechanisms that, in contrast to variants of
automatic, and others’ beliefs should be com- erenced to the current state of affairs. Crucially, if the standard false belief task, is implicit, makes
no reference to others’ beliefs, and requires no
behavioral predictions of what agents will do on
the basis of their beliefs. Specifically, we use an
object detection task to investigate two questions.
First, are belief computations automatically trig-
gered by the mere presence of an agent in adults
and in infants as young as 7 months, even when
the beliefs are entirely irrelevant to the task par-
ticipants have to perform? Second, are beliefs
about others’ beliefs stored in a format sufficient-
ly similar to our own representations about the
environment that both types of representations
can affect our behavior?
In Experiment 1, adults (N = 24) performed a
visual detection task while watching 40 animated
movies (13). As shown in Fig. 1, movies started
with an agent placing a ball on a table in front of
an occluder. Then the ball rolled behind the
occluder. After this, the movies could continue in
four ways depending on the experimental con-
ditions. Our critical manipulations involved the
participant’s beliefs about the ball’s presence and
the “beliefs” of the agent, such that the agent, the
participant, both, or neither could believe that the
ball was behind the occluder. This was achieved
by varying (i) the final location of the ball and (ii)
the time at which the agent left the scene. Spe-
cifically, (i) participants saw the ball either staying
behind the occluder or leaving the scene and (ii)
the agent left the scene either before or after the
ball had reached its final location (leading to a
true/false belief). That is, the agent had a true
belief about the ball’s location if he left the scene
Fig. 1. Logical structure of events in Experiment 1. (A) In all four conditions, the agent enters the scene,
after the ball had reached its final location; if he
placing a ball on a table (13) (Movie S1). The ball then rolls behind an occluder. (B) In the agent’s presence,
the ball stays behind the occluder (a and c), or leaves the scene (b and d). As a result, the agent (A) “believes” left the scene before the ball reached its final lo-
either that the ball is behind the occluder or that there is no ball behind the occluder. Then, the agent leaves cation, his belief was false.
the scene. (C) In the agent’s absence, the ball leaves the scene (c), returns behind the occluder (d), or does At the end of each movie, the agent reentered
not move (a and b). Thus, the participant (P) either believes the ball to be behind the occluder (a and d), or to the scene and the occluder was lowered. The four
have left (b and c). (D) The agent reenters the scene, and the occluder is lowered. In half of the trials of all conditions were paired with two outcomes, in
conditions, participants see the ball behind the occluder. We measure ball detection latencies as a function which the ball was either present or absent behind
of (i) the participant’s belief (P+, ball behind occluder, versus P–, no ball behind occluder) and (ii) the the occluder. Participants were instructed to press
agent’s “belief” (A+, ball behind occluder, versus A–, no ball behind occluder), resulting in two true belief a button as soon as they detected the ball. Notably,
conditions and two false belief conditions. The figure does not reflect the actual timing of the events. To the agent’s beliefs were never mentioned and were
control for the timing differences, we used pairs of conditions matched for their timing properties (13). irrelevant to the task.

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1831

 REPORTS
We will discuss the experimental conditions
in terms of the beliefs of the participant and the
agent, respectively, rather than in terms of the events
displayed in the animations. (The agent’s “belief ” is
what he might believe based on what he “expe-
riences” in the scenes if he were a real person.)
We predicted that participants should be faster
to detect the ball when they believed that the ball
is behind the occluder than when they do not
believe so. Crucially, and as mentioned above,
the agent’s beliefs were completely irrelevant to
the task. As a result, if others’ beliefs are com-
puted through explicit processes requiring effort-
ful computations, the agent’s “beliefs” should
have no effect on reaction times (RTs), because
participants were not required to perform belief
computations. However, if participants automat-
ically compute the agent’s beliefs and store them
in a way similar to that of their own representa-
tions about the environment, their detection la-
tencies should also be faster when only the agent
“believes” that the ball is behind the occluder.
We will compare the experimental conditions
to a baseline condition where neither the partic-
ipant nor the agent believed the ball to be behind
the occluder; as a result, there were no belief rep-
resentations that could speed up RTs. To validate
our paradigm, we compared RTs in this base-
line condition to the condition where both the
participant and the agent believed the ball to be
behind the occluder.
Compared to the baseline condition, partic-
ipants detected the ball faster when they (and
the agent) believed that the ball was behind the
occluder [t(23) = 3.47, P = 0.002] (Fig. 2A) (13).
Likewise, participants were faster than in the
baseline when they but not the agent believed
that the ball was behind the occluder [t(23) =
3.43, P = 0.002].
Our critical comparison involves the baseline
condition and the condition where only the agent
believed that the ball was behind the occluder.
Results showed that participants were faster than
in the baseline condition when only the agent be-
lieved the ball to be behind the occluder [t(23) =
2.42, P = 0.02]. This suggests that they computed
Fig. 2. (A) Results of Experiment
1 (adults; agent present in the last
scene). Ball detection latencies in
adults. Bars represent average la-
tencies, and error bars show SEM
(see Fig. 1 for condition labels). (B)
Results of Experiment 2 (adults; agent
absent in the last scene). Ball de-
tection latencies in adults when a
pile of boxes replaced the agent in
the very last scene (corresponding
to Fig. 1D). (C) Results of Experi-
ment 3 (adults; agent absent in
the entire movie). Ball detection
latencies in adults when no agent was present at all but a stationary pile of
boxes (represented by B in the panel) was present during the entire movie.
The signs in square brackets indicate the beliefs the agent would have had if
he had been present, as in Experiments 1 and 2. Thus, P+B[–] indicates the
1832 24 DECEMBER 2010
the agent’s belief and that this belief influenced
their behavior even though it was inconsistent
with their own belief. Moreover, RTs did not
differ significantly between the condition where
the participants themselves believed that the ball
was behind the occluder and the condition where
only the agent believed so [t(23) = 0.99, P = 0.33].
Thus, both types of belief representations speeded
up the participants’ RTs to similar extents, a result
consistent with the view that the agent’s beliefs
are stored similarly to participants’ own represen-
tations about the environment.
In Experiment 1, participants simply had to
detect a ball after watching a scene involving an
agent whose presence was irrelevant to the task.
Nevertheless, the agent’s beliefs influenced the
participants’ RTs to the same extent as their own
beliefs, suggesting that just seeing the agent au-
tomatically made participants compute his beliefs
and that these beliefs were represented and sus-
tained similarly to participants’ own beliefs.
These results may also provide an important
clue to a question that has remained elusive for
the last three decades, namely how false beliefs
might be computed. In Experiment 1, the agent
had a false belief because he left the scene before
the ball reached its final position. Possibly, par-
ticipants compute online the agent’s last belief and,
unless there is evidence that he updated it, maintain
it in parallel with their own beliefs. If so, the agent’s
belief should continue to influence their RTs even
in the agent’s absence; if the environment changes,
the agent’s beliefs will necessarily become false.
We tested this prediction in Experiment 2 with a
new group of participants (N = 24). This exper-
iment was identical to Experiment 1 except that
the agent did not return in the last phase of the
movies; instead, a pile of boxes entered the scene
before the occluder was lowered (13).
The results of Experiment 2 were similar to
those of Experiment 1. Participants were faster
than in the baseline condition when they and the
agent (who was not present when the occluder
was lowered) believed the ball to be behind the
occluder [t(23) = 2.83, P = 0.009] (Fig. 2B).
Likewise, RTs were faster than in the baseline
condition where the participant believes that the ball is behind the
occluder, and the motion of the ball corresponds to a condition in
Experiments 1 and 2 where the agent did not believe that the ball was
behind the occluder.
VOL 330 SCIENCE www.sciencemag.org
condition when they but not the agent believed
that the ball was there [t(23) = 3.2, P = 0.004].
Crucially, RTs were also faster than in the base-
line condition when only the agent believed that
the ball was behind the occluder [t(23) = 2.1, P =
0.04]. Thus, although the agent was not present
when participants detected the ball, his beliefs
continued to influence participants’ behavior.
Taken together, data from Experiments 1 and
2 show that when participants had to detect the
presence of the ball, their RTs were faster not
only when they believed the ball to be behind the
occluder but also when the agent believed so, ir-
respective of whether or not the agent was
present when participants gave their responses.
Although these results appear to reflect compu-
tations of the agent’s beliefs, differences in the
ordering of events in the different experimental
Downloaded from www.sciencemag.org on December 23, 2010
conditions might possibly affect participants’ RTs
as well. Specifically, in our experimental design,
some conditions required the agent to leave the
scene before the ball reached its final location
(resulting in a false belief), whereas other condi-
tions required the agent to leave after the ball
reached its final location (resulting in a true be-
lief). Despite its plausibility, further analysis of
our results did not support this possibility, as the
ordering differences (e.g., the number of events
that occur after the ball reached its target location)
did not predict participants’ RTs in the four
experimental conditions (13).
Experiment 3 was designed to further confirm
that the RT differences between the crucial condi-
tions in Experiments 1 and 2 were due to the
agent’s beliefs and not to other perceptual dif-
ferences between the conditions. In Experiment
3, participants (N = 24) were presented with movies
that were similar to those shown in Experiments
1 and 2, except that the agent did not appear in
the movies at all. Instead, a stationary pile of
boxes was present in all movies during their en-
tire duration. However, all other events (e.g., the
movements of the ball) were as in Experiments
1 and 2. Hence, while participants in Experiment
3 could not compute another agent’s beliefs (be-
cause there was no agent present), they experienced

motion paths of the ball identical to those in the
different belief conditions from Experiments
1 and 2. If the differences between the critical
conditions in Experiments 1 and 2 were due to
perceptual differences rather than to belief com-
putations, Experiment 3 should yield the same
results, because the motion of the ball is identical.
In contrast, if these differences were due to belief
computations rather than to perceptual differ-
ences, RTs in Experiment 3 should be affected
only by participants’ own beliefs, since participants
never saw an agent and, therefore, could not com-
pute his beliefs.
Results showed that RTs were faster in the
two conditions when participants believed the ball
to be behind the occluder compared to the two
conditions when participants did not believe
the ball to be behind the occluder, with no other
esis was about 3 times more likely than the
alternative hypothesis (13). Hence, when the
agent was not present, participants’ RTs were
influenced only by their own beliefs about the
presence of the ball but not by other perceptual
differences between the conditions.
Together, Experiments 1 to 3 suggest that
adults automatically compute and store the be-
liefs of other agents; the resulting representations
appear to be similarly accessible to other cog-
nitive processes as are their own beliefs. Once a
belief is computed, it seems to remain active even
in the absence of the agent, possibly to be used
for future predictions about the agent’s behavior.
If adults track others’ beliefs automatically,
such processes may well be present in young
infants. We thus asked whether 7-month-olds
would automatically compute the beliefs of an
REPORTS
prise” (indicated by longer looking times) when
no ball was found, although the participant and/or
the agent believed the ball to be behind the
occluder. In each experiment, the condition where
the participant and/or the agent believed the ball
to be behind the occluder was compared to a
condition where neither the participant nor the
agent believed the ball to be behind the occluder.
In Experiment 4, infants watched two movies
from Experiment 1 (13). In the baseline condi-
tion, neither the infant nor the agent believed that
the ball was behind the occluder. We compared
this baseline to a condition where both the infant
and the agent believed that the ball was behind
the occluder. When the occluder revealed no ball,
infants looked longer in this condition than in the
baseline condition [F(1,13) = 5.65, P = 0.03] (Fig.
3A), which suggests that their expectations mod-

Downloaded from www.sciencemag.org on December 23, 2010

differences [all P’s < 0.05; see (13) for details] agent and whether these beliefs would influence ulated their looking behavior.
(Fig. 2C). Contrary to Experiments 1 and 2, RTs their looking times in a violation of expectation Experiment 5 presents the crucial comparison
did not differ between the P–B[–] and P–B[+] paradigm. We tested this possibility in four ex- from Experiment 1. In this experiment, infants’
conditions [t(23) = 0.76, P = 0.45], which cor- periments involving four different groups of looking times were compared in the baseline
respond to the P–A– and P–A+ conditions in the infants (N = 56). Whereas Experiments 1 and 2 condition and in a condition where only the agent
first two experiments. To further confirm that RTs measured how beliefs about the presence of a ball believed the ball to be behind the occluder (Movie
did not differ between these conditions, likeli- influenced RTs when adult participants saw the S1). When no ball appeared behind the occluder,
hood ratio analyses showed that the null hypoth- ball, Experiments 4 to 7 measured infants’ “sur- infants looked longer in this condition than in the
baseline [F(1,13) = 7.29, P = 0.01] (Fig. 3B).
This suggests that infants computed the agent’s
belief and looked longer when this belief was not
confirmed, possibly also expecting the agent to
be surprised. Thus, the beliefs of the agent influ-
enced the infants’ looking behavior, even though
they clashed with the infants’ own beliefs.
Experiment 6 controls for the possibility that
infants’ looking times in Experiment 5 were not
influenced by the agent’s beliefs but rather by
some visual differences between the movies
occurring before the occluder was lowered. (The
movies were identical after the occluder was
lowered.) We exclude this possibility by replicat-
ing Experiment 5, but without lowering the occluder
at the end. Thus, infants did not see whether the
ball was present behind the occluder, and neither
their own nor the agent’s beliefs were confirmed
or disconfirmed. If the results of Experiment 5
were due to visual differences rather than to be-
lief computations, Experiment 6 should yield the
same results. In contrast to this prediction, no
differences were observed when the occluder was
not lowered [F(1,13) = 0.05, P = 0.81] (Fig. 3C).
This suggests that the differences in Experiment
Fig. 3. Results of Experiments 4 to 7. Looking times in 7-month-old infants. Bars represent average 5 were not due to visual differences between the
looking times, and error bars show SEM (see Fig. 1 for condition labels). (A) Results of Experiment 4 (true movies, but rather that infants did indeed com-
belief). Looking times for the condition when infants (and the agent) believed the ball to be behind the
pute the agent’s beliefs.
occluder (P+A+) and for the condition when neither the infants nor the agent believed the ball to be
In analogy to Experiment 2, Experiment 7
behind the occluder (P–A–). (B) Results of Experiment 5 (false belief; agent present in the last scene).
Looking times for the condition when only the agent (falsely) believed the ball to be behind the occluder asked whether infants would maintain others’
(P–A+) (Movie S1), and for the condition when neither they nor the agent believed the ball to be behind beliefs even in the agent’s absence. Specifically,
the occluder (P–A–). (C) Results of Experiment 6 (no outcome control). Looking times for two conditions infants were presented with the baseline condi-
that were identical to the ones used in Experiment 4, except that the occluder was not lowered at the end tion (where both the infant and the agent believed
of the movies. Thus, infants did not see whether the ball was present behind the occluder. As a result, there that the ball was not there) and a condition where
were no confirmed nor violated beliefs. (D) Results of Experiment 7 (false belief; agent absent in the last only the agent believed the ball to be behind the
scene). Looking times for the two conditions where the agent was replaced with a pile of boxes in the very occluder. Before the occluder was lowered, how-
last scene (corresponding to Fig. 1D). We compared the condition where only the agent (falsely) believed ever, a pile of boxes, rather than the agent, entered
the ball to be behind the occluder (P–A+) with the condition where neither the infants nor the agent the scene. As in Experiment 5, infants looked lon-
believed the ball to be behind the occluder (P–A–). ger than in the baseline condition when the agent

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1833

 REPORTS
(who was not present when the occluder was
lowered) believed that the ball was behind the
occluder [F(1,13) = 6.75, P = 0.02] (Fig. 3D).
Hence, like adults in Experiment 2, infants seem to
compute others’ beliefs online and to maintain
them even in the absence of the agent. Possibly,
the boxes could have prompted participants to
think of the agent and his beliefs, although there
was no relation between the boxes and the agent.
However, even if the boxes reminded participants
the agent, our results can be explained only if par-
ticipants computed the agent’s beliefs and sus-
tained them even though the agent was not present.
Together, our results suggest that the mere
presence of social agents is sufficient to automat-
ically trigger online belief computations not only
in adults, but also in 7-month-old infants. Once
the beliefs have been computed, adults and in-
least in implicit tasks like ours, others’ (false) be-
liefs can influence infants’ and adults’ behavior
similarly to their own (true) beliefs. The finding
that others’ beliefs can be similarly accessible as
our own beliefs might seem problematic for an
individual, because it may make one’s behavior
susceptible to others’ beliefs that do not reliably
reflect the current state of affairs. However, the
rapid availability of others’ beliefs might allow
for efficient interactions in complex social groups.
These powerful mechanisms for computing others’
beliefs might, therefore, be part of a core human-
specific “social sense,” and one of the cognitive
preconditions for the evolution of the uniquely
elaborate social structure in humans.
References and Notes
6. A. M. Leslie, O. Friedman, T. P. German, Trends Cogn. Sci.
8, 528 (2004).
7. P. Bloom, T. P. German, Cognition 77, B25
(2000).
8. J. A. Fodor, Cognition 44, 283 (1992).
9. Á. M. Kovács, Dev. Sci. 12, 48 (2009).
10. K. H. Onishi, R. Baillargeon, Science 308, 255
(2005).
11. V. Southgate, A. Senju, G. Csibra, Psychol. Sci. 18,
587 (2007).
12. L. Surian, S. Caldi, D. Sperber, Psychol. Sci. 18, 580
(2007).
13. Materials and methods are available as supporting
material on Science Online.
14. G. Csibra, Cognition 107, 705 (2008).
15. This research was supported by the New and Emerging
Science and Technology PATHFINDER initiative CALACEI
and the Marie Curie Disorders and Coherence of the
Embodied Self Research Training Network. We thank
G. Csibra, G. Gergely, J. Perner, H. Wellman, M. Hauser,
P. Jacob, L. Bonatti, and J. Mehler for comments.

Downloaded from www.sciencemag.org on December 23, 2010

fants maintain them even in the absence of the
agent, presumably for later use in social inter-
actions. Hence, from 7 months on, an age by
which infants attribute goals and intentionality
(14), humans automatically compute other’s be-
liefs and seem to hold them in mind as alternative
representations of the environment. As a result, at
1. E. Herrmann, J. Call, M. V. Hernàndez-Lloreda, B. Hare,
M. Tomasello, Science 317, 1360 (2007). Supporting Online Material
2. D. Premack, G. Woodruff, Behav. Brain Sci. 1, 515 www.sciencemag.org/cgi/content/full/330/6012/1830/DC1
Materials and Methods
(1978).
SOM Text
3. S. Baron-Cohen, A. M. Leslie, U. Frith, Cognition 21,
37 (1985). References
4. J. Perner, Understanding the Representational Mind Movie S1
(MIT Press, Cambridge, MA, 1991). 12 April 2010; accepted 22 November 2010
5. B. J. Scholl, A. M. Leslie, Child Dev. 72, 696 (2001). 10.1126/science.1190792

granule layer (IGL). In this study, we examined

Siah Regulation of Pard3A Controls the roles of the partitioning-defective (PAR) polarity-
signaling complex and an upstream regulator in
Neuronal Cell Adhesion During controlling CGN migration from the EGL, a GZ
niche (fig. S1).

Germinal Zone Exit The PAR complex is an evolutionarily con-

served multiprotein complex containing orthologs
of partitioning defective-6 (Pard6), partitioning
Jakub K. Famulski,*† Niraj Trivedi,* Danielle Howell, Yuan Yang,‡ Yiai Tong, defective-3 (Pard3) and protein kinase Cz (PKCz)
Richard Gilbertson, David J. Solecki§ that regulates many polarized cellular processes,
like cell motility, asymmetric cell division, and
The brain’s circuitry is established by directed migration and synaptogenesis of neurons during epithelial junction formation (21). Because Pard3A
development. Although neurons mature and migrate in specific patterns, little is known about how protein expression is low in the EGL (Fig. 1, A
neurons exit their germinal zone niche. We found that cerebellar granule neuron germinal zone to C), we examined whether elevated Pard3A
exit is regulated by proteasomal degradation of Pard3A by the Seven in Absentia homolog (Siah) activity induces CGN GZ exit. Expression con-
E3 ubiquitin ligase. Pard3A gain of function and Siah loss of function induce precocious radial structs for Pard3A and the fluorescent nuclear
migration. Time-lapse imaging using a probe to measure neuronal cell contact reveals that Pard3A reporter H2B-mCherry were co-electroporated
promotes adhesive interactions needed for germinal zone exit by recruiting the epithelial tight into the cerebellar cortices of postnatal day 8 (P8)
junction adhesion molecule C to the neuronal cell surface. Our findings define a Siah-Pard3A mice, and cerebellar slices were cultured ex vivo
signaling pathway that controls adhesion-dependent exit of neuronal progenitors or immature (22). Whereas control CGNs remained within the
neurons from a germinal zone niche. EGL after 24 hours (fig. S2A), CGNs expressing
elevated Pard3A entered the ML and IGL (fig.
he migration of neurons from a germinal cognitive disorders (4). Although the substrates S2B), suggesting that elevated Pard3A expres-

T zone (GZ) to their final laminar positions

is essential for morphogenesis of the de-
veloping brain (1–3); aberrations in this process
are linked to profound neurodevelopmental and
Department of Developmental Neurobiology, St. Jude Chil-
dren’s Research Hospital, 262 Danny Thomas Place, Memphis,
TN 38105, USA.
*These authors contributed equally to this work.
†Present address: Department of Biological Sciences, Univer-
sity of Alberta, Edmonton, Alberta T6G 2E9, Canada.
‡Present address: Department of Physiology, Development and
Neuroscience, Anatomy Building, University of Cambridge,
Downing Street, Cambridge CB2 3DY, UK.
§To whom all correspondence should be addressed. E-mail:
david.solecki@stjude.org
(5–7), guidance mechanisms (8–10), cytoskeletal
elements (11–13), and posttranslational modifica-
tions (14–16) required for neuronal migrations are
well established, the cell-intrinsic machinery reg-
ulating when neurons gain access to permissive
migration pathways to exit their GZs are uniden-
tified (17). Developing cerebellar granule neurons
(CGNs) are an excellent model to analyze the
mechanisms regulating GZ exit and to elucidate
migration pathway selection, because they under-
go two migration phases (18–20): tangential mi-
gration near the cerebellar surface followed by
radial migration away from the external granular
layer (EGL) where CGNs cross the molecular
layer (ML) to eventually reside within the internal
sion is sufficient to induce precocious GZ exit.
We next examined the role of Siah, a PAR
complex–interacting E3 ubiquitin ligase (fig. S3)
expressed in the EGL (Fig. 1D), in regulating
Pard3A protein level and PAR complex–dependent
GZ exit. The role of Siah ligases in the morpho-
genesis of the vertebrate nervous system has not
previously been examined (23). Epitope-tagged
Siah1B immunoprecipitated Pard3A when coex-
pressed in human embryonic kidney 293 (HEK293)
cells (fig. S3B), an interaction that required an
intact Siah substrate-binding domain (fig. S3D).
Furthermore, Siah1B expression reduced expres-
sion of Pard3A, but not Pard6 or PKCz, protein
(Fig. 1E) and induced Pard3A ubiquitination (Fig.

1834 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

 REPORTS
(who was not present when the occluder was
lowered) believed that the ball was behind the
occluder [F(1,13) = 6.75, P = 0.02] (Fig. 3D).
Hence, like adults in Experiment 2, infants seem to
compute others’ beliefs online and to maintain
them even in the absence of the agent. Possibly,
the boxes could have prompted participants to
think of the agent and his beliefs, although there
was no relation between the boxes and the agent.
However, even if the boxes reminded participants
the agent, our results can be explained only if par-
ticipants computed the agent’s beliefs and sus-
tained them even though the agent was not present.
Together, our results suggest that the mere
presence of social agents is sufficient to automat-
ically trigger online belief computations not only
in adults, but also in 7-month-old infants. Once
the beliefs have been computed, adults and in-
least in implicit tasks like ours, others’ (false) be-
liefs can influence infants’ and adults’ behavior
similarly to their own (true) beliefs. The finding
that others’ beliefs can be similarly accessible as
our own beliefs might seem problematic for an
individual, because it may make one’s behavior
susceptible to others’ beliefs that do not reliably
reflect the current state of affairs. However, the
rapid availability of others’ beliefs might allow
for efficient interactions in complex social groups.
These powerful mechanisms for computing others’
beliefs might, therefore, be part of a core human-
specific “social sense,” and one of the cognitive
preconditions for the evolution of the uniquely
elaborate social structure in humans.
References and Notes
6. A. M. Leslie, O. Friedman, T. P. German, Trends Cogn. Sci.
8, 528 (2004).
7. P. Bloom, T. P. German, Cognition 77, B25
(2000).
8. J. A. Fodor, Cognition 44, 283 (1992).
9. Á. M. Kovács, Dev. Sci. 12, 48 (2009).
10. K. H. Onishi, R. Baillargeon, Science 308, 255
(2005).
11. V. Southgate, A. Senju, G. Csibra, Psychol. Sci. 18,
587 (2007).
12. L. Surian, S. Caldi, D. Sperber, Psychol. Sci. 18, 580
(2007).
13. Materials and methods are available as supporting
material on Science Online.
14. G. Csibra, Cognition 107, 705 (2008).
15. This research was supported by the New and Emerging
Science and Technology PATHFINDER initiative CALACEI
and the Marie Curie Disorders and Coherence of the
Embodied Self Research Training Network. We thank
G. Csibra, G. Gergely, J. Perner, H. Wellman, M. Hauser,
P. Jacob, L. Bonatti, and J. Mehler for comments.

Downloaded from www.sciencemag.org on December 23, 2010

fants maintain them even in the absence of the
agent, presumably for later use in social inter-
actions. Hence, from 7 months on, an age by
which infants attribute goals and intentionality
(14), humans automatically compute other’s be-
liefs and seem to hold them in mind as alternative
representations of the environment. As a result, at
1. E. Herrmann, J. Call, M. V. Hernàndez-Lloreda, B. Hare,
M. Tomasello, Science 317, 1360 (2007). Supporting Online Material
2. D. Premack, G. Woodruff, Behav. Brain Sci. 1, 515 www.sciencemag.org/cgi/content/full/330/6012/1830/DC1
Materials and Methods
(1978).
SOM Text
3. S. Baron-Cohen, A. M. Leslie, U. Frith, Cognition 21,
37 (1985). References
4. J. Perner, Understanding the Representational Mind Movie S1
(MIT Press, Cambridge, MA, 1991). 12 April 2010; accepted 22 November 2010
5. B. J. Scholl, A. M. Leslie, Child Dev. 72, 696 (2001). 10.1126/science.1190792

granule layer (IGL). In this study, we examined

Siah Regulation of Pard3A Controls the roles of the partitioning-defective (PAR) polarity-
signaling complex and an upstream regulator in
Neuronal Cell Adhesion During controlling CGN migration from the EGL, a GZ
niche (fig. S1).

Germinal Zone Exit The PAR complex is an evolutionarily con-

served multiprotein complex containing orthologs
of partitioning defective-6 (Pard6), partitioning
Jakub K. Famulski,*† Niraj Trivedi,* Danielle Howell, Yuan Yang,‡ Yiai Tong, defective-3 (Pard3) and protein kinase Cz (PKCz)
Richard Gilbertson, David J. Solecki§ that regulates many polarized cellular processes,
like cell motility, asymmetric cell division, and
The brain’s circuitry is established by directed migration and synaptogenesis of neurons during epithelial junction formation (21). Because Pard3A
development. Although neurons mature and migrate in specific patterns, little is known about how protein expression is low in the EGL (Fig. 1, A
neurons exit their germinal zone niche. We found that cerebellar granule neuron germinal zone to C), we examined whether elevated Pard3A
exit is regulated by proteasomal degradation of Pard3A by the Seven in Absentia homolog (Siah) activity induces CGN GZ exit. Expression con-
E3 ubiquitin ligase. Pard3A gain of function and Siah loss of function induce precocious radial structs for Pard3A and the fluorescent nuclear
migration. Time-lapse imaging using a probe to measure neuronal cell contact reveals that Pard3A reporter H2B-mCherry were co-electroporated
promotes adhesive interactions needed for germinal zone exit by recruiting the epithelial tight into the cerebellar cortices of postnatal day 8 (P8)
junction adhesion molecule C to the neuronal cell surface. Our findings define a Siah-Pard3A mice, and cerebellar slices were cultured ex vivo
signaling pathway that controls adhesion-dependent exit of neuronal progenitors or immature (22). Whereas control CGNs remained within the
neurons from a germinal zone niche. EGL after 24 hours (fig. S2A), CGNs expressing
elevated Pard3A entered the ML and IGL (fig.
he migration of neurons from a germinal cognitive disorders (4). Although the substrates S2B), suggesting that elevated Pard3A expres-

T zone (GZ) to their final laminar positions

is essential for morphogenesis of the de-
veloping brain (1–3); aberrations in this process
are linked to profound neurodevelopmental and
Department of Developmental Neurobiology, St. Jude Chil-
dren’s Research Hospital, 262 Danny Thomas Place, Memphis,
TN 38105, USA.
*These authors contributed equally to this work.
†Present address: Department of Biological Sciences, Univer-
sity of Alberta, Edmonton, Alberta T6G 2E9, Canada.
‡Present address: Department of Physiology, Development and
Neuroscience, Anatomy Building, University of Cambridge,
Downing Street, Cambridge CB2 3DY, UK.
§To whom all correspondence should be addressed. E-mail:
david.solecki@stjude.org
(5–7), guidance mechanisms (8–10), cytoskeletal
elements (11–13), and posttranslational modifica-
tions (14–16) required for neuronal migrations are
well established, the cell-intrinsic machinery reg-
ulating when neurons gain access to permissive
migration pathways to exit their GZs are uniden-
tified (17). Developing cerebellar granule neurons
(CGNs) are an excellent model to analyze the
mechanisms regulating GZ exit and to elucidate
migration pathway selection, because they under-
go two migration phases (18–20): tangential mi-
gration near the cerebellar surface followed by
radial migration away from the external granular
layer (EGL) where CGNs cross the molecular
layer (ML) to eventually reside within the internal
sion is sufficient to induce precocious GZ exit.
We next examined the role of Siah, a PAR
complex–interacting E3 ubiquitin ligase (fig. S3)
expressed in the EGL (Fig. 1D), in regulating
Pard3A protein level and PAR complex–dependent
GZ exit. The role of Siah ligases in the morpho-
genesis of the vertebrate nervous system has not
previously been examined (23). Epitope-tagged
Siah1B immunoprecipitated Pard3A when coex-
pressed in human embryonic kidney 293 (HEK293)
cells (fig. S3B), an interaction that required an
intact Siah substrate-binding domain (fig. S3D).
Furthermore, Siah1B expression reduced expres-
sion of Pard3A, but not Pard6 or PKCz, protein
(Fig. 1E) and induced Pard3A ubiquitination (Fig.

1834 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org


1F). Pard3A protein levels were also reduced by
Siah2 but not by a dominant negative mutant li-
gase lacking the catalytic RING domain (Siah1B-
DRING, fig. S3E). The Pard3A protein sequence
contains two Siah degron recognition sequences
Pro-x(Ala,Thr,Arg)x-Val-x-Pro (where x is any amino
acid) (fig. S3C) (24). Mutation of the Val-x-Pro
(VxP) core of both Pard3A degrons to Asn-x-Asn
(NxN) or treatment with MG132 attenuated Siah-
mediated reduction of Pard3A protein (fig. S3E)
and blocked Siah-induced ubiquitination (fig. S3F;
see figs. S2, B and C, and S4 for additional func-
tional differences between Pard3A and the NxN
mutant in CGNs). Lastly, Venus-Pard3A fluores-
cence signal in purified CGNs was diminished by
Siah1B coexpression but not by Siah1B-DRING
(fig. S3G). Therefore, Siah regulates Pard3A protein
level through VxP degron-dependent ubiquitination.
Siah1 and 2 immunoreactivity in the EGL is
elevated at P6, the peak time of CGN neuro-
genesis (Fig. 1D) and declines at P15 (see fig. S5
for comparison of Siah1 versus Siah2 expres-
sion). To detect sites of Siah activity in the de-
veloping cerebellum, we fused the Siah degron
motif to the Venus fluorescent protein (fig. S6)
and electroporated expression vectors for this
REPORTS
Siah sensor in conjunction with H2B-mCherry
into P8 EGL. Siah sensor fluorescence was scant
in the EGL but visible in the ML and IGL (fig.
S7), whereas a negative control sensor containing
the NxN mutations showed fluorescence in all
cerebellar layers. These results suggest that Siah
activity is high in CGNs within the EGL and
decreases during GZ exit.
To assess Siah function in GZ exit, we elec-
troporated short hairpin RNAs (shRNAs) silenc-
ing Siah1B and Siah2 (fig. S8A) or an expression
construct for dominant-negative Siah1B-DRING
into P8 EGL. After 1 day of ex vivo culture, CGNs

Fig. 1. Siah is highly ex- Pard3A B sense E FLAG-Pard3A FLAG-Par6 FLAG-PKCζ

pressed in the EGL and A
ubiquitinates Pard3A pro- α−FLAG
tein. (A and B) In situ

Downloaded from www.sciencemag.org on December 23, 2010

hybridization shows that α-myc
Pard3A mRNA is ex-
pressed throughout the β actin
P6 EGL. Scale bar indicates
80 mm. (C) Immunohisto-
+ - + - + Siah1B-myc -
C D F Siah1B-myc + - -
chemistry of P6 mouse
cerebellum for Pard3A Siah1B-∆RING-myc + - -
(green) and Tuj1 (red). Venus-Pard3A + + +
Pard3A expression is low oEGL HA-Ubq + + +
in the outer EGL (oEGL) oEGL
and higher in differenti- iEGL Venus-Pard3A
(α GFP)
ated CGNs in the inner iEGL
EGL (iEGL). Scale bar, 50
mm. (D) Immunohisto- α HA
chemistry of P6 mouse
cerebellum for Siah1 and Pard3A/ Tuj1 Siah1/2/ Tuj1
2 (green) and Tuj1 (red). IP-α-GFP
Siah expression is high
in the oEGL and absent in the iEGL. (E) Expression of Siah1B-myc reduces Pard3A but not Par6 or PKCz protein levels in HEK293 cells. (F) Siah1B-myc but
not Siah1B-DRING-myc can induce ubiquitination of Venus-Pard3A in HEK293 cells. IP, immunoprecipitation; GFP, green fluorescent protein.

70

A EGL
Control Siah1 shRNA Siah2 shRNA EGL Siah1B-∆RING B 60

EGL 50
% Frequency

EGL ML
ML
40

ML 30
ML IGL
IGL
20
IGL
10
IGL
0
0 50 100 150 200 250
Distance migrated (µm)
60

C EGL Control Siah1B Siah2 Siah1B + Pard3A D 50

EGL EGL
% Frequency

ML 40
ML ML
EGL 30

IGL 20
IGL ML IGL
10

IGL 0
0 50 100 150 200 250
Distance migrated (µm)

Fig. 2. Siah activity attenuates GZ exit by negatively regulating Pard3A.
P8 EGL was co-electroporated with the indicated expression constructs and
H2B-mCherry. After 24 (A) or 48 (C) hours of ex vivo culture, the migration
distance of H2B-labeled CGNs from the pial layer (outer dashed line) was
analyzed in three imaging experiments. (A) Most control and Siah1B-
silenced cells remain within the EGL (dashed lines) at 24 hours, whereas
Siah2-silenced and Siah1B-DRING–overexpressing cells prematurely en-
tered the ML and IGL. (B) Migration distance versus frequency plot of con-
trol (n = 872, black), Siah1B-silenced (n = 960, red), Siah2-silenced (n =
926, blue), and Siah1B-DRING–overexpressing (n = 927, green) CGNs.
(C) Whereas control cells entered the ML and IGL after 48 hours, Siah1B-
and Siah2-overexpressing cells remained in the EGL. Addition of Pard3A to
Siah-expressing CGNs restored migration. (D) Migration distance versus
frequency plot of control (n = 909, black), Siah1B-expressing (n = 970,
red), Siah2-expressing (n = 921, blue), and Siah1B+Pard3A-expressing (n =
919, green) CGNs.

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1835

 REPORTS
transfected with a control shRNA remained in
the EGL. Silencing of Siah2, but not of Siah1B,
and expression of Siah1B-DRING increased
CGN migration toward the IGL (Fig. 2, A and
B, and fig. S9A for additional analysis). We next
evaluated whether Siah activity regulates CGN
GZ exit by using a gain-of-function approach.
Expression vectors for mouse Siah1B or Siah2
were electroporated into P8 EGL. After 2 days
of ex vivo culture, control CGNs entered the ML
and IGL (Fig. 2, C and D, and fig. S9B for
additional analysis), whereas Siah1B- or Siah2-
expressing CGNs remained within the EGL. EGL
exit was similarly blocked by Pard3A silencing
(fig. S8, B and C). Elevated Pard3A expression
in Siah-expressing cells rescued the Siah pheno-
type, confirming that Siah-mediated blockade of
EGL exit is not only reversible but also Pard3A-
corporation. Whereas elevated Siah2 activity has
no effect on EdU incorporation, Siah2 silencing,
and Pard3A gain of function, conditions that
induce early GZ exit significantly reduced EdU
incorporation, suggesting a linkage between mi-
gration initiation and cell cycle exit (fig. S10).
Having found that Siah activity controls CGN
GZ exit, we next examined whether Siah regu-
lates CGN migration mode. We introduced ex-
pression vectors for mouse Siah1B or Siah1B
plus Pard3A into purified CGNs via nucleofection
and examined migration by time-lapse micros-
copy of microcultures. Control CGNs exhibited
radial-like migration that persisted in the direc-
tion of leading process extension (Fig. 3A, movie
S1, and figs. S11 and S12). Siah-expressing cells
were motile but did not elaborate long leading
processes or neurites and were less directionally
not alter migration (movie S3). Lastly, increased
Pard3A expression restored normal leading pro-
cess extension (fig. S11C) and directional per-
sistence to Siah-expressing neurons (Fig. 3A
and movie S4). Therefore, the antagonistic rela-
tion between Siah activity and Pard3A regulates
the directional persistence of migrating CGNs
but not cell motility.
To examine CGN migration pathway selec-
tion, we electroporated P8 EGL with Siah1B or
Siah1B plus Pard3A expression constructs and
examined the migration of H2B-mCherry–labeled
CGNs by long-term time-lapse microscopy.
Control CGNs migrated extensively parallel to
the cerebellar slice surface before migrating ra-
dially toward the IGL (Fig. 3B, movie S5, and
fig. S12E). Elevated Siah activity reduced radial
migration and increased the percentage of neu-

Downloaded from www.sciencemag.org on December 23, 2010

dependent (Fig. 2, C and D). Lastly, we assessed persistent (Fig. 3A and movie S2). Ubiquitin rons migrating toward the pial surface, whereas
proliferation status of CGNs in our experiments ligase activity was required for the Siah pheno- the percentage of CGNs migrating tangentially
by using 5-ethynyl-2'-deoxyuridine (EdU) in- type, because Siah1B-DRING expression did within the EGL was unaffected; this phenotype

Fig. 3. Siah activity and

Pard3A regulate transi- A In vitro migration analysis B Ex vivo migration analysis
Minutes 0 25 50 75 100 125 150
tion from tangential to
radial migration. (A) Pu-
0 0
rified CGNs were nucleo- 12 25
fected with Centrin2-Venus, 20
8
H2B-mCherry, and the in- 15
10
dicated constructs. After EGL

% Frequency
% Frequency

4
Control

5
18 hours in culture, frame-
0 90 270 0 90 270
to-frame migration angle ML
of H2B-mCherry–labeled
nuclei was analyzed in
three separate experiments. IGL
Pseudocolored nuclei show 180 180
representative migration
angle
paths. Control cells (n = 0 0
3140) and Siah1B+Pard3A- 12 25
angle 20
expressing cells (n = 8
4743) migrated predomi- EGL 15
10
% Frequency
% Frequency

4
nantly forward (315° to
Siah1B

5
45°), whereas the migra- 0 90 270
ML 0 90 270
tion pattern of Siah1B-
angle
expressing cells (n =
4480) was randomized. IGL
The quadrant (315° to 45°,
45° to 135°, 135° to 225°, 180 180
and 225° to 315°) distribu-
tion of Siah1B-expressing 0 0
12 25
cells differed significantly
20
from that of controls (P =
Siah1B+Pard3A

8
15
0.001, c2 test), whereas 10
% Frequency

4 EGL
% Frequency

that of Siah1B+Pard3A did 5

not (P = 0.670, c2 test). (B) 0 90 270 0 90 270
P8 EGL was electroporated ML
with H2B-mCherry and
the indicated constructs.
Cerebellar slices were incu- IGL
bated for 28 hours and 180 180
then imaged for 20 hours.
Migration endpoint angles of H2B-mCherry–labeled nuclei were tracked, whereas Siah1B-expressing cells (n = 454) display predominantly tangential
binned, and plotted in three separate experiments. Colored lines indicate and pial-directed migration (285° to 75°). Migration of cells expressing
migration paths; arrowheads indicate direction of migration. Control cells Siah1B differed significantly from that of controls (P = 2.056 × 10−8, c2
(n = 431) display tangential migration parallel to the EGL (255° to 285°/75° test), whereas that of cells expressing Siah1B+Pard3A did not (n = 452, P =
to 105°) and radial migration perpendicular to the EGL (105° to 255°), 0.301, c2 test).

1836 24 DECEMBER 2010 VOL 330 SCIENCE www.sciencemag.org

 REPORTS
was rescued by increased Pard3A expression with Pard3A and is required for epithelial adhe- tween JAM-C and Pard3A is required for nor-
(Fig. 3B and movies S6 and S8). Expression of sion (25). Immunostaining revealed that JAM-C mal and Siah2 loss of function–induced GZ exit
Siah1B-DRING, which appeared to drive GZ is not only expressed in differentiating CGNs but (fig. S15C).
exit (Fig. 2A), did not adversely affect move- also labels sites of CGN contact (fig. S14, A and Having found that JAM-C is required for GZ
ment to the IGL, in vitro directional persistence, B). We examined JAM-C function in GZ exit by exit, we next asked whether JAM-C mediated
or migration pathway selection (figs. S12 and electroporating P8 EGL with a shRNA silencing cell contacts are regulated by the Siah-Pard3A
S13). These results show that Siah activity in- JAM-C (fig. S15A) or with a dominant-negative pathway in Madin-Darby canine kidney (MDCK)
trinsically regulates CGN migration mode and fragment of the JAM-C cytoplasmic domain polarized epithelial cells, a well-established sys-
suggest that Siah regulation of Pard3A activity (JAM-C-DN) previously shown to competitively tem to assay PAR complex–dependent adhesion.
constitutes a switch controlling tangential versus inhibit Pard3A binding to endogenous JAM-C Elevated Siah1B expression dissolved ZO-1,
radial migration as CGNs exit the EGL. and block epithelial junction formation (25) (fig. JAM-C, and Pard3A-labeled MDCK junctions,
Although the PAR complex activates cyto- S15B). After 2 days of ex vivo culture, both JAM- and these effects were reversed by Pard3A ex-
skeletal elements that propel migration (13), we C–silenced and JAM-C-DN–expressing CGNs pression (fig. S16A). Elevated Siah1B expres-
hypothesized that the Siah-Pard3A module con- remained predominantly within the EGL, but sion also altered the localization of endogenous
trols adhesion during GZ exit because Pard3A is migration was unaffected by expression of JAM- JAM-C in cultured CGNs (fig. S16B). To direct-
essential for junction formation in epithelial cells. C-DND9, a JAM-C cytoplasmic domain lacking ly measure JAM-C contact in live CGNs, we de-
We therefore focused our approach on the anal- the Pard3A binding motif (fig. S15C). JAM-C- veloped a novel JAM-C–based fluorescent probe
ysis of junctional adhesion molecule C (JAM-C), DN expression inhibited migration induced by by fusing super ecliptic pHluorin, a pH-sensitive

Downloaded from www.sciencemag.org on December 23, 2010

a tight-junction component that directly interacts Siah2 silencing, indicating that interaction be- conditional fluorophore (26), with the extracel-

A 28 min 38 min 41 min 58 min

JAM-C-pHluorin

B Con Siah C JAMC JAMC-Nect3 D Siah+JAMC-Nect3

L
EG

ML
EGL
L
M

ML IGL
IG

IGL
Cell 1 Cell 1
Cell 2 Cell 2
Siah1B+Pard3A
Siah∆R
Average JamC-pHluorin puncta

Siah activity negatively regulates JAM-C JAM-C-Nectin3 Siah1B +

JamC-pHluorin adhesions
p = 6.4E-05 JAM-C-Nectin3
(µm2/cell)

p = 2.3E-08

0 45 90 135 180 225 270 315 360 0 45 90 135 180 225 270 315 360 0 45 90 135 180 225 270 315 360

Fig. 4. Siah activity regulates GZ exit by modulating the formation of P8 EGL were co-electroporated with the indicated constructs and H2B-
Pard3A-dependent JAM-C adhesions. (A) Time-lapse imaging of CGNs mCherry. After 24 or 48 hours of culture, the distance of H2B-labeled cells
electroporated to express JAM-C–pHluorin. Fluorescence is low before cell from the pial layer (outer dashed line) was analyzed in three separate
contact. Upon establishment of stable contacts, JAM-C-pHluorin signal experiments. Gray shading shows percentage of cells found in the EGL; red
fluorescence intensifies. (B) Purified CGNs were electroporated to coexpress overlay indicates the average migration distribution of control cells (error
JAM-C–pHluorin and the indicated constructs. After 18 hours, the abun- bars, SD). (C) JAM-C overexpression (n = 891) did not induce migration from
dance of JAM-C–pHluorin contacts was analyzed. Control cells (n = 64 cells, the EGL, but the JAM-C–Nectin3 fusion molecule (n = 874), a Pard3A-
16045 puncta) displayed robust JAM-C contacts. Siah expression (n = 37 independent JAM-C variant, induced CGN migration from the EGL at 24 hours.
cells, 2579 puncta) significantly reduced JAM-C–pHluorin–positive puncta. Control versus JAM-C, P = 0.58, and versus JAM-C–Nectin3, P = 5.50 × 10−26
Siah1B-DRING (n = 30 cells, 7589 puncta) and Siah1B+Pard3A (n = 32 cells, (c2 test). (D) JAM-C–Nectin3 (n = 971) expression rescues migration of
8069 puncta) cells were similar to controls, whereas Siah2 silencing (n = 34 Siah1B-expressing cells. Control versus Siah1B+JAM-C-Nectin3, P = 0.95 (c2
cells, 10709 puncta) increased JAM-C–pHluorin contact. (C and D) CGNs in test). Error bars indicate SD.

www.sciencemag.org SCIENCE VOL 330 24 DECEMBER 2010 1837

 REPORTS
lular domain of JAM-C (fig. S17A). High JAM-
C-pHluorin signal revealed sites of neuron-neuron
or neuron-glial contact (as indicated by f-actin
and Pard3A accumulation), validating the JAM-C
probe (Fig. 4A, fig.S17B, and movies S9 and
S10). We expressed JAM-C-pHluorin and Siah1B,
Siah1B-DRING, Siah1B plus Pard3A, or Siah2
shRNA into purified CGNs via nucleofection and
assayed cell contacts by time-lapse microscopy
of microcultures. Control CGNs displayed robust
contacts with neighboring cells, whereas Siah1B
gain of function inhibited JAM-C contacts (Fig.
4B and movies S11 and S12). Increased Pard3A
expression restored contact formation to near wild-
type levels (movie S14). Lastly, Siah2 silencing
increased JAM-C contact nearly twofold (movie
S15). Therefore, Siah activity inhibits Pard3A-
dependent JAM-C adhesion.
If Pard3A binding to the JAM-C cytoplasmic
domain is essential for JAM-C–dependent GZ
exit, as it is for epithelial tight junction forma-
tion (25), we hypothesized that swapping the
JAM-C cytoplasmic domain with that of a Pard3-
independent adhesion receptor would create a
Siah-insensitive receptor. We fused extracellular
domains of JAM-C to the cytoplasmic domain
of Nectin-3, a non-Pard3A–binding adhesion
receptor required for epithelial cell adherens junc-
tions (see fig. S15B for schematic) and electro-
porated P8 EGL with expression vectors for
wild-type mouse JAM-C and JAM-C–Nectin3.
Whereas CGNs expressing wild-type JAM-C
remained in the EGL, expression of JAM-C–
Nectin3 spurred early EGL exit (Fig. 4C). JAM-
C–Nectin3 expression also fully rescued the EGL
exit of CGNs expressing Siah1B, directly demon-
strating that elevated JAM-C cell contact over-
comes Siah inhibition of GZ exit (Fig. 4D).
Newborn neurons must exercise plasticity in
order to integrate into the vertebrate brain (27–29).
1838 24 DECEMBER 2010
Our results show that posttranslational ubiquiti-
nation of Pard3A by Siah controls whether ce-
rebellar granule neurons will follow the tangential
migration that keeps them within the EGL of the
mouse cerebellum or the radial migration path
they use to exit their germinal zone and migrate
to the IGL. Migration pattern plasticity is invoked
by the production of tight junction JAM-C cell
contacts necessary for the integration of new neu-
rons into the developing cerebellar cortex. These
cell interactions resemble those known to reg-
ulate mesenchymal epithelial transitions as new
epithelial cells integrate into developing epithe-
lia (30, 31).
References and Notes
1. P. Rakic, J. Comp. Neurol. 145, 61 (1972).
2. M. E. Hatten, Science 297, 1660 (2002).
3. C. Métin, R. B. Vallee, P. Rakic, P. G. Bhide, J. Neurosci.
28, 11746 (2008).
4. M. Kato, W. B. Dobyns, Hum. Mol. Genet. 12, R89 (2003).
5. G. Fishell, M. E. Hatten, Development 113, 755 (1991).
6. E. S. Anton, J. A. Kreidberg, P. Rakic, Neuron 22, 277
(1999).
7. L. A. Elias, D. D. Wang, A. R. Kriegstein, Nature 448, 901
(2007).
8. F. Polleux, K. L. Whitford, P. A. Dijkhuizen, T. Vitalis,
A. Ghosh, Development 129, 3147 (2002).
9. P. Zhou et al., Neuron 55, 53 (2007).
10. J. Renaud et al., Nat. Neurosci. 11, 440 (2008).
11. B. T. Schaar, S. K. McConnell, Proc. Natl. Acad. Sci. U.S.A.
102, 13652 (2005).
12. J. W. Tsai, K. H. Bremner, R. B. Vallee, Nat. Neurosci. 10,
970 (2007).
13. D. J. Solecki et al., Neuron 63, 63 (2009).
14. G. N. Patrick, P. Zhou, Y. T. Kwon, P. M. Howley,
L. H. Tsai, J. Biol. Chem. 273, 24057 (1998).
15. S. Suetsugu et al., Biochem. J. 384, 1 (2004).
16. O. Karakuzu, D. P. Wang, S. Cameron, Development 136,
943 (2009).
17. P. Rakic, in The Cell in Contact: Adhesion and Junctions
of Morphogenetic Determinants, G. M. Edelman,
J. P. Thiery, Eds. (Wiley, New York, 1985), pp. 67–91.
18. J. C. Edmondson, M. E. Hatten, J. Neurosci. 7, 1928
(1987).
19. E. F. Ryder, C. L. Cepko, Neuron 12, 1011 (1994).
VOL 330 SCIENCE www.sciencemag.org
20. H. Komuro, E. Yacubova, E. Yacubova, P. Rakic,
J. Neurosci. 21, 527 (2001).
21. D. J. Solecki, E. E. Govek, T. Tomoda, M. E. Hatten,
Genes Dev. 20, 2639 (2006).
22. Materials and methods are available as supporting
material on Science Online.
23. C. M. House, A. Möller, D. D. Bowtell, Cancer Res. 69,
8835 (2009).
24. C. M. House et al., Structure 14, 695 (2006).
25. K. Ebnet et al., EMBO J. 20, 3738 (2001).
26. G. Miesenböck, D. A. De Angelis, J. E. Rothman, Nature
394, 192 (1998).
27. S. C. Noctor, V. Martínez-Cerdeño, L. Ivic, A. R. Kriegstein,
Nat. Neurosci. 7, 136 (2004).
28. R. S. Bultje et al., Neuron 63, 189 (2009).
29. J. J. Yi, A. P. Barnes, R. Hand, F. Polleux, M. D. Ehlers,
Cell 142, 144 (2010).
30. J. P. Thiery, J. P. Sleeman, Nat. Rev. Mol. Cell Biol. 7, 131
(2006).
31. C. L. Chaffer, E. W. Thompson, E. D. Williams,
Cells Tissues Organs 185, 7 (2007).
32. We are grateful to J. Morgan, M. Dyer, M. Roussel, and
Downloaded from www.sciencemag.org on December 23, 2010
N. Ayad for critically reading the manuscript; A. Sawa,
S. Zakharenko, and B. Schulman for insightful
discussions; S. Connell, R. Baird, and K. Kilborn
(Intelligent Imaging Innovations) for timely data analysis
advice and support; G. Chan for providing space for some
of the revision experiments; and S. Naron for excellent
editorial assistance. G. Miesenbock and Memorial
Sloan-Kettering Cancer Center provided the pHluorin
reporter, A. Miyawaki provided Venus, F. Polleux
provided pCIG2, M. Aurrand-Lions provided anti-JAM-C
antiserum, I. Dikic provided the hemagglutinin-ubiquitin
(HA-Ubq) construct, and A. Sawa shared Siah reagents.
Supported by American Lebanese Syrian Associated
Charities (ALSAC, D.J.S.), a National Cancer Institute
Cancer Center Support Grant (D.J.S.), and a March of Dimes
Basil O’Connor Starter Scholar Research Award (D.J.S.).
Supporting Online Material
www.sciencemag.org/cgi/content/full/science.1198480/DC1
Materials and Methods
Figs. S1 to S17
References
Movies S1 to S15
30 September 2010; accepted 10 November 2010
Published online 25 November 2010;
10.1126/science.1198480
 LIFE SCIENCE TECHNOLOGIES
AAAS/Science Business Office

NEW PRODUCTS

HPLC Columns
The new Syncronis range of high-performance liquid chromatography (HPLC) columns
unerringly delivers consistent, predictable separations, from run to run and column to
column. When developing a new method, one of the key goals for the chromatogra-
pher is to achieve a consistent, reproducible separation. The selection of a highly re-
producible HPLC column is essential to attaining this goal. The new Syncronis column
range has been engineered to deliver exceptional reproducibility by providing highly
pure, high surface-area silica, dense bonding, and double endcapping, all controlled
and characterized through rigorous testing. New enhanced, automated column pack-
ing methods used within the Syncronis range deliver greater consistency, and every
column is individually tested to ensure that it meets the required specifications. These
extensive testing and quality control procedures ensure the delivery of a completely
consistent product, column after column. Syncronis columns are available with 5 µm

Downloaded from www.sciencemag.org on December 23, 2010

particle size for conventional HPLC applications and with 1.7 µm particle size for high-
speed, high-efficiency UHPLC separations.

Thermo Fischer Scientific

For info: 800-532-4752 www.thermoscientific.com/Syncronis

Pressure/Flow Rate Sensors
The new Mitos Sensor Units provide a flexible system for measur-
ing and displaying pressure and flow rates in microfluidic systems.
With a real-time display and low internal volumes, each sensor is
designed to minimize interference with the liquid flow. All data is
clearly displayed on the Mitos Sensor Display unit, which interfaces
with each of the sensor units using a simple push-and-click action.
Comprising five flow rate sensors and one pressure sensor (-0.5–30
bar), the complete range is fully interchangeable and each device
can run stand-alone or integrated with a Mitos P-Pump for the ef-
ficient logging of flow rate output. All data is easily transferrable to
a PC for further analysis via USB. Ongoing enhancements to the
system will soon enable the flow rate sensors to communicate with
the Mitos P-Pump to create a closed loop pumping system. This will
enable the user to automatically control flow rates using pressure,
providing the ultimate in smooth liquid flow.
Dolomite
For info: +44-1763-242491 www.dolomite-microfluidics.com
controlled rate freezer
The Bio-Cool is the only controlled rate freezer that does not require
expendable liquid nitrogen and the associated pumping, refilling,
and storage challenges posed by a cryogenic liquid. The Bio-Cool,
simply plugs into a standard electrical outlet and quietly provides
low temperature cooling to either -40ºC or -80ºC. Unlike liquid
nitrogen systems, which surround samples with a cold vapor phase,
the Bio-Cool immerses samples into a well-circulated, cold liquid
bath. The circulating liquid allows more efficient heat transfer to
the samples and maintains a more consistent temperature profile
at all locations. The Bio-Cool comes standard with an easy-to-use
programmable microprocessor controller that enables programming
and storage of up to eight different protocols, each with up to
eight distinct segments specifying ramp rate, hold temperature,
and hold time. Elimination of liquid nitrogen handling and usage,
better freezing results, simple straightforward operation, and
world-class service support all make the Bio-Cool the controlled
rate freezer of choice for most clinical, commercial, and research
applications.
SP Scientific
For info: 845-255-5000 www.spscientific.com
IMAGING CAMERA
Engineered for high-sensitivity and high-speed performance in
low-light  imaging applications, the Rolera EM-C2  camera aids
researchers in applications such as spinning-disc confocal imaging,
total internal reflection fluorescence (TIRF) microscopy, ratiometric
ion imaging, and  fluorescence  recovery after photobleaching
(FRAP). Fast frame rates essential to cutting-edge live cell imaging
studies are enabled by the camera’s 40 MHz pixel clock rate. Up to
34.2 full-resolution (one megapixel) frames per second can be read
out and transferred over an optimized 800 Mb/s implementation of
the IEEE 1394b FireWire protocol. The Rolera EM-C2 delivers very
low noise and high-sensitivity across a broad spectrum  common
to fluorescence experiments. This camera  introduces an  Easy-
EM mode, which optimizes camera EM Gain settings with a
single click. Researchers no longer have to gauge sensitivity using
the arbitrary  sliders that are required  with the use of existing
cameras to estimate necessary EM Gain. The camera includes the
QCapture Suite software, which can be installed and  operational
within minutes.
QImaging
For info: 800-874-9789 www.qimaging.com

Electronically submit your new product description or product literature information! Go to www.sciencemag.org/products/newproducts.dtl for more information.
Newly offered instrumentation, apparatus, and laboratory materials of interest to researchers in all disciplines in academic, industrial, and governmental organizations are
featured in this space. Emphasis is given to purpose, chief characteristics, and availability of products and materials. Endorsement by Science or AAAS of any products or
materials mentioned is not implied. Additional information may be obtained from the manufacturer or supplier.

www.sciencemag.org/products SCIENCE VOL 330 24 DECEMBER 2010 1839

GE Healthcare
Life Sciences

Thank you ECL Plus, ™

you’ve been great.

For every success there’s a successor. It’s called
progress. So while we knew we had something
special in Amersham™ ECL Plus, our Western
blotting team was quietly working on the next
generation of detection reagent. The result: a new
substrate that operates with superior levels of
sensitivity, signal intensity and stability than even
its famous predecessor, making it an excellent
choice for CCD imagers. Welcome to ECL Prime.

Signal intensity sustained at higher antibody dilutions

Ab dilution Amersham ECL Prime Amersham ECL Plus

Primary Secondary 10 µg 156 ng 10 µg 156 ng

1:3000 1:300 000

1:10 000 1:500 000

Comparison of Amersham ECL Prime and Amersham ECL Plus

detection of ß-catenin in NIH 3T3 whole cell lysates using different
dilutions of rabbit anti-ß-catenin and HRP-conjugated anti-rabbit IgG.

Find out more about Amersham ECL Prime at

www.gelifesciences.com/eclprime

Amersham and ECL are trademarks of GE Healthcare companies.

© 2010 General Electric Company – All rights reserved. First published November 2010.
GE Healthcare Bio-Sciences AB, Björkgatan 30, 751 84 Uppsala, Sweden.
 New Products, Services, and Solutions
MPC-200
Multi-manipulator system
Versatile: User friendly interface controls up to two manipulators
with one controller. Select components to tailor a
system to fit your needs.

Learn how current events Expandable: Daisy chain a second controller and operate up to
are impacting your work. four manipulators with one input device.

Stable: Stepper motors and cross-rolled bearings quarantee

ScienceInsider, the new policy blog from the journal
Science, is your source for breaking news and instant reliable, drift-free stability.
analysis from the nexus of politics and science.
Doubly Quiet: Linear stepper-motor drive reduces electrical noise.
Produced by an international team of science journalists, Thermostatically-controlled cooling
ScienceInsider offers hard-hitting coverage on a range of
issues including climate change, bioterrorism, research fans barely whisper.
funding, and more.
Before research happens at the bench, science policy is
formulated in the halls of government. Make sure you
understand how current events are impacting your work.
Read ScienceInsider today.
Make the
www.ScienceInsider.org
right move!

SCIENCELAB SOLUTIONS
Breaking news and analysis from
the world of science policy

PHONE: 415.883.0128 | FAX: 415.883.0572

EMAIL: INFO@SUTTER.COM | WWW.SUTTER.COM

GrantsNet.
1,000s The first comprehensive
science grants database.
OF GRANTS GrantsNet is expanding its listings of
some 900 funding programs from private

MILLIONS
foundations and not-for-profit organizations
to include 400 to 500 new entries from the
grants.gov site. This provides the first

IN FUNDING
comprehensive database of funding
opportunities to research scientists and
administrators, career counselors, financial
aid specialists, and undergraduate and
graduate students. For listings, go to
www.grantsnet.org
 INTRODUCING

AAAS MemberCentral

The exclusive new website for the AAAS member community.

AAAS MemberCentral is a new website focused on helping you — the scientists, engineers,
educators, students, policymakers, and concerned citizens who make up the AAAS
community — connect like never before.

On MemberCentral you can contribute to discussion groups or blogs, participate in a

webinar, or share photos of your field research. You can exchange ideas, learn about
your fellow members, and gain fresh insights into issues that matter to you the most.
MemberCentral is also an easy access point for a wide variety of other AAAS membership
benefits, like discounts on cars and books, travel opportunities, and more.

Experience MemberCentral for yourself. Visit MemberCentral.aaas.org today and log in

using your Science online username and password. MemberCentral.aaas.org
 LIFE SCIENCE TECHNOLOGIES
AAAS/Science Business Office

NEW PRODUCTS

HPLC COLUMNS
The new Syncronis range of high-performance liquid chromatography (HPLC) columns
unerringly delivers consistent, predictable separations, from run to run and column to
column. When developing a new method, one of the key goals for the chromatogra-
pher is to achieve a consistent, reproducible separation. The selection of a highly re-
producible HPLC column is essential to attaining this goal. The new Syncronis column
range has been engineered to deliver exceptional reproducibility by providing highly
pure, high surface-area silica, dense bonding, and double endcapping, all controlled
and characterized through rigorous testing. New enhanced, automated column pack-
ing methods used within the Syncronis range deliver greater consistency, and every
column is individually tested to ensure that it meets the required specifications. These
extensive testing and quality control procedures ensure the delivery of a completely
consistent product, column after column. Syncronis columns are available with 5 µm
particle size for conventional HPLC applications and with 1.7 µm particle size for high-
speed, high-efficiency UHPLC separations.

Thermo Fischer Scientific

For info: 800-532-4752 www.thermoscientific.com/Syncronis

PRESSURE/FLOW RATE SENSORS
The new Mitos Sensor Units provide a flexible system for measur-
ing and displaying pressure and flow rates in microfluidic systems.
With a real-time display and low internal volumes, each sensor is
designed to minimize interference with the liquid flow. All data is
clearly displayed on the Mitos Sensor Display unit, which interfaces
with each of the sensor units using a simple push-and-click action.
Comprising five flow rate sensors and one pressure sensor (-0.5–30
bar), the complete range is fully interchangeable and each device
can run stand-alone or integrated with a Mitos P-Pump for the ef-
ficient logging of flow rate output. All data is easily transferrable to
a PC for further analysis via USB. Ongoing enhancements to the
system will soon enable the flow rate sensors to communicate with
the Mitos P-Pump to create a closed loop pumping system. This will
enable the user to automatically control flow rates using pressure,
providing the ultimate in smooth liquid flow.
Dolomite
For info: +44-1763-242491 www.dolomite-microfluidics.com
CONTROLLED RATE FREEZER
The Bio-Cool is the only controlled rate freezer that does not require
expendable liquid nitrogen and the associated pumping, refilling,
and storage challenges posed by a cryogenic liquid. The Bio-Cool,
simply plugs into a standard electrical outlet and quietly provides
low temperature cooling to either -40ºC or -80ºC. Unlike liquid
nitrogen systems, which surround samples with a cold vapor phase,
the Bio-Cool immerses samples into a well-circulated, cold liquid
bath. The circulating liquid allows more efficient heat transfer to
the samples and maintains a more consistent temperature profile
at all locations. The Bio-Cool comes standard with an easy-to-use
programmable microprocessor controller that enables programming
and storage of up to eight different protocols, each with up to
eight distinct segments specifying ramp rate, hold temperature,
and hold time. Elimination of liquid nitrogen handling and usage,
better freezing results, simple straightforward operation, and
world-class service support all make the Bio-Cool the controlled
rate freezer of choice for most clinical, commercial, and research
applications.
SP Scientific
For info: 845-255-5000 www.spscientific.com
IMAGING CAMERA
Engineered for high-sensitivity and high-speed performance in
low-light imaging applications, the Rolera EM-C2 camera aids
researchers in applications such as spinning-disc confocal imaging,
total internal reflection fluorescence (TIRF) microscopy, ratiometric
ion imaging, and fluorescence recovery after photobleaching
(FRAP). Fast frame rates essential to cutting-edge live cell imaging
studies are enabled by the camera’s 40 MHz pixel clock rate. Up to
34.2 full-resolution (one megapixel) frames per second can be read
out and transferred over an optimized 800 Mb/s implementation of
the IEEE 1394b FireWire protocol. The Rolera EM-C2 delivers very
low noise and high-sensitivity across a broad spectrum common
to fluorescence experiments. This camera introduces an Easy-
EM mode, which optimizes camera EM Gain settings with a
single click. Researchers no longer have to gauge sensitivity using
the arbitrary sliders that are required with the use of existing
cameras to estimate necessary EM Gain. The camera includes the
QCapture Suite software, which can be installed and operational
within minutes.
QImaging
For info: 800-874-9789 www.qimaging.com

Electronically submit your new product description or product literature information! Go to www.sciencemag.org/products/newproducts.dtl for more information.
Newly offered instrumentation, apparatus, and laboratory materials of interest to researchers in all disciplines in academic, industrial, and governmental organizations are
featured in this space. Emphasis is given to purpose, chief characteristics, and availability of products and materials. Endorsement by Science or AAAS of any products or
materials mentioned is not implied. Additional information may be obtained from the manufacturer or supplier.

www.sciencemag.org/products SCIENCE VOL 330 24 DECEMBER 2010 1839

 2011 AAAS ANNUAL MEETING

Science Without Borders

17–21 February, Washington, D.C.

AAAS, publisher of Science,

thanks the sponsors and supporters of
the 2011 Annual Meeting

Presenting sponsor

To become an exhibitor or
sponsor contact:
Jill C. Perla
Manager of Marketing, Exhibits,
and Sponsors
E-mail: jperla@aaas.org
Phone: 202-326-6736

Web site:
http://www.aaasmeeting.org
 “How do we know
this lead molecule
is novel?”

“ SciFinder—
of course.
”
Need to assess the novelty of substances?
SciFinder is the answer.

It includes CAS REGISTRY,SM the most comprehensive

substance information available, integrated with relevant
journal articles and patents.

Give your research team the highest quality and most timely
scientific information resource.

Make SciFinder an essential part of your research process.

For more information about SciFinder, visit www.cas.org or

e-mail help@cas.org.

SciFinder®—Part of the process.TM

www.cas.org
CAS is a division of the American Chemical Society
online @sciencecareers.org
Multiply The Power of Science

Download
your free copy
today.
ScienceCareers.org/booklets
Science Careers
Classified Advertising
For full advertising details, go to
ScienceCareers.org and click For Employers,
or call one of our representatives.
reers Away
Tracy Holmes
Worldwide Associate Director CAREER Cfraom the Bench
TRENDS
Science Careers tions for Scientists
Phone: +44 (0) 1223 326525 Advice and Op

UNITED STATES & CANADA

E-mail: advertise@sciencecareers.org
Fax: 202-289-6742
Tina Burks
Midwest/West Coast/
South Central/Canada
Phone: 202-326-6577
Elizabeth Early
East Coast & Industry
Phone: 202-326-6578
Marci Gallun
Sales Administrator
Phone: 202-326-6582
Online Job Posting Questions
Phone: 202-326-6577

EUROPE & REST OF WORLD

E-mail: ads@science-int.co.uk
Fax: +44 (0) 1223 326532
Alex Palmer
Phone: +44 (0) 1223 326527
Susanne Kharraz Tavakol
Phone: +44 (0) 1223 326529
you by
Dan Pennington is brought to
This booklet ce Business Office
ien
Phone: +44 (0) 1223 326517 the AAAS/Sc

Lisa Patterson
Phone: +44 (0) 1223 326528

JAPAN
ASCA Corporation From technology specialists to patent
Jie Chin attorneys to policy advisers, learn more
Phone: +81-3-6802-4616
Fax: +81-3-6802-4615 about the types of careers that scientists
E-mail: careerads@sciencemag.jp can pursue and the skills needed in order
To subscribe to Science: to succeed in nonresearch careers.
In United States call 866-434-2227
In the rest of the world call +1 202-326-6417

All ads submitted for publication must comply

with applicable U.S. and non-U.S. laws. Science
reserves the right to refuse any advertisement
at its sole discretion for any reason, including
without limitation for offensive language or
inappropriate content, and all advertising is
subject to publisher approval. Science encour-
ages our readers to alert us to any ads that
they feel may be discriminatory or offensive.
 online @sciencecareers.org
POSTDOCTORAL POSITIONS IN NEUROSCIENCE
Pfizer Neuroscience and Stanford University seek one or more postdoctoral
fellows with an interest in using two-photon fluorescence microscopy in awake
behaving mice to study animal models of brain disease. Postdoctoral fellows
will be co-mentored by Dr. Michael Ehlers (Pfizer, Chief Scientific Officer,
Where will your Neuroscience) and Prof. Mark Schnitzer (Stanford University, HHMI).
research lead? Candidates should have interests in the neurobiology of disease,
neuropharmacology, and optical imaging, and should have earned a Ph.D.,
preferably in relevant areas of neuroscience or engineering such as synaptic
biology, learning and memory, behavioral pharmacology, optical imaging, or
neuronal plasticity.

Additional information can be found on the Schnitzer Lab’s website

(http://pyramidal.stanford.edu) as well as the Pfizer Neuroscience website
(www.pfizerneuroscience.com).

Applicants should send a CV and contact information for three references to

kate.m.yannacci@pfizer.com and mschnitz@stanford.edu.

We are proud to be an equal opportunity employer and welcome applications from people with different
experiences, backgrounds and ethnic origins.

Working together for a healthier world™

Department of Health and Human Services

Food and Drug Administration
Center for Biologics Evaluation and Research
Immediate Office of the Center Director
Deputy Center Director
The FDA’s Center for Biologics Evaluation and Research (CBER), Immediate
Office of the Center Director, is searching for a Deputy Center Director. CBER
has the responsibility for protecting the public health of the U.S. population
by planning, developing and administering policies and programs related to
the safety, effectiveness and labeling of biological products for human use,
including blood and blood products, vaccines, and cellular, tissue, and gene
therapies. CBER is responsible for an annual budget of approximately $260
million and supports more than 1,000 employees.
As the Deputy, the incumbent will serve as the principal advisor and spokesper-
son for the Director in matters related to planning, developing and administering
the Center’s broad national and international programs and activities.
Qualifications: Eligible individuals must be U.S. citizens with an M.D. and/
or Ph.D. is desired with relevant training and extensive experience. Preferred
candidates possess specialized knowledge and experience in the development
of biological products, including the evaluation of safety, effectiveness,
and quality; knowledge of the FDA’s regulatory and review process; strong
leadership and management experience; excellent interpersonal skills to
deal effectively with multi-disciplinary teams and diverse stakeholders; and
outstanding oral and written communication skills. This position may also be
filled by appointment in the U.S. Public Health Service, Commissioned Corps.
Physician candidates for Civil Service or U.S. Commissioned Corps must
possess a valid license to practice medicine in any state in the U.S. Candidates
may also be eligible for an excepted service Title 42 appointment.
Salary: Salary is commensurate with education and experience. An excellent
benefits package is also available.
Location: Rockville, Maryland
How to Apply: Submit resume or curriculum vitae with cover letter by
February 18, 2011 to: CBER.Employment@fda.hhs.gov. Please reference
Job Code: OD/Dep.
The Department of Health and Human Services is an
Equal Opportunity Employer with a smoke free environment.
online @sciencecareers.org
GRANTS
Faculty Position in Environmental Pathology
UNIVERSITY OF VERMONT COLLEGE OF MEDICINE

The Department of Pathology at the University of Vermont College of Medicine seeks an MD or

MD/PhD pathologist with fellowship training in a subspecialty area of Surgical Pathology for a
tenure track position at the Assistant or Associate Professor level. The applicant must be Board
certified (or eligible) in anatomic pathology with subspecialty fellowship training (or have equivalent
qualification/experience). The successful applicant will devote 25% of their effort to clinical practice
and 75% time to research. We are most interested in individuals with research programs relevant to
environmental pathology and carcinogenesis, particularly in the lung. The Department of Pathology
has strong research programs in cardiovascular biology, redox signaling, fiber carcinogenesis,
pulmonary fibrosis and other aspects of environmental pathology. The Department has had an
Environmental Pathology Training Grant from the NIEHS for 25 years. The College of Medicine
and University support modern core facilities for microscopic imaging, genomics/proteomics, mouse
transgenics and other contemporary technologies.

The applicant is expected to direct an independently funded translational research program,

participate in teaching, and contribute to Departmental leadership. Salary, laboratory space and
startup support will be commensurate with qualifications, professional accomplishments and research
activities. Clinical responsibilities will include interpretation of surgical pathology specimens and
training of housestaff and medical students, activities that are shared with the other full-time
surgical pathologists. The surgical pathology volume is approximately 37,000 specimens per year,
with large active subspecialty practices. The University is especially interested in candidates who
can contribute to the diversity and excellence of the academic community through their research,
teaching, and/or service. Applicants are requested to include in their cover letter information about
how they will further this goal.

Applications will be accepted until the position is filled, but we strongly encourage the submission
of application materials by March 1, 2011. Applicants should submit a letter of interest and detailed
curriculum vitae with three references to Ms. Jennifer Diaz, Faculty Search Coordinator,
Department of Pathology, University of Vermont College of Medicine, the Courtyard at Given
S267 BeaumontAve, Burlington, Vermont 05405.Applications may also be submitted electronically
to Jennifer Diaz at Jennifer.Diaz@uvm.edu or you may apply online at www.uvmjobs.com.

The University of Vermont is an Affirmative Action/Equal Opportunity Employer; applications

from women and people from diverse racial, ethnic and cultural backgrounds are encouraged.

Associate Director
of Basic Sciences

The Medical College of Wisconsin is actively recruiting for an Associ-

ate Director for Basic Sciences in its newly developing Cancer Center.
The Associate Director should have a PhD and/or MD degree and a
distinguished record of achievement in a basic science discipline of
cancer research. This position will be responsible for assisting the basic
science program leaders in program development, in identifying areas
of potential research collaboration, and in facilitating translational
research from the basic science laboratories to the clinic. The Associate
Director will also assist in the development of and oversee all basic
science core facilities in the Center. The successful candidate will be a
member of the Executive Committee of the Cancer Center and report
directly to the Center Director. The academic appointment will be in a
mutually agreed upon department at the Medical College of Wisconsin
where teaching, research and service requirements of a departmental
appointment are expected.
Rank is open and commensurate with experience. This leader could also
potentially qualify for an endowed chair. The successful candidate will
bring a significant extramural funding portfolio and focus on research
funding. In addition to an active research program, this position will
include limited administrative activities within the Cancer Center and
service to the institution. Salary is competitive and will depend on rank
and experience. Qualified individuals are encouraged to send, via e-mail
(andreabrown@mcw.edu), a letter of interest, CV, and contact informa-
tion for three references to:
ATTN: Andrea Brown
AD of Basic Science Search
Cancer Center
Medical College of Wisconsin
8701 Watertown Plank Road
Milwaukee, WI 53226
ENDOCRINOLOGY FACULTY POSITIONS
The Endocrinology, Diabetes and Nutrition Division is currently seeking six
new faculty members to join a vibrant group of 27 current faculty (see http:
//medschool.umaryland.edu/medicine/default.asp). We seek two physi-
cian-scientists at the Assistant/Associate and Professor or Associate Pro-
fessor level (Position #’s 03-309-591-592). The successful candidates will
be accomplished independently funded investigators in clinical/translational
or basic science research who will also contribute to the clinical, teaching
and service missions of the University. The additional four positions are for
clinician educators/investigators specializing in diabetes management and
thyroid/pitituary disorders at the Assistant Professor to Associate Profes-
sor level (Position #’s 03-309-525,526,579,593). Clinical responsibilities
will include management of diabetes in a comprehensive multidisciplinary
setting at the University of Maryland Medical Center, and general endo-
crinology in both outpatient and inpatient settings. Successful applicants
will also have opportunities to engage in clinical/translational research and
contribute to the teaching duties of the Division. All candidates must be
board eligible/certified in Endocrinology. Tenure status/salary dependent
upon candidate background.
Send CV and list of 4 references to Kristi Silver, M.D., Associate Pro-
fessor, Division of Endocrinology, Diabetes and Nutrition, c/o JoAnn
Gibbs, Academic Programs Office, Department of Medicine, N3E09,
University of Maryland Medical Center, 22 S. Greene St., Baltimore,
MD 21201-1595. Please reference appropriate position # when submitting
your application.
The UM,B encourages women and minorities to apply and is an
AA/EEO/ADA Employer.
 MEETINGS

The 2011 Department of Energy
Joint Genome Institute (DOE JGI)
sixth annual
GENOMICS OF ENERGY
online @sciencecareers.org
topics include: confirmed speakers include:
Synthetic Biology Peer Bork, European Molecular Christopher Scholin, Monterey
Ecogenomics and
Biology Laboratory (EMBL) Bay Aquarium Research
Institute (MBARI)
Ed Buckler, Cornell University
Ecoresilience of the
Gulf Oil Spill Dan Distel, Ocean Genome
Legacy
Stephan Schuster, Penn State
University

& ENVIRONMENT Hardware and Software

Trends in Genomics
meeting Supercomputing
Computational Approaches
to Massive Short Read
Metagenomic Data Sets
Genomics of Biofuel Crops
Behavioral Genetics of
Pollinating Bees
Microbiome Analyses from
Humans to Shipworms
Metatranscriptomics of
Marine Microbial
Communities
Successful Transposable
Dusko Ehrlich, French National
Institute for Agricultural
Research (INRA)
Terry Hazen, Lawrence
Berkeley National
Laboratory (LBNL)
Scott Hodges, University of
California, Santa Barbara
Tom Juenger, University of
Texas at Austin
Rob Knight, University of
Colorado
Ruth Ley, Cornell University
Mary Ann Moran,
University of Georgia
Magnus Nordborg, Gregor
Pam Silver, Harvard
Jim Tiedje, Michigan State
University
Mike Thomashow, Michigan
State University
Jerry Tuskan, Oak Ridge
National Laboratory/DOE JGI
Sue Wessler, University of
California, Riverside
Katherine Yelick, National
Energy Research Scientific
Computing Center (NERSC)
NL
L
at LBNL

March 22-24 Elements Secrets

Great Prairie Soil
Mendel Institute
Gene Robinson, University of
Illinois at Urbana-Champaign
Walnut Creek, Metagenomics

California workshops include: Integrated

Microbial Genomes; Mycocosm Fungal
Genomics Portal; Phytozome; RNA
http://go.usa.gov/1Gl Technologies & Analysis

POSITIONS OPEN
DNA Sequencing and Computational Biology
Core Facility Director
Health and Human Services (HHS)
National Institutes of Health (NIH)
National Heart, Lung and Blood Institute (NHLBI)
An expert is sought in the area of Computational Biology with an emphasis on next-generation DNA
sequence analysis at the DNA Sequencing and Computational Biology (DSCB) Core Facility, Division
of Intramural Research (DIR), National Heart, Lung and Blood Institute (NHLBI), NIH in Bethesda,
Maryland USA. The successful applicant will participate in analyzing large-scale data sets consisting of
a wide spectrum of sequencing applications, including but not limited to chromatin immunoprecipitation
sequencing (ChIP-Seq), RNA-seq, targeted and whole-genome DNA sequencing, microRNA sequencing.
He/she is expected to work closely with the core director and interact with DIR Principal
Investigators in establishing a comprehensive sequence analysis pipeline, in basic bioinformatics
studies for the presentation of sequence output to end users, and in original collaborative research in
genomics and systems biology.
The DSCB Core Facility is part of an NHLBI DIR Initiative in Systems Biology, and the candidate is
also expected to interact closely with scientists within the DSCB Core and/or other independently
operated DIR Facilities (e.g. proteomics) to facilitate data integration. Although the DSCB Core Facility
is oriented toward providing service and conducting collaborative research, the position will also
have the opportunity to undertake research initiative in the area of computational biology. The mission
of the DIR is to improve the health of all Americans through basic and clinical research, research training,
and translation of discoveries to new tools to be applied directly to the field of medicine.
at
free copy today
Download your ookle ts We are seeking an experienced scientist (with Ph.D. or equivalent) with an outstanding track record
.org/b
ScienceCareers
in computational biology research. Salary will be commensurate with qualifications and experience.
More detailed information about the NHLBI Division of Intramural Research may be found at:
http://dir.nhlbi.nih.gov/ .
Applicants should submit the following: cover letter highlighting key qualifications; current curriculum
vitae with complete bibliography; names and addresses of four references; and a one-page description of
current and future research interests.
Applications should be received by Feb. 1, 2011 but the advertisement will remain open until the
position is filled. PDF versions of documents sent by electronic mail are strongly preferred.
Materials should be sent to Dr. Jun Zhu c/o: Trina Gregory, Administrative Officer, NHLBI, by
Brought to you by the
AAAS/Science Business Office email: gregoryp@nhlbi.nih.gov; or by regular mail: Building 10, Room 7N220, 10 Center Drive
MSC 1670, Bethesda, MD 20892-1670.
HHS and NIH are Equal Opportunity Employers
online @sciencecareers.org

Dive ience
r s it y :
Women in S c a ry 2 1 , 2
anu ntee space.*01 1
engineer

F r e : J
Special Carere ad by January 4 to guara
r e at u
.
available
Reserve you *Ads accepted u
ntil Janu
ary 14 if spac
e is still

technologis
i t

Why you should advertise in this issue of Science:

Context: Job seekers are drawn to information that speaks to their concerns. When
candidates read the article, they’ll see your ad.
Reach: Context without an audience is meaningless. With 700,000 weekly readers
in print, and thousands more online, Science offers the largest readership
of any general scientific journal.
Results: When you combine context and reach you get results. There’s a reason why
hundreds of PIs and HR groups continue to advertise their open positions
via print ads in Science. It works.

Place your ad today and get a head start on making your next hire.

To Bo o k Y o ur Ad , C o n t act :
ada,
States, Can
In the United ica
mer careers.org
and Latin A e@science
l: ad vert is
E-mai
d RoW
In Europe an ience-int.co.uk
l: ad s@ sc
E-mai
In Asia ncemag,jp
erads@scie
E-mail: care

equa ScienceCareers.org
lity
 PRIZES

online @sciencecareers.org
THE 2011 LOUISA GROSS HORWITZ PRIZE
FOR BIOLOGY OR BIOCHEMISTRY
The Louisa Gross Horwitz Prize was established under the will of the late S. Gross Horwitz through a bequest to Columbia University and is named to honor the
donor’s mother. Louisa Gross Horwitz was the daughter of Dr. Samuel David Gross (1805-1889), a prominent surgeon of Philadelphia and author of the out-
standing Systems of Surgery who served as President of the American Medical Association.
Each year since its inception in 1967, the Louisa Gross Horwitz Prize has been awarded by Columbia University for outstanding basic research in the fields of
biology or biochemistry. The purpose of this award is to honor a scientific investigator or group of investigators whose contributions to knowledge in either of
these fields are deemed worthy of special recognition.
The Prize consists of an honorarium and a citation which are awarded at a special presentation event. Unless otherwise recommended by the Prize Committee,
the Prize is awarded annually. Dr. Elizabeth H. Blackburn, University of California, San Francisco, CA; Dr. Joseph G. Gall, Carnegie Institute, Baltimore, MD; Dr.
Carol W. Greider, Johns Hopkins University, Baltimore, MD were the 2007 awardees.

QUALIFICATIONS FOR THE AWARD

The Prize Committee recognizes no geographical limitations. The prize may be awarded to an individual or a group. When the prize is awarded to a group,
the honorarium will be divided among the recipients, and each member will receive a citation. Preference will be given to work done in the recent past.
Nominations must be submitted electronically at: http://www.cumc.columbia.edu/horwitz/

Nominations should include:

1. A summary, preferably less than 500 words, of the research on which this nomination is based.
2. A summary, preferably less than 500 words, of the significance of this research in the fields of biology or biochemistry.
3. A brief biographical sketch of the nominee, including positions held and awards received by the nominee.
4. A listing of up to ten of the nominee’s most significant publications relating to the research noted under item 1.
5. A copy of the nominee’s curriculum vitae.

Deadline date: January 31, 2011

POSITIONS OPEN

Dean, College of Arts and Sciences

Georgia State University, a leading research university located
in the heart of downtown Atlanta, is conducting a search for the
Dean of the College of Arts and Sciences. The Search Committee
invites nominations, applications (letter of interest, complete CV,
and references), or expressions of interest to be submitted to the
search firm assisting Georgia State University. Confidential review
of materials will begin immediately. It is preferred that all nomina-
tions and applications be submitted prior to February 28, 2011. For
a complete position description, refer to Current Opportunities on
www.parkersearch.com. For additional information, please visit
GSU online at www.gsu.edu or the College of Arts and Sciences
at www.cas.gsu.edu.

Laurie C. Wilder, Senior Vice President

Katie Bain, Principal
770-804-1996 ext: 108
kbain@parkersearch.com

Georgia State University, a unit of the University System of

Georgia, is an Equal Opportunity Educational Institution and is
an Equal Opportunity/Affirmative Action Employer.
AAAS is here – connecting government to the scientific community.
As a part of its efforts to introduce fully open government, the White House is reaching out to the scientific community for a
conversation around America’s national scientific and technological priorities.
To enable the White House’s dialogue with scientists, AAAS launched Expert Labs, under the direction of blogger and tech guru
Anil Dash. Expert Labs is building online tools that allow government agencies to ask questions of the scientific community and
then sort and rank the answers they receive.
On April 12, 2010, AAAS asked scientists everywhere to submit their ideas to the Obama administration and at the same time
launched the first of Expert Labs tools, Think Tank, to help policy makers collect the subsequent responses. The result was
thousands of responses to the White House’s request, many of which are already under consideration by the Office of Science
and Technology Policy.
As a AAAS member, your dues support our efforts to help government base policy on direct feedback from the scientific
community. If you are not already a member, join us. Together we can make a difference.

To learn more, visit aaas.org/plusyou/expertlabs

AAAS is here – helping scientists achieve career success.
Every month, over 400,000 students and scientists visit ScienceCareers.org in search of the information, advice, and opportuni-
ties they need to take the next step in their careers.
A complete career resource, free to the public, Science Careers offers a suite of tools and services developed specifically for
scientists. With hundreds of career development articles, a grants and scholarships database, webinars and downloadable
booklets filled with practical advice, a community forum providing real-time answers to career questions, and thousands of
job listings in academia, government, and industry, Science Careers has helped countless individuals prepare themselves for
successful careers.
As a AAAS member, your dues help AAAS make this service freely available to the scientific community. If you’re not a member,
join us. Together we can make a difference.

To learn more, visit aaas.org/plusyou/sciencecareers

AAAS is here – promoting universal science literacy.
In 1985, AAAS founded Project 2061 with the goal of helping all Americans become literate in science, mathematics, and
technology. With its landmark publications Science for All Americans and Benchmarks for Science Literacy, Project 2061 set out
recommendations for what all students should know and be able to do in science, mathematics, and technology by the time they
graduate from high school. Today, many of the state standards in the United States have drawn their content from Project 2061.
Every day Project 2061 staff use their expertise as teachers, researchers, and scientists to evaluate textbooks and assessments,
create conceptual strand maps for educators, produce groundbreaking research and innovative books, CD-ROMs, and profes-
sional development workshops for educators, all in the service of achieving our goal of universal science literacy.
As a AAAS member, your dues help support Project 2061 as it works to improve science education. If you are not yet a AAAS
member, join us. Together we can make a difference.

To learn more, visit aaas.org/plusyou/project2061

 The science and engineering challenges that society faces today are far more complex than those
of 40 to 50 years ago. The best available scientific, technical, and economic information is required
to establish priorities, make decisions, and develop best practices. AAAS manages the Science &
Science & Technology Technology Policy Fellowships in four areas to provide the opportunity for accomplished scientists
Policy Fellows and engineers to contribute to the federal policymaking process while learning firsthand about the
intersection of science and policy. And this is just one of the ways that AAAS is committed to advanc-
ing science to support a healthy and prosperous world. Join us.
Together we can make a difference. aaas.org/plusyou/fellows
 POSITIONS OPEN
online @sciencecareers.org
TENURE-TRACK POSITION
ASSISTANT PROFESSOR
CHEMIST/BIOCHEMIST
Department of Food Science and Technology
University of California, Davis
The Department of Food Science and Technology is
currently seeking to fill a faculty position. We are in-
terested in individuals who have or can establish a strong
extramurally funded research program in an advanced
contemporary area of food chemistry/biochemistry.
The specific areas of research investigation we seek in
the broad area of food chemistry include, but are not
limited to: (i) food components and their interactions
(including polymers, colloids, and emulsions); (ii) car-
bohydrates, especially carbohydrate polymers; (iii) food
formulations; and, (iv) health-promoting bioactive food
components, including encapsulation and delivery of
nutrients or otherwise functional components. Candi-
dates are expected to have a Ph.D. (or equivalent) and
demonstrated ability in chemistry, biochemistry, food
science, or a related discipline. Postdoctoral experience
is highly desirable. Selection will be based in part on a
record of research publications in internationally recog-
nized peer-reviewed journals, and the ability to obtain
extramural funding. The specific research program will
depend upon the expertise and interests of the can-
didate. The successful applicant will be expected to
develop an independent, internationally recognized re-
search program and contribute to the mission of the
CAES and its associated Agricultural Experiment Sta-
tion, teach at the undergraduate and graduate level,
and supervise graduate student thesis research.
Applicants should submit online at website: https://
secure.caes.ucdavis.edu/Recruitment a letter of ap-
plication, curriculum vitae (including list of publica-
tions), a statement of research, a separate statement
describing teaching interests, and background; reprints
of three publications; academic transcripts; and names,
addresses including e-mail, and telephone numbers of
three references. The position is open until filled; but
to assure full consideration, completed online applica-
tions should be submitted no later than February 28,
2011, for a targeted start date of July 1, 2011.
UC Davis is an Affirmative Action/Equal Employment Op-
portunity Employer and is dedicated to recruiting a diverse faculty
community. We welcome all qualified applicants to apply, includ-
ing women, minorities, individuals with disabilities, and veterans.
ASSISTANT, ASSOCIATE, AND
FULL PROFESSOR OF PHYSIOLOGY
University of California, Merced
The University of California, Merced invites appli-
cants for a faculty position in Physiology. The appoint-
ment will be made at either the tenure-track Assistant
Professor or tenured Associate or Full Professor rank. We
seek an outstanding individual with research interests
in and expertise in any area of Physiology. We welcome
applicants using experimental approaches working at
the cellular and/or organism level. We seek distinguished
scholars who will help establish a program of inter-
national repute in Physiology research at UC Merced,
and who will participate actively in the development of
innovative, interdisciplinary curricula and in the teach-
ing and mentoring of a diverse student population. For
more information and to apply visit website: http://
jobs.ucmerced.edu/n/academic/listings.jsf?seriesId01.
The application deadline is February 1, 2011. Affirmative
Action/Equal Opportunity Employer.
EAST TENNESSEE STATE UNIVERSITY—
College of Medicine Department of Physiology—
POSTDOCTORAL positions available to study the
molecular mechanisms of cardiac myocyte apoptosis
and myocardial remodeling. Highly motivated individ-
uals (M.D. and/or Ph.D. required) with experience in
molecular signaling, cardiac physiology, biochemistry,
or cell biology will be preferred. Apply to the positions
at website: https://jobs.etsu.edu. Contingent upon
grant funding. Affirmative Action/Equal Opportunity Employer.
1850 24 DECEMBER 2010
POSITIONS OPEN
TENURE-TRACK FACULTY POSITION in
Health Sciences
Purdue University, School of Health Sciences
Your
The School of Health Sciences invites applications
for a tenure-track position at the rank of ASSISTANT
PROFESSOR. The successful candidate is expected to
develop and maintain extramurally funded research pro-
career
grams in neurotoxicology, metal/pesticide toxicology,
or related research areas. Applicants with expertise in
molecular, cellular, genetic, or neuroimaging approaches
in understanding the mechanisms of neurodegenera-
is our
tive diseases are encouraged to apply. Candidates must
have a Ph.D., M.D., or equivalent degree and at least
two years of relevant postdoctoral research experience.
The position is competitive with regard to salary, start-
cause.
up funds, and laboratory space. Please electronically
send curriculum vitae, a brief statement of current and
future research interests, and contact information for
three references to Dr. Wei Zheng, Head of the School
of Health Sciences, at e-mail: wzheng@purdue.edu.
Review of applicants will begin February 1, 2011,
and will continue until the position is filled. Applicants
are encouraged to apply by January 31, 2011 for full
Get help
consideration.
Purdue University is an Equal Opportunity/Equal Access/
Affirmative Action Employer fully committed to achieving a
from the
diverse workforce.
THE METHODIST HOSPITAL
experts.
RESEARCH INSTITUTE
Weill Cornell Medical College
The Molecular Imaging Program at the Department
of Radiology develops novel agents and new technol-
ogies to image molecular processes and treat diseases.
The research focuses on cancer, cardiovascular disease,
neurodegeneration, cell therapy, and nanomedicine. www.
Several POSTDOCTORAL FELLOW positions are
currently open for application. Self-motivated scientists sciencecareers.org
with expertise in liposomes, peptides, nanoparticles,
radiochemistry, photodynamic therapy, and MR phys-
ics are encouraged to join our dynamic research team.
Please electronically send curriculum vitae and contact
information of three references to Dr. Ching H. Tung
at e-mail: ctung@tmhs.org. The Methodist Hospital
Research Institute is centrally located in the world largest
medical center in Houston, Texas. • Job Postings
• Job Alerts
• Resume/CV
Do what Database
you love. • Career Advice
• Career Forum
Love what • Graduate
you do. Programs
• Meetings and
Announcements
www.sciencecareers.org
VOL 330 SCIENCE www.sciencecareers.org