Professional Documents
Culture Documents
Science - June 2018 PDF
Science - June 2018 PDF
CONTENTS
2 9 J U N E 2 0 1 8 • V O LU M E 3 6 0 • I S S U E 6 3 9 6
Ionic control of friction
NEWS
FEATURES
1388 THE POWER OF MANY
A series of simple steps can explain
the momentous transition from
IN BRIEF single cells to multicellular life
1380 News at a glance By E. Pennisi
▶ VIDEO
CREDITS: (CLOCKWISE FROM TOP) A. KITTERMAN/SCIENCE; WANG CHI LAU/EMBRYOLOGY COURSE AT THE MARINE BIOLOGICAL LABORATORY; ADAM SIMPSON/HEART AGENCY
achieve sustainable materials production
and use By E. A. Olivetti and J. M. Cullen PERSPECTIVES
1385 NEWBORN SCREENING URGED FOR
FATAL NEUROLOGICAL DISORDER 1399 MORE FRICTION FOR
BOOKS ET AL. Decision on spinal muscular atrophy POLYELECTROLYTE BRUSHES
1407 GETTING IT RIGHT ON GMOS due next week By M. Wadman Trivalent yttrium cations increase
A protester’s change of heart sheds friction greatly compared with
light on the public’s reservations about 1386 SEE-THROUGH SOLAR CELLS COULD monovalent cations By M. Ballauff
genetic engineering By J. R. Dinneny POWER OFFICES ▶ REPORT P. 1434
Solar windows absorb ultraviolet and
LETTERS infrared light while letting visible light
1400 LEARNING FROM PAST
pass through By R. F. Service
1409 EDUCATION FOR THE FUTURE CLIMATIC CHANGES
1387 BLOOD TEST MAY PREDICT CANCER 66 million years ago, sea temperatures
REVIEW IMMUNOTHERAPY BENEFIT rose rapidly as a result of environmental
Counting tumor mutations could perturbations By C. Lécuyer
1419 ENERGY
identify right treatment By K. Garber ▶ REPORT P. 1467
Net-zero emissions energy systems
S. J. Davis et al.
REVIEW SUMMARY; FOR FULL TEXT:
dx.doi.org/10.1126/science.aas9793
ON THE COVER
Growing human
populations are
transforming our
planet at an increas-
ing rate, leading to
climatic changes,
diminished resources,
and loss of biodiver-
sity. Continuing on
the current path is likely to endanger our
own well-being and that of other species,
but changing course involves tough choices.
In the Tomorrow’s Earth special series, we
explore paths to a more sustainable future.
See http://scim.ag/TomorrowsEarth.
Illustration: Adam Simpson/Heart Agency 1388
1376 29 JUNE 2018 • VOL 360 ISSUE 6396 sciencemag.org SCIENCE
Published by AAAS
1423 1447
Spliceosome Watching mice see
ready for action
HUMAN DEMOGRAPHY
LIMITS BIODIVERSITY RESEARCH
National laws fearing biopiracy squelch 1459 The plateau of human mortality:
taxonomy studies 1422 PALEOGENOMICS Demography of longevity pioneers
The first horse herders and the impact E. Barbi et al.
By K. Divakaran Prathapan et al.
of early Bronze Age steppe expansions 1462 Predictive modeling of U.S.
BOOKS ET AL. into Asia P. de Barros Damgaard et al. health care spending in late life
RESEARCH ARTICLE SUMMARY; FOR FULL TEXT:
1408 BEYOND EPIGENETICS L. Einav et al.
dx.doi.org/10.1126/science.aar7711
▶ PODCAST
A pair of evolutionary biologists takes a
closer look at nongenetic inheritance 1423 MOLECULAR BIOLOGY
By K. N. Laland 1465 PSYCHOLOGY
Structures of the fully assembled
Prevalence-induced concept change
Saccharomyces cerevisiae spliceosome
in human judgment D. E. Levari et al.
before activation R. Bai et al.
SCIENCE (ISSN 0036-8075) is published weekly on Friday, except 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 © 2018 by the American Association for the Advancement of Science. The title SCIENCE is a registered trademark of the AAAS. Domestic
individual membership, including subscription (12 months): $165 ($74 allocated to subscription). Domestic institutional subscription (51 issues): $1808; Foreign postage extra: Mexico, Caribbean (surface mail) $55; other countries
(air assist delivery): $89. First class, airmail, student, and emeritus rates on request. Canadian rates with GST available upon request, GST #125488122. 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: $15 each plus shipping
and handling; bulk rate on request. Authorization to reproduce 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 others who
use Copyright Clearance Center (CCC) Pay-Per-Use services provided that $35.00 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.
Published by AAAS
ED ITORIAL
Tomorrow’s Earth
O
ur planet is in a perilous state. The combined and propellants for aerosol cans could catalyze the de-
effects of climate change, pollution, and loss of struction of ozone. Less than a decade later, these con-
biodiversity are putting our health and well- cerns were exacerbated by the discovery of seasonal
being at risk. Given that human actions are largely ozone depletion over Antarctica. International discus-
responsible for these global problems, humanity sions on controlling the use of these chemicals culmi-
must now nudge Earth onto a trajectory toward a nated in the Montreal Protocol in 1987. Three decades
more stable, harmonious state. Many of the chal- later, research has shown that ozone depletion appears
lenges are daunting, but solutions can be found. In this to be decreasing in response to industrial and domes- Editor-in-Chief,
issue of Science, we launch a series of monthly articles tic reforms that the regulations facilitated.
Science Journals.
that call attention to some of the choices we can still More recent efforts include the Paris Agreement of
jberg@aaas.org
make for shaping tomorrow’s Earth—commentaries and 2015, which aims to keep a global temperature rise this
analyses that will hopefully century well below 2°C and
the stresses and possible solutions to these growing lation growth is escalating, we have never been so afflu-
human-induced impacts on the Earth system. Donald ent. Along with affluence comes increasing use of energy
Kennedy, then Science’s Editor-in-Chief, aptly asserted: and materials, which puts more pressure on the environ-
“The big question in the end is not whether science can ment. How can humanity maintain high living standards
help. Plainly it could. Rather, it is whether scientific evi- without jeopardizing the basis of our survival?
dence can successfully overcome social, economic, and As our “Tomorrow’s Earth” series (see scim.ag/
political resistance.” TomorrowsEarth) will highlight, rapid research and
Through collective action, we can indeed achieve technology developments across the sciences can help
planetary-scale mitigation of harm. A case in point is to facilitate the implementation of potentially corrective
TOMORROW’S
the Montreal Protocol on Substances that Deplete the options. There will always be varying expert opinions on EARTH
Ozone Layer, the first treaty to achieve universal rati- what to do and how to do it. But as long as there are op- Read more articles
fication by all countries in the world. In the 1970s, sci- tions, we can hope to find the right paths forward. online at scim.ag/
entists had shown that chemicals used as refrigerants –Jeremy Berg TomorrowsEarth
10.1126/science.aau5515
T
he Arctic’s 180,000 known archaeological sites record more federal government more efficient by
than 4000 years of human survival and culture at high lati- reorganizing agencies, including many
tudes, from wood houses to ivory sculptures. Most have not that fund research. But like previous
reshuffling attempts by his Democratic
been excavated, but climate change threatens to degrade or and Republican predecessors, the plan
destroy them before they are discovered and studied, authors is likely to face a tepid response from
write in a review this month in Antiquity. Frozen and wet con- Congress. Environmentalists fear that a
ditions at Arctic sites have provided “extraordinary long-term preser- proposal to merge the National Marine
Fisheries Service with the U.S. Fish and
vation of archeological material,” including animal bones, clothing, Wildlife Service within the Department
and ancient trash piles known as middens. Yet the ribbon of sea of the Interior could weaken enforcement
ice that until recently protected Arctic coastal soils has receded, al- of laws affecting endangered species and
lowing storms to erode sites near shores. As ground temperatures marine mammals. The National Science
Foundation would manage small graduate
rise, the permafrost is thawing, allowing oxygen to penetrate the fellowship programs at several agencies
ground and feed microbes that can damage artifacts. The authors, an without receiving more money to do so.
international team of scientists who have worked across the Arctic, Creating an energy innovation office
call for using remote sensing and developing protocols to priori- to handle applied research programs
at the Department of Energy (DOE)
tize sites, including those that require urgent excavation. Co-author
could improve coordination—but some
Vladimir Pitulko, at the Institute for the History of Material Culture groups see it as a way to dismantle DOE’s
in St. Petersburg, Russia, says he hopes the issue will get attention in
PHOTO: MAX FRIESEN
Published by AAAS
Knight to lead fusion lab
| Steven Cowley, the
P L A S M A P H YS I C S
astrophysicist who once led the United
Kingdom’s Culham Centre for Fusion
Energy (CCFE), has been busy. On 1 July, he
takes over as the director of the Princeton
Plasma Physics Laboratory (PPPL) in New
Jersey, the premier U.S. fusion research
lab, and earlier this month, the United
Kingdom’s Queen Elizabeth II knighted him
“for services to science and the development
of nuclear fusion.” Cowley, or Sir Steven,
most recently was president of Corpus
Christi College at the University of Oxford in
the United Kingdom, and he led CCFE from
2008 to 2016. At PPPL, where he worked as
a staff scientist in the early 1990s, his first
task will be to get the National Spherical
Torus Experiment back on track. The unusu-
ally shaped fusion reactor was shut down in
K
oko, the western lowland gorilla who captivated the world with deep emotional
displays and an ability to communicate with human handlers, died in her sleep
| College students are 56%
WO R K P L AC E
last week at age 46. Koko, born in 1971 at the San Francisco Zoo in California,
CREDITS: (PHOTO) BETTMANN/CONTRIBUTOR/GETTY IMAGES; (GRAPHIC) J. BRAINARD/SCIENCE; (DATA) S. ATIR AND M. J. FERGUSON, PNAS 10.1073/PNAS.1805284115 (2018)
more likely to refer to male professors is best known for learning a modified version of American Sign Language from
than female professors by their last names animal psychologist and The Gorilla Foundation Founder Francine “Penny”
alone—and that form of address may confer Patterson (above). Although Koko mastered some 1000 signs, she ignited a fierce
greater respect. Researchers uncovered the debate among scientists, who questioned whether she was actually using a “lan-
gender disparity in part by looking at 4494 guage,” or simply responding to Patterson’s prompts. Nevertheless, Koko’s abilities
comments by students in online reviews of helped transform how the human world viewed animal emotion—and intelligence.
their professors in five disciplines at 14 uni-
versities. The largest gender disparity was
in computer science, where only surnames
were mentioned in 48% of reviews for male mentioned by last name alone as more It also downplays efforts by federal and
professors and 18% of reviews for female famous and eminent than scientists men- state agencies to collaborate on long-term
professors, according to a study published tioned by full name. They conclude that regional management plans. The rewrite
this week in the Proceedings of the National women may be short-changed on profes- is “a significant step backward” and would
Academy of Sciences. In a set of parallel sional benefits such as research funding “aggressively expand” ocean exploita-
studies, the authors—Stav Atir and Melissa based on nothing more than how people tion, says biologist Jane Lubchenco of
Ferguson of Cornell University—report men utter their names. Oregon State University in Corvallis, who
were far more likely to garner last-name- helped develop the Obama plan and led
only recognition in other contexts as well, the National Oceanic and Atmospheric
such as when pundits discuss politicians on Trump axes Obama ocean policy Administration from 2009 to 2013.
talk shows. The authors’ experimental evi- | President Donald
M A R I N E M A N AG E M E N T
dence suggests people regard researchers Trump last week issued a new national
policy that replaces U.S. leads in supercomputing
How often professors are called only by surnames former President | The United States is back
T E C H N O L O GY
As collected from students commenting on the Rate My Professor website. Barack Obama’s 2010 atop the supercomputing heap for the
plan for managing U.S. first time since November 2012, though its
coastal waters and the reign may not last long. The Summit
Male Female
Great Lakes. Whereas supercomputer at Oak Ridge National
50% that plan emphasized Laboratory in Tennessee, with its
40 conservation and the 122.3 quadrillion (1015) floating point
30
need to address climate operations per second (petaflops), is
change, Trump’s now the world’s fastest, according to the
20 19 June executive ranking project TOP500. It bumped a
10 order is largely silent 93-petaflop machine located in China’s
0 on those issues, and National Supercomputing Center in Wuxi
Computer Psychology Biology* History Economics* instead emphasizes into second place, while another new
science economic development U.S. machine at Lawrence Livermore
*Di0erences lacked statistical signifcance. and national security. National Laboratory in California grabbed
THREE QS the third spot. Even so, the United States convincing explanation is that water vapor
has only 124 of the top 500 machines—its and gases are jetting from ‘Oumuamua,
No more lab cages? poorest showing ever, and down from affecting its course and putting it firmly
145 6 months ago. China has 206, and back in the “comet” camp.
Should lab animals be studied outdoors
within 2 years a Chinese machine is
instead? Garet Lahvis, a behavioral
expected to be the first to reach exascale—
neuroscientist at Oregon Health &
a billion billion calculations per second. U.K. cancels tidal power
Science University in Portland, thinks
E N E R GY| The United Kingdom this week
so. He spoke with Science about the
pulled the plug on plans for the world’s
advantages—and challenges—of A new peer-review experiment first tidal energy lagoon. The project was
conducting lab research outside the
SCIENTIFIC PUBLISHING | The life sciences to use the movement of the tides in the
cage. (See a longer version of this
journal eLife will experiment with guaran- Swansea Bay in Wales to generate up to
interview at https://scim.ag/QALahvis.)
teeing publication of all manuscripts that 320 megawatts of electricity—enough to
are sent out for peer review. Starting this power 155,000 homes. The project’s devel-
Q: What’s wrong with current lab
week, submitting authors can opt into the oper, the Gloucester, U.K.–based company
animal housing?
system, which aims to eliminate the gate- Tidal Lagoon Power, envisioned five more
A: Animals have to deal with a massive
keeping role of outside reviewers. Once lagoons providing 8% of U.K. electricity.
landscape in the wild. They need to hunt
eLife’s editors decide a manuscript’s “scien- But energy minister Greg Clark balked at
for food, avoid predators, seek mates, and
tific standards and potential significance” the pilot protect’s £1.3 billion price tag,
deal with complex environmental vari-
make it worthy of peer review, authors and announced on 25 June that he will
ables. If you live in a cage, everything is
can choose to make changes in response not subsidize it. Proponents argue that the
Published by AAAS
IN D EP TH
COMPUTER SCIENCE
By Sophia Chen way they can find,” says Rene Peralta, a orderly data such as real-time earthquake
computer scientist at NIST in Gaithersburg, measurements, online Twitter posts, radio
I
n Chile, politicians resent the Comptrol- Maryland, who leads the U.S. effort. “I think streams, and cryptocurrency transactions.
ler General, which audits government of- of it as digital infrastructure.” Readily available, trustworthy random
ficials to prevent corruption. The audits A sequence of truly random numbers ex- numbers could be used to deliver public
are supposed to be random—but scruti- hibits no patterns or predictability. Knowing services more fairly, computer scientists say.
nized officials sometimes complain about the sequence one day should not provide Besides trying to help the Comptroller Gen-
unfair targeting. “The auditors have to any hints to the sequence published a min- eral maintain its credibility, Hevia is push-
convince the public they’re doing their work ute or a day later. But that ideal is not easy ing to use Chile’s beacon to assign students
honestly,” says Alejandro Hevia, a computer to attain. Some online random number gen- to schools through a lottery. In the United
scientist at the University of Chile in San- erators rely on algorithms, which means States, the government could use public
tiago. Along with researchers around the their output is, in principle, predictable; random numbers to assign visas. “If you’re
world, he is developing technology that could others depend on random physical phe- an applicant and you were not chosen, you
persuade critics that audits are truly random: nomena. The NIST beacon, which gener- would like to know that it was because you
public random number generators. ates a string of 512 0s and 1s, or bits, every weren’t lucky enough, and not because
On 10 July, Hevia’s team will unveil an on- 60 seconds, combines the two approaches. you are Muslim,” Peralta says. Brazil wants
line random number service. Later in July, It starts with output from two commercial to use its service to assign court cases to
the U.S. National Institute of Standards and random number generators that rely on judges, says Raphael Machado, a computer
Technology (NIST) will launch its Random- electronic noise in circuits, then increases scientist at Brazil’s National Institute of
ness Beacon as a permanent service, up- the numbers’ unpredictability by combin- Metrology Standardization and Industrial
grading a pilot program that began in 2013. ing them in a mathematical operation Quality in Rio de Janeiro.
Brazil, too, is planning a beacon, by the end that reduces underlying bias. Chile’s bea- The numbers could also boost security
PHOTO: ISTOCK.COM/TOSTPHOTO ADAPTED BY C. AYCOCK/SCIENCE
of 2019. All aim to improve on commercial con combines circuit noise with other dis- by serving as timestamps to authenticate
random number generators, not only digital documents. The idea resembles
by being free, but by generating the the classic kidnapper’s protocol: To
random numbers through transparent Taking one’s chances prove a hostage is alive, a kidnapper
protocols and permanently archiving Several countries are set to unveil public random number
photographs the hostage with that day’s
them. The services could benefit every- generators, or beacons. newspaper. Similarly, a random number
day applications such as cryptography linked to a digital document proves the
and lotteries—and also research. Some COUNTRY BEACON START RANDOMNESS SOURCE document was not modified any earlier
scientific simulation methods rely on United States July Circuit noise
than the number was generated.
random numbers, and clinicians could An early motivation for public ran-
use them in drug trials to fairly assign Chile July Circuit noise, earthquakes, crypto- domness was to help develop elegant
who gets a treatment or placebo. currency, Twitter, radio streams cryptographic techniques called zero-
“We want to put randomness on the Brazil End of 2019 Circuit noise, radioactive knowledge protocols, which require
internet for people to use in whatever decay statistics each party to have access to the same
S
refer to an incorrect hash, and it would be ix years ago, Emmanuel Unuabonah, sities to make greater efforts to detect
obvious that someone altered the log. In ad- a chemist at Redeemer’s University in plagiarism—such as by installing software
dition, the U.S., Chilean, and Brazilian bea- Ede, Nigeria, read a scientific paper that can detect plagiarized material—and
cons all use the same format, so users can that made him feel “betrayed.” A col- to penalize those who copy. Last year, NYA
mix and match as they please. “People don’t league from Germany had shown him itself ejected a member for plagiarism, and
have to trust the randomness from NIST be- the study, which was published in a it has formally made improper copying a
cause they can combine it with the ones from Nigeria-based journal. In it, four Nigerian dismissible offense.
Chile and Brazil,” Hevia says. researchers presented data copied from a There’s no conclusive evidence that pla-
Eventually, the computer scientists want paper by the German researcher as their giarism is more common in poorer nations
to move to a hack-proof, gold standard of own. Although Unuabonah had nothing to like Nigeria than in wealthier countries.
randomness: quantum-generated random do with the blatant plagiarism, “I felt hu- But a 2017 survey of attitudes toward re-
numbers. Quantum objects don’t have a de- miliated,” he recalls. “It was not good for the search misconduct in low- and middle-
fined state until you measure them, which image of Nigerian science.” income countries found that respondents
means that a random outcome is guaran- The experience led Unuabonah to be- perceived plagiarism as “common,” a team
teed by the laws of nature. In April, NIST come a leader in a growing movement to led by researchers at Stellenbosch Univer-
said it had developed a quantum random combat academic plagiarism in Nigeria, sity in South Africa reported last year in The
number generator based on single photons, Africa’s most populous nation and home BMJ. Similar views emerged from a 2010
which it plans to integrate with the public to more than 150 public and private uni- survey of 133 Nigerian scientists conducted
beacon in the future. versities and colleges. Since 2012, the by physician Patrick Okonta of Delta State
NIST is planning to hold workshops in Nigerian Young Academy (NYA)—an off- University Teaching Hospital in Otefe, Ni-
the next year to brainstorm new uses for its shoot of the Nigerian Academy of Sciences geria. The survey, published in 2014 in BMC
Randomness Beacon. Peralta expects some (NAS) for scientists younger than 45 that Medical Ethics, found that 88% believed
creative ideas. During the Randomness Bea- Unuabonah helped found—has made edu- plagiarism and other forms of misconduct
con’s prototype years, one man thought God cating academics about the pitfalls of were common at their institutions.
ILLUSTRATION: ISTOCK.COM/DANE_MARK
spoke to him through the beacon. He chose plagiarism a major focus of its work. The Also fueling concerns about shoddy schol-
Bible passages based on its output—and group will hold a session on preventing pla- arship in Nigeria is the large number of
when the scripture sequences didn’t make giarism in August at its annual meeting in researchers who publish in low-quality, fee-
sense, he wrote to Peralta complaining Ondo City, Nigeria. This past February, a re- based journals—including a few titles based
about it. “I get funny mail like that,” Peralta cord 350 participants showed up for a day- within the country—that don’t peer review
says. He is expecting better ideas. j long, NYA-run plagiarism workshop, and manuscripts or screen for plagiarized mate-
the group soon hopes to arrange at least rial. An analysis of 2000 papers appearing in
Sophia Chen is a science journalist in six more, one in each of Nigeria’s six geo- such journals, published in Nature in 2017,
Tucson, Arizona. political regions. found that researchers based in Nigeria
Published by AAAS
made up the third largest group of authors, PUBLIC HEALTH
behind authors from India and the United
States. NYA and NAS are now discussing
creating a journal index that would help aca-
demics identify “which are good and which
Newborn screening urged for
are a waste of time,” says NYA President
Temitope Olomola, a chemist at Obafemi fatal neurological disorder
Awolowo University in Ile-Ife, Nigeria, who
is on a 1-year sabbatical at the University of Decision on spinal muscular atrophy due next week
South Africa in Johannesburg.
Some high-profile plagiarism cases have By Meredith Wadman prove if treatment begins before symptoms
involved Nigerians: In 2017, publisher Tay- appear. In this case, the key data come from
R
lor & Francis retracted 10 publications by oughly once a day in the United an ongoing, Biogen-sponsored trial called
Oluwaseun Bamidele, who began publishing States, a child is born with a fatal NURTURE in which 25 newborns with
papers about terrorism as an undergraduate. genetic disorder that destroys motor confirmed SMA got the drug before symp-
Bamidele later told Retraction Watch that neurons in the brain stem and spinal toms developed. In July 2017, when the old-
he didn’t learn about plagiarism rules until cord. In its worst and most common est baby had been followed for 25 months,
he enrolled in a master’s degree program, form, spinal muscular atrophy (SMA) all the children were still alive, none were
and he took responsibility for his missteps. kills children when they are still toddlers, as ventilator-dependent, and those old enough
That lack of training is common among Ni- their respiratory muscles fail. sit up without support could do so—an
gerian students, says Olomola, who recalls But 18 months ago, the Food and Drug achievement unheard of in babies that die
MATERIALS SCIENCE
By Robert F. Service efficiency of 0.5%. Although that’s fathoms that traditional solar cells capture. The re-
below the efficiency of the best perovskite emitted light is concentrated and shunted
L
ance Wheeler looks at glassy sky- cells, Lunt says it’s high enough to power sideways, through the glass, to solar cell
scrapers and sees untapped poten- another window technology: on-demand strips embedded in the window frame.
tial. Houses and office buildings, he darkening glass that halts intense light Because quantum dots are cheap to make
says, account for 75% of electricity in the heat of the day, thereby reducing a and only a small amount of solar cell ma-
use in the United States, and 40% of building’s need for air conditioning. Lunt terial is needed to capture the re-emitted
its energy use overall. Windows, be- believes his team has a clear path to get to light, these solar windows promise to be
cause they leak energy, are a big part of the efficiencies of 4% in the next few years. At inexpensive. Moreover, solar cells work
problem. “Anything we can do to mitigate that rate, the cells could power some of the better under intense, concentrated light.
that is going to have a very large impact,” building’s lighting and air conditioning. Already these windows have reached effi-
says Wheeler, a solar power expert at the At the other end of the spectrum is in- ciencies of 3.1%, Victor Klimov, a chemist
National Renewable Energy Laboratory in frared light, which strikes Earth’s surface at Los Alamos National Laboratory in New
materials, called perovskites, are closing in conductors and metals. Lunt says future sys- ity: Glass breaks, and many solar window
on silicon with top efficiencies of 22%. Not tems that yoke UV-capturing perovskites to technologies contain a small amount of
only are the perovskites cheaper than sili- infrared-capturing organics could reach ef- toxic materials. The technologies also have
con, they can also be tuned to absorb spe- ficiencies of 20%, while still being nearly to be durable enough to last decades, as
cific frequencies of light by tweaking their entirely transparent. demanded by the building industry. But
chemical recipe. A third approach to clear solar windows he says it’s a safe bet to expect that fu-
This week in Joule, a team led by Richard relies on so-called luminescent solar con- ture buildings won’t draw all their power
Lunt, a chemical engineer from Michigan centrators. In these windows, quantum from the grid. They will generate it, too.
State University in East Lansing, reports dots, which are tiny semiconductor par- “Builders have to put in windows anyway,”
that it tuned the materials to develop a UV- ticles, absorb light at UV and infrared fre- Wheeler says. “Why not piggyback on
absorbing perovskite solar window with an quencies and re-emit it at the wavelengths those windows?” j
Published by AAAS
Immune cells (orange) may more readily attack tumor
cells (brown) with many protein-altering mutations.
Blood test may predict cancer tumors. When the TMB was high, as shown
by the same tissue test, the tumor response
rate to the drug doubled. “The future is now
S
ome cancers generate the seeds of search oncologist Antoni Ribas notes that in that the whole process can take 3 weeks, too
their own destruction. Certain random May, half the cancer patients admitted to his long to wait for newly diagnosed patients.
mutations that accumulate in rapidly hospital had been on checkpoint inhibitors The blood TMB test, also from Foundation
dividing tumor cells can spur the im- in the previous 6 months. “It’s a remarkable Medicine, may prove just as effective as the
mune system to attack the cancer. Re- thing that we’re using these agents so much,” tissue test. At the ASCO meeting, Vamsidhar
searchers are now learning that the he says. In some patients the response is dra- Velcheti of the Cleveland Clinic in Ohio re-
extent of such mutations can predict whether matic, but most still don’t benefit, and others ported early results from a prospective trial
a cancer will respond to new, powerful, are never prescribed the drugs. And except of Tecentriq in lung cancer patients who took
immune-based therapies. A recently unveiled for the 4% of patients whose tumors have a a blood test for TMB. The drug shrunk more
blood test for this so-called tumor mutational specific DNA repair defect (Science, 16 June than 36% of tumors that had a high muta-
burden (TMB) could help make it a practical 2017, p. 1111), doctors cannot reliably tell who tional load but only 6% of low-TMB tumors.
tool for guiding cancer treatment. will benefit. Patients with high-TMB tumors went three
Cancer researchers can already gauge TMB Enter the TMB tests. Most assays esti- times longer without their cancer growing
by sequencing a panel of select genes in biop- mate the number of protein-altering muta- back that those with low-TMB tumors did.
sied tissue, an approach that recently dem- tions in a tumor by sequencing a limited Velcheti only reported on the first 58 pa-
onstrated strong predictive power in a large number of genes from its DNA; that tally tients, making any conclusion tentative, Hos-
lung cancer trial. Some cancer physicians likely reflects the density of mutant pro- sein Borghaei, an oncologist at the Fox Chase
now use tissue TMB tests in select cases. But tein fragments, known as neoantigens, on Cancer Center in Philadelphia, Pennsylvania,
the less-invasive blood test, which analyzes the surface of cancer cells. Such fragments cautioned at the meeting. A 580-patient trial
tumor DNA shed into a person’s circulation, aren’t helping the tumor grow; they’re just is underway. Rimm agrees the initial results
could reveal TMB in the many patients where a byproduct of error-prone tumor cell divi- need validation. “They’re just doing pilot
tissue testing doesn’t work. “We’ll see [TMB] sion. But they do appear foreign to the im- studies and saying, ‘Wow, look what we’ve
more and more,” says Naiyer Rizvi, an onco- mune system—and the more neoantigens, found.’ And it is cool what they’ve found.”
logist at Columbia University Medical Center. the more likely that immunotherapy will In April, FDA designated the blood TMB
Still, TMB testing currently takes too long for shrink the tumor and keep it at bay. test a “breakthrough device” that merits a
IMAGE: STEVE GSCHMEISSNER/SCIENCE SOURCE
routine clinical practice, he adds, and some In the lung cancer trial, which was re- priority review. But whether from blood or
in the cancer field question how useful it’ll ported in April at the American Association biopsies, it’s not clear TMB will give doctors
ultimately prove. for Cancer Research (AACR) annual meeting and patients the outcomes or certainty they
Tests that can predict whether immuno- in Chicago, Illinois, researchers found that crave. Rimm points out that trials haven’t yet
therapy will work in a patient are badly mutational load in tumor tissue predicts shown that high-TMB patients live longer on
needed, especially for so-called checkpoint whether a checkpoint inhibitor combination immunotherapy than on chemotherapy. And
inhibitors, which release a brake on immune will help lung cancer patients more than Ribas predicts TMB “will be one component”
cells and enable them to attack tumors. Since standard chemotherapy does. More than of a future combination biomarker. j
the Food and Drug Administration (FDA) in 40% of lung cancers showed a high TMB,
2014 approved the first antibody drug tar- and the patients with those tumors, on aver- Ken Garber is a science journalist in Ann
geting the “checkpoint” protein called PD-1, age, did much better on the immunotherapy. Arbor, Michigan.
VEILED CHAMELEON
AMPHIPOD
Published by AAAS
NEWS
B
illions of years ago, life crossed nings of multicellularity in just a few hun- posted 8 December 2017 on bioRxiv, based on
a threshold. Single cells started dred generations—an evolutionary instant. a review of how different species of fungi—
to band together, and a world of Evolutionary biologists still debate what some single-celled, some multicellular—are
formless, unicellular life was on drove simple aggregates of cells to become related to one another. The same goes for al-
course to evolve into the riot of more and more complex, leading to the won- gae: Red, brown, and green algae all evolved
shapes and functions of multi- drous diversity of life today. But embarking their own multicellular forms over the past
cellular life today, from ants to on that road no longer seems so daunting. billion years or so.
pear trees to people. It’s a transi- “We are beginning to get a sense of how it Nicole King, a biologist at the University
tion as momentous as any in the might have occurred,” says Ben Kerr, an evo- of California (UC), Berkeley, found a reveal-
history of life, and until recently we had no lutionary biologist at the University of Wash- ing window on those ancient transitions:
idea how it happened. ington in Seattle. “You take what seems to be choanoflagellates, a group of living protists
The gulf between unicellular and multi- a major step in evolution and make it a series that seems on the cusp of making the leap to
cellular life seems almost unbridgeable. A of minor steps.” multicellularity. These single-celled cousins
single cell’s existence is simple and limited. of animals, endowed with a whiplike flagel-
Like hermits, microbes need only be con- HINTS OF MULTICELLULARITY date back 3 bil- lum and a collar of shorter hairs, resemble
cerned with feeding themselves; neither lion years, when impressions of what seem the food-filtering “collar” cells that line the
coordination nor cooperation with others to be mats of microbes appear in the fossil channels of sponges. Some choanoflagel-
is necessary, though some microbes occa- record. Some have argued that 2-billion- lates themselves can form spherical colonies.
sionally join forces. In contrast, cells in a year-old, coil-shaped fossils of what may More than 2 decades ago, King learned to
multicellular organism, from the four cells be blue-green or green algae—found in the culture and study these aquatic creatures,
MAGGIE RIGNEY AND NIPAM PATEL; LONGHUA GUO; JAKE HINES AND NATE PETERS/EMBRYOLOGY COURSE AT THE MARINE BIOLOGICAL LABORATORY; (THIS PAGE) WIM VAN EGMOND/SCIENCE PHOTO LIBRARY
in some algae to the 37 trillion in a human, United States and Asia and dubbed Grypa- and by 2001 her genetic analyses were start-
ported on 31 May in eLife. If, as she and Multicellularity made easy and instead use the centrioles full time
others believe, choanoflagellates offer a Researchers got single-cell yeast to evolve multicellularity in for cell division.
glimpse of the one-celled ancestor of an- the lab, demonstrating the relative ease of the transition. Volvox has repurposed other features
imals, that organism was already well- of the single cell ancestor as well. In
equipped for multicellular life. King and 1 Selection Chlamydomonas, an ancient stress re-
her lab “have put protists at the front As single yeast cells sponse pathway blocks reproduction at
of research to address animal origins,” grow, the larger night, when photosynthesis shuts down
ones sink faster.
says Iñaki Ruiz-Trillo, an evolutionary and resources are scarcer. But in Volvox,
Only those cells are
biologist at the Spanish National Re- allowed to reproduce; the same pathway is active all the time in
search Council and Pompeu Fabra Uni- repeated rounds of its swimming cells, to keep their repro-
versity in Barcelona, Spain. selection result in duction permanently at bay. What was
The ancestral versions of those genes ever-bigger yeast. a response to an environmental signal
might not have done the same jobs they in the single cell ancestor has been co-
later took on. For example, choano- opted for promoting division of labor in
flagellates have genes for proteins cru- its more complex descendent, Kerr says.
cial to neurons, and yet their cells don’t 2 Multicellularity A third set of organisms hints at how
A single mutation
resemble nerve cells, King says. Like- this repurposing of existing genes and
causes a repro-
wise, their flagellum has a protein that ducing yeast’s functions could have taken place. Over
in vertebrates helps create the body’s daughter cells the past decade, Ruiz-Trillo and his
left-right asymmetry, but what it does in to stick together. colleagues have compared more than
the single-celled organism is unknown. Branching snowflake a dozen protist genomes with those of
multicellular relatives, says Corina Tarnita, flagella. Chlamydomonas can both swim Ratcliff’s lab. Instead of looking at the fos-
a theoretical biologist at Princeton Univer- and reproduce, but not at the same time. sil record or comparing genomes of existing
sity. For example, in a unicellular relative of Multicellular Volvox can do both at once, organisms, he has recreated evolution in lab
Volvox, Chlamydomonas, organelles called because its cells have specialized. The cultures. “My own research has been not try
centrioles do double duty. For much of the smaller cells always have flagella, which to find out what happened in the real world,
cell’s lifetime they anchor the two whirl- sweep nutrients over the Volvox’s surface but to look at the process of how cells evolve
ing flagella that propel the cell through the and help it swim. Larger cells lack flagella increased complexity,” he explains.
Published by AAAS
As a postdoc working with Michael pends on a bottleneck between generations, 750 generations—about a year—two of five
Travisano at the University of Minnesota in which a single cell (a fertilized egg, for experimental populations had started to
in St. Paul, Ratcliff subjected yeast cultures example) serves as the starting point for the form and reproduce as groups, the team
to a form of artificial selection. He allowed next generation. The result is that all cells wrote on 12 January in a preprint on bioRxiv.
only the biggest cells—measured by how fast in the new generation start out genetically
they settled to the bottom of the flask—to identical. Snowflake yeasts have their own IF MULTICELLULARITY comes so easy, why
survive and reproduce. Within 2 months, way of purging themselves of deviant cells. did it take several billion years after the ori-
multicellular clusters began to appear, as Because mutations accumulate over time, gin of life for complex organisms to become
newly formed daughter cells stuck to their the most aberrant cells are found at the tips firmly established? Traditionally, researchers
mothers and formed branching structures of the snowflakes. But they break off to form have blamed the early atmosphere’s low oxy-
(Science, 18 November 2011, p. 893). new colonies before they have a chance to gen levels: To get enough oxygen, organisms
As each culture continued to evolve— become cheaters. needed the highest possible ratio of surface
some have now been through more than This mechanism also enables group traits to volume, which forced them to stay small.
3000 generations—the snowflakes got big- to evolve in the yeast. Mutations in the cells Only after oxygen levels rose about 1 billion
ger, the yeast cells became more durable and released from each snowflake branch are years ago could larger, multicellular organ-
more elongated, and a new mode of repro- passed on to all cells in the next colony. isms arise.
duction evolved. In large snowflake yeast, a Consequently, subsequent snowflakes start In 2015, however, Nicholas Butterfield, a
few cells along long branches undergo a form out with new group traits—in the size and paleontologist at the University of Cambridge
of suicide, releasing the cells at the tip to number of cells or the frequency and loca- in the United Kingdom, proposed that low
start a new snowflake. The dying cell sacri- tions of suicide cells, for example—that be- oxygen levels actually favored the evolution
fices its life so that the group can reproduce. come grist for further evolution. From that of multicellularity in ancient marine organ-
A
decade ago, a radical theory of cancer emerged: that this plague of multicellular
back. In many lineages, the number of types
organisms arises when their cells rewind the evolutionary clock and revert to act-
of cells and organs continued to grow, and
ing like unicellular life. Recently, David Goode, a computational cancer biologist at
they developed ever-more-sophisticated ways
the Peter MacCallum Cancer Centre in Melbourne, Australia, and colleagues have
to coordinate their activities. Ratcliff and
found evidence to support that idea. They examined gene expression in seven
Eric Libby, a theoretical biologist at Umeå
types of solid tumors—including breast, stomach, and liver cancers—and traced the
University in Sweden, proposed 4 years ago
ancestry of the active genes they found. Genes that date all the way back to early single-
that a ratcheting effect took over, driving an
celled eukaryotic organisms were revved up, Goode’s team reported last year in the
inexorable increase in complexity (Science,
Proceedings of the National Academy of Sciences. In contrast, genes unique to many-
24 October 2014, p. 426). The more special-
celled animals had gone quiet.
ized and dependent on one another the cells
When an organism makes the leap to multicellularity, it must evolve gene regulatory
of complex organisms became, the harder it
networks to ensure its cells stop dividing at the appropriate time and function in step
was to revert to a single-cell lifestyle. Evo-
with their neighbors (see main story, p. 1388). Goode and his colleagues suggested that
lutionary biologists Guy Cooper and Stuart
in cancer, mutations that cripple the networks cause those constraints to break down,
West at the University of Oxford in the United
giving genes suited to a unicellular lifestyle free rein to drive growth. Cancers seem to
Kingdom recently confirmed that picture in
“undo the molecular constraints and controls that evolved to enable multicellular life,”
mathematical simulations. “Division of labor
Goode says.
is not a consequence but a driver” of more
“The idea that cancer represents a release of ancient genes from multicellularity is very
complex organisms, Cooper and West wrote
appealing,” says Mark Vincent, a medical oncologist at the London Health Sciences Centre
on 28 May in Nature Ecology & Evolution.
in Canada. “It explains a lot about what is otherwise mysterious about cancer,” such as
Touched off by the initial transition from
how drug resistance arises, he says, explaining that “a lot of the treatments we have mimic
one cell to many, a cycle of increasing com-
ancient threats, which eukaryotic cells had to have survived.” —Elizabeth Pennisi
plexity took hold, and the richness of multi-
cellular life today is the result. j
A
s soon as the big yellow school sophomore in a hoodie, imagining that the we don’t speak up, there’s only one side be-
bus pulls into the parking lot of center is set up like a video game or Juras- ing heard,” she says. “The side that wants to
the Oregon National Primate Re- sic Park. shut us down.”
search Center (ONPRC) here, it’s Diana Gordon is here to disabuse him of That side has been racking up victories
clear that many of the high school both notions. As the education and outreach recently. In the past 6 months, animal activ-
students on board don’t know coordinator of the country’s largest pri- ist groups have won bipartisan support in
what they’ve signed up for. They mate research center, she spends her days Congress to scuttle monkey and dog stud-
know that science happens some- guiding students, Rotary clubs, and even ies at top U.S. research facilities; they have
PHOTO: ROGER WERTH
where on this wooded, 70-hectare wedding parties through the facility. Here, also helped pass two state bills that compel
campus west of Portland—and that they visitors see monkeys in their habitats and researchers to adopt out lab animals at the
may get to see monkeys—but everything meet scientists—all while learning, Gordon end of experiments. The public itself seems
else is a mystery. “Are we going to go into hopes, that the animals are well-treated and to be turning against animal research: A
some giant underground lair?” asks a lanky the research is critical for human health. “If Gallup poll released last year revealed that
Published by AAAS
Corrected 29 June 2018. See full text.
NEWS
Kingdom, where public support for animal research. As time ticks by, a blue “morally science-fiction characters.”
research is up for the first time in years. acceptable” line creeps downhill, while an Ken Gordon doesn’t blame activists, how-
But will stepping back into the limelight orange “morally wrong” line climbs higher. ever. He blames the biomedical community.
win converts in the United States—or play According to his extrapolations, the lines will Today, most U.S. universities post little, if
into the opposition’s hands? Labs can ma- intersect in 2023 (see graph, p. 1394). anything, on their websites about their ani-
mal research. And many scientists are reluc- to the tour. Gordon says those animals may Inspired, nearly 100 animal facilities in
tant to discuss their animal work because of be susceptible to human diseases and, unlike Spain signed a similar agreement, and last
their own fears or university pressure. the others, aren’t used to seeing large groups week 16 institutions in Portugal did the
“In the old days, researchers at my univer- of people and would be stressed by visitors. same. In February, about 100 U.S. scientists,
sity used to take their spider monkeys out for She tries to tackle head-on any veterinarians, and university administrators
walks,” says Susan Larson, an anatomist at misconceptions the students may have. gathered in San Francisco, California, to call
the State University of New York (SUNY) in “If you see these animals smacking each for more transparency from the country’s an-
Stony Brook. “Now, everything’s a secret.” other, they’re just establishing dominance. imal labs. One upshot: a proposed U.S. Ani-
Larson says SUNY Stony Brook urged Some are losing their hair, some have red mal Research Openness Agreement, which
her not to talk to outsiders about her work bottoms—this is normal during mating sea- if formalized would bind signatories to be
studying locomotion in chimpanzees, “even son. And here’s what monkey chow looks more candid about the animal research they
though most of what we were doing was like,” she says, passing around a plastic bag- do, much like the U.K. concordat.
videotaping them walking around.” Once gie filled with brown pellets. “Yes, it looks a “You could go through the halls of our uni-
animal activists found out about the re- bit like poop, but it isn’t.” versity and not find any information about
search, she says, “they made it sound like ONPRC’s approach echoes one many U.K. where our medical advances came from,” says
I was doing awful things, like sticking elec- research facilities have taken to heart. After Larry Carbone, director of the animal care
trodes in their heads.” Activists and use program at the University
also launched a 2-year legal bat- of California, San Francisco. He
tle to free the animals (Science, Collision course says his university will try to put
6 December 2013, p. 1154). “In the U.S. support for animal research is declining, alarming research groups. more of its animal research on-
end, by not talking to people, it line. “It should be the first thing a
Published by AAAS
Corrected 29 June 2018. See full text.
adds that outreach efforts like ONPRC’s are FOR NOW, the public relations battle between Back at ONPRC, Diana Gordon continues
just a whitewash. “ONPRC’s tour skips the the animal research and activist communi- her own campaign. The students end their
research monkeys,” he says. “It’s essentially ties rages on. The Rescue + Freedom Proj- day in an auditorium with three scientists
a day at the zoo.” He says the research com- ect is pushing Beagle Freedom Bills in three sitting at a table up front. Reproductive
munity has, in fact, been resisting trans- more states, and last month the White Coat physiologist Carrie Hanna tells the group
parency. He points to a U.S. Government Waste Project began a new campaign target- she once wanted to be a veterinarian. At
Accountability Office analysis, released ing USDA for allegedly killing dozens of cats ONPRC, she says, she’s using baboons to
last month, showing that a variety of U.S. a year for parasite research. The group calls it develop a compound that blocks fallopian
research organizations don’t want federal “taxpayer-funded kitten slaughter.” tubes, potentially leading to a permanent
agencies to release more data on animal ex- Meanwhile, Speaking of Research, an contraceptive for women. She explains that
perimentation. “They’re fighting transpar- international organization that supports her work is heavily regulated and that she
ency at every turn.” using lab animals, has launched a Rapid cares about the primates. “We take animal
Even if U.S. institutions do become more Response Network, which sends out email welfare very seriously,” she says. “We’re
open, that doesn’t guarantee it will sway the alerts to counter animal rights campaigns. animal advocates, too.”
public. U.K. polls showing increased support The goal is to engage scientists by prompt- The hoodie-wearing sophomore seems
for animal research as the openness initiative ing them to send letters or sign petitions in content, even though he didn’t get to see
took hold don’t prove the two are related. And support of animal research. The network an underground lair or meet any wild-eyed
Jarrett admits that the United Kingdom may launched its first offensive last week with scientists. “They just seem,” he says, a bit
not be a perfect model for the United States. an open letter published in USA Today disappointed, “like average people.” j
By Elsa A. Olivetti1 species biodiversity, land-use change, climate ecosystem pollution, often in ecologically
and Jonathan M. Cullen2 impacts, and biogeochemical flows. Mitigat- sensitive areas. Furthermore, energy use
ing the impact of materials use is urgent and during this stage is substantial. For exam-
G
lobal annual resource use reached complex, necessitates proactive assessment ple, primary production of metals accounts
nearly 90 billion metric tons in 2017 of unintended consequences, and requires for ~8% of total global energy consumption;
and may more than double by 2050. multidisciplinary systems approaches. this energy consumption is expected to in-
This growth is coupled with a shift of Materials consumption trends provide crease because of decreasing ore grade.
materials extraction from Europe and context to inform strategies for impact miti- The impacts of processes such as refin-
North America to Asia. In 2017, 60% gation. Beginning in the mid-1950s, there ing and manufacturing mainly derive from
of all materials were extracted in Asia, and has been a shift from biomass or renewable energy use. However, direct emissions from
extraction is expected to rise substantially in materials to nonrenewable substances, such process chemical reactions can also be signif-
PHOTO: BLOOMBERG/CONTRIBUTOR/GETTY IMAGES
Africa over the next decade. Local extraction as metals, fossil fuels, and minerals. Effective icant. For example, in cement production, di-
and processing helps to improve standards of strategies for mitigating their impacts are dif- rect CO2 emissions are caused by calcination
living in the developing world, but also leads ferent for high-volume materials with struc- of limestone into lime—accounting for 50%
to important environmental concerns. Glob- tural applications than for specialty materials of cement production–related emissions, the
ally, materials production and consumption with functional uses. remainder resulting from electricity and fuel
is coming up against environmental con- consumption (1). Other direct air, water, and
straints in almost every domain, including MATERIALS IMPACTS FROM soil emissions can also be substantial, caus-
EXTRACTION TO DISPOSAL ing damage to ecosystems and human health.
1
Department of Materials Science and Engineering, Impacts of materials extraction include Most impacts during materials use origi-
Massachusetts Institute of Technology, Cambridge, MA,
USA. 2Department of Engineering, University of Cambridge, landscape degradation, habitat loss, waste nate from the fuel and electricity needed
Cambridge CB2 1PZ, UK. Email: elsao@mit.edu generation, decreased water quality, and to power machines. However, metals may
Published by AAAS
The circuit board of modern mobile phones contains FIVE LEVERS TO MINIMIZE
more than 60 elements from the periodic table. MATERIALS IMPACT
Even as tractable, environmentally ben-
materials) to provide for human activities eficial, revolutionary business models or
such as sustenance, shelter, communica- technologies emerge, they take time to
tion, and education. As much as half of an- implement. Therefore, more evolutionary
nual global materials extraction is used to strategies must also be pursued with ur-
build up or renew in-use stocks. A sustain- gency. We examine five levers for scientists
able materials system creates and main- and engineers to consider to minimize im-
tains these stocks with minimized material pacts: lifetime extension, dematerialization,
and energy flows. manufacturing efficiency, substitution, and
However, the material flows to support recovery (see the figure) (6). These individ-
physical stock vary vastly by region. North ual strategies must be evaluated in concert,
America consumes 30 tons of material per in practice, and from a life-cycle perspec-
capita, Europe 21 tons per capita, and all tive, because efficiencies in one dimension
other regions under 10 tons per capita. may be correlated with increases elsewhere.
This variation underscores the imperative
to decouple economic development and Lifetime extension
enhanced quality-of-life from materials Because so much of materials consumption
consumption. Such a decoupling requires feeds physical infrastructure, reductions in
a profound transition in technology design demand can be achieved through extended
ig
n
developing material substitutes; dresses both sources and sinks for
n•
however, the aim has typically materials streams (15).
Su
Waste
bst
been to improve technical perfor- Fe
26
Materials and production sys-
ituti
22
mance or reduce use of toxic or dif- Ti tems are driven by economic
on
Dispose
Recovery
ficult-to-source materials, rather Improving Waste incentives, but given current
than more comprehensive envi- sustainability understanding of associated ex-
e n cy
ronmental impact. The substan- ternalities, these incentives only
tial time lag between invention tell part of the story. Governance
f ci
and impact assessment has led to and engagement have become
ge
e
myriad cases of ad hoc, reaction- increasingly important; the vast
rin
tur
ac
L if uf a
uf
of toxicity and human health im- e ti
me M an
M an tioned in this perspective require
ex t e n s i o n
pacts. There are promising initial policy to support technology
Waste Use
approaches to predict life-cycle transition. Researchers should
and risk assessment of emerging Waste evaluate their technologies using
materials by scaling laboratory- multiple performance objectives
level data on materials and energy inputs, with recycled material depends on the alloy, and then communicate their findings in a
coupled with industry handbooks (10). These product, and overall market (13). way that policy-makers find actionable. As a
can be combined with evolving data science The scientific community needs to con- society, we should educate scientists and en-
methods to probe and suggest material syn- sider potential trade-offs quantitatively so as gineers on how to perform these assessments,
thesis routes early in development (11). These to avoid unintended consequences, in which engage with stakeholders, and then provide
methods should integrate more closely with we improve one aspect to the detriment of incentives for systems-based environmental
experimental research and be validated with another part of the materials system or life analysis coupled with fundamental research.
production-level information. cycle. For example, although there are some With over 60% of the urban infrastructure
key opportunities in dematerialization, in- that is expected to exist by 2050 yet to be
Recovery creased materials efficiency is typically built and urban population doubling in the
Because materials or components derived coupled with increased demand or function- coming decades, the opportunity exists now
from nonprimary sources typically require ality—such as larger or more accessorized to shape the future of humanity. j
less energy in manufacturing, another strat- cars negating fuel savings. The need to scale RE FERENCES AND NOT ES
egy is to increase recovery (including reuse these approaches must also be considered. 1. G. Habert et al., Cem. Concr. Res. 40, 820 (2010).
or recycling). The ability to retain value and 2. V. Agrawal, I. Bellos, Manage. Sci. 63, 1545, (2016).
3. A. Allanore, Electrochem. Soc. Interf. 26, 63 (2017).
any impact that results will vary by mate- CHARGE TO THE TECHNICAL COMMUNITY 4. M. C. G. Juenger et al., Cem. Concr. Res. 41, 1232 (2011).
rial. Component reuse and repair have long The strategies presented here are not novel; 5. S. Schneiderman, M. Hillmyer, Macromol. 50, 3733 (2017).
been in decline due to rising product com- we can find individual examples of success 6. J. Allwood et al., Sustainable Materials: With Both Eyes
Open (Cambridge Univ. Press, 2012).
plexity, shortened life cycles, de-emphasis throughout the life cycle and for different 7. T. Pollock, Science 328, 986 (2010).
by manufacturers, and societal norms. De- materials. The current sustainability chal- 8. Z. Li et al., Nature 534, 227 (2016).
9. M. J. Eckelman et al., Environ. Sci. Technol. 46, 2902 (2012).
sign efforts should focus on current per- lenge is that scientists and engineers must 10. M. Tsang et al., Green Chem. 18, 4924 (2016).
ceived limits in the degree of modularity, embrace the complexity, anticipate potential 11. E. Kim et al., npj Comput. Mater. 3, 53 (2017).
what parts can be made accessible for re- trade-offs, quantify multiple performance ob- 12. J. Dahmus, T. Gutowski, Environ. Sci. Technol. 41, 7543
(2007).
placement, and consumer uptake. jectives, and estimate scaled impact during 13. T. Zink, R. Geyer, J. Industr. Ecol. 21, 593 (2017).
Materials recovery from products is easier initial research and development. 14. M. O’Connor et al., ACS Sust. Chem. Eng. 4, 5879 (2016).
15. M. Reuter, Metall. Mater. Trans. B 47B, 3194 (2016).
when they are pure and valuable; however, Recovery practices and technologies have
many products in use today are mixed. This not kept pace with the acceleration of com- ACKNOWL EDGMENTS
increase in entropy makes mixed low-value plexity and scale in materials development. A We acknowledge useful discussions with J. S. Krones, H. J. Uvegi,
and R. J. Myers.
materials expensive and energy-intensive to demonstration of this complexity can be seen
recycle (12). Energy reclamation may be the in the number of elements found in a circuit SUPP LEMENTARY MATE RIA LS
best use of mixed plastics given the substan- board: This number has increased from 11 www.sciencemag.org/content/360/6396/1396/suppl/DC1
GRAPHIC: N. CARY/SCIENCE
tial materials degradation with each recovery in the 1980s to more than 60 in the circuit 10.1126/science.aat6821
cycle, the energy used in sorting, and trans- board for modern mobile phones (14). Many
portation required to consolidate volumes. of these constituents are lost at end-of-life;
For metals, recycling efficiency is limited by this loss includes not only the materials but TOMORROW’S EARTH
the thermodynamics in the processing stage, also the value added by the manufacturing Read more articles online
and the displacement of primary material process. To overcome this problem, compat- at scim.ag/TomorrowsEarth
Published by AAAS
POLYMERS
By Matthias Ballauff two polyelectrolyte brush layers, dramati- erties of the system are mainly determined
cally so for Y3+ ions. by the strong osmotic pressure of the coun-
S
keletal joints must provide lubrica- The osmotic pressure that leads to lubri- terions. Thus, the system can be described
tion under considerable load. Lubri- cation can be probed directly when two pla- in terms of a mean-field model; that is, the
cation between two sliding surfaces nar brushes are brought into contact and counterions act as a structureless medium
in aqueous environments can be compressed. Pressing two polyelectrolyte that exerts a certain pressure. The cele-
greatly enhanced by polyelectrolyte brushes toward each other will decrease the brated Derjaguin-Landau-Verwey-Overbeek
brushes (1): Long macromolecular volume available for the confined counter- (DLVO) theory of colloidal stability (2) be-
chains that bear charges at each repeat- ions and increase their osmotic pressure longs to this class of models.
ing unit are grafted densely to a planar or (see the figure, middle). The forces neces- Increasing the valency of the counter-
curved surface. In the so-called osmotic sary to slide two such surfaces when only ions, however, causes the ions to become
limit (low salt concentrations), a large frac- monovalent cations are present are surpris- spatially correlated. For trivalent ions, the
tion of the counterions are confined within ingly small. The forces increase with the mean-field description breaks down, in that
Normal force
Frictional force
– – –
+ + + – –+ + – – + –
– + + – +
– – + – –
–+ + – – + – – +
– + – + – + + – +– + – + +– –
+– Macromoleculee + – –
+ –+ – + –+ + 3+
– – + + – – – 3+
3+ – – +
+ – – – –
–
– + –+ – + +
Counterions – –
+
–
3+
3+ – + – –
+ – + – + – + – – + – + – + +
– + + – – + –
+ – – + +
– – + – + + – – –
–+ + – – – – – +
+ + + – + + +– – – 3+
3+
– –+ – – Polyelectrolyte
tee + – 3+
3 + –
–
+ + – + + – 3
3+
GRAPHIC: A. KITTERMAN/SCIENCE
+– + brush – – –
+ –
– –
– –
+
– +– – –
– + –
Mica surfacee
High lo
loads
oads
oads
dss but
b slick
slick
slic High
h lo
loads
loads
oadss b
but stuck
stu
uckk
uck
Polyelectrolyte
Polye
lye
yeelect
lect
ectrolyte
olyte
y brbrushes
brushe
brush in aqu
usheeess in aqueous
aqueou phase
eous phas
eou hase F monovalent
For t rio ((e.g., Na+)), pressing
alent counterions F trivalent
For tri lent counterions,
c t the chains
chain will
ill partially
tiall
E h surface
Each s f carries
rri a highly
high
ighl swollen
llen layer
l off charged
c the brushes together is resisted by osmotic collapse. The layers become highly inhomoge-
macromolecules. The counterions are mostly confned pressure and requires a large force. The opposing neous, and chains of opposing layers will greatly
within the brush layer and act as a structureless gas. brush layers barely overlap and still slide easily. overlap, which leads to high friction.
O
dynamics of the chains in such a polyelec- ver the course of the past 540 million ing that the oxygen isotope composition of
trolyte brush for higher-valency counter- years, five catastrophic mass extinc- seawater remained steady through the K/Pg
ions. The dynamics of single chains is a tion events occurred as a result of boundary, which can be considered as a
well-known problem, and theories operat- global climate changes. Those periods robust hypothesis, the decrease of isotopic
ing at this level may be valid to a zeroth ap- of large-magnitude warming or cool- ratios that MacLeod et al. measure corre-
proximation even for dense polyelectrolyte ing resulted from catastrophic events sponds to an increase in seawater tempera-
brushes. However, the strong cross-linking such as asteroid impacts, paroxysmal volca- ture of about 5°C.
Published by AAAS
years ago) and at the end of the Devonian have been documented by paleontologists. STEM CELLS
(380 to 360 million years ago) were linked For example, it would be of great interest
to global cooling (9), with about 57% of ma-
rine genera wiped out in both cases (10, 11).
In the case of the P/T and K/Pg boundar-
to quantify seawater temperature changes
during the Triassic/Jurassic transition.
More generally, knowledge of the magni-
Macrophages
ies, warming was most likely caused by a
paroxysmal volcanic activity that released
tude and duration of past global warming
events and their impact on Earth’s life is of stimulate
huge amounts of carbon dioxide into the
atmosphere. Such extraordinary volcanic
activity is documented in the geological re-
paramount importance for developing cred-
ible scenarios of the global warming we are
facing today. This warming is expected to
mammary
cord with the deposits of massive lava flows
in the Siberian and Deccan traps for the P/T
and K/Pg boundaries, respectively.
deeply reshape the size and distribution of
food and water resources for humans, as
well as to seriously threaten the marine and
stem cells
Climate warming can also cause forma- terrestrial biodiversity of our planet. j Macrophages help mediate
tion of suboxic and anoxic seawater masses.
The solubility of gaseous phases decreases
RE FERENCES AND NOTES hormone-controlled
with increasing temperatures. This means
1. R. K. Bambach, Annu. Rev. Earth Planet. Sci. 34, 127
(2006). changes in the mouse
that the availability of CO2, a nutrient for
photosynthetic organisms, decreases with
2. K. G. MacLeod et al., Science 360, 1467 (2018).
3. Y. Kolodny et al., Earth Planet. Sci. Lett. 64, 398 (1983). mammary gland
4. S. Picard et al., Geology 26, 975 (1998).
increasing water temperature; similarly, less 5. M. J. Benton et al., Trends Ecol. Evol. 18, 358 (2003).
O2 is dissolved in warmer waters, restricting By Nagarajan Kannan1 and
T
causing rivers to carry more nutrients (dis- 9. M. R. Saltzman, S. A. Young, Geology 33, 109 (2005). he adult mammary gland is a bilay-
solved nitrate and phosphate) to estuaries 10. P. M. Sheehan, Ann. Rev. Earth Planet. Sci. 29, 331 (2001). ered branching epithelial structure
11. P. Hull, Curr. Biol. 25, R941 (2015).
and coastal waters, where the resulting eu- consisting of an outer layer of basal
trophic conditions threaten the large biodi- ACKNOWLEDGME NTS cells and an inner layer of luminal
versity of aquatic animals they host. The author is grateful to R. Amiot, who provided the figure, to G. cells. The neonatal structure expands
The isotopic analysis of fish remains used Cuny for the valuable information about the Phanerozoic biotic rapidly during puberty and then un-
crises, and to J. Goedert for a careful reading of this manuscript.
by MacLeod et al. could be applied to other dergoes cyclic growth in response to the
key periods of the past 540 million years changing hormonal stimuli (progesterone
for which mass extinction events already 10.1126/science.aau1690 and estrogen) in each menstrual cycle (1).
These dynamic responses of the mammary
gland involve important interactions with
Learning from past climatic changes various surrounding nonepithelial cells that
66 million years ago, sea temperatures rose rapidly as a result of environmental perturbations. The result was constitute its “niche.” Circulating macro-
a mass extinction event that led to the demise of nonavian dinosaurs and the rise of mammals and birds. phages are important constituents of this
niche although the mechanism through
1 The nonavian dinosaur Tyrannosaurus which they influence mammary cell prolif-
2 The mammal Didelphodon eration has remained unclear. On page 1421
3 The marine crocodylomorph Dyrosaurus of this issue, Chakrabarti et al. (2) show that
3 9
4 The marine lizard Mosasaurus mouse mammary stem cells [cells with an
5 The marine sauropterygian Elasmosaurus ability to regenerate an entire mammary
6 The shark Cretolamna gland (3, 4)] are enriched within a subset
Mass extinction
7 The bird Hesperornis event
4 10 of cells with a phenotype of basal layer cells
8 The amphibian Piceoerpeton
9 The bird Gastornis
and express the delta-like 1 (DLL1) NOTCH
10 The mammal Chriacus ligand, which allows them to interact with
nearby NOTCH-expressing macrophages.
Extinct Survived Emerged 5 8 This interaction triggers the production of
multiple Wnt (Wingless-related integration
site) ligands that, in turn, induce an expan-
sion of the mammary stem cell population.
6 3 This finding is important because it has im-
plications for understanding how breast can-
Deccan traps cer may develop.
1
The demonstration that mammary stem
7 6 cells lack steroid hormone receptors [pro-
Volcanic paroxysm gesterone receptor (PR) and estrogen re-
GRAPHIC: N. DESAI/SCIENCE
2 8 +5°C –5°C
1
Division of Experimental Pathology, Department of Laboratory
CRETACEOUS PALEOGENE Medicine and Pathology, Mayo Clinic, Rochester, MN 55905,
USA. 2Terry Fox Laboratory, British Columbia Cancer Agency,
66.25 Ma 66 Ma 65.9 Ma Vancouver, BC V5Z 1L3, Canada. Email: ceaves@bccrc.ca
to respond to changing amounts of Proposed macrophage control of MaSCs stem cells during and after puberty
progesterone and estrogen depends Mammary stem cell (MaSC) numbers oscillate during the estrous cycle, appears to be under the joint con-
on a multistep mechanism involv- mirroring similar oscillations of serum progesterone and macrophage trol of ovarian hormones and mac-
ing the production of intermediate numbers in the breast. NOTCH+ macrophages interact with DLL1 rophages. What then regulates the
signals by hormone-responsive cells. expressed by a subpopulation of MaSCs with basal features. This triggers observed changes in macrophage
Such a process would then explain macrophages to secrete Wnt ligands. In response to progesterone, numbers? And will macrophages
how oscillations in blood concentra- the ER+PR+ ductal cells secrete WNT4 and RANKL. Together with affect mammary stem cell prolif-
tions of these key ovarian hormones macrophages, this brings Wnt ligands above a threshold that stimulates erative responses in vitro to enable
during the estrous cycle cause MaSC proliferation. further dissection of their effects?
changes in the size of the entire Although mammary-macrophage
gland (see the figure). Further sup- MaSC signaling via a DLL1-Wnt pathway
port for this concept came from the may create a positive-feedback loop
subsequent demonstration of a sig- that is important for the rapid ex-
nificant expansion and contraction Macrophage pansion of mammary stem cells in
of this functionally defined mouse vivo, are there separate mechanisms
Progesterone
mammary stem cell population dur- that counterbalance and fine tune
ing the estrous cycle and pregnancy the restoration of baseline mam-
(6, 7), mediated indirectly by PR- Estrus Metestrus Diestrus Proestrus mary stem cell numbers?
expressing luminal cells stimulated Wnt Notably, these findings are also rel-
to produce receptor activator of nu- MaSC WNT4 evant to understanding the potential
clear factor kB ligand (RANKL) and NOTCH Bipotent progenitor RANKL role of macrophages in the genesis
Published by AAAS
NEUROSCIENCE
By Silvia Arber1,2 and Rui M. Costa3,4 primary sensory systems such as olfaction, vided by function aligned to different swim-
whether an organizational logic exists at the ming speeds (3). Although much work lies
M
ovement is the most common final output steps within the motor system was ahead, one can speculate that specialized
output of nervous system activity unclear. We now know that circuits regulat- spinal microcircuits and their associated
and is essential for survival. But ing the functionally opposing extensor and sensory feedback loops are the recipients of
what makes this seemingly trivial flexor muscles are connected into distinct motor commands from the brain, and that
statement so scientifically chal- and spatially separate spinal submodules in these microcircuits are essential substrates
lenging? Neurons that contribute mice (2), and that in zebrafish, even within to produce diverse actions as the behavioral
to when and how our body moves are dis- one genetic class, neurons can be subdi- output of the brain (see the figure).
tributed throughout the nervous sys- The spinal cord alone cannot gen-
tem. Thus, even a simple movement erate sustained movement. Best sup-
such as arm flexion requires the coor- Circuits for body movements port for this statement comes from
dinated activation of many different Movement requires the coordinated activation of many different patients with complete spinal cord
sion to stratify neuron function, a specific upstream circuits involved in choice and The motor cortex is the evolutionarily
descending glutamatergic brainstem popula- coordination also connect to more than one most recent structure of the motor system
tion was implicated in halting locomotion (7). command line. but also the most controversial one to which
As a last, specialized example, a glutamater- It is easier to comprehend action choice to assign behavioral functions. Early stud-
gic population marked by expression of cor- when it is driven by strong external stimuli ies demonstrate that cats without a motor
ticotropin-releasing hormone controls urine than by cognitive decision-making processes cortex can still perform a large behavioral
release by regulating bladder contraction (8). guiding voluntary behaviors. A classic ex- repertoire, supporting a model in which
These exemplary studies convey two main ample of the first category is the innate subcortical structures play dominant roles
messages. First, brainstem areas use division behavioral interaction between prey and in movement control. An intriguing feature
of labor to control diversification of body predator. Two opposing motor programs— of cortical neurons is that they are the only
movement. Executed behaviors range from predation or evasion—are regulated by dis- class of supraspinal neurons that directly
full-body forms, including locomotion and tinct intermingled neuronal subpopulations communicate with most other motor struc-
orientation, to skilled movement sequences within the hypothalamus (9). Additionally, tures, including the basal ganglia, thalamus,
of body parts during forelimb manipula- the superior colliculus is an important hub midbrain, brainstem, and spinal circuits.
tion behaviors. Second, within a brainstem to compute inputs from many brain areas, Through this broadcasting mechanism, they
region, functionally diverse neurons are including sensory modalities, and mediate can convey and distribute context-dependent
frequently intermingled. This is the likely choice, as exemplified for different forms of and cognitive information widely. Conceptu-
reason why progress in assigning functions defensive behavior in mice in response to ally, motor cortex output can both serve as
to neurons was slow, and only application looming stimuli mediated by subpopulations a command line and influence action choice
of additional criteria—including projection of Parvalbumin-expressing neurons (10). In via the modulation of other command lines.
targets, neurotransmitter identity, or genetic both cases—responses to predator or prey A growing body of evidence suggests that cor-
Published by AAAS
Taxonomy of many species of Rhododendron from
the Himalaya (four are seen flowering) is uncertain.
T
(CBD) commits its 196 nation parties The opportunity to ever know about many munity interest could not be reasonably or
to conserve biological diversity, use of the kinds of organisms with which we practically asserted. Eight are derived from
its components sustainably, and share share this world is rapidly slipping through fungi common in soil or similar environ-
fairly and equitably the benefits from our fingers. Of the estimated 12 million spe- ments, and two are obtained from geneti-
the utilization of genetic resources. The cies of eukaryotes on Earth, fewer than 2 mil- cally engineered bacteria or ovarian cells (2).
last of these objectives was further codified lion have been named. Current estimates are Additionally, high-throughput screening,
in the Convention’s Nagoya Protocol (NP), that 20% of the species on Earth are in dan- combinatorial chemistry, synthetic biology,
which came into effect in 2014. Although ger of extinction, driven primarily by a range and other advanced methodologies have
these aspirations are laudable, the NP and of human activities. Although biological re- largely replaced the role of natural products
resulting national ambitions on Access and sources had long been treated as a common in the discovery of new molecules for devel-
Benefit Sharing (ABS) of genetic resources heritage of humankind, the CBD reinforced oping new drugs, rendering physical access
have generated several national regulatory the notion of sovereign rights of nation states to biological material less important than it
regimes fraught with unintended conse- over biological resources within their politi- has been in the past. Modern technologies,
quences (1). Anticipated benefits from the cal boundaries and entrusted the nation par- including CRISPR gene editing, are redefin-
commercial use of genetic resources, espe- ties to take measures to share benefits arising ing the modalities of access and utilization
cially those that might flow to local or in- from the utilization of genetic resources. of biological resources in ways that were not
digenous communities because of regulated In most countries, particularly developing foreseeable during NP negotiations.
access to those resources, have largely been countries, the agendas of numerous pressure Overall, examples of financially significant
exaggerated and not yet realized. Instead, groups, many of them well-intended but not ABS agreements, a quarter-century after the
national regulations created in anticipation prioritizing science, get mixed up during the CBD was signed, are scarce. Often-mentioned
of commercial benefits, particularly in many legislative process, while conservation biolo- cases are marginal arrangements for the use
countries that are rich in biodiversity, have gists and taxonomists, a vanishingly small of plant extracts for treatment of bone frac-
curtailed biodiversity research by in-country constituency, hold little leverage. Thus, the tures as is traditional in the Cook Islands, the
scientists as well as international collabora- resulting national legislations vary greatly, failed Merck-INBio initiative in Costa Rica,
tion (1). This weakens the first and foremost from being extremely prohibitive of research, and the now discredited case of the “Indian
to a very few that are relatively enabling, ginseng.” A survey of mostly megadiverse
1
Kerala Agricultural University, Thiruvananthapuram such as Costa Rica and South Africa. The countries having functional ABS legislation
695522, Kerala, India. 2Ichthyology Section, Australian
problem is particularly acute where there is showed that very few commercial ABS agree-
PHOTO: SANDESH KADUR
OBSTACLES TO RESEARCH go beyond physical access to genetic mate- by-case negotiations, the Seed Treaty adopted
The principles underlying the CBD and NP rials and run counter to the larger overall a multilateral system for access and benefit
are laudable, and underscore that interna- goals of the CBD. Scientific information in sharing (MLS) through a Standard Material
tional collaboration in research is crucial for the form of DSI is increasingly being pub- Transfer Agreement, averting the need for
conservation of biodiversity and that access to lished through portals of the International bilateral negotiations. The MLS established
genetic resources should be facilitated. How- Nucleotide Sequence Database Collabora- under the Seed Treaty has been viewed as
ever, even as national governments, following tion (INSDC) such as GenBank. Unlimited a very successful model in terms of volume
the CBD, began to enact legislation to regu- and open access of DSI encourages collabo- of material exchanged (8500 transfers every
late access to their biological resources and ration to gain insights into the evolution, week) (12), in contrast to the very limited per-
benefit-sharing from the derived products, maintenance, conservation, and sustainable formance of the bilateral system of CBD and
consequences of such actions on biodiver- use of biological diversity. NP (3). Exchange of genetic material under
sity research and food security were pointed Although NP Article 8(a) appears to en- the Seed Treaty is exempted from the NP, and
out by the science community (4–6). About courage regulations that do not impede the benefit-sharing requirement arises only
100 countries have enacted, or are consider- bona fide scientific research, the NP’s when access for further research and breed-
ing, laws that regulate access by scientists to definition of the “utilization of genetic re- ing is restricted through intellectual property
biological material and benefit sharing. Since sources” as the “means to conduct research rights. One possible course of action for the
the CBD came into effect, and especially after and development on the genetic and/or bio- CoP to the CBD might be to add an explicit
the NP led nation states to step up legisla- chemical composition of genetic resources” treaty or annex to promote and facilitate
tive processes to tighten their control over (Article 2c) makes no exceptions for purely biodiversity research, conservation, and in-
genetic resources (1, 7 ), obtaining permits academic or conservation-related biodiver- ternational collaboration. Such a treaty will
for access to specimens for noncommercial sity research, such as taxonomic studies. address legal uncertainties in the governance
The recent decision to consider the use of essential food and fodder crops. Whereas www.sciencemag.org/content/360/6396/1405/suppl/DC1
of digital sequence information (DSI) under the CBD and NP necessitate access to genetic
the framework of the CBD and NP (10) may resources on a bilateral basis through case- 10.1126/science.aat9844
Published by AAAS
A merger of agribusiness giants Bayer and Monsanto,
approved on 29 May 2018, has some concerned.
A
s a biologist working to understand comes face to face with evidence that con- panies due to the tremendous cost associated
how plants sense and survive in stress- tradicts what he thought he knew about GM with the regulatory process. Lynas points
ful environments, I hope that some of technology. “Certainly it was very worrying if out how efforts by Greenpeace have created
my laboratory’s findings will contrib- real scientists—not to mention the scientific a system in which “only the most profitable
ute to a more sustainable society by community in general—were on the other mass-market global commodity crops have
reducing the environmental costs of side from me on this issue,” he writes. By the been worth investing in.” “Activism has been
growing food. Any success we achieve will end of chapter 7, science has won the debate. the most successful in locking out small and
likely involve the use of genetic Most major global scientific public sector players … thus cementing ex-
modification (GM) technology. organizations have firmly stated actly the monopolistic situation that many
But this method of crop improve- that science backs the efficacy and campaigners say they are fighting against.”
ment has become the subject of a safety of genetic engineering. Yet In the end, Lynas draws a line in the sand.
contentious debate that has tem- in the minds of many, consuming If Greenpeace and other environmental ad-
pered the enthusiasm of many food with a GM organism (GMO)– vocacy organizations are going to fight the
governments and food producers. free label is a must. So what went use of genetic engineering in agriculture,
In Seeds of Science, Mark Lynas wrong? Lynas argues that applying the old arguments—that GMO crops are
gives readers a firsthand look at GM technology first to herbicide- unsafe for consumption or ecologically haz-
both sides of the discourse. Ly- Seeds of Science resistant crops was a mistake ardous—need to be abandoned. “We have al-
Why We Got It So
nas, formerly a dyed-in-the-wool that aligned the chemical manu- ready wasted 20 years fighting over a mere
Wrong on GMOs
anti-GM activist for Greenpeace, facturing industry—which was seed-breeding technique that—used sensibly
PHOTO: GEORGES GOBET/STAFF/GETTY IMAGES
Mark Lynas
is now an advocate for the safe Bloomsbury Sigma, already regarded with skepti- and in the public interest—can certainly help
use of GM technology. The book 2018. 304 pp. cism—with the burgeoning tech- global efforts to fight poverty and make ag-
begins with heart-racing accounts nology. If pest-resistant crops that riculture more sustainable,” he writes. “Let’s
of the law-breaking activities Lynas engaged allowed farmers to apply fewer chemical not waste 20 more.” j
in as one of the pioneers of the anti-GM pesticides had been introduced first, the nar- 10.1126/science.aat8772
movement. We follow along as he slashes rative might have been different.
corn plants in a research field, runs from The book’s subtitle, “Why we got it so
wrong on GMOs,” may, at first, seem to refer to TOMORROW’S EARTH
The reviewer is at the Department of Biology, Stanford University, the activists at Greenpeace. But Lynas might Read more articles online
Stanford, CA 94305, USA. Email: dinneny@stanford.edu believe the people who “got it so wrong” are at scim.ag/TomorrowsEarth
I
n the 19th century, August Weismann sev- literate readers, and their evaluation of key authors’ skepticism is based on the mistaken
ered the tails of mice, observed no reduc- arguments put forward against extended inference that any such claim requires NGI
tion in tail length among their offspring, heredity persuasively demonstrates how to spread without selection through high
and declared Lamarckian inheritance NGI can no longer be dismissed as “lim- rates of adaptive “mutation.” However, the
refuted. Had he instead removed “teeth” ited,” “functionally unimportant,” or always debate, in my view, is not over whether NGI
from the amoeba Difflugia corona, he “under genetic control.” can drive adaptive evolution without selec-
would have found reliable inheritance of the Scholars of human evolution may be frus- tion (selection is almost always important),
disfigurement. The amoeba experiment was trated that Bonduriansky and Day have not nor over whether NGI generates adaptive
conducted by Herbert Jennings in 1937, but engaged much with the cultural evolution variation more frequently than nonadaptive
by that time research into nongenetic inheri- literature. Culture is by far the most exten- variation (aside from social learning, it prob-
Published by AAAS
A Boston University summer intern conducts climate
research. Education that fosters connection to nature
may help prepare students for future challenges.
Interdisciplinary collaboration
To address the increasingly complex and
interdisciplinary challenges of the future,
U.S. education must give up the current
conveyor-belt model of education, which
focuses on testing and standardization in
LET TERS isolated subjects, each with its own termi-
nology. Instead, we should move toward a
Education for the future no single and simple answers. Students form
groups to investigate the question, explore
the question from diferent viewpoints, and
We asked young scientists: Are our schools and universities adequately finally form a joint decision. Rather than
abstract knowledge, students work on real-
prepared to educate young people for future challenges? What is the
life problems that are relevant to them.
most pressing issue in your field, and what one improvement could Beat A. Schwendimann
your country make to its current education system to prepare students Graduate School of Education, University
of California, Berkeley, Berkeley, CA 94720, USA.
to face it? The responses expressed concerns about the current state of edu- Email: beat.schwendimann@gmail.com
cation in countries around the world. Many students lack access to the infor-
Despite decades of modern research,
mation they need, and those with access are often constrained by curriculum
complex disorders such as cancer and
that emphasizes rote learning and isolated subjects. Our respondents sug- cardiovascular disease remain major health
gested a variety of improvements to prepare the next generation for success. issues and claim millions of lives worldwide
every year. To address these challenges,
Turkey’s education system should emphasize
memorization and testing less and focus
Connection with nature education system must develop programs instead on group projects that encourage
There is a growing disconnect between that reconnect students in urban and rural students to think broadly and work together
humans and nature. For many people Australia. Through tourism and sustain- efectively. We must prepare students for
in Poland, nature is just a boring word able agriculture, regional communities and cross-disciplinary work, which is inherently
associated with a subject taught at primary environmental assets will continue to bring collaborative, as it brings together individu-
jobs and growth to future generations, als with a variety of expertise.
school. I propose creating opportunities
including those in the cities.
for positive outdoor experiences by taking Gurkan Mollaoglu
teaching outside the classroom. Positive Adrian Ward Department of Oncological Sciences, University
PHOTO: CYDNEY SCOTT FOR BOSTON UNIVERSITY PHOTOGRAPHY
University of Queensland, St. Lucia, QLD 4072, of Utah School of Medicine, Salt Lake City, UT 84112,
outdoor experiences not only benefit Australia. Email: a.ward@uq.edu.au USA. Email: gurkan.mollaoglu@hci.utah.edu
students’ mental and physical health, they
create the bonds that lead those students Because nature conservation is the respon- U.S. schools are not properly preparing
to care for the global environment. sibility of all citizens, and because there students to become the scientists of tomor-
Barbara Pietrzak is no better way to learn than to experi- row. Whereas liberal arts and languages are
Department of Hydrobiology, Faculty of Biology,
ence, Hong Kong’s universities should often required courses for students across
Biological and Chemical Research Centre,
University of Warsaw, 02-089, Warsaw, Poland. implement a compulsory field course for all fields, classes like computer science and
Email: b.pietrzak@uw.edu.pl undergraduate students from all disci- machine learning are often taken only by
plines of study. Seeing is believing; through students focused on physics, engineering,
Australia’s growing urban population is field expeditions, students will learn to or technology. A generation of chemical,
becoming culturally disconnected from appreciate the beauty of biodiversity. We molecular, cellular, organismal, and natural
the efects of land and water degrada- can then tell our future pillars of society biologists have had to learn how to adapt
tion and the loss of native wildlife. The how humans have afected the natural and create technology when they start
Equal access
The biology curricula in Pakistan largely
ignore the topic of biological evolution,
hindering students from fully understanding
such diverse fields as medicine, pharmacol-
ogy, nutrition, and agricultural sciences.
Teaching biological evolution, especially with
the support of theologians who acknowledge
its importance, would help students connect
Boys in China play an online the dots between the development, spread,
game. Future education may and prevention of antibiotic resistance.
address internet addiction. Saima Naz
Lahore, Punjab, 54000, Pakistan.
Email: saimanaz85@gmail.com
role of genetic information in society education system is the assumption that regardless of income, race, location, or
will continue to expand, both within students will automatically acquire the socioeconomic status. In the United States,
and beyond the medical health arena skills necessary to cope with difculties programs for gifted children are overwhelm-
(including direct-to-consumer testing when they arise. Yet skills such as emotion ingly targeted toward wealthy communities,
and applications in law enforcement regulation and interpersonal communi- whereas low-income and high-diversity
and insurance). Although the Australian cation can be taught. To truly prepare schools often face difculties simply
secondary education curriculum covers students for future challenges, these skills securing faculty with adequate teaching
Published by AAAS
credentials. Equal education opportuni- care providers will still be required to relay challenges such as climate change, food
ties at the elementary level will ensure that information to team members and provide insecurity, antimicrobial resistance, or
all talented children have the potential to patient counseling. artificial intelligence. To design a better
pursue higher education. Such higher educa- Cody Lo future, students need to design solutions for
tion should be afordable to individuals of all Faculty of Medicine, University of British Columbia, those who are hit hardest. The future will
Vancouver, BC V6T 1Z2, Canada. only be prosperous if it is inclusive to all.
income levels. Only if we embrace the gifts of
Email: codylo94@gmail.com
every individual in the upcoming generation Rense Nieuwenhuis
can we be confident in confronting the great- Swedish Institute for Social Research (SOFI),
est challenges the world has yet to face. Social responsibility University of Stockholm, 10691, Stockholm, Sweden.
Email: rense.nieuwenhuis@sofi.su.se
Kyle Isaacson In India, where there is unrestricted sale
Department of Bioengineering, University of Utah, of over-the-counter antibiotics, no national
Salt Lake City, UT 84112, USA. Egypt consumes energy in a careless way.
Email: kyle.isaacson@utah.edu antibiotic policy, and no national anti- Students must be taught the importance
microbial research agenda, problem-based of energy and how to mitigate its usage in
Despite the obvious benefits of computa- training on antimicrobial resistance and their daily life. Schools should raise aware-
tional knowledge in today’s society, many antibiotic stewardship should be made ness about energy challenges, and students
secondary schools in the United States do mandatory for all undergraduate and post- should be rewarded for reducing energy
not ofer a single computer science course. graduate medical students. waste and using renewable sources.
This lack of opportunity disproportionately Prashant Sood Basant A. Ali
afects minority and low-income students, MRC Centre for Medical Mycology, Energy Materials Laboratory, School of Science and
University of Aberdeen, Aberdeen, AB25 2ZD, UK. Engineering, American University in Cairo, Cairo,
systematically excluding would-be innovators Email: drprashantsood@gmail.com 11835, Egypt. Email: basantali@aucegypt.edu
Email: fbeardsley@laverne.edu
Tyler Jones
School of Engineering, Brown University, Providence, The Indian education system needs to be
Canada’s education system must be proac- RI 02912, USA. Email: tojjones@gmail.com revamped to include practical learning.
tive in training professionals for potential Educators should be trained to encourage
shifts in responsibilities. Training medical Students need to learn to think about students to think and not just memorize.
students to be efective communicators and inequality. Sweden’s schools should Project-based learning will bring out cre-
team members will be even more impor- emphasize that not everyone will be ativity among children. Real-world scenarios
tant in the wake of artificial intelligence, as able to avoid the consequences of future must be discussed to sensitize students to
the severity of issues such as water short- Critical thinking To close the gap between research and
policy, Australian universities should
age, climate change, and food security.
Schools should visit nearby research insti- Avoiding violent conflicts, especially encourage all undergraduate students
tutes, where discussions with scientists can among developed countries, is a necessary to participate in cutting-edge research
motivate students to pursue research and precondition for tackling future challenges programs and provide them with experi-
solve problems through innovation. such as climate change and food security. ence that emphasizes the importance of
The Czech Republic needs to start teaching fundamental as well as applied scientific
Brijesh Kumar
Dr. Sneh Lata Singla-Pareek’s Lab, Plant Stress children the value of cooperation, peace, research. This is critical for students in
Biology Group, International Centre for Genetic and democracy, while highlighting the the humanities, business, and law, who
Engineering and Biotechnology, New Delhi, 110067, costs of lies, corruption, and conflict. In traditionally transition more readily to
India. Email: brijeshkumar2412@outlook.com
doing so, educators must teach students political life. Similarly, science students
how to distinguish facts from deliberate should be given insights into politi-
Because creativity often requires proficiency misinformation. As support for totalitarian cal decision-making to encourage more
in other fields, U.S. medical school admis- regimes soars, the risk of serious conflict participation of scientifically educated
sions ofces should place more value on increases. The older generation in the people in politics. If we educate larger
nonmedical work experience. Encouraging Czech Republic grew up in a totalitarian cohorts of students who appreciate how
applicants with real-world work experience environment; we must teach the next gen- evidence-based decisions can advanta-
rather than applicants who just completed eration a better way. geously inform policy, our future will be
an undergraduate degree would result in
Lubomír Cingl much brighter.
future physicians with the backgrounds and Department of Institutional, Environmental, and Anthony Peter O’Mullane
skills from which creativity is born. Experimental Economics, Faculty of Economics, Department of Chemistry, Physics, and Mechanical
10.1126/science.aau3877
TOMORROW’S EARTH
Read more articles online
Current education about genetically modified organisms, such as this rapeseed crop, is inadequate. at scim.ag/TomorrowsEarth
Published by AAAS
AAAS NEWS & NOTES
By Anne Q. Hoy GM crops are developed to express specific traits such as disease
and insect resistance and herbicide tolerance, alterations shown to
Scientists are using technology to expand global food production and protect crop yields and decrease the use of insecticides and herbicides
ease its environmental impact, but advances are being challenged necessary to grow the crops. In the United States, such crops are
by claims that lack scientific evidence and raise public distrust and largely used for animal feed and as ingredients for some consumer
concern, a leading agricultural scientist told an American Association products, including cereals and corn chips. Slowing integration of the
for the Advancement of Science audience. technology, Van Eenennaam said, dents production levels and requires
Alison Van Eenennaam traced the advent of campaigns against additional acreage and more fertilizer, pesticide, and insecticide use.
agricultural innovations related to areas from cattle and chicken Provocative in defense of agricultural science, Van Eenennaam said
production systems to plant biotechnology. The impact such efforts many scientists avoid jumping into topics like the safety of GM crops
are having on agricultural advances was the focus of the ninth annual out of a “fear of isolation.” Yet, leaving false claims unanswered creates
AAAS Charles Valentine Riley Memorial Lecture on 5 June at the a “spiral of silence” that GM opponents leverage, she said, widening the
AAAS headquarters in Washington, D.C. perception gap between scientific knowledge and the general public’s
By way of illustration, Van Eenennaam examined the controversy views. “We need to defend these objective truths around science, irre-
over the adoption of genetically modified crops, known as GM crops spective of the subject area,” she said. “Quite often with agriculture it’s
or GMOs (genetically modified organisms), and the emergence of a a lonely road out there if you’re trying to correct misinformation.”
“parallel science” that has led to public opposition and misconcep- A 2015 Pew study, for instance, found that 88% of AAAS member sci-
tions about the safety of GM plants. “There is just example after entists consider GM foods safe to consume, while only 37% of the gen-
CREDIT (TOP TO BOTTOM): GETTY IMAGES/SCOTT OLSON
example of this as it relates to agriculture where bad decisions are eral public consider them safe and 57% deem GM foods unsafe to eat.
being made that ignore the evidence, based on some people’s world- The resulting 51% gap between the views of scientists and those of the
view and gut instinct that there must be a better system,” said Van public on GMO food safety amounts to an opinion difference greater
Eenennaam, a Ph.D. animal genomics and biotechnology cooperative than divisions over other controversial issues such as climate change,
extension specialist and researcher at the University of California, childhood vaccines, and human evolution, study authors reported.
Davis. “But there is no discussion of the really important trade-offs.” AAAS has defended the validity of scientific evidence on GM crops.
Controversy over the use of GM crops persists today, decades The AAAS Board of Directors issued a statement on 20 October 2012
after they were first commercialized in the mid-1990s and despite describing GM crops as safe. “Indeed, the science is quite clear: crop
widespread use of the technology. In 2017, 92% of planted corn, 96% improvement by the modern molecular techniques of biotechnology
of cotton, and 94% of soybeans grown in the United States were is safe,” it said.
GM varieties, according to data collected by the U.S. Department of “The World Health Organization, the American Medical Association,
Agriculture’s National Agricultural Statistics Service. the U.S. National Academy of Sciences, the British Royal Society, and
By Becky Ham sciences, and pollution and climate research that have both local and
global impacts,” said Rush Holt, AAAS CEO and executive publisher
Rich in sunshine, oil, and fertile soils and flush with eastern trans- of the Science family of journals. “Our Division meetings are unique
plants seeking new job opportunities, California became a hub of because they bring together a local community with its scientists
science and technology in the 20th century—leading the United and engineers and create opportunities to discuss how new tech-
States in energy extraction, aerospace engineering, and innovative nologies and discoveries can be put to work on behalf of the region.”
entertainment from Hollywood to Disneyland. From the earliest days of its southern California theme park, for
To maintain this cutting-edge reputation, however, West Coast sci- instance, the Walt Disney Company has relied on a “shockingly deep
entists and engineers now rely less on the region’s natural resources base of technology” to develop its immersive attractions, said Jon
and more on creative, sometimes counterintuitive, collaborations to Snoddy, head of the Research and Development Studio for Walt
develop new technologies while protecting the environment, speakers Disney Imagineering, in a plenary address at the start of the meeting.
said at the American Association for the Advancement of Science’s But the days when a Disney mechanical engineer might work
Pacific Division meeting, held 12 to 15 June at Cal Poly Pomona. separately from a software developer or even a psychologist are over,
“The Pacific Division conference has long been a showcase for in- replaced by diverse teams that “surround a problem” in unexpected
novative research in fields such as aerospace engineering, materials ways, he said. During the development of the Pirates of the Carib-
Published by AAAS
bean boat ride for the Shanghai Disney resort, he explained, his carbon nanotube materials to biofuels, shows how these diverse
studio needed an interdisciplinary cast of researchers to puzzle out collaborations could help mitigate the environmental impacts of
the challenge of combining massive physical sets with a 50-foot-tall flying, Robinson-Berry said.
video screen featuring images from 28 projectors stitched together The meeting featured several presentations by scientists contend-
using computer vision software, to ensure that visitors would experi- ing with California’s changing climate and environmental pressures.
ence the ride as a “seamless space.” At Cal Poly Pomona, graduate student Sebastian Olarte is manufac-
“I often find that when you put everyone of the same discipline on turing a fluoride-based polymer membrane filter to turn seawater
a problem, you get predictable, interesting progress forward,” said into fresh water, spurred on by the 2011–2014 drought, the worst
Snoddy. “But if you put a diverse mix of people on a problem, you get in the state’s history since 1895. Engineering professor Mikhail
cool leaps, you get things that are non-obvious.” Gershfeld is exploring new technologies for producing laminated
The pool of West Coast research and development talent also has timber and plywood panels that may provide material for commer-
expanded in the 21st century, most notably with the rise of Silicon cial buildings that is earthquake-resilient and has a lower carbon
Valley and the Seattle technology corridor, said plenary speaker footprint than concrete. Shelton Murinda, an animal and veterinary
Joan Robinson-Berry, vice president and general manager of Boeing sciences professor, is working on a handheld device that can test for
South Carolina and Boeing Commercial Airplanes. toxic E. coli bacteria contamination directly in the fields of southern
In the past, Boeing’s market dominance ensured that it could de- California produce farms.
velop future flight technologies at its own pace, Robinson-Berry said. Jeanette Cobian Iñiguez, a Ph.D. student at the University of Cali-
“But the whole world has changed, because we have all these new fornia, Riverside, shared her research using wind tunnels to study the
entrants in this field, like Tesla and Virgin Galactic, so we have to kind spread of wildfire in dry chaparral shrublands. For Cobian Iñiguez,
of disrupt ourselves or be disrupted by these other companies.” the Pacific Division meeting was a good venue to discuss her unusual
Disruption for Boeing, she said, means moving beyond a workforce approach to a common California problem.
Published by AAAS
RESEARCH
Friction between
polyelectrolyte brushes
Yu et al., p. 1434
PSYCHOLOGY
D
o we think that a problem persists even
when it has become less frequent? Levari
et al. show experimentally that when the
“signal” a person is searching for becomes
rare, the person naturally responds by
broadening his or her definition of the signal—
and therefore continues to find it even when it
is not there. From low-level perception of color
to higher-level judgments of ethics, there is a
robust tendency for perceptual and judgmental
standards to “creep” when they ought not to. For
example, when blue dots become rare, partici-
pants start calling purple dots blue, and when
threatening faces become rare, participants start
calling neutral faces threatening. This phenome-
non has broad implications that may help explain
why people whose job is to find and eliminate
problems in the world often cannot tell when their
work is done. —AMS
Science, this issue p. 1465
MOLECULAR BIOLOGY the early steps of spliceosome high fidelity and resolution. In over long time periods in awake,
assembly and activation. —SYM the cortex, two-photon imaging active mice. Both orientation-
Structural basis for Science, this issue p. 1423 with these indicators was used and direction-selective retinal
spliceosome assembly to map dopamine activity at cel- ganglion cells can be monitored,
The spliceosome removes lular resolution. —PRS as can the circadian modulation
noncoding sequences from NEUROSCIENCE Science, this issue p. 1420 of retinal ganglion cell activity.
precursor RNA and ligates —MSL
coding sequences into useful
Imaging dopamine Science, this issue p. 1447
mRNA. The pre-spliceosome release in the brain BIOMEDICAL MATERIALS
(A complex) associates with a Neuromodulator release alters
small nuclear ribonucleopro- the function of target circuits in
Eye can see SOLAR CELLS
CREDITS: (FROM TOP) J. YU ET AL.; C. SMITH/SCIENCE
tein (snRNP) complex called poorly known ways. An essential neural activity Perovskite layers
U4/U6.U5 tri-snRNP to form step to address this knowledge Organisms take up a tremen-
the pre-B complex, which is gap is to measure the dynam- dous amount of information make the grade
converted into the precatalytic ics of neuromodulatory signals through the visual system, which Inverted planar perovskite solar
B complex. Bai et al. solved while simultaneously manipulat- is then processed by the neural cells offer opportunities for
the cryo–electron microscopy ing the elements of the target circuitry. Hong et al. developed a simplified device structure
structures of the pre-B and B circuit during behavior. Patriarchi a mesh electronics implant that compared with conventional
complexes isolated from yeast. et al. developed fluorescent pro- is delivered by injection into mesoporous titanium oxide
These structures show the U1 tein–based dopamine indicators mice retinas. With these devices, interlayers. However, their lower
and U2 snRNPs and allow mod- to visualize spatial and temporal it is possible to obtain record- open-circuit voltages result
eling of the A complex to reveal release of dopamine directly with ings from retinal ganglion cells in lower power conversion
Published by AAAS
efficiencies. Using mixed-cation roles in development. Gonen et Edited by Caroline Ash
lead mixed-halide perovskite and al. find that a single far-upstream IN OTHER JOURNALS and Jesse Smith
a solution-processed secondary 557–base pair element is critical
growth method, Luo et al. cre- for up-regulating Sox9. Without
ated a surface region in the it, XY mice develop as females
perovskite film that inhibited instead of males. The 557–base Damage to one eye
nonradiative charge-carrier pair enhancer is conserved, causes harmful responses
recombination. This kind of solar likely to be relevant to human in the uninjured eye.
cell had comparable perfor- disorders of sex differentiation,
mance to that of conventional and probably essential because
cells. —PDS it acts early in a time-critical
Science, this issue p. 1442 process, and any failure allows
ovary-specific factors to domi-
nate. —BAP
CONDENSED MATTER
Science, this issue p. 1469
Golden ultrafast melting
Understanding fast melting of
metals is important for appli- EVOLUTIONARY BIOLOGY
cations such as welding and
micromachining. However, fast
Human influence on
melting leaves simulation as orangutans
the only option for probing the The numbers of orangutans and
process. Mo et al. performed their geographic distribution
ultrafast electron diffraction declined dramatically after the
experiments on laser-pulsed late Pleistocene. Experts have
gold films. This allowed detailed proposed climate change and
mapping of the melting process, human activities as possible
which proceeds through two dis- causes. Synthesizing available
tinct regimes while the bonding archaeological, genetic, and
behavior changes in unexpected behavioral data, Spehar et al.
ways. The results require adding concluded that over the past
new physical processes to high- 70,000 years, hunting especially
energy melting models. —BG played a role. Some adaptable IMMUNOLOGY
Science, this issue p. 1451 orangutan populations continue
to live in human-dominated A site for sore eyes
environments, which challenges
M
ucosal tolerance arises when exposure to foreign
SEX DETERMINATION the long-held belief that orang- antigens at mucosal sites results in suppressed
utans require pristine habitats.
Sox9 regulation during —PJB
immune responses mediated by regulatory T cells
and tolerogenic antigen-presenting cells (APCs).
sex determination Sci. Adv. 10.1126/
In mammals, immune responses of the retina and
Sex determination is regu- sciadv.1701422 (2018).
cornea in both eyes are interdependent: Damage in one eye
lated by the Sox9 gene. During causes a response in the uninjured eye, too. Using a mouse
testis differentiation, this gene is
T CELL ACTIVATION model of ocular injury, Guzmán et al. show that damage to
directly targeted by the product the conjunctival mucosa in one eye also leads to a loss of
of the Y chromosome–encoded Controlled activation mucosal tolerance in the opposite, undamaged conjunctiva.
gene Sry. The regulatory region At the intestinal barrier, lym-
PHOTOS: (FROM LEFT) GRETA KEENAN, THE FRANCIS CRICK INSTITUTE; JENA ARDELL
TRPV1 channels in the injured eye signal via the central ner-
of Sox9 is complex, which is phocyte activation is a tightly vous system, which leads to the neuropeptide substance P
typical of genes with multiple regulated process that enables being released in the uninjured eye. Consequently, epithelial
rapid responses to pathogens nuclear factor kB signaling and APC maturation direct
but avoids destructive inflam- antigen-specific T cells to an effector phenotype and poten-
mation. Konjar et al. examined tially damaging inflammation in the intact eye. —STS
how intraepithelial lymphocytes
Mucosal Immunol. 10.1038/s41385-018-0040-5 (2018).
(IELs) maintain a controlled
activation state, which is influ-
enced by the composition of
the mitochondrial membrane. CELL BIOLOGY droplets maintained the spatial
Inflammation triggers changes pattern of an inhibitor of droplet
in the mitochondrial mem-
Memories: Just a phase formation long after the inhibi-
Recently, several proteins have
branes of IELs, particularly the tor had been removed. Despite
cardiolipin composition, and been shown to phase-separate this persistence, individual
these changes support rapid into liquid droplets within the droplets were highly dynamic,
proliferation and effector func- cell. Dine et al. found that such continuously exchanging their
If Sox9 is not up-regulated, XY mice tions. —CNF protein droplets exhibit a robust constituents with the cytosolic
develop as females. Sci. Immunol. 3, eaan2543 (2018). form of spatial memory. The phase. The authors exploited
EDUCATION
Sibon nebulatus, one of a
tribe of diverse South and
A CURE for undergraduate
Central American snakes research
Course-based undergraduate
research experiences (CUREs)
are designed to engage an
entire class in a research ques-
tion within the context of the
course itself. Current research
suggests that five distinct core
components come together
to define a CURE. Ballen et al.
used a backward-elimination
experimental design to test the
importance of two CURE com-
ponents for non–biology majors:
experience of discovery and the
production of data. They did
not find significant impacts of
either component on nonmajors’
academic performance, science
self-efficacy, sense of project
ownership, or perceived value
of the CURE. These findings
HERPETOLOGY
challenge the current definition
Snail-snacking snakes of what constitutes a CURE and
suggest future studies aimed
at understanding why different
laboratory environments can
D
ipsadini are snakes that consume snails by sucking them out of their shells. They constitute
be effective for both major and
a diverse tribe of more than 70 recognized species of tree-living snakes with striking skin
nonmajor populations. —MMc
patterns. Arteaga et al. have sampled nuclear and mitochondrial genes from a new collec-
J. Microbiol. Biol. Educ. 10.1128/
tion of snakes from Central and South America to reexamine their taxonomy. It appears that
jmbe.v19i2.1515 (2018).
these snakes’ specialized lifestyle has resulted in a large adaptive radiation. Unfortunately,
the five species newly discovered in this analysis are all vulnerable or endangered owing to habitat
fragmentation. —CA
SURFACE SCIENCE
Zookeys 10.3897/zookeys.766.24523 (2018).
Tipping the vibrational
spectrum
their system to drive persistent, in vitro. Surprisingly, they found nonsilicon integrated electronic Scanning tunneling micros-
local regulation of cytoskeletal that male cells also undergo XCI, and photonic devices. However, copy can be used to measure
activity via dynamic clusters of albeit transiently, at the onset methods for growing devices the vibrational spectrum of
receptor tyrosine kinases. Thus, of stem cell differentiation. In on amorphous and nonepitaxial adsorbed molecules as loss fea-
protein phase separation may addition, both X chromosomes substrates are limited. Sarkar et tures in the inelastic tunneling
underlie the persistent polariza- are ephemerally inactivated in al. overcome this by using stand- of electrons. Okabayashi et al.
tion observed in many cellular female cells before one X is ran- ing evaporation or sputtering explored perturbations caused
and developmental processes. domly inactivated permanently. techniques to deposit a metal, by the close proximity of the
—SMH Although it remains to be shown such as indium, with a capping microscope tip to the adsorbed
Cell Syst. 10.1016/ in vivo, these results suggest oxide layer. The metal is heated molecule—in this case, CO
j.cels.2018.05.002 (2018). gender-independent XCI initia- in a hydrogen environment, and adsorbed on the Cu(111) surface.
tion and additional, unknown a precursor is added to convert From a series of force-sensing
PHOTO: SIBONS PHOTOGRAPHY/ALAMY STOCK PHOTO
female-specific mechanisms to the metal to the desired target, and frequency-shift scans, they
MOLECULAR BIOLOGY maintain XCI. —SYM such as InP, under conditions could determine the changes in
where only a single nucleation frequency arising from the force
Inactivating sex Cell Stem Cell 10.1016/
j.stem.2018.05.001 (2018). site forms in each patterned site. exerted by the tip by modeling
chromosomes The versatility of the method the surface, molecule, and tip
To compensate for sex chro- is demonstrated through the as a mechanical system. The
mosome dosage, XX females MATERIALS SCIENCE growth of InP, GaP, InAs, InGaP, tip weakened and lengthened
undergo epigenetic inactivation SnP, and Sn4P3 crystals directly the C–O bond and shifted the
(XCI) of one X chromosome.
Growing compound on SiO2, Si3N4, TiO2, Al2O3, Gd2O3, frustrated translational mode of
Using a mouse embryonic stem semiconductors SrTiO3, and graphene. —MSL CO to higher energies. —PDS
cell culture system, Sousa et al. Epitaxial growth is the main ACS Nano 10.1021/ Proc. Natl. Acad. Sci. U.S.A. 10.1073/
investigated this process technique used for depositing acsnano.8b01819 (2018). pnas.1721498115 (2018).
Published by AAAS
RESEARCH
Published by AAAS
Cretaceous-Paleogene bound- governance and education to
ary, also initiated a long period drive research and practice in
of strong global warming. Using a more sustainable direction.
data from phosphatic micro- —JFU
fossils, including fish teeth, Science, this issue p. 1396
scales, and bone, MacLeod et
al. estimated global average
MALARIA
temperature. Immediately after
the asteroid strike, temperatures Uncomplicating malaria
increased by ~5°C and remained Severe malaria is caused by
high for about 100,000 years the parasite Plasmodium
(see the Perspective by Lécuyer). falciparum. Infections can
These results are relevant to result in organ failure and life-
current climate projections, threatening hematological or
because the Chicxulub impact metabolic abnormalities. Lee
perturbed Earth systems on time et al. sequenced patient and
scales even shorter than the cur- parasite transcriptomes from 46
rent rate of change. —HJS P. falciparum–infected Gambian
Science, this issue p. 1467; children to better understand
see also p. 1400 host-pathogen interactions.
The immune response in severe
malaria, compared with that in
NEUROSCIENCE uncomplicated malaria, was not
necessarily dysregulated but
Behavior with movement instead reflected high parasite
The neuronal circuits required loads, although there was a dis-
for movement reside in the tinct neutrophil response. —CAC
spinal cord. But how does the Sci. Transl. Med. 10, eaar3619 (2018).
nervous system coordinate
multiple neuronal populations
across different brain regions to SEPSIS
fulfil an organism’s behavioral
needs? In a Perspective, Arber
Inflammatory decoy
and Costa explore the orga- control
nizational logic of specialized Bacterial infection can lead to
spinal microcircuits, sensory sepsis, inflammation, and death.
feedback loops, and brain motor Li et al. found that the long
commands. How action choice noncoding RNA MEG3-4 and the
occurs, however, remains a mRNA encoding the proinflam-
mystery. —GKA matory cytokine interleukin-1b
Science, this issue p. 1403 (IL-1b) competitively bound to
the microRNA miR-138 in the
lungs of bacterially infected
MATERIALS mice. Initially, MEG3-4 binding to
miR-138 facilitated IL-1b produc-
A more sustainable tion, but it ultimately shut down
materials system IL-1b–dependent inflammation.
Large-scale use and consump- Lung-specific overexpression
tion of materials are central to of MEG3-4 prolonged infection
modern lifestyles but increas- and exacerbated inflammation
ingly cause environmental and lung injury in mice, whereas
problems. In a Perspective, intravenously delivering miR-
Olivetti and Cullen outline the 138 mimics to infected mice
impacts and indicate that a more enhanced their survival. —LKF
sustainable materials system Sci. Signal. 11, eaao2387 (2018).
can be achieved by reducing
consumption and changing
technology. Economic incentives
alone will not be sufficient to
achieve the required changes.
In addition, lifetime extension,
higher manufacturing efficiency,
and recovery are essential. Such
reforms must be supported by
Steven J. Davis*, Nathan S. Lewis*, Matthew Shaner, Sonia Aggarwal, Doug Arent, carbon from the atmo- ON OUR WEBSITE
Inês L. Azevedo, Sally M. Benson, Thomas Bradley, Jack Brouwer, Yet-Ming Chiang, sphere (carbon manage- Read the full article
Christopher T. M. Clack, Armond Cohen, Stephen Doig, Jae Edmonds, Paul Fennell, ment) is also likely to be at http://dx.doi.
Christopher B. Field, Bryan Hannegan, Bri-Mathias Hodge, Martin I. Hoffert, an important activity of org/10.1126/
Eric Ingersoll, Paulina Jaramillo, Klaus S. Lackner, Katharine J. Mach, any net-zero emissions science.aas9793
..................................................
Michael Mastrandrea, Joan Ogden, Per F. Peterson, Daniel L. Sanchez, energy system. The spe-
Daniel Sperling, Joseph Stagner, Jessika E. Trancik, Chi-Jen Yang, Ken Caldeira* cific technologies that will be favored in
future marketplaces are largely uncertain,
BACKGROUND: Net emissions of CO 2 by search, development, demonstration, and de- but only a finite number of technology choices
human activities—including not only en- ployment. It may take decades to research, exist today for each functional role. To take
ergy services and industrial production but develop, and deploy these new technologies. appropriate actions in the near term, it is
also land use and agriculture—must ap- imperative to clearly identify desired end
Net-zero emissions energy systems logical opportunities and barriers for eliminat-
ing and/or managing emissions related to the
difficult-to-decarbonize services; pitfalls in which
Steven J. Davis1,2*, Nathan S. Lewis3*, Matthew Shaner4, Sonia Aggarwal5, near-term actions may make it more difficult or
Doug Arent6,7, Inês L. Azevedo8, Sally M. Benson9,10,11, Thomas Bradley12, costly to achieve the net-zero emissions goal;
Jack Brouwer13,14, Yet-Ming Chiang15, Christopher T. M. Clack16, Armond Cohen17, and critical areas for research, development,
Stephen Doig18, Jae Edmonds19, Paul Fennell20,21, Christopher B. Field22, demonstration, and deployment. Our scope is
Bryan Hannegan23, Bri-Mathias Hodge6,24,25, Martin I. Hoffert26, Eric Ingersoll27, not comprehensive; we focus on what now seem
the most promising technologies and pathways.
Paulina Jaramillo8, Klaus S. Lackner28, Katharine J. Mach29, Michael Mastrandrea4,
Our assertions regarding feasibility throughout
Joan Ogden30, Per F. Peterson31, Daniel L. Sanchez32, Daniel Sperling33,
are not the result of formal, quantitative econo-
Joseph Stagner34, Jessika E. Trancik35,36, Chi-Jen Yang37, Ken Caldeira32*
mic modeling; rather, they are based on compar-
ison of current and projected costs, with stated
Some energy services and industrial processes—such as long-distance freight transport,
assumptions about progress and policy.
air travel, highly reliable electricity, and steel and cement manufacturing—are particularly
A major conclusion is that it is vital to integrate
difficult to provide without adding carbon dioxide (CO2) to the atmosphere. Rapidly
currently discrete energy sectors and industrial
P
of global CO2 emissions from fossil fuel com-
eople do not want energy itself, but rather renewable sources (including nuclear energy bustion and industry sources [estimated by
the services that energy provides and the and fossil fuels with carbon capture and storage). using (7–9)]. Similarly long trips in light-duty
products that rely on these services. Even However, other energy services essential to mo- vehicles accounted for an additional 40 Mt CO2,
with substantial improvements in efficiency, dern civilization entail emissions that are likely and aviation and other shipping modes (such
global demand for energy is projected to to be more difficult to fully eliminate. These as trains and ships) emitted 830 and 1060 Mt
increase markedly over this century (1). Mean- difficult-to-decarbonize energy services include CO2, respectively. Altogether, these sources were
while, net emissions of carbon dioxide (CO2) from aviation, long-distance transport, and shipping; responsible for ~6% of global CO2 emissions
human activities—including not only energy production of carbon-intensive structural materi- (Fig. 2). Meanwhile, both global energy demand
and industrial production, but also land use and als such as steel and cement; and provision of for transportation and the ratio of heavy- to
agriculture—must approach zero to stabilize glo- a reliable electricity supply that meets varying light-duty vehicles is expected to increase (9).
bal mean temperature (2, 3). Indeed, interna- demand. To the extent that carbon remains in- Light-duty vehicles can be electrified or run
tional climate targets, such as avoiding more volved in these services in the future, net-zero on hydrogen without drastic changes in perfor-
than 2°C of mean warming, are likely to require emissions will also entail active management mance except for range and/or refueling time.
an energy system with net-zero (or net-negative) of carbon. By contrast, general-use air transportation and
emissions later this century (Fig. 1) (3). In 2014, difficult-to-eliminate emissions related long-distance transportation, especially by trucks
Energy services such as light-duty transpor- to aviation, long-distance transportation, and or ships, have additional constraints of revenue
tation, heating, cooling, and lighting may be shipping; structural materials; and highly reliable cargo space and payload capacity that mandate
relatively straightforward to decarbonize by electricity totaled ~9.2 Gt CO2, or 27% of global energy sources with high volumetric and grav-
electrifying and generating electricity from var- CO2 emissions from all fossil fuel and industrial imetric density (10). Closed-cycle electrochemical
iable renewable energy sources (such as wind sources (Fig. 2). Yet despite their importance, batteries must contain all of their reactants and
and solar) and dispatchable (“on-demand”) non- detailed representation of these services in in- products. Hence, fuels that are oxidized with
1
Department of Earth System Science, University of California, Irvine, Irvine, CA, USA. 2Department of Civil and Environmental Engineering, University of California, Irvine, Irvine, CA, USA.
3
Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, CA, USA. 4Near Zero, Carnegie Institution for Science, Stanford, CA, USA. 5Energy Innovation, San
Francisco, CA, USA. 6National Renewable Energy Laboratory, Golden, CO, USA. 7Joint Institute for Strategic Energy Analysis, Golden, CO, USA. 8Engineering and Public Policy, Carnegie Mellon
University, Pittsburgh, PA, USA. 9Global Climate and Energy Project, Stanford University, Stanford, CA, USA. 10Precourt Institute for Energy, Stanford University, Stanford, CA, USA. 11Department
of Energy Resource Engineering, Stanford University, Stanford, CA, USA. 12Department of Mechanical Engineering, Colorado State University, Fort Collins, CO, USA. 13Department of Mechanical
and Aerospace Engineering, University of California, Irvine, Irvine, CA, USA. 14Advanced Power and Energy Program, University of California, Irvine, CA, USA. 15Department of Material Science and
Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA. 16Vibrant Clean Energy, Boulder, CO, USA. 17Clean Air Task Force, Boston, MA, USA. 18Rocky Mountain Institute,
Boulder, CO, USA. 19Pacific National Northwestern Laboratory, College Park, MD, USA. 20Department of Chemical Engineering, South Kensington Campus, Imperial College London, London, UK.
21
Joint Bioenergy Institute, 5885 Hollis Street, Emeryville, CA, USA. 22Woods Institute for the Environment, Stanford University, Stanford, CA, USA. 23Holy Cross Energy, Glenwood Springs, CO,
USA. 24Department of Electrical, Computer, and Energy Engineering, University of Colorado Boulder, Boulder, CO, USA. 25Department of Chemical and Biological Engineering, Colorado School of
Mines, Golden, CO, USA. 26Department of Physics, New York University, New York, NY, USA. 27Lucid Strategy, Cambridge, MA, USA. 28The Center for Negative Carbon Emissions, Arizona State
University, Tempe, AZ, USA. 29Department of Earth System Science, Stanford University, Stanford, CA, USA. 30Environmental Science and Policy, University of California, Davis, Davis, CA, USA.
31
Department of Nuclear Engineering, University of California, Berkeley, Berkeley, CA, USA. 32Department of Global Ecology, Carnegie Institution for Science, Stanford, CA, USA. 33Institute of
Transportation Studies, University of California, Davis, Davis, CA, USA. 34Department of Sustainability and Energy Management, Stanford University, Stanford, CA, USA. 35Institute for Data,
Systems, and Society, Massachusetts Institute of Technology, Cambridge, MA, USA. 36Santa Fe Institute, Santa Fe, NM, USA. 37Independent researcher.
*Corresponding authors: Email: sjdavis@uci.edu (S.J.D.); nslewis@caltech.edu (N.S.L.); kcaldeira@carnegiescience.edu (K.C.)
ambient air and then vent their exhaust to the range, heavy-duty trucks powered by current and volumetric energy density likely preclude
atmosphere have a substantial chemical advan- lithium-ion batteries and electric motors can car- battery- or hydrogen-powered aircraft for long-
tage in gravimetric energy density. ry ~40% less goods than can trucks powered distance cargo or passenger service (12). Auto-
Battery- and hydrogen-powered trucks are now by diesel-fueled, internal combustion engines. nomous trucks and distributed manufacturing
used in short-distance trucking (11), but at equal The same physical constraints of gravimetric may fundamentally alter the energy demands of
Fig. 1. Schematic of an integrated system that can provide mission; blue, hydrogen production and transport; purple,
essential energy services without adding any CO 2 to the atmo- hydrocarbon production and transport; orange, ammonia production
sphere. (A to S) Colors indicate the dominant role of specific and transport; red, carbon management; and black, end uses of
technologies and processes. Green, electricity generation and trans- energy and materials.
Table 1. Key energy carriers and the processes for interconversion. Processes listed in each cell convert the row energy carrier to the column energy
carrier. Further details about costs and efficiencies of these interconversions are available in the supplementary materials.
To
–
From e H2 CxOyHz NH3
............................................................................................................................................................................................................................................................................................................................................
e– Electrolysis ($5 to 6/kg H ) 2 Electrolysis + methanation Electrolysis + Haber-Bosch
........................................................................................................................................................................................................................................................................................................................
Electrolysis + Fischer-Tropsch
............................................................................................................................................................................................................................................................................................................................................
H2 Combustion Methanation Haber-Bosch ($0.50 to
($0.07 to 0.57/m3 CH ) 4 0.60/kg NH ) 3
........................................................................................................................................................................................................................................................................................................................
Oxidation via fuel cell Fischer-Tropsch ($4.40
to $15.00/gallon of
gasoline-equivalent)
............................................................................................................................................................................................................................................................................................................................................
CxOyHz Combustion Steam reforming Steam reforming +
($1.29 to 1.50/kg H ) 2 Haber-Bosch
........................................................................................................................................................................................................................................................................................................................
Biomass gasification
($4.80 to 5.40/kg H ) 2
............................................................................................................................................................................................................................................................................................................................................
NH3 Combustion Metal catalysts Metal catalysts + methanation/
(~$3/kg H 2 ) Fischer-Tropsch
........................................................................................................................................................................................................................................................................................................................
the freight industry, but if available, energy-dense may be directly used in an engine or may be portation fuels at costs roughly competitive with
liquid fuels are likely to remain the preferred cracked to produce hydrogen. Its thermolysis gasoline (for example, U.S. $19/GJ or U.S. $1.51/
energy source for long-distance transportation must be carefully controlled so as to minimize gallon of ethanol) (22). As technology matures
services (13). production of highly oxidized products such as and overall decarbonization efforts of the energy
Options for such energy-dense liquid fuels in- NOx (17). Furthermore, like hydrogen, ammo- system proceed, biofuels may be able to largely
clude the hydrocarbons we now use, as well as nia’s gravimetric energy density is considerably avoid fossil fuel inputs such as those related to
hydrogen, ammonia, and alcohols and ethers. lower than that of hydrocarbons such as diesel on-farm processes and transport, as well as emis-
In each case, there are options for producing (Fig. 3A). sions associated with induced land-use change
carbon-neutral or low-carbon fuels that could (23, 24). The extent to which biomass will supply
be integrated to a net-zero emissions energy Biofuels liquid fuels in a future net-zero emissions energy
system (Fig. 1), and each can also be intercon- Conversion of biomass currently provides the system thus depends on advances in conversion
verted through existing thermochemical processes most cost-effective pathway to nonfossil, carbon- technology, competing demands for bioenergy
(Table 1). containing liquid fuels. Liquid biofuels at present and land, the feasibility of other sources of carbon-
represent ~4.2 EJ of the roughly 100 EJ of energy neutral fuels, and integration of biomass produc-
Hydrogen and ammonia fuels consumed by the transport sector worldwide. tion with other objectives (25).
The low volumetric energy density of hydrogen Currently, the main liquid biofuels are ethanol
favors transport and storage at low temperatures from grain and sugar cane and biodiesel and re- Synthetic hydrocarbons
(–253°C for liquid hydrogen at atmospheric pres- newable diesel from oil seeds and waste oils. Liquid hydrocarbons can also be synthesized
sure) and/or high pressures (350 to 700 bar), They are associated with substantial challenges through industrial hydrogenation of feedstock
thus requiring heavy and bulky storage contain- related to their life-cycle carbon emissions, cost, carbon, such as the reaction of carbon monoxide
ers (14). To contain the same total energy as a and scalability (18). and hydrogen by the Fischer-Tropsch process
diesel fuel storage system, a liquid hydrogen Photosynthesis converts <5% of incident ra- (26). If the carbon contained in the feedstock
storage system would weigh roughly six times diation to chemical energy, and only a fraction is taken from the atmosphere and no fossil en-
more and be about eight times larger (Fig. 3A). of that chemical energy remains in biomass (19). ergy is used for the production, processing, and
However, hydrogen fuel cell or hybrid hydrogen- Conversion of biomass to fuel also requires en- transport of feedstocks and synthesized fuels,
battery trucks can be more energy efficient than ergy for processing and transportation. Land the resulting hydrocarbons would be carbon-
those with internal combustion diesel engines used to produce biofuels must have water, nu- neutral (Fig. 1). For example, emissions-free elec-
(15), requiring less onboard energy storage to trient, soil, and climate characteristics suitable tricity could be used to produce dihydrogen (H2)
achieve the same traveling range. Toyota has for agriculture, thus putting biofuels in competi- by means of electrolysis of water, which would
recently introduced a heavy-duty (36,000 kg), tion with other land uses. This has implications be reacted with CO2 removed from the atmo-
500-kW fuel cell/battery hybrid truck designed for food security, sustainable rural economies, and sphere either through direct air capture or photo-
to travel 200 miles on liquid hydrogen and stored the protection of nature and ecosystem services synthesis (which in the latter case could include
electricity, and Nikola has announced a similar (20). Potential land-use competition is heightened CO2 captured from the exhaust of biomass or
battery/fuel cell heavy-duty truck with a claimed by increasing interest in bioenergy with carbon biogas combustion) (27, 28).
range of 1300 to 1900 km, which is comparable capture and storage (BECCS) as a source of nega- At present, the cost of electrolysis is a major
with today’s long-haul diesel trucks (16). If hy- tive emissions (that is, carbon dioxide removal), barrier. This cost includes both the capital costs
drogen can be produced affordably without CO2 which biofuels can provide (21). of electrolyzers and the cost of emissions-free
emissions, its use in the transport sector could Advanced biofuel efforts include processes that electricity; 60 to 70% of current electrolytic hy-
ultimately be bolstered by the fuel’s importance seek to overcome the recalcitrance of cellulose to drogen cost is electricity (Fig. 3C) (28, 29). The
in providing other energy services. allow use of different feedstocks (such as woody cheapest and most mature electrolysis technology
Ammonia is another technologically viable crops, agricultural residues, and wastes) in order available today uses alkaline electrolytes [such as
alternative fuel that contains no carbon and to achieve large-scale production of liquid trans- potassium hydroxide (KOH) or sodium hydroxide
Fig. 2. Difficult-to-eliminate
emissions in current context.
(A and B) Estimates of CO2
emissions related to different
energy services, highlighting
[for example, by longer pie
pieces in (A)] those services
that will be the most difficult
to decarbonize, and the
magnitude of 2014 emissions
from those difficult-to-
eliminate emissions. The
shares and emissions shown
here reflect a global energy
system that still relies
primarily on fossil fuels and
that serves many developing
regions. Both (A) the shares
and (B) the level of emissions
related to these difficult-to-
decarbonize services are
(NaOH)] together with metal catalysts to pro- of the relative simplicity of large, long-term frastructure. Between 2000 and 2015, cement and
duce hydrogen at an efficiency of 50 to 60% and storage of chemical fuels. Hence, using emissions- steel use persistently averaged 50 and 21 tons per
a cost of ~U.S. $5.50/kg H2 (assuming industrial free electricity to make fuels represents a critical million dollars of global GDP, respectively (~1 kg
electricity costs of U.S. $0.07/kWh and 75% uti- opportunity for integrating electricity and trans- per person per day in developed countries) (4).
lization rates) (29, 30). At this cost of hydrogen, portation systems in order to supply a persistent Globally, ~1320 and 1740 Mt CO2 emissions em-
the minimum price of synthesized hydrocarbons demand for carbon-neutral fuels while boosting anated from chemical reactions involved with the
would be $1.50 to $1.70/liter of diesel equivalent utilization rates of system assets. manufacture of cement and steel, respectively
[or $5.50 to $6.50/gallon and $42 to $50 per GJ, (Fig. 2) (8, 38, 39); altogether, this equates to
assuming carbon feedstock costs of $0 to 100 per Direct solar fuels ~9% of global CO2 emissions in 2014 (Fig. 1,
ton of CO2 and very low process costs of $0.05/ Photoelectrochemical cells or particulate/molecular purple and blue). Although materials intensity
liter or $1.50 per GJ (28)]. For comparison, H2 photocatalysts directly split water by using sunlight of construction could be substantially reduced
from steam reforming of fossil CH4 into CO2 and to produce fuel through artificial photosynthesis, (40, 41), steel demand is projected to grow by 3.3%
H2 currently costs $1.30 to 1.50 per kg (Fig. 3D, without the land-use constraints associated with per year to 2.4 billion tons in 2025 (42), and ce-
red line) (29, 31). Thus, the feasibility of syn- biomass (35). Hydrogen production efficiencies ment production is projected to grow by 0.8 to
thesizing hydrocarbons from electrolytic H2 may can be high, but costs, capacity factors, and life- 1.2% per year to 3.7 billion to 4.4 billion tons in
depend on demonstrating valuable cross-sector times need to be improved in order to obtain an 2050 (43, 44), continuing historical patterns of
benefits, such as balancing variability of renew- integrated, cost-advantaged approach to carbon- infrastructure accumulation and materials use seen
able electricity generation, or else a policy-imposed neutral fuel production (36). Short-lived labora- in regions such as China, India, and Africa (4).
price of ~$400 per ton of CO2 emitted (which tory demonstrations have also produced liquid Decarbonizing the provision of cement and
would also raise fossil diesel prices by ~$1.00/liter carbon-containing fuels by using concentrated steel will require major changes in manufac-
or ~$4.00/gallon). CO2 streams (Fig. 1H) (37), in some cases by turing processes, use of alternative materials
In the absence of policies or cross-sector coor- using bacteria as catalysts. that do not emit CO2 during manufacture, or
dination, hydrogen costs of $2.00/kg (approaching carbon capture and storage (CCS) technologies
the cost of fossil-derived hydrogen and synthe- Outlook to minimize the release of process-related CO2
sized diesel of ~$0.79/liter or $3.00/gallon) could Large-scale production of carbon-neutral and to the atmosphere (Fig. 1B) (45).
be achieved, for example, if electricity costs were energy-dense liquid fuels may be critical to achiev-
$0.03/kWh and current electrolyzer costs were ing a net-zero emissions energy system. Such fuels Steel
reduced by 60 to 80% (Fig. 3B) (29). Such reduc- could provide a highly advantageous bridge be- During steel making, carbon (coke from coking
tions may be possible (32) but may require central- tween the stationary and transportation energy pro- coal) is used to reduce iron oxide ore in blast
ized electrolysis (33) and using less mature but duction sectors and may therefore deserve special furnaces, producing 1.6 to 3.1 tons of process
promising technologies, such as high-temperature priority in energy research and development efforts. CO2 per ton of crude steel produced (39). This
solid oxide or molten carbonate fuel cells, or is in addition to CO2 emissions from fossil fuels
thermochemical water splitting (30, 34). Fuel Structural materials burned to generate the necessary high temper-
markets are vastly more flexible than instan- Economic development and industrialization atures (1100 to 1500°C). Reductions in CO2 emis-
taneously balanced electricity markets because are historically linked to the construction of in- sions per ton of crude steel are possible through
the use of electric arc furnace (EAF) “minimills” are high (40 to 50% and 35% by volume, re- as steel and cement, or close substitutes, without
that operate by using emissions-free electricity, spectively) (Fig. 1, G and E) (51, 52). adding CO2 to the atmosphere. Although alter-
efficiency improvements (such as top gas recovery), native processes might avoid liberation and use
new process methods (such as “ultra-low CO2 Cement of carbon, the cement and steel industries are
direct reduction,” ULCORED), process heat fuel- About 40% of the CO2 emissions during cement especially averse to the risk of compromising the
switching, and decreased demand via better production are from fossil energy inputs, with the mechanical properties of produced materials.
engineering. For example, a global switch to remaining CO2 emissions arising from the calcina- Demonstration and testing of such alternatives
ultrahigh-strength steel for vehicles would avoid tion of calcium carbonate (CaCO3) (typically lime- at scale is therefore potentially valuable. Unless
~160 Mt CO2 annually. The availability of scrap stone) (53). Eliminating the process emissions and until such alternatives are proven, eliminating
steel feedstocks currently constrains EAF pro- requires fundamental changes to the cement- emissions related to steel and cement will de-
duction to ~30% of global demand (46, 47), and making process and cement materials and/or pend on CCS.
the other improvements reduce—but do not installation of carbon-capture technology (Fig. 1G)
eliminate—emissions. (54). CO2 concentrations are typically ~30% by Highly reliable electricity
Prominent alternative reductants include char- volume in cement plant flue gas [compared with Modern economies demand highly reliable elec-
coal (biomass-derived carbon) and hydrogen. ~10 to 15% in power plant flue gas (54)], improv- tricity; for example, demand must be met >99.9%
Charcoal was used until the 18th century, and the ing the viability of post-combustion carbon cap- of the time (Fig. 1A). This requires investment in
Brazilian steel sector has increasingly substituted ture. Firing the kiln with oxygen and recycled CO2 energy generation or storage assets that will be
charcoal for coal in order to reduce fossil CO2 is another option (55), but it may be challenging used a small percentage of the time, when demand
emissions (48). However, the ~0.6 tons of char- to manage the composition of gases in existing is high relative to variable or baseload generation.
coal needed per ton of steel produced require cement kilns that are not gas-tight, operate at As the share of renewable electricity has grown
0.1 to 0.3 ha of Brazilian eucalyptus plantation very high temperatures (~1500°C), and rotate (56). in the United States, natural gas-fired generators
(48, 49). Hundreds of millions of hectares of A substantial fraction of process CO2 emis- have increasingly been used to provide generat-
the flexibility, scalability, and low capital costs Energy storage stationary batteries. Not shown in Fig. 3D, less-
of electricity that can currently be provided by Reliable electricity could also be achieved through efficient (for example, 70% round-trip) batteries
natural gas–fired generators—but without emit- energy storage technologies. The value of today’s based on abundant materials such as sulfur might
ting fossil CO2. This might be accomplished by a energy storage is currently greatest when frequent reduce capital cost per unit energy capacity to
mix of flexible generation, energy storage, and cycling is required, such as for minute-to-minute $8/kWh (with a power capacity cost of $150/kW),
demand management. frequency regulation or price arbitrage (72). Cost- leading to a levelized cost of discharged electri-
effectively storing and discharging much larger city for the grid-scale use case in the range of
Flexible generation $0.06 to 0.09/kWh ($17 to 25 per GJ), assuming
quantities of energy over consecutive days and less
Even when spanning large geographical areas, frequent cycling may favor a different set of 20 to 100 cycles per year over 20 years (81).
a system in which variable energy from wind innovative technologies, policies, and valuation Utilization rates might be increased if elec-
and solar are major sources of electricity will (72, 73). tric vehicle batteries were used to support the
have occasional but substantial and long-term electrical grid [vehicle-to-grid (V2G)], presuming
mismatches between supply and demand. For Chemical bonds that the disruption to vehicle owners from dim-
example, such gaps in the United States are Chemical storage of energy in gas or liquid fuels inished battery charge would be less costly than
commonly tens of petajoules (40 PJ = 10.8 TWh = is a key option for achieving an integrated net- an outage would be to electricity consumers (82).
24 hours of mean U.S. electricity demand in 2015) zero emissions energy system (Table 1). Stored For example, if all of the ~150 million light-duty
and span multiple days, or even weeks (61). Thus, electrolytic hydrogen can be converted back to vehicles in the United States were electrified,
even with continental-scale or global electricity electricity either in fuel cells or through com- 10% of each battery’s 100 kWh charge would
interconnections (61–63), highly reliable electricity bustion in gas turbines [power-to-gas-to-power provide 1.5 TWh, which is commensurate with
in such a system will require either very sub- (P2G2P)] (Figs. 1, F and P, and 3D, red curve); ~3 hours of the country’s average ~0.5 TW power
stantial amounts of dispatchable electricity sources commercial-scale P2G2P systems currently exhibit demand. It is also not yet clear how owners
(either generators or stored energy) that operate a round-trip efficiency (energy out divided by would be compensated for the long-term impacts
Thermal energy Capture and storage will be distinct carbon cooperation among regulators and disparate, risk-
Thermal storage systems are based on sensible management services in a net-zero emissions averse businesses. We thus suggest two parallel
heat (such as in water tanks, building envelopes, energy system (for example, Fig. 1, E and J). broad streams of R&D effort: (i) research in
molten salt, or solid materials such as bricks and Carbon captured from the ambient air could be technologies and processes that can provide these
gravel), latent heat (such as solid-solid or solid- used to synthesize carbon-neutral hydrocarbon difficult-to-decarbonize energy services, and (ii)
liquid transformations of phase-change materials), fuels or sequestered to produce negative emis- research in systems integration that would allow
or thermochemical reactions. Sensible heat storage sions. Carbon captured from combustion of bio- for the provision of these services and products
systems are characterized by low energy densities mass or synthesized hydrocarbons could be in a reliable and cost-effective way.
[36 to 180 kJ/kg or 10 to 50 watt-hour thermal recycled to produce more fuels (98). Storage of We have focused on provision of energy ser-
(Whth)/kg] and high costs (84, 87, 88). Future captured CO2 (for example, underground) will vices without adding CO2 to the atmosphere.
cost targets are <$15/kWhth (89). Thermal stor- be required to the extent that uses of fossil car- However, many of the challenges discussed here
age is well suited to within-day shifting of heat- bon persist and/or that negative emissions are could be reduced by moderating demand, such
ing and cooling loads, whereas low efficiency, needed (20). as through substantial improvements in energy
heat losses, and physical size are key barriers to For industrial CO2 capture, research and de- and materials efficiency. Particularly crucial are
filling week-long, large-scale (for example, 30% of velopment are needed to reduce the capital costs the rate and intensity of economic growth in
daily demand) shortfalls in electricity generation. and costs related to energy for gas separation developing countries and the degree to which
and compression (99). Future constraints on such growth can avoid fossil-fuel energy while
Demand management land, water, and food resources may limit bio- prioritizing human development, environmental
Technologies that allow electricity demand to be logically mediated capture (20). The main chal- protection, sustainability, and social equity
shifted in time (load-shifting or load-shaping) or lenges to direct air capture include costs to (4, 107, 108). Furthermore, many energy services
curtailed to better correlate with supply would manufacture sorbents and structures, energize rely on long-lived infrastructure and systems so
improve overall system reliability while reducing the process, and handle and transport the cap- that current investment decisions may lock in
3. J. Rogelj et al., Zero emission targets as long-term global 27. F. S. Zeman, D. W. Keith, Carbon neutral hydrocarbons. manufacture and CO2 abatement using the calcium looping
goals for climate protection. Environ. Res. Lett. 10, 105007 Philos. Trans. A Math Phys. Eng. Sci. 366, 3901–3918 (2008). cycle. Energy Environ. Sci. 4, 2050–2053 (2011).
(2015). doi: 10.1088/1748-9326/10/10/105007 doi: 10.1098/rsta.2008.0143; pmid: 18757281 doi: 10.1039/c1ee01282g
4. J. C. Steckel, R. J. Brecha, M. Jakob, J. Strefler, 28. C. Graves, S. D. Ebbesen, M. Mogensen, K. S. Lackner, 54. D. Barker et al., “CO2 capture in the cement industry”
G. Luderer, Development without energy? Assessing future Sustainable hydrocarbon fuels by recycling CO2 and H2O with (IEA Greenhouse as R&D Programme, 2008).
scenarios of energy consumption in developing countries. renewable or nuclear energy. Renew. Sustain. Energy Rev. 15, 55. F. S. Zeman, K. S. Lackner, The zero emission kiln. Int.
Ecol. Econ. 90, 53–67 (2013). doi: 10.1016/ 1–23 (2011). doi: 10.1016/j.rser.2010.07.014 Cement Rev. 2006, 55–58 (2006).
j.ecolecon.2013.02.006 29. M. R. Shaner, H. A. Atwater, N. S. Lewis, E. W. McFarland, A 56. L. Zheng, T. P. Hills, P. Fennell, Phase evolution,
5. S. Collins et al., Integrating short term variations of the power comparative technoeconomic analysis of renewable hydrogen characterisation, and performance of cement prepared in an
system into integrated energy system models: A production using solar energy. Energy Environ. Sci. 9, oxy-fuel atmosphere. Faraday Discuss. 192, 113–124
methodological review. Renew. Sustain. Energy Rev. 76, 2354–2371 (2016). doi: 10.1039/C5EE02573G (2016). doi: 10.1039/C6FD00032Kpmid: 27477884
839–856 (2017). doi: 10.1016/j.rser.2017.03.090 30. J. D. Holladay, J. Hu, D. L. King, Y. Wang, An overview of 57. F. Xi et al., Substantial global carbon uptake by cement
6. S. Yeh et al., Detailed assessment of global transport-energy hydrogen production technologies. Catal. Today 139, carbonation. Nat. Geosci. 9, 880–883 (2016). doi: 10.1038/
models’ structures and projections. Transp. Res. Part D 244–260 (2009). doi: 10.1016/j.cattod.2008.08.039 ngeo2840
Transp. Environ. 55, 294–309 (2017). doi: 10.1016/ 31. U.S. Department of Energy (DOE), H2A (Hydrogen Analysis) 58. M. Jarre, M. Noussan, A. Poggio, Operational analysis of
j.trd.2016.11.001 Model (DOE, 2017). natural gas combined cycle CHP plants: Energy performance
7. S. C. Davis, S. W. Diegel, R. G. Boundy, Transportation Energy 32. O. Schmidt et al., Future cost and performance of water and pollutant emissions. Appl. Therm. Eng. 100, 304–314
Data Book. (Center for Transportation Analysis, ed. 34, 2015). electrolysis: An expert elicitation study. Int. J. Hydrogen (2016). doi: 10.1016/j.applthermaleng.2016.02.040
8. International Energy Agency (IEA), “CO2 emissions from fuel Energy 42, 30470–30492 (2017). doi: 10.1016/ 59. Q. Wang, X. Chen, A. N. Jha, H. Rogers, Natural gas from
combustion,” (IEA, 2016). j.ijhydene.2017.10.045 shale formation – The evolution, evidences and challenges of
9. IEA, Energy Technology Perspectives 2017 (IEA, 2017). 33. DOE, “Technical targets for hydrogen production from shale gas revolution in United States. Renew. Sustain.
10. L. M. Fulton, L. R. Lynd, A. Körner, N. Greene, L. R. Tonachel, electrolysis” (2018); www.energy.gov/eere/fuelcells/doe- Energy Rev. 30, 1–28 (2014). doi: 10.1016/j.rser.2013.08.065
The need for biofuels as part of a low carbon energy future. technical-targets-hydrogen-production-electrolysis. 60. U.S. Energy Information Administration (EIA), “Monthly
Biofuels Bioprod. Biorefin. 9, 476–483 (2015). doi: 10.1002/ 34. S. M. Saba, M. Muller, M. Robinius, D. Stolten, The investment generator capacity factor data now available by fuel and
bbb.1559 costs of electrolysis—A comparison of cost studies from the technology” (EIA, 2014).
11. J. Impullitti, “Zero emission cargo transport II: San Pedro Bay past 30 years. Int. J. Hydrogen Energy 43, 1209–1223 (2018). 61. M. R. Shaner, S. J. Davis, N. S. Lewis, K. Caldeira, Geophysical
ports hybrid & fuel cell electric vehicle project”; www.energy. doi: 10.1016/j.ijhydene.2017.11.115 constraints on the reliability of solar and wind power in the
78. Amaerican Gas Association, Transitioning the Transportation 95. D. P. van Vuuren et al., The role of negative CO2 emissions 109. S. J. Davis, K. Caldeira, H. D. Matthews, Future CO2 emissions
Sector: Exploring the Intersection of Hydrogen Fuel Cell and for reaching 2°C—Insights from integrated assessment and climate change from existing energy infrastructure.
Natural Gas Vehicles (Sandia National Laboratory, 2014). modelling. Clim. Change 118, 15–27 (2013). doi: 10.1007/ Science 329, 1330–1333 (2010). doi: 10.1126/
79. DOE, “Goals for batteries” (DOE, Vehicle Technologies Office, s10584-012-0680-5 science.1188566; pmid: 20829483
2018); https://energy.gov/eere/vehicles/batteries. 96. E. Kriegler et al., The role of technology for achieving climate 110. K. C. Seto et al., Carbon lock-in: Types, causes, and policy
80. R. E. Ciez, J. F. Whitacre, The cost of lithium is unlikely to policy objectives: Overview of the EMF 27 study on global implications. Annu. Rev. Environ. Resour. 41, 425–452 (2016).
upend the price of Li-ion storage systems. J. Power Sources technology and climate policy strategies. Clim. Change 123, doi: 10.1146/annurev-environ-110615-085934
320, 310–313 (2016). doi: 10.1016/j.jpowsour.2016.04.073 353–367 (2014). doi: 10.1007/s10584-013-0953-7 111. D. E. H. J. Gernaat et al., Understanding the contribution of
81. Z. Li et al., Air-breathing aqueous sulfur flow battery for 97. C. Azar et al., The feasibility of low CO2 concentration targets non-carbon dioxide gases in deep mitigation scenarios.
ultralow cost electrical storage. Joule 1, 306–327 (2017). and the role of bio-energy with carbon capture and storage Glob. Environ. Change 33, 142–153 (2015). doi: 10.1016/
doi: 10.1016/j.joule.2017.08.007 (BECCS). Clim. Change 100, 195–202 (2010). doi: 10.1007/ j.gloenvcha.2015.04.010
82. C. Quinn, D. Zimmerle, T. H. Bradley, The effect of s10584-010-9832-7 112. D. P. van Vuuren et al., Energy, land-use and greenhouse gas
communication architecture on the availability, reliability, and 98. J. M. D. MacElroy, Closing the carbon cycle through rational emissions trajectories under a green growth paradigm.
economics of plug-in hybrid electric vehicle-to-grid ancillary use of carbon-based fuels. Ambio 45 (Suppl 1), S5–S14 Glob. Environ. Change 42, 237–250 (2017). doi: 10.1016/
services. J. Power Sources 195, 1500–1509 (2010). (2016). doi: 10.1007/s13280-015-0728-7; pmid: 26667055 j.gloenvcha.2016.05.008
doi: 10.1016/j.jpowsour.2009.08.075 99. H. de Coninck, S. M. Benson, Carbon dioxide capture and 113. EIA, “Levelized Cost and Levelized Avoided Cost of New
83. J. I. Pérez-Díaz, M. Chazarra, J. García-González, G. Cavazzini, storage: Issues and prospects. Annu. Rev. Environ. Resour. 39, Generation Resources in the Annual Energy Outlook 2018”
A. Stoppato, Trends and challenges in the operation of 243–270 (2014). doi: 10.1146/annurev-environ-032112-095222 (2018); www.eia.gov/outlooks/aeo/pdf/
pumped-storage hydropower plants. Renew. Sustain. Energy 100. R. Socolow et al., “Direct air capture of CO2 with chemicals: A electricity_generation.pdf.
Rev. 44, 767–784 (2015). doi: 10.1016/j.rser.2015.01.029 technology assessment for the APS Panel on Public Affairs,”
84. A. B. Gallo, J. R. Simões-Moreira, H. K. M. Costa, (American Physical Society, 2011). AC KNOWLED GME NTS
M. M. Santos, E. Moutinho dos Santos, Energy storage in the 101. K. S. Lackner et al., The urgency of the development of CO2 The authors extend a special acknowledgment to M.I.H. for
energy transition context: A technology review. Renew. capture from ambient air. Proc. Natl. Acad. Sci. U.S.A. 109, inspiration on the 20th anniversary of publication of (1). The
Sustain. Energy Rev. 65, 800–822 (2016). doi: 10.1016/ 13156–13162 (2012). doi: 10.1073/pnas.1108765109; authors also thank M. Dyson, L. Fulton, L. Lynd, G. Janssens-Maenhout,
j.rser.2016.07.028 pmid: 22843674 M. McKinnon, J. Mueller, G. Pereira, M. Ziegler, and
85. T. Letcher, Storing Energy with Special Reference to 102. Z. Kapetaki, J. Scowcroft, Overview of carbon capture and M. Wang for helpful input. This Review stems from an
Renewable Energy Sources (Elsevier, 2016). storage (CCS) demonstration project business models: Risks Aspen Global Change Institute meeting in July 2016 convened
Ruqiang Liang, Min Jee Jang, Haining Zhong, Daniel Dombeck, Mark von Zastrow, ON OUR WEBSITE fluorescence intensity of a
Axel Nimmerjahn, Viviana Gradinaru, John T. Williams, Lin Tian† circularly permuted green
Read the full article
at http://dx.doi. fluorescent protein. The
org/10.1126/ high sensitivity and tem-
INTRODUCTION: Neuromodulators, such as very little is known about how these signals science.aat4422 poral resolution of dLight1
..................................................
Footshock
Fluorescence Extinction
Response to cue
circuits.
▪
High-resolution dopamine imaging in vivo. dLight1 permits robust detection of physiolog- The list of author affiliations is available in the full article online.
ically and behaviorally relevant dopamine (DA) transients with high sensitivity and spatio- *These authors contributed equally to this work.
†Corresponding author. Email: lintian@ucdavis.edu
temporal resolution, including dynamic learning-induced dopamine changes in the nucleus Cite this article as T. Patriarchi et al., Science 360,
accumbens (bottom) and task-specific dopamine transients in the cortex (top). eaat4422 (2018). DOI: 10.1126/science.aat4422
A
site in dLight1.1 and dLight1.2 (fig. S1, C to G)
nimal behavior is influenced by the release circuits while also measuring and manipulating greatly increased the dynamic range but also re-
of neuromodulators such as dopamine the elements of the circuit during behavior. duced the affinity to micromolar range (dLight1.3a:
(DA), which signal behavioral variables that Analytical techniques such as microdialysis DF/Fmax = 660 ± 30%, Kd = 2300 ± 20 nM, fig. S2,
are relevant to the functioning of circuits and electrochemical microsensors have provided A and B; dLight1.3b: DF/Fmax = 930 ± 30%, Kd =
brainwide. Projections from dopaminergic useful insights about neuromodulator presence 1680 ± 10 nM; Fig. 1, D and E). Insertion of the
nuclei to the striatum and cortex, for example, (12, 13) but suffer from poor spatial and/or tem- cpGFP module into DRD4 and DRD2 produced
play important roles in reinforcement learning, poral resolution and cannot be targeted to cells dLight1.4 and dLight1.5, respectively, which ex-
decision-making, and motor control. Loss of DA of interest. Optical approaches such as injected hibited nanomolar affinity with a relatively small
or dysfunction of its target circuits has been cell-based systems (CNiFERs) (14) and reporter dynamic range [dLight1.4: DF/Fmax = 170 ± 10%,
linked to disorders such as Parkinson’s disease, gene–based iTango (15) can reveal DA release Kd = 4.1 ± 0.2 nM, Fig. 1, B, D, and E; dLight1.5:
schizophrenia, and addiction (1–3). with high molecular specificity. However, these DA, DF/Fmax = 180 ± 10%, Kd = 110 ± 10 nM; quin-
Much work has been devoted to determining systems are limited by poor temporal resolution pirole (synthetic agonist of D2 dopamine recep-
how neural representations of behavioral states (seconds to hours), preventing direct detection tors), DF/Fmax = 124 ± 19%, fig. S2, A to C]. In
are encoded in the firing patterns of neuro- of DA release events that occur on a subsecond addition, we engineered a control sensor by in-
modulatory neurons (4–9), but very little is time scale (16, 17). corporating a D103A mutation in dLight1.1 to
known about how the precise release of neuro- High-quality single fluorescence protein (FP)– abolish DA binding (control sensor: DF/F = 0.4 ±
modulators alters the function of their target based sensors that report calcium or glutamate 4%, Fig. 1E) (22). Because dLight1.1 and dLight1.2
circuits (10, 11). To address this problem, an transients with subsecond temporal resolution produced large responses at low DA concentra-
essential step is to monitor the spatiotemporal have recently been developed and are widely tion (e.g., 100 nM) without approaching response
dynamics of neuromodulatory signals in target used (18, 19). Here, we report the development saturation (Fig. 1E, inset) and had submicro-
of a set of single FP–based DA sensors, named molar affinity, we further characterized these two
dLight1, that enables imaging of DA transients sensors.
1
Department of Biochemistry and Molecular Medicine, with high spatiotemporal resolution in behav-
University of California, Davis, 2700 Stockton Boulevard,
ing animals. Sensor characterization
Sacramento, CA 95817, USA. 2Division of Biology and
Biological Engineering, California Institute of Technology, These two sensors showed peak emissions at
Pasadena, CA 91125, USA. 3Waitt Advanced Biophotonics Sensor engineering 516 nm and 920 nm for one- and two-photon
Center, Salk Institute for Biological Studies, La Jolla, CA Sensitive optical readout of changes in DA con- illumination in HEK cells, respectively (fig. S3).
92037, USA. 4Department of Neurobiology, Northwestern
University, Evanston, IL 60208, USA. 5Department of Cellular
centration was achieved by directly coupling the In situ titration on dissociated hippocampal neu-
and Molecular Pharmacology, University of California, San DA binding–induced conformational changes in rons and on HEK293 cells showed similar ap-
Francisco, CA 94131, USA. 6Vollum Institute, Oregon Health human DA receptors to changes in the fluores- parent affinities to DA (Fig. 1E and fig. S4, A to C).
& Science University, Portland, OR 97239, USA. cence intensity of circularly permuted green fluo- Single 5-ms pulses of uncaged DA were robustly
*These authors contributed equally to this work.
†Present address: Department of Psychological and Brain
rescent protein (cpGFP). We did this by replacing detected on the dendrites of cultured neurons,
Sciences, Boston University, Boston, MA 02215, USA. the third intracellular loop (IL3) of the human and the fluorescence response tracked uncaging
‡Corresponding author. Email: lintian@ucdavis.edu dopamine D1 receptor (DRD1), D2 receptor pulse duration (fig. S4, D to F). In cultured
hippocampal slices, dLight1 could reliably detect the presence of the DRD1 antagonists SKF-83566 nously expressed DRD1 (U2OS), dLight1 did not
submicromolar DA concentration changes at and SCH-23390 but was unaffected by the DRD2 significantly alter the dose-response curve for DA
dendrites and single dendritic spines (fig. S4, antagonists haloperidol and sulpiride (Fig. 1F). (P = 0.96, fig. S6B). dLight1 also showed a signif-
G to I). To investigate the possible interference of icant reduction in agonist-induced internaliza-
We then investigated the endogenous and phar- sensor expression with G protein signaling, we tion, a readout of DRD1 engagement of b-arrestin
macological molecular specificity of the sensor. first measured the effect of sensor expression (24), when compared to wild-type DRD1 (fig. S6C).
dLight1 was less sensitive to norepinephrine on the ligand-induced cyclic adenosine mono- Total internal reflectance fluorescence (TIRF)
and epinephrine than to DA by factors of ~70 phosphate (cAMP) response (fig. S6) (23). Tran- imaging verified that dLight1 remained diffusely
and ~40, respectively; negligible responses were siently transfected dLight1.1 and dLight1.2 triggered distributed in the plasma membrane, without any
observed to all other neuromodulators tested no significant cAMP response in HEK cells, sim- detectable internalization, during a complete cycle
(fig. S5). The amplitude of the response to each ilar to the negative control (EGFP), whereas wild- of ligand-dependent fluorescence change (fig. S6,
pharmacological compound reflected the efficacy type DRD1 receptor significantly did (fig. S6A). D to F). Taken together, these results indicate
of drugs on the wild-type receptors, with the The conversion of DRD1 to a fluorescent sensor that the dLight sensors are suitable for use on
largest response to the full agonist dihydrexidine thus apparently blocked the scaffold’s ability to the cell membrane without affecting endogenous
(DF/F = 300 ± 10%), followed by partial agonists bind G protein and trigger the signaling cascade. signaling through G proteins or engagement of
(Fig. 1F). The response to DA was abolished in When introduced into a cell line that endoge- b-arrestins.
Fig. 1. Development and characterization of dLight1. (A) Simulated Fluorescence intensity and signal-to-noise ratio of apo and sat state are shown.
structure of dLight1 consisting of DRD1 and cpGFP module. (B) Sequence Scale bars, 10 mm. (E) In situ titration of DA on HEK cells. Data were fitted
alignment of transmembrane (TM) domain 5 and 6 in b2AR, DRD1, and DRD4. with the Hill equation (n = 5). (F) Pharmacological specificity of dLight1.1.
Library design is shown. Amino acid abbreviations: A, Ala; D, Asp; E, Glu; DRD1 full agonist (dihydrexidine, 295 ± 8%, n = 5); DRD1 partial agonists
F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; Q, Gln; R, Arg; S, Ser; T, Thr; (SKF-81297, 230 ± 7.7%, n = 5; A77636, 153 ± 7.8%, n = 7; apomorphine, 22 ±
V, Val; W, Trp; Y, Tyr. (C) Screening result of 585 linker variants. Red and 0.8%, n = 6); DRD1 antagonists (SCH-23390, –0.04 ± 0.01%, n = 7; SKF-
blue vertical bars indicate fluorescence changes (DF/F) in response to 10 mM 83566, 0.04 ± 0.03%, n = 7); DRD2 antagonists (sulpiride, 213 ± 5.1%, n = 5;
DA; significance values of DF/F are shown by colored bars and scale (n = 3 haloperidol, 219 ± 11%, n = 6). Data are means ± SEM. ****P < 0.0001
trials, two-tailed t test). (D) Expression of dLight variants in HEK cells. [one-way analysis of variance (ANOVA), Dunnett posttest]; n.s., not significant.
Versatile application to opioid receptors (KOR, MOR) and a2 adre- fluorescence responses to their respective agonists
other neuromodulators nergic receptor (A2AR); and Gq-coupled 5- (fig. S7B).
We next applied the design strategy of dLight1 hydroxytryptamine (serotonin) receptor-2A
to modularly develop a class of intensity-based (5HT2A) and melatonin type-2 receptor (MT2). Two-photon imaging of DA release in
sensors for various neuromodulators and neuro- As with dLight1, we replaced IL3 with cpGFP, dorsal striatum ex vivo and in vivo
peptides. We selected a subset of GPCRs, includ- with insertion sites chosen to preserve the con- We next used dLight1 to measure the time course
ing Gs-coupled b1 and b2 adrenergic receptors served positive charges (fig. S7A). All sensors and concentration of endogenous DA release
(B1AR and B2AR); Gi -coupled k- and m-type localized to the membrane and showed positive triggered by electrical stimulation and drug
modification in acute striatal slices with two- frequency stimuli elicited significantly smaller higher than previously reported in ventral stri-
photon imaging (Fig. 2A). Two to four weeks after responses (Fig. 2, E and F), indicating strong atum using fast-scan cyclic voltammetry (FSCV)
injection of an adeno-associated virus encod- depression from an initially high probability of (25) and is similar to that reported by measuring
ing dLight1 (AAV9.hSynapsin1.dLight1.2) into release. Blockade of DA reuptake with cocaine DRD2 activation (26). Addition of saturating
the dorsal striatum, we observed both broadly significantly prolonged the decay of fluorescence amphetamine (10 mM in the presence of 400 mM
distributed and localized fluorescence transients from peak to baseline (Fig. 2, G and H), but with sulpiride) increased tonic DA to 3.3 mM (fig. S8,
across the field of view (Fig. 2, B and C, and fig. equivocal effect on response amplitude (Fig. 2, G F and G).
S8, A to C) in response to a single electrical sti- and H). Application of the competitive antagonist We then examined the action of known mod-
mulus. Fast line scan at these hotspots (Fig. 2C) SKF83566 eliminated the responses (fig. S8F), ulators of DA release using dLight1 (Fig. 2, J to L).
revealed a rapid onset of fluorescence increase confirming that fluorescent signals are indeed Activation of D2 autoreceptors with quinpirole
(rise t1/2 = 9.5 ± 1.1 ms) followed by a plateaued attributable to DA binding. decreased the electrically evoked fluorescence
peak (averaged DF/F = 220 ± 50%) for about We next used dLight1 to estimate released DA transients; this effect was significantly reversed
150 ms, which decayed to baseline in about concentration induced by a brief electrical stim- by the application of sulpiride (Fig. 2J). Perfusion
400 ms (decay t1/2 = 90 ± 11 ms, Fig. 2D). We ob- ulus. By comparison with a concentration-response with a k-opioid receptor agonist (U69,593) caused
served robust and reproducible fluorescent tran- curve (fig. S8, D, E, and G), the fluorescence a small decrease in the amplitude, which was
sients to low-frequency stimuli over a prolonged response suggested a DA release of 10 to 30 mM completely blocked by naloxone (Fig. 2J). We then
imaging period, whereas subsequent higher- (Fig. 2I), which is one to two orders of magnitude imaged the effects of nicotinic receptor activation
Fig. 3. Deep brain imaging of DA release triggered by optogenetic stimulation of VTA GABA neurons that inhibits VTA DA neurons. (J and
stimulation and combined with calcium imaging in freely behaving K) Averaged fluorescence decrease in response to optogenetic stimulation
mice. (A) Schematics showing fiber photometry recording of dLight1.1 or at 40 Hz (n = 4 mice) and quantification of mean fluorescence. (L) Dual-
control sensor in NAc while stimulating VTA DA neurons by optogenetics. color fiber photometry recording of DA release with dLight1.1 and local
(B) Expression of dLight1.1 in NAc around fiber tip location and neuronal activity with jRGECO1a. (M and N) Increase of dLight1.1 (green)
ChrimsonR-expressing axons from midbrain. (C) ChrimsonR-expressing and jRGECO1a (magenta) fluorescence during 5% sucrose consumption
TH+DA neurons in VTA. (D) Averaged fluorescence increase in response with lick rate (black, n = 5 mice) and quantification of mean fluorescence.
to optogenetic stimuli (n = 5 mice). (E) Quantification of peak fluorescence (O and P) Fluorescence decrease in dLight1.1 (green) and increase in
at 20 Hz. (F) Fluorescence fold changes relative to 5 Hz. (G and jRGECO1a (magenta) during unpredictable footshock delivery (0.6 mA for 1 s,
H) Optogenetically induced fluorescence increase of dLight1.1 after systemic n = 5 mice) and quantification of mean fluorescence. Data shown are
administration of saline, D1 antagonist (SCH-23390, 0.25 mg/kg), and means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 (paired or unpaired
DA reuptake inhibitor (GBR-12909, 10 mg/kg) (n = 5 mice). (I) Schematics t tests for two-group comparisons; one-way ANOVA by post hoc Tukey test
showing fiber photometry recording of dLight1.1 in NAc and optogenetic for multiple-group comparisons).
in mediating the probability of DA release. consistent regardless of the stimulation protocol hSynapsin1.flex.tdTomato. We measured DA tran-
Blockade of nicotinic receptors with hexametho- (Fig. 2, K and L) (27). sients with two-photon imaging during rest and
nium profoundly reduced the fluorescence tran- Next, we asked whether dLight1 could reliably self-initiated locomotion (fig. S9). Consistent with
sient, which depended on the number of stimuli report DA signals associated with mouse loco- in vivo two-photon calcium imaging of substantia
(Fig. 2, K and L). In the absence of hexametho- motion in dorsal striatum, which was labeled nigra pars compacta (SNc) axon terminals in
nium, the amplitude of the fluorescence remained with AAV1.hSynapsin1.dLight1.1/1.2 and AAV1. dorsal striatum (10), dLight1 reliably showed
Fig. 4. Dynamic changes of NAc DA signaling during appetitive and US-evoked [(G), left] average fluorescence and US-triggered licks [(G),
Pavlovian conditioning and reward prediction error. (A) Pavlovian right] across learning and extinction sessions. (H) Quantification of peak
conditioning procedures involved learning to associate neutral cues fluorescence across learning and extinction. (I) Reward prediction error
(CS; house light and 5-kHz tone) with a sucrose reward (US; 50 ml of procedure. (J) Fluorescence response during expected (red) versus
5% sucrose) and subsequent extinction. (B) Change of CS-evoked licks unexpected (black) reward consumption (n = 4 mice). (K) Peak fluores-
across cue-reward learning (left) and extinction (right). (C and D) dLight1.1 cence evoked by expected (red) and unexpected (black) reward con-
dynamics in response to CS and US in first and last sessions of cue-reward sumption. (L) Fluorescence response during expected (red) versus
learning, shown in single (gray) and averaged (blue) trials (n = 20 trials) unexpected (brown) reward omission (n = 4 mice). Second and third
from a single animal (C) or averaged across all trials and animals (n = dotted lines indicate US onset and CS offset, respectively. (M) Mean
5 mice) (D). Lick rate is shown in black. (E) Same as (D) for cue-reward fluorescence during baseline and after unexpected reward omission.
extinction (n = 5 mice). In (D) and (E), dotted lines indicate CS onset, US Data are means ± SEM. **P < 0.01 (Pearson correlation coefficient and
onset, and CS offset, respectively. (F to H) Evolution of CS-evoked (F) paired t test).
widespread and synchronous subsecond slice. In addition, dLight1 enables direct visual- and fig. S10, A and B). dLight1 revealed visible
transients associated with spontaneous locomo- ization of locomotion-triggered DA release in spontaneous DA transients, which were absent
tion, which was clearly distinguishable from behaving mice. in the imaging sessions using the control sensor
motion artifacts (fig. S9, A to E). The DA tran- (fig. S10C).
sients were rapidly and bidirectionally modu- Deep-brain recording of DA dynamics To optically activate VTA dopaminergic neu-
lated with respect to locomotion. Accelerations simultaneously with optogenetics or rons, we infected VTA of TH::IRES-Cre mice
were associated with an increase and deceler- calcium imaging with AAV5.hSynapsin1.flex.ChrimsonR.tdTomato
ations with a decrease in fluorescence (peak mean The nucleus accumbens (NAc) receives projec- (28) (Fig. 3, A to C, fig. S11, A and B, and fig. S12, A
cross-correlation 240 ms fig. S9, F to L). tions from dopaminergic neurons in the ventral and D). The high temporal resolution of dLight1
In summary, dLight1 faithfully and directly tegmental area (VTA). To directly probe DA re- enabled detection of individual peaks of DA
reports the time course and concentration of lease in freely moving mice, we delivered AAV9. transients in response to 5-, 10-, and 20-Hz
local DA release and drug-dependent modula- CAG.dLight1.1 or AAV9.CAG.control_sensor in photostimulation (Fig. 3D and fig. S13, A to C).
tory effects on DA release in an acute striatum NAc, followed by fiber photometry imaging (Fig. 3 The amplitude of fluorescence increase was
correlated with the frequency of photostimula- US response, when aligned to the consumption during reward expectation or correlate with loco-
tion (Fig. 3, D and F). In contrast, no fluorescence onset, showed a monotonic decrease across learn- motion, we aligned the trials at running onset
changes were observed with the control sensor ing sessions (Fig. 4, G and H, and fig. S14D) (9, 33). (Fig. 5E, group averages; fig. S16G, single ROIs)
using 20-Hz stimuli (Fig. 3, D and E). Relative to During extinction, we observed an attenuated and compared the DA transients of runs trig-
saline-injected controls, systemic administration phasic CS response (Fig. 4, E, F, and H). The am- gered by the “Go” stimulus (when the animals
of SCH-23390 significantly reduced optogeneti- plitude of the phasic CS response was correlated expected a reward) with spontaneous runs that
cally induced dLight1 responses, whereas the re- with CS-triggered licking behavior during both erroneously occurred during the standstill phase
uptake inhibitor GBR-12909 enhanced them learning and extinction sessions (fig. S14E). (with no reward expectation). A small subset of
(Fig. 3, G and H). We further investigated whether dLight1 can responsive ROIs (5%) showed significant increases
Next, we examined whether dLight1 can re- report signals correlated with “reward prediction in DA transients during reward expectation but
port inhibition of DA transients. To induce tran- error” (4). After the animals had fully learned not spontaneous running (Fig. 5E, center), whereas
sient inhibition of VTA dopaminergic neurons, we CS-US association, mice underwent “unexpected the other 32% of ROIs correlated with locomotion
optogenetically stimulated VTA g-aminobutyric reward availability” sessions (in which the US (Fig. 5E, left). The 63% of ROIs responsive to
acid–releasing (GABAergic) neurons in VGAT:: was occasionally made available without the CS) reward only (Fig. 5D, right) also showed increased
IRES-Cre mice (29) (Fig. 3I). Histology confirmed between normal paired trials (Fig. 4I). Unexpected DA transients during the early stimulus pre-
ChrimsonR expression in VTA GABAergic neu- availability of reward elicited significantly higher sentation phase consisting of both locomotion-
rons (fig. S12, B, C, and E). We observed rapid fluorescence than did expected consumption and reward expectation–related responses (Fig.
and reversible reductions in dLight1 fluorescence (Fig. 4, J and K). In the “unexpected reward omis- 5E, right). All three types of responses were
in response to VTA GABAergic neuron photo- sion” session, where the US was occasionally consistently seen across animals. Comparing the
activation at 40 Hz (Fig. 3, J and K, and fig. S13D), omitted after the predictive CS, fluorescence de- heterogeneity of response transients between
indicating that dLight1 can report bidirectional creased below the pre-CS baseline after the time layer 1 and layers 2/3 of cortical area M1 (fig. S16,
changes in local DA release. at which the US would have normally become E and F), we found that layer 2/3 showed more
functionally heterogeneous DA signals during a 13. M. Ganesana, S. T. Lee, Y. Wang, B. J. Venton, Analytical 33. J. J. Day, M. F. Roitman, R. M. Wightman, R. M. Carelli,
visuomotor learning task. Simultaneous cal- Techniques in Neuroscience: Recent Advances in Imaging, Associative learning mediates dynamic shifts in
Separation, and Electrochemical Methods. Anal. Chem. dopamine signaling in the nucleus accumbens. Nat.
cium imaging can further determine how 89, 314–341 (2017). doi: 10.1021/acs.analchem.6b04278; Neurosci. 10, 1020–1028 (2007). doi: 10.1038/nn1923;
spatiotemporal differences in DA levels relate to pmid: 28105819 pmid: 17603481
ongoing neural activity and influence associa- 14. A. Muller, V. Joseph, P. A. Slesinger, D. Kleinfeld, Cell-based 34. J. J. Clark et al., Chronic microsensors for longitudinal,
tive learning or goal-directed behavior. reporters reveal in vivo dynamics of dopamine and subsecond dopamine detection in behaving animals.
norepinephrine release in murine cortex. Nat. Methods 11, Nat. Methods 7, 126–129 (2010). doi: 10.1038/nmeth.1412;
dLight1.1 and dLight1.2 are optimized sensor 1245–1252 (2014). doi: 10.1038/nmeth.3151; pmid: 25344639 pmid: 20037591
variants that can be immediately applied to 15. D. Lee et al., Temporally precise labeling and control of 35. J. A. da Silva, F. Tecuapetla, V. Paixão, R. M. Costa, Dopamine
ex vivo or in vivo studies, as they offer a good neuromodulatory circuits in the mammalian brain. neuron activity before action initiation gates and invigorates
balance between dynamic range and affinity. Nat. Methods 14, 495–503 (2017). doi: 10.1038/nmeth.4234; future movements. Nature 554, 244–248 (2018). doi: 10.1038/
pmid: 28369042 nature25457; pmid: 29420469
Other dLight variants may be suitable for mea- 16. C. P. Ford, P. E. Phillips, J. T. Williams, The time course of
36. K. D. Piatkevich et al., A robotic multidimensional directed
suring synaptic release (dLight1.3) or tonic DA dopamine transmission in the ventral tegmental area.
evolution approach applied to fluorescent voltage reporters.
transients (dLight1.4). Given the broadly tunable J. Neurosci. 29, 13344–13352 (2009). doi: 10.1523/
Nat. Chem. Biol. 14, 352–360 (2018). doi: 10.1038/
affinity and dynamic range of dLight1, protein JNEUROSCI.3546-09.2009; pmid: 19846722
s41589-018-0004-9
17. C. P. Ford, S. C. Gantz, P. E. Phillips, J. T. Williams, Control of
engineering and high-throughput screening ef- extracellular dopamine at dendrite and axon terminals.
AC KNOWLED GME NTS
forts can further optimize the signal-to-noise J. Neurosci. 30, 6975–6983 (2010). doi: 10.1523/
We thank B. P. McGrew for assistance during in vitro sensor library
ratio and molecular specificity (36) as well as the JNEUROSCI.1020-10.2010; pmid: 20484639
18. T. W. Chen et al., Ultrasensitive fluorescent proteins for screening; H. Cheng for producing AAV virus; L. Lavis (Janelia
performance of other neuromodulator indicators. Research Campus) for providing NV-caged DA; E. Carey for
imaging neuronal activity. Nature 499, 295–300 (2013).
In combination with calcium imaging and op- doi: 10.1038/nature12354; pmid: 23868258 performing cortical viral vector injections; and B. Mensh for critical
togenetics, our sensors are well poised to permit 19. J. S. Marvin et al., An optimized fluorescent probe for advice in writing and revising the manuscript. This project was
visualizing glutamate neurotransmission. Nat. Methods 10, made possible with generous help from L. Looger (Janelia
direct functional analysis of how the spatio-
162–170 (2013). doi: 10.1038/nmeth.2333; pmid: 23314171 Research Campus). Funding: Supported by NIH BRAIN Initiative
temporal coding of neuromodulatory signaling
between mammary stem cells and the reporter mice, we found that Dll1 expres-
sion is enriched in MaSCs, and Dll1+ MaSCs
have a greater regenerative potential than
macrophageal niche Dll1− MaSCs. Lineage tracing with Dll1-Cre-
ERT2;dTomato reporter mice revealed that
Rumela Chakrabarti*, Toni Celià-Terrassa†, Sushil Kumar†, Xiang Hang, Yong Wei, Dll1+ cells can produce both basal and lu-
Abrar Choudhury, Julie Hwang, Jia Peng, Briana Nixon, John J. Grady, Christina DeCoste, minal cells. Dll1cKO mice exhibit a significant
Jie Gao, Johan H. van Es, Ming O. Li, Iannis Aifantis, Hans Clevers, Yibin Kang* reduction in the number of mammary gland
◥
macrophages. The mam-
ON OUR WEBSITE mary macrophages have
INTRODUCTION: The stem cell niche plays a understanding both normal tissue homeosta- Read the full article
molecular features, includ-
crucial role in regulating key stem cell proper- sis and disease conditions such as breast cancer. at http://dx.doi. ing enrichment of Wnt and
ties, including self-renewal, differentiation, and org/10.1126/ Notch signaling pathway
Notch ligand Dll1 mediates cross-talk mary gland was confirmed by significant reduc-
tion of Dll1 mRNA (Fig. 1C) and Dll1 protein
(Fig. 1D) in MECs of K14-Cre/Dll1f/f (Dll1cKO) mam-
between mammary stem cells and the mary glands and by immunofluorescence in P4
cells (fig. S1A). Notably, a significant reduction in
M
droplets and milk production, as compared with
ammary epithelial cells (MECs) are com- established for sustaining adult stem cells in WT alveoli (fig. S1C). We performed immuno-
posed of two major cell types—basal and many organs (23), including MaSCs (3, 18–20, 24). histochemistry with Ki67 antibodies to confirm
luminal cells—both of which derive from Several studies have shown that Wnt ligands such reduced proliferation in the Dll1cKO mammary
mammary gland stem cells (MaSCs) dur- as Wnt3a and Wnt4 are important for the self- glands, compared with WT mammary glands,
ing puberty and each round of pregnancy renewal of MaSCs (18, 25); however, these Wnt during lactation (fig. S1C). Because the alveoli
and lactation (1, 2). The existence of MaSCs has ligands are not expressed by the basal stem cells, in the Dll1cKO mammary glands were lacking in
been demonstrated by transplantation (3, 4) and which suggests that an adjacent MaSC niche lipid droplets and milk secretion, we next tested
lineage tracing experiments (5, 6). MECs are sur- might be responsible for secretion of the ligands. whether the proliferation defect was associated
rounded by various stromal cell types, including Indeed, recent studies have shown that Wnt4 con- with secretory differentiation failure by staining
adipocytes, fibroblasts, macrophages, endothe- trols MaSC function through luminal-myoepithelial for Np2b (Na-Pi cotransporter) protein, whose ab-
lial cells, and lymphoid cells (7). Although some cross-talk (25). It remains unclear whether stro- sence at parturition indicates a lack of secretory
of these cells contribute to mammary gland de- mal niche cells can also produce Wnt ligands to function (29). The apical membranes of secretory
velopment and homeostasis (8–14), little is known regulate MaSCs. alveoli in Dll1cKO mammary glands showed re-
about their functional involvement in regulat- In this study, we demonstrate that Notch li- duced Npt2b staining compared with the WT
ing MaSCs. gand Dll1 expression is enriched in MaSCs and mammary glands that exhibited intense Npt2b
Notch and Wnt signaling pathways are key that Dll1 activates Notch signaling in stromal staining (fig. S1C). Overall, our data indicate that
regulators of essential developmental processes macrophages to induce their expression of Wnt reduced proliferation and a block in secretory
in the mammary gland, including stem cell main- ligands, which feed back to MaSCs to promote differentiation may both play a critical role in
tenance, cell fate decisions, and dedifferentiation stem cell activity. Our study defines a Dll1-mediated contributing to the defects in lobuloalveolar de-
(3, 15–20). Notably, studies on Notch signaling in MaSC niche in which the survival and function of velopment of Dll1cKO mice.
mammary gland development have primarily MaSCs and stromal macrophages is mutually
focused on the receptors (15–17, 21, 22), whereas regulated by cross-talk through Notch and Wnt Dll1 is critical for maintaining
relatively little is known about the role of specific signaling. MaSC numbers
Notch ligands. Similarly, Wnt signaling is well Because Dll1 is predominantly expressed in basal
Results cell populations in which MaSCs are thought to
1
Dll1 is required for mammary reside (3, 5), we next probed for possible alteration
Department of Molecular Biology, Princeton University, morphogenesis in virgin and pregnant
Princeton, NJ 08544, USA. 2Department of Biomedical of MaSC number or function in Dll1cKO mice.
Sciences, University of Pennsylvania, Philadelphia, PA 19104, mammary glands Fluorescence-activated cell sorting (FACS) anal-
USA. 3Immunology Program, Memorial Sloan Kettering Our recent gene expression profiling analysis of ysis demonstrated a significant decrease in the
Cancer Center, New York, NY 10065, USA. 4Department of
Pathology, NYU Langone Medical Center, New York City, NY
different populations of mammary gland cells (26) MaSC-enriched P4 population in Dll1cKO mice
10016, USA. 5Hubrecht Institute and University Medical revealed that Dll1 is predominantly expressed in (Fig. 2, A and B, and fig. S2A). Limiting dilution
Center Utrecht, Utrecht, Netherlands. 6Rutgers Cancer basal cells [population four (P4), Lin−CD24+CD29hi] cleared–fat-pad repopulation assay with either
Institute of New Jersey, New Brunswick, NJ 08903, USA. that have been reported to be enriched for MaSCs total live cells (propidium iodide–negative) or
*Corresponding author. Email: ykang@princeton.edu (Y.K.);
rumela@vet.upenn.edu (R.C.) †These authors contributed
(3), compared with lower expression of Dll1 in lineage-negative (Lin−) live cells (CD31−, Ter119−,
equally to this work. ‡Present address: Cancer Research Program, luminal (P5, Lin−CD24+CD29lo) and stromal- and CD45−) revealed a significantly reduced
Hospital del Mar Research Institute (IMIM), Barcelona, Spain. enriched cells (P6, Lin−CD24loCD29lo) (Fig. 1, A repopulating frequency by cells obtained from
Dll1cKO mice (Fig. 2, C and D). Conversely, over- that Dll1 plays a critical role in maintaining Assessment of the reconstitution potential of
expression of Dll1 in MECs by lentiviral trans- MaSC number during different stages of mam- Dll1+ and Dll1− cells from both lineage-negative
duction before transplantation increased MaSC mary gland development. and basal cells (P4) by transplantation assay
repopulation frequency (Fig. 2E and fig. S2B). revealed that Lin−Dll1+ cells generated mamma-
Similar repopulation assays using isolated P4 and Dll1+ cells are enriched for MaSCs ry outgrowths more efficiently than did either
P5 cells from WT and Dll1cKO mice revealed that To further characterize the function and expres- Lin−Dll1− cells or total Lin− populations (fig. S4,
luminal cells (P5) from either WT or Dll1cKO mice sion of Dll1 in mammary glands, we used a Dll1- A and B). Similarly, P4-Dll1+ cells had a much higher
were unable to generate ductal growth, as expected mCherry transgenic mouse model in which the repopulation frequency (Fig. 3, C to E), which
(fig. S2, C and D). Surprisingly, no significant mCherry reporter gene is driven by the Dll1 ge- suggests that the Dll1+ cells represent a subset of
difference was observed between WT and Dll1cKO nomic regulatory sequences. In the mammary MaSC-enriched population. Furthermore, P4-Dll1hi
P4 cells (Fig. 2F). Moreover, only a modest dif- gland, the Dll1-mCherry reporter is expressed basal cells have increased reconstitution potential
ference was observed in the serial transplant predominantly in basal cells at all develop- relative to P4-Dll1lo basal cells (Fig. 3, D and E).
take rate of WT and Dll1cKO basal cells (fig. S2, E mental stages (Fig. 3A and fig. S3, C to F). FACS Gene set enrichment analysis (GSEA) confirmed
and F). These results indicate that the primary analysis indicated that ~12% of Lin− cells are that P4-Dll1hi cells were enriched for MaSC
reason for the reduced ductal growth in the Dll1mCherry positive in virgin mice (Fig. 3B and signatures (26, 30), whereas P4-Dll1lo cells were
Dll1cKO mice is a reduction in the number of fig. S3D), and this population increases substan- enriched for luminal signatures (Fig. 3F). Finally,
MaSCs rather than a defect in their function. tially during pregnancy and lactation (fig. S3D). in serial transplantation assays, both Dll1+ and
This reduction of the MaSC-enriched P4 popu- Notably, Dll1mCherry expression is predominantly Dll1hi cells continued to be more efficient in re-
lation was also observed during lactation (fig. S2, enriched in the upper right portion of the P4 constitution compared with Dll1− and Dll1lo cells,
G and H). Taken together, our studies suggest population (Fig. 3B and fig. S3, E and F). respectively (fig. S4, C to E), further supporting
the notion that Dll1 is enriched in the MaSC confocal imaging further confirmed tdTomato ing further confirmed reduced F4/80+ macro-
population. expression in K14+K8− basal cells (Fig. 4D). At 2 phages in Dll1cKO terminal end buds (TEBs) and
and 6 weeks after induction, tdTomato expres- ducts at different developmental time points
Dll1+-enriched MaSCs can produce both sion was observed in both basal and luminal (Fig. 5D and fig. S7, A to I), with corresponding
basal and luminal cells cells, indicating that Dll1+ cells can produce both reduction of P4 cells (fig. S7, A to I) and increased
To examine the function of Dll1+ cells during lineages (Fig. 4, E and F, and fig. S6A). Lineage apoptotic activity in macrophages (Fig. 5E). Im-
mammary gland development, we performed a tracing also confirmed that Dll1+ cells generated munofluorescence analysis of mammary gland
lineage tracing experiment using the previously both basal and luminal tdTomato+ cells at preg- sections and mammospheres from a 3D in vitro
described Dll1-GFP-IRES-Cre-ERT2 (GFP, green nancy day 14.5 (Fig. 4, G and H) and in adult coculture system demonstrated that Dll1+ basal
fluorescent protein; IRES, internal ribosomal mammary glands (fig. S6, B to F). cells localized close to F4/80+ stromal cells (Fig. 5,
entry site) knock-in mouse model (31). Similar to F, G, and H), suggesting potential cross-talk be-
our observation in Dll1mCherry mice, Dll1GFP was Mammary gland macrophages have tween the two populations via juxtacrine or
predominantly expressed in basal cells, which are distinctive molecular properties and are paracrine signaling. Using the 3D coculture assay,
positive for K14 and DNp63 and negative for K8 regulated by Dll1+ MaSCs we next tested the impact of macrophages on
(fig. S5, A to D). To trace the fate of Dll1GFP+ cells, Because Notch signaling is involved in inter- Dll1+ MaSC activity. Notably, mammary gland
Dll1-GFP-IRES-Cre-ERT2 mice were mated with cellular signaling, we next determined whether macrophages can induce MaSC activity, as re-
tdTomato reporter mice (Fig. 4, A and B), and Dll1 knockout in MECs affected specific stromal flected by increased mammosphere numbers,
tdTomato expression was traced at different time cell populations in the mammary glands. FACS whereas peritoneal macrophages could not (Fig.
points after the initial induction with tamoxifen analysis showed reduced F4/80+ macrophage 5I), indicating a tissue-specific function. Further-
in 4-week-old-mice (Fig. 4B). As expected, FACS and moderately reduced PDGFRa+ fibroblast more, enhancement of stem cell activity was much
analysis at early an time point (2 days after populations in Dll1cKO mammary glands com- more prominent in P4-Dll1+ cells than in P4-Dll1−
induction) revealed tdTomato expression pre- pared with WT glands (Fig. 5, A to C), whereas cells (Fig. 5I), suggesting that MaSCs depends on
Fig. 2. Dll1 is required for maintaining MaSC activity. (A) Representa- stitution efficiency at limiting dilution of total Lin− cells from WT and
tive FACS profile of Lin− MECs from WT and Dll1cKO mice at 5 to 6 weeks of Dll1-overexpressing (Dll1 OE) mammary glands injected into cleared
age. Numbers within the plots are percentages. (B) Box plot showing mammary fat pads of recipient mice. Representative alum carmine–
percentage of P4 (basal) cells in WT and Dll1cKO mice after FACS. n = 18 stained mammary outgrowths from transplantation with 10,000 Lin−
samples for both WT and Dll1cKO animals. See fig. S2A for individual values cells are shown at bottom. (F) Reconstitution efficiency at limiting
for the indicated groups. The P value was computed by paired t test. dilution of Lin−CD24+CD29hi (P4) cells from WT and Dll1cKO mouse
(C) Reconstitution efficiency of total live cells from WT and Dll1cKO mammary glands injected into cleared mammary fat pads of recipient
mammary glands injected into cleared mammary fat pads of recipient mice. Representative alum carmine–stained mammary outgrowths
mice. (D) Reconstitution efficiency of total Lin− cells from WT and Dll1cKO from transplantation are shown at bottom. n indicates the number of
mammary glands injected into cleared mammary fat pads of recipient mammary fat pad injections, as shown in (C) to (F). P values were
mice. Representative alum carmine–stained mammary outgrowths from obtained by Pearson’s chi-square test by using ELDA software. Scale bars,
transplantation with 10,000 Lin− cells are shown at bottom. (E) Recon- 2 mm in (D) to (F).
A gene expression study showed that the overall we first used clodronate liposomes (CLs) to de- apoptosis and depletes macrophages (34). Both
gene signature of the mammary macrophages is plete macrophages in Dll1-mCherry transgenic Csf1r-blocking antibody treatment in WT mice
more similar to that of the resting peritoneal mice (32). Systemic ablation of macrophages and AP20187 treatment in MaFIA mice signifi-
macrophages than to that of activated peritoneal decreased Dll1mcherry+ MaSC numbers (fig. S9, A cantly reduced the number of macrophages (Fig. 6,
macrophages (fig. S8). GSEA showed that, com- and B) and increased their apoptosis (fig. S9, C C, D, G, and H) without affecting dendritic cells
pared with peritoneal macrophages, mammary and D). Next, we performed a cotransplantation and neutrophils (fig. S9G) and blocked the re-
gland macrophages are enriched for Wnt- and assay by injecting Dll1mcherry+ MaSCs into cleared population of the mammary gland by Dll1+ P4
Notch-related gene signatures (Fig. 5, J and K), mammary fat pads with either clodronate- cells (Fig. 6, A, B, E, and F).
including several Wnt ligands and Notch recep- containing or control liposomes. We observed a We further used two in vivo models to inves-
tors (Fig. 5L). Elevated expression of Notch1, -3, nearly complete inhibition of reconstitution in tigate the importance of Notch signaling in ma-
and -4 in mammary macrophages compared with the tissue injected with CL-containing Dll1mcherry+ crophages for sustaining MaSC activity. First, we
peritoneal macrophages was further confirmed by MaSCs compared with control tissue, indicat- used the previously reported RbpjkcKO (CD11c-
Western blot analysis (Fig. 5, M to O). Together, ing the dependence of MaSCs on macrophages Cre; Rbpjk floxed) mouse model (35) in which
these data suggest that MaSCs depend on Dll1 to (fig. S9, E and F). Rbpjk, a mediator of Notch signaling, is condi-
engage and respond specifically to resident macro- To more specifically test whether macrophages tionally deleted in macrophages. Five- to six-week-
phages in the mammary gland. are necessary for Dll1+ MaSC activity, we used two old RbpjkcKO mice showed a significant reduction
additional approaches to deplete macrophages in mammary ductal elongation, branching, and
Depletion of macrophages reduces in vivo: (i) Csf1r-blocking antibody treatment (33) TEB counts compared with their WT littermates
function of Dll1+ MaSCs (Fig. 6, A to D), and (ii) macrophage Fas-induced (Fig. 6, I to L). This phenotype was also associated
To investigate Dll1- and Notch-dependent func- apoptosis (MaFIA) mice (34) (Fig. 6, E to H), in with a decreased basal (P4) population (Fig. 6M),
tion of macrophages within the MaSC niche, which administration of the drug AP20187 induces phenocopying Dll1cKO mice. Next, we used an
Fig. 3. Dll1+ population is enriched in cells with MaSC activity. mice. For sorting of P4-Dll1hi and P4-Dll1lo, the top and bottom 10 to 12% of the
(A) Immunofluorescence image of Dll1-mCherry reporter mouse mammary population were chosen from the P4-Dll1+ cell population. n indicates the
gland section stained with mCherry antibody. (B) (Left) Representative FACS number of mammary fat pad injections. P values were obtained by Pearson’s
profile of MECs from Dll1-mCherry reporter mice at 6 weeks of age. (Middle) chi-square test by using ELDA software. (E) Representative alum carmine–
mCherry+ cells in the Lin− population. (Right) Distribution of Dll1-mCherry+ stained mammary outgrowths from transplantation, as indicated in (C)
cells in different epithelial populations (left), and Dll1-mCherry+ and Dll1- and (D). (F) GSEA demonstrating enriched MaSC signatures in P4-Dll1hi
mCherry− cells in the P4 population (right). (C and D) Reconstitution efficiency populations compared with P4-Dll1lo populations (26, 30). In contrast,
at limiting dilution of different groups of P4 cells from Dll1-mCherry reporter luminal progenitor cell signatures are enriched in Dll1lo populations. NES,
mouse mammary glands injected into cleared mammary fat pads of recipient normalized enrichment score. Scale bars, 40 mm in (A); 2 mm in (E).
ex vivo transplant method (fig. S9H) in which RNAs (shRNAs) (36). Rbpjk-KD and control was significantly reduced when P4-Dll1mcherry+
mammary gland macrophages were isolated mammary macrophages were then mixed with cells were mixed with Rbpjk KD macrophages
from actin-GFP mice, and Rbpjk was knocked Dll1mcherry+ P4 cells and transplanted into re- compared with control macrophages (Fig. 6, N
down by lentiviral transduction (fig. S9I) by cipient NSG mice, which have defective macro- and O, and fig. S9, H and I). Taken together, these
using two previously reported short hairpin phages. The take rate of mammary outgrowths studies demonstrate a functional dependence of
Dll1+ MaSCs on mammary macrophages through of Dll1-mediated Notch signaling between macro- in Dll1mCherry− basal cells when cocultured with
Notch signaling. phages and Dll1+ MaSCs. The addition of P4- macrophages (fig. S10, E and F). Such Dll1-
Dll1mcherry+ cells to the culture induced Hes1 dependent Notch downstream gene activation
Dll1 regulates Notch signaling in and Hey1 expression in F4/80+ macrophages was suppressed with a Dll1-blocking monoclonal
neighboring macrophages but not in fibroblasts and endothelial cells (fig. antibody (Fig. 7A). In F4/80+ macrophages from
We developed an in vitro coculture assay (fig. S10, S10, C and D). Hes1 and Hey1 expression was WT mammary glands, Notch2 and Notch3 are
A and B) to further investigate the molecular role more pronounced in Dll1mCherry+ basal cells than the most abundantly expressed Notch receptors
(fig. S10F). Treatment of the coculture of P4-Dll1+ was a significant increase in mammosphere num- by antibodies against Dll1, Notch2, or Notch3
mammary stem cells and macrophages with either ber (Fig. 7C), suggesting a MaSC-promoting prop- reduced the number of mammospheres (Fig. 7D).
Notch2- or Notch3-blocking antibody reduced erty of mammary gland macrophages. Conversely, Overall, our data indicate a Dll1-mediated Notch
Hey1 expression (Fig. 7B), indicating that Notch2 macrophages from Dll1cKO mice (MfcKO) were signaling pathway between MaSCs and macro-
and Notch3 mediate Dll1-dependent cross-talk be- less competent in promoting mammosphere phages that is crucial for supporting MaSC activity.
tween MaSCs and macrophages. formation of either WT or Dll1cKO P4 cells, in-
To examine the functional importance of Dll1- dicating an altered cellular property of the macro- Dll1-dependent expression of Wnt
Notch signaling within the MaSC-macrophage phages from Dll1cKO mammary glands. Consistent ligands in macrophages
niche, we again used mammosphere coculture with the role of Dll1 and Notch2 and -3 in me- To gain mechanistic insight as to how macro-
assays. When macrophages from WT mice were diating the cross-talk between MaSCs and macro- phages dictate the cell fate of MaSCs, we per-
mixed with either WT or Dll1cKO P4 cells, there phages, treatment of the mammosphere coculture formed global transcriptomic analysis of F4/80+
macrophages isolated from WT and Dll1cKO Wnt16, and Wnt3 (Fig. 7E). These data are con- fluorescence polymerase chain reaction (qRT-
mammary glands. Focusing on extracellular sistent with our earlier finding (Fig. 5, K and L) PCR) and immunofluorescence analyses of Wnt3,
secreted factors and cytokines, we found that showing that mammary macrophages are en- Wnt10a, and Wnt16 in the coculture system
among the 10 most differentially expressed genes riched for Wnt signaling genes as compared with further confirmed Dll1-Notch2– or Dll1-Notch3–
were three genes coding for Wnt ligands: Wnt10A, peritoneal macrophages. Quantitative real-time dependent stimulation of Wnt ligand expression
Fig. 7. Dll1-mediated cross-talk between MaSCs and macrophages (I to K) Representative immunofluorescence images of coculture cells
promotes Wnt ligand expression in macrophages to support MaSC (macrophages cultured for 3 days followed by addition of P4-Dll1+ cells for 5 h)
activity. (A and B) Hey1 mRNA levels in F4/80+ cells after coculture with stained with Wnt3,Wnt10a, and Wnt16 antibodies.The control was macrophage
P4-Dll1+ cells with and without blocking antibody against Dll1, Notch2, and cultured alone without P4-Dll1+ cells. (L) Quantification of Wnt3, Wnt10a,
Notch3 receptors. (C) Mammosphere assay with P4 cells from WT and Dll1cKO and Wnt16 immunofluorescence intensity in indicated groups from (I) to (K).
MECs with and without WT and Dll1cKO macrophages. n = 5 samples. (M) Mammosphere assay of WT P4 cells with and without coculture with
(D) Mammosphere assay with P4-Dll1+ cells with and without macrophages macrophages along with Wnt inhibitor Dkk1. n = 4 samples. For macrophage
and treatment of antibodies against Dll1, Notch2, and Notch3. n = 3 samples. isolation, a combination of F4/80 and CD140 antibodies was used. (N) Model
(E) Fold change in gene expression of the most differentially expressed genes showing cross-talk of a Dll1+ MaSC–enriched population with macrophages
encoding secreted factors or extracellular proteins between macrophage through Notch and Wnt signaling. All qRT-PCR experiments were performed
populations from WTand Dll1cKO mammary glands. (F to H) Wnt3, Wnt10a, and three times. Data are presented as mean ± SD. ***P < 0.001, **P < 0.01,
Wnt16 mRNA levels in F4/80+ cells after coculture with P4-Dll1+ cells with and and *P < 0.05 by Student’s t test in (A) to (D), (F) to (H), (L), and (M).
without blocking antibody against Dll1, Notch2, and Notch3. n = 3 samples. Scale bars, 10 mm in (I) to (K).
in macrophages (Fig. 7, F to L). Furthermore, results collectively indicate reciprocal interac- Notch and Wnt pathways have been reported to
stimulation of mammosphere forming activity tions between macrophages and MaSCs: Dll1- be key oncogenic pathways in breast cancer and
of P4 cells by macrophages was largely blocked Notch signaling from MaSCs to macrophages macrophages are a major component of the tumor
by the presence of the Wnt signaling inhibitor maintains the number and niche-related activ- microenvironment, future studies of Notch-Wnt–
Dkk1 (Dickkopf-1) in the coculture (Fig. 7M). These ity of the macrophages; conversely, the ma- dependent interaction between MaSCs and macro-
results indicate that the regulation of the MaSC crophageal niche is crucial for sustaining the phages may provide insights into tumor initiation
population by macrophages is likely mediated MaSC pool. and progression in breast cancer.
by increased Wnt ligand production by macro- Consistent with these notions, we established
phages in response to Dll1-Notch signaling. and used a 3D coculture assay to show that Dll1+ Materials and methods
MaSC-enriched basal cells interact with stromal Animal studies
Discussion macrophages through Notch2 and -3 receptors. Animal procedures were conducted in compli-
Intercellular signaling between stem cells and This organoid coculture system aims to mimic ance with Institutional Animal Care and Use
stromal cells within the stem cell niche dictates the point at which there is substantial contact Committee (IACUC) of Princeton University,
stem cell number and function, including self- between MaSCs and macrophages. With the use University of Pennsylvania, and Memorial Sloan
renewal activity. Although other stem cell niches of this system, coculturing MaSCs with macro- Kettering Cancer Center. Dll1 floxed mice (28),
have been extensively studied, the current inability phages resulted in a significant increase in stem Dll1-GFP-IRES-Cre-ERT2 mice (31), and CD11c-
to identify MaSCs within their associated stromal cell activity of Dll1+ cells. Conditional knockout Cre;RbpjkcKO mice (35) have been described pre-
niche has hindered similar studies in the mam- of Dll1 in the MECs not only renders the Dll1cKO viously. The Dll1-mCherry transgenic mice were
mary gland. In our study, we first showed that basal cells less responsive to macrophage activa- generated using a genomic BAC clone with
Dll1 is a marker and crucial regulator of MaSCs. tion but also reduces the potency of macrophages mCherry cDNA inserted after the start codon
Using transplantation assays, we showed that from Dll1cKO mice in supporting MaSC function. of Dll1. tdTomato mice [B6;129S6-Gt(ROSA)
Dll1+/hi cells are enriched for MaSCs with in- Macrophages isolated from Dll1cKO mice have 26Sortm9(CAG-tdTomato)Hze/J], Actin-GFP, Actin-
mammary macrophages with P4-Dll1mCherry+ cells, by sorting and plated on gelatin-coated plates for manufacturer. For EdU assay in FACS along with
see schematic in fig. S9H for the detailed process. 3 to 5 days. Dll1 (0.75 mg/ml) or Notch2 or Notch3 other antibodies, samples were first stained with
Frequency of MaSCs in the transplanted cell (1.5 mg/ml) blocking antibodies were added alone CD24, CD29, Ter119, CD45, and CD31 antibodies,
suspension was calculated using L-calc software or in combination followed by the addition of fixed and then stained with EdU following the
(StemCell Technologies) or ELDA (extreme limit- control (no P4 cell), P4- Dll1mCherry+ cells (P4-Dll1+) protocol from manufacturer (Invitrogen). For
ing dilution assay) (30, 45). Single-hit model was or P4- Dll1mcherry− cells (P4-Dll1) for 90 min. Cells immunofluorescence, paraffin embedded sections
also tested using ELDA and value of slope was 1. are then washed, trypsinized and sorted for either were first rehydrated using standard protocol and
MaSC abundances were assumed to follow a mCherry+ and mCherry- population or mCherry+ then stained with EdU followed by other anti-
Poisson distribution in LDAs, and generalized and GFP+ population followed by RNA isolation bodies following the manufacturer’s instructions.
linear models utilizing a log-log link function for gene expression analysis. For IF with Wnt
were used to derive repopulation frequency param- antibodies, macrophages were cocultured for qRT-PCR analyses
eters. Self-renewal activity of MaSCs after trans- 5 hours with P4-Dll1mCherry+ cells. Coculture was Total RNA was isolated from primary cells using
plantation was tested by their ability to regenerate washed extensively to remove P4 cells. Attached Qiagen RNA extraction kit in accordance with
functional mammary glands in virgin mouse. macrophages were stained with respective Wnt the manufacturer’s instructions. Real-time RT-
antibodies. PCR was performed on ABI 7900 96 HT series
Clondronate liposome (CL) assay and StepOne Plus PCR machines (Applied Bio-
The CLs are nontoxic until ingested by macro- Protein extraction and Western system) using SYBR Green Supermix (Bio-Rad
phages. Once ingested, they are then broken down blot analysis Laboratories). The gene-specific primer sets were
by liposomal phospholipases to release the drug Proteins were extracted from primary epithelial used at a final concentration of 0.2 mM, and their
that subsequently induces cell death in macro- cell cultures and cell lines in RIPA buffer as pre- sequences are listed in table S3. All qRT-PCR
phages by apoptosis (32). For systemic treatment viously described (27). Western blot analysis was assays were performed in duplicate in at least
of Dll1mCherry+ reporter mice, the animals were performed using the standard protocol. Anti- three independent experiments using three dif-
Both are defined by significantly up-regulated (2013). doi: 10.1007/s10911-013-9284-6; 24. N. M. Badders et al., The Wnt receptor, Lrp5, is expressed by
genes (P < 0.05 and FC > 3) in MaSC-enriched pmid: 23624881 mouse mammary stem cells and is required to maintain the
3. M. Shackleton et al., Generation of a functional mammary basal lineage. PLOS ONE 4, e6594 (2009). doi: 10.1371/journal.
subpopulations from MECs of WT mice. Among gland from a single stem cell. Nature 439, 84–88 (2006). pone.0006594; pmid: 19672307
which, the “MaSC signature and luminal cell doi: 10.1038/nature04372; pmid: 16397499 25. R. D. Rajaram et al., Progesterone and Wnt4 control mammary
signature from Chakrabarti et al., 2014” (Fig. 3F) 4. J. Stingl et al., Purification and unique properties of mammary stem cells via myoepithelial crosstalk. EMBO J. 34, 641–652
is derived from the microarray data collected epithelial stem cells. Nature 439, 993–997 (2006). (2015). doi: 10.15252/embj.201490434; pmid: 25603931
from our lab as described in previous study (26) pmid: 16395311 26. R. Chakrabarti et al., DNp63 promotes stem cell activity in
5. A. C. Rios, N. Y. Fu, G. J. Lindeman, J. E. Visvader, In situ mammary gland development and basal-like breast cancer by
(GSE47493). The genes showing >3 folds up- identification of bipotent stem cells in the mammary gland. enhancing Fzd7 expression and Wnt signalling. Nat. Cell Biol.
regulation in P4 comparing to both P5 and P6 Nature 506, 322–327 (2014). doi: 10.1038/nature12948; 16, 1004–1015 (2014). doi: 10.1038/ncb3040; pmid: 25241036
of WT mice were included in the MaSC gene set. pmid: 24463516 27. Y. S. Choi, R. Chakrabarti, R. Escamilla-Hernandez, S. Sinha,
For luminal signature, the genes showing >3 6. A. Van Keymeulen et al., Distinct stem cells contribute to Elf5 conditional knockout mice reveal its role as a master
mammary gland development and maintenance. Nature 479, regulator in mammary alveolar development: Failure of Stat5
folds up-regulation in P5 comparing to both P4 189–193 (2011). doi: 10.1038/nature10573; pmid: 21983963 activation and functional differentiation in the absence of Elf5.
and P6 of WT mice were included in the luminal 7. J. L. Inman, C. Robertson, J. D. Mott, M. J. Bissell, Mammary Dev. Biol. 329, 227–241 (2009). doi: 10.1016/
gene set. The other MaSC and luminal cell sig- gland development: Cell fate specification, stem cells and j.ydbio.2009.02.032; pmid: 19269284
nature is derived from published dataset (30). For the microenvironment. Development 142, 1028–1042 (2015). 28. K. Hozumi et al., Delta-like 1 is necessary for the generation of
doi: 10.1242/dev.087643; pmid: 25758218 marginal zone B cells but not T cells in vivo. Nat. Immunol. 5,
gene expression in macrophages, the genes show- 638–644 (2004). doi: 10.1038/ni1075; pmid: 15146182
8. R. C. Hovey, L. Aimo, Diverse and active roles for adipocytes
ing >3 folds up-regulation in mammary resting during mammary gland growth and function. J. Mammary 29. K. Miyoshi et al., Signal transducer and activator of
macrophages comparing to resting peritoneal Gland Biol. Neoplasia 15, 279–290 (2010). doi: 10.1007/ transcription (Stat) 5 controls the proliferation and
macrophage of WT mice were included. s10911-010-9187-8; pmid: 20717712 differentiation of mammary alveolar epithelium. J. Cell Biol.
9. M. F. Gregor et al., The role of adipocyte XBP1 in metabolic 155, 531–542 (2001). doi: 10.1083/jcb.200107065
regulation during lactation. Cell Rep. 3, 1430–1439 (2013). pmid: 11706048
Lineage tracing
doi: 10.1016/j.celrep.2013.03.042; pmid: 23623498 30. M. L. Asselin-Labat et al., Control of mammary stem cell
Lineage tracing experiment was performed follow- 10. K. L. Schwertfeger, J. M. Rosen, D. A. Cohen, Mammary gland function by steroid hormone signalling. Nature 465, 798–802
cells in vivo reveals stem cell dynamics throughout pregnancy. supported by a DOD Postdoctoral Fellowship (BC103740) and a Competing interests: The authors declare no competing interests.
PLOS ONE 4, e8035 (2009). doi: 10.1371/journal.pone.0008035; NCI-K22 grant (K22CA193661) to R.C.; a Susan G. Komen Data and material availability: All data needed to understand
pmid: 19946375 Fellowship (PDF15332075) to T.C.-T.; grants from the Brewster and assess the conclusions of this research are available in the
47. V. K. Mootha et al., Integrated analysis of protein composition, Foundation, the Breast Cancer Research Foundation, DOD main text and supplementary materials. Microarray data reported
tissue diversity, and gene regulation in mouse mitochondria. (BC123187), and NIH (R01CA141062) to Y.K; and NIH grants (R01 herein have been deposited at the NCBI Gene Expression
Cell 115, 629–640 (2003). doi: 10.1016/S0092-8674(03) CA198280-01 and P30 CA008748) to M.O.L. This research was Omnibus GSE77504.
00926-7; pmid: 14651853 also supported by the Genomic Editing and Flow Cytometry Shared
48. A. Subramanian et al., Gene set enrichment analysis: A Resources of the Cancer Institute of New Jersey (P30CA072720). SUPPLEMENTARY MATERIALS
knowledge-based approach for interpreting genome-wide Author contributions: R.C. and Y.K. designed all experiments. www.sciencemag.org/content/360/6396/eaan4153/suppl/DC1
expression profiles. Proc. Natl. Acad. Sci. U.S.A. 102, R.C., S.K., T.C.-T., X.H., A.C., J.H., and J.P. performed the Figs. S1 to S10
15545–15550 (2005). doi: 10.1073/pnas.0506580102; experiments. C.D. and J.J.G provided technical advice and helped Tables S1 to S3
pmid: 16199517 with FACS analysis and sorting. Y.W. performed all microarray
and statistical analyses. B.N. and M.O.L. participated in RbpjkcKO- 10 April 2017; resubmitted 4 January 2018
ACKN OW LEDG MEN TS related experiments. J.G., J.H.v.E., I.A., and H.C. provided mouse Accepted 2 May 2018
We thank L. King (University of Pennsylvania) for critical reading of strains and advice. R.C and Y.K. wrote the manuscript. All authors Published online 17 May 2018
the manuscript and helpful discussions. Funding: This work was discussed the results and commented on the manuscript. 10.1126/science.aan4153
impact of early Bronze Age steppe ent amounts of Eastern hunter-gatherer (EHG)
and Ancient East Asian genetic ancestry rep-
resented by Baikal_EN.
expansions into Asia In Anatolia, Bronze Age samples, includ-
ing from Hittite speaking settlements asso-
Peter de Barros Damgaard et al.* ciated with the first written evidence of IE
languages, show genetic continuity with pre-
INTRODUCTION: According to the common- Central Asia, whereas direct evidence for ceding Anatolian Copper Age (CA) samples
ly accepted “steppe hypothesis,” the initial Yamnaya equestrianism remains elusive. and have substantial Caucasian hunter-gatherer
spread of Indo-European (IE) languages into (CHG)–related ancestry but no evidence of
both Europe and Asia took place with migra- RATIONALE: We investigated the genetic im- direct steppe admixture.
tions of Early Bronze Age Yamnaya pasto- pact of Early Bronze Age migrations into Asia In South Asia, we identified at least two
ralists from the Pontic-Caspian steppe. This is and interpret our findings in relation to the distinct waves of admixture from the west,
EHG CA EMBA EMBA LNBA EN CONCLUSION: Our findings reveal that the
early spread of Yamnaya Bronze Age pasto-
ralists had limited genetic impact in Anatolia
as well as Central and South Asia. As such, the
Asian story of Early Bronze Age expansions
differs from that of Europe. Intriguingly, we
Sidelkino Botai find that direct descendants of Upper Paleo-
CHG MLBA IA CA EBA Central Okunevo lithic hunter-gatherers of Central Asia, now
Steppe Baikal extinct as a separate lineage, survived well
into the Bronze Age. These groups likely en-
gaged in early horse domestication as a prey-
Yamnaya
Anatolia
CA - Copper Age route transition from hunting to herding, as
Steppe CHG - Caucasus Hunter-Gatherer otherwise seen for reindeer. Our findings fur-
Namazga EHG - Eastern Hunter-Gatherer
EBA - Early Bronze Age
ther suggest that West Eurasian ancestry en-
EMBA- Early-Middle Bronze Age tered South Asia before and after, rather than
South Asian EN - Early Neolithic during, the initial expansion of western steppe
populations LNBA - Late Neolithic/Bronze Age pastoralists, with the later event consistent
MLBA- Middle-Late Bronze Age
North-west IA - Iron Age
with a Late Bronze Age entry of IE languages
N - Neolithic into South Asia. Finally, the lack of steppe
N CA EBA MLBA IA Ottoman
ancestry in samples from Anatolia indicates
East that the spread of the earliest branch of IE
languages into that region was not associ-
ated with a major population migration from
South the steppe.
▪
Model-based admixture proportions for selected ancient and present-day individuals, The list of author affiliations is available in the full article online.
assuming K = 6, shown with their corresponding geographical locations. Ancient groups *These authors contributed equally to this work.
are represented by larger admixture plots, with those sequenced in the present work sur- †Corresponding author. Email: rd109@cam.ac.uk (R.D.);
ewillerslev@snm.ku.dk (E.W.)
rounded by black borders and others used for providing context with blue borders. Present-day Cite this article as P. de Barros Damgaard et al., Science 360,
South Asian groups are represented by smaller admixture plots with dark red borders. eaar7711 (2018). DOI: 10.1126/science.aar7711
The first horse herders and the Bronze Age (3000 to 2500 BCE) cultural dynam-
ics (9). For Asia, however, several conflicting inter-
pretations have long been debated. These concern
impact of early Bronze Age steppe the origins and genetic composition of the local
Asian populations encountered by the Yamnaya-
T
Copper Age individuals (~3500 to 3300 BCE) from
he vast grasslands making up the Eur- “steppe hypothesis,” this expansion of groups Botai in northern Kazakhstan (Botai_CA; 13.6X,
asian steppe zones, from Ukraine through in the western steppe related to the Yamnaya 3.7X, and 3X coverage, respectively); 1 Early Bronze
Kazakhstan to Mongolia, have served as a and Afanasievo cultures was associated with Age (~2900 BCE) Yamnaya sample from Karagash,
crossroad for human population movements the spread of Indo-European (IE) languages into Kazakhstan (16) (YamnayaKaragash_EBA; 25.2X);
during the last 5000 years (1–3), but the dy- Europe and Asia (1, 2, 4, 6). The peoples who 1 Mesolithic (~9000 BCE) EHG from Sidelkino,
namics of its human occupation—especially of formed the Yamnaya and Afanasievo cultures Russia (SidelkinoEHG_ML; 2.9X); 2 Early/Middle
the earliest period—remain poorly understood. The belonged to the same genetically homogeneous Bronze Age (~2200 BCE) central steppe individ-
domestication of the horse at the transition from population, with direct ancestry attributed to both uals (~4200 BP) (CentralSteppe_EMBA; 4.5X and
the Copper Age to the Bronze Age, ~3000 BCE, Copper Age (CA) western steppe pastoralists, de- 9.1X average coverage, respectively) from burials
enhanced human mobility (4, 5) and may have scending primarily from the European Eastern at Sholpan and Gregorievka that display cultural
triggered waves of migration. According to the hunter-gatherers (EHG) of the Mesolithic and similarities to Yamnaya and Afanasievo (12);
1
Centre for GeoGenetics, Natural History Museum, University of Copenhagen, Copenhagen, Denmark. 2Wellcome Trust Sanger Institute, Wellcome Genome Campus, Cambridge CB10 1SA, UK.
3
Department of Genetics, University of Cambridge, Downing Street, Cambridge CB2 3EH, UK. 4Department of Nordic Studies and Linguistics, University of Copenhagen, Copenhagen, Denmark.
5
Leiden University Centre for Linguistics, Leiden University, Leiden, Netherlands. 6Department of Near Eastern Languages and Civilizations, Harvard University, Cambridge, MA, USA. 7Department
of Bio and Health Informatics, Technical University of Denmark, Kongens Lyngby, Denmark. 8Shejire DNA project, Abai ave. 150/230, 050046 Almaty, Kazakhstan. 9Institute of Archaeology and
Steppe Civilization, Al-Farabi Kazakh National University, Almaty, 050040, Kazakhstan. 10S. Toraighyrov Pavlodar State University, Joint Research Center for Archeological Studies named after
A.Kh. Margulan, Pavlodar, Kazakhstan. 11Department of Human Genetics, University of Chicago, Chicago, IL, USA. 12Saryarkinsky Institute of Archaeology, Buketov Karaganda State University,
Karaganda. 100074, Kazakhstan. 13Department of Anthropology, University of Alaska, Fairbanks, AK, USA. 14The Institute of Forensic Sciences, Istanbul University, Istanbul, Turkey. 15Department
of Genetics, Hazara University, Garden Campus, Mansehra, Pakistan. 16Department of Historical Studies, University of Gothenburg, 40530 Göteborg, Sweden. 17Peter the Great Museum of
Anthropology and Ethnography (Kunstkamera) RAS, St. Petersburg, Russia. 18Institute for the History of Material Culture, Russian Academy of Sciences, St. Petersburg, Russia. 19Japanese
Institute of Anatolian Archaeology, Kaman, Kırşehir, Turkey. 20Department of Archaeology, Faculty of Arts, Gazi University, Ankara, Turkey. 21Center of Omic Sciences, Islamia College, Peshawar,
Pakistan. 22Department of Archaeology, University of Exeter, Exeter, EX4 4QE, UK. 23Department of Archeology, Hazara University, Garden Campus, Mansehra, Pakistan. 24Directorate of
Archaeology and Museums Government of Khyber Pakhtunkhwa, Pakistan. 25Archaeological Museum Harappa at Archaeology Department Govt. of Punjab, Pakistan. 26Institute of Archaeology
and Ethnography, Siberian Branch of the Russian Academy of Sciences, Academician Lavrent’iev Ave. 17, Novosibirsk, 630090, Russia. 27Department of History, Irkutsk State University,
Karl Marx Street 1, Irkutsk 664003, Russia. 28Department of Ecology and Evolution, University of Chicago, Chicago, IL, USA. 29Department of Anthropology, University of Alberta, Edmonton,
Alberta, T6G 2H4, Canada. 30Laboratoire d’Anthropobiologie Moléculaire et d’Imagerie de Synthèse, CNRS UMR 5288, Université deToulouse, Université Paul Sabatier, 31000 Toulouse, France.
31
Departments of Integrative Biology and Statistics, University of Berkeley, Berkeley, CA, USA. 32Department of Zoology, University of Cambridge, Cambridge, UK.
*These authors contributed equally to this work.
†Corresponding author. Email: rd109@cam.ac.uk (R.D.); ewillerslev@snm.ku.dk (E.W.)
pairwise covariances. We find that Botai_CA, Furthermore, we observe a shift in genetic ances- tinct origins of central steppe hunter-herders
CentralSteppe_EMBA, Okunevo_EMBA, and try between the Early Neolithic (Baikal_EN) and related to Botai of the central steppe and those
Baikal populations are deeply separated from the Late Neolithic/Bronze Age hunter-gatherers related to Altaian hunter-gatherers of the east-
other ancient and present-day populations and (Baikal_LNBA) (Fig. 2A), with the Baikal_LNBA ern steppe (30). Furthermore, this population
are best modeled as mixtures in different pro- cluster showing admixture from an ANE-related structure, which is best described as part of the
portions of ANE ancestry and an Ancient East source. We estimate the ANE related ancestry in the ANE metapopulation, persisted within Inner
Asian (AEA) ancestry component represented by Baikal_LNBA to be ~5 to 11% (qpAdm) (table S12) Asia from the Upper Paleolithic to the end of
Baikal_EN, with mixing times dated to ~5000 BCE. (2), using MA1 as a source of ANE, Baikal_EN as a the Early Bronze Age. In the Baikal region, the
Although some modern Siberian samples lie source of AEA, and a set of six outgroups. How- results show that at least two genetic shifts oc-
under the Baikal samples in Fig. 2A, these are ever, neither MA1 nor any of the other steppe pop- curred: first, a complete population replacement
separated out in a more limited PCA, involving ulations lie in the direction of Baikal_LNBA from of the Upper Paleolithic hunter-gatherers be-
just those populations and the ancient sam- Baikal_EN on the PCA plot (fig. S23). This sug- longing to the ANE by Early Neolithic communi-
ples (fig. S23). Our momi model infers that the gests that the new ANE ancestry in Baikal_LNBA ties of Ancient East Asian ancestry, and second,
ANE lineage separated ~15,000 years ago in stems from an unsampled source. Given that this an admixture event between the latter and addi-
the Upper Paleolithic from the EHG lineage to source may have harbored East Asian ances- tional members of the ANE clade, occurring during
the west, with no independent drift assigned try, the contribution may be larger than 10%. the 1500-year period that separates the Neolithic
to MA1. This suggests that MA1 may represent These serial changes in the Baikal popula- from the Early Bronze Age. These genetic shifts
their common ancestor. Similarly, the AEA lin- tions are reflected in Y-chromosome lineages complement previously observed severe cultural
eage to the east also separated ~15,000 years ago, (Fig. 5A, figs. S24 to S27, and tables S13 and changes in the Baikal region (18–22).
with the component that leads to Baikal_EN S14). MA1 carries the R haplogroup, whereas
and the AEA component of the steppe sepa- the majority of Baikal_EN males belong to N Relevance for history of
rating from the lineage leading to present- lineages, which were widely distributed across horse domestication
day East Asian populations represented by Han Northern Eurasia (29), and the Baikal_LNBA The earliest unambiguous evidence for horse
Chinese (figs. S19 to S21). The ANE and AEA males all carry Q haplogroups, as do most of husbandry is from the Copper Age Botai hunter-
lineages themselves are estimated as having the Okunevo_EMBA as well as some present- herder culture of the central steppe in Northern
separated approximately 40,000 years ago, rela- day Central Asians and Siberians. Mitochon- Kazakhstan ~3500 to 3000 BCE (5, 10, 23, 31–33).
tively soon after the peopling of Eurasia by mod- drial haplogroups show less turnover (Fig. 5B There was extensive debate over whether Botai
ern humans. and table S15), which could either indicate male- horses were hunted or herded (33), but more
Because the ANE MA1 sample comes from the mediated admixture or reflect bottlenecks in the recent studies have evidenced harnessing and
same cis-Baikal region as the AEA-derived Neo- male population. milking (10, 17), the presence of likely corrals,
lithic samples analyzed here, we document evi- The deep population structure among the and genetic domestication selection at the horse
dence for a population replacement between local populations in Inner Asia around the Copper TRPM1 coat-color locus (32). Although horse hus-
the Paleolithic and the Neolithic in this region. Age/Bronze Age transition is in line with dis- bandry has been demonstrated at Botai, it is
Namazga Namazga
Namazga + S. Indian + S. Indian TurkmenistanIA
Steppe + S. Indian + SteppeMLBA + Xiongnu + Xiongnu Anatolia
o_ BA
Xi his in
A
rk bi
Ja _IA
ar r
ep Ko iya
Gu A
Th in
An W r
Pa s
Us EM A
lB h n
hm i
M Iyer
Ta gno A
a e
am ma BA
ka BA
am tiv N
ka N
ga N
jik Tna_IA
at An _M A
An ato lia_ _N
a_ ol A
to IA
Ok tepp ota MA1L
Sh Kur da_ A
ko da A
an
ev EM A
m Iran N
at lia_ CA
ga Khmi l
Pa ha
gn ni
riB riptia
Kallan
Ir ar
m ufz i
en O or
Goula
_M a
Br kh i
Ko hm i
u
Sa Ho
iev _E G
Pa nd
irh l
An at HG
ra ur
a e
GuUth Yus lan
an a
a
ra atr
i
ja
s_ jik
m
ist ng
lS B _M
pe rw
ar
EB
_C
o_ B
Ya LB
B
oli at LB
tI B
Lo UstI EB
un e_ i_C
Sh mo _L
an _E
az _
_E
on ta
as sh CH
HGEH
nt
d
Ot ia_
ipu ma
m
at
An at olia
an _E
oli E
u
a
_
B
T
a
E
an ga
t i
no
Ta
Af ra
m
r
Na
en
j
lki
St
rk
ya Ka
an
tB
de
Tu
M
ra
na a
An
es
Si
m ay
nt
W
Ya amn
Ce
Y
Fig. 3. Model-based clustering analysis of present-day and ancient individuals assuming K = 6 ancestral components. The main ancestry
components at K = 6 correlate well with CHG (turquoise), a major component of Iran_N, Namazga_CA and South Asian clines; EHG (pale blue), a
component of the steppe cline and present in South Asia; East Asia (yellow ochre), the other component of the steppe cline also in Tibeto-Burman South
Asian populations; South Indian (pink), a core component of South Asian populations; Anatolian_N (purple), an important component of Anatolian
Bronze Age and Steppe_MLBA; Onge (dark pink) forms its own component.
also now clear from genetic studies that this was present-day Eastern Europeans, one of the two gion Bronze Age (~2200 to 1800 BCE) samples
not the source of modern domestic horse stock Botai_CA males belongs to the basal N lineage, (fig. S14).
Fig. 5. Y-chromosome and mitochondrial lineages identified in ancient phylogenetic tree estimated with 182 present-day and ancient individ-
and present-day individuals. (A) Maximum likelihood Y-chromosome uals. The phylogenies displayed were restricted to a subset of clades
phylogenetic tree estimated with data from 109 high-coverage samples. relevant to the present work. Columns represent archaeological groups
Dashed lines represent the upper bound for the inclusion of 42 low- analyzed in the present study, ordered by time, and colored areas
coverage ancient samples in specific Y-chromosome clades on the basis indicate membership of the major Y-chromosome and mitochondrial
of the lineages identified. (B) Maximum likelihood mitochondrial DNA (mtDNA) haplogroups.
and admixtures from the Upper Paleolithic to with historical models of cultural hybridity and 8. J. P. Mallory, D. Q. Adams, The Oxford introduction to
the Bronze Age (1, 2, 27, 65, 66), with a strong “middle ground” in a multicultural and multi- Proto-Indo-European and the Proto-Indo-European world
(Oxford Univ. Press, Oxford, 2006).
influence across the continent from the Early lingual but genetically homogeneous Bronze 9. K. Kristiansen et al., Re-theorising mobility and the formation
Bronze Age (~3000 to 2500 BCE) western steppe Age Anatolia (69, 70). of culture and language among the Corded Ware Culture
Yamnaya. In contrast, for Central Asia, continu- Current linguistic estimations converge on in Europe. Antiquity 91, 334–347 (2017). doi: 10.15184/
ity is observed from the Upper Paleolithic to the dating the Proto-Anatolian split from residual aqy.2017.17
10. A. K. Outram et al., The earliest horse harnessing and milking.
end of the Copper Age (~3500 to 3000 BCE), with PIE to the late fifth or early fourth millennia Science 323, 1332–1335 (2009). doi: 10.1126/science.1168594
descendants of Paleolithic hunter-gatherers per- BCE (53, 71) and place the breakup of Anatolian 11. E. Kaiser, Der Übergang zur Rinderzucht im nördlichen
sisting as largely isolated populations after the IE inside Turkey before the mid-third millennium Schwarzmeerraum. Godišnjak Centar za Balkanološka
Yamnaya and Afanasievo pastoralist migrations. (51, 54, 72). In (49) we present new onomastic Ispitivanja 39, 23–34 (2010).
12. D. Reich, K. Thangaraj, N. Patterson, A. L. Price, L. Singh,
Instead of western pastoralists admixing with material (73) that pushes the period of Proto- Reconstructing Indian population history. Nature 461, 489–494
or replacing local groups, we see groups with East Anatolian linguistic unity even further back in (2009). doi: 10.1038/nature08365
Asian ancestry replacing ANE populations in the time. We cannot at this point reject a scenario in 13. K. Kristiansen, Europe Before History (Cambridge Univ. Press,
Lake Baikal region. Thus, unlike in Europe, the which the introduction of the Anatolian IE lan- Cambridge, 1998).
14. See the supplementary materials.
hunter/gathering/herding groups of Inner Asia guages into Anatolia was coupled with the CHG- 15. A. K. Outram, A. Polyakov, A. Gromov, V. Moiseyev, A. W. Weber,
were much less affected by the Yamnaya and derived admixture before 3700 BCE, but note V. I. Bazaliiskii, O. I. Goriunova, Supplementary discussion of
Afanasievo expansion. This may be due to the that this is contrary to the standard view that the archaeology of Central Asian and East Asian Neolithic
rise of early horse husbandry, likely initially orig- PIE arose in the steppe north of the Caucasus (4) to Bronze Age hunter-gatherers and early pastoralists,
including consideration of horse domestication. 10.5281/
inated through a local “prey route” (40) adap- and that CHG ancestry is also associated with zenodo.1240521 (9 May 2018).
tation by horse-dependent hunter-gatherers at several non-IE-speaking groups, historical and 16. V. V. Evdokimov, V. G. Loman, Raskopki yamnogo kurgana v
Botai. Work on ancient horse genomes (32) in- current. Indeed, our data are also consistent with Karagandinskoj oblasti, in Voprosy Arheologii Central'nogo i
dicates that Botai horses were not the main the first speakers of Anatolian IE coming to the Severnogo Kazahstana (Karaganda, 1989), pp. 34–46.
33. S. L. Olsen, S. Grant, A. M. Choyke, L. Bartosiewicz, Horses and 52. J. Puhvel, “Whence the Hittite, whither the Jonesian vision?” AC KNOWLED GME NTS
Humans: the Evolution of Human-Equine Relationships in Sprung from a common source, S. M. Lamb, E. D. Mitchell, Eds. We thank K. Magnussen, L. Petersen, and C. Mortensen at the Danish
(British Archeological Reports, Oxford, 2006). (Stanford Univ. Press, Stanford, Calif., 1991), pp. 51–66. National Sequencing Centre for conducting the sequencing and
34. N. M. Myres et al., A major Y-chromosome haplogroup R1b 53. B. Darden, “On the question of the Anatolian origin of P. Reimer and S. Hoper at the 14Chrono Center Belfast for providing
Holocene era founder effect in Central and Western Europe. Indo-Hittite” in Greater Anatolia and the Indo-Hittite Language the AMS dating. We thank S. Ellingvåg, B. E. Heyerdahl, and the
Eur. J. Hum. Genet. 19, 95–101 (2011). doi: 10.1038/ Family, R. Drews, Ed. (Journal of Indo-European Studies, Explico-Historical Research Foundation team, as well as N. Thompson,
ejhg.2010.146 Washington, 2001), pp. 184–228. for involvement in field work. We thank the Turkish Ministry of
35. D. J. Lawson, G. Hellenthal, S. Myers, D. Falush, Inference of 54. C. Watkins, “An Indo-European linguistic area and its Culture and Tourism, Kaman-Kalehöyük Archeology Museum, and
population structure using dense haplotype data. PLOS Genet. characteristics: Anatolia” in Areal Diffusion and Genetic Nevşehir Museum for permission to use samples of Kaman-Kalehöyük
8, e1002453 (2012). doi: 10.1371/journal.pgen.1002453 Inheritance: Problems in Comparative Linguistics, and Ovaören. We thank J. Stenderup, P. V. Olsen, and T. Brand for
36. S. Schiffels et al., Iron Age and Anglo-Saxon genomes from A.Y. Aikhenvald, R. M. W. Dixon, Eds. (Oxford Univ. Press, technical assistance in the laboratory. We thank T. Korneliussen
East England reveal British migration history. Nat. Commun. 7, 2001), pp. 44–63. for helpful discussions. We thank St. John’s College, Cambridge, for
10408 (2016). doi: 10.1038/ncomms10408 55. H. C. Melchert, Ed., The Luwians (Brill, 2003). providing the setting for fruitful scientific discussions. We thank all
37. A. A. Kovalev, D. Erdenebaatar, “Discovery of new cultures of 56. F. Kortlandt, C. C. Uhlenbeck on Indo-European, Uralic and involved archaeologists, historians, and collaborators from Pakistan
the Bronze Age in Mongolia according to the data obtained Caucasian. Hist. Sprachforsch 122, 39–47 (2009). who assisted I.U. in the field. We thank G. Baimbetov (Shejire DNA),
by the International Central Asian Archaeological Expedition” in 57. S. M. Winn, Burial evidence and the kurgan culture in Eastern I. Baimukhan, B. Daulet, A. Kusaev, A. Kopbassarova, Y. Yousupov,
Current Archaeoligcal Research in Mongolia, J. Bemmann, Anatolia c. 3000 BC: An interpretation. J. Indo-Eur. Stud. 9, M. Akchurin, and V. Volkov for important assistance in the field.
H. Parzinger, E. Pohl, D. Tseveendorzh, Eds. (Vor- und 113–118 (1981). Funding: The study was supported by the Lundbeck Foundation
Frühgeschichtliche Archäologie Uni-Bonn, Bonn, 2009), 58. I. Lazaridis et al., Genetic origins of the Minoans and (E.W.), the Danish National Research Foundation (E.W.), and KU2016
pp. 149–170. Mycenaeans. Nature 548, 214–218 (2017). (E.W.). Research at the Sanger Institute was supported by the
38. G. Larson et al., Rethinking dog domestication by integrating 59. L. Rahmstorf, “Indications of Aegean-Caucasian relations Wellcome Trust (grant 206194). R.M. was supported by an EMBO
genetics, archeology, and biogeography. Proc. Natl. Acad. during the third millennium BC,” in Von Maikop bis Trialeti. Long-Term Fellowship (ALTF 133-2017). J.K. was supported by the
Sci. U.S.A. 109, 8878–8883 (2012). doi: 10.1073/ Gewinnung und Verbreitung von Metallen und Obsidian in Human Frontiers Science Program (LT000402/2017). Botai fieldwork
pnas.1203005109 Kaukasien im 4.-2. Jt. v. Chr., S. Hansen, A. Hauptmann, was supported by University of Exeter, Archeology Exploration Fund,
39. K. H. Røed et al., Genetic analyses reveal independent I. Motzenbäcker, E. Pernicka, Eds. (Verlag Rudolf Habelt, Bonn, and N. Thompson, Clearwater Documentary. A.B. was supported
domestication origins of Eurasian reindeer. Proc. Biol. Sci. 275, 2010), pp. 263–295. by NIH grant 5T32GM007197-43. G.K. was funded by Riksbankens
1849–1855 (2008). doi: 10.1098/rspb.2008.0332 60. A. T. Smith, The Political Machine: Assembling Sovereignty in Jubileumsfond and European Research Council. M.P. was funded by
A
and Snu17). The protein components of U2 snRNP
ssembly and activation of the spliceosome Structure elucidation of the yeast spliceosome are distributed in the SF3a complex (Prp9, Prp11,
take place in an ordered process (1–3). First, has led to major advances in the mechanistic and Prp21), the SF3b complex (Cus1, Hsh49,
the 5′ splice site (5′SS) and the branch point understanding of pre-mRNA splicing (8–10). Hsh155, Rse1, Rds3, and Ysf3), and the U2 core
sequence (BPS) are recognized by the U1 Since determination of the 3.6-Å structure of (Lea1, Msl1, and the U2 Sm ring). Proteins of the
and U2 small nuclear ribonucleoproteins the Schizosaccharomyces pombe ILS complex tri-snRNP include 11 in U4 snRNP (Prp3, Prp4,
(snRNPs), respectively, through duplex formation in 2015 (11, 12), cryo–electron microscopy (cryo- Prp31, Snu13, and the U4 Sm ring), 11 in U5 snRNP
with U1 and U2 snRNAs in the pre-spliceosome EM) structures at atomic or near-atomic resolu- (Brr2, Dib1, Prp8, Snu114, and the U5 Sm ring), the
(known as the A complex). Then, the A complex tions have been reported for the Saccharomyces U6 LSm ring, and 2 tri-snRNP–specific proteins
associates with the U4/U6.U5 tri-snRNP to form cerevisiae B, Bact, C, C*, P, and ILS complexes. (Prp6 and Snu66).
the pre-B complex, the first fully assembled splice- Here we report the cryo-EM structure of the U1 snRNP is relatively compact and well defined
osome that contains all five snRNPs (4). The S. cerevisiae pre-B complex at resolutions of by the 3.3-Å EM map (fig. S6). U2 snRNP has an
adenosine triphosphatase (ATPase)/helicase Prp28 3.3, 3.6 to 4.6, and 3.4 Å for U1 snRNP, U2 snRNP, elongated shape and exhibits considerable inter-
drives the dissociation of U1 snRNP, freeing and the tri-snRNP, respectively. We also report nal flexibility; SF3a bridges SF3b and the U2 core
the 5′SS and 5′ exon for recognition by the U6 the structure of the S. cerevisiae B complex at (fig. S7). The pre-mRNA retention and splicing (RES)
and U5 small nuclear RNAs (snRNAs), respectively 3.9-Å resolution. complex binds Hsh155 and the 3′-end sequences
(5–7). The resulting B complex is converted by of the intron (fig. S7, A and C). The tri-snRNP is
the ATPase/helicase Brr2 into the activated Electron microscopy of the endogenous well characterized by the EM map (fig. S8). The
spliceosome (Bact complex). The Bact complex is pre-B complex structures and locations of most protein and RNA
remodeled to become the catalytically activated The endogenous pre-B and B complexes were components in the tri-snRNP are nearly identical
spliceosome (B* complex), where the branching individually derived from two different strains to those in the isolated yeast tri-snRNP (16, 17).
reaction occurs. The resulting catalytic step I of S. cerevisiae. In both cases, the spliceosome The U1 and U2 snRNPs loosely interact with each
spliceosome (C complex) is converted into the was purified through two steps of affinity chro- other (fig. S7, A and B), and they only make lim-
step II catalytically activated spliceosome (C* matography (fig. S1, A and B), and its identity ited contacts with the tri-snRNP. Consequently, the
complex), and exon ligation follows. The ligated was confirmed by snRNA analysis (fig. S1, C and entire pre-B complex exhibits considerable flexibil-
exon in the postcatalytic spliceosome (P complex) D). Chemical cross-linking was used to stabilize ity, with five rigid parts (U1 snRNP, SF3a, SF3b, U2
is released, and the resulting intron lariat splice- the otherwise highly dynamic pre-B and B com- core, and tri-snRNP) loosely bound together to gen-
osome (ILS) is disassembled, completing one plexes. To overcome the transient nature of the erate a highly asymmetric assembly (Fig. 1A).
cycle of precursor messenger RNA (pre-mRNA) pre-B complex, we engineered a mutant Prp28 The structure of the S. cerevisiae B complex
splicing. that blocks the dissociation of U1 snRNP (13). contains 55 proteins, 4 snRNA molecules, and
Cryo-EM samples were imaged by a K2 Summit the pre-mRNA (Fig. 1B). Compared with the pre-
1
Beijing Advanced Innovation Center for Structural Biology, detector (Gatan) mounted on a Titan Krios elec- B complex, Prp38, Snu23, and Spp381 are re-
Tsinghua-Peking Joint Center for Life Sciences, Schools of tron microscope (FEI) (fig. S1, E and F). cruited further into the B complex (20–23). These
Life Sciences and Medicine, Tsinghua University, Beijing
100084, China. 2Technology Center for Protein Sciences,
Low-resolution references of the pre-B and B proteins form a subcomplex and appear to ori-
Ministry of Education Key Laboratory of Protein Sciences, complexes were derived from a preliminary anal- ent the pre-mRNA and facilitate the recognition
School of Life Sciences, Tsinghua University, Beijing 100084, ysis of the EM data (fig. S2). For the S. cerevisiae of pre-mRNA by U6 snRNA. Except for Brr2, the
China. 3Institute of Biology, Westlake Institute for Advanced pre-B complex, 1.85 million particles were auto- U4 Sm ring, and the RNaseH-like and Jab1/MPN
Study, Westlake University, 18 Shilongshan Road, Xihu
District, Hangzhou 310064, Zhejiang Province, China.
picked and classified using a guided multireference domains of Prp8, all other proteins in the tri-snRNP
*These authors contributed equally to this work. procedure, as reported previously (14) (fig. S3). of the B complex remain structurally identical to
†Corresponding author. Email: shi-lab@tsinghua.edu.cn Owing to the motions of U1 and U2 snRNPs those in the pre-B complex. U2 snRNP appears to
collapse onto the tri-snRNP in the B complex, generally conserved in the S. pombe and human these specific H-bonds, mostly made to the back-
forging closer interactions than in the pre-B orthologs (Fig. 3D). Consistent with the crystal bone of the 5′SS/U1 duplex, contribute to its
complex (Fig. 1B). structure of the human U1 snRNP (27), our struc- specific accommodation by U1 snRNP (Fig. 3H).
tural findings explain the observation that muta-
RNA elements in the pre-B and tions of Ser19 and Val20 compromise the ability of Structure of U2 snRNP
B complexes Yhc1 to stabilize the 5′SS/U1 duplex (25). In the published cryo-EM structure of the
In the pre-B complex, the 5′SS and BPS are recog- A C-terminal fragment (residues 198 to 230) of S. cerevisiae B complex (18), the local resolution
nized by the U1 and U2 snRNPs, respectively, in Luc7 also forms a C2H2-type zinc finger (fig. S9, around U2 snRNP was estimated to be 17.2 Å,
part through duplex formation with U1 and U2 B and C), which is known to promote the splicing which allowed docking of known structures. The
snRNAs (Fig. 2, A and B). Eleven consecutive nu- of pre-mRNA with a weak 5′SS (26, 30, 31). In our improved local resolution (3.6 to 4.6 Å) around
cleotides (A1UACUUACCUU11) at the 5′ end of U1 structure, three charged residues from this zinc U2 snRNP in our structure of the pre-B com-
snRNA base-pair with the 5′SS (GUAUGU) and its finger—Asp212, Arg216, and Lys224—directly con- plex offers considerably more structural details
surrounding nucleotides (Fig. 2C). Fifteen consec- tact the 5′SS/U1 duplex through H-bonds (Fig. (Fig. 4A). The SF3b complex is connected to the
utive nucleotides (G32UGUAGUAUCUGUUC46) of 3E). These three residues are invariant in the U2 core by the SF3a complex. Within the SF3a
U2 snRNA form a duplex with the BPS (UACUAAC) Luc7 orthologs Usp106 (S. pombe) and Luc7L complex (32), Prp11 binds the SF3b complex,
and its surrounding nucleotides (Fig. 2D). The (Homo sapiens) (Fig. 3F). In addition, the C- Prp9 interacts with the U2 core, and Prp21 con-
5′-end sequences of U2 snRNA already form terminal residues of SmB and SmD3 interact nects Prp11 and Prp9. Specifically, the helices a7
helix II with the 3′-end sequences of U6 snRNA, with the 5′SS/U1 duplex (Fig. 3G). Together, and a9 from Prp9 associate with SmD1 and
and five consecutive nucleotides of U6 snRNA
(A26U27U28U29G30) are anchored to loop I of U5
snRNA through duplex formation (Fig. 2E and
fig. S8D).
SmD2 of the U2 Sm ring (Fig. 4B). The N- and Hsh155, whereas the zinc-binding motif other through three loose interfaces, which yield
terminal 90 residues (residues 15 to 105) of directly binds the BPS/U2 duplex (Fig. 4C). considerable flexibility (Fig. 5A). At the interface
Prp11 form a folded domain that is stabilized between the U1 and U2 snRNPs, Lea1 contacts
by a C2H2-coordinated (Cys68, Cys71, His84, and The transition from pre-B to B Prp39 through a small interface (Fig. 5B, left
His90) zinc ion. The N-terminal fragment of Prp11 In the pre-B complex, U1 snRNP, U2 snRNP, and panel, and fig. S7B). An RNA duplex from U2
(residues 15 to 66) closely interacts with Cus1 the U4/U6.U5 tri-snRNP interact with each snRNA binds the positively charged surface of
Fig. 2. The RNA elements in the S. cerevisiae pre-B and B complexes. (D) A close-up view of the RNA duplex between the BPS and the complementary
(A) Structural comparison of the RNA elements of the pre-B and B complexes. U2 snRNA sequences in the pre-B complex. The nucleophile-containing
The RNA elements in the pre-B complex are colored identically to those in adenine base is already flipped out of the duplex registry. (E) A close-up view of
Fig. 1A, whereas in the B complex, the pre-mRNA is colored wheat and all other the duplex between U6 snRNA and loop I of U5 snRNA in the pre-B complex.
RNA elements are shown in gray. (B) Schematic diagrams of the base-pairing Five consecutive nucleotides (A26U27U28U29G30) of U6 snRNA form a duplex
interactions among the RNA elements in the pre-B complex (left panel) and with U96U97U98U99A100 of U5 loop I. (F) A close-up view of the interactions
the B complex (right panel). (C) A close-up view of the RNA duplex between the between the 5′-exon–5′SS sequences and U6 snRNA in the B complex. The
5′SS and the complementary U1 snRNA sequences in the pre-B complex. 5′-exon–5′SS sequences are located close to U6 snRNA.
Prp42 (Fig. 5B, right panel). At the interface Compared with its position in the pre-B com- Hsh155 moves closer to the LSm ring in the B
between U2 snRNP and the tri-snRNP, Hsh155 plex, the entire U2 snRNP is translocated toward complex (Fig. 5F). Cus1 and Rse1 directly interact
is positioned close to the LSm ring, and U2 the tri-snRNP in the B complex, resulting in a con- with the N- and C-terminal RecA2 domains of Brr2,
snRNA connects U2 snRNP to the tri-snRNP siderably more compact assembly (Fig. 5E). Brr2 respectively (Fig. 5G). The B-specific proteins Prp38,
(Fig. 5C). In addition, Hsh49 and Rse1 of the SF3b undergoes a rotation of about 30° and a trans- Snu23, and Spp381 are recruited into the B complex
complex are positioned close to Brr2, although location of 40 to 50 Å in the pre-B–to–B transition. and interact with each other (Fig. 5H). Prp38 and
direct interactions may be lacking (Fig. 5D). Relative to its place in the pre-B complex (Fig. 5C), Snu23 are positioned close to the 5′-exon–5′SS
Fig. 4. Structure of U2 snRNP in the S. cerevisiae pre-B complex. green, forest green, and pale green, respectively. (B) A close-up view of the
(A) Overall structure of U2 snRNP in the pre-B complex.The protein components interface between Prp9 of the SF3a complex and SmD1 and SmD2 of the U2 Sm
in the SF3b complex and the U2 core are colored purple and cyan, respectively. ring. (C) A close-up view of the interface between Prp11 of the SF3a complex
The three proteins of the SF3a complexes—Prp11, Prp21, and Prp9—are colored and Cus1 and Hsh155 of the SF3b complex. H15, H16, and H17 are helices.
sequences and U6 snRNA, and their positive standing of pre-mRNA splicing by the spliceo- sient interfaces, the interactions between U1 and
electrostatic surface potential may help orient some. The local resolutions in the core of U1 U2 snRNPs in the A complex are likely preserved
the 5′-end sequences of pre-mRNA (Fig. 5I). snRNP and the tri-snRNP reach 3.0 Å (fig. S5B), in the pre-B complex. Therefore, we propose that
which allows assignment of atomic features. The the structure of the A complex may be faithfully
Discussion pre-B complex is assembled from the A complex represented in the pre-B complex (Fig. 6A).
Structural elucidation of the pre-B complex fills and the tri-snRNP. Because U1 and U2 snRNPs The determination of the pre-B structure,
an important void in the mechanistic under- associate with the tri-snRNP only through tran- along with other published information, reveals
Fig. 5. Structural changes in the S. cerevisiae pre-B–to–B transition. the pre-B and B complexes. Compared with its position in the pre-B complex,
(A) Surface representation of the yeast pre-B complex. The two interfaces the entire U2 snRNP moves closer to the tri-snRNP in the B complex. The
between U2 snRNP and the tri-snRNP are indicated by red and yellow boxes. B complex is color-coded, and the pre-B complex is shown in gray. (F) A close-up
The interface between the U1 and U2 snRNPs is indicated by a blue box. view of the interface between Hsh155 and the LSm ring in the B complex.
(B) The interface between the U1 and U2 snRNPs of the pre-B complex is shown The contact surface area between Hsh155 and the LSm ring is considerably
in cartoon representation (left panel) and electrostatic surface potential (right larger than in the pre-B complex. (G) A close-up view of the interface between
panel). Lea1 directly contacts Prp39, and an RNA duplex of U2 snRNA binds the the SF3b complex and the tri-snRNP. Cus1 and Rse1 directly contact the
positively charged surface of Prp42. (C) A close-up view of the interface N-terminal and the C-terminal RecA2 domains, respectively, of Brr2 in the
between U2 snRNP and the tri-snRNP in the pre-B complex. Hsh155 is located B complex. (H) The three B-specific proteins (Prp38, Snu23, and Spp381)
close to the LSm ring, and U2 snRNA links U2 snRNP to the tri-snRNP. stabilize the RNA elements in the B complex. (I) The positive charges
(D) Rse1 of the SF3b complex is located close to, but may not directly interact on the surface of the B-specific proteins may help orient the pre-mRNA and
with, Brr2 of the tri-snRNP in the pre-B complex. (E) Structural overlay of U6 snRNA.
REFE RENCES A ND N OT ES
1. C. L. Will, R. Lührmann, Cold Spring Harb. Perspect. Biol. 3,
a003707 (2011).
2. Y. C. Liu, S. C. Cheng, J. Biomed. Sci. 22, 54 (2015).
3. O. Cordin, D. Hahn, J. D. Beggs, Curr. Opin. Cell Biol. 24,
431–438 (2012).
4. C. Boesler et al., Nat. Commun. 7, 11997 (2016).
5. J. P. Staley, C. Guthrie, Mol. Cell 3, 55–64 (1999).
6. J. Y. Chen et al., Mol. Cell 7, 227–232 (2001).
7. R. Hage et al., Mol. Cell. Biol. 29, 3941–3952 (2009).
8. Y. Shi, Nat. Rev. Mol. Cell Biol. 18, 655–670 (2017).
9. Y. Shi, J. Mol. Biol. 429, 2640–2653 (2017).
10. S. M. Fica, K. Nagai, Nat. Struct. Mol. Biol. 24, 791–799
(2017).
(Fig. 6B). In the first phase, Brr2 pulls on U4 structurally identified in the S. cerevisiae pre-B We thank the Tsinghua University Branch of the China National
Center for Protein Sciences (Beijing) for providing facility
snRNA, triggering unwinding of the U4/U6 du- complex. Nonetheless, the structure of the pre-B support. The computation was completed on the Explorer
plex, dissociation of U4 snRNP and the LSm complex, together with other published informa- 100 cluster system of the Tsinghua National Laboratory for
ring, and rearrangement of U6 snRNA and the tion (18, 19, 40), allows us to track the movement Information Science and Technology. Funding: This work was
supported by funds from the National Natural Science competing financial interests. Data and materials availability: SUPPLEMENTARY MATERIALS
Foundation of China (31621092 and 31430020) and the The atomic coordinates have been deposited in the www.sciencemag.org/content/360/6396/1423/suppl/DC1
Ministry of Science and Technology (2016YFA0501100 to J.L.). Protein Data Bank with the following accession codes: Materials and Methods
Author contributions: R.B. and R.W. purified the yeast 5ZWM for the tri-snRNP and U2 snRNP of the pre-B complex, Figs. S1 to S10
spliceosomes and prepared the cryo-EM samples. R.B., R.W., 5ZWN for U1 snRNP of the pre-B complex, and 5ZWO Tables S1 to S3
J.L., and C.Y. collected and processed the EM data. for the B complex. The EM maps have been deposited in References (42–59)
C.Y. generated the EM map and built the atomic model. the EMDB with the following accession codes: EMD-6972
All authors contributed to structure analysis. R.B., for the tri-snRNP and U2 snRNP of the pre-B complex,
R.W., and C.Y. contributed to manuscript preparation. EMD-6973 for U1 snRNP of the pre-B complex, and EMD-6974 30 April 2018; accepted 16 May 2018
Y.S. designed and guided the project and wrote the for the B complex. Requests for materials should be Published online 24 May 2018
manuscript. Competing interests: The authors declare no addressed to Y.S. 10.1126/science.aau0325
@An @Am
Topological order is often quantified in terms of Chern numbers, each of which classifies Fmn ðqÞ ¼ i½Am ; An ð1Þ
@qm @qn
a topological singularity. Here, inspired by concepts from high-energy physics, we use
quantum simulation based on the spin degrees of freedom of atomic Bose-Einstein
condensates to characterize a singularity present in five-dimensional non-Abelian gauge
where i is an imaginary unit; in three spatial
theories—a Yang monopole. We quantify the monopole in terms of Chern numbers
dimensions, the components of the magnetic
measured on enclosing manifolds: Whereas the well-known first Chern number vanishes,
field Bm ¼ Dmnl Fnl =2, where l is an integer, can
the second Chern number does not. By displacing the manifold, we induce and observe
be determined from the elements of the Fnl
a topological transition, where the topology of the manifold changes to a trivial state.
matrices (Dmnl is the rank-3 Levi-Civita symbol,
and we used Einstein’s implied summation con-
a divergence in the field strength and can con- Experimental Hamiltonian ces commutes with the time-reversal operator, the
tribute a unit of flux through any enclosing We realized a non-Abelian gauge field by cyclical- system has time-reversal symmetry (15); Kramers
manifold. This generalized flux is quantized and ly coupling four levels within the hyperfine ground theorem then implies that the system has two
is given by the Chern numbers. In particular, for states of rubidium-87 using radio-frequency and pairs of degenerate energy states, here with en-
Yang monopoles, the first Chern number is zero, microwave fields (Fig. 2, A and B), essentially ergies ET ¼ Tħjqj=2. Thus, each energy, labeled by
but the second Chern number is either +1 or −1 forming a square plaquette. The four couplings + or −, has two independent eigenstates j↑T ðqÞi
(Fig. 1). were parameterized by two Rabi frequencies and j↓T ðqÞi; each of these pairs define a degen-
Many quantum systems can be described by a WA and WB and two phases fA and fB arranged so erate subspace (DS). As shown in Fig. 2B, these
Hamiltonian Ĥ(q) that depends on position q in that the sum of the phases around the plaquette DSs are characterized by a generalized magneti-
parameter space. At each position, the system is was p. This configuration of control fields, along zation vector hGi ¼ ðhG ^ 2 i; hG
^ 1 i; hG ^ 3 i; hG
^ 4 i; hG
^ 5 iÞ
characterized by energies Ek(q) and eigenstates with a detuning d ¼ jgF jmDB =ħ, where gF is the on a unit 4-sphere in our 5D space. Different
jkðqÞi, where k∈f1; …; Kg is an index that iden- Landé g factor, m is the Bohr magneton, DB is the configurations within each DS share the same
tifies the eigenstate in ourK -dimensional Hilbert shift in the magnetic field from resonant cou- magnetization vector, which can be pictured in
space. A gauge potential called the non-Abelian pling condition, and ħ = h/2p, gave us an experi- terms of an additional 3D Bloch sphere (green
Berry connection Abg m ðqÞ ¼ ihbðqÞj@=@qm jgðqÞi, mentally controllable 5D parameter space labeled sphere in Figs. 1 and 2B). An eigenstate is fully
where b; g∈f1; …; Kg , is encoded in the wave by the Cartesian coordinates q = (−WB cos fB, depicted by assigning the two such “Bloch” vec-
functions; thus, for any position q, each vector −WA cos fA, −WA sin fA, d, −WB sin fB). In much tors. The Yang monopole (16, 17) resides at the
component Am is represented by a matrix. Chern the same way that a two-level atom in a magnetic Hamiltonian’s degeneracy point at q = 0, a singular
numbers and curvatures can be then defined by field can be understood in terms of three Pauli point where the non-Abelian Berry’s connection
Eqs. 1 to 3 for each well-separated energy level. matrices, our four-level system is governed by the diverges. The non-Abelian Berry’s curvatures from
Because of these gauge fields, an initial quantum Hamiltonian our experimental Hamiltonian (Eq. 4) quantum
^ m i ¼ vn hF^ mn i þ constant
hM ð6Þ
qT ðtÞ ¼ q0 ½cosð2pt=T pffiffiffi Þ; cosð2pt=T Þ; ∓sinð2pt= associated with the state within the DS at q0. The four initial states used in (A) to (C) are also shown
T Þ; 0; Tsinð2pt=T Þ= 2 by ramping f2 and evalu- in the theoretical (top sphere) and the experimental (bottom sphere) plots. (E) 1/q2 scaling in the
ating the deflection along the q2 direction at q0, strength of the curvature. The matrix components of the curvature hF ^ f q i are evaluated for various q0.
2 2
we obtained 2q20 F^q2 f2 ¼ 0:08ð3Þ^I 0 þ ½0:12ð5Þ; The data show excellent agreement with the theory that exhibits 1/q2 dependence (solid lines).
previous section, we directly obtained C2 ¼ individual curvatures, was also zero. Thus, the caused by the breakdown of the linear response
2q40 tr½Fq1 f1 ðq0ÞFf2 q2 ðq0Þ ¼ 0:97ð6Þ for the ground nontrivial topology of the monopole field is not near the transition point. Our theory [continuous
state, consistent with the theoretical value C2 = 1. expressed by a first Chern number. curves in Fig. 6B, and see (15)] shows that slower
We repeated the measurements for the excited ramps enlarge the region in which linear re-
state and found C2 ¼ 2q40 tr½Ff1 q2 ðq 0 ÞFf2 q1 ðq0 Þ ¼ Topological transition sponse is valid and make the transition sharper
0:93ð6Þ, also in agreement with the theoret- We concluded our measurements by inducing (Fig. 6B). Topological transitions have been ob-
ical value C2 = –1. These nonzero Chern num- a topologically nontrivial-to-trivial transition served in a range of experiments (6, 7, 25); how-
bers inform us that the manifold is topologically of the manifold by displacing the 4-sphere in ever, in all of these cases, the observed topological
nontrivial. parameter space from the origin by an amount phases were only identified by a Dirac monopoles’
Because the system is time-reversal symmet- qoffset = qoffset(q0/q0) (Fig. 6A). The topological first Chern number and enclosing 2D manifolds.
ric, the first Chern form is zero, and therefore transition occurs at the critical displacement By contrast, in our system, the first Chern number
Eq. 3 for the first Chern number should be zero qcrit = q0 when the Yang monopole departs the is zero everywhere and the second Chern number
for both degenerate manifolds. Indeed, all the manifold. Figure 6B shows our observed tran- characterizes the topological phase, arising from
measured non-Abelian Berry curvatures were sition of the second Chern number from ±1, for a Yang monopole at the origin of parameter space.
traceless (q20 tr½F^f2 q2 ¼ 0:08ð3Þ and q20 tr½F^q1 f1 ¼ the ground and excited states, to zero as the off- The opposite topological charges observed in the
0:01ð3Þ for the ground state, and q20 tr½F^q1 f2 ¼ set coupling qoffset was increased. This transition ground or excited manifolds result from a mono-
0:02ð3Þ and q20 tr½F^f1 q2 ¼ 0:00ð3Þ for the ex- is associated with the topology of the manifold pole in one manifold acting as an antimonopole
cited state), so that the first Chern number, changing from topologically nontrivial to trivial. in the other. With these Chern number measure-
which is the surface integral of the trace of the The smoothness of the observed transition was ments, we confirmed that the engineered topolog-
ical singularity in our system indeed simulated a
Yang monopole.
11. M. V. Berry, Proc. R. Soc. Lond. Ser. A 392, 45–57 (1984). 28. H. M. Price, O. Zilberberg, T. Ozawa, I. Carusotto, N. Goldman, Institute. S.S. acknowledges the Japan Society for the
12. F. Wilczek, A. Zee, Phys. Rev. Lett. 52, 2111–2114 (1984). Phys. Rev. B 93, 245113 (2016). Promotion of Science (fellowship for research abroad). Author
13. K. Toyoda, K. Uchida, A. Noguchi, S. Haze, S. Urabe, Phys. Rev. 29. S. Murakami, N. Nagosa, S.-C. Zhang, Phys. Rev. B 69, 235206 contributions: S.S. conducted the experiment, performed the
A 87, 052307 (2013). (2004). theoretical work, and analyzed the data. S.S., A.R.P., F.S.-C.,
14. T. Li et al., Science 352, 1094–1097 (2016). 30. Y. E. Kraus, Z. Ringel, O. Zilberberg, Phys. Rev. Lett. 111, and I.B.S. contributed to the rubidium Bose-Einstein condensate
15. See supplementary materials. 226401 (2013). apparatus. All authors substantially participated in the
16. Y. Hatsugai, New J. Phys. 12, 065004 (2010). 31. C.-H. Chern, H.-D. Chen, C. Wu, J.-P. Hu, S.-C. Zhang, Phys. discussion and the writing of the manuscript. S.S. and I.B.S.
17. Our non-Abelian gauge field is SU(2) symmetric, where the Rev. B 69, 214512 (2004). conceived of the project. Competing interests: The authors
generator can be, for example, the Pauli matrices. 32. X.-L. Qi, T. L. Hughes, S.-C. Zhang, Phys. Rev. B 78, 195424 (2008). declare no competing interests. Data and materials
18. V. Gritsev, A. Polkovnikov, Proc. Natl. Acad. Sci. U.S.A. 109,
33. Y.-J. Lin et al., Phys. Rev. Lett. 102, 130401 (2009). availability: All data are available in the manuscript or the
6457–6462 (2012).
34. J. Ruseckas, G. Juzeliūnas, P. Öhberg, M. Fleischhauer, Phys. supplementary materials.
19. J. E. Avron, M. Fraas, G. M. Graf, O. Kenneth, New J. Phys. 13,
Rev. Lett. 95, 010404 (2005).
053042 (2011).
35. P. Hauke et al., Phys. Rev. Lett. 109, 145301 (2012).
20. H.-I. Lu et al., Phys. Rev. Lett. 116, 200402 (2016). SUPPLEMENTARY MATERIALS
21. M. Lohse, C. Schweizer, O. Zilberberg, M. Aidelsburger, I. Bloch, 36. E. A. Martinez et al., Nature 534, 516–519 (2016).
www.sciencemag.org/content/360/6396/1429/suppl/DC1
Nat. Phys. 12, 350–354 (2016). Materials and Methods
22. S. Nakajima et al., Nat. Phys. 12, 296–300 (2016). ACKN OWLED GMEN TS
Supplementary Text
23. N. Nagaosa, J. Sinova, S. Onoda, A. H. MacDonald, N. P. Ong, The authors would like to thank A. Polkovnikov and M. Kolodrubetz
Figs. S1 to S3
Rev. Mod. Phys. 82, 1539–1592 (2010). for discussion. Funding: This work was partially supported by
References (37, 38)
24. G. Jotzu et al., Nature 515, 237–240 (2014). the U.S. Army Research Office’s Atomtronics MURI, and by the
Database S1
25. M. Aidelsburger et al., Nat. Phys. 11, 162–166 (2015). U.S. Air Force Office of Scientific Research’s Quantum Matter
26. M. Kolodrubetz, Phys. Rev. Lett. 117, 015301 (2016). MURI, National Institute of Standards and Technology, and 15 February 2017; accepted 27 April 2018
27. S.-C. Zhang, J. Hu, Science 294, 823–828 (2001). NSF through the Physics Frontier Center at the Joint Quantum 10.1126/science.aam9031
Multivalent counterions diminish the valent ions, we measured the normal and fric-
tion forces of apposing brushes at constant
contact point and ionic strength (6 mM) for
lubricity of polyelectrolyte brushes four concentrations of Y3+ (0, 0.01, 0.1, and 0.5 mM)
and four concentrations of Ca2+ and Ba2+ (0, 0.01,
0.1, and 1.0 mM) (Fig. 1, C to E, and figs. S4 to S8).
J. Yu1,2,3*, N. E. Jackson1,2*, X. Xu4, Y. Morgenstern5, Y. Kaufman5, M. Ruths4,
Sliding was studied at low loads to avoid surface
J. J. de Pablo1,2†, M. Tirrell1,2† damage. In 6 mM NaNO3, a repulsive force was
detected at a separation distance of 600 nm
Polyelectrolyte brushes provide wear protection and lubrication in many technical, and rapidly increased upon compression (Fig. 1C).
medical, physiological, and biological applications. Wear resistance and low friction Friction forces measured after the normal force
are attributed to counterion osmotic pressure and the hydration layer surrounding measurement were extremely low (<10 mN)
the charged polymer segments. However, the presence of multivalent counterions in (Fig. 1B) for the range of load (0 to 6 mN) in-
solution can strongly affect the interchain interactions and structural properties of brush vestigated (Fig. 1D). The presence of Y3+, even
layers. We evaluated the lubrication properties of polystyrene sulfonate brush layers sliding at low concentration (0.01 mM), decreased the
against each other in aqueous solutions containing increasing concentrations of counterions.
S
oft interfaces consisting of charged macro- water or monovalent salt solutions (10, 13, 27). However, the addition of 0.01 mM Y3+ signifi-
molecules are the norm in biology, includ- Considering the prevalence of multivalent ions in cantly affected the friction force (Fig. 1D), which
ing the cellular glycocalyx (1), the surface of biological systems and industrial formulations was very small at loads up to 3 mN and then
articular cartilage (2, 3), and the interfaces (28, 29), a concrete understanding of their role in rapidly increased. Despite showing distinct dif-
between mineralized collagen in bone (4). polyelectrolyte brush lubrication is of importance. ferences in the friction forces at loads above 3 mN,
Commercial products, such as those for personal Densely tethered polystyrene sulfonate (PSS) the PSS brushes had similar heights in pure 6 mM
care, medical prostheses (e.g., joint replacement), brushes were grown from mica surfaces by using NaNO3 and in 0.01 mM Y3+ (fig. S2). We hypoth-
anti-fogging surfaces, and DNA brushes for syn- a surface-initiated atom transfer radical po- esize that this increase was due to enhanced
thetic biology and gene chips (5, 6), also often lymerization method (21). The kinetic friction chain interpenetration of the apposing brushes
rely on highly hydrated polyelectrolyte brush force (Fk) between apposing brush surfaces was at higher compression (31), leading to more
interfaces (7–11). Characterization of the tribo- measured as a function of their absolute sepa- bridging events between the negatively charged
logical and dynamical properties of polyelec- ration distance (D), compression load (L), and sulfonate groups, mediated by the multivalent
trolyte brushes has been of interest because of sliding velocity (V||) in a droplet of aqueous ions. This will be addressed in the MD simu-
their tunable wetting properties (12), strong solution with a surface forces apparatus (model lations described below.
osmotic repulsion (13), compression-induced SFA 2000, SurForce LLC) (Fig. 1A). In 6 mM In solutions containing higher concentrations
interpenetration (14), durability (15), and solvent NaNO3, the PSS brushes showed excellent lu- (0.1 and 0.5 mM) of Y3+, the range of separation
response (16). Polyelectrolyte brush structure brication properties. The friction forces between distances over which repulsion was measured
can be influenced by environmental modulations two brush layers sliding against each other at during compression rapidly decreased to less than
(e.g., pH, temperature, and added salt) (17–19). a velocity of ~4.8 mm/s were just above the de- 300 nm, and strong hysteresis was measured
Sharp changes in brush height and morphology tection limit (~5 mN) of the friction sensor up to during the separation (Fig. 1C). In both condi-
were observed as a function of the concentration a load of 13 mN (Fig. 1B, inset, and fig. S1). The tions, the friction forces were larger than those
of added multivalent ions in experiments (20, 21), coefficient of friction, m, between the two brush measured at lower Y3+ concentration: A high
molecular dynamics (MD) simulations (22–24), layers, defined as dFk/dL, was smaller than 0.001. friction force of ~0.1 mN was evident even at
and theory (25). Surface forces apparatus (SFA) The pressure between apposing surfaces in very low load (0.2 mN) and then rapidly in-
experiments have detected attractive interactions this load regime was estimated to be ~5 MPa. creased with the load (Fig. 1D). Similar trends,
consistent with multivalent ion–induced bridging Both m and the pressure are within the range with quantitative differences depending on the
between polyelectrolyte brush layers on apposing encountered in human knee joints (30). As the ion, in normal and friction forces were observed
surfaces (26). However, most tribological studies load was increased above 13 mN, the friction for divalent ions (Ca2+ and Ba2+), with small
of polyelectrolyte brushes have been limited to pure force was observed to increase linearly with a friction forces at low concentrations (0.01 mM)
slope of m = 0.005 ± 0.001, which is still a very and sharply increasing forces at higher concen-
1
Institute for Molecular Engineering, University of Chicago, low friction coefficient. Increasing the mono- trations (0.1 and 1 mM) (Fig. 1E and figs. S4 to
Chicago, IL 60637, USA. 2Institute for Molecular Engineering, valent salt concentration to 0.15 M screened S8). Polyelectrolyte brushes can form pinned-
Argonne National Laboratory, Lemont, IL 60439, USA. electrostatic repulsion between charged PSS micelle–like inhomogeneous structures in the
3
School of Materials Science and Engineering, Nanyang
Technological University, Singapore 639798. 4Department of
segments, which facilitated interchain penetra- presence of multivalent counterions as a result
Chemistry, University of Massachusetts Lowell, Lowell, MA tion of the apposing brushes, leading to an in- of electrostatic bridging (22–24). AFM mea-
01854, USA. 5Zuckerberg Institute for Water Research, The creased m of 0.03 (fig. S1B). Introducing trivalent surements confirmed that in 0.1 and 0.5 mM Y3+,
Jacob Blaustein Institutes for Desert Research, Ben-Gurion Y3+ ions into the solution sharply increased the and 1 mM Ba2+ solution, the brush layer col-
University of the Negev, Midreshet Ben-Gurion, Israel.
*These authors contributed equally to this work.
friction forces. In a solution containing 3 mM lapsed into pinned-micelle–like structures (Fig. 2,
†Corresponding author. Email: depablo@uchicago.edu (J.J.dP.); NaNO3 and 0.5 mM Y(NO3)3 (maintaining the C and D, and figs. S10 to S12) accompanied by
mtirrell@uchicago.edu (M.T.) total ionic strength at 6 mM), large friction a stark decrease in brush height (Fig. 2E and
figs. S2 and S8). Although the PSS brush also 0.1 MPa to 13.7 ± 2.7 MPa when increasing the structure and mechanical properties of the PSS
collapsed in 1 mM Ca2+ solution (fig. S7), no clear concentration of Y3+ from 0 to 0.5 mM (Fig. 2E brushes, dictating their boundary lubrication
micelle-like structure was observed by AFM and figs. S11 and S12). The agreement between properties (Fig. 2E and fig. S13).
(fig. S10). Additionally, AFM revealed that the SFA and AFM measurements shows that multi- We propose that multivalent ions affect the lubri-
elastic modulus of the brush increased from 1.2 ± valent counterions substantially influence the city of the PSS brushes through two mechanisms:
Fig. 4. Multivalent ion–induced friction mechanisms. MD simulation separation distance (z), which is normalized by the polymer contour
showing multivalent ion–induced bridging between apposing polyelectrolyte length (50s). Error bars represent ±SD, averaged over five replicas.
chains at low and high salt concentrations. (A) Percentage of multivalent (B to D) Schematics of apposing PSS brushes sliding against each other in
cations bridging apposing polyelectrolyte chains plotted against the scaled (B) 6 mM NaNO3, (C) 0.01 mM Y(NO3)3, and (D) 0.1 mM Y(NO3)3 solution.
(i) At low concentrations of multivalent salt, resulting in increased surface roughness. The geometry were employed to investigate these
multivalency induces three-body electrostatic bridging mechanism is expected to be stronger hypotheses.
bridging interactions between PSS chains, and for trivalent ions than for divalent ions owing MD simulations were performed at two mul-
(ii) at high concentrations of multivalent salt, to the higher valency. Coarse-grained MD sim- tivalent ion concentrations: “low,” where multi-
the brushes undergo a heterogeneous collapse, ulations of polyelectrolyte brushes in the SFA valent ions neutralized 10% of the polyelectrolyte
brush charge, and “high,” where multivalent ions at large separation (z > 0.55) (Fig. 4A). At low 16. O. Al-Jaf, A. Alswieleh, S. P. Armes, G. J. Leggett, Soft Matter
neutralized 100% of the brush charge (32). Given concentrations, bridging interactions decrease 13, 2075–2084 (2017).
17. Y. Mei et al., Phys. Rev. Lett. 97, 158301 (2006).
the experimental sliding rate of 5 mm/s, ~100 ns gradually with increased separation, whereas at
18. S. Sanjuan, P. Perrin, N. Pantoustier, Y. Tran, Langmuir 23,
of simulation time corresponds to 5 × 10−4 nm high concentrations, a sharp transition occurs 5769–5778 (2007).
relative motion of the apposing surfaces, which when the brush separation distance cannot sup- 19. M. A. C. Stuart et al., Nat. Mater. 9, 101–113 (2010).
makes comparison with equilibrium simulations port a single spanning collapsed phase. Divalent 20. J. Yu et al., Macromolecules 49, 5609–5617 (2016).
reasonable. Brushes were simulated at a large ions exhibit similar concentration dependencies, 21. A. Jusufi, O. Borisov, M. Ballauff, Polymer (Guildf.) 54,
2028–2035 (2013).
separation distance where the apposing brushes but with a smaller overall magnitude of bridging.
22. J. Yu et al., Sci. Adv. 3, o1497 (2017).
were relatively noninteracting, and also at high Simulation results support the hypothesis that 23. N. E. Jackson, B. K. Brettmann, V. Vishwanath, M. Tirrell,
compression. Trivalent results are presented in electrostatic bridging increases the friction be- J. J. de Pablo, ACS Macro Lett. 6, 155–160 (2017).
Fig. 3 and figs. S14 and S15, with divalent results tween apposing polyelectrolyte brushes at low 24. L. Liu, P. Pincus, C. Hyeon, Macromolecules 50, 1579–1588
found in figs. S16 and S17. multivalent ion concentrations (Fig. 4, B and (2017).
25. B. K. Brettmann, P. Pincus, M. Tirrell, Macromolecules 50,
Figure 3, A and B, shows the time-averaged C) and that structural change of the polyelec-
1225–1235 (2017).
brush density and a representative snapshot of trolyte brushes dominates the friction at much 26. R. Farina, N. Laugel, J. Yu, M. Tirrell, J. Phys. Chem. C 119,
chain conformations at low trivalent ion concen- higher concentrations (Fig. 4D). 14805–14814 (2015).
tration and large separation, where the brushes Polyelectrolyte brushes have been proposed 27. M. Chen, W. H. Briscoe, S. P. Armes, J. Klein, Science 323,
exhibited a uniform, extended morphology. When as efficient boundary lubricants (2). This work 1698–1701 (2009).
28. R. W. Wilson, V. A. Bloomfield, Biochemistry 18, 2192–2196
apposing brush layers were brought into close confirms that the charged polymer chains and (1979).
contact (Fig. 3C and fig. S14), they maintained the osmotic pressure of the counterions provide 29. D. Bracha, R. H. Bar-Ziv, J. Am. Chem. Soc. 136, 4945–4953
a uniform extended morphology. When the tri- excellent lubrication properties in monovalent (2014).
valent ion concentration was increased (Fig. 3, D salt solution. However, the lubrication can break 30. S. Jahn, J. Seror, J. Klein, Annu. Rev. Biomed. Eng. 18, 235–258
(2016).
E
namine catalysis is a powerful activation competition for the active a-amino anion 7 side chain will activate the imine substrate during
mode in organocatalysis (Fig. 1A) (1–3). The under the reaction conditions; and (iii) for an catalysis (33, 34). Treatment of 21 with methyl
process involves conversion of carbonyl asymmetric version, carbonyl catalyst 1 should iodide (MeI) followed by deprotection with HCl
compound 1 into enamine 4 through imine control positioning of the incoming electro- yielded N-quaternized pyridoxals 16a to 16f as
intermediate 3 (an iminium intermediate phile. Development of such carbonyl cata- dark brown salts.
if a secondary amine is applied) and is catalyzed lysts is thus a formidable challenge in organic In the presence of 1 mole % (mol %) pyridoxal
by amine 2 (4). Nucleophilic addition of the chemistry. (R,S)-16b, reaction of tert-butyl glycinate (17)
activated enamine to an electrophile can proceed In biological systems, pyridoxal-dependent with N-diphenylphosphinyl imine 18a occurred
to yield substituted carbonyl 6. On the other aldolases can act as carbonyl catalysts for di- successfully, generating diamino acid ester 19a
hand, the formation of imine 3 also increases rect a-functionalization, as in the aldol reac- in a 90% yield with a >20:1 diastereomeric ratio
the a-H acidity of amine 2 (5–7) to facilitate for- tion of glycine (16–18). The enzymatic process (dr) and 99% enantiomeric excess (ee) (SM and
mation of the a-amino carbanion 7 (8, 9), which can provide a straightforward way to synthesize table S1, entry 2). The structure and absolute
can also react with an electrophile to produce 8 pharmaceutically important chiral b-hydroxy-a- configuration of the major diastereomer anti-19a
(10). If the product 8 can be hydrolyzed under amino acids and derivatives (16), such as Parkinson were confirmed by x-ray analysis of its acetyl
the reaction conditions to regenerate 1, it would drug L-threo-3,4-dihydroxyphenylserine (19) and derivative (2R,3R)-19a-acetyl (SM and fig. S3).
be possible to use the carbonyl compound 1 as a drug candidate (2R,3S)-2-amino-3-hydroxy-3- No reaction was observed without 16 present
catalyst (11) to promote a-functionalization of (pyridin-4-yl)-1-(pyrrolidin-1-yl)propan-1-one (20), (table S1, entry 7). Catalyst (R,S)-16b is the most
amine 2. without protecting manipulations. The mech- efficient in terms of yield and selectivities among
a-Functionalization of primary amines has anistic pathway is exemplified by the widely in- the pyridoxals (16a to 16f) examined (table S1,
been used extensively to make pharmaceutical- vestigated threonine aldolase–catalyzed aldol entry 2 versus entries 1 and 3 to 6). A mixed sys-
ly relevant compounds (12, 13). This transfor- reaction of glycine depicted in Fig. 1B (21). The tem of CHCl3-H2O (1:1) was the choice of solvent.
mation usually involves three steps, including enzymatic condensation of glycine and pyridoxal- Further studies showed that ethyl and benzyl
protection of the NH2 group with stoichiomet- 5′-phosphate (PLP) forms aldimine 13. Depro- glycinates both produced the corresponding di-
ric aldehydes or ketones before reaction and tonation of 13 produces resonance-stabilized amino acid esters with 99% ee but with some-
deprotection to release the desired free amines carbanion 14, which undergoes addition to what lower yields than tert-butyl glycinate (65
after reaction. In contrast, the carbonyl-catalyzed aldehyde 11 and subsequent hydrolysis to form and 64% versus 90%) (SM and table S2).
process eliminates the protecting manipula- b-hydroxy-a-amino acid 12 and regenerate Under the optimized conditions, we exam-
tions and is atom economical (14, 15). An ideal PLP. Studies on imitating the biological process, ined the substrate scope of the imine fragment
reaction should meet the following require- contributed by Kuzuhara (22, 23) and Breslow (Fig. 3). Monosubstituted (for 19a to 19m)
ments: (i) Carbonyl catalyst 1 must be electron- (24) and coworkers, have shown that stoichi- and polysubstituted (for 19n to 19p) phenyl
withdrawing enough to promote deprotonation ometric chiral pyridoxals can promote aldol imines as well as fused-aromatic imines (for
to form a-amino anion 7 for further reaction addition of glycine to aldehydes in the pres- 19q to 19t) reacted smoothly with glycinate 17
with an electrophile; (ii) both the carbonyl ence of metal salts, producing b-hydroxy-a- in the presence of 1 mol % pyridoxal (R,S)-16b,
catalyst 1 and the imine intermediate 3 should amino acids with moderate enantioselectivity generating the corresponding a,b-diamino acid
be much less reactive than the electrophile in and low diastereoselectivity. Recently, Richard esters 19a to 19t in good to excellent yields
and coworkers found that 5′-deoxypyridoxal with excellent diastereo- and enantioselectiv-
1
The Education Ministry Key Lab of Resource Chemistry and
promotes Mannich addition of glycine to glycine- ities. With 0.2 mol % of 16b, reaction of imine
Shanghai Key Laboratory of Rare Earth Functional Materials, 5′-deoxypyridoxal imine to form the correspond- 18a and 17 can still be completed in 11 hours
Shanghai Normal University, Shanghai 200234, China. ing Mannich adduct (25, 26). These studies, to yield product 19a without any loss of selec-
2
Chengdu Institute of Organic Chemistry, Chinese Academy together with those on the enzymatic aldol re- tivities. Imines containing electron-withdrawing
of Sciences, Chengdu 610041, China.
*These authors contributed equally to this work.
action of glycine (16–21), have laid a foundation substituents such as CN (for 19f), CF3 (for 19l),
†Corresponding author. Email: yuanwc@cioc.ac.cn (W.Y.); for the development of asymmetric carbonyl and pyrazole (for 19h) exhibit somewhat lower
zhaobg2006@shnu.edu.cn (B.Z.) catalysts. enantioselectivity. Heteroaromatic imines (for
Fig. 3. Substrate scope of the biomimetic Mannich reaction. Reactions chromatography (HPLC) analysis. The absolute configuration of 19a was
were carried out with 18 (0.20 mmol), 17 (0.30 mmol), 16b (0.0020 mmol), assigned as (2R,3R) by x-ray analysis of its acetyl derivative (2R,3R)-19a-acetyl
and NaHCO3 (0.50 mmol) in CHCl3-H2O (0.3 ml/0.3 ml) at 10°C for 4 to (SM and fig. S3). The absolute configurations of 19b to 19ae and 23 are
39 hours unless otherwise stated. For 19p, the reaction was carried out in half proposed by analogy. † indicates the ratios of the four theoretic diastereomers
scale. Isolated yields were based on imine 18 or 22. The dr (anti/syn) values determined by 1H NMR and HPLC analyses; ‡ indicates that only one
were determined by 1H nuclear magnetic resonance (NMR) analysis of crude isomer was observed, as judged by 1H NMR and HPLC analyses. TMS,
reaction mixtures. The ee values were determined by high-performance liquid trimethylsilyl; TBS, tert-butyldimethylsilyl.
Fig. 4. Applications
and mechanism.
(A) Synthetic trans-
formations. (B) Pro-
posed mechanism.
(C) Comparison of
catalytic selectivities.
(D) Comparison of
catalytic activities.
RE FE RENCES AND N OT ES 7. R. Casasnovas et al., J. Phys. Chem. B 116, 10665–10675 12. M. J. O’Donnell, Acc. Chem. Res. 37, 506–517 (2004).
1. P. I. Dalko, L. Moisan, Angew. Chem. Int. Ed. Engl. 40, (2012). 13. T. Hashimoto, K. Maruoka, Chem. Rev. 107, 5656–5682
3726–3748 (2001). 8. P. Beak, A. Basu, D. J. Gallagher, Y. S. Park, S. Thayumanavan, (2007).
2. B. List, Chem. Rev. 107, 5413–5415 (2007). Acc. Chem. Res. 29, 552–560 (1996). 14. B. Xu et al., Chem. Sci. 5, 1988–1991 (2014).
3. D. W. C. MacMillan, Nature 455, 304–308 (2008). 9. E. A. Mitchell, A. Peschiulli, N. Lefevre, L. Meerpoel, 15. B. M. Trost, Angew. Chem. Int. Ed. Engl. 34, 259–281
4. S. Mukherjee, J. W. Yang, S. Hoffmann, B. List, Chem. Rev. 107, B. U. W. Maes, Chemistry 18, 10092–10142 (2012). (1995).
5471–5569 (2007). 10. Y. Wu, L. Hu, Z. Li, L. Deng, Nature 523, 445–450 16. N. Dückers, K. Baer, S. Simon, H. Gröger, W. Hummel,
5. K. Toth, J. P. Richard, J. Am. Chem. Soc. 129, 3013–3021 (2007). (2015). Appl. Microbiol. Biotechnol. 88, 409–424 (2010).
6. J. Crugeiras, A. Rios, E. Riveiros, J. P. Richard, J. Am. Chem. Soc. 11. B.-J. Li, C. El-Nachef, A. M. Beauchemin, Chem. Commun. 53, 17. T. Kimura, V. P. Vassilev, G.-J. Shen, C.-H. Wong, J. Am. Chem. Soc.
133, 3173–3183 (2011). 13192–13204 (2017). 119, 11734–11742 (1997).
18. H. Nozaki, S. Kuroda, K. Watanabe, K. Yokozeki, J. Mol. 30. Y. Xie, H. Pan, M. Liu, X. Xiao, Y. Shi, Chem. Soc. Rev. 44, experiments. J.L., Y.L., J.M., and C.H. synthesized intermediates for
Catal. B Enzym. 59, 237–242 (2009). 1740–1748 (2015). catalyst development and also performed parts of substrate
19. H.-J. Gwon et al., Prep. Biochem. Biotechnol. 42, 143–154 31. L. Shi et al., Org. Lett. 17, 5784–5787 (2015). screening experiments. G.Z. searched for reaction conditions for
(2012). 32. Y. E. Liu et al., J. Am. Chem. Soc. 138, 10730–10733 (2016). selective removal of the diphenylphosphinyl group of 19a.
20. M. A. Schmidt et al., Org. Process Res. Dev. 19, 1317–1322 33. A. G. Doyle, E. N. Jacobsen, Chem. Rev. 107, 5713–5743 (2007). W.Y. and B.Z. cowrote the manuscript. Competing
(2015). 34. B. Zhao, Z. Han, K. Ding, Angew. Chem. Int. Ed. Engl. 52, interests: The authors declare no competing interests.
21. M. L. di Salvo et al., FEBS J. 281, 129–145 (2014). 4744–4788 (2013). Data and materials availability: Crystallographic data for
22. H. Kuzuhara, N. Watanabe, M. Ando, J. Chem. Soc. Chem. compounds (S,S)-38-oxime in fig. S2 and (2R,3R)-19a-acetyl in fig.
35. Z. Hussain et al., Res. J. Pharm. Biol. Chem. Sci. 7 (5),
Commun. 1987 (2), 95–96 (1987). S3 are available from the Cambridge Crystallographic Data Centre
1008–1025 (2016).
23. M. Ando, H. Kuzuhara, Bull. Chem. Soc. Jpn. 63, 1925–1928 under reference numbers CCDC 1837630 and CCDC 1837631,
36. L.-Q. Lu, X.-L. An, J.-R. Chen, W.-J. Xiao, Synlett 2012,
(1990). respectively. All other data to support the conclusions are included
24. J. T. Koh, L. Delaude, R. Breslow, J. Am. Chem. Soc. 116, 490–508 (2012).
in the main text or supplementary materials.
11234–11240 (1994).
25. K. Toth, L. M. Gaskell, J. P. Richard, J. Org. Chem. 71, ACKN OWLED GMEN TS
SUPPLEMENTARY MATERIALS
7094–7096 (2006). This work is in memory of R. Breslow for his outstanding
www.sciencemag.org/content/360/6396/1438/suppl/DC1
26. K. Toth, T. L. Amyes, J. P. Richard, J. P. G. Malthouse, contributions to the field of biomimetic chemistry and enzymology.
Materials and Methods
M. E. NíBeilliú, J. Am. Chem. Soc. 126, 10538–10539 (2004). Funding: We are grateful for the generous financial support
Figs. S1 to S4
27. A. Viso, R. Fernández de la Pradilla, A. García, A. Flores, from the NSFC (21672148 and 21472125) and the Ministry of
Tables S1 to S4
Chem. Rev. 105, 3167–3196 (2005). Education of China (PCSIRT_IRT_16R49). Author contributions:
28. R. G. Arrayás, J. C. Carretero, Chem. Soc. Rev. 38, 1940–1948 References (37–52)
B.Z. conceived of and directed the project. J.C. and X.G. developed
(2009). both the chiral N-quaternized pyridoxal catalysts and the biomimetic 24 February 2018; accepted 2 May 2018
29. R. Breslow, Acc. Chem. Res. 28, 146–153 (1995). asymmetric Mannich reaction and conducted most of the 10.1126/science.aat4210
R
surface of the film and a more n-type perovskite
ecently, perovskite solar cells (PSCs) with by the relatively low open-circuit voltages (Voc), film by using a solution-processed secondary
inverted planar heterojunction structures, typically <1.10 V, in comparison with >1.20 V growth (SSG) technique, leading to a substan-
wherein a polycrystalline perovskite film reported for regular PSCs using similar bandgap tial increase in Voc .
is sandwiched between a hole- and an perovskites (2, 5). The less-than-ideal Voc of in- The SSG procedure includes two steps: (i) the
electron-extraction layer, have gained at- verted PSCs is attributed to the nonradiative preparation of perovskite films by solution pro-
tention because they offer the promise of easy recombination losses both inside the perovskite cessing and (ii) the secondary growth with the
fabrication, compatibility with flexible substrates, bulk material and at the interfacial contacts assistance of guanidinium bromide (which we
versatility of energy-band engineering, and the (6–8), owing to the presence of a considerable abbreviate SSG-G). We started with the prepara-
possibility of fabricating multijunction cells (1–4); density of defects or recombination centers (9–15). tion of a mixed-cation lead mixed-halide perov-
moreover, they have achieved power conversion Critically, although “isolated” perovskite films skite layer using a nonstoichiometric recipe of
efficiencies (PCEs) exceeding 20%. Further en- can exhibit very high radiative efficiencies, the (FA0.95PbI2.95)0.85(MAPbBr3)0.15 (14, 26), where
hancement of their PCEs is now mainly hampered radiative efficiency drops considerably when MA and FA denote methylammonium and form-
amidinium, respectively. In Fig. 1A, we show
the scanning electron microscopy (SEM) image
Fig. 1. Improved of the control perovskite film. The bright crys-
morphologies and tals on the surface indicate regions of higher
crystal structures. electron density, or regions that accumulated
(A) Top-view SEM charges during the measurement. From energy-
images of the control dispersive x-ray (EDX) analysis (fig. S1), we
and SSG-G films. determined that the lead halide complex mainly
Scale bar, 2 mm. dominated the composition of the bright crystals
(B) XRD patterns (a, d, in the control sample, absent of carbon, so we
and the black square assigned it as PbI1.50Br0.50. We further confirmed
denote the identified this through x-ray diffraction (XRD) analysis of
diffraction peaks
corresponding to the 1
State Key Laboratory for Artificial Microstructure and
black perovskite phase,
Mesoscopic Physics, Department of Physics, Peking
the nonperovskite University, Beijing 100871, China. 2Collaborative Innovation
phase, and PbI1.50Br0.50, Center of Quantum Matter, Beijing 100871, China. 3Clarendon
respectively). a.u., Laboratory, Department of Physics, University of Oxford,
Parks Road, Oxford OX1 3PU, UK. 4Cavendish Laboratory,
arbitrary units.
JJ Thomson Avenue, Cambridge CB3 0HE, UK. 5The Surface
Analysis Laboratory, Faculty of Engineering and Physical
Sciences, University of Surrey, Guildford GU2 7XH, UK.
6
Advanced Technology Institute, University of Surrey,
Guildford GU2 7XH, UK. 7Collaborative Innovation Center of
Extreme Optics, Shanxi University, Taiyuan, Shanxi 030006,
China.
*These authors contributed equally to this work.
†Corresponding author. Email: wz0003@surrey.ac.uk (W.Z.);
henry.snaith@physics.ox.ac.uk (H.J.S.); iamzhurui@pku.edu.cn (R.Z.)
Fig. 3. Surface
photovoltage and
ultraviolet photo-
electron spectros-
copy. (A) Surface
photovoltage measure-
ments of the perov-
skite films coated on
ITO (structure 1) or
ITO/PTAA (structure
2) and of close-to-
complete cells consist-
ing of ITO/PTAA/
perovskite/PC61BM
(structure 3). (B) Helium
Ia (hn = 21.22 eV) spec-
tra of secondary elec-
tron cutoff and valence
band of control and
SSG-G perovskite thin
films. The blue-green
vertical lines indicate
To corroborate the PLQY results with the im- from the PMMA (top) side, and this peak shifted sistent with establishing a wider bandgap close
proved Voc in the solar cells, we measured the from 735 to 767 nm over a few nanoseconds (Fig. to the film surface.
electroluminescence from the PSCs. We observed 4C). However, when we excited the film from the In EDX bromide elemental mapping (fig. S15)
an electroluminescence spectrum at 778 nm for glass side, we only observed the redshifted band and time-of-flight secondary ion mass spectros-
the SSG-G device (fig. S14A), with an external (~767 nm). We further performed transient ab- copy (ToF-SIMS) depth profiles (fig. S16), we
radiative efficiency (ERE) of 1.2% under a forward sorption (TA) measurements by exciting the film observed a bromide-rich region close to the top
bias current density of 23 mA cm−2, but we could with 400-nm (90-fs) pump pulses and probing surface of the SSG-G film. The Br enrichment
not detect electroluminescence from the control with broadband visible pulses. For the control near the top surface is consistent with a wider
device. The ERE value is lower than the PLQY of film, we did not observe a considerable shift in bandgap existing in this region of the film. Our
the perovskite films contacting both PTAA and the ground-state bleach upon exciting from either studies of the spectroscopy and the stability of
PC61BM. This is likely due to the isolated film side (Fig. 4B). In contrast, again, for the SSG-G the PSCs suggest that the presence of the wider
emitting in both directions in the PLQY measure- film, we observed a redshift of the ground-state bandgap top layer is stable. Unexpectedly, we
ments, whereas the PSCs can only emit in the bleach maxima from 730 to 750 nm over the first observed that the guanidinium (GA+) is well
forward direction, implying more reabsorption few nanoseconds when exciting and probing from distributed across the perovskite layer. Because
of emitted light for the latter. Furthermore, there the PMMA side (Fig. 4D); we did not observe such GA+ is a large cation, and because in XRD mea-
were likely imperfections in the deposition of the a redshift when exciting and probing from the surements, we observed negligible shifts in char-
films, resulting in some shunting paths between glass side. acteristic perovskite reflection peaks, we would
the top metal electrode and the perovskite film, These time-resolved PL and TA results suggest not expect a considerable fraction of GA+ to
which would increase dark current and reduce the presence of a wider bandgap (~80 meV wider) be incorporated into the perovskite crystal lat-
the ERE of the PSCs. We estimated a Voc of 1.21 V in the SSG-G film, close to the surface. The re- tice. Therefore, its location is most likely at the
on the basis of the measured radiative efficiency sulting redshift in the time-resolved PL and TA surface and grain boundaries. A substantive
(fig. S14B), which is in direct agreement with spectra over time indicates that the charge pop- contribution to the reduction in nonradiative
our device measurements under simulated 1-Sun ulation migrates from the wider bandgap region recombination may therefore be through a
illumination. to the narrower bandgap bulk. Because we pumped passivation role (33), given that GA+ may be
To examine whether the electronic changes to the perovskite films with high-energy photons present at both top and bottom surfaces and at
the film are homogeneous throughout the thick- (400 nm), we expect that the light was absorbed the grain boundaries.
ness of the perovskite layer or solely present in a strongly in the top region (1/e absorption depth, Multiple factors are thus responsible for the
surface region, we measured time-resolved photo- ~62 nm), and the initial photoexcited charge pop- reduced nonradiative recombination. First, the
luminescence (PL) spectra of perovskite films ulation was hot. The time scale that we determined predominant trap species leading to trap-assisted
[coated on quartz substrates: quartz glass/ for the transfer of the charge population from the recombination in perovskite films are electron
perovskite/poly(methyl methacrylate) (PMMA)] top region to the narrower bandgap bulk region is traps (34). The more n-type perovskite film could
and excited the film from the glass and PMMA on the order of a few nanoseconds—much longer result in a larger fraction of occupied versus
sides, using a 400-nm laser with a ~90-fs pulse than the charge population cooling time, which vacant traps and a reduction in the rate of trap-
width. We observed a single band emission of is expected to occur within 100 fs (32). Therefore, assisted recombination. Secondly, a wider band-
the control film when exciting the film from both we expect that an equilibrium population of free gap near the top surface of the SSG-G film might
the PMMA and the glass sides (Fig. 4A). In con- carriers is created in the topmost region of the reduce the total electron-hole recombination
trast, for the SSG-G film, we observed a dual peak film, before charge transfer toward the bulk, most rate in this region of the film where surface de-
at an early time (0 ns) when exciting the film likely driven by diffusion. These results are con- fects are likely to exist, by making it energetically
favorable for one or both charge carriers to reside 19. W. Nie et al., Science 347, 522–525 (2015). acknowledges a Newton-Bhabha international fellowship. Author
predominantly within the bulk of the film, away 20. Y. Shao, Y. Yuan, J. Huang, Nat. Energy 1, 15001 (2016). contributions: D.L. and R.Z. conceived of the work. D.L. and W.Y.
21. Y. Wu et al., Nat. Energy 1, 16148 (2016). fabricated and characterized solar cells. D.L. and W.Y. conducted
from the surface. Thirdly, the presence of gua- 22. M. Liu, M. B. Johnston, H. J. Snaith, Nature 501, 395–398 UPS, KPFM, electroluminescence, ERE, and EQE measurements.
nidinium halide throughout the films and at the (2013). Z.W. conducted SPV measurements and estimated microstrain
surfaces may inhibit unwanted trap-assisted re- 23. T. Leijtens et al., ACS Nano 8, 7147–7155 (2014). and the voltage radiative limit. Z.W. and R.Sh. conducted PLQY
combination at the heterojunctions by trap pas- 24. X. Zhang et al., J. Phys. Chem. Lett. 7, 4602–4610 (2016). measurements. A.S. and R.H.F. contributed to the photothermal
25. P. Schulz et al., Adv. Mater. Interfaces 2, 1400532 (2015). deflection spectroscopy data. A.S., R.Sh., and R.H.F contributed to
sivation. Hence, the SSG-G process that we have 26. Y. C. Kim et al., Adv. Energy Mater. 6, 1502104 (2016). time-resolved PL and TA data. Q.G. analyzed and discussed time-
presented to improve the perovskite film quality 27. Q. Chen et al., Nano Lett. 14, 4158–4163 (2014). resolved PL and TA spectra. Q.H. conducted SEM experiments on
results in considerable improvements to the elec- 28. K. Tvingstedt et al., Sci. Rep. 4, 6071 (2014). the perovskite films, performed SEM analysis, and conducted
tronic nature at the heterojunctions between the 29. C. Quarti, F. De Angelis, D. Beljonne, Chem. Mater. 29, the XRD measurements. L.Z. and F.Y. analyzed the XRD data.
958–968 (2017). R.Su., T.L., K.C., and D.L. contributed to the certification of solar
perovskite film and charge-extraction layers. Our 30. J. J. Li et al., ACS Appl. Mater. Interfaces 7, 28518–28523 cells. R.Su. and P.W. conducted thermal stability tests on the
findings should be broadly applicable to PSCs and (2015). devices. Y.T. and J.W. conducted EDX measurements, and L.Z.
perovskite light-emitting diodes. 31. C. Bi et al., Nat. Commun. 6, 7747 (2015). contributed to the EDX analysis. Y.Z. prepared metal oxide buffer
32. M. B. Price et al., Nat. Commun. 6, 8420 (2015). layers. Y.Z. and X.Y. measured ultraviolet–visible light absorption
RE FE RENCES AND N OT ES 33. X. Hou et al., J. Mater. Chem. A. 5, 73–78 (2017). spectra. G.F.T., J.F.W., R.Su., D.L., and W.Z. contributed to the
1. K. A. Bush et al., Nat. Energy 2, 17009 (2017). 34. S. D. Stranks et al., Phys. Rev. Appl. 2, 034007 (2014). ToF-SIMS measurement and data analysis. Z.X. conducted
2. J. Zhao et al., Energy Environ. Sci. 9, 3650–3656 (2016). statistical analysis of device efficiencies. D.L. and Z.W. estimated
3. X. Zheng et al., Nat. Energy 2, 17102 (2017). ACKN OWLED GMEN TS the voltage deficit from bandgap to voltage. W.Z., H.J.S., and
4. W. Chen et al., Science 350, 944–948 (2015). We thank S. Hinder from the Faculty of Engineering and Physical R.Z. directed and supervised the project. D.L. and W.Y. wrote the
5. M. Saliba et al., Science 354, 206–209 (2016). Sciences, University of Surrey (UK), for his kind assistance with first draft of the paper. Z.W., W.Z., H.J.S., and R.Z. revised the
6. L. Ling et al., Adv. Funct. Mater. 26, 5028–5034 (2016). ToF-SIMS measurements and helpful discussion; B. Wenger from paper. All authors analyzed their data and reviewed and
7. J. Huang, Y. Yuan, Y. Shao, Y. Yan, Nat. Rev. Mater. 2, 17042 the University of Oxford (UK) for his assistance with PLQY commented on the paper. Competing interests: H.J.S. is chief
(2017). measurements; S. Mahesh from the University of Oxford (UK) for scientific officer of Oxford PV, a company commercializing
perovskite solar cells. Data and materials availability: All data
A
syringe injection through capillary needles and
s an approachable part of the brain, the with organismic responses or behaviors. In vivo to minimize interference with the retina.
retina provides an excellent model for RGC electrophysiology could offer insight into The three-dimensional highly concave mouse
analyzing the assembly and function of the interaction between the retina and related retina precludes using conventional methods, such
information-processing circuits in the cen- brain regions involved in vision processing and as silicon, glass, or metal electrodes (7, 16, 23) or
tral nervous system (CNS) (1, 2). Interneurons regulation (9, 13–15), yet existing technologies planar microelectrode arrays (8), to form a con-
receive signals from light-sensitive photorecep- either have been unable to achieve recordings formal and chronically stable retina interface.
tors (rods and cones) and pass it to retinal gan-
glion cells (RGCs), which send axons through the Fig. 1. Noncoaxial intravitreal
optic nerve to visual areas of the brain. Whereas injection and conformal coating of
photoreceptors are akin to pixels, information mesh electronics on the mouse retina.
processing by interneurons renders each of (A) I: Schematic showing the layout of
≥40 types of RGCs selectively responsive to mesh electronics comprising 16 recording
specific visual features such as motion or color electrodes (green dots indicated by a
contrasts (2–4). However, whereas in vivo single- green arrow) and I/O pads (red dots
neuron recordings in awake, behaving animals indicated by a red arrow). II: Schematic
are routine for many parts of the brain (5, 6), showing noncoaxial intravitreal injection
analysis of RGCs has relied primarily on ex vivo of mesh electronics onto the RGC layer.
electrophysiological recording (7, 8) and calcium Multiplexed recording electrodes are
imaging (4). Although these ex vivo studies have shown as yellow dots. III: Schematic
provided deep insights into retinal computations, of noncoaxial injection that allows
they are limited in several respects. First, sys- controlled positioning of mesh elec-
temic effects such as neuromodulation, altera- tronics on the concave retina surface
tions in hormonal milieu, and circadian variation (cyan arc). The blue and red dotted
are difficult to study ex vivo (9–12). Second, record- arrows indicate the motion of the needle
ings are limited to the short lifetime of the and desired trajectory of the top end of
preparation, typically a few hours, so their ability the mesh, respectively (see fig. S1 for
to detect plasticity in activity patterns is com- details) (22). (B) In vivo through-lens
promised. Third, rod function is prone to rapid images of the same mouse eye fundus
loss in explants, partly because of its dependence on days 0 and 14 after injection of mesh
on pigment epithelium, which is generally re- electronics, with electrode indexing in
moved during explantation. Therefore, ex vivo the day 14 image (22). (C) Ex vivo
recordings of rod activities over extended times imaging of the interface between
have remained challenging (4). Finally, it is ob- injected mesh electronics (red, mesh
viously infeasible to correlate retinal activity ex vivo polymer elements) and the retina (green
dots, RGCs) on days 0 (I) and 7 (II) after
1
injection.The inset of II shows the region
Department of Chemistry and Chemical Biology, Harvard
indicated by a yellow arrow where the
University, Cambridge, MA, USA. 2Center for Brain Science
and Department of Molecular and Cellular Biology, Harvard high-resolution image was taken (22).
University, Cambridge, MA, USA. 3John A. Paulson School of (D) Comparison of pupillary reflex (n = 3),
Engineering and Applied Sciences, Harvard University, OKR (n = 5), and visual acuity (n = 3)
Cambridge, MA, USA. 4Department of Physics, Korea
between control and injected mouse eyes.
University, Seoul, Republic of Korea.
*These authors contributed equally to this work. Error bars denote SD; NS, not significant
†Corresponding author. Email: cml@cmliris.harvard.edu (P > 0.05) by one-way ANOVA test.
(Fig. 1D, II), both of which were also consistent the mesh-injected eyes to exhibit the same visual minimal blockage of incoming light, as evidenced
with previous reports (30). Last, to evaluate the acuity as the control (~0.4 cycles per degree; Fig. by the ~95% light transmittance in the 400- to
impact of the injected mesh on visual acuity, we 1D, III) (31). It is also noteworthy that the trans- 600-nm spectral window visible to the mouse
measured the OKR in response to moving grat- parent polymer constituting the mesh scaffold (32) (fig. S5C, II, inset), resulting in negligible
ings with varying spatial frequencies, and found with <5% space occupied by metal features yields distortion of visual input. Taken together, these
data demonstrate that injection of the mesh elec-
tronics causes minimal damage to the orbicularis
oculi, iris dilator, and extraocular muscles as well
as negligible interference with light perception
and visual acuity of the retina.
Having shown that the mesh electronics probe
has negligible effect on normal visual functions,
we conducted a series of tests to investigate its
ability to detect the diverse RGC activities. First,
we asked how many of the 16 channels in the
implanted probe were sufficiently close to RGCs
to record single-unit activity. Figure 2A and fig.
S6A show examples from two mice in which we
obtained high-quality recordings from all 16 chan-
nels, with a signal-to-noise ratio (SNR) of >7 for
single-unit spikes. Moreover, single-unit activity
Overall, of the 28 RGCs from three mice we into the dynamic information processing be- 27. Y. Zhang, I. J. Kim, J. R. Sanes, M. Meister, Proc. Natl. Acad.
recorded in this regime, 20 exhibited higher tween the retina and other parts of the CNS. Sci. U.S.A. 109, E2391–E2398 (2012).
28. N. Suematsu et al., Front. Syst. Neurosci. 7, 103 (2013).
firing rates during the diurnal phase. Four others 29. D. Zoccolan, B. J. Graham, D. D. Cox, Front. Neurosci. 4, 193 (2010).
exhibited decreased firing rates during the di- 30. K. Yonehara et al., Neuron 89, 177–193 (2016).
urnal phase, and the remaining four showed RE FERENCES AND NOTES 31. G. T. Prusky, N. M. Alam, S. Beekman, R. M. Douglas,
minimal circadian modulation, as assessed 1. R. H. Masland, Neuron 76, 266–280 (2012). Invest. Ophthalmol. Vis. Sci. 45, 4611–4616 (2004).
2. M. Hoon, H. Okawa, L. Della Santina, R. O. Wong, 32. G. H. Jacobs et al., Science 315, 1723–1725 (2007).
from their circadian modulation indices (CMi’s; 33. R. Q. Quiroga, Z. Nadasdy, Y. Ben-Shaul, Neural Comput. 16,
Prog. Retin. Eye Res. 42, 44–84 (2014).
Fig. 4, C and D) (22). Cells that were tracked for 3. J. R. Sanes, R. H. Masland, Annu. Rev. Neurosci. 38, 221–246 1661–1687 (2004).
three complete circadian cycles demonstrated (2015). 34. N. Schmitzer-Torbert, A. D. Redish, J. Neurophysiol. 91,
that RGCs remained in the same circadian mod- 4. T. Baden et al., Nature 529, 345–350 (2016). 2259–2272 (2004).
5. C. M. Lewis, C. A. Bosman, P. Fries, Curr. Opin. Neurobiol. 32, 35. A. Jackson, E. E. Fetz, J. Neurophysiol. 98, 3109–3118 (2007).
ulation categories, despite slight variations in 36. W. Sun et al., Nat. Neurosci. 19, 308–315 (2016).
68–77 (2015).
CMi values (Fig. 4E). Interestingly, of six cells 6. E. J. Hamel, B. F. Grewe, J. G. Parker, M. J. Schnitzer, Neuron 37. L. Gurevich, M. M. Slaughter, Vision Res. 33, 2431–2435 (1993).
for which ON-OFF preferences were measured 86, 140–159 (2015).
carefully, two of three diurnal-high cells were 7. I. J. Kim, Y. Zhang, M. Yamagata, M. Meister, J. R. Sanes, AC KNOWLED GME NTS
ON and one was ON-OFF, both of the nocturnal- Nature 452, 478–482 (2008). We thank J. E. Dowling for helpful discussions, M. Meister for useful
8. G. D. Field et al., Nature 467, 673–677 (2010). suggestions, and J. Huang for help with recording instrumentation.
high cells were OFF, and the sole invariant cell 9. T. A. LeGates, D. C. Fernandez, S. Hattar, Nat. Rev. Neurosci. Funding: Supported by Air Force Office of Scientific Research grant
was ON-OFF, suggesting a correlation between 15, 443–454 (2014). FA9550-14-1-0136, Harvard University Physical Sciences and
RGC polarity and day/night modulation of ac- 10. K. S. Korshunov, L. J. Blakemore, P. Q. Trombley, Front. Cell. Engineering Accelerator award, and a NIH Director’s Pioneer Award
tivity that will be interesting to investigate. The Neurosci. 11, 91 (2017). 1DP1EB025835-01 (C.M.L.); American Heart Association Postdoctoral
11. C. R. Jackson et al., J. Neurosci. 32, 9359–9368 (2012). Fellowship 16POST27250219 and NIH Pathway to Independence
pattern of increased diurnal firing activity for the 12. C. K. Hwang et al., J. Neurosci. 33, 14989–14997 (2013). Award from NIA 1K99AG056636-01 (G.H.); a NIH R37 grant from
majority of RGCs is consistent with results of a 13. B. H. Liu, A. D. Huberman, M. Scanziani, Nature 538, 383–387 NINDS NS029169 (M.Q. and J.R.S.); and the Harvard University Center
for Nanoscale Systems supported by NSF. Author contributions:
M
(Fig. 1, A to D), we first observed the decrease
odern ultrafast laser techniques can bring cess limited by the subsonic melt-front propagation of Laue diffraction peaks (LDPs) intensity owing
materials into states far from thermal speed. However, higher energy densities can to the Debye-Waller effect immediately after laser
equilibrium. These ultrafast processes cause extremely high heating rates that exceed excitation. At 2 ps delay, the heights of diffraction
yield extreme material conditions with 1014 K/s, producing a superheated state in which peaks relative to the adjacent backgrounds show
thermal energy comparable with the homogeneous nucleation occurs catastrophically obvious drops compared with the reference data
Fermi energy and the ion-ion coupling parameter throughout the sample. Early electron-diffraction taken before the arrival of the laser pulse (–2 ps)
exceeding unity, which is referred to as warm experiments observed long melt times in alumi- (Fig. 1D). At 7 ps delay, the data shows a weak
dense matter (1, 2). These conditions exist as a num but did not observe the heterogeneous co- liquid diffraction ring, which is a signature of
transient state in a variety of processes ranging existence (10). Other experiments were performed the formation of a disorderd state. At 17 ps, the
from laser micromachining (3) to inertial confine- in the homogeneous melting regime (7, 8, 11, 12), complete disappearance of the LDPs and the
ment fusion experiments (4). but determining melt times and testing theoretical appearance of the two liquid Debye-Scherrer
In the case of semiconductors, ultrafast optical predictions (9, 13–15) have been elusive. In addi- rings demonstrate that the sample is completely
irradiation can cause strong bond softening and tion, whether the highly excited electron system molten. Such a fast melting process is indicative
nonthermal melting owing to the changes in the can cause bond hardening (8, 16) or softening of homogeneous melting according to MD sim-
potential-energy surface of the lattice by the (17) in gold remains controversial, further com- ulations (9, 27).
excited valence electrons (5, 6). By contrast, melt- plicating the understanding of ultrafast laser– At intermediate energy density of 0.36 MJ/kg
ing of metals is a purely thermal process governed induced solid-liquid phase transitions in metals. (Fig. 1, E to H), the low-order LDPs from regions
by the energy coupling between the excited Visualizing solid-liquid phase transitions and of solid gold and the primary diffraction ring
electrons and relatively cold lattices (7, 8). Two- accurately measuring melt times in the heter- from liquid gold are simultaneously visible at
temperature modeling coupled to molecular ogeneous and homogeneous melting regimes the delay time of 20 ps. Such heterogeneous
dynamics (TTM-MD) simulations predicted the required the development of ultrafast electron coexistence persists over long time scales until
existence of distinct melting regimes in ultrafast diffraction (UED) with mega–electron volt ener- a 800 ps delay, long after electron-ion equilibra-
laser–excited gold (9). At low energy densities, gies (18–20). Because of the reduced space charge tion time of ~50 ps, demonstrating the solid-
the simulations predict that the slow ion heating effect, this device provides high peak currents liquid coexistence at heterogeneous melting
rate will allow the solid-liquid phase transition (~100 mA), enabling measurements with extreme- conditions.
to occur as heterogeneous melting initiated on ly high signal-to-noise ratios. The electron beam At an even lower energy density of 0.18 MJ/kg
liquid nucleation sites on surfaces, grain bound- is produced by means of ultrafast ultraviolet (Fig. 1, I to L), the data show strong LDPs over a
aries, or defects, resulting in a slow melting pro- laser irradiation of a copper cathode and accel- longer duration even at 100 ps, when ion tem-
erated with a linac accelerator-type radio fre- perature Ti should have reached its apex, but
1
SLAC National Accelerator Laboratory, Menlo Park, CA
quency gun; the same laser is split off to heat the the melt front is propagating at a very slow rate.
94025, USA. 2Institut für Physik, Universität Rostock, 18051 sample, providing accurate cross timing between At 1000 ps, the sample is still in a solid-liquid
Rostock, Germany. 3Los Alamos National Laboratory, Bikini laser pump and the electron probe of <30 fs coexistence regime and does not show complete
Atoll Road, Los Alamos, NM 87545, USA. 4Department of [root mean square (RMS)] (21, 22). Furthermore, disappearance of solid diffraction peaks, even at
Physics and Astronomy, University of British Columbia,
Vancouver, BC V6T 1Z1, Canada. 5Department of Electrical
mega–electron volt electrons form a nearly flat delay times as large as 3000 ps. We categorized
and Computer Engineering, University of Alberta, Edmonton, Ewald sphere on the reciprocal space, allowing this case as incomplete melting regime because
AB T6G 2V4, Canada. 6Faculty of Physics and Centre for simultaneous access to multiple orders of diffrac- the energy density deposited in the sample is
Nanointegration Duisburg-Essen, University of Duisburg-Essen, tion peaks (23). Last, multiple elastic scattering below the requirement of complete melting ex-
Lotharstrasse 1, D-47048 Duisburg, Germany.
*Corresponding author. Email: mmo09@slac.stanford.edu
effects are less probable in nanometer-thin films pected at ~0.22 MJ/kg (28).
(M.Z.M.); glenzer@slac.stanford.edu (S.H.G.) at these energies because of their relatively large Our experiment provided high-quality liquid
†These authors contributed equally to this work. elastic mean-free-path (24). diffraction data spanning over a large reciprocal
space that allowed us to determine its corre- respectively. To first order, we assumed a tem- (16) and those values used in (8). Neither of
sponding ion temperature. We realized this porally constant gei and determined its value by these models agreed with our measured (220)
by comparing with the theoretical liquid scat- solving for the ion temperature at complete melt, LDP decay (Fig. 2, A to C). Below the nominal
tering signal based on density functional theory taking into account the energy consumed by latent melting temperature (Tmeltnom
¼ 1340 K), QD showed
(DFT)–MD simulations (23). We performed this heat. The TTM yields gei = (4.9 ± 1) × 1016 W/m3/K striking agreement with the x-ray measurements
analysis for e = 1.17 MJ/kg, which yields a best at 1.17 MJ/kg. We compared the temporal evolu- of gold under thermal equilibrium (31), suggest-
fit T i = 3500 K ± 500 K at the delay time of tion of Te and Ti using this value for gei with those ing that Ti is still the dominant factor for QD in
17 ps, indicating a superheated state. The error based on simulated Te-dependent values for gei nonequilibrium gold at much higher Te. Our find-
bar here represents one standard deviation (SD) from (29) at 1.17 MJ/kg (Fig. 2D). We found that ing is thus different from the previously reported
uncertainty. Te-dependent gei overestimates the ion temper- bond-softening model of gold (17), which ascribed
We characterized the initial and final temper- ature at complete melting by more than 60%. the effect to the highly elevated Te.
atures of the dynamic melting process, and thus We estimated the temporal evolution of the We arrived at the following picture for ultra-
the electron-ion coupling rate gei is constrained Debye temperature QD (23), a manifestation fast melting of gold. We fit our entire dataset over
with a pair of coupled equations of the commonly of interatomic potential (16), directly from the three melting regimes with only one single assump-
used TTM to describe the temperature evolution measured LDP decay using Ti determined from tion, that gei is weakly dependent on energy den-
of both electron and ion subsystems in ultra- the TTM. We observed rapid decay of QD (Fig. 2, sity, modestly increasing from 2.2 × 1016 W/m3/K
fast laser–excited materials (23). We used the G to I). This differs dramatically from both the at the lowest energy density to 4.9 × 1016 W/m3/K
temperature-dependent electron- and ion-specific bond-hardening model based on the Te-dependent for the highest energy density. For example, using
heat Ce(Te) and Ci(Ti) of gold from (29) and (30), phonon spectrum in nonequilibrium conditions gei = 2.2 × 1016 W/m3/K for the lowest energy
ranging from 150 to 300 m/s (35) results in RE FERENCES AND NOTES 33. J. R. Dwyer et al., Phil. Trans. R. Soc. A 364, 741–778
an average distance between nucleation seeds 1. S. Ichimaru, Rev. Mod. Phys. 54, 1017–1059 (1982). (2006).
2. A. L. Kritcher et al., Science 322, 69–71 (2008). 34. J. R. Dwyer et al., J. Mod. Opt. 54, 905–922 (2007).
ranging from 35 to 70 nm, corresponding to
35. F. Celestini, J.-M. Debierre, Phys. Rev. E 65, 041605
a nucleation seed density ranging from 1 × 104 3. R. R. Gattass, E. Mazur, Nat. Photonics 2, 219–225
(2002).
(2008).
to 7 × 104 mm–3. 4. S. H. Glenzer et al., Science 327, 1228–1231 (2010). 36. B. Holst et al., Phys. Rev. B 90, 035121 (2014).
Our data from SC samples showed functional 5. C. W. Siders et al., Science 286, 1340–1342 (1999).
agreement with the TTM-MD simulation results 6. M. Harb et al., Phys. Rev. Lett. 100, 155504 (2008). AC KNOWLED GME NTS
of SC gold by Lin et al. (9) and Mazevet et al. 7. B. J. Siwick, J. R. Dwyer, R. E. Jordan, R. J. D. Miller, Science We thank the SLAC management for the strong support. The
(13). However, the threshold of the transition 302, 1382–1385 (2003). technical support by the SLAC Accelerator Directorate, Technology
8. R. Ernstorfer et al., Science 323, 1033–1037 (2009). Innovation Directorate, LCLS Laser Science Technology Division,
between heterogeneous melting and homoge-
9. Z. Lin, L. V. Zhigilei, Phys. Rev. B 73, 184113 (2006). and Test Facilities Department is gratefully acknowledged. We also
neous melting was found to be higher than pre- 10. S. Williamson, G. Mourou, J. C. M. Li, Phys. Rev. Lett. 52, 2364 thank the technical support on sample manufacturing from the
dicted (9). We speculate that this could be in (1984). Center for Integrated Nanotechnologies, a U.S. Department of
large part due to the embedded-atom method 11. T. Ao et al., Phys. Rev. Lett. 96, 055001 (2006). Energy (DOE) nanoscience user facility jointly operated by Los
potential used in the simulations: The resultant 12. Z. Chen, V. Sametoglu, Y. Y. Tsui, T. Ao, A. Ng, Phys. Rev. Lett. Alamos and Sandia National Laboratories. Funding: This work was
108, 165001 (2012). supported by DOE contract DE-AC02-76SF00515 and the DOE
melting temperature is 963 K, and the thresh- Fusion Energy Sciences under FWP 100182 and partially supported
13. S. Mazevet, J. Clérouin, V. Recoules, P. M. Anglade, G. Zerah,
old for complete melting is 0.13 MJ/kg, both of Phys. Rev. Lett. 95, 085002 (2005). by DOE BES Accelerator and Detector program, the SLAC UED/
which are much lower than experimental obser- 14. K. Lu, Y. Li, Phys. Rev. Lett. 80, 4474–4477 (1998). UEM Initiative Program Development Fund. The support from
vations. Meanwhile, for homogeneous melting, 15. B. Rethfeld, K. Sokolowski-Tinten, D. von der Linde, Natural Sciences and Engineering Research Council of Canada is
also acknowledged. K.S.-T. acknowledges financial support from
the slightly lower tmelt calculated from Mazevet’s S. I. Anisimov, Phys. Rev. B 65, 092103 (2002).
the German Research Council through project C01 “Structural
16. V. Recoules, J. Clérouin, G. Zérah, P. M. Anglade, S. Mazevet,
simulations could be due to the thin-film geom- Phys. Rev. Lett. 96, 055503 (2006). Dynamics in Impulsively Excited Nanostructures” of the
etry not being considered. Moreover, in the Collaborative Research Center SFB 1242 “Non-Equilibrium
17. S. L. Daraszewicz et al., Phys. Rev. B 88, 184101 (2013).
Dynamics of Condensed Matter in the Time Domain.” B.B.L.W. and
T
material with little disorder, especially given the
he search for materials with extreme ther- A number of materials, mostly with complex simple crystal structure (body-centered cubic;
mal properties continues because of their structure, have been proposed for thermoelec- space group I 43m) with only eight atoms per
importance for thermal management appli- tric applications, including half-Heuslers (17, 18), primitive cell, most with relatively small mass
cations. Materials with low k are used in skutterudites (19, 20), and clathrates (21, 22) where [fig. S6 (36)]. Other low-k crystals typically
data storage devices, thermal barrier coat- anharmonic rattling modes give strong intrinsic have much more complex unit cells with many
ings, and thermoelectrics, whereas high-k mate- phonon resistance and suppressed k. The lowest heavy atoms, which provide more phonon scat-
rials are useful for thermal energy transmission reported room temperature k value among bulk tering channels and lower sound velocities. The
and heat dissipation. Energy costs, carbon emis- crystalline TEs is 0.47 W/m-K in SnSe (23). Ther- ultralow k values in Tl3VSe4 prompted us to in-
sions from fossil fuels, and energy wasted as mal conductivity values similar to this are rare vestigate its origin, phonon characteristics (fre-
heat in the world’s energy economy (>60%) drive in crystalline materials, although lower values quencies and l), and relation to the CWP formula.
tremendous interest in advanced processes and are typical in highly disordered or amorphous In Tl3VSe4, Tl atoms are very weakly bonded,
materials for thermal energy transmission, stor- materials where the phonon picture is not ap- having nearest-neighbor distances of 3.22 and
age, and conversion. Toward these goals, discov- plicable. Disordered transport regimes are not 3.95 Å with Se and V atoms, respectively, slightly
ery of higher-efficiency thermoelectric materials yet fully understood, although great strides larger than the sum of the individual ionic radii.
(TEs) for use in waste heat recovery is highly have been taken, experimentally and theoret- Additionally, Tl atoms exhibit nonoverlapping
desirable (1). The efficiency of TEs for thermal- ically (24), to gain insights into these. Most symmetric spherical electron densities with little
to-electric energy conversion is characterized relevant to this work, Cahill, Watson, and Pohl deformation, representative of their 6s2 lone-pair
by the figure of merit; ZT ¼ sS 2 T =k, where s (25) put forth a k model (CWP formula) to de- character (fig. S6). The V and Se atoms form
is electrical conductivity, S is the Seebeck co- scribe a lower bound to k based on random hops VSe34 tetrahedral units with equal V–Se bond
efficient, and k is thermal conductivity. Numerous of “instantaneously” localized (not to be confused distances of 2.29 Å (see tables S1 and S2 for
schemes for improving the power factor (sS 2 ) with Anderson localization) vibrational thermal structural information), and the total charge den-
(2–8) and minimizing k (9–16) have been devised energy among uncorrelated oscillators in glassy sity between V and Se atoms shows considerable
to achieve enhanced ZT. However, the complex and amorphous materials. This was first pro- overlap of the electronic densities, representative
interplay of s, S, and k remains a challenging posed by Einstein to explain the measured k of of hybridization in the VSe3 4 tetrahedra. These
bottleneck toward further advances. Disordered KCl (25, 26) and failed dramatically as phonons electronic and structural features suggest ionic
and amorphous materials typically exhibit ultra- carry the majority of heat in that material. In bonding between Tl and Se or V atoms and co-
low k values owing to lack of lattice periodicity, disordered materials, k monotonically increases valent bonding between V and Se atoms. Given
but these also give low s. Therefore, theoretical with temperature (T) before saturating above the the weak atomic bonding to the Tl3VSe4 lattice,
and experimental efforts targeting quality crys- Debye temperature, thus having an effective Tl atoms have relatively large thermal displace-
talline materials with very low k and large power minimum k nearly independent of T for a broad ments at room temperature, as shown by the
factors have been a strong driver in thermoelec- range of T (27). In contrast, phonon k in crys- calculated mean square displacements (MSDs)
tric research. talline materials typically increases with T from (fig. S7). At 300 K, the calculated MSD for Tl
T = 0 following a Debye T 3 behavior, peaks at atoms is 0.067 Å2, larger than that of V (0.017 Å2)
intermediate T, often dictated by isotope or and Se (0.027 Å2) atoms. MSD data from powder
1
Materials Science and Technology Division, Oak Ridge
defect scattering, and then exhibits a near 1/T x-ray diffraction measurements give similar ratios
National Laboratory, Oak Ridge, TN 37831, USA. 2NRC behavior with increasing T above the Debye but smaller values (also shown in fig. S7). The
Research Associate at Naval Research Laboratory, temperature owing to intrinsic umklapp scat- large MSD values indicate that the Tl atoms reside
Washington, DC 20375, USA. 3Center for Nanophase tering. For complex crystalline materials (28–30) in a relatively flat potential energy environment
Materials Sciences, Oak Ridge National Laboratory, Oak
Ridge, TN 37831, USA.
whose phonon mean free paths (l) are strongly and that displacing them from their equilibrium
*Corresponding author. Email: lindsaylr@ornl.gov (L.L.); salesbc@ suppressed, approaching lengths on the order sites costs little energy. This scenario is similar to
ornl.gov (B.C.S.); saikat.mukhopadhyay.ctr.in@nrl.navy.mil (S.M.) of interatomic spacings [Ioffe-Regel limit (31)] Zintl-type TlInTe2 (37, 38) where it was argued
Fig. 1. Thermal conduc- thus short l (Fig. 3A) and ultralow k (Fig. 1). As
tivity of Tl3VSe4. revealed by the projected phonon density of states
Measured thermal con- (Fig. 2A), the acoustic phonons with short l are
ductivity (kmeasured) and primarily derived from vibration of the heavy Tl
calculated values from atoms. Therefore, we conclude that a combination
the PBT (kphonon). The of very low group velocities and strong anharmo-
proposed ktwo-channel nicity governs the ultralow k behavior in Tl3VSe4.
values using the CWP This extends to the other compounds that we
formula and the Einstein investigate in the Tl3XSe4 (X is V, Nb, Ta) family
k formula give the range (figs. S10 to S13) and highlights that the Tl atoms
of values shown by the and their associated chemical bonding environ-
shaded area between ment dictate the lattice dynamical properties in
the red curves. A small Tl3XSe4.
“radiation tail” can be These ultralow-k semiconducting thallium-
seen in the measured based systems may also show promise as efficient
Tl3VSe4 k data near low-temperature TEs, provided that the electron-
300 K. For comparison, ic properties and other factors (stability, ability
k values of low-k to dope, and so forth) are also beneficial. Our
materials at 300 K calculations of the electronic structure of Tl3VSe4
from the literature are reveal a bandgap (Eg) of 1.5 eV (fig. S9) lying
shown: PbTe (47), within the range of good TEs. The calculated Eg
KSbS2 (48), Cu2Se (49),
temperature dependence that we measured for and disordered materials with so-called prop- stant) for the heavy Tl atom [48 K (measured)
k strongly suggests that another transport mecha- agons, locons, and diffusons (24, 44). Further- and 32 K (calculated) at 300 K (36)] or the Raman
nism is at play. Electronic k contributions can more, guest atoms in weakly bonded systems (e.g., lifetimes (t) using qE ¼ pħ=kB t (36 K), all of
be eliminated as Tl3VSe4 is a semiconductor with clathrates and filled skutterudites) were previ- which are moderately constistent. We note that
a notable electronic bandgap (figs. S2 and S9). ously suggested as Einstein oscillators in host specific heat from measurements and calculations
We propose that a phonon conduction channel substructures that are treated within the Debye compare favorably for T > 20K and give similar
and a hopping channel of “instantaneously” local- model. This Debye-Einstein combination was “Einstein peaks” at low T; however, the purely
ized vibrations coexist (ktwo-channel = kphonon + used to explain various thermodynamic prop- harmonic phonon theory overpredicts the mea-
khop) in crystalline materials, particularly with erties of this class of materials (21, 38). sured specific heat at very low T (fig. S3). Combining
loosely bound atoms or rattling modes. The lo- We calculated khop via the CWP formula (25) kphonon (excluding phonons with mean free paths
calized vibrational energy from “phonon” modes and from the Einstein (26) k formula from which below the Ioffe-Regel limit) and khop from these
with mean free paths below the Ioffe-Regel limit it is derived (36). (As an interesting side note, models gives values that are consistently in better
(Fig. 3A) is conducted in the khop channel (model Einstein’s original estimation of k ~ 0.3 W/m-K agreement with measured k and its temperature
described below), and kphonon represents the usual for KCl is similar to the calculated and mea- dependence (Fig. 1), with the CWP formula giving
conduction from “well-defined” phonons with sured values that we obtained for Tl3VSe4.) The the highest values and the Einstein k formula
mean free paths greater than this limit. The khop CWP formula requires sound velocities as input, with qE ¼ 22 K giving the lowest. Indeed, using
channel only becomes apparent when the pho- whereas the Einstein k formula requires defining qE ¼ 65 K gives excellent agreement with the
non contribution to k is very small, as in Tl3VSe4. a constant frequency/temperature (qE ) for the measured data, though not justified by other
The idea of a phenomenological two-channel oscillators (36). This can be estimated from the measurements. We examined measured and
model has been previously introduced to describe low-temperature specific heat [22 K (measured) calculated k in other low-k materials from the
amorphous glasses (from which the “hop” term is and 18 K (calculated); fig. S3], from the MSD = literature (28, 45, 46) and found large improve-
ħ 2 T =mkB q2E (ħ is the reduced Planck’s constant, ments (see Table 1 and fig. S14 for T dependence)
formula. This suggests that two vibrational heat- 18. J. R. Sootsman, D. Y. Chung, M. G. Kanatzidis, Angew. Chem. 48. E. J. Skoug, D. T. Morelli, Phys. Rev. Lett. 107, 235901
conduction channels coexist in crystalline mate- Int. Ed. 48, 8616–8639 (2009). (2011).
19. W. Ren, H. Geng, Z. Zhang, L. Zhang, Phys. Rev. Lett. 118, 49. H. Kim et al., Acta Mater. 86, 247–253 (2015).
rials, and the hopping channel becomes evident
245901 (2017). 50. C. B. Satterthwaite, R. W. Ure, Phys. Rev. 108, 1164–1170
in materials with small phonon k. 20. B. C. Sales, D. Mandrus, R. K. Williams, Science 272, (1957).
In this study, we found that crystalline Tl3VSe4 1325–1328 (1996). 51. G. A. Slack, P. Andersson, Phys. Rev. B 26, 1873–1884
has ultralow thermal conductivity (kmeasured = 21. T. Takabatake, K. Suekuni, T. Nakayama, E. Kaneshita, (1982).
0.30 W/m-K, kcalculated = 0.16 W/m-K at room Rev. Mod. Phys. 86, 669–716 (2014). 52. B. Wölfing, C. Kloc, J. Teubner, E. Bucher, Phys. Rev. Lett. 86,
22. G. S. Nolas, J. L. Cohn, G. A. Slack, S. B. Schujman, 4350–4353 (2001).
temperature), despite lacking disorder. The ultra-
Appl. Phys. Lett. 73, 178–180 (1998). 53. D. C. Schmitt et al., J. Am. Chem. Soc. 134, 5965–5973
low k stems from very small acoustic group ve- 23. L.-D. Zhao et al., Nature 508, 373–377 (2014). (2012).
locities from heavy, loosely bound Tl atoms and 24. W. Lv, A. Henry, Sci. Rep. 6, 35720 (2016). 54. H. Lin et al., Angew. Chem. Int. Ed. 55, 11431–11436
strong anharmonicity likely arising from Tl s2 25. D. G. Cahill, S. K. Watson, R. O. Pohl, Phys. Rev. B Condens. (2016).
lone-pair repulsion. Describing k as the sum of Matter 46, 6131–6140 (1992). 55. X. Wang, C. D. Liman, N. D. Treat, M. L. Chabinyc, D. G. Cahill,
26. A. Einstein, Ann. Phys. 340, 679–694 (1911). Phys. Rev. B 88, 075310 (2013).
two separate vibrational transport channels, pho- 56. J. C. Duda, P. E. Hopkins, Y. Shen, M. C. Gupta, Phys. Rev. Lett.
27. J. J. Freeman, A. C. Anderson, Phys. Rev. B Condens. Matter
nons, and random walk among uncorrelated 34, 5684–5690 (1986). 110, 015902 (2013).
oscillators gives substantially better agreement 28. W. Lee et al., Proc. Natl. Acad. Sci. U.S.A. 114, 8693–8697
with measured k data and its temperature de- (2017).
AC KNOWLED GME NTS
pendence. The latter channel is only apparent 29. J. He et al., Phys. Rev. Lett. 117, 046602 (2016).
30. W. Qiu et al., Proc. Natl. Acad. Sci. U.S.A. 111, 15031–15035 We acknowledge helpful discussion with D. Cahill,
in crystals for which phonons carry little heat and D. Broido, P. Allen, and D. Mandrus. Funding: This work
(2014).
explains some previous discrepancies of measured 31. A. F. Ioffe, A. R. Regel, in Progress in Semiconductors,
was supported by the U.S. Department of Energy (DOE), Office
and calculated k values in other low-k materials. A. F. Gibson, F. A. Kroger, R. E. Burgess, Eds. (Wiley, New York, of Science, Basic Energy Sciences, Materials Sciences and
The improvement in theory for ultralow-k crys- Engineering Division. S.M. was in part supported by the
1960), vol. 4.
NRC-NRL Research Associateship Program for supplemental
32. A. Jain, A. J. H. McGaughey, Comput. Mater. Sci. 110, 115–120
S
pling variation. Within the limits of precision, su-
0.8
1.1 0.7
0.6 LEGEND
1 LEGEND
0.5
HMD HMD
0.9 0.4
istat Predicted
0.8 0.3
Hazard (log scale)
0.7 0.2
0.6
0.1
0.5
0.4 0.05
0.3
0.02
Fig. 1. Yearly hazards on a logarithmic scale for the cohort of Italian women born in 1904. Confidence intervals were derived from Human Mortality
Database (HMD) data for ages up to 105 and from ISTAT data beyond age 105. (A) Closeup with 95% confidence intervals based solely on single-cohort data.
(B) Broad view with estimated plateau beyond age 105 (black dashed line) and 95% confidence bands (orange) predicted from the model parameters based on the
full ISTAT database, along with a straight-line prediction (black) from fitting a Gompertz model to ages 65 to 80.
Table 1. Distribution of the observed cases. Age at entry into study is given in computed years.
1896 112 0 0 0 4 ≥4
............................................................................................................................................................................................................................................................................................................................................
1897 111 0 0 0 1 ≥1
............................................................................................................................................................................................................................................................................................................................................
1898 110 0 0 0 5 ≥5
............................................................................................................................................................................................................................................................................................................................................
1899 109 0 0 1 12 ≥13
............................................................................................................................................................................................................................................................................................................................................
1900 108 0 0 0 23 ≥23
............................................................................................................................................................................................................................................................................................................................................
1901 107 0 7 0 46 ≥53
............................................................................................................................................................................................................................................................................................................................................
1902 106 0 17 2 134 ≥153
............................................................................................................................................................................................................................................................................................................................................
1903 105 0 23 2 195 ≥220
............................................................................................................................................................................................................................................................................................................................................
1904 105 0 35 5 302 342
............................................................................................................................................................................................................................................................................................................................................
1905 105 2 40 10 331 383
............................................................................................................................................................................................................................................................................................................................................
1906 105 2 48 19 348 417
............................................................................................................................................................................................................................................................................................................................................
1907 105 11 55 40 354 460
............................................................................................................................................................................................................................................................................................................................................
1908 105 19 57 106 345 527
............................................................................................................................................................................................................................................................................................................................................
1909 105 28 33 219 296 576
............................................................................................................................................................................................................................................................................................................................................
has recently collected and validated the indi- fore 31 December 1903 (birth cohorts 1896–1903) to alternatives with a nonzero Gompertz slope
vidual survival trajectories of all inhabitants entered the study at ages older than exactly 105 parameter. We include a (modest) exponential
of Italy aged 105 and older in the period from and, as such, provided left-truncated survival cohort trend and a proportional gender effect,
1 January 2009 to 31 December 2015; these data trajectories. Death during the follow-up was ob- setting the hazard at age x years beyond 105
were used for the present study. For several rea- served in 2883 cases; as a result, 953 individuals equal to
sons, these data allow estimation of mortality at were right-censored (i.e., still alive at the end of aebx eb1 C þ b2 M
extreme ages with accuracy and precision that the study). Table 1 displays the distribution of
were not possible before. First, individual trajec- observed deaths and censored trajectories with b constrained to zero for the null model.
tories provide information on survival in contin- across gender and cohort. Increases in samples Here, C is cohort birth year minus 1904, and
uous time, therefore avoiding possibly misleading from row to row bear testimony to improvements M = 1 for males but is otherwise set to 0. Param-
patterns of death rates that are computed on in survival from cohort to cohort at ages before eters include initial hazard a at 105, Gompertz
prespecified age intervals and are often ob- 105 and lead us to expect the downward cohort slope b, cohort effect b1, and gender effect b2.
tained by aggregating heterogeneous birth co- trend in hazards beyond age 105 in our data to Parameters estimated by standard maximum
horts. Second, the validation procedure has be described below. likelihood methods for truncated and censored
been developed specifically for this population For context, Fig. 1A shows confidence intervals survival data (14) are shown in Table 2. A like-
segment and meets the highest validation crite- for logarithms of yearly hazards for the single- lihood ratio test fails to reject the constant-
ria provided by the IDL protocol. It is based on year cohort of Italian women born in 1904. For hazard null model at a level as generous as 0.44.
the resident population of the Italian munic- ages before 105, intervals were derived from Under the alternative hypothesis, the Gompertz
ipalities that is recorded yearly on 1 January. vital statistics in the Human Mortality Database slope parameter estimate b = 0.013 [standard
Each municipality where individuals aged ≥105 (www.mortality.org). These widening intervals, error (SE) = 0.017] is not statistically significant
have been reported was contacted by ISTAT. A also likely distorted by age misreporting, only hint at the 5% level and is practically indistinguish-
death certificate was required for each deceased at decreasing slopes. Beyond age 105, intervals able from 0. This near-negligible slope stands
subject. This certificate includes, among other were derived from ISTAT data restricted to this in contrast to the slope as large as 0.103 at younger
information, the date of birth of the deceased single cohort, with separate intervals for each ages (65 to 80) in Fig. 1B, which is paired with
individual, as certified by the civil status officer. year of age. Even with these high-quality data, a log hazard at 65 of log(0.015). For variant mod-
A certificate of survival was required for all separating out cohorts and ages leaves too much els and power calculations, see tables S1 and S2.
individuals who were expected to still be alive at uncertainty to tell whether hazards continue The estimated cohort effect b1 = −0.020 (SE =
the end of the study period. For supercentenar- upward, level out, or decrease beyond 105. Hence, 0.008), though small, is in line with expectations,
ians, those most problematic in terms of age we fit a model that combines cohorts and ages statistically significant, and noteworthy (13). The
reporting, birth certificates were also collected. to circumvent this challenge. Our best-estimated 463 male survivors older than 105 are too few for
Hence, age misreporting is believed to be min- trajectory for the 1904 cohort from our modeling the gender effect to come out statistically signif-
imal in these data. The project includes all is the flat curve (the plateau) shown in Fig. 1B. icant, though the estimate b2 = 0.033 is plausible.
individuals 105 and older in the period from On a log scale, exponential curves become straight For the baseline cohort born in 1904, the es-
1 January 2009 to 31 December 2015, so the data lines. A straight-line fit based on ages 65 to 80, timated level of the plateau is a = 0.645. It
are also free of age ascertainment bias. where the Gompertz model does appear to hold, corresponds to an annual probability of dying of
The present study based on ISTAT data in- fails at older ages and far overshoots our esti- 1 − e−0.645 = 0.475 and an expectation of further
cludes 3836 cases, 463 of whom are males, across mated plateau beyond age 105. life of 1/0.645 = 1.55 years. This outcome is con-
15 birth cohorts (one for each year from 1896 to To determine from the full ISTAT data wheth- sistent with the probability estimated elsewhere
1910). Fewer than 4% of these individuals were er log-hazard slopes are level, upward, or down- for supercentenarians (12). With 90% of person-
born abroad. Of those, many have clear Italian ward after age 105, our modeling approach years at risk (a measurement of total time at
heritage (13). Altogether, 472 individuals born be- compares a null hypothesis of constant hazards risk) coming before age 108, the ISTAT data do
b.....................................................................................................................................................................................................................
2 0.033 (0.058) REFERENCES AND NOTES
Gompertz hazard model
..................................................................................................................................................................................................................... 1. B. Gompertz, Philos. Trans. R. Soc. Lond. 115, 513–583 (1825).
a 0.629 (0.026) −4250.370 8508.740
.....................................................................................................................................................................................................................
2. S. Horiuchi, A. J. Coale, Math. Popul. Stud. 2, 245–267 (1990).
3. V. Kannisto, Development of Oldest-Old Mortality, 1950-1990:
b 0.013 (0.017)
..................................................................................................................................................................................................................... Evidence from 28 Developed Countries (Odense Univ. Press,
b.....................................................................................................................................................................................................................
1 −0.016 (0.009) 1994).
b.....................................................................................................................................................................................................................
2 0.034 (0.058) 4. S. Horiuchi, J. R. Wilmoth, Demography 35, 391–412 (1998).
5. A. R. Thatcher, V. Kannisto, J. W. Vaupel, The Force of Mortality
at Ages 80 to 120 (Odense Univ. Press, 1998).
6. J. Robine, J. W. Vaupel, Exp. Gerontol. 36, 915–930 (2001).
7. E. Barbi, G. Caselli, J. Vallin, Population 58, 43–65 (2003).
8. J. W. Vaupel et al., Science 280, 855–860 (1998).
5
9. L. A. Gavrilov, N. S. Gavrilova, N. Am. Actuar. J. 15, 432–447
LEGEND (2011).
Constant 10. N. S. Gavrilova, L. A. Gavrilov, J. Gerontol. Ser. A 70, 1–9 (2015).
pp. 1–24.
12. J. Gampe, in Supercentenarians, H. Meier, J. Gampe, B. Jeune,
3 J. M. Robine, J. W. Vaupel, Eds. (Springer, 2010), pp. 219–230.
13. Materials and methods are available as supplementary
materials.
14. J. P. Klein, M. L. Moeschberger, Survival Analysis (Springer,
2
2003), pp. 74–95, 91–98.
15. X. Dong, B. Milholland, J. Vijg, Nature 538, 257–259 (2016).
16. A. Lenart, J. W. Vaupel, Nature 546, E13–E14 (2017).
1 17. H. Rootzén, D. Zholud, Extremes 20, 713–728 (2017).
18. J. W. Vaupel, Nature 464, 536–542 (2010).
19. J. W. Vaupel, K. G. Manton, E. Stallard, Demography 16,
439–454 (1979).
0 20. R. E. Beard, in The Lifespan of Animals, G. E. W. Wolstenholme,
M. O’Connor, Eds. (Little, Brown, 1959), pp. 802–811.
105 106 107 108 109 110 111 112
21. M. S. Finkelstein, V. Esaulova, Adv. Appl. Probab. 38, 244–262
Age (years) (2006).
22. D. R. Steinsaltz, K. W. Wachter, Math. Popul. Stud. 13, 19–37
Fig. 2. Cumulative hazard beyond age 105 for the cohort of Italian women born in 1904, as (2006).
23. T. I. Missov, M. Finkelstein, Theor. Popul. Biol. 80, 64–70 (2011).
determined by the Nelson-Aalen estimator. Straight lines represent cumulative hazards of the
24. T. I. Missov, J. W. Vaupel, SIAM Rev. 57, 61–70 (2015).
estimated plateau predicted from ISTAT data, under a constant hazard (light blue) and a Gompertz 25. K. W. Wachter, D. Steinsaltz, S. N. Evans, Proc. Natl. Acad.
hazard model (darker blue). The shaded area indicates the 95% confidence bands of the Sci. U.S.A. 111 (suppl. 3), 10846–10853 (2014).
Nelson-Aalen estimate.
AC KNOWLED GME NTS
We thank M. Battaglini and G. Capacci at ISTAT for collecting and
not enable us to rule out alternatives, such as a fixed maximum life span for humans, under- validating the data used in this paper. Funding: K.W.W. was
plateau followed somewhat later by a decline, writing doubt that any limit is as yet in view. supported by grant 5P30AG012839 from the U.S. National Institute
but supercentenarian estimates provide indica- Our findings further provide fundamental on Aging. Author contributions: E.B. wrote the paper; F.L.
performed the statistical analyses; M.M. designed the data
tions against such alternatives. knowledge about the biodemography of human
validation procedure and supervised the data collection; J.W.V.
Our estimates based on all ISTAT cohorts to- longevity. By using clean data from a single initiated the research project and suggested revisions to
gether produce excellent fits for single cohorts. nation and straightforward estimation methods, subsequent drafts; and K.W.W. suggested extensions and
We examine the cumulative hazard, the integral we have shown that death rates, which increase revisions. All authors contributed to the interpretation of results.
Competing interests: The authors declare no competing interests.
under the hazard curve, for which nonparametric exponentially up to about age 80, do decelerate
Data and materials availability: The data that support the
confidence bounds are available (14). Hazards thereafter and reach or closely approach a pla- findings of this study are owned by ISTAT and are not publicly
that are constant imply cumulative hazards that teau after age 105. Thus, these well-estimated available. However, the data can be obtained directly from ISTAT
are linearly increasing; poor fit would stand out hazard curves share the qualitative pattern ob- by registering at the Contact Center (https://contact.istat.it)
and mentioning the Semisupercentenarian Survey and M.M. as
as curvature. The plot in Fig. 2 shows absence served for extreme ages in widely differing species
contact person. The computer codes used to generate the results
of curvature in the data for Italian women born (8, 18), regularities calling for common structural reported in the manuscript are available at https://
in 1904. and evolutionary explanations. scienzepolitiche.uniroma3.it/flagona/publications-en/.
The increasing number of exceptionally long- An important structural contributor to mor-
lived people (Table 1) and the fact that their tality rate deceleration must be the impact of
SUPPLEMENTARY MATERIALS
mortality beyond 105 is seen to be declining selective survival in heterogeneous populations.
www.sciencemag.org/content/360/6396/1459/suppl/DC1
across cohorts—lowering the mortality plateau The fixed-frailty proportional hazard model of Materials and Methods
or postponing the age when it appears—strongly Vaupel et al. (19) [with precursor (20)] implies Fig. S1
suggest that longevity is continuing to increase approach to plateaus (5, 8, 18), and Gamma- Tables S1 to S3
References (26, 27)
over time and that a limit, if any, has not been Gompertz distributions for deaths arise natu-
reached. Our results contribute to a recently re- rally in the framework (21–24). Enhanced care 13 February 2018; accepted 9 May 2018
kindled debate (15–17) about the existence of a for the extremely old may help to mitigate in- 10.1126/science.aat3119
health care spending in late life the concentration of spending at the end of life.
We report results for two spending measures.
The first, which we refer to as “backfilling,” follows
Liran Einav1,2, Amy Finkelstein1,3*, Sendhil Mullainathan1,4, Ziad Obermeyer5 the approach of the end-of-life literature (13). For
survivors, it measures spending over the relevant
That one-quarter of Medicare spending in the United States occurs in the last year of life is time interval from 1 January 2008 going forward;
commonly interpreted as waste. But this interpretation presumes knowledge of who will for decedents, it measures spending starting from
die and when. Here we analyze how spending is distributed by predicted mortality, the date of death in 2008 and going backward
based on a machine-learning model of annual mortality risk built using Medicare claims. over the same length of time. Using this approach,
Death is highly unpredictable. Less than 5% of spending is accounted for by individuals we estimate that the 5% of Medicare beneficiaries
with predicted mortality above 50%. The simple fact that we spend more on the sick—both who died accounted for 21% of Medicare spend-
on those who recover and those who die—accounts for 30 to 50% of the concentration of ing, closely matching prior estimates (13).
spending on the dead. Our results suggest that spending on the ex post dead does not This standard analysis suffers from two related
necessarily mean that we spend on the ex ante “hopeless.” biases: We do not know who will die in a given
O
time interval, or when, within that interval, they
nly 5% of Medicare beneficiaries in the (6, 7); see also (8) for early empirical analysis. will die. We therefore also analyze what we refer
to as “unadjusted spending,” for which we mea-
Share of individuals
their trajectories over the prior 12 months. This
produces thousands of potential predictors. We 0.003
use an ensemble (of random forest, gradient 0.3
boosting, and LASSO)—a standard and pop- 0.002
ular machine-learning technique—to generate 0.001
mortality predictions. To avoid overfitting, we
randomly split the data into a “training” sub- 0
Share of individuals
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
sample, for which we develop the prediction
algorithm, and a “test” subsample, for which we 0.2
Predicted mortality
apply the resulting algorithm to generate pre-
dicted mortalities. All subsequent results are for
this test subsample, which is one-third of our
original sample. How we construct the potential
mortality predictors and the prediction algorithm
is described in detail in the supplementary ma- 0.1
terials, section B. It shows that predicted mor-
tality varies in sensible ways with individual
characteristics and that our algorithm’s perform-
ance is comparable to other recent mortality-
prediction endeavors.
0.8
above 46% percent, accounts for under 5% of
total spending, and 45% of these individuals are Survivors 0.08
survivors. To capture a group of decedents who
account for at least 5% of total spending, we must Decedents
0.04
set a threshold of predicted mortality of 39% or 0.6
higher. These results are based on the backfilled 0
measure of decedent spending; when using the
0.25
0.30
0.35
0.40
0.45
0.50
0.55
0.60
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
unadjusted measure, spending on decedents is
even lower, so that a smaller share of spending 0.4
above each mortality prediction threshold is
accounted for by decedents.
A natural question is whether these results
would change if we had better predictions, for 0.2
example, made with higher quality data such as
electronic medical records. The available evidence,
although limited, suggests that, relative to using
only (detailed) claims data, the incremental pre-
0
dictive power obtained from electronic medical 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
records (14) or subjective physician predictions Predicted mortality
(15, 16) is relatively small. Moreover, such data
are arguably less relevant for national policy, Fig. 3. Concentration of spending by ex ante mortality. For each level of predicted annual mortality
which needs to be based on standardized, na- (x axis), the share of total annual Medicare spending accounted for by individuals with predicted
tionally available data. mortality of that value or greater is shown. Each bar stacks the share accounted for by decedents (black)
There is also the possibility of better prediction and by survivors (gray), so that the height of the bar represents total annual Medicare spending
algorithms. Indeed, some cutting-edge machine- accounted for by individuals (decedents and survivors) with predicted mortality of that value or greater.
learning methods (17, 18) do better in select patient All results use the backfilled measure of decedent spending. All data are from the test subsample.
groups. To study how a hypothetical, better pre- The inset provides more detail about the corresponding section of the distribution.
dictor might plausibly affect our results, we pro-
duce an artificial “oracle” predictor by adjusting or does not). If we put a weight of 0.1 on the only account for 5% of total spending. This hap-
predicted probabilities toward realized outcomes realized outcome, this generates an area under pens because, at low baseline mortality rates (i.e.,
(i.e., increasing predictions for the dead and the curve (AUC) of 0.963—a level of algorithm annual mortality rate of 5%), models can be
lowering them for survivors); our hypothetical performance well above any in the literature— extremely good at identifying those at high risk
predictor is thus a weighted average of our actual but our results do not qualitatively change: Indi- (i.e., AUC can be extremely high), but the highest
predictor and the realized outcome (death occurs viduals with predicted mortality above 47% still percentiles can still have modest absolute rates
A 30,000
B 75,000
Decedents -- backfilled
60,000 Decedents -- unadjusted
Survivors
Total spending ($US)
30,000
10,000
15,000
0 0
0.008 0.016 0.031 0.063 0.125 0.250 0.500 1.000 0.008 0.016 0.031 0.063 0.125 0.250 0.500 1.000
Predicted mortality (log scale) Predicted mortality (log scale)
Fig. 4. Spending by predicted mortality. (A) Kernel density spending separately for survivors and decedents. Spending
of total Medicare spending in the 12 months after 1 January 2008 measures are as defined in Fig. 1. All data are from the test
of predicted mortality (under 50%). As a result, subsequent decedents and survivors. Using these care interventions on survival rates and, just as
there is little concentration of spending on indi- estimates, we find that survivors randomly sam- importantly, on palliation of symptoms; such re-
viduals with high absolute rates of predicted pled from the decedents’ distribution of predicted search should focus not just on averages but also
mortality. More details are provided in the sup- mortality spend about twice as much on health on potentially heterogeneous impacts across dif-
plementary materials, section C. care as a randomly sampled survivor. As a re- ferent individuals (19–21).
Nor do our conclusions change when we view sult, 30 to 50% of the concentration of spend-
the prediction task from an arguably more ing on decedents relative to survivors would REFERENCES AND NOTES
“decision-relevant” time point: when potentially be eliminated (depending on whether the un- 1. G. F. Riley, J. D. Lubitz, Health Serv. Res. 45, 565–576 (2010).
2. A. Gawande, “Letting go,” New Yorker, 2 August 2010;
costly medical treatment decisions are made, adjusted or backfilled spending measure was www.newyorker.com/magazine/2010/08/02/letting-go-2.
at hospital admission. In the supplementary ma- used), simply by accounting for the fact that spend- 3. E. Porter, “Rationing health care more fairly,” New York Times,
terials, section D, we reestimate the prediction ing is higher on those with higher mortality risk. 21 August 2012; www.nytimes.com/2012/08/22/business/
algorithm to generate 12-month mortality pre- More details are provided in the supplementary economy/rationing-health-care-more-fairly.html.
4. E. J. Emanuel, L. L. Emanuel, N. Engl. J. Med. 330, 540–544 (1994).
dictions at the time of hospital admission for the materials, section E. 5. Medicare Payment Advisory Commission, “Report to the
subsample of individuals admitted to the hospital However, Fig. 4B also shows that, even for Congress: Selected Medicare Issues," June 1999; https://babel.
during 2008. We use the same predictors, now individuals with the same predicted mortality hathitrust.org/cgi/pt?id=mdp.39015046749704;view=1up;seq=3.
measured in the 12 months before admission, probability, spending is higher for those who sub- 6. A. A. Scitovsky, Milbank Q. 83, 825–841 (2005).
7. J. P. Newhouse, Health Aff. (Millwood) 12 (suppl. 1), 152–171 (1993).
as well as the admitting diagnosis. Even from the sequently die, particularly for individuals with 8. A. S. Detsky, S. C. Stricker, A. G. Mulley, G. E. Thibault, N. Engl.
vantage point of admission to the hospital, where the lowest predicted mortality. This may be be- J. Med. 305, 667–672 (1981).
annual mortality is about 20%, the 95th percentile cause of ex ante differences across patients that 9. M. M. Desai, S. T. Bogardus Jr., C. S. Williams, G. Vitagliano,
of annual death probabilities is still only 67%. Less our current prediction algorithm does not in- S. K. Inouye, J. Am. Geriatr. Soc. 50, 474–481 (2002).
10. J. J. Gagne, R. J. Glynn, J. Avorn, R. Levin, S. Schneeweiss,
than 4% of those who end up dying in the sub- corporate, or it may be related to the process by J. Clin. Epidemiol. 64, 749–759 (2011).
sequent year have a predicted mortality above which individuals die or even the basic mechanics 11. M. Makar, M. Ghassemi, D. M. Cutler, Z. Obermeyer,
80% at the time of admission. Even if we zoom of death. More work is needed to fully understand Int. J. Mach. Learn. Comput. 5, 192–197 (2015).
in further on the subsample of individuals who why death remains expensive, even conditional 12. L. C. Yourman, S. J. Lee, M. A. Schonberg, E. W. Widera,
A. K. Smith, JAMA 307, 182–192 (2012).
enter the hospital with metastatic cancer—63% die on mortality risk. 13. C. Hogan, J. Lunney, J. Gabel, J. Lynn, Health Aff. (Millwood)
over the subsequent 12 months, but they account In sum, although spending on the ex post dead 20, 188–195 (2001).
for only 7% of annual Medicare deaths—we find is very high, we find there are only a few indi- 14. S. K. Inouye et al., Med. Care 41, 70–83 (2003).
that only 12% of decedents have an annual pre- viduals for whom, ex ante, death is near certain. 15. P. Glare et al., BMJ 327, 195–198 (2003).
16. J. R. Lakin et al., JAMA Intern. Med. 176, 1863–1865 (2016).
dicted mortality of more than 80%. Qualitatively Moreover, a substantial component of the concen- 17. A. Elfiky et al., bioRxiv 204081 [Preprint]. 18 October 2017.
similar findings hold if we look at mortality in the tration of spending at the end of life is mechan- 18. A. Rajkomar et al., arXiv:1801.07860 [cs.CY] (11 May 2018).
month, rather than year, after hospital admission. ically driven by the fact that those who end up 19. M. D. Aldridge, A. S. Kelley, Am. J. Public Health 105,
Figure 4 shows the distribution of spending by dying are sicker, and spending, naturally, is high- 2411–2415 (2015).
20. A. S. Kelley et al., Health Serv. Res. 52, 113–131 (2017).
predicted mortality and illustrates another key er for sicker individuals. Of course, we do not— 21. A. S. Kelley, E. Bollens-Lund, K. E. Covinsky, J. S. Skinner,
finding: A large share of the concentration of and cannot—rule out individual cases where R. S. Morrison, J. Palliat. Med. 21, 44–54 (2018).
spending at the end of life can be explained by the treatment is performed on an individual for
AC KNOWLED GME NTS
concentration of spending on the sick. Decedents whom death is near certain. But our findings
We are grateful to P. Friedrich, D. Hernandez, A. Olssen, and
have higher predicted mortality than survivors indicate that such individuals are not a mean- A. Russell for superb research assistance and to J. Skinner,
and, as Fig. 4A shows, spending is increasing in ingful share of decedents. H. Williams, participants in the Stanford brown-bag lunch, and
predicted mortality. This simple observation goes These findings suggest that a focus on end-of- participants in the NBER Aging conference for helpful comments.
a long way toward explaining the concentration life spending is not, by itself, a useful way to Funding: L.E. and A.F. gratefully acknowledge support from the
National Institute on Aging (R01 AG032449), and Z.O.
of spending at the ex post end of life. identify wasteful spending. Instead, research- acknowledges support from the Office of the Director of the
Figure 4B shows the relationship between ers must focus on quality of care for very sick National Institutes of Health (DP5 OD012161), the National Institute
spending and predicted mortality separately for patients—identifying the impact of specific health on Aging (R56 AG055728), and the National Institute for Health
Care Management. Author contributions: All authors participated a standard application process described at www.resdac.org. Supplementary Text
in design of the study, analysis and interpretation of data, and Analysis code is available at http://web.stanford.edu/~leinav/ Figs. S1 to S9
the drafting and critical revision of the manuscript. Competing pubs/Science2018_Programs.zip. Tables S1 to S6
interests: L.E. is an adviser to Nuna Health, a data analytics References (22–39)
start-up company that specializes in health insurance claims. The SUPPLEMENTARY MATERIALS
authors declare no other competing interests. Data and materials www.sciencemag.org/content/360/6396/1462/suppl/DC1 15 November 2017; accepted 30 April 2018
availability: The data used in this paper can be accessed through Materials and Methods 10.1126/science.aar5045
change” over trials, we told them that the prev- in Study 7 we asked participants to play the role When blue dots became rare, purple dots began
alence of blue dots would “definitely decrease” of a reviewer on an Institutional Review Board. to look blue; when threatening faces became
over trials. The effect seen in Study 1 was repli- We showed participants a series of 240 proposals rare, neutral faces began to appear threatening;
cated (see fig. S2). In Study 3, we replicated the for scientific studies that (according to raters) and when unethical research proposals became
procedure for Study 1, except that this time a varied on a continuum from very ethical to very rare, ambiguous research proposals began to
third of the participants in the decreasing prev- unethical, and we asked participants to decide seem unethical. This happened even when the
alence condition were explicitly instructed to whether researchers should or should not be change in the prevalence of instances was abrupt,
“be consistent” and to not allow their concept allowed to conduct the study. After 96 trials, even when participants were explicitly told that
of blue to change over the course of the study, we decreased the prevalence of unethical pro- the prevalence of instances would change, and
and another third were given the same instruc- posals for participants in the decreasing preva- even when participants were instructed and paid
tion and were also offered a monetary incentive lence condition but not for participants in the to ignore these changes.
for following it. In all conditions, the effect seen stable prevalence condition. Figure 3 shows the These results may have sobering implications.
in Study 1 was replicated (see fig. S3). In Study 4, percentage of proposals that participants rejected Many organizations and institutions are dedicated
we replicated the procedure for Study 1, except on the initial 48 trials and on the final 48 trials. to identifying and reducing the prevalence of
that this time we decreased the prevalence of blue Participants in the stable prevalence condition social problems, from unethical research to un-
dots gradually for some participants (as we did (Fig. 3A) were just as likely to reject ethically warranted aggressions. But our studies suggest
in the previous studies) and abruptly for others. ambiguous proposals that appeared on a final that even well-meaning agents may sometimes
In all conditions, the effect seen in Study 1 was trial and on an initial trial, but participants in fail to recognize the success of their own efforts,
replicated (see fig. S4). Finally, in Study 5, we the decreasing prevalence condition (Fig. 3B) simply because they view each new instance in
replicated the procedure for Study 1, except that were more likely to reject ethically ambiguous the decreasingly problematic context that they
this time instead of decreasing the prevalence of proposals that appeared on a final trial than on themselves have brought about. Although mod-
RE FE RENCES AND N OT ES 17. M. Carandini, D. J. Heeger, Nat. Rev. Neurosci. 13, 51–62 (2011). J. Green, L. Harris, U. Heller, T. Hicks, S. Hoffman, E. Kemp,
1. S. O. Lilienfeld, Perspect. Psychol. Sci. 12, 138–169 (2017). 18. K. Louie, M. W. Khaw, P. W. Glimcher, Proc. Natl. Acad. Sci. U.S.A. B. Kollek, R. Lisner, Z. Lu, B. Martinez, T. Murphy, D. Park, D. Peng,
2. N. Haslam, Psychol. Inq. 27, 1–17 (2016). 110, 6139–6144 (2013). M. Powell, M. Sanders, C. Shaw, G. Stern, R. Stramp, A. Strauss,
3. A. Parducci, Psychol. Rev. 72, 407–418 (1965). 19. A. Parducci, Sci. Am. 219, 84–90 (1968). L. Symes, H. Tieh, A. Wang, I. Wang, C. Wu, X. Zeng, and Y. Zheng
4. H. Helson, Adaptation-Level Theory (Harper & Row, New York, 20. J. M. Wolfe, T. S. Horowitz, N. M. Kenner, Nature 435, 439–440 for research assistance. Statistical support was provided by
1964). (2005). data science specialists S. Worthington and I. Zahn at the Institute
5. C. W. G. Clifford et al., Vision Res. 47, 3125–3131 (2007). 21. M. C. Hout, S. C. Walenchok, S. D. Goldinger, J. M. Wolfe, for Quantitative Social Science, Harvard University. Funding: We
6. C. Summerfield, F. P. de Lange, Nat. Rev. Neurosci. 15, J. Exp. Psychol. Hum. Percept. Perform. 41, 977–994 (2015). acknowledge the support of National Science Foundation grant
745–756 (2014). 22. M. Roser, The short history of global living conditions and BCS-1423747 to T.D.W. and D.T.G. Author contributions: All
7. X. X. Wei, A. A. Stocker, Nat. Neurosci. 18, 1509–1517 (2015). authors contributed to the conceptual development of the work,
why it matters that we know it (2017); available at https://
8. N. Stewart, N. Chater, G. D. A. Brown, Cognit. Psychol. 53, 1–26 D.E.L. collected and analyzed the data, and D.E.L. and D.T.G.
ourworldindata.org/a-history-of-global-living-conditions-
(2006). designed the experiments and wrote the manuscript. Competing
in-5-charts.
9. A. Pepitone, M. DiNubile, J. Pers. Soc. Psychol. 33, 448–459 interests: None. Data and materials availability: The complete
(1976). 23. S. Pinker, Enlightenment Now: The Case for Reason, Science,
materials and data for all studies are available at Zenodo,
10. G. Rodríguez, S. Blanco, Anu. Psicol. Juridica 26, 107–113 (2016). Humanism, and Progress (Viking Press, New York, 2018).
https://zenodo.org/record/1219833 (25).
11. D. L. Chen, T. J. Moskowitz, K. Shue, Q. J. Econ. 131, 1181–1242 24. YouGov Survey (2015); available at https://d25d2506sfb94s.
(2016). cloudfront.net/cumulus_uploads/document/z2knhgzguv/
12. U. Simonsohn, F. Gino, Psychol. Sci. 24, 219–224 (2013). GB_Website.pdf. SUPPLEMENTARY MATERIALS
13. S. Bhargava, R. Fisman, Rev. Econ. Stat. 96, 444–457 25. D. Levari, Prevalence-induced concept change in human judgment.
www.sciencemag.org/content/360/6396/1465/suppl/DC1
(2014). Zenodo, Version v1.0.0 (2018); 10.5281/zenodo.1219833.
Materials and Methods
14. U. Simonsohn, Rev. Econ. Stat. 88, 1–9 (2006). Figs. S1 to S6
15. U. Simonsohn, G. Loewenstein, Econ. J. (Lond.) 116, 175–199 ACKN OWLED GMEN TS
References (26–31)
(2006). We thank T. Brady and M. Thornton for technical assistance and
16. L. Damisch, T. Mussweiler, H. Plessner, J. Exp. Psychol. Appl. M. Banker, S. Carroll, P. Chan, R. Chmielinksi, A. Collinsworth, 5 September 2017; accepted 30 April 2018
12, 166–178 (2006). K. da Silva, I. Droney, C. Fitzgerald, S. Ganley, M. Graether, 10.1126/science.aap8731
warming shown by fish debris oxygen varies as a function of temperature at the time
the phosphate was secreted and is more resistant
to diagenetic alteration than carbonate d18O
isotopes (El Kef, Tunisia) values (16, 17). Thus, fish debris from El Kef is an
excellent material with which to examine post-
K. G. MacLeod1*, P. C. Quinton2, J. Sepúlveda3, M. H. Negra4
K/Pg temperature history.
Fish debris d18O values show little variability
Greenhouse warming is a predicted consequence of the Chicxulub impact, but supporting in the uppermost 2 m of the Cretaceous [avg. =
data are sparse. This shortcoming compromises understanding of the impact’s effects, 20.7‰VSMOW (per mil) on the Vienna standard
and it has persisted due to an absence of sections that both contain suitable material for mean ocean water scale ± 0.33; n = 10], decrease
traditional carbonate- or organic-based paleothermometry and are complete and expanded
by ~1‰ at the boundary, and remain low over
enough to resolve changes on short time scales. We address the problem by analyzing the the first 3 m of the earliest Paleogene (avg. =
oxygen isotopic composition of fish debris, phosphatic microfossils that are relatively 19.5‰VSMOW ± 0.42, n = 21), and increase by ~1‰
resistant to diagenetic alteration, from the Global Stratotype Section and Point for the (avg. = 20.6‰VSMOW ± 0.63, n = 9) in the top 4 m
Cretaceous/Paleogene boundary at El Kef, Tunisia. We report an ~1 per mil decrease in oxygen
of section studied. The mean value for the older
isotopic values (~5°C warming) beginning at the boundary and spanning ~300 centimeters Paleocene samples is statistically distinct from the
means of the subjacent and superjacent groups,
T
distinct (Fig. 3 and supplementary materials). A
he Cretaceous/Paleogene (K/Pg) mass been found. In fact, postimpact cooling lasting ~1‰ decrease in d18O values across the boundary
extinction is unique among major extinc- thousands of years has been suggested from suggests ~5°C of warming, assuming seawater
tion events in that its ultimate cause (the paleontological data at the El Kef K/Pg Global d18O values remained approximately constant.
Chicxulub impact) perturbed Earth systems Stratotype Section and Point (GSSP) and nearby We treat the d18O results as an integrated sig-
on time scales shorter than those of current sections (11, 12). nal of the outer shelf water column. Fish are
anthropogenic changes (1). Study of the impact In this study, we report oxygen isotopic values mobile organisms, and different individuals could
aftermath provides a perspective on the response (d18O) of phosphate isolated from sand-sized record conditions from a range of environments,
of Earth systems to extremely rapid global per- (~0.1 to 2 mm) remains of fish teeth, scales, and depths, and seasons. Mitigating these concerns
turbations, making the K/Pg event an unusually bone (herein “fish debris”) from El Kef, Tunisia are estimated water depths (200 to 400 m) that
relevant natural experiment to compare to mod- (Fig. 1), indicating ~5°C warming beginning at restrict the depths at which the fish likely lived, and
ern climatic and environmental changes. Prom- the K/Pg boundary and lasting for ~100,000 years. the tropical paleolatitude of the site, so expected
inent among the proposed effects of the Chicxulub The El Kef section is dominated by hemipelagic seasonal variability is low. In addition, each mea-
impact are wide swings in temperature that in- marls (~40% CaCO3) and was deposited at ~20°N surement was based on separates of typically 50
clude a brief infrared pulse due to frictional heat- in waters 200 to 400 m deep on the Tethyan mar- to 75 individual microfossils, and the measurements
ing of ejecta re-entering the atmosphere, lasting gin of northern Africa. The boundary is placed at were grouped into three stratigraphic bins—2 m
for as little as 10 minutes (2, 3); an impact winter the base of a thin red layer that contains geo- of section representing deposition during the last
due to atmospheric loading of dust, soot, and chemical, mineralogical, and sedimentological in-
sulfate aerosols lasting for months to years (4–6); dicators of impact and coincides with the level of 1
Department of Geological Sciences, University of Missouri,
and greenhouse warming lasting ~100,000 years the K/Pg mass extinction. Above the red layer is a Columbia, MO, USA. 2Department of Geology, SUNY
or more due to increased atmospheric CO2 con- 50-cm-thick, carbonate-poor (<10%) claystone in Potsdam, Potsdam, NY, USA. 3Department of Geological
centrations from impact-volatilized carbonates which bulk carbonate d C values exhibit the ex-
13 Sciences and Institute of Arctic and Alpine Research,
University of Colorado Boulder, Boulder, CO, USA.
and wildfires (7, 8). pected K/Pg negative excursion. Carbonate content 4
Department of Geology, Faculty of Sciences of Tunis,
These shifts have been hypothesized for dec- increases back to pre-impact levels gradually over a University of Tunis El Manar, 2092 Manar II, Tunis, Tunisia.
ades and are invoked as examples that should ~2-m interval above the boundary claystone (13, 14) *Corresponding author. Email: macleodk@missouri.edu
inform thinking about the collateral consequences We analyzed samples from 2 m
of activities as different as nuclear war and an- below to 6.6 m above the bound-
thropogenic emissions (1, 6, 9), but testing pre- ary. The El Kef section is the
dictions is difficult. K/Pg changes occurred over GSSP for the K/Pg Boundary and,
short time scales relative to the typical resolu- thus, is formally accepted as being
tion of the rock record, and samples suitable for stratigraphically complete with
generating meaningful paleotemperature esti- well-resolved chronostratigraphic
mates are scarce. Thus, progress has been slow, control. Average sedimentation
and data are often ambiguous. rates are ~3.5 cm/1000 years from
Observational evidence cited in support of the 12 m below the boundary to 7 m
extreme heat pulse are high concentrations of above the boundary (Fig. 2). 3He
soot above the boundary and selective survivor- concentrations suggest that the
ship among terrestrial taxa (3). For impact winter, clay layer accumulated more rap-
an immediately post-K/Pg, 2° to 4°C cold pulse is idly than suggested biostratigraph-
suggested by some TEX86-based paleotemperature ically (15), but this detail does not
estimates (6, 10). Evidence for postimpact green- affect our analysis as observed
house warming using samples that meet current d18O changes span several dated Fig. 1. K/Pg paleogeography. K/Pg paleogeography (21)
criteria for quality of sample preservation has not events. showing El Kef and Chicxulub.
LO Parvularugo- 600
Averages ± 1 SD for each
0 globigerina bin are shown to the right
K/Pg boundary longiapertura 500 20.6 ± 0.63 with P values of t tests
marl
PALEOGENE
mentary materials). The
10 -6
400
-500 difference in means between
the earliest Paleogene
p = 4.6
300 100
(orange) and both the
underlying Cretaceous (blue)
p = 0.97
200 and overlying Paleogene
-1000 30
19.5 ± 0.42 (green) is significant at the
LO Plummerita 5s and 4s level, respectively.
100
10-8
hantkeninoides The difference in means
10
clay between the blue and green
p = 4.2
-1500 0 CRETACEOUS 0 bins is not significant. Ages
marl
Age (Ma) 100 20.7 ± 0.33 are based on Fig. 2 age model;
temperature calculated using
Fig. 2. Age-depth model of El Kef foraminiferal avg. ± 1 S.D.
(26), assuming seawater
events. Age-depth model of El Kef foraminiferal 200 50 error (1 S.D.)
d18O = –1‰VSMOW; error bar
events (22) dated relative to a K/Pg boundary
19 20 21 ± 0.3‰ (supplementary
(23, 24) using a boundary age of 66.043 million
δ18O (‰ V-SMOW) materials).
years (Ma) (25). LO, lowest occurrence; HO,
highest occurrence.
50,000 years of the Cretaceous, 3 m of section and geochemical records that contributed to the volatilization of carbonates, suggest more mod-
representing the first 100,000 years of the Pa- selection of El Kef as the K/Pg GSSP. est CO2 increase (≤50%). If the match between
leogene, and 4 m of section representing 100,000 Our results correspond remarkably well with stomatal estimates of CO2 change and El Kef
to 300,000 years after the boundary—of 10, 21, trends in d18O values predicted for greenhouse temperature increase are meaningful, widespread
and 9 samples, respectively. The average of mea- warming starting within decades after the impact, wildfires (4, 8) or other large CO2 sources seem
surements within each bin are compared to but they contradict conclusions based on paleon- required to augment CO2 from impact volatil-
assess temperature history (Fig. 3 and supple- tological data suggesting thousands of years of ization, and paleosol-based interpretations implic-
mentary materials). postimpact cooling (11, 12). We offer facies con- itly would be questioned. Alternatively, if CO2
Potential sampling biases and temporal vari- trol as an alternative explanation for the paleon- concentrations only increased modestly, either
ability should still be considered despite large tological results. Relatively high abundance of our interpretation of the El Kef results greatly
sample sizes. The high standard deviation among taxa with boreal affinities (the observation sup- overestimates global post-K/Pg warming or the
the samples from 3 to 7 m above the boundary porting the inference of cooler temperatures) climate sensitivity used in postimpact warming
suggest that uncertainty is greatest for this bin. are restricted to the 50-cm-thick boundary clay- models is too low. Discriminating among these
Additional analyses might reduce apparent vari- stone. The K/Pg event could have perturbed many alternatives would be a step forward in under-
ability, might reveal stratigraphic trends within variables (e.g., precipitation, nutrient loading, standing the post-K/Pg world, with implications
the interval, or might reinforce the impression pH, oxygen levels, carbonate production, carbon for temperature responses to modern climatic
that high variability among samples is a feature cycling, niche occupancy, and food web structure) perturbations (especially increasing atmospheric
of this part of the section. In contrast, the low that would plausibly affect both lithology and CO2 levels) imposed on very short time scales.
standard deviation of the d18O measurements microfossil abundance without a priori temper-
REFERENCES AND NOTES
(close to analytical precision of 0.3‰) in the lower ature implications. The lack of a lithological shift
1. P. Schulte et al., Science 327, 1214–1218 (2010).
two bins argues that these bins have quite stable at the end of the isotopic excursion, in contrast, 2. J. Morgan, N. Artemieva, T. Goldin, J. Geophys. Res. Biogeosci.
isotopic signatures that our sampling scheme renders ad hoc explanations that would explain 118, 1508–1520 (2013).
adequately captured. away apparent warming as an artifact of a strat- 3. D. S. Robertson, W. M. Lewis, P. M. Sheehan, O. B. Toon,
J. Geophys. Res. Biogeosci. 118, 329–336 (2013).
Evidence for increased temperatures begins igraphically restricted interval of altered ocean 4. W. S. Wolbach, R. S. Lewis, E. Anders, Science 230, 167–170
at the K/Pg boundary. We see no evidence of circulation, changed moisture balance, and/or (1985).
an impact winter, but finding evidence for this diagenesis (supplementary materials). 5. K. O. Pope, K. H. Baines, A. C. Ocampo, B. A. Ivanov, Earth
≤decade-long interval was unlikely. Collected Our results suggest ~5°C of postimpact warm- Planet. Sci. Lett. 128, 719–725 (1994).
6. J. Vellekoop et al., Geology 44, 619–622 (2016).
samples spanned 2.5 to 10 cm of section (i.e., on ing lasting ~100,000 years (Fig. 3). This magni- 7. D. J. Beerling, B. H. Lomax, D. L. Royer, G. R. Upchurch Jr.,
average, each represents ≥1000 years of deposi- tude and duration of warming matches well with L. R. Kump, Proc. Natl. Acad. Sci. U.S.A. 99, 7836–7840
tion), and bioturbation and physical reworking modeled climate response to post-K/Pg increase (2002).
8. D. A. Kring, Palaeogeogr. Palaeoclimatol. Palaeoecol. 255, 4–21
(14) would introduce additional time averaging. in atmospheric CO2 concentrations to ≥2300 (2007).
Reworking could also have smeared the evidence parts per million (ppm) from background levels 9. D. A. Kring, GSA Today 10, 1–7 (2000).
for the initiation of greenhouse warming. How- of 350 to 500 ppm, as estimated from stomatal 10. J. Vellekoop et al., Proc. Natl. Acad. Sci. U.S.A. 111, 7537–7541
ever, the sharp K/Pg d18O decrease found (Fig. 3) densities on fossil leaves (7). In contrast, paleosol (2014).
11. S. Galeotti, H. Brinkhuis, M. Huber, Geology 32, 529–532 (2004).
indicates reworking was not that severe, a finding paleobarometry (18), as well as modeling (19) 12. J. Vellekoop et al., Palaeogeogr. Palaeoclimatol. Palaeoecol.
consistent with the well-resolved biostratigraphic and experimental results (20) of impact-induced 426, 216–228 (2015).
13. G. Keller, M. Lindinger, Palaeogeogr. Palaeoclimatol. Palaeoecol. 23. S. Abramovich, G. Keller, Palaeogeogr. Palaeoclimatol. analyzed samples and participated in interpretation, writing, and
73, 243–265 (1989). Palaeoecol. 178, 145–164 (2002). editing of the manuscript; and J.S. and M.H.N. collected samples
14. E. Molina et al., Episodes 2006, 263–273 (2006). 24. B. S. Wade, P. N. Pearson, W. A. Berggren, H. Pälike, and participated in interpretation, writing, and editing of the
15. S. Mukhopadhyay, K. A. Farley, A. Montanari, Science 291, Earth Sci. Rev. 104, 111–142 (2011). manuscript. Competing interests: None declared. Data and
1952–1955 (2001). 25. P. R. Renne et al., Science 339, 684–687 (2013). materials availability: All data are presented in the
16. B. Wenzel, C. Lécuyer, M. M. Joachimski, Geochim. supplementary materials.
26. C. Lécuyer, R. Amiot, A. Touzeau, J. Trotter, Chem. Geol. 347,
Cosmochim. Acta 64, 1859–1872 (2000).
217–226 (2013).
17. D. Bassett, K. G. Macleod, J. F. Miller, R. L. Ethington, Palaios SUPPLEMENTARY MATERIALS
22, 98–103 (2007). ACKN OWLED GMEN TS www.sciencemag.org/content/360/6396/1467/suppl/DC1
18. C. Huang, G. J. Retallack, C. Wang, Q. Huang, Materials and Methods
We thank I. Arenillas, J. Arz, M. Katz, and J. Smit for useful
Palaeogeogr. Palaeoclimatol. Palaeoecol. 385, 95–105 (2013). Supplementary Text
discussion; two reviewers for editorial suggestions; and
19. E. Pierazzo, D. A. Kring, H. J. Melosh, J. Geophys. Res. 103, S. Haynes, G. O’Neil, L. Alegret, S. Bey, M. Baroumi, M. Giron, Fig. S1
28607–28625 (1998). R. Summons, H. Theyri, and J. Wendler for laboratory and/or Tables S1 and S2
20. M. S. Bell, Meteorit. Planet. Sci. 51, 619–646 (2016). field help. Funding: NSF-EAR 1323444 (K.G.M.), MU References (27–30)
21. C. R. Scotese, in Atlas of Late Cretaceous Paleogeographic Undergraduate Research Fund (K.G.M.), and the MIT MISTI-
Maps, PALEOMAP Atlas for ArcGIS (PALEOMAP Project, Spain Program (J.S.). Author contributions: K.G.M. conceived 1 September 2017; accepted 10 May 2018
Evanstan, IL, 2014), vol. 2. the study, processed and analyzed samples, interpreted results, Published online 24 May 2018
22. I. Arenillas, J. A. Arz, E. Molina, GFF 124, 121–126 (2002). and wrote and edited the manuscript; P.C.Q. processed and 10.1126/science.aap8525
T
Development, Institute of Human Genetics, CNRS-University
of Montpellier UMR9002, Montpellier, France.
he regulation of genes with important roles tive feed-forward loops. This complexity is likely *Present address: Center for Genetic Medicine, Northwestern
in embryonic development can be complex, to be amplified when the gene has functions in University, Chicago, IL 60611, USA. †Present address: Department
involving multiple, often redundant en- more than one tissue, given that the regulatory of Oncological Sciences, Huntsman Cancer Institute, University of
hancers, silencers, and insulators (1, 2). elements required for each are often interspersed Utah, Salt Lake City, UT 84112, USA. ‡Present address: Institute
of Ophthalmology, University College London, 11-34 Bath Street,
The genes may have a poised epigenetic and necessitate dynamic alterations in chromatin London EC1V 9EL, UK. §These authors contributed equally to this
state prior to their expression, and their activa- conformation (1, 2). The developing gonads con- work. ||Deceased.
tion or repression may involve positive or nega- stitute an interesting system in which to explore ¶Corresponding author. Email: robin.lovell-badge@crick.ac.uk
A XY SR REV SEX
Sox99
1 4 5 7 8 9 11 13 14 16 18 19 22 29 30 TES 32
kb
1500 1000 900 800 700 600 500 400 300 200 100 0
Tg 1 E13.5 Tg 2
10 kb mm9 C XY XY
B Enh13
Enh13 557 bp XY SR
E15.5 Sertoli DNaseI peaks
E13.5
E13.5 Sertoli DNaseI
Fig. 1. Enh13 is a testis-positive enhancer of Sox9 located within the Sertoli cells (blue) and granulosa cells (purple), as well as E10.5 sorted
XY SR region. (A) A schematic representation of the gene desert somatic cells, at the Enh13 genomic region are presented. Peaks
upstream of the mouse Sox9 gene and the locations of the putative correspond to nucleosome-depleted regions and are marked by a black
enhancers identified by ATAC-seq and DNaseI-seq that were screened horizontal line if they are significantly enriched compared to flanking
in vivo with transgenic reporter mice. Enhancers that did not drive gonad regions, as determined by model-based analysis of ChIP-seq, and present
expression of LacZ are shown in gray. Enhancers that drove testis-specific in at least two biological replicates. The gray box overlaying the peak
and ovary-specific LacZ expression are shown in blue and pink, respec- indicates the cloned fragment. Green areas represent sequence conser-
tively. The mouse regions that show conserved synteny with the human vation among mice, humans (Hu), rhesus monkeys (Rh), cows (Co), and
XY SR and REV SEX are depicted in green and purple boxes, respectively. chickens (Ch) (sequence conservation tracks were obtained from the
(B) Enh13 (gray box) is located at the 5′ side of the 25.7-kb mouse University of California–Santa Cruz). (C) b-Gal staining (blue) of E13.5
equivalent XY SR locus (heavy black line). Data from DNaseI-seq (black) testes and ovaries from two representative independent stable Enh13
on E15.5 and E13.5 XY sorted Sertoli cells and ATAC-seq on E13.5 sorted transgenic (Tg) lines. Scale bars, 100 mm.
loss-of-function studies in mice and humans 1 (Enh1) and Enh14 (Fig. 1A and fig. S1). Chromatin fig. S6, B and C). ATAC-seq, DNaseI-seq, and
demonstrate that Sox9 plays a key role in testis immunoprecipitation sequencing (ChIP-seq) was H3K27ac ChIP-seq data suggest that Enh32 is a
determination (8–13). Notably, humans hetero- also performed for H3K27ac, a histone modifica- Sertoli cell enhancer, although weak peaks were
zygous for null mutations develop campomelic tion that marks active enhancers (fig. S1). seen in the granulosa cell samples (figs. S1
dysplasia (CD) [Online Mendelian Inheritance All 16 putative enhancers were cloned upstream and S6A).
in Man (OMIM) entry 114290] (11), a severe syn- of an Hsp68 minimal promoter and the reporter The remaining enhancer, Enh13 (557 bp long,
drome where 70% of XY patients show female gene LacZ and used to generate transgenic mice 565 kb 5′), is highly conserved among mammals
development (12, 13). (2, 19) (table S1). For initial screens, we performed and is located toward the distal 5′ end of a 25.7-kb
Sox9 functions in many embryonic and adult transient analyses at E13.5. Twelve enhancers region in mice that shows conserved synteny
cell types (14), and genetic and molecular evi- failed to produce any gonadal b-galactosidase with a 32.5-kb region upstream of human SOX9
dence suggests that its regulatory region is spread (b-Gal) activity, although many showed staining termed XY SR, the deletion of which is asso-
over a gene desert of at least 2 Mb 5′ to the in other tissues in which Sox9 is normally ex- ciated with sex reversal (20) (Fig. 1, A and B,
coding sequence (15). The only enhancer known pressed, such as chondrocytes, brain, and spinal and fig. S1). Enh13 shows the strongest Sertoli
to be relevant for expression in Sertoli cells was cord (fig. S3). The remaining four showed gonad cell–specific peak within this region in both the
TES, a 3.2-kb element mapping 13 kb 5′ from expression, and these constructs were reinjected DNaseI-seq and ATAC-seq data. H3K27ac ChIP-
the transcriptional start site, and its 1.4-kb core, to generate stable lines in order to better study seq data mark Enh13 as active in both Sertoli
TESCO (16). Targeted deletion of TES or TESCO their activity in both males and females during and granulosa cells, which may support the ob-
reduced Sox9 expression levels in the early and development. Enh8 [672 base pairs (bp) long, servation that some transgenic lines also exhibit
postnatal mouse testis to about 45% of normal 838 kb 5′] conferred robust b-Gal activity in the b-Gal activity in the ovary (Fig. 1C and figs. S1
but did not result in XY female development (17). ovary, whereas it was barely present in the testis and S7A). b-Gal expression is clearly within Ser-
We therefore used several unbiased approaches at E13.5 (Fig. 1A and fig. S4B). This may be due to toli and granulosa cells (fig. S7C). ATAC-seq data
to systematically screen for additional gonad Enh8 being taken out of its original genomic from E10.5 genital ridges show that Enh13 is not
Chr11 10 kb XY XY XY XX
A C
Sox9 Enh13 +/+ Sox9 Enh13 -/+ Sox9 Enh13 -/- Sox9 Enh13 +/+
Allele 1 Enh13 TES Sox9
557 bp 3194 bp
565 kb
XY
FOXL2
B XY XY XX
E13.5
Sox9 Enh13 +/+ Sox9 Enh13 -/+ Sox9 Enh13 -/- Sox9 Enh13 +/+
BF
Merge+DAPI
E13.5
H&E
Fig. 2. Deletion of Enh13 leads to complete XY male-to-female sex Enh13+/−, and Enh13−/− and XX Enh13+/+ gonads. (C) Immunostaining of
reversal. (A) A schematic representation of the locations of Enh13 and E13.5 XY wild-type, Enh13+/−, and Enh13−/− and XX wild-type gonads.
TES upstream of Sox9. Blue and purple arrows represent the external and Gonads were stained for Sertoli marker SOX9 (green), granulosa marker
internal single guide RNAs, respectively, used to delete Enh13. Black FOXL2 (red), and 4′,6-diamidino-2-phenylindole (DAPI) (blue). Sex-
arrows represent the PCR primers used to genotype embryos and mice reversed gonads are indistinguishable from wild-type XX gonads,
with Enh13 deletion. Chr11, chromosome 11. (B) Bright-field (BF) pictures whereas the heterozygous deletion does not appear to alter testis
and hematoxylin and eosin (H&E)–stained sections of E13.5 XY Enh13+/+, morphogenesis. Scale bars in (B) and (C), 100 mm.
Nevertheless, whereas XY Enh13+/− embryos still begins (5). We therefore analyzed Sox9, Sry, Sf1, with SRY and later with SOX9 to regulate Sox9
undergo normal testis development, XY Enh13−/− and Foxl2 mRNAs at E11.5, during the brief expression levels and also many of their down-
embryos produce ovaries indistinguishable from period when gonadal sex is being determined. stream target genes (16). We found no significant
those of XX wild-type embryos, with no signs of Real-time quantitative polymerase chain reac- changes in levels of Sf1 mRNA with any of the
testis cords or a coelomic vessel (Fig. 2, B and C, tion (RT-qPCR) revealed that XY Enh13+/− and enhancer deletions at E11.5 (Fig. 3C). Using Foxl2
and fig. S10). Immunofluorescence analysis of Enh13−/− genital ridges expressed 58 and 21% of as an early marker of granulosa cell differentia-
E13.5 and 6-week-old XY Enh13+/− and Enh13−/− the wild-type levels of Sox9 mRNA, whereas XY tion (24), we found that mRNA levels in XX wild-
gonads for SOX9 and FOXL2 showed that the TES+/− and TES−/− genital ridges showed 55 and type gonads at E11.5 were 3.6 times as high as
former are still testes whereas the latter are fully 50%, respectively (Fig. 3A). Control XX genital those in XY gonads (Fig. 3D). Compared with the
sex-reversed ovaries (Fig. 2C and fig. S10D). Similar ridges contained 18% of the Sox9 mRNA levels latter, Enh13+/− and Enh13−/− XY gonads showed
analysis of XX gonads with Enh13 deletion did found in XY genital ridges (Fig. 3A). Therefore, two- and threefold increases, respectively, with
not show any obvious phenotype (fig. S11). E11.5 XY Enh13−/− gonads express Sox9 at levels the homozygotes having mRNA levels very close
The Enh13 deletion was generated in a C57BL/ close to those of XX gonads at the same stage, to XX control levels. Therefore, Enh13−/− XY go-
6J genetic background, which is sensitized toward explaining the observed complete sex reversal. nads reveal an early commitment to the ovarian
XY female sex reversal (21). To test the strength Deleting one or two copies of TES had relatively pathway.
of the deleted allele, we therefore backcrossed little effect at E11.5, especially compared with the There was a 30 to 50% decrease in Sox9 mRNA
the deletion into a mixed C57BL/6J × CBA back- effect at E13.5, in contrast to the results with Enh13 levels in E11.5 XX Enh13−/− gonads compared to
ground. As before, XY heterozygotes presented deletions at E11.5 (Fig. 3A) (17). This again supports the wild type, as reflected by reduced immuno-
as normal fertile males whereas homozygotes the conclusion that Enh13 plays a more substantial fluorescence for SOX9 protein (fig. S11). These
showed full male-to-female sex reversal (fig. S12). role than TES during early gonadal development. data indicate that Enh13 also plays a role in the
We could detect no difference in gonadal pheno- Sry expression is normally down-regulated as very early expression of Sox9 in the XX gonad,
types between Enh13−/−:TES+/+ and Enh13−/−: SOX9 levels increase, but it can persist if testis consistent both with the small peak seen with
*
** **
1.00 **
2.0
bryos by using a specific antibody against the
Sox9 Sry MYC tag (22). SRY-MYC–positive gonads had an
0.75 1.5 11-fold enrichment versus SRY-MYC–negative
gonads with primers spanning the SOX9 consen-
0.50 1.0 sus site and a sixfold enrichment with primers
0.25 0.5
spanning the SRY site, whereas primers against
the strongest SRY binding site in TESCO (22)
0.00 0.0 showed fivefold enrichment (Fig. 4, A and B).
XY XY
XY XY XY XY
XY XY
XY XX
XX XY
XY XY XY XY XY XY XX This reveals the strong binding of SRY to Enh13
WildEnh13
WT Enh13 Enh13 TES
TES TES Wild Wild Enh13
Enh13 Enh13 TES TES
TES Wild
type +/- Enh13
-/- +/- TES
-/- WT
type WT
type +/- Enh13
-/- TES
+/- -/- WT
type at E11.5, with a preference for the SOX9 consen-
n=6 n=1
+/-1 n=3
-/- n=5
+/- n=7
-/- n=5 n=6 n=1
+/-1 n=3
-/- n=5
+/- n=7
-/- n=5
n=6 n=11 n=3 n=5 n=7 n=5 n=6 n=11 n=3 n=5 n=7 n=5 sus site, possibly due to the adjacent SF1 bind-
ing site. Preferential binding of SRY to Enh13
over TESCO at E11.5 supports the hypothesis
C 1.5 D 4 ***
that the former is more critical because it ini-
mRNA Relative to Hprt
SOX9
(Sry Myc+ / Sry Myc-)
ChIP
% of unbound
Fold enrichment
0-29
Input
Sox9Enh13
60 060 kb 60 080 kb
Chr 19
(Bov)
Enh13-1 Enh13-2 TESCO-2 Contr Enh13-1 TESCO-1 Contr 59,492-59,494 kb Sox9
Fig. 4. Enh13 is bound by SRY and SOX9 in vivo. (A) A schematic immunoprecipitation with anti-SOX9 antibody. The primers used span
representation of the locations of the primers used for ChIP experiments the putative SOX9 binding site in Enh13 and TESCO [around the SOX9 R1 site;
in Enh13. BS, binding site. (B) ChIP-qPCR assay of E11.5 mouse genital see (16)] and a negative control region on chr11. Data are the means ± SD
ridges after immunoprecipitation with anti-cMYC antibody. Data are (n = 3). ****P < 0.0005; ns, not significant (Student t test). Unr ab, unrelated
a similar role, heterozygosity for Enh13 deletions have at most subtle if not undetectable effects 9. F. Barrionuevo et al., Biol. Reprod. 74, 195–201 (2006).
in humans should mimic heterozygosity for null (30, 31). It is therefore remarkable to see that 10. V. P. Vidal, M. C. Chaboissier, D. G. de Rooij, A. Schedl,
Nat. Genet. 28, 216–217 (2001).
mutations in SOX9, with XY female sex reversal deleting Enh13 alone phenocopies the loss of 11. C. S. Houston et al., Am. J. Med. Genet. 15, 3–28 (1983).
occurring in about 70% of cases but perhaps Sox9 itself within the supporting cell lineage 12. J. W. Foster et al., Nature 372, 525–530 (1994).
without other CD phenotypes. (9, 32). Substantial evidence points to the time- 13. T. Wagner et al., Cell 79, 1111–1120 (1994).
It is clear from our data and those of others dependent action of SRY on Sox9. If Sox9 fails 14. A. Jo et al., Genes Dis. 1, 149–161 (2014).
15. A. Symon, V. Harley, Int. J. Biochem. Cell Biol. 87, 18–22
that the upstream regulatory region of Sox9 is to reach a critical threshold within a few hours, (2017).
very complex. Our screens with gonadal cells then ovary-determining and/or antitestis factors, 16. R. Sekido, R. Lovell-Badge, Nature 453, 930–934 (2008).
revealed 33 potential enhancers distributed such as Wnt signaling, accumulate to a sufficient 17. N. Gonen, A. Quinn, H. C. O’Neill, P. Koopman, R. Lovell-Badge,
over 1.5 Mb. Transgenic assays used to test the level to repress Sox9 and make it refractory to PLOS Genet. 13, e1006520 (2017).
18. D. M. Maatouk et al., Development 144, 720–730 (2017).
most promising 16 enhancers revealed 4 that male-promoting factors, including SRY, even 19. E. Z. Kvon, Genomics 106, 185–192 (2015).
produced expression in the gonads, whereas the though expression of the latter persists in XY 20. G. J. Kim et al., J. Med. Genet. 52, 240–247 (2015).
majority did not. This “hit rate” of 25% agrees gonads when Sertoli cells fail to differentiate 21. S. M. Correa et al., PLOS Genet. 8, e1002569 (2012).
with the results of some other studies (19) and (33). We suggest that Enh13 is an early-acting 22. R. Sekido, I. Bar, V. Narváez, G. Penny, R. Lovell-Badge,
Dev. Biol. 274, 271–279 (2004).
merits caution in interpreting data based solely enhancer, such that without it Sox9 transcription 23. C. H. Lee, T. Taketo, Dev. Biol. 165, 442–452 (1994).
on the accessibility of chromatin and histone fails to increase to a level where the other en- 24. D. Schmidt et al., Development 131, 933–942 (2004).
marks. However, several of these putative en- hancers can act before the gene is silenced. It is 25. M. Rahmoun et al., Nucleic Acids Res. 45, 7191–7211 (2017).
hancers are bona fide enhancers in other loca- only later that TES, and perhaps other enhanc- 26. T. Ohnesorg, B. Croft, J. Tan, A. H. Sinclair, Sex Dev. 10, 59–65
tions; for example, Enh29, mapping about 70 kb ers such as Enh14 and Enh32, begins to act in a (2016).
27. T. J. Mead et al., Nucleic Acids Res. 41, 4459–4469
5′, is equivalent to SOM, an enhancer active in more redundant fashion, although it is con- (2013).
many tissues, excluding the gonads (27). Others ceivable that each enhancer has a major role 28. M. Lagha, J. P. Bothma, M. Levine, Trends Genet. 28, 409–416
may have distinct roles; Enh11 contains a puta- to play during distinct phases of Sertoli cell (2012).
tive CTCF binding site. In addition, several puta- development from the fetal to the adult testis. 29. A. Visel, E. M. Rubin, L. A. Pennacchio, Nature 461, 199–205
(2009).
tive enhancers, notably Enh4, -5, -8, and -9, appear It will be of interest to determine how Enh13 ac-
30. D. E. Dickel et al., Cell 172, 491–499.e15 (2018).
to be open in both granulosa and Sertoli cells, tivity cascades into the recruitment of the other 31. M. Osterwalder et al., Nature 554, 239–243 (2018).
with Enh8 even more so in the former. These enhancers. 32. R. Lopes, G. Korkmaz, R. Agami, Nat. Rev. Mol. Cell Biol. 17,
may contribute to the low level of Sox9 expres- 597–604 (2016).
sion seen in supporting cell precursors, but they RE FERENCES AND NOTES 33. R. Hiramatsu et al., Development 136, 129–138 (2009).
may also represent sequences required to repress 1. W. de Laat, D. Duboule, Nature 502, 499–506 (2013).
2. F. Spitz, Semin. Cell Dev. Biol. 57, 57–67 (2016). AC KNOWLED GME NTS
Sox9, which might not be detected by transgenic 3. S. A. Garcia-Moreno, M. P. Plebanek, B. Capel, Mol. Cell. We dedicate this paper to the memory of Danielle M. Maatouk,
reporter assays. Endocrinol. 468, 19–30 (2018). a much-missed colleague. We are grateful to the Biological
The notion of redundancy or “shadow en- 4. B. Capel, Nat. Rev. Genet. 18, 675–689 (2017). Research Facility, Genetic Modification Service, Advanced
hancers” within a regulatory region is well es- 5. S. Jakob, R. Lovell-Badge, FEBS J. 278, 1002–1009 (2011). Sequencing Facility, and Experimental Histopathology Facility of
6. D. Wilhelm, S. Palmer, P. Koopman, Physiol. Rev. 87, 1–28 (2007). the Francis Crick Institute and the Flow Cytometry Core at
tablished (28, 29), and recent data suggest that 7. P. Koopman, J. Gubbay, N. Vivian, P. Goodfellow, Northwestern University for technical assistance. We thank
deletion of single, even “ultraconserved” enhancers R. Lovell-Badge, Nature 351, 117–121 (1991). members of our labs for advice, support, and helpful comments.
from developmentally important genes can 8. M. C. Chaboissier et al., Development 131, 1891–1901 (2004). Funding: This work was supported by the Francis Crick
Institute, which receives its core funding from Cancer Research the TES mutant (TESMS)–cyan fluorescent protein (CFP) SUPPLEMENTARY MATERIALS
UK (FC001107), the U.K. Medical Research Council (FC001107), transgenic mice. F.P. performed the mouse and bovine SOX9 www.sciencemag.org/content/360/6396/1469/suppl/DC1
and the Wellcome Trust (FC001107), and by the U.K. Medical and SRY ChIP. N.G. performed the rest of the experiments. Materials and Methods
Research Council (U117512772). F.P. was funded by the Agence N.G. and R.L.-B. analyzed and interpreted the results and wrote Figs. S1 to S13
Nationale pour la Recherche (ANR blanc TestisDev). D.M.M. the manuscript. All authors reviewed and added input to the Tables S1 to S4
was funded by the Northwestern University School of Medicine. manuscript. Competing interests: The authors have no References (34–44)
Author contributions: N.G. and R.L.-B. designed the study. competing interests. Data and materials availability: All data
C.R.F., S.A.G.-M., I.M.S., and D.M.M. performed ATAC-seq and are available in the main text or the supplementary materials.
H3K27ac ChIP-seq and cloned the enhancer-LacZ plasmids. ATAC-seq and H3K27Ac ChIP-seq data have been deposited in 8 January 2018; accepted 31 May 2018
S.C.S. helped with analyzing the reporter mice. S.W. performed the Gene Expression Omnibus under accession number Published online 14 June 2018
cytoplasmic and pronuclear zygote injections. R.S. provided GSE99320. 10.1126/science.aas9408
If
they tell you about the ˇeld? Would they Radmanís collaborator, Jean-NoÎl Thorel, a pharmacist and founder of the skin-
point to its interdisciplinary approach, care group NAOS, has been talking about an ecobiologyóor holisticóapproach for
or to a focus on the health of the whole several years. His philosophy is centered on human well-being, and on an ethical
organism, or perhaps to its ultimate approach to business. ìI see so many new products that have absolutely no use,î he
goal, advancing human well-being? says. ìThe skincare industry should strive to innovate to be useful.î
Such were the questionsófrom the He wants ecobiologists ìto create new products that will allow us to help the skin
nitty-gritty of research to the philosophicalóthat to adapt to our environment,î he insists.
consumed scientists gathered at the ecobiology Whether a science or not, could ecobiology be a useful approach to tackling
summit, held April 18ñ19 in Split, Croatia. They some of the big issues in human health, beginning with aging? Radman comments
were meeting to thrash out a deˇnition for ecobi- that much research has been focused on diseases of aging as opposed to the aging
ology and to assess the impact it could have on process itself. Too much of the science is concerned with treating the consequen-
human health. Ecobiology, say its proponents, is ces rather than trying to understand the causes, he asserts.
an approach that investigates the interconnec- The body has many mechanisms to repair and maintain cells, and yet aging still
tions and communication between cells, and be- occurs. The key to unlocking treatments to extend healthy life may come from exten-
tween cells and their external environment. sive research on mechanisms that provide protection against protein damage, which
For Errol Friedberg, emeritus professor of are being studied in ˇelds such as cancer research.
pathology at University of Texas Southwestern The conference heard from Peter Karran, a former principal scientist at the Francis
Medical Center in Dallas, that raises the ques- Crick Institute in the United Kingdom, whose focus has been on skin cancer. Karran
tion, ìWhy donít we just stick with integrative biol- points out that because the skin provides a barrier between us and our environ-
ogy, with what is axiomatic of all biology?î ment, it is ìexposed to threats not seen by other cells in the bodyóthe most obvious
Molecular biologist Miroslav Radman, one of example is sunlight. This is in addition to the internal threats [of] water and highly
the conference organizers, sees ecobiology as damaging oxygen that affect all cells.î Humans have sophisticated mechanisms that
ìmore a vision, a movement. Do we even need recognize and repair DNA in order to prevent mutations. But repair is imperfect and
a sharp deˇnition?î he asks. ìResults and new mutations accumulate with age, occasionally leading to skin cancer.
knowledge will eventually arrive independent of Proteins, which carry out all cellular processes, are also vulnerable to damage
the buzzwords used.î from oxygen. Earlier research by Radman and others suggests that some proteins,
including those expressed from mutated genes, are particularly vulnerable to mis-
folding and oxidative damage. Karranís work with cells from patients displaying ac-
celerated photoaging builds on this ˇnding. He posits that mutations accumulated
over time, which fuel the inevitably growing burden of oxidized proteins, contribute
ALL PHOTOS: ALEKSANDAR TOPALOVIC
Meeting of minds
The complexity of the skin ecosystem and
its interaction with the skin microbiome is what
drives Thorel to argue for a multidisciplinary ap-
proach to skin care. His quest to have science
inform product developmentóa philosophy
shared with Radmanóhas prompted him to
invest in Radmanís Mediterranean Institute for
Life Sciences (MedILS), located in Split, Croa-
tia. Set up more than 10 years ago, the institute
was the ˇrst in continental Europe dedicated to
the biology of aging and age-related diseases,
says Radman. Now the two hope to attract young
scientists to this very practical challenge, and
Thorel plans to launch a scholarship in ecobiol-
ogy. MedILS currently has around 20 scientists
This approach was evident when Thorel launched engaged in understanding the role of protein
his ˇrst productófor sun protectionóin the late 1970s. stability in longevity, and how the biological ìro-
While most cosmetics companies were working on bustnessî of bacterial species that survive ex-
products with high sun protection factors (SPFs) or oils treme environments could be applied to improve
to help tanning, Thorel chose a molecule that would trig- human health.
ger the skin to produce the pigment melanin. Melanin Thorelís investment funds both pure and ap-
gives some protection against damage from sunlight. plied research. FranÁois-Xavier Pellay is the re-
searcher charged with identifying commercial
Under the skin: Reexamining ingredients and prospects for new molecules. Pellayówith his
investigating microbiomes background in biology, chemistry, biochemistry,
AurÈlie Guyoux NAOSís three brandsóBioderma, Institut Esthe- and bioinformaticsóembodies the multidisci-
derm, and Etat Puróare billed as ecobiology in the plinary approach Radman and Thorel advocate.
service of dermatology, aesthetics, and personalized ìOur goal is to turn brilliant ideas into some-
skincare, respectively. thing that can be applied to the skin ... we know
ìWe consider the skin as an ecosystem because it the targets are proteins, so [our approach is]
is composed of different kinds of cells, each with their ëhow do we protect these proteins?íî explains
own speciˇc environment. They continually interact with Pellay.
each other, and with our environment,î explains AurÈlie One strand of research is particularly prom-
Guyoux, director of R&D for NAOS. ising. Pellay has been exploring the mecha-
With that in mind, NAOS carefully selected just 600 nisms that a species of cold- and ultraviolet (UV)-
of the more than 30,000 ingredients frequently used in resistant bacteria uses to protect its proteins
modern cosmetics. ìWe looked at the structure of these from oxidation.
ingredients, [and asked:] are they identical to mole- His team tested very speciˇc molecules be-
Eric Perrier
cules already found in the skin? If the skin recognizes longing to the carotenoid family, isolated from the
them, they will be better tolerated and help to restore bacteria. Carotenoids, which originally evolved
skin health,î asserts Guyoux. Moreover, with skin sen- in plants, are very good at capturing reactive
sitivity on the rise, NAOS wants to reduce skin ìpollut- oxygen species. Pellayís unpublished research
ants,î she explains. suggests that the isolated carotenoid molecules
NAOS has plans for some relaunches of its products, he tested bind to proteins, functioning as anti-
especially those designed for sensitive skin, in light of oxidants while also protecting the proteome of
new discoveries about interaction and communication human skin cells against stress such as UV light
between skin cells as well as the interaction of skin cells and pollutants.
with the skin microbiome. Eric Perrier, Innovation direc- It may be many years before this avenue of
tor for NAOS, points out that ìcell-to-cell interactions research delivers advances in skincare or even
are complicated enough, but we now have another helps push back the ravages of aging. Will eco-
layer of communication between the skin cells and mi- biology then be recognized as a science? For
crobiome as well as between the bacteria in the micro- Radman and Thorel, the answer is perhaps not
biome themselves.î Changes in the environment (such that important, as the ˇeld already provides a
as pollution) cause alterations in the microbiome, as will methodological approach, both to the acquisi-
drugs used to treat skin conditions, explains dermatolo- tion of knowledge through the cross-fertilization
gist Brigitte DrÈno. She heads the Department of Der- of scientiˇc research, and to commercial appli-
MedILS and NAOS matology at the University Hospital of Nantes. DrÈno cations aimed at advancing human well-being.
recently collaborated anticipates that the concept of stratiˇed medicine will
to hold the ecobiology inform the development of personalized probiotics that
summit. will help to maintain the health and function of the skin
microbiome.
LIFE SCIENCE TECHNOLOGIES Produced by the Science/AAAS Custom Publishing Office
new pro d uc ts
CD28, which makes it a promising target for antitumor therapy.
data generated from the same sample. The isolated gDNA, total RNA, ACROBiosystems
and total denatured proteins are suitable for genomic and proteomic For info: 800-810-0816
applications such as PCR, restriction-enzyme digestion, sequencing, www.acrobiosystems.com
array CGH, reverse transcription PCR, gene expression microarray,
SDS-PAGE, Western blotting, 2D DIGE, and LC/MS. IVD Raw Materials and Customized Services
GE Healthcare Life Sciences Cloud-Clone is a supplier of raw materials to the in vitro diagnostics (IVD)
For info: 800-526-3593 industry and to research laboratories globally, and furnishes various
www.gelifesciences.com human proteins and antibodies covering a wide range of disease areas,
!
!!!!
!
!!
!
reproductive endocrinology, infection, and cardiovascular diseases. We
! !
! !
Immune Checkpoint Antibody Sets ! ! !
! ! !! !
ProSci has released Risk-Free Immune Checkpoint Antibody Sets test services for SDS-PAGE, IP, ICC, IHC, and WB, as well as ELISA/CLIA
#&&&" !$ ) )
(
assays, among others. Cloud-Clone can deliver thousands of animal
caused by the fact that many antibodies work well in one application modelsóhuman disease animal models, knockout animal models, and
but not in others, despite detecting the same proteins. Because !!
!
! -
of this characteristic of antibodies, scientists must search for and vices, such as animal feeding, conservation of strains, animal ethology,
'))
) ())
' and animal imaging. We also provide cell-based experiments including
(
( )()())%
) cell culture, transfection, proliferation, infection, and apoptosis. Further-
of Risk-Free monoclonal antibodies developed with recombinant more, drug discovery services are available, including drug screening,
protein antigens expressed in mammalian cells, to best recognize pharmacokinetics, toxicology, and pharmacology.
immune checkpoint targets in their most native form. Because Cloud-Clone
of this development process, Risk-Free monoclonal antibodiesí !"
epitopes have a broad range of binding features and work in multiple www.cloud-clone.com
applications. Rigorously tested in a variety of applications, the best
antibodies are carefully selected and paired together to provide the Skin Toxicology Assay
broadest selection of binding features. DermaChip is a high-throughput assay technology that allows
ProSci scientists to reliably measure the genotoxic risk of a whole range of
For info: 888-513-9525 environmental agents (including cleaning agents, household products,
www.prosci-inc.com industrial chemicals, and cosmetics) to our skin. The assay traps
basal keratinocytes, which potentially harbor damaged DNA from
Cell Sorter the analysis and reduce the background for some 3D models, before
!
##
#
"!" "!! ! !"
cytometry and sorting surprisingly simple, saving you time and technology is based on the incorporation of collagen I into our
$#$
"$
$$ $ $ CometChips, forming a disposable gel attached to the specially treated
all while facilitating consistent results and high-purity sorting. With 96-well-format glass slides. Instead of being randomly dispersed in
BD FACSMelody, sorting into plates is as easy as sorting into tubes agarose on a glass slide as in the traditional Comet Assay, cells in the
and requires no extra setup steps. Select the collection format and !"! ! !!
" !
" !
desired number of cells, and the system automatically aligns the 30-µm diameter). Leveraging this advance, a single 96-well DermaChip
stream precisely to the center of the multiwell plate. Operators of all can produce 20,000 data points per chip, based on an average of 208
skill levels can quickly learn to use it. cells imaged per well.
BD Biosciences AMS Biotechnology
For info: 877-232-8995
www.bdbiosciences.com www.amsbio.com/dermachip.aspx
Electronically submit your new product description or product literature information! Go to www.sciencemag.org/about/new-products-section 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.
Chemistry
Issue date: July 27
Book ad by July 12
Ads accepted until July 20 if space allows
The Americas Why choose this Chemistry Focus for your advertisement?
+ 202 326 6577 § Relevant ads lead of the career section with a special
Europe “Chemistry” banner
+44 (0) 1223 326527
§ Bonus distribution to:
Japan
American Chemical Society Fall, August 19–23, Boston, MA.
+81 3 6459 4174
China/Korea/Singapore/
Expand your exposure.
Taiwan
Post your print ad online to beneft from:
+86 131 4114 0012
§ Link on the job board homepage directly
to chemistry jobs
Produced by the Science/AAAS
Custom Publishing Ofce. § Dedicated landing page for jobs
in chemistry
§ Additional marketing driving relevant
job seekers to the job board.
SCIENCECAREERS.ORG
FO R R EC R U I T M E N T I N S C I E N C E , T H E R E ’S O N LY O N E S C I E N C E.
HOW FAR
WILL YOUR ESSAY
TAKE YOU?
Apply for the Science & SciLifeLab Prize for Young Scientists —
an annual prize awarded to early-career scientists. The prize is
presented in four categories: Cell and Molecular Biology, Genomics
and Proteomics, Ecology and Environment, and Translational Medicine.
SCIENCEPRIZE.SCILIFELAB.SE
online @sciencecareers.org
POSITIONS OPEN
myIDP: A career plan TENURE TRACK FACULTY POSITIONS
customized for you, by you. Physiology
With the rapid revitalization of Detroit, Wayne State
University (WSU) is undergoing a major growth in
its research enterprise. The Department of Physiology
at WSU School of Medicine (SOM) (http://physiology.
med.wayne.edu) invites applications for two tenure-
track Assistant/Associate Professor positions. We seek
candidates with research interests in cardiovascular,
metabolic, respiratory and other areas of physiology,
SCIENCE CAREERS pathophysiology and biophysics, applying molecular,
cellular and systems approaches. WSU SOM is a state
ADVERTISING of the art research environment, rated in the top third
of all US Research Institutions by the Carnegie Foun-
For full advertising details, dation. The Department of Physiology, ranked #40
in 2017 of the ~120 Physiology departments in US, has
go to ScienceCareers.org one of the most active research programs among the
and click For Employers, basic science departments at WSU-SOM. Start-up
or call one of our package and salary are highly competitive.
representatives. Candidates should hold Ph.D., M.D. or equivalent
There’s only one Science. from a relevant area. The selected candidates are ex-
pected to establish extramurally funded active research
programs and participate in teaching medical and
graduate students. Please apply to website: https://
jobs.wayne.edu/applicants/jsp/shared/position/
Recommended by JobDetails_css.jsp?postingId=542621, posting 043648,
leading professional by uploading a curriculum vitae, a detailed future re-
societies and the NIH search plan, and names/contact information of three
AMERICAS references. Please submit inquiry with Curriculum Vitae
to Physiologyfacultysearch@wayne.edu. Review of
+1 202 326-6577 applications will begin after July 15, 2018 and con-
+1 202 326-6578 Features in myIDP include: tinue until the positions are filled.
Wayne State University is a premier, public, urban
advertise@sciencecareers.org research university located in the heart of Detroit where
§ Exercises to help you examine students from all backgrounds are offered a rich, high
E U R O P E , I N D I A , AU S T R A L I A , your skills, interests, and values. quality education. Our deep-rooted commitment to
NEW ZEAL AND, REST OF excellence, collaboration, integrity, diversity and in-
clusion creates exceptional educational opportunities
WO R L D § A list of 20 scientific career paths preparing students for success in a diverse, global so-
with a prediction of which ones ciety. WSU encourages applications from women, people
+44 (0) 1223 326527 of color and other underrepresented people. WSU is an
best fit your skills and interests. Affirmative Action/Equal Opportunity Employer.
advertise@sciencecareers.org
§ A tool for setting strategic goals
C H I N A , KO R E A , S I N G A P O R E ,
for the coming year, with optional
Search more
TA I WA N , T H A I L A N D
reminders to keep you on track.
+86 131 4114 0012
advertise@sciencecareers.org § Articles and resources to guide
you through the process.
jobs online
JA PA N
§ Options to save materials online Access hundreds
+81 3-6459-4174
and print them for further review of job postings on
advertise@sciencecareers.org and discussion.
ScienceCareers.org.
C U S TO M E R S E RV I C E § A certificate of completion for
users that finish myIDP Expand your
AMERICAS and more. search today.
+1 202 326-6577
R E S T O F WO R L D
Start planning today!
myIDP.sciencecareers.org
+44 (0) 1223 326528
advertise@sciencecareers.org
In partnership with:
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 encourages our readers to alert us to any ads
that they feel may be discriminatory or offensive.
ScienceCareers.org
ways that Science Careers
can help advance your career
Visit ScienceCareers.org
today — all resources are free
SCIENCECAREERS.ORG
online @sciencecareers.org
SRON, the Dutch national expertise institute for
scientific space research, is seeking a
career advice & more! with an international reputation in the field of scientific
space research. The director is well connected to the
international and national space research community
and has a clear vision on the future scientific direction
of space research. The director has proven managerial
@ScienceCareers skills at a research organization and has the natural
ability and authority for leading scientific and support
staff. The director is a teamplayer and has a proven
track record of leading international scientific research
collaborations. The director promotes diversity, is a
strong communicator and has a keen eye for
/ScienceCareers opportunities to connect science with industry and
society.
Featured Topics:
§ Networking
§ Industry or Academia
§ Job Searching
§ Non-Bench Careers
§ And More
WORKING LIFE
By Elise A. Kikis
“A
re you really going to cross the picket line?” my mother asked. She had called after reading that the
clerical union workers were on strike, and she could hear in the background the tell-tale honking
horns and ringing bells of the picket line. “Yes,” I responded. Despite her impassioned pleas, I was
not going to boycott my first day of graduate school. A few weeks later, I said no when asked to
join the graduate student union. Why would I, a paid student, need union representation? If only
I had known then what I know now, 16 years later, this is what I would have told myself.
ing dries up, you will move out of your apartment and onto thing in your power to lessen the financial hurdles that
friends’ sofas. This may sound like a fun adventure now, but students and early career scholars face. Doing science is
it will be demoralizing when you are 32 years old, homeless, difficult enough. Your students don’t need to be hazed just
collecting unemployment, and hoping for a job at Walgreens because you were. j
(they don’t call you back), all while doing your science for
free. You will earn a few thousand dollars adjunct teaching, Elise A. Kikis is an assistant professor of biology at the
which will put food on the table for a few months. University of the South in Sewanee, Tennessee. Send your
You will then be one of the lucky ones to hit pay dirt on career story to SciCareerEditor@aaas.org.
Published by AAAS