Science - 22 July 2022

Misconduct suspicions cloud Lasing from time A screening approach to improve
key Alzheimer’s research p. 358 crystals pp. 368 & 425 cancer nanomedicine pp. 371 & 384
$15
22 JULY 2022
science.org
SHAPING
RANGES
Competition limits bird
distributions on
tropical mountains p. 416
QUALITY CONTENT FOR THE GLOBAL SCIENTIFIC COMMUNITY
Multiple ways to stay informed on issues related to your research
Sponsored
Collection Booklets
Podcasts Advertorials
Posters Webinars
Scan the code and start exploring

the latest advances in science and
technology innovation!
Science.org/custom-publishing
Brought to you by the Science/AAAS Custom Publishing Office.
Posters Podcasts Sponsored Advertorials Webinars

Collection Booklets
0722Product.indd 346 7/14/22 7:45 AM

CONTENTS
2 2 J U LY 2 0 2 2 • VO LU M E 3 7 7 • I S S U E 6 6 0 4
FEATURES
358 Blots on a field?
A neuroscience image sleuth finds signs of
fabrication in scores of Alzheimer’s articles,
threatening a reigning theory of the disease
By C. Piller
363 Research backing experimental
Alzheimer’s drug was
first target of suspicion
By C. Piller
PODCAST
INSIGHTS
PERSPECTIVES
364 Reference genomes for conservation
High-quality reference genomes for
non–model species can benefit conservation
By S. Paez et al.
366 In the glare of the Sun

Searches during twilight toward the Sun have
found several asteroids near Venus’ orbit
By S. S. Sheppard
368 To make a mirrorless laser

Periodic temporal modulation of a photonic
crystal can be used to produce laser light
By D. Faccio and E. M. Wright
REPORT p. 425
369 Improving catalysis by moving water

The conversion of gases into building blocks
for synthesizing plastics is enhanced
By M. Ding and Y. Xu
REPORT p. 406
364 370 The quest for more food

Rice yield is increased by boosting nitrogen
A reference genome for the kākāpō, an endangered parrot in New Zealand, could help conservation breeding programs. uptake and photosynthesis By S. Kelly
RESEARCH ARTICLE p. 386
354 Consortium seeks to expand

NEWS human gene catalog
Finding sequences that code for short
371 One step closer to cancer
nanomedicine
High-throughput tool uncovers links between
proteins could add thousands of genes
cell signaling and nanomaterial uptake
IN BRIEF By R. F. Service
By J. O. Winter
350 News at a glance RESEARCH ARTICLE p. 384
355 Russian scientist facing treason
charges dies in custody
IN DEPTH Advocates say state’s zeal for arrests has POLICY FORUM
destroyed the lives of researchers working in 373 What will it take to stabilize
352 As Omicron rages on, virus’ path
sensitive fields By O. Dobrovidova the Colorado River?
remains unpredictable
PHOTO: STEPHEN BELCHER/MINDEN PICTURES
Fast-spreading subvariants are A continuation of the current 23-year-long

356 Deadly pest reaches Oregon, drought will require difficult decisions to
coming and going. But an entirely new sparking fears for ash trees
variant could still emerge prevent further decline By K. G. Wheeler et al.
Emerald ash borer has already killed millions
By K. Kupferschmidt of trees By G. Popkin
BOOKS ET AL.
353 Cleaner air is adding to 357 Half of Americans anticipate a 376 Setting college students
global warming U.S. civil war soon, survey finds up for success
Satellites capture fall Findings suggest rising gun violence will spill A pair of researchers outline strategies
in light-blocking pollution into the political sphere, driven by conspiracy for ensuring that postsecondary
By P. Voosen theories By R. Pérez Ortega courses are inclusive By J. Hsu
SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 347

CONTENTS
377 The virtual worlds of the metaverse 385 Coronavirus 411 Organic chemistry
An immersive internet is just around the Pathogen-sugar interactions revealed A concise synthesis of tetrodotoxin
corner, for better or worse By D. Greenbaum by universal saturation transfer analysis D. B. Konrad et al.
C. J. Buchanan et al.
LETTERS RESEARCH ARTICLE SUMMARY; FOR FULL TEXT: 416 Biogeography
DOI.ORG/10.1126/SCIENCE.ABM3125 Interspecific competition limits bird
379 Better preparation for Iran’s
species’ ranges in tropical mountains
forest fires 386 Plant science B. G. Freeman et al.
By M. Tavakoli Hafshejani et al. A transcriptional regulator that boosts
grain yields and shortens the growth duration 420 Coronavirus
379 China’s restoration fees of rice S. Wei et al. Shifting mutational constraints in the
require transparency RESEARCH ARTICLE SUMMARY; FOR FULL TEXT:
SARS-CoV-2 receptor-binding
By S. Gao et al. DOI.ORG/10.1126/SCIENCE.ABI8455
PERSPECTIVE p. 370 domain during viral evolution T. N. Starr et al.
380 Global goals overlook
387 Protein design 425 Optics
freshwater conservation
Scaffolding protein functional sites using Amplified emission and lasing in photonic
By D. V. Gonçalves and V. Hermoso
deep learning J. Wang et al. time crystals M. Lyubarov et al.
PERSPECTIVE p. 368
394 Surface chemistry
RESEARCH Quantum effects in thermal reaction rates
at metal surfaces D. Borodin et al.
428 Coronavirus
Pathogenicity, transmissibility, and fitness
of SARS-CoV-2 Omicron in Syrian hamsters
399 Evolution S. Yuan et al.
IN BRIEF
A chromosomal inversion contributes
381 From Science and other journals to divergence in multiple traits between Semiconductors
deer mouse ecotypes E. R. Hager et al. 433 High ambipolar mobility in cubic
RESEARCH ARTICLES
boron arsenide revealed by transient
384 Nanomedicine REPORTS
reflectivity microscopy S. Yue et al.
Massively parallel pooled screening reveals 406 Catalysis
Physical mixing of a catalyst and a 437 High ambipolar mobility in cubic boron
genomic determinants of nanoparticle
hydrophobic polymer promotes arsenide J. Shin et al.
delivery N. Boehnke et al.
RESEARCH ARTICLE SUMMARY; FOR FULL TEXT: CO hydrogenation through dehydration
DOI.ORG/10.1126/SCIENCE.ABM5551 W. Fang et al.
PERSPECTIVE p. 371 PERSPECTIVE p. 369 DEPARTMENTS
349 Editorial
Confronting 21st-century monekypox
By M. T. Osterholm and B. Gellin
442 Working Life

From weakness comes strength
By S. X. Pfister
ON THE COVER
Species tend to live in narrower slices of
mountainside on tropical versus temperate
mountains. Stronger competition in the
tropics explains this pattern for birds.
For example, the habitable range of this
white-tipped sicklebill
(Eutoxeres aquila) is
limited as a result of
ILLUSTRATOIN: IAN HAYDON, INSTITUTE FOR PROTEIN DESIGN

competition with its close
relative, the buff-tailed
sicklebill (Eutoxeres
condamini). See page
416. Photo: ©Juan Carlos
387 Vindas/Getty Images
Computer renderings of protein structures designed using deep learning methods Science Careers ......................................... 441
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 © 2022 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): $2212; Foreign postage extra: Air assist delivery: $98. 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 available from backissues.science.org; 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 can be obtained
through the Copyright Clearance Center (CCC), www.copyright.com. The identification code for Science is 0036-8075. Science is indexed in the Reader’s Guide to Periodical Literature and in several specialized indexes.
348 22 JULY 2022 • VOL 377 ISSUE 6604 science.org SCIENCE

EDITORIAL
Confronting 21st-century monkeypox
T
he World Health Organization (WHO) hasn’t lic health measures, including increased vaccination Michael T.
called the current monkeypox outbreak a Pub- and diagnostic testing and extensive education cam- Osterholm
lic Health Emergency of International Concern paigns targeted at populations at risk and minimiz- is director of
(PHEIC), but as a worldwide epidemic, it is clearly ing social stigma. In addition to a massive scaling up the Center for
an emerging pandemic. More than 12,556 mon- of vaccine production, other immediate dose-sparing Infectious Disease
keypox cases and three deaths have been reported actions can be taken: administration of a single dose Research and Policy
in 68 countries since early May, and these num- per person instead of two doses (or a first dose fol- at the University
bers will rise rapidly with improved surveillance, access lowed by a delayed second dose when supplies allow) of Minnesota,
to diagnostics, and continuing global spread of infection. or intradermal (versus intramuscular) administration Minneapolis, MN,
Although many tools are needed to control this unfold- of a smaller dose. However, research will be needed
USA. [email protected]
ing pandemic, it’s clear that limiting ongoing spread will to determine whether such dose-sparing approaches
require a comprehensive international vaccination strat- provide adequate immune protection.
Bruce Gellin
egy and adequate supplies. Determining how vaccine will be allocated to coun-
is chief of global
People 40 years old and younger who have not bene- tries and within countries to have the most impact on
fitted from the immunization campaign that eradicated transmission is essential. Expect major shortages of vac- public health
smallpox by 1980 are now susceptible cine among frustrated at-risk individuals strategy at The
to monkeypox (which is in the same vi- for many months to come. To dampen the Rockefeller
rus family as smallpox), and this lack of current outbreak will require vaccination Foundation,
population immunity has contributed to
the current outbreak. Most of the cases
to date have occurred among men who
have sex with men (MSM), particularly
>12K
Monkeypox cases
of those at highest risk, with global es-
timates of the number of MSM ranging
from 1 to 3%. The needed global vaccine
supply just for MSM is similar to those
Washington, DC,
USA. bgellin@
rockfound.org
those with new or multiple partners. Epi- considered for HIV oral preexposure pro-
demiologic investigations indicate that
the predominant mode of transmission is
through skin-to-skin and sexual contact,
not contact with contaminated cloth-
3 Deaths
phylaxis (PrEP). It is estimated that by
2023, 2.4 million to 5.3 million people
worldwide should receive PrEP.
Monkeypox is a zoonotic disease;
ing or bed linens. Although respiratory thus, another critical step is to greatly
droplet transmission might occur, there
is no evidence of airborne transmission
as there is with COVID-19. And because
monkeypox is a self-limited infection with
68 Countries
reduce transmission of the virus from
current rodent reservoirs and to prevent
spillovers in areas of the world where
monkeypox isn’t endemic. Long-term
symptoms lasting 2 to 4 weeks, there isn’t control of monkeypox will require vac-
a chronic carrier state as there is with HIV, cinating as many as possible of the 327
which would increase the risk for ongoing transmission. million people 40 years of age and younger living in
Although many tools are needed, it is clear that lim- the 11 African countries where monkeypox is endemic
iting ongoing spread will require widely available vac- in an animal (rodent) reservoir. This effort should in-
cination. The ACAM2000 vaccine is licensed by the US clude childhood vaccine programs. Surveillance will
Food and Drug Administration for smallpox and allowed be needed to identify new animal reservoirs, which
for use against monkeypox on an expanded access ba- might be established in other countries as a result of
sis (so-called “compassionate use” for an investigational infected humans inadvertently transmitting the virus
drug use). It is associated with potentially serious side to domestic rodents that have subsequent contact with
effects. A newer vaccine with an improved safety pro- wild rodents.
file was approved for monkeypox and smallpox in 2019. The smallpox eradication program was a 12-year ef-
This two-dose vaccine, produced by Bavarian Nordic, is fort that involved 73 countries working with as many
a modified vaccinia virus Ankara (MVA; Jynneos in the as 150,000 national staff. Because of its animal reser-
United States, Imvanex in the European Union, and Ima- voir, monkeypox can’t be eradicated. Unless the world
mune in Canada). Its supply, however, is limited. develops and executes an international plan to contain
How can the world leverage these vaccines to control the current outbreak, it will be yet another emerging
the spread of monkeypox? Transmission among MSM infectious disease that we will regret not containing.
populations must be reduced through aggressive pub- – Michael T. Osterholm and Bruce Gellin
Published online 19 July 2022; 10.1126/science.add9651

NEWS
IN BRIEF
Edited by Jeffrey Brainard
GLOBAL HEALTH
Pandemic contributes to big drop in childhood vaccinations
I
n what UNICEF Executive Director Catherine Russell against the three dangerous diseases. The majority of chil-
called a “red alert,” childhood vaccination rates in many dren who missed shots live in India, Nigeria, Indonesia,
countries worldwide have dropped to the lowest level Ethiopia, and the Philippines, but the largest relative
since 2008, in part because of the COVID-19 pandemic. drops occurred in two countries with much smaller popu-
UNICEF and the World Health Organization together lations: Myanmar and Mozambique. A similar number
track inoculations against diphtheria, pertussis, and of children did not get their first dose of the measles
tetanus—which are administered as one vaccine— People queue vaccine, and millions also missed polio and hu-
as a marker for vaccination coverage overall. In for vaccinations in man papillomavirus inoculations. The pandemic
2021, only 81% of children worldwide received the June 2021 in India. has limited the ability of health care workers to
The country had
recommended three doses of the combined vac- provide immunizations and disrupted supply
the largest drop in
cine, down from 86% in 2019. As a result, some inoculated children chains, UNICEF says; armed conflicts and vaccine
25 million children remain insufficiently protected during the pandemic. misinformation also contributed to the declines.
PHOTO: SUNIL GHOSH/HINDUSTAN TIMES/GETTY IMAGES

a “protein subunit” vaccine, containing agency’s manufacturing standards. The
At last, U.S. OKs Novavax vaccine the coronavirus spike protein and an authorization only applies to the primary
| After a long wait, Novavax’s
C OV I D -1 9 immune stimulant; the company hopes series of inoculations, but the company
COVID-19 vaccine last week joined the this will appeal to people who worry about hopes that in coming months FDA will
short list of pandemic shots authorized in side effects from Pfizer’s and Moderna’s authorize a booster dose of the vaccine.
the United States. The U.S. Food and Drug messenger RNA vaccines and Johnson Uptake of the Novavax vaccine in the
Administration (FDA) issued an emer- & Johnson’s adenovirus-based vaccine. European Union, where it was authorized
gency use authorization for the two-dose FDA’s blessing was delayed in part because in December 2021, has been slow: Only
vaccine on 13 July. Novavax’s product is Novavax struggled for months to meet the about 250,000 people have gotten it so far.

train visiting scientists on computational THEY SAID IT
What makes an old growth forest? tools, such as machine learning, to make
I was worried about my
“
| President Joe Biden’s
C O N S E R VAT I O N sense of the vast data being collected by
administration last week asked for public efforts such as the NSF-funded National
comment to help it define what constitutes Ecological Observatory Network and the goldfish getting too hot.
an “old growth” forest, to inform its efforts Ocean Observatories Initiative. Now I’m worried about the
to inventory and protect those on federal
lands. A better, “universal” definition that survival of my family and
Disputed fossil to return to Brazil my neighbors.
”
reflects evolving scientific understanding
of these “unique” ecosystems is needed, | A German science
PA L E O N T O L O GY
Hannah Cloke, a natural hazards researcher
says the formal request for comment ministry plans to repatriate an unusual
at the University of Reading, about Europe’s
issued on 15 July by the U.S. Forest Service dinosaur fossil to Brazil, where scientists
worsening heat wave, which she calls “a
and the Bureau of Land Management, alleged the artifact had been removed
wake-up call about the climate emergency.”
which together manage nearly 80 million illegally. Scientists at the State Museum
hectares of forests. Although big, ancient of Natural History Karlsruhe (SMNK)
trees are often seen as key markers of old published a paper in 2020 describing the
I want young people
growth, forests containing them “dif-
fer widely in character” depending on
factors such as geography and climate,
the agencies note. Suggestions are due
chicken-size dinosaur with spearlike feath-
ers that they called Ubirajara jubatus. But
after the team failed to provide proper
documentation of the fossil’s legality, the
“
to see that this is possible
for them, and that
by 15 August, and disagreement is likely: journal withdrew the paper. Last year, it’s not off limits because
Environmentalists want the new definition a Science investigation (1 October 2021,
they are Black.
to be expansive, and the timber industry
prefers a narrower one.
p. 14) prompted the science ministry of
Germany’s Baden-Württemberg state,
which manages SMNK, to investigate;
this week, authorities concluded that the
”
Marine geologist Dawn Wright, in Nature,
after becoming the first Black person
to visit Challenger Deep, Earth’s
U.S. ‘superbug’ infections rise museum provided the ministry with false deepest spot, aboard a submersible this
| Infections and deaths caused
C OV I D -1 9 information regarding the fossil’s acquisi- month. She and a crewmate used
by some of the most harmful antibiotic- tion, prompting the decision to return it. side-scan sonar for seafloor mapping.
resistant pathogens in U.S. hospitals leapt
by at least 15% during the first year of
the coronavirus pandemic, the Centers Mauna Kea’s summit is both
for Disease Control and Prevention (CDC) a premier site for astronomy
reported last week. The spike boosted and sacred ground.
2020 deaths from these infections to
29,400 and was a turnabout from declines
in “superbug” infections during the previ-
ous decade. Causes included overworked
hospital workers forced to let sanitation
precautions slip and shortages of personal
protective equipment, CDC said. Resistant
microbes deemed among the most danger-
ous drove the largest reported increases in
rates of hospital-acquired infections. For
example, the rate for Acinetobacter bac-
teria resistant to carbapenem antibiotics
increased by 78%, with 7500 such cases.
The microbe commonly infects patients
on ventilators, such as those hospitalized
for COVID-19. ASTRONOMY
Native Hawaiians gain voice in managing Mauna Kea

Center tackles ecological data
T
he state of Hawaii this month created a new management body for Mauna Kea,
FUNDING | The University of Colorado, one of the world’s best sites for astronomy, that could help resolve a long-running
Boulder, will host a new research center dispute over telescopes on its summit. Many Native Hawaiians consider the
to synthesize large amounts of data about mountain sacred and have long objected to the observatories, especially the pro-
PHOTO: THOMAS LINKEL/LAIF/REDUX
environmental change, such as increas- posed construction of the Thirty Meter Telescope (TMT), a U.S.-led international
ing wildfires and biodiversity loss. The project. Under a new state law, control of the summit will be transferred over 5 years
Environmental Data Science Innovation from the University of Hawaii to a new body whose 11 members will be appointed by
and Inclusion Lab will fill “an enormous the governor and include representatives of Native Hawaiian groups, the Mauna Kea
need,” said a statement last week from the observatories, and others. Mauna Kea Anaina Hou, one of the Native Hawaiian groups
U.S. National Science Foundation (NSF) that has opposed TMT’s construction, objected that the panel’s Native Hawaiian
announcing it will fund the new center with members will not be chosen by the groups and may be heavily outnumbered.
$20 million over 5 years. The project will

NEWS
A nurse prepares a COVID-19 vaccine in Guwahati,

India, on 10 April. A new subvariant named BA.2.75
that was first detected in India has surfaced in many
other countries.
winding down. “Those systems are being

dismantled, they are being defunded, peo-
ple are being fired,” she says.
The variants that ruled in 2021 did not
arise one out of the other. Instead, they
evolved in parallel from SARS-CoV-2 vi-
ruses circulating early in the pandemic.
In the viral family trees researchers draw
to visualize the evolutionary relationships
of SARS-CoV-2 viruses, these variants ap-
peared at the tips of long, bare branches.
The pattern seems to reflect virus lurk-
ing in a single person for a long time and
evolving before it emerges and spreads
again, much changed.
More and more studies seem to confirm
that this occurs in immunocompromised
people who can’t clear the virus and have
long-running infections. On 2 July, for ex-
IN DEP TH ample, Yale University genomic epidemio-
logist Nathan Grubaugh and his team
posted a preprint on medRxiv about one
COVID-19 such patient they found accidentally. In
the summer of 2021, their surveillance
As Omicron rages on, virus’ program at the Yale New Haven Hospi-
tal kept finding a variant of SARS-CoV-2
called B.1.517 even though that lineage was
path remains unpredictable supposed to have disappeared from the

community long ago. All of the samples,
it turned out, came from the same per-
Fast-spreading subvariants are coming and going. But an son, an immunocompromised patient in
his 60s undergoing treatment for a B cell
entirely new variant could still emerge lymphoma. He was infected with B.1.517 in
November 2020 and is still positive today.
By Kai Kupferschmidt or BA.5 will spawn additional descendants By following his infection to observe
with more mutations and that one or more how the virus changed over time, the team
I
n the short history of the COVID-19 of those subvariants will spread and will be found it evolved at twice the normal speed
pandemic, 2021 was the year of the the next thing,” says Jesse Bloom, an evo- of SARS-CoV-2. (Some of the viruses circu-
new variants. Alpha, Beta, Gamma, and lutionary biologist at the Fred Hutchinson lating in the patient today might be quali-
Delta each had a couple of months in Cancer Research Center. fied as new variants if they were found
the Sun. But others believe a new variant differ- in the community, Grubaugh says.) That
But this was the year of Omi- ent enough from Omicron and all other supports the hypothesis that chronic infec-
cron, which swept the globe late in 2021 variants to deserve the next Greek letter tions could drive the “unpredictable emer-
and has continued to dominate, with designation, Pi, may already be develop- gence” of new variants, the researchers
subvariants—given more prosaic names ing, perhaps in a chronically infected pa- write in their preprint.
such as BA.1, BA.2, and BA.2.12.1—appearing tient. And even if Omicron is not replaced, Other viruses that chronically infect pa-
in rapid succession. Two closely related sub- its dominance is no cause for complacency, tients also change faster within one host
variants named BA.4 and BA.5 are now says Maria Van Kerkhove, technical lead than when they spread from one person to
driving infections around the world, but for COVID-19 at the World Health Organi- the next, says Aris Katzourakis, an evolu-
new candidates, including one named zation. “It’s bad enough as it is,” she says. tionary biologist at the University of Ox-
BA.2.75, are knocking on the door. “If we can’t get people to act [without] a ford. This is partly a numbers game: There
Omicron’s lasting dominance has evolu- new Greek name, that’s a problem.” are millions of viruses replicating in an in-
tionary biologists wondering what comes Even with Omicron, Van Kerkhove em- dividual, but only a handful are passed on PHOTO: ANUPAM NATH/AP IMAGES
next. Some think it’s a sign that SARS- phasizes, the world may face continuing during transmission. So a lot of potential
CoV-2’s initial frenzy of evolution is over waves of disease as immunity wanes and evolution is lost in a chain of infections,
and it, like other coronaviruses that have fresh subvariants arise. She is also alarmed whereas a chronic infection allows for end-
been with humanity much longer, is set- that the surveillance efforts that allowed less opportunities to evolve.
tling into a pattern of gradual evolution. researchers to spot Omicron and other But since Omicron emerged in November
“I think a good guess is that either BA.2 new variants early on are scaling back or 2021, no new variants have appeared out

of nowhere. Instead, Omicron has accu- it’s something worth keeping a close eye ATMOSPHERIC SCIENCE
mulated small changes, making it better at on,” says Emma Hodcroft, a virologist at
evading immune responses and—together
with waning immunity—leading to succes-
sive waves. “I think it’s probably harder and
the University of Bern.
Keeping an eye on anything is getting
harder, however, because surveillance is
Cleaner air
harder for these new things to emerge and
take over because all the different Omicron
lineages are stiff competition,” Grubaugh
decreasing. Switzerland, for example, now
sequences about 500 samples per week,
down from 2000 at its peak, Hodcroft
is adding to
says, given how transmissible and immune-
evading they already are.
If so, the U.S. decision to update
says; the United States went from more
than 60,000 per week in January to about
10,000. “Some governments are anxious
global warming
COVID-19 vaccines by adding an Omicron to cut back on the money they dedicated Satellites capture fall in
component is the right move, Bloom says; to sequencing,” Hodcroft says. Defending light-blocking pollution
even if Omicron keeps changing, a vaccine the expense is a “hard sell,” she says, “es-
based on it is likely to provide more pro- pecially if there’s a feeling the countries
tection than one based on earlier variants. around you will continue sequencing even By Paul Voosen
But it’s still possible that an entirely new if you stop.”
I
variant unrelated to Omicron will emerge. Even if a variant emerges in a place t’s one of the paradoxes of global warm-
Or one of the previous variants, such as with good surveillance, it may be harder ing. Burning coal or gasoline releases
Alpha or Delta, could make a comeback af- than in the past to predict how big a the greenhouse gases that drive cli-
ter causing a chronic infection and going threat it poses, because differences in past mate change. But it also lofts pollu-
through a bout of accelerated evolution, COVID-19 waves, vaccines, and immunization particles that reflect sunlight and
says Tom Peacock, a virologist at Imperial tion schedules have created a global check- cool the planet, offsetting a fraction of
College London: “This is what we would erboard of immunity. That means a new the warming. Now, however, as pollution-
call second-generation variants.” Given variant might do well in one place but run control technologies spread, both the nox-
those possibilities, “Studying chronic in- into a wall of immunity elsewhere. “The ious clouds and their silver lining are start-
fections is now more important than ever,” situation has become even less predicting to dissipate.
says Ravindra Gupta, a microbiologist at able,” Katzourakis says. Using an array of satellite observa-
the University of Cambridge. “They might Given that Omicron appears to be milder tions, researchers have found that the cli-
tell us the kind of mutational direction the than previous variants, surveillance efforts matic influence of global air pollution has
virus will take in the population.” should aim to identify variants that cause dropped by up to 30% from 2000 levels.
BA.2.75, which was picked up recently, severe disease in hospitalized patients, Although this is welcome news for public
already has some scientists concerned. Gupta says. “I think that that’s where we health—airborne fine particles, or aerosols,
Nicknamed Centaurus, it evolved from should be focusing our efforts, because if we are believed to kill several million people
Omicron but seems to have quickly ac- keep focusing on new variants genomically, per year—it is bad news for global warm-
cumulated a whole slew of important we may get a bit fatigued, and then kind of ing. The cleaner air has effectively boosted
changes in its genome, more like an en- drop the ball when things do happen.” the total warming from carbon dioxide
tirely new variant than a new Omicron Many virologists acknowledge that emitted over the same time by anywhere
subvariant. “This looks exactly like Alpha SARS-CoV-2’s evolution has caught them from 15% to 50%, estimates Johannes
did, or Gamma or Beta,” Peacock says. by surprise again and again. “It was really Quaas, a climate scientist at Leipzig Uni-
BA.2.75 appears to be spreading in India, in part a failure of imagination,” Grubaugh versity and lead author of the study. And
where it was first identified, and has been says. But whatever scenario researchers as air pollution continues to be curbed, he
found in many other countries. Whether can imagine, Bloom acknowledges the vi- says, “There is a lot more of this to come.”
it’s really outcompeting other subvariants rus will chart its own course: “I think in “I believe their conclusions are correct,”
CREDITS: (GRAPHIC) C. BICKEL/SCIENCE AND N. DESAI/SCIENCE; (DATA) NEXTSTRAIN/GISAID
is unclear, Van Kerkhove says: “The data is the end, we just kind of have to wait and says James Hansen, a retired NASA climate
superlimited right now.” “I certainly think see what happens.” j scientist who first called attention to the
“Faustian bargain” of fossil fuel pollution
in 1991. He says it’s impressive scientific
Making waves detective work because no satellite could
A series of Omicron subvariants has appeared in rapid succession around the world since the beginning directly measure global aerosols over this
of this year. Some scientists say that pattern will likely continue—but an entirely new variant could still arise. whole period. “It’s like deducing the proper-
ties of unobserved dark matter by looking at
Delta BA.1 BA.2 BA.4 BA.5 BA.2.12.1
100
its gravitational effects.” Hansen expects a
flurry of follow-up work, as researchers seek
to quantify the boost to warming.
Percentage of samples
80
Some aerosols, such as black carbon, or
60
soot, absorb heat. But reflective sulfate and
nitrate particles have a cooling effect. For
40 many years, they formed from polluting
gases escaping from car tailpipes, ship flues,
20 and power plant smokestacks. Technologies
to scrub or eliminate this pollution have
0 spread slowly from North America and Eu-
February March April May June July rope to the developing world. Only in 2010

GENOMICS
Consortium
seeks to expand
human gene
catalog
Finding sequences that code
for short proteins could add
thousands of genes
Pollution in China has fallen with the spread of smokestack scrubbers. But the cleanup is boosting warming. By Robert F. Service
T
did air pollution in China begin to decline, tion should undo the effect—and using the he relatively small universe of hu-
for example, and international restrictions same instruments, Quaas and his team man genes could grow by up to one-
on sulfur-heavy ship fuel have come just in found a clear decrease in cloud droplet third, if a concerted effort to search
the past few years. concentrations in the same regions where for new genes that encode short
The new study, submitted as a preprint aerosols declined. proteins is successful. Many known
to Atmospheric Chemistry and Physics in The evidence in the paper is clear, says miniproteins have already been
April and expected for publication in the Joyce Penner, an atmospheric scientist at the shown to play key roles in cellular me-
next few months, grew directly out of last University of Michigan, Ann Arbor. “It’s re- tabolism and disease, so the international
year’s U.N. climate assessment. It included markable that we’re seeing this already,” she effort to catalog new ones and determine
studies showing aerosol declines in North says. “This is contributing a lot to the climate their functions, announced last week in
America and Europe but no clear global changes we’re seeing in the current era.” Nature Biotechnology, could shed light on
trends. Quaas and his co-authors thought Just how much this declining reflectivity a vast array of biochemical processes and
two NASA satellites, Terra and Aqua, oper- has boosted recent warming is hard to quan- provide targets for novel medicines.
ating since 1999 and 2002, might be able tify, says Stuart Jenkins, a doctoral student “The microproteome is a potential gold
to help. at the University of Oxford who is also study- mine of unexplored biology,” says Eric
The satellites tally Earth’s incoming and ing the aerosol decline. In forthcoming work, Olson, a molecular biologist at the Univer-
outgoing radiation, which has enabled sev- Jenkins will show there’s just too much natu- sity of Texas Southwestern Medical Center
eral research groups, including Quaas and ral variability in the past 20 years to pick out who is not involved with the new consor-
his colleagues, to track the increase in in- the effect of clearer skies. tium. Anne O’Donnell-Luria, an expert in
frared heat trapped by greenhouse gases. Whatever the exact contribution, it is sure the genetics of rare diseases at Boston Chil-
But one instrument on Aqua and Terra to grow as air quality continues to improve dren’s Hospital, adds that the expanded
has also shown a decline in reflected light. around the world. The answer isn’t to keep catalog could be a rich source of clues to ge-
Models suggested a decrease in aerosols polluting, says Jan Cermak, a remote-sensing netic links to disease. “Everyone will be able
is partly responsible, says Venkatachalam scientist at the Karlsruhe Institute of Tech- to use this data set to make progress in
Ramaswamy, director of the National Oce- nology. “Air pollution kills people. We need their area.”
anic and Atmospheric Administration’s clean air. There is no question about that.”
CREDITS: (PHOTO) CHINATOPIX/AP IMAGES; (GRAPHIC) V. ALTOUNIAN/SCIENCE

Geophysical Fluid Dynamics Laboratory. Instead, efforts to reduce greenhouse gases
“It’s very hard to find alternate reasons for need to be redoubled, he says.
this,” he says. But with Earth having warmed by some
Quaas and his co-authors have now 1.2°C since preindustrial times, Hansen
taken things a step further with two in- thinks there’s little hope of cutting emis-
struments on Terra and Aqua that record sions fast enough to meet the 1.5°C target
the haziness of the sky—and therefore its he and other scientists have called for. And
aerosol load. From 2000 to 2019, haze over so the solution, he says, could come back
North America, Europe, and East Asia to aerosols, this time ones spread deliber-
clearly declined, although it continued to ately through solar geoengineering—the
thicken over coal-dependent India. controversial idea of lofting sulfate par-
Aerosols don’t just reflect light on their ticles into the stratosphere and creating a
own; they can also alter clouds. By serving global, reflective haze. “It will be necessary
as nuclei on which water vapor condenses, to take temporary corrective measures,” he
pollution particles reduce cloud droplet says, “almost surely including temporary An international consortium will search for RNAs
size and increase their number, making purposeful use of aerosols to avoid cata- (blue) that are converted to functional small proteins
clouds more reflective. Reducing pollu- strophic implications.” j (orange) by ribosomes (center).

NE WS | I N D E P T H
Only 19,370 human genes are known EUROPE

to code for proteins. But current catalogs
only include genes for proteins contain-
ing at least 100 amino acids each, a cut-
off chosen in part because longer DNA
Russian scientist facing treason
sequences make it easier for geneticists to
look for commonalities between species.
Many smaller proteins are known to exist,
charges dies in custody
but they’ve largely flown under the radar Advocates say state’s zeal for arrests has destroyed
even though some have been shown to play
crucial roles in regulating the immune sys-
the lives of researchers working in sensitive fields
tem, blocking other proteins, and destroy-
ing faulty RNAs. “The fact that these have By Olga Dobrovidova says the arrests are driven by perverse in-
been excluded represents a large hole in centives at FSB, where agents are eager to
L
genetics and developmental biology,” says ast month, Dmitry Kolker, 54, direc- supply “enemies of the state” in return for
consortium member John Prensner, a pedi- tor of the Laboratory of Quantum Op- bonuses and promotions.
atric oncologist at Boston Children’s. tics at Novosibirsk State University, Eugene Chudnovsky, a physicist at
When genes are translated into proteins, was dealing with late-stage pancre- Lehman College and co-chair of the Com-
they are first transcribed into snippets of atic cancer. But on 30 June, agents with mittee of Concerned Scientists, believes the
messenger RNA (mRNA). Cellular organ- Russia’s Federal Security Service (FSB) prosecutions may also be “an intimidation
elles called ribosomes then read those removed him from a cancer clinic, flew him tactic” directed at scientists more deeply
mRNA sequences and follow their instruc- to Moscow, and detained him on charges of involved in sensitive research, which the
tions to string together amino acids into treason. By 2 July, he was dead. His family Russian government is careful not to dis-
proteins. When scientists scan for genes, learned of his fate via a curt telegram. rupt too much.
they typically look for distinctive DNA se- Kolker’s colleagues at the Russian Acad- Pavlov says the criteria for classifying in-
quences flanked by start and stop signals emy of Sciences (RAS) expressed outrage. formation as state secrets are purposefully
for the protein assembly process, so-called A group of RAS members signed an open vague, with all details themselves classi-
open reading frames (ORFs). letter protesting FSB’s handling of the case fied, so it is easy to manufacture an accu-
In recent years, researchers have come and called for “those guilty of our
up with other ways to identify protein- colleague’s death to be held ac-
coding sequences. One called Ribo-seq uses countable.” Kolker’s family told
high-throughput sequencing technology to local media he was accused of
catalog all the RNAs in a sample that are leaking state secrets to China.
bound to a ribosome at a given time. Those But the RAS group posted a
RNA sequences point to likely genes, al- photo of an expert report from
though the technique can’t prove that any an RAS institute concluding that
one sequence makes a stable, functional optics lectures Kolker gave in
protein. Ribo-seq databases now contain China in 2018 included no classi-
thousands of ORFs, many of which don’t fied information.
code for known proteins and therefore The case is far from unusual.
may represent new ones. Three days before Kolker’s ar-
In the consortium’s first phase, members rest, FSB arrested another re-
scanned seven Ribo-seq databases for can- searcher in Siberia: Anatoly
didate ORFs that might correspond with Maslov, 75, an aerodynamicist
small proteins. After weeding out redundant at the Khristianovich Institute
entries they came up with 7264 candidates. of Theoretical and Applied Me-
Next, the group will try to identify which chanics, who now faces up to
of those yield proteins with actual cellular 20 years in prison on treason
functions. Techniques such as mass spectro- charges. A 2020 investigation Laser physicist Dmitry Kolker died this month after being
metry can help determine whether particu- from independent Moscow accused of divulging state secrets.
lar RNAs are translated into stable proteins. newspaper Novaya Gazeta found
Others, such as epitope tagging, use antibod- that more than 30 scientists had been ac- sation. Viktor Kudryavtsev, an aerospace
ies to track marked proteins, revealing their cused of treason since 2000. Like Maslov, engineer who collaborated with European
location and abundance in cells and provid- many worked on hypersonics, a research researchers on a hypersonics project, was
ing hints about their function. area at the center of a new arms race arrested in 2018 even though a military
For now, the 35 investigators involved (Science, 10 January 2020, p. 136). review panel had previously approved the
PHOTO: ALEXANDER FEFELOV/REUTERS
are funding the effort from their own lab Scientists are “prime targets” for FSB work; FSB classified the work 5 years after
budgets, and don’t have immediate plans because they have access to sensitive in- the project ended.
to seek dedicated funding. “There is so formation and often travel to conferences The relationship between scientists and
much there, this just needs to be done,” and meet with foreign colleagues, says the Russian security services has long been
says consortium member Sebastiaan van Ivan Pavlov, a defense lawyer for opposi- fraught, says David Holloway, a historian
Heesch, a systems biologist at the Princess tion leader Alexei Navalny’s foundation of the Soviet nuclear program at Stanford
Máxima Center for Pediatric Oncology in and several treason suspects who fled Rus- University. In the Soviet era, “There was
the Netherlands. j sia himself after being detained by FSB. He certainly an incentive to find guilty people

NEWS | I N D E P T H
and targets to be met,” he says. “If you are INVASIVE SPECIES

not arresting people, you aren’t doing your
job.” But during the Cold War, prominent
scientists could leverage their usefulness
in the nuclear weapons program to gain
Deadly pest reaches Oregon,
some protection with party bosses. “The
physicists were somehow protected by the
bomb, they were needed.”
sparking fears for ash trees
Alexander Fedulov, Kolker’s lawyer, says Emerald ash borer has already killed millions of trees
the family intends to fight to clear the
physicist’s name. Yaroslav Kudryavtsev,
a polymer scientist and Viktor’s son, also By Gabriel Popkin documented in 36 states; Washington, D.C.;
kept up efforts to vindicate his father in and parts of Canada.
T
court, even after he died in 2021 while hree years ago, forest scientists on the Once the borer shows up, “You cannot,
under court-mandated travel restrictions. U.S. West Coast launched an effort to generally speaking, get rid of [it],” says Leigh
But the family gave up this year, after the gather nearly 1 million seeds of the Greenwood, a forest specialist at the Nature
Ukraine war began and Russia passed laws Oregon ash. The ecologically valuable Conservancy. In Oregon, officials will likely
to end the jurisdiction of the European tree, found from southern California try to slow its spread and reduce its popula-
Court of Human Rights. to British Columbia in Canada, often tion through selective use of insecticides and
For Chudnovsky, the futility of seeking grows along streams and in wetlands, an- by releasing tiny wasps that parasitize and
an acquittal in Russian courts sets these choring rich ecosystems. kill the beetles. Such strategies have been
cases apart from the China Initiative in In part, the collecting effort represented used elsewhere with limited success.
the United States, a law enforcement cam- disaster insurance: The emerald ash borer, In the longer term, some researchers hope
paign that was launched in 2018 to prevent an invasive, iridescent green beetle that has to breed trees that can resist the beetle. Of
China from stealing U.S. technologies and wiped out ash trees throughout much of the now-imperiled western ash species, Or-
was recently rethought (Science, 4 March, the eastern and midwestern United States, egon ash (Fraxinus latifolia) is the most
p. 945). Still, Pavlov’s team has managed to was spreading westward, and the saved immediate concern. In sensitive wetlands
secure pardons and shorter prison terms seeds might one day help where it can form nearly
for several defendants. “In today’s Russia, restore the species if the pure stands, no other tree
freedom is much more valuable than any pest ever arrived. can readily take its place.
available justice,” he says. Now, it appears that res- “In some areas it’s the only
Private and public support from the cue mission began none [tree] species there,” says
scientific community was vital to Viktor too soon. Last week, the Richard Sniezko, a USFS
Kudryavtsev and his family, but ultimately U.S. Department of Agri- geneticist and a leader of
could not do much to protect the scientist, culture confirmed that the the seed collecting project.
his son says. Boris Altshuler, a theoretical emerald ash borer (Agrilus Sniezko began working
physicist and human rights activist at RAS’s planipennis) had reached on ash trees after attend-
P.N. Lebedev Physical Institute, says that in Oregon—and likely has ing a 2019 conference. He
Soviet times, international pressure from been there for up to 5 years. is now growing seedlings
researchers could sometimes bring the se- The discovery marked the The emerald ash borer now has from a number of Or-
curity apparatus to heel. “Now, I’m not sure borer’s first appearance a foothold in the western United States. egon ash populations at
whether the man at the top would listen.” west of the Rocky Moun- a research station in the
At home, public displays of support have tains; previously it had only gotten as far as state; colleagues are overseeing a similar
become scarce since the beginning of the Boulder, Colorado. Forest managers now fear set of plantings in Washington and Ohio.
war in Ukraine and a government crack- for the future of the Oregon ash and at least Once the ash borer arrives, researchers will
down on protests and dissent. RAS President eight other ash species found only in western observe how the trees fare. Individual trees
Alexander Sergeev, who just a few years ago North America. that hold up better than others might ulti-
publicly called for Viktor Kudryavtsev to be “It’s extremely grave and sobering to have mately help scientists breed new, hardier
released from jail, has remained quiet about the situation upon us,” Karen Ripley, a forest varieties. Such breeding efforts are already
Kolker and Maslov. In a June speech, he told health monitoring coordinator with the U.S. underway for other ash species in Ohio
colleagues to stop “insulting the state” with Forest Service (USFS), wrote last week in an (Science, 13 November 2020, p. 756).
antiwar declarations. email to colleagues. Researchers are also beginning to collect
PHOTO: MACROSCOPIC SOLUTIONS/SCIENCE SOURCE

In Akademgorodok, the enclave of On 30 June, a biologist with the city of seeds from up to eight other endemic ash
Novosibirsk research institutes where Portland alerted officials to adult beetles species that only live in the southwestern
Kolker and Maslov worked, short-lived me- he saw emerging from a tree in nearby For- United States. That effort is challenging
morials and graffiti about them keep pop- est Grove, Oregon. The next day, Wyatt because several of the species are rare and
ping up despite police efforts. At the edge Williams, an invasive species specialist with grow in remote areas, says Tim Thibault, a
of a forest, a note of protest was taped on the Oregon Department of Forestry, con- curator at the Huntington, a botanical gar-
top of an official tick warning. It read, firmed that an Oregon ash was infested. “My den in San Marino, California, who is co-
“Kolker and Maslov are victims of Moscow heart just sank,” he says. leading the project.
occupants, Siberia is not a colony.” j The report opened a new front in the The seed collections are only the begin-
nearly 2-decade-old fight against the borer, ning of a long and expensive process, scien-
Olga Dobrovidova is a science journalist in Paris an Asian species that was first found in tists warn. Rescuing a tree through breeding,
who does climate communications work. 2002 outside Detroit and has since been Sniezko says, “is not for the faint of heart.” j

The insurrection at the U.S. Capitol on 6 January 2021
showed how politics could motivate violence.
you’re [still] getting these high numbers …

is really quite concerning.”
She is less alarmed by the shaky sup-
port for democracy, noting that political
gridlock—as in U.S. politics today—can of-
ten distort attitudes. “What people mean
by ‘democracy’ is pretty fuzzy,” she says. Po-
litical paralysis, she adds, can quickly lead
people who think, “Yeah, I like democracy,”
to also say, “Yeah, I want a strong man”
in leadership.
“The findings are scary, but not sur-
prising,” Kurt Braddock, who studies the
psychology of extremist communication at
American University, wrote in an email to
GUN VIOLENCE Science. In recent years, he says, the United
States has seen an increase in individual
Half of Americans anticipate a willingness to engage in violence—homicides

in cities increased 44% between 2019 and
2021, for instance—an attitude he says is
U.S. civil war soon, survey finds likely to spill into the political sphere.
Researchers have criticized the sam-
pling and survey methodology of previous
Findings suggest rising gun violence will spill into the studies that found increasing support for
political sphere, driven by conspiracy theories political violence. But the new study gen-
erally agrees with earlier efforts, Kleinfeld
says. A small survey from 2021, for in-
By Rodrigo Pérez Ortega violence. The respondents were part of the stance, found about 46% of voters thought
Ipsos KnowledgePanel—an online research the United States would have another civil
V
iolence can seem to be everywhere panel that has been used widely, including war, and another showed more than one-
in the United States, and political by Wintemute for research on violence and third of Americans agree that “The tradi-
violence is in the spotlight, with the firearm ownership. The team then applied tional American way of life is disappearing
6 January 2021 insurrection as exhibit statistical methods to extrapolate the sur- so fast that we may have to use force to
A. Now, a large study confirms one vey results to the entire country. save it.”
in five Americans believes violence Although almost all respondents thought Wintemute and colleagues found that
motivated by political reasons is—at least it’s important for the United States to re- conspiracy theories, some rooted in rac-
sometimes—justified. Nearly half expect a main a democracy, about 40% said having ism, are helping shape views about po-
civil war, and many say they would trade de- a strong leader is more important. Half ex- litical violence. They found roughly two in
mocracy for a strong leader, a preprint sub- pect a civil war in the United States in the five adults agreed with the white national-
mitted last week to medRxiv found. next few years. (The survey didn’t specify ist “great replacement theory,” or the idea
“This is not a study that’s meant to when.) “The fact that basically half the coun- that native-born white voters are being re-
shock,” says Rachel Kleinfeld, a political vio- try is expecting a civil war is just chilling,” placed by immigrants for electoral gains.
lence expert at the Carnegie Endowment for Wintemute says. And many expect to take And one in five respondents believed the
International Peace who was not involved part. If found in a situation where they think false QAnon conspiracy theory that U.S. in-
in the research. “But it should be shocking.” violence is justified to advance an impor- stitutions are controlled by an elite group
Firearm deaths in the United States grew tant political objective, about one in five re- of Satan-worshipping pedophiles.
by nearly 43% between 2010 and 2020, and spondents thinks they will likely be armed To reduce the threat of political vio-
gun sales surged during the coronavirus with a gun. About 7% of participants— lence, Braddock says, the first step is to
pandemic. Garen Wintemute, an emer- which would correspond to about 18 mil- call out the disinformation online and in
gency medicine physician and longtime lion U.S. adults—said they would be willing right-wing media, some of which is taken
gun violence researcher at the University to kill a person in such a situation. directly from extremist propaganda. “We
PHOTO: SHAY HORSE/NURPHOTO/GETTY IMAGES
of California, Davis, wondered what those Kleinfeld says the study’s findings are need to call that out for what it is before
trends portend for civil unrest. “Sometimes compelling because of the large number we can begin to address the problems it
being an ER [emergency room] doc is like of participants and because it asked about is causing.” Kleinfeld adds that leaders—
being the bow man on the Titanic going, specific scenarios in which participants from politicians and media personalities
‘Look at that iceberg!’” he says. think violence is justified—such as for self- to church pastors—can also make a dif-
He and his colleagues surveyed more defense or to stop people with different ference. Experiments show courageous
than 8600 adults in English and Spanish political beliefs from voting. The sample leaders can deter their communities from
about their views on democracy in the does slightly overrepresent older people, engaging in violence. “Now’s the time to
United States, racial attitudes in U.S. soci- who are not known to commit much vio- take this seriously and not put our heads
ety, and their own attitudes toward political lence worldwide, she says. “So the fact that in the sand,” Kleinfeld says. j

NEWS
FEATURES
BLOTS ON A FIELD?
A neuroscience image sleuth finds signs of fabrication in scores of Alzheimer’s
articles, threatening a reigning theory of the disease By Charles Piller
I
n August 2021, Matthew Schrag, a scientists who are also short sellers who Neuroscientist and physician Matthew Schrag
neuroscientist and physician at Vander- profit if the company’s stock falls—believed found suspect images in dozens of papers
bilt University, got a call that would some research related to Simufilam may have involving Alzheimer’s disease, including Western
plunge him into a maelstrom of pos- been “fraudulent,” according to a petition blots (projected in green) measuring a protein
sible scientific misconduct. A colleague later filed on their behalf with the U.S. Food linked to cognitive decline in rats.
wanted to connect him with an at- and Drug Administration (FDA).
torney investigating an experimental Schrag, 37, a softspoken, nonchalantly So he applied his technical and medical
drug for Alzheimer’s disease called rumpled junior professor, had already gained knowledge to interrogate published images
Simufilam. The drug’s developer, Cas- some notoriety by publicly criticizing the about the drug and its underlying science—
PHOTO: JOSEPH ROSS
sava Sciences, claimed it improved cognition, controversial FDA approval of the anti-Ab for which the attorney paid him $18,000. He
partly by repairing a protein that can block drug Aduhelm. His own research also con- identified apparently altered or duplicated
sticky brain deposits of the protein amyloid tradicted some of Cassava’s claims. He feared images in dozens of journal articles. The at-
beta (Ab), a hallmark of Alzheimer’s. The volunteers in ongoing Simufilam trials faced torney reported many of the discoveries in
attorney’s clients—two prominent neuro- risks of side effects with no chance of benefit. the FDA petition, and Schrag sent all of them

NE WS
to the National Institutes of Health (NIH), by Lesné. Schrag’s work, done independently While prepping for medical school at the
which had invested tens of millions of dollars of Vanderbilt and its medical center, implies University of North Dakota, Schrag spent
in the work. (Cassava denies any misconduct millions of federal dollars may have been long hours in a neuropharmacology lab ab-
[see sidebar, p. 363].) misspent on the research—and much more sorbing the patient rhythms of science. He
But Schrag’s sleuthing drew him into a dif- on related efforts. Some Alzheimer’s experts repeated experiments over and over, refining
ferent episode of possible misconduct, lead- now suspect Lesné’s studies have misdirected his skills. These included a protein identifica-
ing to findings that threaten one of the most Alzheimer’s research for 16 years. tion method known as the Western blot. It
cited Alzheimer’s studies of this century and “The immediate, obvious damage is wasted uses electricity to drive protein-rich tissue
numerous related experiments. NIH funding and wasted thinking in the field samples through a gel that acts like a sieve to
The first author of that influential study, because people are using these results as a separate the molecules by size. Distinct pro-
published in Nature in 2006, was an ascend- starting point for their own experiments,” teins, tagged and illuminated by fluorescent
ing neuroscientist: Sylvain Lesné of the Uni- says Stanford University neuroscientist antibodies, appear as stacked bands.
versity of Minnesota (UMN), Twin Cities. His Thomas Südhof, a Nobel laureate and expert In 2006, Schrag’s first publication exam-
work underpins a key element of the domi- on Alzheimer’s and related conditions. ined how feeding a high-cholesterol diet to
nant yet controversial amyloid hypothesis Lesné did not respond to requests for com- rabbits seemed to increase Ab plaques and
of Alzheimer’s, which holds that Ab clumps, ment. A UMN spokesperson says the univer- iron deposits in one part of their brains. Not
known as plaques, in brain tissue are a pri- sity is reviewing complaints about his work. long afterward, when he was an M.D.-Ph.D.
mary cause of the devastating illness, which student at Loma Linda University, another
afflicts tens of millions globally. In what research group found support for a link be-
looked like a smoking gun for the theory “You can’t cheat to cure a tween Alzheimer’s and iron metabolism.
and a lead to possible therapies, Lesné and
his colleagues discovered an Ab subtype and
disease. Biology doesn’t care.” Encouraged, Schrag poured his energy into
trying to confirm the connection in people—
seemed to prove it caused dementia in rats. If Matthew Schrag, Vanderbilt University and failed. The experience introduced him
Schrag’s doubts are correct, Lesné’s findings to a disquieting element of Alzheimer’s re-
were an elaborate mirage. To Schrag, the two disputed threads of Ab search. With this enigmatic, complex disease,
Schrag, who had not publicly revealed research raise far-reaching questions about even careful experiments done in good faith
his role as a whistleblower until this article, scientific integrity in the struggle to under- can fail to replicate, leading to dead ends and
avoids the word “fraud” in his critiques of stand and cure Alzheimer’s. Some adherents unexpected setbacks.
Lesné’s work and the Cassava-related studies of the amyloid hypothesis are too uncritical of One of its biggest mysteries is also its most
and does not claim to have proved miscon- work that seems to support it, he says. “Even distinctive feature: the plaques and other
duct. That would require access to original, if misconduct is rare, false ideas inserted into protein deposits that German pathologist
complete, unpublished images and in some key nodes in our body of scientific knowledge Alois Alzheimer first saw in 1906 in the brain
cases raw numerical data. “I focus on what we can warp our understanding.” of a deceased dementia patient. In 1984, Ab
can see in the published images, and describe was identified as the main component of
them as red flags, not final conclusions,” he IN HIS MODEST OFFICE, steps away from a the plaques. And in 1991, researchers traced
says. “The data should speak for itself.” buzzing refrigerator, Schrag displays an family-linked Alzheimer’s to mutations in the
A 6-month investigation by Science pro- antique microscope—an homage to pre- gene for a precursor protein from which am-
vided strong support for Schrag’s suspi- decessors who applied painstaking bench yloid derives. To many scientists, it seemed
cions and raised questions about Lesné’s science to medicine’s endless enigmas. A clear that Ab buildup sets off a cascade of
research. A leading independent image ana- small sign on his desk reads, “Everything damage and dysfunction in neurons, causing
lyst and several top Alzheimer’s researchers— is figureoutable.” dementia. Stopping amyloid deposits became
including George Perry of the University of So far, Alzheimer’s has been an exception. the most plausible therapeutic strategy.
Texas, San Antonio, and John Forsayeth of But Schrag’s background has left him com- Hundreds of clinical trials of amyloid-
the University of California, San Francisco fortable with the field’s contradictions. His targeted therapies have yielded few glimmers
(UCSF)—reviewed most of Schrag’s findings father hails from a family of Mennonites, of promise, however; only the underwhelm-
at Science’s request. They concurred with known for their philosophy of peacemaking— ing Aduhelm has gained FDA approval. Yet
his overall conclusions, which cast doubt on but joined the military. The family moved Ab still dominates research and drug de-
hundreds of images, including more than from Arizona to Germany to England be- velopment. NIH spent about $1.6 billion on
70 in Lesné’s papers. Some look like “shock- fore settling in Davenport, a tiny cow town projects that mention amyloids in this fiscal
ingly blatant” examples of image tampering, in eastern Washington. After leaving the year, about half its overall Alzheimer’s fund-
says Donna Wilcock, an Alzheimer’s expert at Air Force, Schrag’s dad became a nurse and ing. Scientists who advance other potential
the University of Kentucky. worked in a nursing home. As a young teen, Alzheimer’s causes, such as immune dys-
The authors “appeared to have composed Schrag volunteered to visit dementia patients function or inflammation, complain they
figures by piecing together parts of photos there. “I remembered being mystified by a lot have been sidelined by the “amyloid mafia.”
from different experiments,” says Elisabeth of the strange behaviors,” he says. It was a for- Forsayeth says the amyloid hypothesis be-
Bik, a molecular biologist and well-known mative experience “to see people struggling came “the scientific equivalent of the Ptol-
forensic image consultant. “The obtained with such unfair symptoms.” emaic model of the Solar System,” in which
experimental results might not have been Home-schooled by his mom, Schrag en- the Sun and planets rotate around Earth.
the desired results, and that data might have tered community college at 16, like many of By 2006, the centenary of Alois Alzheimer’s
been changed to … better fit a hypothesis.” the town’s studious kids—including his teen- epic discovery, a growing cadre of skeptics
Early this year, Schrag raised his doubts age sweetheart and future wife, Sarah. They wondered aloud whether the field needed a
with NIH and journals including Nature; now live on a small ranch outside Nashville reset. Then, a breathtaking Nature paper en-
two, including Nature last week, have pub- with their two young children and three ag- tered the breach.
lished expressions of concern about papers ing horses that Sarah grew up with. It emerged from the lab of UMN physi-

NEWS | F E AT U R E S
cian and neuroscientist Karen Ashe, who had mer, and Alzheimer’s” has risen from near lems with other blots in the same articles.
already made a remarkable series of discov- zero to $287 million in 2021. Lesné and Ashe He also found some blot backgrounds that
eries. As a medical resident at UCSF, she con- helped spark that explosion, experts say. seemed to have been improperly duplicated.
tributed to Nobel laureate Stanley Prusiner’s The paper provided an “important boost” Three of the papers listed Lesné, whom
pioneering work on prions—infectious pro- to the amyloid and toxic oligomer hypotheses Schrag had never heard of, as first or senior
teins that cause rare neurological disorders. when they faced rising doubts, Südhof says. author. Schrag quickly found that another
In the mid-1990s, she created a transgenic “Proponents loved it, because it seemed to be Lesné paper had also drawn scrutiny on Pub-
mouse that churns out human Ab, which an independent validation of what they have Peer, and he broadened his search to Lesné
forms plaques in the animal’s brain. The been proposing for a long time.” papers that had not been flagged there. The
mouse also shows dementia-like symptoms. “That was a really big finding that kind of investigation “developed organically,” he says,
It became a favored Alzheimer’s model. turned the field on its head,” partly because of as other apparent problems emerged.
By the early 2000s, “toxic oligomers,” Ashe’s impeccable imprimatur, Wilcock says. “So much in our field is not reproducible,
subtypes of Ab that dissolve in some bodily “It drove a lot of other investigators to … go so it’s a huge advantage to understand when
fluids, had gained currency as a likely chief looking for these [heavier] oligomer species.” data streams might not be reliable,” Schrag
culprit for Alzheimer’s—potentially more As Ashe’s star burned more brightly, Le- says. “Some of that’s going to happen repro-
pathogenic than the insoluble plaques. sné’s rose. He joined UMN with his own ducing data on the bench. But if it can hap-
Amyloid oligomers had NIH-funded lab in 2009. pen in simpler, faster ways—such as image
been linked to impaired Ab*56 remained a primary analysis—it should.” Eventually Schrag ran
communication between research focus. Megan across the seminal Nature paper, the basis
neurons in vitro and in ani- Larson, who worked as a ju- for many others. It, too, seemed to contain
mals, and autopsies have nior scientist for Lesné and multiple doctored images.
shown higher levels of the is now a product manager Science asked two independent image
oligomers in people with at Bio-Techne, a biosciences analysts—Bik and Jana Christopher—to re-
Alzheimer’s than in cogni- supply company, calls him view Schrag’s findings about that paper and
tively sound individuals. passionate, hardworking, others by Lesné. They say some supposed ma-
But no one had proved that and charismatic. She and nipulation might be digital artifacts that can
any one of the many known others in the lab often ran occur inadvertently during image process-
oligomers directly caused experiments and produced ing, a possibility Schrag concedes. But Bik
cognitive decline. Western blots, Larson says, found his conclusions compelling and sound.
In the brains of Ashe’s but in their papers together, Christopher concurred about the many du-
transgenic mice, the UMN Lesné prepared all the implicated images and some markings suggest-
team discovered a previ- ages for publication. ing cut-and-pasted Western blots flagged by
ously unknown oligomer He became a leader of Schrag. She also identified additional dubi-
species, dubbed Ab*56 (pro- UMN’s neuroscience gradu- ous blots and backgrounds he had missed.
nounced “amyloid beta star Sylvain Lesné, ate program in 2020, and in In the 16 years following the landmark pa-
56”) after its relatively heavy University of Minnesota, Twin Cities May 2021, 4 months after per, Lesné and Ashe—separately or jointly—
molecular weight compared Schrag delivered his con- published many articles on their stellar
with other oligomers. The group isolated cerns to NIH, Lesné received a coveted R01 oligomer. Yet only a handful of other groups
Ab*56 and injected it into young rats. The rats’ grant from the agency, with up to 5 years of have reported detecting Ab*56.
capacity to recall simple, previously learned support. The NIH program officer for the Citing the ongoing UMN review of Lesné’s
information—such as the location of a hid- grant, Austin Yang—a co-author on the 2006 work, Ashe declined via email to be inter-
den platform in a maze—plummeted. The Nature paper—declined to comment. viewed or to answer written questions posed
2006 paper’s first author, sometimes cred- by Science, which she called “sobering.” But
ited as the discoverer of Ab*56, was Lesné, IN DECEMBER 2021, Schrag visited PubPeer, she wrote, “I still have faith in Ab*56,” not-
a young scientist Ashe had hired straight out a website where scientists flag possible ering her ongoing work studying the structure
of a Ph.D. program at the University of Caen rors in published papers. Many of the site’s of Ab oligomers. “We have promising initial
Normandy in France. posts come from technical gumshoes who results. I remain excited about this work, and
Ashe touted Ab*56 on her website as “the deconstruct Western blots for telltale marks believe it has the potential to explain why Ab
first substance ever identified in brain tissue indicating that bands representing proteins therapies may yet work despite recent fail-
in Alzheimer’s research that has been shown could have been removed or inserted where ures targeting amyloid plaques.”
to cause memory impairment.” An accompa- they don’t belong. Such manipulations can But even before Schrag’s investigation, the
nying editorial in Nature called Ab*56 “a star falsely suggest a protein is present—or al- spotty evidence that Ab*56 plays a role in Al-
suspect” in Alzheimer’s. Alzforum, a widely ter the levels at which a detected protein is zheimer’s had raised eyebrows. Wilcock has
read online hub for the field, titled its cov- apparently found. Schrag, still focused on long doubted studies that claim to use “puri-
erage, “Ab Star is Born?” Less than 2 weeks Cassava-linked scientists, was looking for ex- fied” Ab*56. Such oligomers are notoriously
after the paper was published, Ashe won the amples that could refine his own sleuthing. unstable, converting to other oligomer types
prestigious Potamkin Prize for neuroscience, In a PubPeer search for “Alzheimer’s,” post- spontaneously. Multiple types can be present
partly for work leading to Ab*56. ings about articles in The Journal of Neuro- in a sample even after purification efforts, ILLUSTRATION: N. DESAI/SCIENCE
The Nature paper has been cited in about science caught Schrag’s eye. They questioned making it hard to say any cognitive effects
2300 scholarly articles—more than all but the authenticity of blots used to differentiate are due to AB*56 alone, she notes—assuming
four other Alzheimer’s basic research reports Ab and similar proteins in mouse brain tis- it exists. In fact, Wilcock and others say, sev-
published since 2006, according to the Web sue. Several bands seemed to be duplicated. eral labs have tried and failed to find Ab*56,
of Science database. Since then, annual NIH Using software tools, Schrag confirmed the although few have published those findings.
support for studies labeled “amyloid, oligo- PubPeer comments and found similar prob- Journals are often uninterested in negative

results, and researchers can be reluctant to How an image sleuth uncovered possible tampering
contradict a famous investigator. Vanderbilt University neuroscientist Matthew Schrag found apparently falsified images in papers by
An exception was Harvard University’s University of Minnesota, Twin Cities, neuroscientist Sylvain Lesné, including a 2006 paper in Nature co-authored
Dennis Selkoe, a leading advocate of the with Karen Ashe and others. It linked an amyloid-beta (Ab) protein, Ab*56, to Alzheimer’s dementia.
amyloid and toxic oligomer hypotheses, who
has cited the Nature paper at least 13 times.
In two 2008 papers, Selkoe said he could not Image in question Mou age
Mouse agess in
i mmonths
onths
find Ab*56 in human fluids or tissues. Ashe uploaded this Western blot 12 12 12 13 13 13 15 15 17 17 17 20
to PubPeer after Schrag said the Heavier
Selkoe examined Schrag’s dossier on
version published in Nature showed
Lesné’s papers at Science’s request, and says cut marks suggesting improper
he finds it credible and well supported. He tampering with bands portraying
Proteins
did not see manipulation in every suspect Ab*56 and other proteins (black
Aβ
A β*56 bands
b d
image, but says, “There are certainly at least boxes added by Ashe). The figure
12 or 15 images where I would agree that shows levels of Ab*56 (dashed red
box) increasing in older mice as
there is no other explanation” than manipu-
symptoms emerge. But Schrag’s
lation. One—an image in the Nature paper analysis suggests this version of
displaying purified Ab*56—shows “very wor- the image contains improperly Lighter
risome” signs of tampering, Selkoe says. The duplicated bands.
same image reappeared in a different paper,
co-authored by Lesné and Ashe, 5 years later. 1 Spot the similarities
Many other images in Lesné’s papers might Some bands looked abnormally similar,
an apparent manipulation that in some
be improper—more than enough to challenge cases (not shown) could have made
the body of work, Selkoe adds. Ab*56 appear more abundant than it
A few of Lesné’s questioned papers de- was. One striking example (red box)
scribe a technique he developed to measure ostensibly shows proteins that emerge
Ab oligomers separately in brain cells, spaces later in the life span than Ab*56.
outside the cells, and cell membranes. Selkoe
recalls Ashe talking about her “brilliant post- 2 Match contrast
doctoral fellow” who devised it. He was skep- Schrag matched the contrast level
in the two sets of bands for
tical of Lesné’s claim that oligomers could be
an apples-to-apples comparison.
analyzed separately inside and outside cells
in a mixture of soluble material from fro-
3 Colorize and align
zen or processed brain tissue. “All of us who
Schrag turned backgrounds black
heard about that knew in a moment that it to make the bands easier to see,
made no biochemical sense. If it did, we’d all then colorized them and precisely
be using a method like that,” Selkoe says. The matched their size and orientation.
CREDITS: (GRAPHIC) C. BICKEL/SCIENCE; (DATA) S. LESNÉ ET AL., NATURE 440, 352 (2006). HTTPS://DOI.ORG/10.1038/NATURE04533
Possibly duplicated bands

Nature paper depended on that method.
Selkoe himself co-authored a 2006 pa- 4 Merge
per with Lesné in the Annals of Neurology. He merged the sets of colorized
They sought to neutralize the effects of toxic bands. The areas of the image
oligomers, although not Ab*56. The paper that are identical appear in yellow.
includes an image that Schrag, Bik, and
Christopher agree was reprinted as if original 5 Calculate similarity This heat map shows one
in two subsequent Lesné articles. Selkoe calls Schrag then calculated the point for each group of pixels
that “highly egregious.” correlation coefficient, showing compared. Red indicates
Given those findings, the scarcity of inde- the strength of the relationship dense areas of the original
between the merged bands. image, such as the center
pendent confirmation of the Ab*56 claims Identical images show a of a band; purple indicates
seems telling, Selkoe says. “In science, once correlation of 1, and display sparse areas.
you publish your data, if it’s not readily repli- as a straight 45° angle line.
cated, then there is real concern that it’s not These bands show a 0.98
correct or true. There’s precious little clear- correlation, highly unlikely to
cut evidence that Ab*56 exists, or if it exists, occur by chance.
correlates in a reproducible fashion with fea-
tures of Alzheimer’s—even in animal models.” Unmistakable Fuzzier, insect-
differences wing shape
IN ALL, SCHRAG OR BIK identified more than These images examine shows both
20 suspect Lesné papers; 10 concerned dissimilar bands using the dense and sparse
Ab*56. Schrag contacted several of the jour- same process. In the merged areas of the
Dissimilar original images
nals starting early this year, and Lesné and image, clear differences
bands
his collaborators recently published two appear in green or red—as have dissimilar
expected when comparing elements.
corrections. One for a 2012 paper in The
naturally produced bands.
Journal of Neuroscience replaced several A degree of correlation is Merge
images Schrag had flagged as problematic, expected, but far lower than
writing that the earlier versions had been in duplicated bands.

NEWS | F E AT U R E S
“processed inappropriately.” But Schrag says tist at Caen, co-authored five Lesné papers NIH said it takes research misconduct seri-
even the corrected images show numerous flagged by Schrag or Bik. Vivien defends the ously, but otherwise declined to comment.
signs of improper changes in bands, and in validity of those articles, but says he had rea- In the fanfare around the Lesné-Ashe
one case, complete replacement of a blot. son to be wary of Lesné. work, some Alzheimer’s experts see a fail-
A 2013 Brain paper in which Schrag had Toward the end of Lesné’s time in France, ure of skepticism, including by journals that
flagged multiple images was also extensively Vivien says they worked together on a pa- published the work. After Schrag contacted
corrected in May. Lesné and Ashe were the per for Nature Neuroscience involving Ab. Nature, Science Signaling, and five other
first and senior authors, respectively, of the During final revisions, he saw immunos- journals about 13 papers co-authored by
study, which showed “negligible” levels of taining images—in which antibodies detect Lesné, a few are under investigation, accord-
Ab*56 in children and young adults, more proteins in tissue samples—that Lesné had ing to emails he received from editors.
when people reached their 40s, and steadily provided. They looked dubious to Vivien, “There are very strong, legitimate
increasing levels after that. It concluded and he asked other students to replicate the questions,” John Foley, editor of Science
that Ab*56 “may play a pathogenic role very findings. Their efforts failed. Vivien says he Signaling, later told Science. He says the
early in the pathogenesis of Alzheimer’s dis- confronted Lesné, who denied wrongdoing. journal has contacted authors and univer-
ease.” The authors said the correction had Although Vivien lacked “irrefutable proof” sity officers of two papers from 2016 and
no bearing on the study’s findings. of misconduct, he withdrew the paper be- 2017 for a response. It also recently issued
Schrag isn’t convinced. Among other fore publication “to preserve my scientific expressions of concern about the articles.
problems, one corrected integrity,” and broke off all A spokesperson for Nature, which pub-
blot shows multiple bands contact with Lesné, he says. lishes image integrity standards, says the
that appear to have been “We are never safe from a journal takes concerns raised about its papers
added or removed artifi- student who would like to seriously, but otherwise had no comment.
cially, he says. deceive us and we must re- Days after an inquiry from Science, Nature
Selkoe calls the appar- main vigilant.” published a note saying it was investigating
ently falsified corrections Schrag spot checked Lesné’s 2006 paper and advising caution
“shocking,” particularly in papers by Vivien or Ashe about its results.
light of Ashe’s pride in the without Lesné. He found The Journal of Neuroscience stands out
2006 Nature paper. “I don’t no anomalies—suggesting with five suspect Lesné papers. A journal
see how she would not Vivien and Ashe were in- spokesperson said it follows guidelines from
hyperscrutinize anything nocent of misconduct. the Committee on Publication Ethics to assess
that subsequently related Yet senior scientists concerns, but otherwise had no comment.
to Ab*56,” he says. must balance the trust es- “Journals and granting institutions don’t
After Science contacted sential to fostering a pro- know how to deal with image manipula-
Ashe, she separately posted tégé’s independence with tion,” Forsayeth says. “They’re not subject-
to PubPeer a defense of prudent verification, Wil- ing images to sophisticated analysis, even
some images Schrag had Karen Ashe, cock says. If you sign off though those tools are very widely avail-
challenged in the Nature pa- University of Minnesota, on images time after time, able. It’s not some magic skill. It’s their job
per. She supplied portions Twin Cities claim credit, speak pub- to do the gatekeeping.”
of a few original, unpub- licly, and win awards for Holden Thorp, editor-in-chief of the
lished versions that do not show the apparent the work—as Ashe has done—you have to Science journals, said the journals have
digital cut marks Schrag had detected in the be sure it’s right, she adds. subjected images to increasing scrutiny,
published images. That suggests the mark- “Ashe obviously failed in that very serious adding that “2017 would have been [near]
ings were harmless digital artifacts. Yet the duty” to ask tough questions and ensure the the beginning of when more attention
original images reveal something that Schrag data’s accuracy, Forsayeth says. “It was a ma- was being paid to this—not just for us,
and Selkoe find even more incriminating: jor ethical lapse.” but across scientific publishing.” He cited
unequivocal evidence that, despite the lack the Materials Design Analysis Reporting
of obvious cut marks, multiple bands were IN HIS WHISTLEBLOWER REPORT to NIH about framework developed jointly by several
copied and pasted from adjacent areas (see Lesné’s research, Schrag made its scope and publishers to improve data transparency
graphic, p. 361). stakes clear: “[This] dossier is a fraction and weed out image manipulation.
Schrag could find no innocent explanation of the anomalies easily visible on review As federal agencies, universities, and jour-
for a 2-decade litany of oddities. In experi- of the publicly accessible data,” he wrote. nals quietly investigate Schrag’s concerns, he
ment after experiment using Western blots, The suspect work “not only represents a decided to try to speed up the process by pro-
microscopy, and other techniques, serious substantial investment in [NIH] research viding his findings to Science. He knows the
anomalies emerged. But he notes that he has support, but has been cited … thousands of move could have personal consequences. By
not examined the original, uncropped, high- times and thus has the potential to mislead calling out powerful agencies, journals, and
resolution images. Authors sometimes share an entire field of research.” scientists, Schrag might jeopardize grants
those with researchers conducting similar The agency’s reply, which Schrag shared and publications essential to his success.
work, although they usually ignore such re- with Science, noted that complaints deemed But he says he felt an urgent need to go
quests, according to recent studies of data- credible will go to the Department of Health public about work that might mislead the ILLUSTRATION: N. DESAI/SCIENCE
sharing practices. Sharing agreements do not and Human Services Office of Research Integ- field and slow the race to save lives. “You
include access for independent misconduct rity (ORI) for review. That agency could then can cheat to get a paper. You can cheat to get
detectives. Lesné and Ashe did not respond instruct grantee universities to investigate a degree. You can cheat to get a grant. You
to a Science request for those images. prior to a final ORI review, a process that can can’t cheat to cure a disease,” he says. “Bio-
Questions about Lesné’s work are not new. take years and remains confidential absent logy doesn’t care.”
Cell biologist Denis Vivien, a senior scien- an official misconduct finding. To Science, Like other anti-Ab efforts, toxic oligomer

Research backing experimental Alzheimer’s drug was first target of suspicion
W
hen Vanderbilt University physician rated with the company for 15 years. expression of concern—reserving judgment
and neuroscientist Matthew Schrag None agreed to answer questions from until CUNY completes its investigation.
first grew suspicious of work under- Science. Cassava CEO Remi Barbier also Schrag received $18,000 from an attorney
lying a major theory of Alzheimer’s declined to answer questions or to name for short sellers behind the FDA petition, who
disease (see main story, p. 358), the company’s current scientific advisers. profit if Cassava’s value falls. Schrag, whose
he was following a different trail. In August He said in an email that Schrag’s dossier is efforts were independent of Vanderbilt, says
2021, he provided analysis for a petition to “generally consistent with prior allegations he worked hundreds of hours on the petition
the Food and Drug Administration (FDA), about our science … such allegations are and independent research and he has never
requesting that it pause two phase 3 clinical false.” Cassava hired investigators to review shorted Cassava stock or earned other
trials of Cassava Sciences’s Alzheimer’s its work, provided “nearly 100,000 pages of money for efforts on that issue, or for similar
drug Simufilam. The petition claimed some documents to an alphabet soup of outside work involving University of Minnesota,
science behind the drug might be fraudulent, investigative agencies,” and asked CUNY Twin Cities, neuroscientist Sylvain Lesné. (In
and the more than 1800 planned trial partici- to investigate, he added. That effort “has either case, if federal authorities determine
pants might see no benefits. yielded an important finding to date: there is fraud occurred and demand a return of grant
That month, Schrag submitted sting- no evidence of research misconduct.” (CUNY money, Schrag might be eligible to receive a
ing reports to the National Institutes of says it takes allegations of misconduct seri- portion of the funds.)
Health (NIH) about 34 published papers by ously, but otherwise declined to comment The most influential Cassava-related
Cassava-linked scientists, describing “seri- because of its ongoing investigation.) paper appeared in The Journal of Clinical
ous concerns of research misconduct.” His Last year, Schrag reached out to most Investigation in 2012. The authors—including
findings, including possibly manipulated of the journals that published questioned Wang; Arnold; David Bennett, who leads a
scientific images and suspect numerical papers. Seven were retracted—including five brain-tissue bank at Rush University; and his
data, challenge work supported by tens of by PLOS ONE in April. Three others received Rush colleague, neuroscientist Zoe Arvanita-
millions of dollars in NIH funds. Some of the expressions of concern; in each case, the kis—linked insulin resistance to Alzheimer’s
studies suggest Simufilam reinstates the editors said they were awaiting completion and the formation of amyloid plaques. Cas-
shape and function of the protein filamin A, of the CUNY investigation. In a few cases, the sava scientists say Simufilam lessens insulin
which Cassava claims causes Alzheimer’s editors told him, reviews were underway. resistance. They relied on a method in which
dementia when misfolded. (Other publica- Cassava has said editors of two suspect dead brain tissue, frozen for a decade and
tions have reported on the FDA petition, papers dismissed misconduct concerns. then partially thawed and chopped, purport-
but not Schrag’s identity. The Wall Street Last year, the editors of a 2005 Neuro- edly transmits nerve impulses.
Journal has reported that the U.S. Securi- science paper co-authored by Wang, Burns, Schrag and others say it contradicts basic
ties and Exchange Commission is also and others found no improper manipulation neurobiology. Schrag adds that he could find
investigating Cassava.) of Western blots, but said in an editorial note no evidence that other investigators have
In February, FDA refused to pause the they would review any concerns from an “in- replicated that result. (None of the authors
trials, calling the petition the wrong way to stitutional investigation,” apparently CUNY’s agreed to be interviewed for this article.)
intervene, but said it might eventually take probe. They did not respond to additional That paper supported the science behind
action. Independent image analysts and findings Schrag raised this year. Simufilam, Schrag says, “and spawned an
Alzheimer’s experts who reviewed Schrag’s Another paper that purportedly validated entire field of research in Alzheimer’s, ‘diabe-
findings at Science’s request generally agree science behind Simufilam—also by Wang, tes of the brain.’” It has been cited more than
with him. Burns, and colleagues—appeared in 2012 in 1500 times. Schrag sent the journal’s editor
Schrag’s sleuthing implicates work by The Journal of Neuroscience. In December his analysis of more than 15 suspect images.
Cassava Senior Vice President Lindsay 2021, the editors corrected one figure. In an email that Schrag provided to Science,
Burns, Hoau-Yan Wang of the City University Barbier said in a statement that they told the editor said the journal had reviewed high-
of New York (CUNY), and Harvard University him they had found no manipulation. But resolution versions of the images when they
neurologist Steven Arnold. Wang and Arnold in January, after Schrag and others raised were originally submitted and declined to
have advised Cassava, and Wang collabo- additional doubts, the editors issued an consider Schrag’s findings. —C.P.
research has spawned no effective therapies. tentially survives this one problem,” Schrag Selkoe’s bigger worry, he says, is that
“Many companies have invested millions and says. “But it makes you pause and rethink the Lesné episode might further under-
millions of dollars, or even billions ... to go the foundation of the story.” cut public trust in science during a time
after soluble Ab [oligomers]. And that hasn’t Selkoe adds that the broader amyloid hy- of increasing skepticism and attacks. But
worked,” says Daniel Alkon, president of the pothesis remains viable. “I hope that people scientists must show they can find and cor-
bioscience company Synaptogenix, who once will not become faint hearted as a result of rect rare cases of apparent misconduct, he
directed neurologic research at NIH. what really looks like a very egregious exam- says. “We need to declare these examples
Schrag says oligomers might still play role ple of malfeasance that’s squarely in the Ab and warn the world.” j
in Alzheimer’s. Following the Nature paper, oligomer field,” he says. But if current phase
other investigators connected combinations 3 clinical trials of three drugs targeting amy- With reporting by Meagan Weiland. This
of oligomers to cognitive impairment in loid oligomers all fail, he notes, “the Ab hy- story was supported by the Science Fund
animals. “The wider story [of oligomers] po- pothesis is very much under duress.” for Investigative Reporting.

INSIGHTS
PERSPECTIVES
GENOMICS
Reference genomes
for conservation
High-quality reference
genomes for non–model species
can benefit conservation
By Sadye Paez, Robert H. S. Kraus, date, most are in draft form. It is proposed nome assemblies (telomere to telomere) (5)
Beth Shapiro, M. Thomas P. Gilbert, that conservation efforts can be enhanced of both maternal and paternal haplotypes.
Erich D. Jarvis, the Vertebrate Genomes by the production of high-quality reference Given the extra computational and financial
Project Conservation Group genome assemblies. costs that such improved genomes incur, a
Conservation genomics leverages genetic legitimate question often asked is, what is
A
s of 2022, the International Union for data, from individual loci to genomic-scale the added value of these high-quality as-
Conservation of Nature (IUCN) Red datasets, to aid preservation of species and semblies, beyond current draft genomes,
List estimates that more than 32% population-level biodiversity. This includes for conservation?
of fungal, plant, and animal species using genomic data to measure effective Species must maintain a certain level of
are threatened with extinction. This population sizes, demographic history, and genetic diversity to adapt to various envi-
sixth mass extinction is caused by genetic diversity and to perform genetic ronmental changes and/or population de-
the activities and expanding biomass of hu- manipulations pre- or postextinction. Many creases, whether natural or human driven.
mans, necessitating a distinct name for this of these efforts have been conducted using Genetic diversity or other genomic health
geological epoch—the Anthropocene (1). first- and second-generation genome se- assessments have historically drawn on
Human population growth and the verte- quencing and assembly technologies with polymerase chain reaction (PCR)–generated
brate extinction rate (2) have been linearly short reads, leading to sequence errors, sequence data from DNA microsatellites,
correlated over the past 500 years (see the structural errors, and missing sequences. which are tandem repetitive sequences that
figure). For some species of conservation Now, third-generation genome technolo- tend to diverge at a higher rate compared
concern, documenting, informing, and miti- gies—with improvements in longer read with single-nucleotide variants. Typically,
PHOTO: STEPHEN BELCHER/MINDEN PICTURES

gating this biodiversity loss has been helped lengths, nucleotide accuracy, chromosomal the greater the diversity in microsatellites
by powerful genomic tools, including a ref- maps, and assembly algorithms—have led among individuals, the healthier the popu-
erence assembly (3). Yet, currently, only to more complete assemblies (4). These lation. Genomic health of a population can
a small fraction (<1%) of the ~35,500 spe- high-quality assemblies have 10- to 200-fold also be assessed by identifying changes in
cies assessed as threatened with extinction improvements in quality metrics, includ- mutational load (the population frequency
have an available genome assembly, and to ing the amount of sequence assignable to of deleterious alleles) and estimating
chromosomes, genes fully assembled, and lengths of runs of homozygosity (ROHs) (6).
recovery of GC-rich regulatory regions (4). The accumulation of ROHs in small and in-
Author affiliations and Vertebrate Genomes Project
Conservation Group authors and affiliations are listed in Method developments are also underway bred populations can fix or drive deleteri-
the supplementary materials. Email: [email protected] for generating complete and error-free ge- ous alleles to high frequencies. Identifying

There were once ~30,000 kākāpō (Strigops quita, like the kākāpō, may have survived
rich sources of adaptive polymorphism (9),
habroptilus) on mainland New Zealand, but there recent population declines because of effec-
representing conservation-relevant biologi-
are now only ~200 on nearby islands. High-quality tive long-term purging of deleterious muta-
cal adaptations. For example, some SVs are
reference genomes are aiding conservation breeding tions in the wild (8). Broadly, these findings
adaptations to certain diseases, and thus
programs of these critically endangered parrots. indicate that low heterozygosity in a popu-
selective breeding of individuals with these
lation will not always be detrimental if del-
variants could potentially enhance popula-
these and other signs of poor genomic eterious mutations have been purged and
tion resistance to environmental changes.
health can provide a warning that a species that introduction of individuals with higher
Developing and implementing conserva-
is becoming critically endangered or iden- heterozygosity but more deleterious alleles
tion management strategies often requires
tify specific populations that are fragile and into a population with less deleterious al-
delineation of populations or species to
deserving of focused conservation efforts. leles needs to be cautiously considered.
distinguish between subspecies and cryptic
Although draft short read–based reference Genomic diversity can also be measured
species and, consequently, to differentiate
genomes have been successfully used for by counting and comparing structural vari-
conservation strategies. This was originally
such analyses, third-generation reference ants (SVs), such as indels (insertions and de-
determined with DNA barcodes—that is,
genomes would lead to a comprehensive letions), inversions, chromosomal fusions,
small fragments of DNA that are divergent
identification of microsatellites, mutational copy-number variations, and transposable
between species, such as fragments of the
load, ROHs, and diverse segments of het- elements. Identification of SVs is more
organellular cytochrome oxidase subunit 1
erozygous variants (6) (see table S1). straightforward in third-generation assem-
and 16S and 12S ribosomal RNA genes—al-
Two examples of third-generation genome blies (4). SVs are increasingly appreciated as
though this is increasingly complemented
assemblies providing key informa- with analysis of whole-organelle se-
tion for conservation are those of quences. The long-read approaches
the kākāpō (Strigops habroptilus) Vertebrate extinction rate and human used in third-generation genome
and vaquita (Phocoena sinus), which assemblies simplifies the reconstruc-
are both critically endangered (4)
population growth tion of entire organelle genomes (4).
The extinction rate of vertebrates, calculated according to historical
(see table S2). The kākāpō is a par- These organelle genomes have re-
and current records of the International Union for Conservation
rot endemic to New Zealand whose of Nature (IUCN) animal extinction list, is shown. All vertebrates as vealed repeat regions and gene du-
population once comprised ~30,000 well as different species combinations are shown. The expected plications that were not assembled
individuals on the mainland. Human cumulative background extinction is based on geological extinction properly or were missed entirely
colonization circa 1360 CE and again estimates between the fifth mass extinction (~65 million years ago) in first- and second-generation ge-
in the 1800s reduced the population and 10,000 years ago. Graphs are modified from (2). nome assemblies (4). Additionally,
to 18 birds by 1977, but it is now recov- these organelle genome assemblies,
ering, with ~200 living on nearby is- 2.5 sometimes generated as a single-
Mammals Other vertebrates
lands. Analyses of a third-generation molecule sequence read, can be used
Extinctions per 10,000 species
Birds Background
kākāpō genome and second-genera- 2 for refined species delineation, phy-
tion resequenced genomes from 49 logeography, and population studies;
per 100 years
individuals representing both extant 1.5 they also reduce problems that arise
and historical populations revealed when, for example, mitochondrial
that the surviving island population 1 genomes are confused with nuclear
has had low genomic heterozygos- All vertebrates mitochondrial sequences (NUMTS).
ity in long ROHs for the past 10,000 0.5 Genetic rescue—which includes ge-
years, whereas the now-extinct main- netically informed translocations of a
land population did not (6). These species from one geographical region
0
findings affect conservation decision- Years
1501– 1601– 1701– 1801– 1901– to another, other breeding strategies,
making, whereby closely related in- 1600 1700 1800 1900 2014 and more extreme interventions such
dividuals can now be bred with less 8 as gene editing—aims to increase di-
Human population
concern for deleterious mutations, versity or prevent the fixation of del-

6
(billions)
allowing a small population an op- eterious alleles by facilitating gene

portunity to thrive. 4 flow from one population to another.
GRAPHIC: N. DESAI/SCIENCE; (HUMAN POPULATION DATA) ECOLOGY.COM
The vaquita is a small porpoise 2 Although applications of gene editing

endemic to the Gulf of California, have been mainly limited to agricul-
0
Mexico, and is, at present, the world’s ture, for example, to augment disease
most endangered marine mammal. resistance in crops (10), proposed
Fewer than 19 individuals survive Human population correlates with extinction applications to conservation include
There is a correlation between human population size and cumulative
today, which is a reduction from a extinction of all vertebrates in the past 500 years; each point is a 100-year
improving a species’ resistance to vi-
historical effective population size mean from the extinction data in the graph above. ral and bacterial infections or toxins
(Ne) of ~5000 (7) caused by bycatch 1.8 and a species’ capacity to adapt to
Cumulative extinctions of
in gillnets for shrimp and finfish anthropogenic and natural changes

IUCN vertebrates (%)
over the past century. Inferred his- 1.4 to their habitats, such as changes in
torical population analyses based on 1 temperature, salinity, or precipitation.
a third-generation vaquita genome R2 = 0.9992 Although these approaches are in
0.6
assembly revealed that the species p < 0.00001 the early stages of development and
has had low genomic heterozygosity 0.2 additional research is needed, third-
0
and a small Ne for the past ~250,000 0 1 2 3 4 5 6 7 8 9 generation genome assemblies may be
years (7). This suggests that the va- Human population (billions) critical to such efforts, for example, by

INSIGHTS | P E R S P E C T I V E S
better assembling translocations and having Addressing biodiversity loss is a complex PLANETARY SCIENCE
nearly all available sequences to determine problem that requires multifaceted solutions.
potential off-target sites of genome editing.
Another potential benefit of third-gener-
ation genome assemblies to conservation
Genomics can be an important component
of conservation management. It is urgent
that high-quality reference genome assem-
In the glare
will be for deextinction, such as resurrecting
extinct traits in a living species or proxies of
extinct species, for example, creating cold-
blies and cryopreserved cells be produced
for endangered species now and, eventually,
for all species. Waiting for technology im-
of the Sun
adapted elephants using genomic diversity provements, policy changes, or outcomes of Searches during twilight
that evolved in the woolly mammoth (11). nongenomic efforts places too many species toward the Sun have
One approach is cloning using somatic cell in peril. Generating high-quality genome as-
nuclear transfer (SCNT), whereby nuclei semblies from poorly preserved tissue or fos- found several asteroids
of cells from extinct sublineages are trans-
ferred into enucleated oocytes that are then
sil remains (DNA can be extracted from sam-
ples up to 1 million years old in permafrost)
near Venus’ orbit
transplanted into a female. Preliminary is impossible because of the short lengths of
reports indicate that this was successfully surviving DNA molecules, highlighting the By Scott S. Sheppard
done for the Przewalski’s horse and black- need for optimized cryopreservation of cells
A
footed ferret with decades-old cryobanked and tissues. When sex chromosomes exist, se- steroid surveys generally operate
cells, with the resulting clones still living quencing the heterogametic sex (e.g., males at night, mostly finding objects be-
in captivity (12). But this approach requires in mammals and females in birds) is prefer- yond Earth’s orbit. This creates a
preserved living cells, which limits its appli- able. Also, material should be preserved from blind spot because many near-Earth
cation, and it is also not straightforward for multiple individuals so that information objects (NEOs) could be lurking in
egg-laying species such as birds and fishes. about population genetic diversity can be ob- the sunlight interior to Earth’s or-
In these cases, gene editing of cells before tained. A notable and continuing challenge bit. New telescopic surveys are braving
SCNT or of early-stage embryos before egg lies with the ethical, legal, and moral implica- the Sun’s glare and searching for asteroids
formation might work better. But this ap- tions of translating genomic data to conser- toward the Sun during twilight. These sur-
proach requires knowledge of which edits vation. Coordination between scientists and veys have found many previously undiscov-
to make. Contiguous and nearly complete other stakeholders is important, especially ered asteroids interior to Earth, including
genomes provide greater resolution to iden- for access and benefit sharing of samples and the first asteroid with an orbit interior to
tify species-specific coding and regulatory the resulting digital sequence information Venus, 'Ayló'chaxnim (2020 AV2), and an
sequences for gene editing. with Indigenous Peoples and local communi- asteroid with the shortest-known orbital
In the absence of frozen cells, complete ties. Genome assemblies by themselves, even period around the Sun, 2021 PH27 (1, 2).
genome sequence data could also be used if complete and error free, cannot fully ad- NEOs are classified into different dy-
to create synthetic chromosomes and place dress the ongoing sixth mass extinction. But namical types (see the figure). Starting
them into viable cells, as was achieved by high-quality reference genome assemblies from the most distant are the Amors, which
the Yeast 2.0 Project, which synthesized the are advantageous for pre- and postconserva- approach Earth but do not cross Earth’s
entire genome of Saccharomyces cerevisiae tion management and monitoring with other orbit. Apollos cross Earth’s orbit but have
(13). Although yeast genomes are 3 to 10% strategies, such as preserving land, forest, semimajor axes greater than that of Earth.
of the size of vertebrate genomes and the and water reserves, and with other protec- Atens also cross Earth’s orbit but have
technology does not yet exist to synthesize tions to the environment. j semimajor axes less than that of Earth.
larger genomes, this highlights the poten- Atiras (also called Apohele) have orbits
REF ERENCES AND NOTES
tial power for synthetic biology in deextinc- completely interior to Earth, and Vatiras
1. C. N. Waters et al., Science 351, aad2622 (2016).
tion efforts. In multicellular organisms, 2. G. Ceballos et al., Sci. Adv. 1, e1400253 (2015). have orbits completely interior to Venus,
synthesized chromosomes could be placed 3. P. Brandies, E. Peel, C. J. Hogg, K. Belov, Genes 10, 846 with 2020 AV2 being the first known.
in enucleated oocytes of another species, (2019). NEOs have dynamically unstable orbits
4. A. Rhie et al., Nature 592, 737 (2021).
similar to the SCNT approach. 5. S. Nurk et al., Science 376, 44 (2022). of ~10 million years. A reservoir must ex-
As genomic analyses and synthetic biol- 6. N. Dussex et al., Cell Genom. 1, 100002 (2021). ist that replenishes the NEOs because their
ogy become components of conservation 7. P. A. Morin et al., Mol. Ecol. Resour. 21, 1008 (2021). numbers have been in a steady state over
8. J. A. Robinson et al., Science 376, 635 (2022).
management, there are challenges to over- 9. D. Weetman, L. S. Djogbenou, E. Lucas, Curr. Opin. Insect the past few billion years (3). Most NEOs are
come, including developing approaches Sci. 27, 82 (2018). likely dislodged objects from the main belt
that consider other complex genome or- 10. M. P. Phelps, L. W. Seeb, J. E. Seeb, Conserv. Biol. 34, 54 of asteroids between Mars and Jupiter (4–
(2020).
ganizations, such as species with germline 11. B. Shapiro, Genome Biol. 16, 228 (2015).
6). Physical observations show that NEOs
cells that have germline-specific chromo- 12. Revive & Restore, Projects (2021); https://reviverestore. are similar to main belt asteroids (MBAs),
somes (e.g., lamprey and songbirds), and org/projects/. with a small fraction being dormant comets
13. I. S. Pretorius, J. D. Boeke, FEMS Yeast Res. 18, foy032
rearranged chromosomes during different from the outer Solar System (7).
(2018).
developmental stages (e.g., single-cell cili- 14. C. M. Kinsella et al., Nat. Commun. 10, 5468 (2019). MBAs with orbital periods near whole
ate protists) (14, 15). Multicellular organ- 15. J. J. Smith et al., Nat. Genet. 50, 270 (2018). number ratios with Jupiter’s period are
isms also rely on microbial symbionts, depleted, which indicates that these areas
ACKNOWL EDGMENTS
some of which are inherited. High-quality are dynamically unstable. Small MBAs con-
S.P. and R.H.S.K. contributed equally to this work.
genome assemblies for symbiotic microbes, See supplementary materials for full acknowledgments. tinually move into these unstable regions
such as those being developed by the Earth
SUPPLEMENTARY MATERIALS
HoloGenome Initiative, are the crucial first
science.org/doi/10.1126/science.abm8127 Earth and Planets Laboratory, Carnegie Institution
step in incorporating this information into for Science, Washington, DC 20015, USA.
conservation management plans. 10.1126/science.abm8127 Email: [email protected]

and are ejected from the main from Yarkovsky drift, colli-
belt of asteroids through the Categorizing near-Earth objects sions, and thermal fracturing
Yarkovsky effect, which slowly The different types of asteroids orbiting close to the Sun are so close to the Sun. Fewer than
changes an asteroid’s orbit classified on the basis of what planetary orbits they cross. expected low-albedo, high-
through the nonisotropic emit- The 'Ayló'chaxnim asteroid is the first Vatira type that has been eccentricity NEOs with perihe-
ting of absorbed sunlight. The observed. In principle, asteroids that only exist inside Mercury’s lia very close to the Sun have
movement depends on the as- orbit are also possible (the Vulcanoids) but have not yet been observed. been found, likely because they
teroid’s rotation, size, albedo, fracture from the extreme ther-
and distance from the Sun. The mal stresses (13). These dust-
smaller an asteroid is and the producing events may be the
more sunlight it absorbs, the source of many meteor showers
larger its movement. Amor
Amor seen annually on Earth, like the
Fewer Atiras should exist Apollo
Apollo Geminids meteor shower that
than the more-distant NEOs, Aten
Ate
Aten occurs in mid-December from
and even fewer Vatiras, because Attiira
Ati
A
Atirara the shedding of material off the
it becomes harder and harder NEO Phaethon (14).
Vat
Vatira
atira
at ira
for an object to move inward From NEO formation models
past Earth’s and then Venus’ and the current NEO survey
orbit. Random walks of a NEO’s Vulcanoids?
canoid
can oid efficiencies, more than 90% of
orbit through planetary gravita- planet-killer NEOs have been
tional interactions can make an Sun found (those larger than 1 km),
Aten into an Atira and/or Vatira Mercury
Mer
Mercur
curyy
cur although only about half of the
orbit and vice versa. Atiras city-killer NEOs are known
should make up some 1.2% and Venus
Venus (those larger than 140 meters).
Vatiras only 0.3% of the total Earth
Earth The last few unknown 1-km
NEO population coming from NEOs likely have orbits close
the main belt of asteroids (4). to the Sun or high inclinations,
2020 AV2 itself will spend only Mars
Mars which keep them away from
a few million years in a Vatira the fields of the main NEO
orbit before crossing Venus’ or- surveys. The 48-inch Zwicky
Main belt asteroids
bit. Eventually, 2020 AV2 will Transient Facility telescope
either collide with or be tidally has found one Vatira and sev-
disrupted by one of the planets, eral Atira asteroids, making it
disintegrate near the Sun, or be ejected of any known object in our Solar System, in- one of the most prolific asteroid hunters
from the inner Solar System. cluding Mercury. 2021 PH27’s surface likely interior to Earth. To combat twilight to
Recent NEO models predict that there gets to 500°C, which is hot enough to melt find smaller asteroids, one can use a big-
should be less than one Vatira of the roughly lead. 2021 PH27 is also apparently ~1 km ger telescope. Large telescopes usually do
1.5-km diameter of 2020 AV2 but many in size, which is relatively large. However, not have big fields of view to efficiently
more smaller ones (4). Only a fraction of because the diameter of these interior aster- survey. The National Science Foundation’s
the sky has been searched where Vatira-like oids is calculated with an assumed albedo Blanco 4-meter telescope in Chile with the
asteroids reside; however, because of the and solar phase function, the actual diam- Dark Energy Camera (DECam) is an excep-
scattered light problem from the Sun, only eters for both of these discoveries could be tion. A new search for asteroids hidden in
the largest are observable. Finding a rela- under 1 km (10). This would put them in a plain twilight with DECam has found a few
tively large Vatira in the little area searched more-expected population and make them Atira asteroids, including 2021 PH27. These
is somewhat unexpected, but small number less of a statistical fluke. continuing twilight surveys are finally un-
statistics has caveats when trying to under- Some Atiras and Vatiras could also have covering the population of small asteroids
stand a whole population. Only a few aster- another source region besides the main near the orbit of Venus. j
oid surveys have imaged interior to Venus belt of asteroids. These could be relatively
with published results (8, 9), but the null stable inner reservoirs, such as objects in
1. B. T. Bolin et al., MPEC 2020-A99 (2020).
results of others may be unpublished, mak- stable long-term resonances with Venus 2. S. S. Sheppard et al., MPEC 2021-Q41 (2021).
ing it hard to determine how much space or Mercury, or even the hypothetical 3 R. G. Strom et al., Res. Astron. Astrophys. 15, 407 (2015).
interior to Venus has actually been well Vulcanoids (8). Vulcanoids are asteroids 4. M. Granvik et al., Icarus 312, 181 (2018).
5. A. Morbidelli et al., Icarus 340, 113631 (2020).
searched. This makes it difficult to get a that could exist with orbits completely inte- 6. A. W. Harris, P. W. Chodas, Icarus 365, 114452 (2021).
true handle on Vatira discovery statistics. rior to Mercury’s orbit that could be stable 7. R. P. Binzel et al., Icarus 324, 41 (2019).
Recently, the asteroid with the smallest- for billions of years, possibly forming there. 8. P. Pokorný, M. J. Kuchner, S. S. Sheppard, Planet. Sci. J.
1, 47 (2020).
known semimajor axis at 0.46 astronomi- Many exoplanets have been found closer to 9. Q. Ye et al., Astron. J. 159, 70 (2020).
cal units (au) was found—2021 PH27 (2). their host stars than the Vulcanoid region 10. A. Mainzer et al., Astrophys. J. 752, 110 (2012).
Because of 2021 PH27’s large eccentricity of in our Solar System. Vulcanoids could also 11. S. Greenstreet, H. Ngo, B. Gladman, Icarus 217, 355
GRAPHIC: N. CARY/SCIENCE
(2012).
0.7, its orbit actually crosses both the orbits come from a random walk from the NEO 12. A. J. Steffl, N. J. Cunningham, A. B. Shinn, D. D. Durda,
of Mercury and Venus, making it an Atira population, but this would be very rare S. A. Stern, Icarus 223, 48 (2013).
and not a Vatira asteroid. 2021 PH27 ap- (11). Spacecraft observations of the near- 13. Q. Ye, M. Granvik, Astrophys. J. 873, 104 (2019).
14. D. Jewitt, J. Li, J. Agarwal, Astrophys. J. Lett. 771, L36
proaches so close to the Sun (0.13 au) that it Sun environment likely rule out Vulcanoids
(2013).
has the strongest general relativity effects, larger than ~5 km (12). Vulcanoids could
at almost 1 arc min precession per century, be destabilized over long periods of time 10.1126/science.abj9820

LASERS tal, Lyubarov et al. provide detailed classical

and quantum models for an atom placed in-
To make a mirrorless laser side such a crystal. When stimulated with a

flash of light, an atom inside the medium re-
mains essentially in a so-called “transparent”
Periodic temporal modulation of a photonic crystal state in which there are an equal number of
electrons in the ground and excited states. In
can be used to produce laser light this state of balance, the stimulation causes
the atom to absorb and emit equal amounts
By Daniele Faccio1 and Ewan M. Wright2 periodically and strategically (4). A com- of light. The periodic modulation can then
mon feature of temporally periodic systems lead to exponential amplification for the
I
nside any laser is a cavity with a “gain is a typical resonance frequency at which emitted light with a narrowed spectrum that
medium” that gives the laser its energy the energy transfers from the time-varying is characteristic of a laser beam. In this view,
to emit light. A typical gain medium parameter—for example, the child periodi- the “transparent” atom acts as a conduit for
contains atoms that can be excited by cally and strategically shifting their center of energy transfer between the periodic modu-
using an external energy source and is mass twice per period, first by bending their lation of the medium and the emitted light.
sandwiched between a pair of mirrors. legs backward as they swing backward and Moreover, this behavior does not appear to
The mirrors impose a periodicity on the light then later extending their legs forward as depend on the specific initial excitation of the
inside the cavity—similar to how the length they swing forward. In technical terms, the atom. The initial stimulation with a flash of
of a guitar string limits what musical notes greatest amplification in energy for any pe- light can be at a substantially different fre-
can be played—and allows the medium to riodic system occurs for light waves with a quency from the bandgap frequency as long
pack more energy into the light each time it frequency equal to twice the parameter mod- as the light is emitted across a broad range of
passes through the gain medium. On page ulation frequency. frequencies. Eventually, the exponential am-
425 of this issue, Lyubarov et al. (1) propose Researchers have been investigating how plification at the bandgap will take over and
a radically new approach to making a laser to use temporally modulated materials. For pin the system to emission at the resonant
in which the cavity is replaced by a medium example, can such materials control the fre- frequency at half the modulation frequency.
with no mirrors. Instead, the optical proper- quency of light or can magnet-free materials Although the periodic time crystal laser
ties of the medium are periodically modu- made of nonreciprocal elements be created in does not rely on cavity mirrors or a gain me-
lated in time. which light can only propagate in one direc- dium, it does rely on a modulation of the me-
The laser device of Lyubarov et al. contains dium that needs to be extremely fast because
no mechanism for recirculating the light at
all. Its operation relies on a slab of trans-
“...the ‘transparent’ atom of the resonance condition, with the expo-
nential amplification rate depending on the
parent material with a refractive index that
varies periodically in time. Because the wave-
acts as a conduit amplitude of the modulation. Typical pho-
tonic materials exhibit small modulations
length of light in a medium varies inversely for energy transfer...” of the refractive index at the femtosecond or
with the refractive index—the shorter the picosecond time scales required for lasers at
wavelength, the higher the effective refractive tion (5)? One can draw on analogies between visible to terahertz wavelengths. Recent prog-
index—the medium modulation produces an the spatial and temporal cases of photon ress in so-called epsilon-near-zero or index-
effect that is akin to periodically compressing modulation in crystals to understand the near-zero materials offers a possibility for
light waves. Lyubarov et al. take advantage physics of periodic time crystals. Spatial pho- ultrafast switching of the medium with near-
of this periodic temporal compression and tonic crystals are crystal-like materials with unity refractive index modulation (7, 8), but
show that it can be used to amplify light that periodic structures that modulate the propa- this typically also has large losses that may
will also be coherent, as is laser light. gation of light (6). These crystals behave for make it harder to achieve laser-like behavior.
Time-modulated systems and amplifica- light in a way similar to what atomic crystals The mechanism presented by Lyubarov et
tion of light from temporal modulation are do for electrons, in that they lead to the for- al. may also be applied for producing light at
not completely new. Although different re- mation of periodic bandgap structures—for- much longer wavelengths by transducing dif-
search fields may trace the origins of these bidden “gaps” in the frequency range where ferent forms of energy into electromagnetic
ideas back to different sources, they all con- the propagation of waves is strongly sup- radiation. For example, a periodic temporal
nect to a series of ideas proposed in the pressed. This suppression of waves occurs mechanical modulation in the form of a pe-
mid-20th century. In 1970, physicist Gerald when the wave vector of the light is equal to riodic pressure applied to the medium could
Moore explained how a temporally modu- half of the periodicity of the modulation and be used to amplify electromagnetic waves,
lated yet otherwise empty cavity can lead to can be used to confine light, similar to the akin to the original proposal by Moore, albeit
the creation of photons (2). This is commonly mirrors of a standard laser. not with a cavity but through a photonic time
referred to as the dynamical Casimir effect, To observe the formation of “gaps” with a crystal—more than half a century after the
which was not experimentally verified in a temporal modulation, one may use a block idea was first proposed. j
superconducting circuit until 2011 (3). In gen- of material that can change its refractive REF ERENCES AND NOTES
eral, any system that has a time-dependent index with the right periodicity. For such a 1. M. Lyubarov et al., Science 377, 425 (2022).
parameter can exhibit some form of amplifi- system, one can expect a bandgap where the 2. G. T. Moore, J. Math. Phys. 11, 2679 (1970).
3. C. M. Wilson et al., Nature 479, 376 (2011).
cation, similar to how a child can increase the frequency is equal to half of the temporal 4. P. D. Nation et al., Rev. Mod. Phys. 84, 1 (2012).
amplitude of a swing by shifting their weight modulation frequency of the material. When 5. E. Galiffi et al., Adv. Photonics 4, 014002 (2022).
this happens, the energy of the system is no 6. J. D. Joannopoulos et al., Molding the Flow of Light
1
School of Physics and Astronomy, University of (Princeton Univ. Press, ed. 2, 2008).
longer conserved, which allows its energy to 7. M. Z. Alam, I. De Leon, R. W. Boyd, Science 352, 795 (2016).
Glasgow, Glasgow, UK. 2Wyant College of Optical Sciences,
University of Arizona, Tucson, AZ 85721, USA. be amplified. Although this was known for a 8. L. Caspani et al., Phys. Rev. Lett. 116, 233901 (2016).
Email: [email protected]; [email protected] wave propagating inside a periodic time crys- 10.1126/science.abq5012

PLASTICS
Improving catalysis by moving water

The conversion of gases into building blocks for synthesizing plastics is enhanced
By Mingyue Ding and Yanfei Xu hydrocarbon products by adjusting zeolite librium on the CoMn surface through its hy-
pore size. A superior selectivity of 80% for drophobicity. To support this interpretation,
L
ight olefins, which include commer- light olefins with 17% CO conversion was Fang et al. examined CoMn during a reaction
cially important chemicals such as achieved over a mixture of partially reduced in which water is added to the syngas feed.
ethylene, propylene, and butylene, are oxide and mesoporous silicoaluminophos- After the deliberate injection of water, the
used to fabricate a wide range of plas- phate zeolite (ZnCrOx/MSAPO) (4). CO conversion over CoMn decreased from 34
tics and synthetic fibers (1). Because Water is a by-product of the CO hydroge- to ~8%, whereas CO conversion over CoMn/
they are typically produced from petro- nation reaction and can limit the efficiency PDVB only decreased slightly, from 64 to
leum, their production is intrinsically linked of syngas to light olefins by covering active ~55%. Further pulse experiments and in situ
to problems stemming from petroleum ex- sites or inducing undesirable reactions. Fang diffuse reflectance infrared Fourier trans-
traction and processing. Researchers have et al. report a catalyst that is a mixture of form spectroscopy revealed the hindrance
been exploring alternative routes for the a CoMn catalyst with polydivinylbenzene of CO adsorption by competition with water
synthesis of light olefins and have found (PDVB). The CoMn/PDVB mixture possesses on the CoMn surface. The addition of hydro-
success in making light olefins from syngas, the advantages of the CoMn catalyst—its high phobic PDVB to CoMn efficiently shifts the
a mixture of carbon monoxide sorption equilibrium of water,
(CO) and hydrogen (H2) that can thereby reducing the negative
be derived from not only coal Reducing the negative effect of water effect of water on the adsorption
and natural gas but also non– and conversion of CO molecules.
fossil fuel–based biomass (2–4).
on light olefin catalysis Fang et al. also performed
The competitive adsorption of water (H2O) and carbon monoxide (CO)
Although considerable progress theoretical simulations to study
on a cobalt manganese (CoMn) catalyst limits syngas conversion (left).
has been made in syngas con- Hydrophobic polydivinylbenzene (PDVB) acts as water-conduction the effect of channel wetta-
version, the ability to produce channels and accelerates the diffusion of water, thereby exposing more active bility on water diffusion. The
light olefins from syngas re- sites on the CoMn catalyst for converting syngas to light olefins (right). hydrophilic channel interacts
mains limited. On page 406 of with water molecules and slows
this issue, Fang et al. (5) present down their diffusion, whereas
a simple and effective method to the weak interaction between
boost the conversion of syngas H2 the hydrophobic channel and
to light olefins. water molecules accelerates the
During syngas conversion, Light escape of water. The simulation
CO H2O olefins
the carbon-carbon coupling results suggest that even though
that is catalyzed by conven- the water-adsorbed region of the
tional metals or carbide-based CoMn catalyst and the channels
metals follows a polymerization are separated from each other,
step in which the carbon chain more water molecules escape
grows without control. This from the hydrophobic channel
produces hydrocarbon products than the hydrophilic one.
with a wide range of carbon Mixing a hydrophobic pro-
numbers, which means a poor moter with a CoMn catalyst is
selectivity for light olefins (6). CoMn PDVB a simple but effective strategy
Designing a catalyst with excel- to enhance the conversion of
lent selectivity for light olefins is challeng- selectivity for light olefins and mild reaction syngas to light olefins. Fang et al. provide a
ing. An iron-based catalyst, in which iron is conditions—whereas the PDVB acts as hydro- method to enhance the conversion efficiency
supported on aluminum oxide (Fe/a-Al2O3) phobic water-conduction channels to move without influencing the selectivity for target
exhibited 53% selectivity for light olefins water away from the CoMn catalyst (see the products. The improved efficiency will re-
with 80% CO conversion (2). Cobalt manga- figure). This combination leads to a substan- duce the manufacturing cost that is incurred
nese (CoMn) as a catalyst showed a slightly tial increase in CO conversion, at 64%, and by the undesirable but necessary repeated
better 61% selectivity for light olefins with a good selectivity for light olefins in hydro- reaction of unreacted syngas during produc-
32% CO conversion (3). In the other cata- carbon products, at 71%, achieved under mild tion. The use of a hydrophobic promoter to
lyst system, called the oxide-zeolite route, reaction conditions of 250°C and 0.1 MPa. accelerate water escape and thereby expose
the carbon-carbon coupling proceeds in After removing the PDVB in the CoMn/ more active sites on the catalyst for reactants
GRAPHIC: N. DESAI/SCIENCE
the confined zeolite micropore, and thus PDVB mixture after use, the remaining CoMn may be applicable in other water-restricted
it is easy to control the carbon number of exhibited catalytic activity similar to that of reactions, such as carbon dioxide (CO2) cata-
fresh CoMn. This observation suggests that lytic hydrogenation, which can be used to
hydrophobic PDVB does not change the produce fossil fuel–free fuel.
School of Power and Mechanical Engineering, The Institute
of Technological Sciences, Wuhan University, Wuhan structure of CoMn during the reaction but Before the findings of Fang et al., there
430072, China. Email: [email protected] instead influences the water-sorption equi- had been other hydrophobization strategies,
22 JULY 2022 • VOL 377 ISSUE 6604 369

such as chemically modifying the catalyst PLANT BIOLOGY

surface (7), which reduces CO2 selectivity by
suppressing the water-gas shift reaction (i.e.,
CO + H2O CO2 + H2). Both the mixing and
the chemically modifying strategies enhance
The quest for more food
the conversion of CO to hydrocarbon prod- Rice yield is increased by boosting nitrogen
ucts, but the two approaches differ because
of the difference in the distance between the
uptake and photosynthesis
hydrophobic entity and the catalytic active
sites in each case. By Steven Kelly (6). Evolution by natural selection took half
For the chemically modifying route, the a billion years to craft this natural diversity
E
hydrophobic compound coats the catalytic nhancing photosynthesis is regarded and distribute it over the ~104 million km2
active sites, inhibiting the readsorption of as one of the most promising av- of habitable land. However, in just the past
water on the catalyst. In this case, most of enues for increasing crop yield (1). 5000 years, humanity has replaced ~50%
the water produced during the reaction can Accordingly, substantial focus has of this wilderness with agriculture (7) (see
escape without participating in the water-gas been drawn to this challenge, with the figure). This rapid destruction of the
shift reaction. The water-gas shift reaction several breakthroughs holding great natural world is causing the sixth global
deviates away from equilibrium, giving a low potential (2). A common feature of these mass extinction (8), with current rates of
CO2 selectivity of 13% with a CO conversion successes is that they have targeted meta- extinction unseen since an asteroid wiped
rate of 56%. For the mixing route, the CO2 bolic processes, altering the rate and/or the out the dinosaurs (8). Moreover, the con-
selectivity over CoMn/PDVB is near 50%, im- path of metabolite flow through the plant version of complex ecosystems into agri-
plying that the water-gas shift reaction is near to achieve higher rates of photosynthesis. cultural systems has released billions of
equilibrium. The hydrophobic compound However, on page 386 of this issue, Wei et tons of CO2 into the atmosphere, accelerat-
and the catalytic active sites are separated al. (3) report an alternative approach. They ing climate change, reducing the capacity
from each other, and the distance between show that photosynthesis and yield can be to store CO2, and placing further pressure
them is nanometers to microns. In this case, improved in rice by overexpressing a tran- on the natural world (9)—all in the quest
most of the water produced on CoMn partici- scriptional regulator that promotes the for food.
pates in the water-gas shift reaction before expression of yield-associated genes. This Although the main way in which hu-
diffusing to the hydrophobic promoter. Fang highlights that there is a substantial latent manity has increased food production has
et al. report that the CoMn catalyst is water capacity for enhancing photosynthesis hid- been through the expansion of agricultural
sensitive, and the small amount of water that den in the genomes of plants. Moreover, land, this expansion has been mitigated by
is not involved in the water-gas shift reaction this latent capacity is present in abun- scientific technological development. For
can hinder the adsorption of CO on CoMn dance even in plants subjected to thou- example, plant breeders have improved
and inhibit CO conversion. The addition of sands of years of improvement through the shape and form of domesticated plants
hydrophobic PDVB accelerates the diffusion plant breeding. so that their canopy intercepts almost all
of this unused water, exposing more active Planet Earth is home to more than 6000 of the light before it hits the ground and
sites on the CoMn catalyst for the adsorption species of mammal (4), 300,000 species of so that the largest possible fraction of the
and conversion of CO. plant (5), and 5,000,000 species of insect carbon captured over the lifetime of the
Decreasing the CO2 release during syngas
conversion can reduce costs and enhance
the productivity of the target product (8, 9). Changing landscapes
As countries propose carbon-neutral goals to The proportion of total habitable land that is wild versus farmed has decreased over time (7) (top left).
address climate change, carbon emission re- Additionally, the amount of people fed per hectare has increased, owing to technological improvements
duction in the chemical industry is impera- (bottom left). However, this is not enough to feed the human population, so further yield gains are
tive. Combining the hydrophobic strategies needed. Photosynthetic rate is highly variable in different plant species (11) (right), indicating that
of mixing or chemical modification to de- improvements to photosynthesis, and thus yield, are possible.
velop new catalysts with low CO2 selectivity
Wild Farmed
and high CO conversion will unlock the full 100 45
Habitable land (%)
Maximum photosynthetic rate (µmol CO2 m−2s−1)
potential of this process to produce valuable

chemicals from syngas. j 75
REFERENCES AND NOTES 50

1. X. Pan, F. Jiao, D. Miao, X. Bao, Chem. Rev. 121, 6588
(2021). 25 30
2. H. M. Torres Galvis et al., Science 335, 835 (2012).
3. L. Zhong et al., Nature 538, 84 (2016). 0
4. F. Jiao et al., Science 351, 1065 (2016). 5000 4000 3000 2000 1000 0
5. W. Fang et al., Science 377, 406 (2022).
People fed per hectare
1.2
6. W. Zhou et al., Chem. Soc. Rev. 48, 3193 (2019).
7. Y. Xu et al., Science 371, 610 (2021). GRAPHIC: K. FRANKLIN/SCIENCE
8. P. Wang et al., Sci. Adv. 4, eaau2947 (2018). 0.8 15
9. Y. Xu, X. Li, M. Ding, Chem 7, 1977 (2021).
0.4
ACKNOWLEDGMENTS
This work was supported by the National Natural Science
Foundation of China (21978225 and U21A20317). 0
5000 4000 3000 2000 1000 0 0
10.1126/science.adc9414 Years before present Flowering plants

plant ends up in the grain (1). These mas- OsDREB1C in wheat and the model plant MEDICINE
terworks of plant engineering, coupled Arabidopsis thaliana similarly improved
with a near-limitless supply of fertilizer
from the Haber-Bosch process (10), have
enabled humanity to feed more people
yield (or biomass) by more than 10%.
Pinning down the mechanism that gave
rise to this yield increase was challenging
One step closer
per unit of land area than ever before (7).
Today, it takes only ~30% as much land to
feed one person as it did 5000 years ago.
because there were thousands of genes
whose expression changed in response
to the overexpression of OsDREB1C.
to cancer
The problem is that there are 138 times
as many people alive today as there were
5000 years ago. Thus, this improvement in
However, chromatin immunoprecipitation
coupled to RNA sequencing revealed that
key nitrogen assimilation enzymes and
nanomedicine
land use is not enough. transporters were direct downstream tar- High-throughput tool
Because photosynthesis provides all gets of OsDREB1C. This enabled plants to uncovers links between
of the carbon and energy that plants use take up more nitrogen from their environ-
to grow, enhancing photosynthesis holds ment and support a higher rate of photo- cell signaling and
great potential to increase crop growth and synthesis, faster growth, and more yield. nanomaterial uptake
yield. Notably, there is extensive natural di- Ultimately, this work revealed that there
versity in the rates of photosynthesis be- is a gold mine of potential for enhancing
tween plants (11). Some plants have evolved growth and yield that can be uncovered by By Jessica O. Winter1,2
high rates of photosynthe- learning how plants respond
“Our ancestors did
T
sis to optimize growth and to their environment. he promise of chemotherapeutic
reproduction in the short- If there is to be both a sus- nanomedicine has tantalized cli-
est time possible. However,
some plants barely grow at
not choose plants tainable future for humanity
and more space for wildlife,
nicians and patients for decades.
Nanoparticles (NPs) can directly tar-
all, only reproducing once a to domesticate then complex issues, such as get tumor cells, which would reduce
century or less. Our ancestors reducing food waste and al- the amount of chemotherapy admin-
did not choose plants to do- based on how tering the relative use of ani- istered and its systemic toxicity, increasing
mesticate based on how good
they were at photosynthesis.
good they were at mal and plant proteins, need
to be addressed. However,
patient quality of life and extending utility
of therapies with lifetime dosing limits.
They domesticated plants be- photosynthesis.” even if these challenging However, these hopes remain largely unre-
cause they tasted good, did goals can be achieved, avert- alized. Liposomal drug carriers, which make
not poison them (mostly), and provided ing the sixth global mass extinction will up nearly all clinically approved nanomedi-
a reliable source of easy-access nutrition. require producing a larger quantity of cines, have not extended overall patient
By chance, some of these plants, such as food than has ever been produced in hu- survival compared with treatment with the
Zea mays (maize), are extremely efficient man history, on much less land than is drugs alone (1). These failures have been at-
at photosynthesis. However others, such being used today. For this reason, increas- tributed to poor delivery to target cells (2)
as rice, are just average. This disparity has ing photosynthesis, and consequently the because NPs must first traverse a series of
inspired researchers to ask how photosyn- amount of food that can be produced per biological barriers (3). Although nanocar-
thesis can be improved in average plants unit of land area, is one of the most impor- rier composition, surface chemistry, size,
like rice. Success has come from thinking tant objectives for the sustainable future of and shape have been optimized to promote
carefully about how photosynthesis works the planet. j cell entry, progress has been confounded by
and what its limitations and bottlenecks heterogeneity in cell uptake signaling (4).
are. Examples of these successes include On page 384 of this issue, Boehnke et al. (5)
1. S. P. Long, A. Marshall-Colon, X.-G. Zhu, Cell 161, 56
speeding up plant responses to fluctuat- (2015). uncover the reciprocal relationship between
ing light (12), the creation of metabolic 2. A. J. Simkin, P. E. López-Calcagno, C. A. Raines, J. Exp. NP material properties and cell internaliza-
bypasses to minimize energy losses (13), Bot. 70, 1119 (2019). tion using nanoPRISM, a high-throughput
and enhancing the capacity of the engine 3. S. Wei et al., Science 377, 386 (2022). screening approach.
4. C. J. Burgin, J. P. Colella, P. L. Kahn, N. S. Upham, J.
of photosynthesis—the chloroplast (14). Mammal. 99, 1 (2018).
The nanoPRISM technology uses the pro-
Wei et al. found a new way to improve 5. M. J. M. Christenhusz, J. W. Byng, Phytotaxa 261, 201 filing relative inhibition simultaneously in
photosynthesis. They discovered it by ask- (2016). mixtures (PRISM) (6) method to generate a
ing, what does rice do when challenged 6. N. E. Stork, Annu. Rev. Entomol. 63, 31 (2018). screening library of ~500 cancer cell lines that
7. K. Klein Goldewijk, A. Beusen, J. Doelman, E. Stehfest,
with stressful growth conditions? They are barcoded with distinct DNA sequences
Earth Syst. Sci. Data 9, 927 (2017).
noticed that expression of the OsDREB1C 8. R. H. Cowie, P. Bouchet, B. Fontaine, Biol. Rev. 97, 640 that permit identification of cells with high-
transcription factor was up-regulated (2022). throughput genomic sequencing. This cell li-
when rice plants were grown under low- 9. S. R. Weiskopf et al., Sci. Total Environ. 733, 137782 brary is combined with a panel of 35 different
(2020).
nitrogen conditions. When they overex- fluorescently labeled NPs with varying core
10. V. Smil, Ambio 31, 126 (2002).
pressed OsDREB1C in rice, they increased 11. J. Gago et al., Trends Plant Sci. 24, 947 (2019). compositions, surface chemistries, and diam-
nitrogen uptake and boosted photosyn- 12. J. Kromdijk et al., Science 354, 857 (2016). eters to identify synergistic interactions for
thesis, the net result being a 12 to 40% in- 13. R. Kebeish et al., Nat. Biotechnol. 25, 593 (2007). cell uptake. PRISM-tagged cells are separated
crease in yield in rice plants grown in the 14. X. Li et al., Commun. Biol. 3, 151 (2020).
field. They also found that overexpressing ACKNOWL EDGMENTS
1
William G. Lowrie Department of Chemical and
Biomolecular Engineering, Ohio State University,
S.K. is a cofounder of Wild Bioscience LTD and provides Columbus, OH, USA. 2Department of Biomedical
Department of Plant Sciences, University of Oxford, Oxford, consultancy to the company. Engineering, Ohio State University, Columbus, OH, USA.
UK. Email: [email protected] 10.1126/science.add3882 Email: [email protected]
SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 37 1

into four groups according to uptake Signatures of cellular uptake nanomedicine approval based on
level, and their DNA is sequenced to The nanoPRISM method combines cell and nanomaterial libraries similarity to an existing product (12).
identify them and screen for key driv- to identify signatures associated with cellular internalization. Given the long timeline for drug de-
ers of NP internalization that can be The ABC and SLC protein families regulate uptake of lipid-based and velopment, which can span a decade
attributed to either NP characteristics polymer nanoparticles differentially, whereas vesicular trafficking, or more, technologies to safely accel-
or cell signaling. ECM, and focal adhesion pathways affected all types of nanoparticles. erate this process are desirable.
Boehnke et al. compared the uptake Core composition, not surface chemistry, was the strongest regulator The nanoPRISM method repre-
efficiency of NPs conjugated to anti- of uptake behavior. sents a substantial advance over the
bodies targeting epidermal growth less rigorous and qualitative studies
factor receptor (EGFR) versus EGFR Solid lipid of NP internalization that charac-
nanoparticle
antibodies alone in cell lines that over- terized the early years of the field.
express this receptor. NanoPRISM re- Studies that examined a few NP
vealed differences in cellular uptake, properties in a single cell line could
most likely resulting from the steric Nanomaterials library DNA-barcoded
DNA
DNA- barcoded
ded can
cancer
ncer not capture the complexities of NP
celll library
l (P
(PRISM
(PRISM)M)
hindrance of NP conjugation. These cell entry. Combined with machine
results suggest that nanoPRISM may learning and iterative simulation
Polymer
be suitable for evaluating antibody- nanoparticle and materials synthesis approaches,
drug conjugates (ADCs), a growing nanoPRISM nanoPRISM could enable screening
therapeutic category. Liposome for nanomaterials that target spe-
ABC/MDR
Boehnke et al. also use nano- ABCA1 cific cell types, similar to current
PRISM to interrogate NPs with com- biopanning methods for peptides or
positions most commonly applied to Vesicle the systematic evolution of ligands
nanomedicine: spherical liposomes trafficking by exponential enrichment (SELEX)
made of lipid bilayers and solid lipid method of aptamer discovery (13).
and polymer NPs consisting of dis- Although the study of Boehnke et al.
ordered, spherical lipid or polymer Lysosome examines only 35 different NPs, addi-
aggregates. They also examine NPs SLC46A3 ECM tional nanomaterials could be added
Focal
with or without polyethylene glycol adhesions ons to the library, such as inorganic NPs
(PEG) modification, which is used to (such as gold, silica, and carbon) and
reduce systemic uptake and improve ABC, adenosine triphosphate (ATP)-binding cassette; ECM, materials with complex geometries
extracellular matrix; MDR, multidrug resistance; SLC, solute carrier.
circulation time (7). They find that (such as DNA origamis). A limitation
NP core composition is a primary of nanoPRISM is its focus on cellu-
determinant in cellular uptake. This un- lysosomal transmembrane protein linked lar entry, the last step of the biodistribution
expected finding upends years of work on to lipid catabolism (10) that influences ly- process. However, it is easy to envision ex-
modulating NP surface chemistries to alter sosomal trafficking of ADCs (11). Expression panding this approach beyond cell uptake
protein adsorption patterns and subsequent of SLC46A3 negatively regulated liposo- to study the relationship between NP mate-
cell adhesion (8). Although cells first detect mal and solid lipid NP cellular uptake, rial properties and gene expression in cell
NPs through their surface chemistry, the whereas polymer NPs that lack lipids were adhesion and trafficking. Additionally, with
findings of Boehnke et al. support early unaffected. SLC46A3 association with lipid- the template provided by Boehnke et al.,
studies that showed that NP stiffness and based NPs was evidenced even when NP similar methods could be integrated with
deformability, which are dictated by core surfaces were coated with nonlipid mole- microfluidics, organ-on-a-chip, or tumor
composition, are stronger modulators of the cules. This further indicates the importance organoid cultures to model other delivery
uptake process (9). of NP core composition in cellular uptake barriers, such as circulation, extravasation,
The power of the nanoPRISM method is processes and also suggests that cells can and tissue diffusion. Thus, the nanoPRISM
further illustrated by combining these find- detect core composition through surface approach could catalyze rapid materials op-
ings with the Cancer Cell Line Encyclopedia, coatings, which better resemble a porous timization, accelerating nanocarrier design
which quantifies mutational genomic signa- net than a wall. This could have important and bringing the promise of cancer nano-
tures of common cancer cell lines. Boehnke implications for predicting the efficacy of medicine closer to reality. j
et al. identify genomic signatures and sig- nucleic acid vaccines and therapies that
naling networks most correlated with NP use lipid-based carriers, such as COVID-19
1. G. H. Petersen, S. K. Alzghari, W. Chee, S. S. Sankari, N. M.
internalization. Many of the results are not mRNA vaccines. For example, SLC46A3 La-Beck, J. Control. Release 232, 255 (2016).
surprising, such as involvement of the sol- biomarker testing could be implemented to 2. S. Wilhelm et al., Nat. Rev. Mater. 1, 16014 (2016).
ute carrier (SLC) transporter or adenosine identify patients most likely to respond to 3. S. Barua, S. Mitragotri, Nano Today 9, 223 (2014).
4. B. D. Chithrani, A. A. Ghazani, W. C. W. Chan, Nano Lett. 6,
triphosphate (ATP)–binding cassette (ABC) lipid-based nanotherapeutics. 662 (2006).
families, which have previously been impli- The results of the nanoPRISM screens 5. N. Boehnke et al., Science 377, eabm5551 (2022).
cated in NP cellular entry and transport. The are also confirmed in animal models, in- 6. C. Yu et al., Nat. Biotechnol. 34, 419 (2016).
7. M. Eugene, Cell. Mol. Biol. 50, 209 (2004).
nanoPRISM screens also highlight gene net- dicating that this technique could be used
GRAPHIC: V. ALTOUNIAN/SCIENCE
8. A. Albanese et al., ACS Nano 8, 5515 (2014).
works associated with the plasma membrane to identify the most promising formula- 9. X. Sun et al., Biomacromolecules 6, 2541 (2005).
and extracellular matrix that contribute to tions for downstream analysis, reducing 10. J.-H. Kim et al., Nat. Commun. 12, 290 (2021).
11. K. J. Hamblett et al., Cancer Res. 75, 5329 (2015).
NP cellular entry processes (see the figure). preclinical animal testing demands. Such 12. S. Soares, J. Sousa, A. Pais, C. Vitorino, Front Chem. 6,
However, the nanoPRISM method also high-throughput approaches are critical to 360 (2018).
reveals involvement of an understudied the rapid advancement of cancer nanomed- 13. C. Tuerk, L. Gold, Science 249, 505 (1990).
gene that has not been associated with NP icine, because US and European regulatory
internalization: SLC46A3. This encodes a agencies have not established criteria for 10.1126/science.add3666

while Upper Basin development lagged;
P OLICY FORUM hence, this apportionment sought a degree
of future equality among the basins. The
WATER GOVERNANCE Compact also required the Upper Basin not
to deplete the river’s flow to less than 75
What will it take to stabilize MAF during any 10 consecutive years (the
“non-depletion obligation”) and required
each basin to equally share any obligations
the Colorado River? to Mexico. The 1944 Treaty established a de-

livery requirement of at least 1.5 MAF/year.
The distinction between the Upper and
A continuation of the current 23-year-long drought will Lower Basin created an institutional divi-
require difficult decisions to prevent further decline sion that endures today. Lake Mead is of-
ten perceived as the water supply for the
Lower Basin, and Lake Powell is primarily
By Kevin G. Wheeler1,2, Brad Udall3, hydropower. Storage downstream in Lake managed to avoid violation of the non-de-
Jian Wang4, Eric Kuhn5, Homa Salehabadi4, Mead, created by Hoover Dam, has a two pletion obligation, even though all stored
John C. Schmidt6 in five chance of falling to its most severe water effectively flows to the Lower Basin.
management condition, which forces large This division is reinforced by Reclamation’s
T
he Colorado River supplies water to reductions on downstream users (5). institutional structure and distinct energy
more than 40 million inhabitants in Municipalities of Los Angeles, San Diego, marketing arrangements between the two
the southwestern United States and Phoenix, Tucson, Las Vegas, Denver, Salt hydropower facilities.
northwestern Mexico. A basin-wide Lake City, Albuquerque, and Tijuana rely Under the Law of the River, a total of 16.5
water supply crisis is occurring be- heavily on the river for their water supplies. MAF/year of the mainstem flow is allocated
cause of decreased watershed run- About 70% of the water is used to irrigate for consumptive use. The primary metric
off caused by a warming climate and legal nearly 5.7 million acres (2.3 million hect- used to evaluate hydrologic conditions is
and water management policies that allow ares) of agriculture. The basin is home to the natural flow at Lee Ferry. The Compact
systematic overuse. By the end of 2022, 30 recognized Native American Tribes that negotiators optimistically presumed a
combined storage in Lake Powell and Lake hold senior legal rights to divert substan- natural flow at Lee Ferry of 17.5 MAF/year
Mead, the two largest reservoirs in the tially more water than they currently use. and more than 20 MAF/year basin-wide.
United States, will have declined from 95% Between 2000 and 2021, the average annual Evidence suggests, however, that they es-
full in 2000 to approximately 25% full. If energy generation from the two major dams chewed scientifically sound estimates that
this “Millennium Drought” persists, then was 7.6 terawatt-hours (TWh)/year, enough the available supply was potentially less (9).
stabilizing reservoir levels to avoid severe to serve 2.5 million people. The river’s This knowledge was dismissed to help reach
outcomes will require reducing water use landscapes and ecosystems provide criti- an agreement; the basin is increasingly pay-
to match diminished runoff. With a process cal habitat for federally protected species ing the price for this strategy. The 20th-cen-
underway to renegotiate interstate and in- (6) and support an extensive recreation- tury natural flows at Lees Ferry averaged
ternational agreements on consumptive based economy. Today, the entire flow is 15.2 MAF/year, an amount nearly sufficient
uses of the river, we describe a promising diverted along its 1400-mile course. In its to meet the Upper Basin’s peak use of 4.0
new management approach based on com- lower reaches, only 10% of the natural flow MAF/year, 9.0 MAF/year of normal alloca-
bined storage of both reservoirs, rather reaches Mexico; rarely does the river flow to tion in the Lower Basin and Mexico, plus
than just Lake Mead as currently used, to the Gulf of California (7). 2.4 MAF/year for typical evaporation losses.
trigger consumptive use reductions to the However, since 2000, the average natural
Lower Basin and Mexico. CONSTRAINTS OF PAST POLICIES flow dropped to 12.3 MAF/year. To continue
Since 2000, the average annual natural Management of the river is governed by a meeting demands, storage in lakes Powell
flows (that which would exist without hu- set of interstate compacts, court decrees, and Mead decreased from 46 to 13.8 MAF.
man interventions) into lakes Powell and federal laws, secretarial guidelines, and If the Millennium Drought continues or in-
Mead have been almost 20% below the an international treaty that is collectively flows decline further, then the only option
20th-century average (1, 2). As a result of referred to as the Law of the River. The will be to reduce consumptive uses to match
these unprecedented low flows and insuf- cornerstones are the 1922 Colorado River the diminished supply.
ficient management adaptations (3, 4), Compact and the 1944 Treaty between the
5-year projections by the US Bureau of United States and Mexico. The Compact is THE RACE TO REDUCE DEMANDS
Reclamation(Reclamation) suggest that an agreement among seven Basin States, The Lower Basin and Mexico have been
Lake Powell, created by Glen Canyon Dam, which divided the watershed into two parts, fully using their combined 9.0 MAF/year ap-
has a one in four chance of falling below the a lower basin that includes portions of Ari- portionment of the Colorado River. Under
minimum elevation necessary to produce zona, Nevada, and California and an upper forceful federal prompting, the Lower Ba-
basin that includes portions of Colorado, sin states committed in 2007 to reductions
1
Environmental Change Institute, University of Oxford, New Mexico, Utah, Wyoming, and a small in consumptive uses (known as “shortages”)
Oxford, UK. 2Water Balance Consulting, Boulder, CO, USA.
3
area in Arizona. The Compact apportioned in stages based on Lake Mead levels through
Colorado Water Center, Colorado State University, Fort
Collins, CO, USA. 4Utah Water Research Laboratory and 7.5 million acre-feet (MAF; 1 acre-foot = 1233 “Interim Guidelines.” Recognizing that these
Department of Civil and Environmental Engineering, Utah m3) per year of consumptive use to each reductions would be insufficient to slow the
State University, Logan, UT, USA. 5Independent consul- basin and specified the division between drawdown of Lake Mead, a 2019 Drought
tant. 6Center for Colorado River Studies, Department of
Watershed Sciences, Utah State University, Logan, UT, USA. them as Lees Ferry in northern Arizona (8, Contingency Plan (DCP) augmented these
Email: [email protected] 9). The Lower Basin was developing rapidly commitments. Mexico agreed to reduce uses

approximately in proportion to US commit- tions—is that a delivery obligation applies the Compact, yet 100 years later, that has not
ments through negotiated implementation only to intermittent drought risk with no occurred. Economic and equity considerations
agreements to the 1944 Treaty (Minutes). underlying change in mean flows, not the also exist. The Lower Basin irrigates less than
Collectively, these agreements require the substantively different and much larger risk half the area irrigated by the Upper Basin, yet
Lower Basin and Mexico to reduce their 9.0 of permanently reduced flows. its agricultural sales are more than three times
MAF/year usage by 2.7 to 15.2% (0.241 to One thing is clear: Additional Upper that of the Upper Basin (13). Because the loss
1.375 MAF/year), with reductions increas- Basin consumptive uses would decrease of an established resource is arguably more
ing as Lake Mead’s storage declines from 41 inflows to Lake Powell and reduce storage harmful than never having developed one,
to 23% full (10.9 to 6.0 MAF). Reductions volumes in lakes Powell and Mead. Lower proposed large new uses are being ques-
in Lower Basin use already occurred in Basin users have indicated that they are un- tioned, and existing uses are facing unprec-
2020 and 2021 under the DCP. Additional likely to reduce their uses to stabilize reser- edented reductions.
voluntary reductions of 0.5 MAF/year by voirs only to see new upstream uses nullify Given many possible solutions, our re-
Lower Basin States and 0.1 MAF/year by these conservation efforts. search identified combinations of Upper
Reclamation were recently proposed (10).
The Interim Guidelines and Treaty
Minutes were triggered for the first time Average combined storage assuming drought conditions continue
in 2022; thus, the combination of required Average end-of-year combined Lake Powell and Lake Mead storage is shown, assuming hydrologic conditions
and newly proposed voluntary Lower Basin of the Millennium Drought continue. Results show combined reservoir contents using a range of Upper Basin
and Mexico reductions will be 13.5% of consumptive use limits (colored ribbons) along with a range of Lower Basin maximum consumptive use
their allocation (1.213 MAF/year). If Lake reductions (line styles) triggered when the combined storage falls below 15 million acre-feet (MAF). The status
Mead storage declines to 6.0 MAF (23% of quo lines use the 2016 Upper Colorado River Commission (UCRC) projections and existing elevation-based
capacity), then the required and voluntary shortage triggers. All water use and shortage values are annual volumes (MAF/year).
reductions would reach 21.9% (1.975 MAF/
year). Citing concerns of hydropower fail- 25
Upper Basin use limits (MAF/year) Maximum Lower Basin shortage (MAF/year)
ure, Reclamation’s commissioner Touton
3.0 3.5 4.0 4.5 1.375 1.5 2.0 2.5 3.0
informed Congress in June that 2 to 4 MAF
Combined Lake Powell and Lake Mead storage (MAF)
UCRC Schedule
of reductions below current commitments
are needed. She did not specify how these 20
reductions should be made among the
states but reiterated the federal authority
to act unilaterally if needed. All interstate
and international shortage agreements
15
will expire by 2026; a renegotiation pro-
cess is underway.
THE UPPER BASIN SQUEEZE

In contrast to the Lower Basin and Mexico, 10
the Upper Basin is not using its full 7.5 MAF/ Status quo Match status quo
year apportionment. Between 2000 and (Lake Mead only) (combined storage)
2020, Upper Basin consumptive uses aver-
aged 3.7 MAF/year plus at least 0.7 MAF/year Both reservoirs at minimum power pool
5
of reservoir evaporation. There are plans for
additional development; the Upper Colorado Selected storage stabilization options
River Commission (UCRC) ambitiously pro- Upper Basin limit of 4.0 / Lower Basin shortage of 2.0
jects 5.4 MAF/year of Upper Basin uses by Upper Basin limit of 4.5 / Lower Basin shortage of 3.0
2060, exclusive of reservoir evaporation (11). 0
Additional Upper Basin water use threatens 2021 2026 2031 2036 2041 2046 2051 2056
to expose the uncertainty around the mean-
ing of the Compact’s non-depletion obliga- WHAT MUST BE DONE? Basin consumptive use limitations and
tion, which in turn could upset basin-wide Considering alarming warming trends and Lower Basin reductions to maintain res-
water delivery expectations. the past 23 years of drought, water managers ervoir storage levels if the Millennium
Under variable year-to-year hydrologic must face the possibility that recent conditions Drought continues (14). If these measures
conditions but with unchanging mean will persist or worsen. Tree ring studies indi- allow the current storage levels to be main-
flows, the non-depletion obligation is fre- cate that longer past droughts have occurred tained, then we consider the system to be
quently interpreted as a firm requirement (12). Up to half of the recent flow decline stabilized under these specific, but highly
for the Upper Basin to deliver a fixed vol- has been attributed to Upper Basin warm- relevant, runoff conditions. Although the
ume downstream. Under declining flows, ing, and additional declines are likely with focus of our study is a scenario of continued
however, the meaning of a non-depletion continued climate trends (1, 2). Under these drought, the insights and approaches can be GRAPHIC: K. FRANKLIN/SCIENCE
obligation becomes unclear. A fixed deliv- conditions and with reservoirs nearly depleted, adapted to plan for other future scenarios.
ery requirement under declining flows puts simple mass balance dictates that consump- We used Reclamation’s Colorado River
the entire burden of climate change on the tive uses must be reduced. But to what extent Simulation System (CRSS) (15), which has
Upper Basin. A more nuanced view of this and how should reductions be allocated? been used for all major basin-wide analyses
obligation—and one that would arguably The Upper Basin emphasizes water use and decisions on the Colorado for the past
align with the Compact negotiators’ inten- equality between the basins as envisioned in 20 years and will be used in forthcoming

renegotiations. To represent future hydro- ing of the non-depletion obligation in ex- tion. Such work is iterative and slow, adding
logic conditions, we developed 100 possible change for defined Upper Basin use limits, to the difficulties and pressures faced by de-
scenarios by randomly resampling natural providing both basins with less risk and cision makers.
flows that occurred from 2000 to 2018 (12). more certainty. Our results show that although current
Mean annual flow at Lee Ferry during this To implement our approach, different policies are inadequate to stabilize the
period was 12.4 MAF/year, and our resam- combined storage trigger thresholds were Colorado River if the Millennium Drought
pling method maintained the annual vari- iterated with varying Lower Basin and continues, various consumptive use strat-
ability (see supplementary materials). Mexico shortage volumes until a reasonable egies can stabilize the system. However,
The first management strategy we as- match with the status quo was discovered these measures must be applied swiftly.
sessed is what would happen if current con- (see the figure). We then incrementally in- Although these concessions by both basins
sumptive use reduction commitments by creased the consumptive use reductions to may seem unthinkable at present, they will
the Lower Basin and Mexico remain in place the Lower Basin and Mexico whenever the be necessary if recent conditions persist. j
and uses in the Upper Basin increase as pro- combined reservoir storage falls below 15
jected by the UCRC. This status quo scenario MAF. Reclamation’s well-established CRSS
1. B. Udall, J. Overpeck, Water Resour. Res. 53, 2404
assumes continued drought conditions but model is thoroughly documented (15), and (2017).
otherwise uses all existing assumptions and our adaptations are described in preceding 2. P. C. D. Milly, K. A. Dunne, Science 367, 1252 (2020).
logic in CRSS, including current obligatory work (14) and the supplemental materials. 3. S. Blumstein, J. D. Petersen-Perlman, Water Int. 46, 306
(2021).
and voluntary consumptive use reduction If the Millennium Drought persists, then 4. J. Fleck, A. Castle, Water 14, 2 (2022).
measures. Following Reclamation’s previ- the combined storage under the status quo 5. US Bureau of Reclamation (USBR), “Colorado River
ous studies, this scenario also assumes will decrease to 6 MAF (12% of total Mead 5-year probabilistic projections report” [US Department
that the Upper Basin non-depletion obliga- and Powell storage) before it stabilizes. At of the Interior (DOI), May 2022]; https://www.usbr.gov/
lc/region/g4000/riverops/crss-5year-projections.
tion in the 1922 Compact is not invoked. this volume, either Glen Canyon Dam or html.
Meeting this obligation would require the Hoover Dam would stop generating hy- 6. K. L. Dibble, C. B. Yackulic, T. A. Kennedy, K. R. Bestgen, J.
Upper Basin to curtail uses, a legally dis- dropower. These impacts show reservoir C. Schmidt, Ecol. Appl. 31, 1 (2021).
7. J. Pitt et al., Ecol. Eng. 106, 629 (2017).
puted issue owing to differing interpreta- contents averaged across conditions since
8. Acre-feet is the volumetric unit historically embedded in
tions of when this obligation is triggered 2000; exceptionally dry years such as 2020 all policies, legal allocations, and operational decisions
and to what extent Upper Basin uses must and 2021 will have an even greater impact. on the Colorado River.
be curtailed. This issue is unlikely to be re- Current reservoir storage levels could, 9. E. Kuhn, J. Fleck, Science be Dammed: How Ignoring
Inconvenient Science Drained the Colorado River (Univ.
solved in a time frame conducive to manag- however, be stabilized if consumptive uses Arizona Press, 2019).
ing further drought. The result of the status decrease under different scenarios (see fig. 10. USBR, Arizona Department of Water Resources,
quo scenario is sharply declining combined S1). If the Upper Basin commits to limit Southern Nevada Water Authority, Press Release:
storage of lakes Mead and Powell (see the water uses to 4.5 MAF/year (60% of their Water agencies announce partnership to invest $200
million in conservation efforts to bolster Colorado
figure), which falls to levels that further 7.5 MAF/year allocation, approximately 0.8 River’s Lake Mead, under 500+ plan (Central Arizona
threaten hydropower production and in- MAF/year higher than recent use), then Project, Phoenix, 2021); https://library.cap-az.com/
creases risk of disruptions to downstream the Lower Basin and Mexico must commit documents/departments/planning/colorado-river-
programs/CAP-500PlusPlan-NewsRelease.pdf.
water deliveries. to more than doubling their current maxi- 11. Upper Colorado River Commission, “Upper Colorado
To evaluate the magnitude of potential mum reductions in existing use to 3.0 MAF/ River division states, current and future depletion
policy changes to stabilize the system, we year (see the figure and fig. S1). In this sce- demand schedule” (2016); http://www.ucrcommission.
conducted a two-dimensional sensitiv- nario, the Lower Basin and Mexico receive com/RepDoc/DepSchedules/CurFutDemandSchedule.
pdf.
ity analysis that mapped out a range of 66.7% of their allocation, nearly matching 12. H. Salehabadi et al., “The future hydrology of the
greater reductions in existing consumptive the Upper Basin percentage. If the Upper Colorado River Basin” (Center for Colorado River
use by the Lower Basin and Mexico when Basin limits their depletions to 4.0 MAF/ Studies, Utah State University, 2020); doi:10.4211/hs.d
3efcf0c930646fd9ef4f17c56436d20.
reservoir storage diminishes, combined year (53.3% of their allocation, 0.3 MAF/ 13. USBR, “Moving Forward Effort: Phase 1 Report” (DOI,
with a range of future Upper Basin uses. year higher than recent use), then the May 2015); https://www.usbr.gov/lc/region/programs/
Our proposal deviated from current man- Lower Basin and Mexico would need to de- crbstudy/MovingForward.
agement practice in which Lower Basin crease uses by approximately 2.0 MAF/year 14. K. Wheeler et al., “Alternative management paradigms
for the future of the Colorado and Green Rivers” (Center
shortages are based only on elevations of to stabilize the reservoirs (see the figure and for Colorado River Studies, Utah State University, 2021);
Lake Mead, a policy that reflects the in- fig. S1), assuring 77.8% of their allocation. doi:10.4211/hs.59175cf99a58462f901cdf56ec79ddbe.
stitutional divisions between the Upper This is close to recently proposed maximum 15. USBR, “DRAFT CRSS: Key modeling assumptions, June
2021 model” (DOI, 2021); http://bor.colorado.edu/
and Lower Basin. We instead used the Lower Basin and Mexico commitments to
Public_web/CRSTMWG/CRSS.
combined storage of the two reservoirs to reduce existing use, which would not be in-
trigger consumptive use reductions to the voked until Lake Mead declines further by 3 ACKNOWL EDGMENTS
Lower Basin and Mexico. Our approach ac- MAF. Delaying these reductions until then We acknowledge the valuable support of A. Castle, D.
Kenney, and D. Tarboton. The authors acknowledge fund-
knowledges the hydrologic reality that wa- would result in greater loss of storage and
ing from the Catena Foundation (grant 202059), Walton
ter stored in both reservoirs is consumed stabilization occurring at lower levels than Family Foundation (grant 2018-585), My Good Fund,
almost exclusively in the Lower Basin and shown in the figure. and David Bonderman fund. J.C.S. acknowledges a Janet
Mexico. Furthermore, the current opera- Water management models such as Quinney Lawson Chair in Colorado River Studies endow-
ment. H.S. acknowledges funding from the Utah Water
tional policies that govern the storage bal- CRSS are only one part of the difficult work Research Laboratory. K.G.W. acknowledges funding from
ance between the reservoirs are likely to needed to achieve real-world solutions. the Oxford Martin Programme on Transboundary Resource
evolve in the forthcoming negotiations. We Resolving complex water supply problems Management.
also assumed that the non-depletion obli- in large transboundary basins also requires SUPPL EMENTARY MATE RIALS
gation is not invoked if the Upper Basin deep understanding of the social and eco- science.org/doi/10.1126/science.abo4452
limits future depletions. This removes the nomic implications of any proposed poli-
longstanding ambiguity over the mean- cies, along with political barriers to adop- 10.1126/science.abo4452

INSIGHTS
Adding structure can offer students more

opportunities to engage, practice, and get feedback.
ing it proceeded too quickly, some saying

too slowly—inspired her to create videos of
herself solving problems that students could
watch at their own pace before class.
Hogan and Sathy also include summaries
of education research—their own as well as
work conducted by others—and fictional sce-
narios that highlight the impact of inclusive
practices. This juxtaposition of educational
research findings with hypothetical scenarios
is jarring at first. However, as the book pro-
gresses, this duality proves effective, allowing
the authors to discuss empirical evidence of
the impacts of inclusive teaching practices
while clearly illustrating how these impacts
B O OKS et al . can play out in the real world. In chapter 2,
for example, the authors discuss how struc-
ture can improve learning outcomes, by
EDUCATION providing more required opportunities for
students to engage, practice, and get feed-
Setting college students back. They illustrate this potentially abstract

concept with a fictional professor, Dr. Slim,
who offers limited chances for students to
up for success practice and get feedback in the course.

They argue that by adding more structure
in the form of pre-class questions, online
A pair of researchers outline strategies for ensuring quizzes, group discussion, in-class activities,
that postsecondary courses are inclusive and post-class reflections, Dr. Slim can guide
and support students more effectively.
At times, however, this strategy is less
By Jeremy L. Hsu offering a timely and much-needed comple- convincing, particularly when the authors
ment to the existing literature. discuss areas with less empirical work from
W
hat does it mean to have an in- The authors shine at translating the some- which to draw. There is a paucity of research
clusive college classroom? How times-dense literature of education research examining instructor practices outside the
can we promote inclusion—both into clear, accessible, and actionable steps for classroom, for example. The authors rely on
inside and outside the classroom— instructors. In chapter 3, for example, Hogan their own experiences to fill such gaps, al-
and help students from different and Sathy focus on language used in course though the distinction between anecdote and
backgrounds succeed in science, syllabi, an element that is oft forgotten. evidence could have been made clearer.
technology, engineering, and mathematics Here, they cite studies showing how student- The book addresses several underexplored,
(STEM)? Authors Kelly Hogan and centered language (e.g., language but critical, areas where instructors can play
Viji Sathy, both STEM professors at that uses “I,” “you,” and “we” rather vital roles in promoting inclusion outside the
the University of North Carolina at than “students will”) can make classroom. A whole chapter, for instance, fo-
Chapel Hill and well-known educa- students feel more connected and cuses on modeling inclusivity with students,
tion researchers, tackle these criti- perceive the instructor as more including suggestions for proactively email-
cal questions in their aptly named competent. They also provide spe- ing struggling students; promoting transpar-
book, Inclusive Teaching. cific examples and rubrics for im- ency and sharing norms about office hours;
Hogan and Sathy cite the large proving syllabi. Each chapter closes reducing bias in grading; and providing
disparities in performance often Inclusive Teaching with a concise summary, framed as structured feedback to encourage students.
observed across different student Kelly A. Hogan a checklist of items for instructors Similarly, another chapter is devoted to in-
demographics in undergradu- and Viji Sathy to consider. stitutional change, discussing how instruc-
ate STEM courses as inspiration West Virginia University Hogan and Sathy weave first- tors can reflect and document their inclusive
Press, 2022. 272 pp.
for their work, regularly return- hand accounts into the narrative, teaching practices and use their efforts to ad-
PHOTO: NOSYSTEM IMAGES/GETTY IMAGES

ing to these disparities to convince readers describing the continual improvements vocate for change.
that more must be done to ensure that all they have made in their teaching over their These strengths make Inclusive Teaching
students are equipped to succeed. The book respective careers. For example, in the first compelling and critical. Given the urgent
provides evidence-based practices and ac- chapter, Hogan describes adding guided need to promote justice, equity, diversity, and
tionable suggestions on inclusive teaching, reading questions, pre-class homework, and inclusion in our communities, the book is
in-class polls to her introductory biology a must-read for all who are in a position to
course after seeing certain students struggle. better support inclusive teaching inside and
The reviewer is at the Schmid College of Science and
Technology, Chapman University, Orange, CA 92866, USA. Sathy, meanwhile, describes how student outside the classroom. j
Email: [email protected] complaints about her class’s pace—some say- 10.1126/science.abq5892

TECHNOLOGY
The virtual worlds of the metaverse

An immersive internet is just around the corner, for better or worse
By Dov Greenbaum initially, of many closed and proprietary sys- The Metaverse: And
tems, resulting in competing virtual worlds. It How It Will Revolutionize
Everything
W
ith the cryptocurrency market spi- will likely exacerbate many of the ongoing ills
Matthew Ball
raling downward and the bursting of the internet, including perpetuating a lack Liveright, 2022. 352 pp.
of the non-fungible token (NFT) of data rights and facilitating manipulation,
bubble, many skeptics believe radicalization, harassment, and misinforma-
that the metaverse—a thoroughly tion online. But according to Ball, with the
immersive emerging version of necessary governance, the benefits will ulti- the hardest nontechnical challenge for the
the internet—is just another fleeting trend. mately outweigh the system’s shortcomings. metaverse to overcome. And while he seems
Mathew Ball’s The Metaverse seeks to re- There are also considerable technical hur- agnostic as to the overall future of distrib-
assure the reader that it is not. Reporting dles that require resolution before the meta- uted ledger technology, he suggests that
that the term “metaverse” was mentioned verse—at least as it is currently conceived of blockchains could become a key platform in
more than 260 times in US Securities and in popular culture—can be achieved. These reducing this control by the tech companies.
Exchange Commission filings in 2021, Ball include “fitting a supercomputer into the There is some debate as to when exactly the
suggests that “the sheer number of compa- frame of…glasses,” which Mark Zuckerberg critical mass of relevant innovations will solve
nies that see potential value in the the technical and logistic barriers
Metaverse speaks to the size and di- now precluding the metaverse. Like
versity of the opportunity.” the internet before it, “Disruption is
The three-part book provides more not a linear process, but a recursive
than just a definitive definition of the and unpredictable one.” But when
metaverse, which, despite mounting it does arise, the book suggests that
interest, many people still fail to fully many may see it simply as a successor
understand. The first section delivers to the current internet.
an overview of the technology in all Ball distinguishes the metaverse
its potential iterations. The second from another proposed internet suc-
examines the ongoing technological cessor—Web3—“a somewhat vaguely
infrastructure expansion that is nec- defined future version of the internet
essary to grow and maintain it. And built around independent developers
the final section predicts the societal and users.” The often-conflated con-
changes that will arise as a result of cept aims to use blockchain technolo-
expansive adoption of the metaverse. gies to become a more decentralized
Ball defines the metaverse as “a internet. Ball notes that while “the
massively scaled and interoperable principles of Web3 are likely critical
network of real-time rendered 3D to establishing a thriving Metaverse,”
virtual worlds that can be experi- confusing the metaverse for Web3 is
enced synchronously and persis- like “conflating the rise of democratic
tently by an effectively unlimited republics with industrialization or
number of users with an individual A visitor plays a virtual game at the 2018 IFA consumer electronics fair. electrification”—one is about gover-
sense of presence, and with continu- nance, the other is about technology.
ity of data.” He refers back to this definition has described as “the hardest technology The book also outlines three nontechni-
throughout the book to signpost various cen- challenge of our time.” Ball personally con- cal, but nonetheless critical, factors that will
tral aspects and elements of the technology siders the current inability to host more than be essential to metaverse growth. The first
as he details its potential. Ball also frequently a limited number of individuals concurrently is regulatory action to open up better pay-
and deferentially references the 1992 novel in one version of a virtual world to be the ment solutions and increase standardization
Snow Crash and its author, Neal Stephenson, hardest problem to solve. and interoperability. Online games—particu-
who coined the term metaverse, as he pre- While much of the book is a breezy read, larly Fortnite, Roblox, Minecraft, and even
sents the technology’s evolution through the Ball sometimes wades into the details of Microsoft Flight Simulator—have been de-
lens of the development of the internet and what systems will be necessary for the meta- scribed as “proto-Metaverses,” and they con-
PHOTO: SHAN YUQI/XINHUA/ALAMY
the online video gaming industry. verse to thrive. He describes, for example, tinue to provide a second essential nontech-
In contrast to the open nature of the inter- the genesis and current state of internet pay- nical factor: social acceptance. The third and
net, the metaverse will be composed, at least ment rails, the “complex series of systems final component crucial to the success of the
and standards, deployed across a wide net- metaverse will be the availability of useful and
The reviewer is at the Zvi Meitar Institute for Legal work and in support of trillions of dollars in desirable experiences, including those related
Implications of Emerging Technologies, Harry Radzyner economic activity.” Here, he argues that the to education, lifestyle, entertainment, fashion,
Law School, Reichman University, Herzliya, Israel, and
Computational Biology and Bioinformatics, Yale University, current control exerted by the major tech advertising, and industrial innovation. j
New Haven, CT, USA. Email: [email protected] companies over payment systems will be 10.1126/science.add5905

Ninth International Congress on Register
Peer Review and Scientific Publication now!
September 8-10, 2022 | Chicago, IL
Congress Advisory Board The Ninth International Congress on Peer Review

and Scientific Publication will be held September
Congress Directors
John P.A. Ioannidis, Meta-Research Innovation Center at 8-10, 2022, in Chicago, Illinois, USA, in person and
Stanford (METRICS)
Michael Berkwits, JAMA and the JAMA Network online. This congress, organized by JAMA Network,
Congress Executive Director The BMJ, and METRICS, will feature 3 days of
Annette Flanagin, JAMA and the JAMA Network
European Director presentations of new and original research.
Theodora Bloom, BMJ
Associate Directors
Steve Goodman, Meta-Research Innovation Center at
What?
Stanford (METRICS)
Véronique Kiermer, PLOS The Congress will provide a forum for the presentation and
David Moher, Ottawa Hospital Research Institute discussion of new research into the quality and credibility of peer
Director Emeritus review and scientific publication, to establish the evidence base
Drummond Rennie, University of California, San Francisco
on which scientists can improve the conduct, reporting, and
Kamran Abbasi, BMJ dissemination of scientific research.
Dianne Babski, National Library of Medicine
Vivienne Bachelet, Medwave, Universidad de Santiago de Chile Who?
Albert-László Barabási, Northeastern University Participants will include editors and publishers of scientific peer-
Howard Bauchner, Boston University School of Medicine
reviewed journals, researchers, funders, bibliometric and informatics
Lisa A. Bero, The University of Sydney
Patrick M. Bossuyt, University of Amsterdam experts, information innovators, librarians, journalists, policymakers,
Lex Bouter, Vrije University ethicists, scientific information disseminators, and anyone interested
Isabelle Boutron, Université de Paris
in the progress of the scientific information enterprise and the quality
An-Wen Chan, University of Toronto
Kay Dickersin, Johns Hopkins Bloomberg School of Public Health
of scientific evidence.
Dan Evanko, American Association for Cancer Research
The Congress embraces a wide range of disciplines, including
James Evans, The University of Chicago
Fiona Godlee, London, UK
biomedicine, health and life sciences, applied sciences, basic
R. Brian Haynes, McMaster University Faculty of Health Sciences sciences, physical and chemical sciences, mathematics, computer
James Kigera, Annals of African Surgery sciences, engineering, economics, and social sciences. New and
Sabine Kleinert, The Lancet
emerging disciplines are also welcome.
Christine Laine, Annals of Internal Medicine
José Florencio F. Lapeña Jr, University of the Philippines
Malcolm MacLeod, The University of Edinburgh
When?
Emilie Marcus, University of California, Los Angeles September 8-10, 2022
Ana Marušić, Journal of Global Health
Bahar Memani, Elsevier
Frank Miedema, University Medical Center Utrecht
Registration will open to attendees in Spring, 2022.
Jigisha Patel, London, UK
Eric J. Rubin, New England Journal of Medicine Questions? [email protected]
David Schriger, University of California, Los Angeles
Emily Sena, BMJ Open Science
Magdalena Skipper, Nature and Nature Research
Victoria Stodden, University of Southern California
Deborah Sweet, Cell Press
Sarah Tegen, American Chemical Society
Martin R. Tramèr, Geneva University Hospitals
Valda Vinson, Science
Erik Von Elm, Cochrane Switzerland
Eric-Jan Wagenmakers, University of Amsterdam
Jasmine Wallace, American Society for Microbiology
peerreviewcongress.org
0722Product.indd 378 7/14/22 7:45 AM

LET TERS
A wildfire rages in
the forests of the
Zagros Mountains in
southwestern Iran.
Edited by Jennifer Sills storage ponds, and helicopter launch pads 1 June (2), will charge a fee to developers
throughout high-risk areas. Finally, nation- whose projects result in wetland area
Better preparation for wide wireless networks would facilitate fire
alerts (3). With the support of the public,
losses. The fees will pay for restoring wet-
lands with comparable qualities and quan-
Iran’s forest fires nongovernmental organizations, and scien-
tists, Iran can reduce the rate of wildfires.
tities elsewhere. When this strategy has
been implemented in the past, transpar-
In Iran, one of the driest countries in Mahmood Tavakoli Hafshejani,1 Mahmoud ency has been insufficient. The wetlands
the world (1–3), about 1500 wildfires are Nasrollahzadeh2, Valiollah Mirkhani1* law, as well as other laws requiring restora-
1
reported each year, destroying thousands Department of Chemistry, Faculty of Science, tion fees, must include data tracking and
University of Isfahan, Isfahan 81746-73441, Iran.
of hectares of forests and pastures annually 2
Department of Chemistry, Faculty of Science,
availability to ensure that the money is
and causing more than US$5.6 million in University of Qom, Qom, Iran. used as intended and that the restored eco-
economic damage (4–6). To mitigate fire *Corresponding author. systems are suitable substitutes for those
Email: [email protected]
damage, Iran should improve containment that have been degraded.
strategies and work toward more effective REF ERENCES AND NOTES China has enacted two previous nation-
fire prevention. 1. A. Shahabfar, A. Ghulam, J. Eitzinger, Int. J. Appl. Earth wide mandatory natural habitat restora-
By the time fires are detected in Iran, Obs. Geoinf. 18, 119 (2012). tion fees, one in 1998 for forest vegetation
2. R. Akbari, M. Nasrollahzadeh, Science 375, 984 (2022).
they are often difficult to control (5, 7). (3, 4) and one in 2003 for grassland veg-
3. H. Ghazanfarpoor, S. Hasanzadeh, M. Hamedi, J. Nat.
Iran lacks specialized manpower, for- Environ. Hazards 5, 61 (2017) [in Farsi]. etation (5). In each case, tracking conser-
est emergency bases, and air relief. 4. R. Abedi, Int. J. Geoherit. Parks 10, 84 (2022). vation outcomes and evaluating whether
Firefighters often lack access to the fire 5. S. Eskandari, S. Eskandari, Hum. Environ. 19, 175 (2021) ecological compensation requirements
[in Farsi].
sites and water reservoirs. As a result, 6. R. Jahdi et al., Nat. Hazards 101, 911 (2020). and targets are being met have proved
time-consuming preparation hinders their 7. S. Aleemahmoodi Sarab, J. Feghhi, H. Goshtasb, Intl. J. challenging. Information on how much
ability to prevent the fire’s spread (3, 7, 8). Mol. Evol. Biodivers. 3, 15 (2013). money various levels of governments
8. S. Ozmehmet Tasan, Y. E. Ergenc, in New Global
Iran could also benefit from better fire Perspectives on Industrial Engineering and Management, have collected and spent, and on what, is
prevention (9). By creating a database of J. Mula, R. Barbastefano, M. Díaz-Madroñero, R. Poler, extremely limited, at least in the public
past fires in the geographic information Eds. (Springer, 2019), pp. 203–211. domain. The lack of financial transpar-
9. A. Alexandridis, D. Vakalis, C.I. Siettos, G.V. Bafas, Appl.
system, the Iranian government could focus Math. Comput. 204, 191 (2008). ency could lead to misuse or misappro-
preventive measures in the areas most at 10. Z. X. Zhang, H. Y. Zhang, D. W. Zhou, J. Arid Environ. 74, priation of restoration funds as well as
risk of wildfire (10). A spatial database could 386 (2010). ineffective use of funds, with money going
track the factors influencing the occurrence 10.1126/science.add5194 toward, for example, projects with no evi-
PHOTO: AHMAD HALABISAZ/XINHUA/GETTY IMAGES
of wildfires and the environmental charac- dence of positive outcomes (6).

teristics of forests in different regions. These A similar approach has been imple-
data would allow Iran to build fire towers
and watchtowers in high-risk areas. In addi-
China’s restoration fees mented for wetlands in the United States
since the 1980s (7) as well as for other
tion, Iran should educate the public to raise
awareness about forest vulnerability and the
require transparency habitats in other countries, including
Australia, Brazil, the United Kingdom,
importance of natural resources (3, 5). China is home to 10% of the world’s wet- and Germany (6, 8). In each case, results
Iran should invest the funds necessary land areas, but many of those wetlands are were mixed (8, 9). Given that success is not
to provide updated firefighting training threatened by development (1). To increase guaranteed, it is even more vital to track
and equipment. The government should conservation efforts, China’s first wetland the progress of the program and adjust its
also build forest road networks, water protection law, which came into force on implementation to maximize benefits.

INSIGHTS | L E T T E R S
In 2021, China committed to enhanc-

ing biodiversity and ecosystem functions Global goals overlook challenges that such ecosystems face. In
the Intergovernmental Panel on Climate
and services by gradually advancing
information disclosure and encouraging
freshwater conservation Change’s sixth assessment report (AR6), the
working group on “impacts, adaptation and
public participation (10). In light of this As global conservation and restoration vulnerability” breaks ecosystem impacts
pledge, China’s government should cre- policies focus on a land and sea frame- into terrestrial, ocean, and fresh water (6).
ate a mechanism to clearly, thoroughly, work, freshwater biodiversity and services In addition, the latest draft of the post-2020
and regularly report the collection and continue to decline at alarming rates (1). Global Biodiversity Framework indicates the
use of forest, grassland, and wetland If freshwater ecosystems are overlooked, possibility of including fresh water in sev-
restoration fees. The information should their sustainability could be compromised eral goals and targets (7).
include government spending, ecological when decision-makers evaluate trade-offs Ahead of keystone events like COP27
assessment before development begins, with land and sea conservation and devel- in November in Sharm El-Sheikh, Egypt
restoration implementation, and out- opment goals. To protect freshwater bio- and the UN Biodiversity Conference
comes (11), and all data should be made diversity and vital services, international (COP15) finally scheduled for December
available for public scrutiny. As a moni- agreements must explicitly acknowledge in Montreal, authors of reports that
toring system model, China could use freshwater ecosystems as a unique realm influence international agreements must
the US Regulatory In-lieu Fee and Bank and set specific goals to address their make the case that freshwater ecosystems
Information Tracking System, a registry problems (2, 3). require attention independent of other
of conservation-related programs that has At the 2021 UN Climate Change conservation efforts. This recognition
been in place for nearly 40 years (12). Conference in Glasgow (COP26), countries could include, if not an additional SDG,
Shuo Gao1,2*, Joseph W. Bull3, Zhiqiang Wu4,5, reaffirmed their commitments to the three targets addressing freshwater-specific area
Renlu Qiao4,5, Li Xia6, Ming K. Lim7 Rio Conventions on Biological Diversity, protection and restoration, the waterflow
1
Interdisciplinary Centre for Conservation Science, Climate Change, and Desertification (4). quality needed to maintain ecosystems
University of Oxford, Oxford, Oxfordshire, UK.
2
St. Hilda’s College, University of Oxford, Oxford,
The three respective panels are preparing and related services, and integrated water
Oxfordshire, UK. 3Durrell Institute of Conservation reports that will shape the 2030 sustain- resources management (2).
and Ecology, University of Kent, Canterbury, Kent, able development goals (SDGs) and the Ground and surface freshwater habitats
UK. 4College of Architecture and Urban Planning, 2021–2030 UN Decade on Ecosystem are home to more than 10% of all known
Tongji University, Shanghai, China. 5Shanghai
Research Institute for Intelligent Autonomous Restoration. Setting explicit objectives for species, including 30% of all vertebrates
Systems, Tongji University, Shanghai, China. freshwater ecosystems in these goals must (8). The ecosystem services they provide
6
School of Management, University of Science and be a priority. are estimated to be worth more than
Technology of China, Hefei, Anhui, China. 7Adam
Smith Business School, University of Glasgow, Unfortunately, the recently released US$4 trillion annually (9). Only by explic-
Glasgow, UK. “Global land outlook” (5), the flagship itly recognizing the value and distinctive-
*Corresponding author. publication of the UN Convention to ness of freshwater ecosystems can we set
Email: [email protected]
Combat Desertification, a convention goals that can effectively protect them.
REFERENCES AND NOTES that defines pathways to sustainable Duarte V. Gonçalves1,2* and Virgilio Hermoso3
1
1. W. Xu et al., Curr. Biol. 29, 3065 (2019). land and water management, still mostly Interdisciplinary Centre of Marine and
2. Ministry of Ecology and Environment of the People’s Environmental Research (CIIMAR), Matosinhos,
treats fresh water as a simple resource
Republic of China, “Wetland Protection Law of the Portugal. 2Research Centre in Biodiversity and
People’s Republic of China” (2022); www.mee.gov.cn/ for services such as irrigation and con- Genetic Resources (CIBIO/InBIO), University
ywgz/fgbz/fl/202112/t20211227_965347.shtml [in sumption rather than a unique ecosystem of Porto, Vairão, Portugal. 3Departamento de
Chinese]. that sustains biodiversity and a range Biología Vegetal y Ecología, Universidad de
3. Ministry of Ecology and Environment, “Forestry Law of Sevilla, Sevilla, Spain.
the People’s Republic of China” (2021); www.mee.gov. of other services and that has particular *Corresponding author.
cn/ywgz/fgbz/fl/202106/t20210608_836755.shtml management needs. The undervaluing of Email: [email protected]
[in Chinese]. freshwater ecosystems is demonstrated
4. B. Madsen, N. Carroll, K. Moore Brands, “State of REF ERENCES AND NOTES
biodiversity markets report: Offset and compensa-
by how rarely the word is used: Fresh
1. World Wildlife Fund, “Living planet report 2020: Bending
tion programs worldwide” (Ecosystem Marketplace, water is mentioned twice in the summary the curve of biodiversity loss” (Gland, Switzerland,
Washington, DC, 2010). for decision-makers, but both times as 2020).
5. Ministry of Ecology and Environment, “Grassland Law of 2. C. B. van Rees et al., Conserv. Lett. 14, e12771 (2021).
“freshwater use,” with no mention of the
the People’s Republic of China” (2021); www.mee.gov. 3. Wetlands-International, “We need wetlands: The urgent
cn/ywgz/fgbz/fl/200212/t20021228_81958.shtml [in associated ecosystems or their manage- case for global wetland targets” (2022).
Chinese]. ment. Land restoration commitments of 4. “Glasgow leaders’ declaration on forests and
6. J. W. Bull, N. Strange, Nat. Sustain. 1, 790 (2018). “1 billion hectares of farms, forests, and land use,” UN Climate Change Conference
7. R. F. Ambrose, Wetlands Australia J. 19, 1 (2010). UK 2021 (2021); https://ukcop26.org/
8. S. O. zu Ermgassen et al., Conserv. Lett. 12, e12664 pastures” make no explicit allusion to glasgow-leaders-declaration-on-forests-and-land-use/.
(2019). rivers or other freshwater ecosystems. 5. UN Convention to Combat Desertification (UNCCD),
9. W. Matthews, A. G. Endress, Environ. Manage. 41, 130 This shortsightedness is consistent with “The Global Land Outlook, second edition” (UNCCD,
(2008). Bonn, 2022).
10. State Council of the People’s Republic of China SDG 15 (“life on land”), which discounts 6. Intergovernmental Panel on Climate Change, “Climate
(State Council), “Opinions on Further Strengthening the uniqueness of the freshwater realm, change 2022: Impacts, adaptation, and vulnerability.
Biodiversity Conservation” (2021); www.gov.cn/ and with SDG 6 (“water and sanitation”), Contribution of Working Group II to the sixth assess-
zhengce/2021-10/19/content_5643674.htm [in ment report of the Intergovernmental Panel on Climate
Chinese]. which prioritizes only the most immediate
Change” (Cambridge University Press, 2022).
11. H. Kujala et al., One Earth 5, 650 (2022). services that freshwater ecosystems pro- 7. Convention on Biological Diversity, “Preparation
12. U.S. Army Corps of Engineers, U.S. Environmental vide. Underestimating the value of fresh of the post-2020 global biodiversity frame-
Protection Agency, U.S. Fish and Wildlife Service, work” (2022); www.cbd.int/doc/c/c949/b2cc/
National Marine Fisheries Service, Natural Resources
water undermines the potential for long-
a311c0c411d3a81134e2c7f3/wg2020-03-l-02-en.pdf.
Conservation Service, “Federal guidance for the estab- term sustainability. 8. R. A. Mittermeier et al., “Fresh water: The essence of life
lishment, use and operation of mitigation banks,” Fed. Some recent reports provide hope that (vol. 18)” (Conservation International, 2010).
Reg. 60, 58605 (1995). 9. A. J. Lynch et al., Environ. Rev. 24, 115 (2016).
we can prioritize freshwater conservation
10.1126/science.add5125 and recognize the unique problems and 10.1126/science.add6295

RESEARCH
reproduces the experiment,
opening up new vistas for the
growing field of computational
heterogeneous catalysis. —YS
Science, abq1414, this issue p. 394
OPTICS
IN S CI
CIENCE
E N CE JOURNAL
JOU RNAL S Amplification in photonic
Edited by Michael Funk time crystals
Regular photonic crystals are
structures in which the refractive
index is spatially periodic and
can suppress the spontaneous
emission of light from an emit-
ter embedded in the structure.
In photonic time crystals, the
refractive index is periodically
modulated in time on ultrafast
time scales. Lyubarov et al.
explored theoretically what
happens when an emitter is
placed in such a time crystal
(see the Perspective by Faccio
and Wright). In contrast to the
regular photonic crystals, the
authors found that time crystals
should amplify emission, leading
to lasing. —ISO
Science, abo3324, this issue p. 425;
see also abq5012, p. 368
ORGANIC CHEMISTRY
Tetrodotoxin by
EVOLUTION cycloaddition
Tetrodotoxin is a potent bacterial
Maintaining difference neurotoxin widely associated with
S
pecies often comprise several ecotypes, distinct populations that occupy different pufferfish and thoroughly studied
habitats. Ecotypes can persist over long time periods, even with substantial gene flow for its sodium channel–blocking
between them, which raises the question of how they maintain their locally adaptive properties. Its intricate structure
phenotypes over time. Hager et al. examined the genetic basis of two traits, tail length of oxygen-rich interconnected
and coat color, that define the forest and prairie ecotypes of deer mice. They found a rings has also long intrigued
large chromosomal inversion that links redder coats and longer tails in the forest ecotype. synthetic chemists. Konrad et al.
Modeling suggests that the inversion originated under divergent selection many thousands of report a comparatively concise
generations ago and likely provided a benefit to the forest ecotype by suppressing recombina- route to the natural product from
tion despite gene flow. —BEL Science, abg0718, this issue p. 399 a glucose derivative. A dipolar
cycloaddition enabled the forma-
Deer mice, such as this individual from central Oregon, are a widely dispersed species with locally adaptive ecotypes.
tion of the cyclohexyl core at a
later stage than prior approaches.
Ruthenium catalysis was then
key in assembling the surround-
PHOTO: MICHAEL DURHAM/MINDEN PICTURES
SURFACE CHEMISTRY constants, which also makes it ion imaging–based calibration ing oxygenated rings. —JSY
impossible to rigorously vali- of absolute molecular beam Science, abn0571, this issue p. 411
Making surface chemistry date theoretical estimates. Even fluxes, Borodin et al. managed
more exact for reactions as simple as ther- to overcome established experi-
SEMICONDUCTORS
Accurate description of elemen- mal recombination of hydrogen mental difficulties and report
tary steps of chemical reactions atoms on platinum surfaces, unprecedentedly accurate rate Swift carriers
at surfaces is a long-standing previous experimental rate con- constants for this reaction over Boron arsenide is a semi-
challenge because of the lack of stants have only been obtained a wide temperature range. They conductor with several
reliable experimental measure- with large uncertainties. Using also demonstrate a parameter- interesting properties, including
ments of the corresponding rate velocity-resolved kinetics and free model that quantitatively a high thermal conductivity.

RESE ARCH | I N S C I E N C E J O U R NA L S
Theoretical calculations also findings revealed that one-third

IN OTHER JOURNALS Edited by Caroline Ash
suggest that it has high ambi- of children in high-transmission
and Jesse Smith
polar mobility, a measure of settings had severe febrile illness
the mobility of electrons and caused by other types of patho-
holes. Yue et al. and Shin et al. gens. —CNF
used different types of mea- Sci. Transl. Med. 14, eabn5040 (2022).
surements to observe a high
ambipolar mobility in very pure
TUMOR IMMUNOLOGY
cubic boron arsenide. Shin et
al. were able to simultaneously Under the skin
measure the high thermal and Although systemic immune
electrical transport proper- checkpoint blockade displays
ties in the same place in their therapeutic efficacy in cancers,
samples. Yue et al. found even it induces immune-related
higher ambipolar mobility than adverse events that can under-
the theoretical estimates at a mine treatment. Van Pul et al.
few locations. Boron arsenide’s performed a phase 1 clinical trial
combination of transport testing the effects of an intra-
properties could make it an dermal injection of antibodies
attractive semiconductor for that block cytotoxic T lympho-
various applications. —BG cyte–associated protein-4 at
Science, abn4290 and abn4727, the site of primary tumor exci-
this issue p. 433 and p. 437. sion in patients with early-stage
melanoma. Seven of 13 patients
immunologically responded to
MALARIA
the treatment without severe
Severe malaria insights adverse events. Responders had
Diagnosis of children with severe more tumor antigen–specific
malaria caused by infection with T cells in the blood, increased
Plasmodium falciparum has been migratory dendritic cell activa-
difficult in high-transmission tion in the sentinel lymph node,
PLASTIC RECYCLING
settings because of the high and decreased regulatory T cells
coincidence of malaria with
other febrile illnesses. Watson
in both the sentinal lymph node
and the blood. This treatment
Unlocking polystyrene
M
any of today’s most common plastics are difficult to
et al. analyzed data from 2649 has promise for patients with
break down for recycling. In polystyrene, for instance,
severely ill children and adults early-stage melanoma. —DAE
the reactive C=C double bond in the monomer
in low- and high-transmission Sci. Immunol. 7, eabn8097 (2022).
becomes an inert C–C single bond in the polymer.
settings enrolled in several
Kiel et al. sought to circumvent this problem by
malaria clinical studies. Using a
introducing a small percentage of alternative monomers
combination of platelet counts PROTEIN DESIGN
into the backbone. Specifically, when styrene was copo-
and plasma concentrations of P.
falciparum histidine-rich protein-2
Designing around lymerized with 2.5% of a thionolactone, treatment of the
in a Bayesian latent class model, function plastic with a thiol cleaved the backbone at the sulfur sites.
The resultant oligomers could then be repolymerized to
PHOTOS (LEFT TO RIGHT): MICHAEL ABBEY/SCIENCE SOURCE; TANAPHONGPICT/SHUTTERSTOCK

the authors achieved a sensitiv- Protein design has had success
show properties similar to those of pure polystyrene. —JSY
ity of approximately 74% and a in finding sequences that fold
J. Am. Chem. Soc. 10.1021/jacs.2c05374 (2022).
specificity of approximately 93% into a desired conformation, but
in identifying severe malaria. Their designing functional proteins Introducing a small percentage of certain alternative monomers
remains challenging. Wang et into the backbone of polystyrene, such as the beads pictured, could
al. describe two deep-learning make it easier to recycle.
methods to design proteins that
contain prespecified functional
sites. In the first, they found
sequences predicted to fold into CANCER types of cancer. Kay et al. exam-
stable structures that contain ined CAFs from a human model
the functional site. In the second,
Targeting the tumor and a mouse model of breast
they retrained a structure pre- microenvironment cancer and found that gluta-
diction network to recover the Cancer-associated fibroblasts mine metabolism drives proline
sequence and full structure of a (CAFs) are stromal cells with synthesis, which is required to
protein given only the functional important roles in modulating produce tumorigenic collagens.
site. The authors demonstrate tumorigenesis, tumor progres- The pyrroline-5-carboxylate
their methods by designing sion, and responses to therapy. reductase 1 (PYCR1) enzyme,
Microscopy image of Plasmodium proteins containing a variety of CAFs produce collagen, a important in proline synthe-
falciparium parasites infecting red functional motifs. —VV component of the extracellular sis, was overexpressed in
blood cells Science, abn2100, this issue p. 387 matrix that can promote some CAFs. Reducing its expression

reduced collagen production, chain reaction assay to detect factor. Indeed, elephants have scenarios, lending further
tumor growth, and progression hrp2/3 deletions. This droplet as many as 20 copies of p53 urgency to efforts to eliminate
in vivo. Numerous subtypes of technology is beginning to be genes, more than any other our greenhouse gas contribu-
CAFs with different functions adopted for malaria diagnosis animal. Therefore, Padariya et tions. —HJS
exist in the tumor microen- in endemic countries. Screening al. combined in silico modeling Nat. Clim. Change 12, 547 (2022).
vironment, so targeting their showed that samples obtained and in vitro assays to explore
pro-tumorigenic effects may be from South America and Africa the p53 isoforms of elephants.
an effective strategy in certain are highly heterogeneous for The authors discovered that SENSORS
cancer types. —GKA hrp deletions. For example, the they are functionally diverse
Nat. Med. 4, 693 (2022). authors found that up to 80% and express a variety of binding
Something smells
of parasites in hospitalized motifs known as BOX-I MDM2. fishy here
patients in Eritrea had hrp2 These motifs enhance sensitiv- Rotting fish release a bouquet
AGING deletion, and high levels of dele- ity to cellular damage that of chemicals, but trimethyl-
tions also occurred in Brazil. prefigure cancer and promote amine (TMA) is the main source
Fat to brain Apart from avoiding false nega- defense pathways. —DJ of the odor that we detect.
longevity signal tives, additional advantages of Mol. Biol. Evol. 10.1093/ Sensors for TMA are thus
Lysosomes, once mainly this approach lie in improved molbev/msac149 (2022). used to indicate freshness but
thought to degrade cellular Plasmodium falciparum identi- struggle to remain flexible and
components, are now impli- fication and a reduction in the sensitive at low temperatures.
GLOBAL WARMING
cated in complex signaling number of assays needed for Li et al. developed a sensor
roles. Savini et al. found that accurate diagnosis. —CA What are the odds? based on the MXene Ti3C2Tx,
overexpression of the lyso- eLife 11, e72083 (2022). How likely is it that we will be where T is a terminating group
somal acid lipase LIPL-4 (which able to limit global warming to consisting of O, OH, or F. The
serves as a fat storage function) 1.5° or 2.0° C above the prein- MXene was modified with Au
CANCER
in the intestine of the worm dustrial average, the goals set nanoparticles and then polym-
Caenorhabditis elegans seems p53 in elephants by the Paris Agreement? Dvorak erized with a hydrogel and
to extend life span through As we age, our cells are more et al. used an emissions-based soaked in a solution of ethylene
communication with the brain. likely to become cancerous climate model to estimate the glycol and HCl to endow it with
In the proposed scheme, a because they accumulate more evolution of global temperature antifreeze properties. The com-
specific polyunsaturated fatty mutations. How very large, long- after an abrupt cessation of posite hydrogel showed high
acid (PUFA) produced by LIPL-4 lived animals such as elephants anthropogenic greenhouse gas strength, stretchability, and
appears to be carried to the and whales protect themselves emissions using a variety of toughness even at low tempera-
brain by the fatty acid–binding against cancer is an interesting emission scenarios. They found tures, and was able to sense
protein LBP-3. In the brain, the question. One hypothesis is that that there is only one chance in TMA at both room temperature
PUFA seemed to enhance tran- these animals have more copies three that peak global warm- and 0°C. —MSL
scription of the neuropeptide or higher expression of the p53 ing can be kept below 1.5°C by ACS Appl. Mater. Interfaces 14,
NLP-11, which is required for the tumor suppressor transcription 2027–2032 under all emissions 30182 (2022).
life-extending effect. Thus, fatty
acids bound to carrier proteins
might have a hormone-like
function, so lysosomes may
be a signaling hub coordinat-
ing communication between
organs. —LBR
Nat. Cell Biol. 24, 906 (2022).
MALARIA
Digital gene deletion
diagnosis
Rapid diagnostic tests for falci-
parum malaria primarily rely on
detecting the virulence-related
PHOTO: YASUYOSHI CHIBA/AFP/GETTY IMAGES
histidine-rich proteins 2 and 3

(HRP2 and HRP3). Sometimes,
the genes hrp2 and hrp3 are
deleted in the parasite and
the diagnosis is missed, which
obviously presents a problem
for treatment, surveillance, and
control. Vera-Arias et al. have
developed a high-throughput Long-lived animals such as elephants (shown here is Tolstoy, a 49-year old “tusker” in Amboseli National Park) thwart cancer
droplet digital polymerase with isoforms of a tumor suppressor.

RESE ARCH
ALSO IN SCIENCE JOURNALS Edited by Michael Funk
GENOMICS the authors identified some capacity, nitrogen utilization, species experience less seasonal
features of the cancer cells and and flowering time. In field tri- variation in temperature and
DNA sequencing to nanoparticles that could predict als, plants overexpressing this thus have narrower climatic
support conservation successful nanoparticle delivery, gene delivered greater yield, niches, but an alternative view
A huge number of fungal, plant, which should facilitate further shortened growth duration, and is that ranges may be limited by
and animal species are threat- therapeutic advances. —YN improved nitrogen use efficiency. competition with other species.
ened with extinction, and this Science, abm5551, this issue p. 384; —PJH Freeman et al. found that species
biodiversity loss is largely caused see also add3666, p. 371 Science, abi8455, this issue p. 386; richness is a better predictor of
by human activity. Conservation see also add3882, p. 370 smaller elevational ranges than
efforts to protect endangered climate seasonality, suggesting
CORONAVIRUS
species are complex and could a larger role of competition in
be supported by genomic tools. Glycans in the spotlight CORONAVIRUS shaping tropical mountain biodi-
In a Perspective, Paez et al. Viral infection of a cell requires versity than has previously been
discuss how high-quality refer- a complex series of molecu-
A shifting landscape recognized. —BEL
ence genomes could be used lar recognition events, often for spike Science, abl7242, this issue p. 416
to aid species preservation and mediated by glycoproteins and As proteins evolve, the muta-
population diversity. Additionally, cell-surface glycans. Buchanan tional landscape changes
CATALYSIS
these efforts might be used in et al. developed a nuclear mag- because a mutation at one
de-extinction measures such as netic resonance analysis method residue position may affect the Channeling water away
genetic rescue to avoid del- to better study such interac- functional consequences of a Heterogeneous catalytic reac-
eterious changes in dwindling tions and applied it to influenza mutation at a second position. tions that produce water as a
populations and even in resur- and severe acute respiratory To explore how this plays out in by-product can be inhibited by
recting extinct traits. Although syndrome coronavirus 2 (SARS- the evolution of the severe acute its presence on the surface. Fang
genomics will not solve biodi- CoV-2) proteins binding sialoside respiratory syndrome corona- et al. found that for the produc-
versity loss on its own, these glycans. For SARS-CoV-2 in virus 2 (SARS-CoV-2) spike tion of light olefins from syngas
tools could provide important particular, they found evidence protein, Starr et al. measured (a 2:1 mixture of hydrogen and
information and new avenues to for a sialoside-binding site in the the effects of all single-amino- carbon monoxide) with a cobalt
explore in multifaceted conser- N-terminal domain of the origi- acid mutations on binding of manganese carbide catalyst
vation approaches. —GKA nal B-origin lineage spike protein the spike protein to the cellular at 250°C, the addition of the
Science, abm8127, this issue p. 364 that was lost in subsequent receptor ACE2 in the context of hydrophobic polymer polydivi-
variants. These results were different SARS-CoV-2 variants nylbenzene as part of a physical
corroborated by cryo–electron (Wuhan-Hu-1, Alpha, Beta, Delta, mixture almost doubled the con-
NANOMEDICINE microscopy structures of the and Eta). They show how evolu- version of carbon monoxide (see
glycan-bound spike protein and tion of the protein is shaping the the Perspective by Ding and Xu).
A systematic view genetic variation analysis from possibility for future mutations, Theoretical models suggest that
of nanoparticles patients early in the pandemic, for example, allowing mutations the polymer formed channels
Nanoparticles are increasingly which uncovered host factors that escape antibodies while that accelerated water diffusion
being tested as vehicles for involved in glycosylation that maintaining binding to ACE2. away from the catalyst. —PDS
delivering therapeutics, and potentially contributed to varia- —VV Science, abo0356, this issue p. 406;
some are already in clinical tion in disease severity. —MAF Science, abo7896, this issue p. 420 see also adc9414, p. 369
use for cancer chemotherapy. Science, abm3125, this issue p. 385
Nanoparticle-based treatments
can offer various therapeutic BIOGEOGRAPHY CORONAVIRUS
advantages such as decreased PLANT SCIENCE
toxicity, longer half-life, and
Tropical mountain Delta and Omicron
improved drug delivery. However,
Genetic improvement biodiversity go toe to toe
there are a multitude of possible drives rice yield Mountains in the tropics are There is growing evidence that
nanoparticle formulations with Improvements in agricultural highly biodiverse, often with the Omicron variant of concern
different physical and biological productivity could lessen the entirely different sets of spe- (VOC) is more transmissible
properties, and it is not readily impact of agriculture on the cies occurring at high and low and infectious than previous
apparent which ones would be environment and perhaps elevations. This high turnover iterations of severe acute respira-
best in a given disease setting. supply more food from less occurs because species occupy tory syndrome coronavirus 2.
Boehnke et al. developed a high- land. Working in rice, Wei et al. narrower elevational ranges in Yuan et al. compared Syrian
throughput screening method identified a transcription factor the tropics than on temperate hamsters exposed to either
that allowed them to systemati- that, when overexpressed, has a mountains. Freeman et al. used Omicron or Delta VOCs. Animals
cally evaluate the interactions variety of useful effects (see the citizen science data from eBird, a infected with Omicron showed
of 35 different nanoparticle Perspective by Kelly). The gene’s global citizen science project, to lower respiratory tract viral
types with hundreds of cancer expression is induced by both test two hypotheses for narrower burdens and reduced clinical
cell lines (see the Perspective light and low-nitrogen status, ranges in the tropics. The prevail- severity. Nevertheless, Omicron
by Winter). Using this approach, and it regulates photosynthetic ing hypothesis is that tropical was at least as transmissible
383-B 22 JULY 2022 • VOL 377 ISSUE 6604 science.org SCIENCE

RE S E ARC H
as Delta, if not more so. When properties of the sequence order

animals were challenged with and motor chunks to maximize
a mixture of both variants, the performance of this task.
Delta outcompeted Omicron This process appeared to decline
in naïve hamsters. This com- in older adults, but it could be
petitive advantage disappeared, partially restored by using non-
however, in vaccinated animals. invasive brain stimulation during
Moreover, vaccinated hamsters training. With this treatment,
were better than unvaccinated motor skill acquisition followed
animals at transmitting Omicron a similar pattern as in young
to co-housed companions. adults, in whom speeding up led
This work helps to clarify how to increased motor chunk forma-
Omicron might have gone on to tion, which in turn allowed for the
become the predominant strain increasing speed. —CK
in populations with high rates of Sci. Adv. 10.1126/
previous infection and vaccina- sciadv. abo3505 (2022).
tion. —STS
Science, abn8939, this issue p. 428
IMMUNOLOGY
Freeing T cells from
a sticky situation
T cells must adhere to and
release from other cells to
migrate into tissues and stimu-
late antibody production. The
integrin lymphocyte function–
associated antigen 1 (LFA-1),
which promotes T cell adhe-
sion, is activated downstream
of phosphoinositide 3-kinase
(PI3K), which generates the lipid
second messenger phosphati-
dylinositol 3,4,5-trisphosphate
(PIP3). Johansen et al. identified
the PIP3-binding, GTPase-
activating protein RASA3 as an
inhibitor of LFA-1. Because of
an increase in adhesion, T cells
lacking RASA3 were impaired in
entering or exiting lymph nodes,
and mice with RASA3-deficient
T cells had defective responses
to immunization. —WW
Sci. Signal. 15, eabl9169 (2022).
NEUROSCIENCE
Improving motor skill
acquisition in older adults
Activities of daily life typically
require executing actions in
sequential order, often both
accurately and at high speed.
In the brain, serial actions
are automated into smaller
groups of co-occurring actions
called motor chunks. Using a
sequence-tapping paradigm,
Maceira-Elvira et al. identified
early consolidation of spatial
SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 383-C

RES EARCH
◥ and uncover previously unrecognized regula-

RESEARCH ARTICLE SUMMARY tors of nanoparticle delivery.
NANOMEDICINE RESULTS: We designed a modular fluorescent

nanoparticle library to capture the effects
Massively parallel pooled screening reveals genomic of a wide range of nanoparticle parameters—
including core composition, surface chemis-
determinants of nanoparticle delivery try, and size—on cancer cell interactions. In a
competitive assay, we screened interactions
Natalie Boehnke*†, Joelle P. Straehla*†, Hannah C. Safford, Mustafa Kocak, Matthew G. Rees, of each nanoparticle formulation with 488
Melissa Ronan, Danny Rosenberg, Charles H. Adelmann, Raghu R. Chivukula, Namita Nabar, pooled, barcoded cancer cell lines across 22
Adam G. Berger, Nicholas G. Lamson, Jaime H. Cheah, Hojun Li, Jennifer A. Roth, lineages and identified cells by strength of nano-
Angela N. Koehler, Paula T. Hammond* particle association. The screen was designed to
probe nanocarrier delivery; thus, no toxic cargo
was incorporated. Unsupervised hierarchical
INTRODUCTION: Nanoparticle-mediated deliv- biological features that mediate successful clustering of resulting interaction profiles for
ery of therapeutic agents has the potential to nanoparticle delivery. each nanoparticle–cell line pair identified core
considerably affect cancer treatments, partic- composition as a strong determinant of cell as-
ularly in the context of personalized cancer RATIONALE: Traditional nanoparticle screens sociation, with the three tested core materials
therapies. Nanoparticles span a diverse range of are designed to optimize materials properties forming distinct clusters. To probe cellular fea-
materials and properties. They can be tailored in isolated biological contexts. In the era of tures that govern nanoparticle association, we
to encapsulate and protect a wide range of precision medicine, with the desire to deliver integrated multiomic data by using correlative
therapeutic cargos, including small molecules, molecularly targeted therapies to specific sub- analyses. We appropriately identified high epi-
biologics, and nucleic acids. One challenge to cellular compartments, it is also important to dermal growth factor receptor (EGFR) gene ex-
successful targeted nanoparticle delivery is an probe the structure-function relationship of pression and protein abundance as biomarkers
incomplete understanding of nano-bio inter- nanoparticles as they relate to cellular and that are predictive of cellular affinity for anti-
actions at the target delivery site. In designing biological heterogeneity. The combination of EGFR formulations. More generally, we observed
this screen, we sought to gain a holistic under- pooled screening with multiomic annotation that nanoparticle core material as well as sur-
standing of both the materials properties and has the potential to accelerate target discovery face modification influence the number and
significance of biomarkers that are predictive
of uptake. Many biomarkers were associated
Nanoparticle screening with pooled, barcoded cell mixtures with established uptake, transport, and adhe-
Diverse library of 488 DNA-barcoded sion gene sets. Using machine learning algo-
35 nanoparticles cancer cell lines rithms, we identified predictive biomarkers
that cluster to form interrelated protein-protein
interaction networks, identifying cellular fea-
tures that mediate nanoparticle trafficking.
-2 -1 0 +1 +2
We also identified formulation-specific bio-
Interaction profiles for each Similarity profiles cluster
markers. We validated expression of SLC46A3,
Liposome PLGA PS nanoparticle–cancer cell line pair by core material a lysosomal transporter, as a negative regulator
and predictive biomarker for lipid-based nano-
particle uptake and downstream functional
Omics integration to identify predictive biomarkers
applications. These applications include trans-
Validation of targeted interactions Shared and formulation-specific trafficking networks lation of our findings to in vivo models and
demonstration of SLC46A3 expression mod-
Nanoparticles
ulating both lipid nanoparticle uptake and

transfection efficacy of nucleic acid cargo.
CONCLUSION: This work represents a high-

Features throughput interrogation of nanoparticle–
Distinct and overlapping biomarkers Discovery of a biologic regulator of nanoparticle delivery cancer cell interactions through the lens of
SLC46A3
multiomics. Our analysis provides a frame-
Weighted average
work that will empower studies of nano-bio

interactions and advance the rational design
Significance
of nanocarriers.
▪
The list of author affiliations is available in the full article online.
Effect size SLC46A3 expression
*Corresponding author. Email: [email protected],
Identification Potential applications [email protected] (N.B.); [email protected] (J.P.S.);
[email protected] (P.T.H.)
These authors contributed equally to this work.
nanoPRISM screen integrates drug delivery and omics. Using a curated nanoparticle library, we screened
Cite this article as N. Boehnke et al., Science 377,
nanoparticle-cell interaction profiles of hundreds of cancer cells simultaneously. By incorporating omics eabm5551 (2022). DOI: 10.1126/science.abm5551
annotation, we identified biological features, or biomarkers, that mediate nanoparticle delivery to cells. We
generated trafficking networks and discovered a biologic regulator of lipid-based nanoparticle delivery. PLGA, READ THE FULL ARTICLE AT
polylactide-co-glycolide; PS, polystyrene. https://doi.org/10.1126/science.abm5551

RES EARCH
◥ ships to cellular targeting and uptake (10–13).

RESEARCH ARTICLE However, there is a need to equally consider
the influence of biological heterogeneity on
NANOMEDICINE interactions at the NP-cell interface—for exam-
ple, by probing cells across cancer cell lineages
Massively parallel pooled screening reveals genomic with a range of genetic drivers and cell states.
In the era of precision medicine, with the
determinants of nanoparticle delivery desire to deliver molecularly targeted and
gene-based therapies to specific subcellular
Natalie Boehnke1,2*†‡, Joelle P. Straehla1,2,3,4*†, Hannah C. Safford1, Mustafa Kocak2, compartments within cancer cells, it is im-
Matthew G. Rees2, Melissa Ronan2, Danny Rosenberg2, Charles H. Adelmann5,6,7, perative to holistically probe the structure-
Raghu R. Chivukula6,8, Namita Nabar1,9, Adam G. Berger1,10,11, Nicholas G. Lamson1, function relationship of NPs as they relate to
Jaime H. Cheah1, Hojun Li1,3,4, Jennifer A. Roth2, Angela N. Koehler1,2,12, Paula T. Hammond1,9,10* cellular interactions.
Inspired by recent advancements in can-
To accelerate the translation of cancer nanomedicine, we used an integrated genomic approach to cer genomics (14), we postulated that apply-
improve our understanding of the cellular processes that govern nanoparticle trafficking. We developed a ing similar techniques to the study of cancer
massively parallel screen that leverages barcoded, pooled cancer cell lines annotated with multiomic nanomedicine would uncover both the cell-
data to investigate cell association patterns across a nanoparticle library spanning a range of and NP-specific features that mediate effi-
formulations with clinical potential. We identified both materials properties and cell-intrinsic features cient targeting and delivery. The combination
that mediate nanoparticle-cell association. Using machine learning algorithms, we constructed genomic of pooled screening with multiomic anno-
nanoparticle trafficking networks and identified nanoparticle-specific biomarkers. We validated one tation has accelerated target discovery and
such biomarker: gene expression of SLC46A3, which inversely predicts lipid-based nanoparticle uptake uncovered previously unrecognized mecha-
in vitro and in vivo. Our work establishes the power of integrated screens for nanoparticle delivery nisms of action in small-molecule screens.
and enables the identification and utilization of biomarkers to rationally design nanoformulations. Specifically, in the profiling relative inhibition
simultaneously in mixtures (PRISM) method,
N
DNA-barcoded mixtures of cells have recently
anoparticle (NP)–based therapeutics toxicity but do not efficiently accumulate in been used for multiplexed viability screening.
have enormous potential for personal- tumors (5, 6). In cell line pools grouped by doubling time,
ized cancer therapy because they can en- Delivery challenges attributed to circula- 500 barcoded cell lines have been screened
capsulate a range of therapeutic cargos, tion, immune detection, and clearance, as well against tens of thousands of compounds to
including small molecules, biologics, and as extravasation and diffusion through tissue, identify genotype-specific cancer vulnerabil-
more recently, nucleic acids. Therapy-loaded all influence NP accumulation at target dis- ities (15, 16).
NPs can be designed to prevent undesired deg- ease sites. Efforts to improve NP accumulation To comprehensively capture pan-cancer
radation of the cargo, increase circulation time, in tumors through active targeting motifs have complexities and enable the statistical power
and direct drugs specifically to target tumors been met with limited success, both in the to link NP association with cell-intrinsic char-
(1–3). There have been notable successes in laboratory and the clinic (1, 7). Fewer efforts acteristics, we developed a competitive pheno-
clinical translation of nanomedicines, includ- have focused on gaining a fundamental under- typic screen to assess associations of a curated
ing liposomal formulations of doxorubicin standing of the biological features that medi- NP library with hundreds of cancer cell lines
(Doxil) and irinotecan (Onivyde) (4). These ate successful NP-cell interaction and uptake. simultaneously. NP-cell association was cor-
formulations extend the half-life of the ac- Although progress has been made in under- related with genomic features to identify can-
tive agent and have the potential to lower standing how specific physical and chemical didate biomarkers. Coupling our biomarker
NP properties affect trafficking and uptake, findings with k-means clustering, we con-
1 comprehensive evaluation of multiple NP structed genomic interaction networks asso-
Koch Institute for Integrative Cancer Research,
Massachusetts Institute of Technology, Cambridge, MA parameters in combination has thus far been ciated with NP engagement, which enabled
02142, USA. 2Broad Institute of MIT and Harvard, elusive. Additionally, the biologic diversity of the identification of genes associated with
Cambridge, MA 02142, USA. 3Department of Pediatric cancer targets makes it prohibitively challeng- the binding, recognition, and subcellular traf-
Oncology, Dana-Farber Cancer Institute, Boston, MA 02115,
USA. 4Division of Pediatric Hematology/Oncology, Boston ing to gain a holistic understanding of which ficking of distinct NP formulations. Moreover,
Children’s Hospital, Boston, MA 02115, USA. 5Cutaneous NP properties dictate successful trafficking through the use of univariate analyses and
Biology Research Center, Massachusetts General Hospital and drug delivery (8, 9). Once NP parameters random forest algorithms, we identified that
Department of Dermatology, Harvard Medical School,
Boston, MA 02114, USA. 6Whitehead Institute for Biomedical are considered in combination, the number the gene SLC46A3 holds value as a predic-
Research, Cambridge, MA 02142, USA. 7Department of of formulations to test increases exponen- tive, NP-specific biomarker. We further val-
Biology, Massachusetts Institute of Technology, Cambridge, tially, particularly because comparisons across idated SLC46A3 as a negative regulator of
MA 02142, USA. 8Division of Pulmonary and Critical Care
Medicine, Department of Medicine, Massachusetts General
several systems need to be drawn. A further liposomal NP uptake in vitro and in vivo.
Hospital, Boston, MA 02114, USA. 9Department of Chemical barrier is the need to adapt the NP formula- The strategy outlined here identifies cellular
Engineering, Massachusetts Institute of Technology, tion of each encapsulated therapy for a given features underlying NP engagement in can-
Cambridge, MA 02142, USA. 10Institute for Soldier
Nanotechnologies, Massachusetts Institute of Technology,
drug or target because each formulation has cer nanomedicine.
Cambridge, MA 02139, USA. 11Harvard-MIT Health Sciences its own distinct biological fate (9). As therapies
and Technology, Massachusetts Institute of Technology, continue to increase in molecular complexity, Results
Cambridge, MA 02139, USA. 12Department of Biological nanoPRISM: Screening NP association with
new nanocarrier formulations capable of de-
Engineering, Massachusetts Institute of Technology, pooled cell lines
Cambridge, MA 02142, USA. livering such entities will need to be developed
*Corresponding author. Email: [email protected], nboehnke@ and examined for their specific trafficking To screen hundreds of cancer cell lines simul-
umn.edu (N.B.); [email protected] (J.P.S.); [email protected] properties. taneously for NP–cancer cell line association
(P.T.H.) †These authors contributed equally to this work. ‡Present
address: Department of Chemical Engineering and Materials We and others have designed panels of NPs patterns, we cultured pooled PRISM cells and
Science, University of Minnesota, Minneapolis, MN 55455, USA. to elucidate the structure-function relation- incubated them with fluorescent NPs. We then
Boehnke et al., Science 377, eabm5551 (2022) 22 July 2022 1 of 13

RES EARCH | R E S E A R C H A R T I C L E
implemented a fluorescence-activated cell sort- lated and electrostatically coated with cationic for the liposome and PLGA formulations be-
ing (FACS) adaptive gating strategy to sort cell poly-L-arginine (PLR) followed by a series of cause the similar sizes would enable cross-core
populations into four bins (quartiles, A to D) polyanions (17–21). The polyanions were se- comparisons. We also included commercially
on the basis of fluorescence signal as a proxy lected for their synthetic [polyacrylic acid manufactured fluorescent carboxylate- and
for the extent of NP-cell association (Fig. 1A). (PAA)], semisynthetic [poly-L-aspartate (PLD) sulfate-modified polystyrene (PS) NPs in a
Experimental parameters were optimized to and poly-L-glutamate (PLE)], or natural [hyal- range of diameters from 20 to 200 nm, which
ensure sufficient cell number and barcode rep- uronate (HA), dextran sulfate (DXS), fucoidan enabled the study of particle size and surface
resentationafter cell sorting (fig. S1), and NPs (FUC), alginate (ALG), and chondroitin sulfate chemistry. Because of the clinical importance
were incubated for 4 and 24 hours. (CS)] origin as well as the inclusion of both of polyethylene glycol (PEG)–containing for-
For this screen, we designed a modular NP carboxylate and sulfate ions (22–24). These mulations (25), PEGylated versions of liposome,
library to capture the effects of NP core com- same electrostatic coatings were used to modify PLGA, and PS particles were prepared, includ-
position, surface chemistry, and size on cell polymeric NP cores [polylactide-co-glycolide ing the drug-free versions of two commercial
interactions. This panel of 35 NPs encompassed (PLGA)] to test the effects of core composi- formulations, liposomal doxorubicin (Doxil)
both clinical and experimental formulations. tion on NP-cell interactions. We optimized and liposomal irinotecan (Onyvide). The latter
Specifically, anionic liposomes were formu- formulations to obtain a diameter of ~100 nm two formulations are denoted as LIPO-5%
A B Zeta (mV) Number average size (nm) C 8000

LIPO - bare
Bin
488 barcoded cell lines LIPO - PLD
6000 A
1 NP formulation
Counts
LIPO - PLE
pooled in each well LIPO - PAA B
dosed per well
LIPO - HA C
LIPO - ALG
4000
LIPO - DXS D
LIPO - FUC
2000
LIPO - CS
LIPO - 5% PEG
x35 fluorescent LIPO - 25% PEG 0
NP formulations LIPO - EGFR
LIPO -IgG 1 2 3 4 5
4000 log10(barcode count)
LIPO- 0.3% PEG*
LIPO- 5% PEG*
D
2 time Bin
PLGA - bare 3000
points A
Counts
PLGA - PLD
screened B
PLGA - PLE 2000
PLGA - PAA C
FACS sort barcoded sequence PLGA - HA D
PLGA - ALG 1000
cells into bins DNA barcodes
PLGA - DXS
F PLGA - FUC
PLGA - CS
0
0.00 0.25 0.50 0.75 1.00
LIPO PLGA PS PLGA - 50% PEG Bin probability
PS-200nm COOH E 100
PS-100nm COOH
PS-40nm COOH
PS-20nm COOH 75
Counts
PS-200nm SO3
PS-20nm SO3 50
PS-200nm PEG
PS-100nm PEG 25
PS-40nm PEG
PS-20nm PEG
0
-100 -50 0 50 100 150 200 250 300
−2 −1 0 1 2
Zeta (mV) Number average size (nm) Weighted Average
G Core composition Cancer cell lineage

0.4 0.1 Lineage
Aut. gang. Ovary
Bil. tract Pancreas
Bone Pleura
PC2 (15.3%)
PC2 (15.1%)
0.0 0 Breast Prostate

CNS Sal. gland
Endomet. Skin
Kidney Soft tis.
Core −0.1 Lg. Intest. Stomach
−0.4 LIPO Liver Thyroid
PLGA
Lung Aerodig.
PS
Pearson correlation 0.0 0.5 1.0 Esoph. Urin. tract
−0.4 0.0 0.4 −0.1 0 0.1
PC1 (28.6%) PC1 (38.2%)
Fig. 1. Assessing NP-cell interactions across hundreds of cancer cell lines counts, with similar numerical distribution of barcodes in each bin, represented
simultaneously. (A) Schematic of the nanoPRISM assay. Fluorescently as a stacked histogram. (D) Accounting for baseline differences in barcode
labeled NPs are incubated with pooled cancer cells before FACS through NP representation yields the probability (P) that each cell line will be found in
association and sequencing of DNA barcodes for downstream analyses. a particular bin. (E) Probabilities are collapsed into a single WA for each
(B) Characterization of the diameter and z potential of the NP library by means NP–cell line pair. (F) A similarity matrix collapsing WA values for 488 cell lines
of dynamic light scattering. Data are represented as the mean and standard reveals clusters of NP formulations with the same core formulation. (G) PCA
deviation of three technical repeats. Formulations marked with an asterisk of NP–cell line WA values at 24 hours confirms distinct clustering of NP
indicate drug-free analogs of clinical liposomal formulations as described in the formulations based on core composition (left), but cell lines do not form
text. (C) Raw data from the screen were obtained in the form of barcode clusters (right).

PEG* and LIPO-0.3% PEG*, respectively. All higher a values (fig. S6), and therefore, a = 2 (q = 6 × 10−21 and q = 2 × 10−18, respectively)
of the NPs examined exhibited negative or was used for subsequent analyses. than the those of IgG control (q = 3 × 10−9 and
neutral net charge because the focus of this q = 3 × 10−6, respectively) (Fig. 2A, bottom).
work is on systemic NP delivery systems. Cancer cells distinguish NPs on the basis of The statistical significance of EGFR bio-
Positively charged NPs have been shown to core composition markers was lower for the antibody-conjugated
undergo nonspecific charge interactions with Pearson-based unsupervised hierarchical clus- liposome than the free antibody, which may
cells and proteins, leading to toxicity and tering of pairwise WAs identified NP core be due to changes in protein concentration
premature clearance in vivo (26). Dynamic material as a strong determinant of cell asso- across samples or steric blockage introduced
light scattering (DLS) was used to characterize ciation, with the three core materials tested by covalently linking an antibody to a NP
the diameter, z potential, and polydispersity (liposomal, PLGA, and PS) forming distinct surface that may interfere with binding to
index (Fig. 1B and tables S1 to S3) of this NP clusters (Fig. 1F and fig. S7A). This result was its target (36). Thus, we demonstrated the
library. unexpected because we hypothesized surface ability to quantitatively compare expected
To ensure that our methods led to robust chemistry to be a larger predictor of NP-cell biomarker targets of both free antibodies
and meaningful data, we selected an anti- interactions. Principal components analysis and antibody-conjugated NPs by using our
body to epidermal growth factor receptor (PCA) similarly identified core-specific trends platform.
(EGFR) as an active targeting control. We at both the 4- and 24-hour time points (Fig. 1G
hypothesized that the design of our screen and fig. S7, B and C) Further analysis within Biomarker number and identity are influenced
would allow us to identify features relevant each core material did reveal surface chemistry– by NP properties
to EGFR expression with a high level of con- dependent trends, although they were more We applied univariate analysis to correlate
fidence. A nonlethal EGFR antibody or im- subtle than core-based clustering (fig. S8). association and CCLE features for each NP
munoglobulin G (IgG) isotype control was By contrast, no clusters were apparent when formulation both quantitatively and quali-
covalently incorporated onto a liposome by PCA was performed according to cell line, in- tatively by using curated gene sets. First, we
means of a PEG tether (27). We elected to dicating that cancer cells of the same lineage thresholded q values at less than 1 × 10−10 to
focus on EGFR because of the wide range of did not have similar NP-association trends compare the absolute number of candidate
native EGFR expression of the 488 cell lines (Fig. 1H and fig. S7, B and C). Heterogeneity biomarkers at varying degrees of significance
included in our screen as well as prior eval- in NP-cell association in proliferating cells (Fig. 2B). Selection of this cutoff was guided by
uation of EGFR-targeting compounds with the has been attributed to various aspects of cell the IgG-conjugated antibody analysis, which
PRISM assay (fig. S2) (15). growth and metabolism (30–33). To ensure returned few hits above this threshold. For
After incubating the cells with the NP li- that differential cell proliferation did not con- liposomal NPs, we observed that the num-
brary, we used FACS to bin cells into quartiles found our results, we performed a parallel ber of significant biomarkers was higher at
according to fluorescence intensity (fig. S3). growth experiment with the same pooled cells 4 hours than 24 hours. We believe that this
Cells were then lysed, and the DNA barcodes and found no correlation between estimated may be indicative of active uptake processes,
were amplified, sequenced, and deconvoluted doubling time and WA (fig. S9). established to take place within the first few
according to previously detailed protocols hours of NP-cell interactions, whereas at
(15, 28). After quality control analysis of tech- Cell-intrinsic features mediate NP trafficking 24 hours, we may be capturing features asso-
nical (n = 2) and biological (n = 3) replicates, We applied data from the Cancer Cell Line ciated with less specific interactions (37, 38).
all 488 cell lines met quality control measures Encyclopedia (CCLE) (34, 35) to identify ge- We next investigated biomarkers associated
and were carried forward for downstream nomic features that act as predictive bio- with established uptake, transport, and adhe-
analyses (fig. S4). This dynamic gating strategy markers for NP-cell association. To do this, we sion gene sets (Fig. 2C) (39–41). To examine
was used to enable comparison of cell line used both univariate analyses and a random the distribution of biomarker significance
representation per bin (quartile) indepen- forest algorithm to correlate the baseline mo- across curated gene sets and NP formula-
dent of fluorophore identity or amount in- lecular features of each cell line (cell lineage; tions, each gene was visualized by using the
corporated into each tested formulation. gene copy number; mRNA, microRNA, pro- –log(q value) for gene expression. As expected,
A probabilistic model was developed and tein, or metabolite abundance; and function- we identified highly significant biomarkers
applied to the data to infer the relative distri- damaging, hotspot, or missense mutations) from gene sets important in drug import and
bution of each cell line into the predetermined with NP association (fig. S10, A and B). export such as solute carrier (SLC) transporter
bins (A to D) for each NP formulation. The family and adenosine 5′-triphosphate (ATP)–
probability of a cell from a given cell line fall- EGFR-targeting compounds identified relevant binding cassette (ABC) family. Our screen
ing into a given bin is used to represent those biomarkers with high confidence provides data on both the significance and
distributions: PA + PB + PC + PD = 1 (Fig. 1, C Using univariate analysis for all CCLE fea- the relationship to NP delivery. For example,
and D). The technical details and the model’s tures, we identified EGFR gene expression we found that ABCA1, which plays a role in
implementation are presented in the supple- and protein abundance as the two most sig- cholesterol transport, has a positive relation-
mentary materials (29). Given the concordance nificantly correlated hits (q = 4 × 10−100 and ship with liposomal NPs, whereas several
of the inferred probabilities among the biologic q = 4 × 10−76, respectively) with antibody to members of the multidrug-resistance sub-
replicates (fig. S5), we collapsed the replicates EGFR, but much less significantly (q = 6 × family (ABCB1/P-GP, ABCC1/MRP, and ABCC4/
through their arithmetic average. Probabilities 10−9 and q = 4 × 10 −10, respectively) asso- MRP4) have a negative relationship with PLGA
were then summarized by using a weighting ciated with the isotype control (Fig. 2A, top). NPs (fig. S11) (42). We also identified bio-
factor alpha (a) to calculate a weighted aver- We also confirmed that fluorophore identity markers important for cell engagement (focal
age (WA) for each NP–cell line pair: WA = does not affect biomarker identification, adhesion and extracellular matrix) as well as
–aPA – PB + PC + aPD, in which a higher WA demonstrating that both AlexaFluor 488– intracellular trafficking (vesicular transport,
implies higher NP-cell association and vice and Cy5-conjugated antibodies to EGFR per- lysosome, and cholesterol transport). This
versa (Fig. 1E). We trialed a range of weighting form similarly (fig. S10, C to E). highlights the ability of our screen to identify
factors (a = 2, 10, 20, and 100) and found that In EGFR-conjugated liposomes, the same expected biomarkers and enable comparison
downstream results were unchanged with the hits were also identified more significantly between drug delivery modalities.

A EGFR Antibody IgG Antibody C -log(q-value)

100
EGFR NP core LIPO PLGA PS
expression 0 4 8 12 16 20
PEGylated NP surface
75
EGFR
RPPA Time 4 hour 24 hour
50
Significance [-log10(q-value)]
25
EGFR
RPPA Solute carrier family (SLC)
EGFR
0 expression
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30 SLC-mediated transmembrane transport
LIPO-EGFR LIPO-IgG
30
20 EGFR
expression
Focal adhesion pathway
EGFR
RPPA Extracellular matrix binding
10
EGFR
expression ATP-binding cassette (ABC) family
EGFR
RPPA ABC-mediated transmembrane transport
0
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30
Effect size (z-score)
B
4 hour 24 hour
LIPO - bare
LIPO - PLD
LIPO - PLE
LIPO - PAA
LIPO - HA
LIPO - ALG
LIPO - DXS
LIPO - FUC
LIPO - CS
LIPO - 5% PEG
LIPO - 25% PEG
LIPO - 0.3% PEG*
LIPO - 5% PEG*
PLGA - bare
Significance
Vesicle-mediated transport
PLGA - PLD (−log[q])
PLGA - PLE
PLGA - PAA > 25
PLGA - HA 20 - 25
PLGA - ALG
PLGA - DXS 15 - 20
PLGA - FUC 10 - 15
PLGA - CS
PLGA - 50% PEG
PS-200nm COOH
PS-100nm COOH
PS-40nm COOH
PS-20nm COOH
PS-200nm SO3
PS-20nm SO3
PS-200nm PEG
PS-100nm PEG
PS-40nm PEG
PS-20nm PEG Cholesterol transport and efflux
0 250 500 750 0 250 500 750
Number of significant biomarkers Lysosome pathway
D Pearson NP core LIPO PLGA PS E

correlation 0.3 0.4 0.5 0.6 0.8
IGFBP7
CALU
ZYX CSF1
1 2 3 4 5
CALU
CD44
MYLK
MYLK
RHOC
LGALS1
CYR61
FLNA
SNAI2
MMP14
VIM
ITGA5
FGF1 NGF
VIM
ACTN1 FN1
ITGB1
VEGFC MET
EGFR
THBS1 ITGA3
PLAU
ANXA2
INHBA
SDC4 TGFB1
INHBA
SERPINE1
CAV1
SMAD3
STRING database network statistics

Input = cluster 1 and 2 biomarkers
Nodes 205
Edges 725
Avg. node degree 7.07
Avg. local cluster coefficient 0.439
Expected edges 215
PPI enrichment p-value <1e-16
Fig. 2. Correlative genomic analysis identifies expected validation biomarkers significance of biomarkers associated with established transport, uptake, and
as well as hundreds of formulation- and time-dependent biomarkers. (A) Cells adhesion gene sets. Gene set headings are bolded, and subsections are listed
with high EGFR antibody association are strongly correlated with EGFR gene below respective headings. (D) A heatmap showing all gene- and protein-
expression and protein abundance [by means of reverse phase protein array expression features, with positive correlation identified by means of random forest
(RPPA)] (top left). These correlations are diminished in the isotype control-treated algorithm in columns, and NP formulations in rows. Features are colored on
sample. The same EGFR-related hits, in addition to NP-specific markers, are the basis of their Pearson correlation and clustered by using k-means clustering,
observed for cells treated with antibody-conjugated liposomes (bottom row). with clusters 1+2 highlighted as features present across multiple NP formulations.
(B) Univariate analysis identifies genomic features correlated with NP association. (E) Visual representation of the STRING network generated by inputting the
All biomarkers meeting a significance threshold of Ðlog10(q value) > 10 are shown 205 features from clusters 1+2, with network statistics. Each node indicates a
as stacked bar graphs separated by NP formulation and time point. PEGylated feature, and the edges indicate predicted functional associations. (Inset) Zoom-in
NP formulations are indicated with a gray background. (C) A heatmap showing the with the most interconnected nodes labeled.

We also observed that liposome surface Tool for the Retrieval of Interacting Genes/ maintained regardless of cancer cell lineage
modification influences the number and sig- Proteins (STRING) database (46–48) to gener- (Fig. 3B and fig. S13).
nificance of biomarkers. Specifically, liposomes ate a protein-protein interaction (PPI) network We selected nine cancer cell lines from the
electrostatically coated with polysaccharides that was found to be highly interconnected (PPI nanoPRISM pool and four additional cell lines,
(HA, ALG, DXS, FUC, and CS) had the highest enrichment P < 1 × 10−16) (Fig. 2E). The network spanning multiple lineages, with a range of
amount of associated biomarkers, which we is enriched in proteins found in the plasma native SLC46A3 expression levels for screening
hypothesize is due to the high degree of in- membrane, extracellular region, and extra- in a nonpooled fashion (Fig. 3, C and D, and
teractions between sugars and cell surface cellular matrix [false discovery rate (FDR) = figs. S3, S14, and S15). Analogous to the pooled
proteins as well as the potential for naturally 8 × 10−12, 3 × 10−9, and 3 × 10−8, respectively] screen, individual cell lines were profiled by
occurring polysaccharides to interact with a on the basis of enrichment analysis with Gene using flow cytometry, and NP-associated fluo-
wide range of cell surface elements (23, 43, 44). Ontology (GO) localization datasets (fig. S12) rescence was quantified after 24 hours incu-
In line with this hypothesis, the addition of (49–51). The identification of overlapping bio- bation; SLC46A3 expression was concurrently
PEG, a well-established antifouling polymer, markers that are localized to the cell surface quantified by using quantitative polymerase
reduces the number and significance of asso- and have established protein-protein interac- chain reaction (qPCR) (Fig. 3D and fig. S9). In
ciated biomarkers almost to zero. In light of tions led us to hypothesize that these proteins line with observations from pooled screen-
the highly specific hits generated from EGFR- are important in early NP trafficking. Enrich- ing, the inverse relationship between liposome
conjugated liposomes (formulated by using ment analyses by using GO molecular functions association and native SLC46A3 expression was
25% PEG liposomes), this abrupt decrease in datasets showed enrichment in numerous maintained, suggesting that SLC46A3 may play
significant biomarkers further indicates the binding processes (data S1 and fig. S12), giving a key role in regulating the degree of liposomal
ability of our platform to identify specific NP further credence to this theory. NP uptake.
binding and recognition elements. In contrast To probe whether SLC46A3 governs cellular
to the liposomal formulations, PLGA formu- SLC46A3 is a negative regulator of liposomal association with NPs, we selected the breast
lations, regardless of surface modification, NP uptake cancer cell line T47D, which exhibited high
resulted in few biomarkers at either time Evaluating univariate results across NP for- native SLC46A3 (Fig. 4A). We deactivated
point. Last, a high number of significant bio- mulations, we identified one biomarker with SLC46A3 with small interfering RNA (siRNA)
markers was associated with both carboxylated a strong, inverse relationship with liposomal and evaluated the effect on liposomal NP as-
and sulfated PS NPs included in our screen, NP association: expression of solute carrier sociation. We observed that T47D cells with
although there was no time dependence, in family 46 member 3 (SLC46A3). A member reduced SLC46A3 had higher NP-cell associ-
contrast to the liposomal formulation. Al- of the solute carrier (SLC) transporter family, ation with both tested formulations, which
though this result was unexpected, because SLC46A3 is a relatively unstudied transporter suggested that modulating SLC46A3 expres-
the PS formulations are made of synthetic that has been localized to the lysosome (52, 53). sion alone can regulate NP-cell association
polystyrene polymers, meaningful biological SLC46A3 was recently identified as a modulator (Fig. 4B).
interactions with anionic polystyrenes both of cytosolic copper homeostasis in hepatocytes, To further functionally evaluate the rela-
in polymer and particle form have been re- connecting hepatic copper concentrations tionship of SLC46A3 expression and NP-cell
ported. Specifically, it was described that with lipid catabolism and mitochondrial func- association, we selected two cancer cell lines
NPs bearing anionic polystyrene motifs have tion (54). This reported relationship between from the pooled screen (Fig. 4A): the T47D cell
the appropriate mix of hydrophobicity and SLC46A3 and lipid catabolism may help to line and the melanoma cell line LOXIMVI. We
anionic charge character to interact favor- explain why SLC46A3 was found to have a developed a toolkit using these two cell lines
ably with trafficking proteins, including the strong relationship with liposomal NP uptake by permanently deactivating SLC46A3 in T47D
caveolins (45). and not uptake of polymeric NPs. In the con- cells and inducing SLC46A3 overexpression in
text of cancer, SLC46A3 was recently shown LOXIMVIs (fig. S16, A to D).
NP biomarkers are connected and create to transport noncleavable antibody-drug con- Because SLC46A3 is a protein associated
trafficking networks jugate (ADC) catabolites from the lysosome to with lysosomal membranes (55, 56, 59), we
We then used an unbiased approach to iden- the cytosol, thereby being necessary for ther- used LysoTracker dye to evaluate the effect
tify predictive biomarkers by using a random- apeutic efficacy (55). Further, down-regulation of SLC46A3 modulation on endolysosomal
forest algorithm, annotated by feature set: gene of SLC46A3 was identified as a resistance compartments in both T47D and LOXIMVI
expression, gene copy number, and protein mechanism for ADC delivery in cancer cells, engineered cell lines (Fig. 4C). We observed an
abundance. Data from the 4-hour time point including in patient samples of multiple mye- SLC46A3-dependent change: cells with lower
were chosen for this analysis on the basis of the loma (55–58). Although the biologic function SLC46A3 expression (T47D-SLC46A3 deactiva-
EGFR-related hits for liposomes, which were of SLC46A3 in cancer is not yet clear, given the tion, LOXIMVI-vector control) exhibited more
more significant at 4 hours than at 24 hours. potential therapeutic implications and the un- brightly dyed endolysosomal compartments
Because we were interested in applying this usual inverse relationship between SLC46A3 as compared with that of their high-SLC46A3-
approach to identify cellular features posi- expression and NP delivery, we sought to vali- expression counterparts (T47D-vector control,
tively correlated with uptake (for example, date the predictive power of SLC46A3 as a bio- LOXIMVI-SLC46A3 OE).
increased expression of trafficking proteins), marker for liposomal NP association. Overexpression of SLC46A3 in LOXIMVI
hits negatively correlated with NP associa- SLC46A3 expression was the most signifi- cells significantly abrogated interaction with
tion were removed from this analysis. Next, cant hit on univariate analysis and also the top bare liposomes (P = 0.006) by using flow cy-
we used k-means clustering to visualize bio- ranked random forest feature for each lipo- tometry profiling (Fig. 4D). The T47D-SLC46A3
markers according to their relative importance somal NP tested at 24 hours, regardless of deactivation cell line demonstrated significant-
and presence across formulations (Fig. 2D). surface modification (q < 10−20) (Fig. 3A and ly increased association with bare liposomes
Clusters 1 and 2 contained 205 hits shared fig. S13). This inverse relationship between compared with that of parental or vector con-
across NP formulations and were especially SLC46A3 expression and NP association was trol lines (P = 0.0017) (Fig. 4D). We further
enriched for liposomal and PS NPs. These found to be specific to liposomal NPs, and confirmed that these trends are generaliz-
genes and proteins were input into the Search not observed with PLGA or PS NPs, and was able across a range of surface functionalized

A Bare LIPO LIPO - HA liposomes (Fig. 4E and fig. S16E). Moreover,

30 no significant changes in NP association were
SLC46A3 SLC46A3 observed for PLGA and PS NPs (Fig. 4E and
fig. S16, F and G). We also confirmed that the
20 presence of serum proteins in cell culture
media does not abrogate this trend (fig. S16H).
Taken together, these data indicate that modu-
10
lation of SLC46A3 alone in cancer cells is suf-
ficient to negatively regulate association and
uptake of liposomal NPs.
Significance [−log10(q-value)]
Because flow cytometry does not provide
0
spatial information with respect to NP-cell
−20 −10 0 10 20 −20 −10 0 10 20
interactions, we used imaging cytometry
LIPO - 0.3% PEG* LIPO - 25% PEG to characterize NP localization in a high-
30
throughput manner (Fig. 5, A to F). We se-
SLC46A3 SLC46A3 lected four representative formulations: three
liposomal NPs to probe the relationship of
20 SLC46A3 expression with liposome trafficking
and one PLGA NP formulation with the same
outer layer (PLD).
10 Consistent with trends observed with flow
cytometry, we observed an inverse relation-
ship between NP intensity and SLC46A3 ex-
0
pression for liposomal, but not PLGA, NPs
−20 −10 0 20 −20 −10 0 20
(Fig. 5, A and D, and fig. S11). Using brightfield
10 10
images, we applied a mask to investigate cellu-
Effect size (z-score)
B lar localization of NPs. All tested formulations
2 LIPO p < 0.001 PLGA p = 0.306 PS p = 0.981
were internalized, and this did not change
with SLC46A3 modulation (Fig. 5, B and E).
1 We investigated localization of NPs by
Weighted average
scoring NP signal according to distribution

within each cell (Fig. 5, C and F, and fig. S17).
0
We observed stark differences in median cell-
ular distribution scores of liposomal NPs in
-1 relation to SLC46A3 expression in T47D cells.
This was not observed for PLGA NPs, mimick-
-2 ing the previously observed core-specific rela-
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 tionship between NP-cell association and
SLC46A3 log2(TPM+1) SLC46A3 expression. Changes in this score,
although less pronounced, were also observed
C D
2 p = 0.002 p = 0.025 for liposomal NPs in LOXIMVI cells.
150
To confirm our findings with higher spatial
LOXIMVI
resolution, we used deconvolution microscopy
Weighted average
NP positive (%)
1
100 LOXIMVI
HepG
DAOY of live cells and incorporated a lysosomal stain
MDAMB231
MCF7 to observe changes in intracellular trafficking
0 HeLa
MDAMB231 (Fig. 5, G and H). NPs appeared uniformly dis-
DAOY
MCF7 50 CAOV3
SW948 tributed within T47D-SLC46A3Ðdeactivation
-1 SW948
CAOV3 cells, colocalizing with endolysosomal vesicles.
HCC1143 HCC1395 SJSA-1 By contrast, LIPO-PLD NPs were localized to
SJSA−1 T47D
-2 T47D HCC1143 large endolysosomal clusters in T47D-vector
0
Jurkat
control cells. This trend was also observed
0 1 2 3 4 -15.0 -12.5 -10.0 -7.5 for LIPO-PLE and LIPO-0.3% PEG* NPs and
SLC46A3 log2(TPM+1) SLC46A3 log2(2-delCt)
at the earlier time point of 4 hours (fig. S18).
Changes in localization were not observed for
Fig. 3. Native expression of the lysosomal transporter SLC46A3 predicts NP-cell interaction for the tested PLGA PLD NPs. This again indi-
liposome formulations. (A) Univariate analysis identified SLC46A3 expression as strongly inversely correlated cates a NP core-dependent relationship with
with liposome association, regardless of liposomal surface modification. (B) Using linear regression to SLC46A3.
evaluate the biomarker relationship across core formulations revealed that SLC46A3 expression was inversely In the engineered LOXIMVI cell lines, we
correlated with NP association in liposome–cell line pairs (P < 0.001) but not PLGA– and PS–cell line pairs also observed colocalization of liposomal
(P > 0.05); n = 488 cell lines for each plot. (C) Cell lines in the nanoPRISM pool exhibited a range of native NPs with endolysosomal signal (Fig. 5H).
SLC46A3 expression and a log linear correlation with uptake of bare liposomes. (D) This same correlation was also However, predictable changes in NP locali-
exhibited when assessing liposome-cell associations with flow cytometry in a nonpooled fashion (P = 0.025). zation were not detected, which is in line with
Cell lines in red were not part of the pooled PRISM screen. Data represented in (D) are shown as the mean and smaller changes in median cellular distribu-
standard deviation of four biological replicates. Error bars are not shown when smaller than data points. tion scores.

A B C
5
2000 ****
T47D Low SLC46A3 High SLC46A3
NP-associated fluorescence
T47D SCL46A3 KO T47D vector control
SLC46A3 log2(TPM+1)
4
1500
LysoTracker signal
3
1000
2
LOXIMVI
1 500
0
siScramble siSLC46A3
Cell line in PRISM assay
(n=488) T47D
D **
Low SLC46A3 High SLC46A3
ns
125 LOXIMVI vector control LOXIMVI SLC46A3 OE
NP-associated fluorescence
100
LysoTracker signal
75
**
50 ns
25
0
Parental Vector SLC46A3 Parental Vector SLC46A3
control overexpressing control knockout
LOXIMVI T47D
E Bare LIPO LIPO-DXS LIPO-PLD LIPO-HA Bare PLGA PS-100 COOH

parental
control
T47D SLC46A3
KO
knockout
parental
LOXIMVI SLC46A3
overexpressing
control
OE
10-3 0 103 104 105 10-3 0 103 104 105 10-3 0 103 104 105 10-3 0 103 104 105 10-3 0 103 104 105 10-3 0 103 104 105
Nanoparticle-associated fluorescence
Fig. 4. Modulating SLC46A3 expression in cancer cell lines is sufficient to 10 mm. (D) Using lentivirus to overexpress SLC46A3 in LOXIMVI cells and CRISPR/
negatively regulate interaction with liposome NP formulations. (A) T47D and Cas9 to permanently deactivate SLC46A3 in T47D cells, we showed that modulation
LOXIMVI cells have high and low SLC46A3 expression, respectively, among the results in significantly changed liposome association, as determined with flow
cells in the nanoPRISM cell line pool. (B) T47D cells treated with siRNA to deactivate cytometry (**P < 0.001, Kruskal-Wallis test). NP-associated fluorescence is defined
SLC46A3 have higher uptake of Lipo-PLD compared with that of T47D cells as median fluorescence intensity normalized to untreated cells. Data are represented
treated with a scrambled siRNA control (****P < 0.0001, Mann-Whitney test). as the mean and standard deviation of four biological replicates. (E) Shifts in NP
(C) Representative micrographs of Lysotracker signal in engineered cell lines association were consistently observed across all tested liposomes, independent of
showed endolysosomal compartments. KO, knockout (deactivated). Scale bars, surface modification. No shifts were observed with PLGA or PS formulations.
Impact of SLC46A3 expression on and recycling endosomes (RAB11) as well as NPs, modest changes in EEA1, RAB7, and
endolysosomal maturation is minimal lysosomes (LAMP1) in engineered LOXIMVI LAMP1 texture were noted (fig. S19, A and B).
To further probe the relationship between cells (figs. S19 and S20 and table S4). Although We then assigned values to the colocalization
intracellular liposomal NP trafficking and no apparent differences in endolysosomal between each endolysosomal marker and NP
SLC46A3 expression, we used imaging cy- marker signal strength, size, and shape were signals and observed increasing colocaliza-
tometry to spatially interrogate markers of observed when comparing LOXIMVI-SLC46A3 tion from EEA1 to RAB5 to RAB7, which is
endolysosomal transport. We elected to study OE and LOXIMVI-vector control cells both in consistent with liposome trafficking from
markers of early (EEA1 and RAB5A), late (RAB7), the absence and in the presence of liposomal early to late endosomes (fig. S19, C to F).

A Low NP intensity High NP intensity

B C
NP NP / NP NP /
Cells with signal brightfield signal brightfield
signal
signal
NP
NP
internalized NPs
brightfield
brightfield
NP /
NP /
20
Cellular distribution score

10
SLC46A3 KO
0
-10
vector control Low SLC46A3 High SLC46A3

-20 T47D SLC46A3 KO T47D vector control
0 104 105 106 -20 -10 0 10 20
Nanoparticle intensity (RFU) Cellular distribution score Cells at median cellular distribution score
D E F
LIPO-PLD
LIPO-PLE
LIPO-0.3% PEG*
PLGA-PLD
1x104 1x105 1x106 0 25 50 75 100 0 4 8 12
Median NP intensity (RFU) % of cells with internalized NPs Median cellular distribution score
vector control T47D SLC46A3 KO LOXIMVI SLC46A3 OE
G LIPO-PLD PLGA-PLD
NP signal LysoTracker merge + CellTracker NP signal LysoTracker merge + CellTracker
T47D SCL46A3 KO
Low SLC46A3
T47D vector control
High SLC46A3
H
LOXIMVI vector control
Low SLC46A3
LOXIMVI SLC46A3 OE
High SLC46A3
Fig. 5. High-throughput imaging cytometry confirmed NP internalization LIPO-PLD NPs. (D) Quantification of median intensity of tested NP formulations
and revealed SLC46A3-dependent changes to intracellular trafficking. in engineered T47D and LOXIMVI cell lines demonstrated SLC46A3-dependent
(A) Imaging cytometry was used to investigate the intensity (x axis) and changes. (E) NPs remained predominantly internalized independent of
distribution (y axis) of NPs in a high-throughput manner. (Bottom) Bivariate SLC46A3 expression. (F) Shifts in the median cellular distribution scores
density plot of n = 10,000 cells (T47D-vector control) after 24 hours incubation were observed in response to SLC46A3 modulation. (G and H) Live cell
with LIPO-PLD NPs, with representative cell images at low and high NP signal. micrographs of (G) T47D-vector control and T47D-SLC46A3 deactivation cells
(B) Cellular distribution patterns of NPs were scored so that scores greater and (H) LOXIMVI-vector control and LOXIMVI-SLC46A3 OE cells incubated
than 0 indicate cells with internalized NPs. Representative data from LIPO-PLD with LIPO-PLD and PLGA-PLD NPs for 24 hours. NP signal is pseudo-colored
NPs in engineered T47D cells are shown. (C) Representative cell images at magenta, LysoTracker signal is yellow, and CellTracker is cyan. Scale bars,
the median cellular distribution score for engineered T47D cells treated with (A) and (C), 7 mm; (G) and (H), 5 mm.

Colocalization between RAB7 and liposomal A C

1x107
NPs was higher in LOXIMVI-SLC46A3 OE single intratumoral 4h and 24h
injection
cells as compared with vector control, and the post injection
Assess 8x106
Tumor nanoparticle signal

(TRE/ tumor mass in mg)
opposite relationship was observed for LAMP1 NP signal
LOXIMVI in tumors
colocalization. Inoculate SLC46A3 OE 6x106
tumors or control
Liposome retention and accumulation remains 4x106
SLC46A3 dependent in vivo daily intravenous 4h post final

injections injection 2x106
To evaluate the potential clinical utility of B
SLC46A3 as a negative regulator of liposo-
LOXIMVI SLC46A3 OE
2.2 0
mal NP delivery, we tested in vivo delivery of
High SLC46A3
SLC46A3 OE Vector control
Intratumoral injection
a US Food and Drug Administration (FDA)–
D
2.0
approved NP analog, the drug-free version of 3x106

1.8x108
liposomal irinotecan (LIPO-0.3% PEG*), in
Tumor nanoparticle signal

(TRE/ tumor mass in mg)
mice bearing subcutaneous LOXIMVI flank 1.6
2x106
tumors. Fluorescently labeled NPs were admin-
LOXIMVI vector control

istered by means of a one-time intratumoral
Low SLC46A3
1.4
injection or repeat intravenous administration

Total radiant 1x106
to evaluate tumor retention and accumulation, efficiency (TRE)
respectively (Fig. 6A and fig. S21). ( _______

p/s/cm2/sr
µW/cm2
)
Min= 1.3x108
NP signal was quantified at both 4 and Max= 2.3x108
0
24 hours after intratumoral administration. SLC46A3 OE Vector control
In line with our hypothesis, as well as with Intravenous injections
in vitro NP-associated fluorescence data (fig. Fig. 6. Retention and accumulation of PEGylated liposomes (LIPO-0.3% PEG*) in LOXIMVI tumors is
S21A), we observed an inverse relationship dependent on SLC46A3 expression. (A) Fluorescently labeled LIPO-0.3% PEG* NPs were administered to
between SLC46A3 expression and LIPO-0.3% mice bearing LOXIMVI flank tumors by means of a one-time intratumoral injection or repeat intravenous
PEG* NP retention that became more pro- injections. (B) Whole-animal fluorescence images of mice (four males, six females per group) 24 hours
nounced over time (P = 0.0115, 4 hours; P = after being intratumorally injected with LIPO-0.3% PEG* NPs. (C) Quantification of LIPO-0.3% PEG*
0.0002, 24 hours) (Fig. 6, B and C, and fig. S21, NP retention 24 hours after intratumoral administration to LOXIMVI flank tumors. (D) Quantification of
B to E). Moreover, these findings also align LIPO-0.3% PEG* NP accumulation after repeat intravenous injections. In (C) and (D), NP signal is expressed
with our initial nanoPRISM findings, in which on the y axis as total radiant efficiency divided by tumor mass; units are provided in the figure. The mean
SLC46A3 expression was a more signifi- and standard deviation of n = 10 mice are shown with the exception of the LOXIMVI-vector control,
cant biomarker at 24 hours (q = 3.49 × 10−30) repeat intravenous injection group, where n = 9 mice (**P < 0.01, ***P < 0.001, Mann-Whitney test).
(data S2 and fig. S13A) than at 4 hours (q =
1.47 × 10−4) (data S2 and fig. S13A).
To determine whether SLC46A3 expression
predictably governs accumulation of nontar- LNP association, as quantified by Cy5 sig- the lens of multiomics. Harnessing the power
geted NPs, which bear no specific functional nal, was significantly lower for LOXIMVI- of pooled screening and high-throughput se-
ligands on their surface, after systemic admin- SLC46A3 OE cells than LOXIMVI-vector quencing, we developed and validated a plat-
istration, we quantified NP signal after intra- control cells, showing the same relationship form to identify predictive biomarkers for NP
venous injections. We observed a significant (lower SLC46A3 expression correlating with interactions with cancer cells. We used this
relationship between SLC46A3 and NP accu- higher association) for LNPs as for liposomal platform to screen a 35-member NP library
mulation (P = 0.0019) (Fig. 6D and fig. S21F). NPs (P = 0.008) (Fig. 7, A and B). A similarly against a panel of 488 cancer cell lines. This
This demonstrates that baseline tumor expres- inverse relationship with SLC46A3 expression enabled the comprehensive study and iden-
sion of SLC46A3 may influence NP delivery in was seen for transfection, as quantified by GFP tification of key parameters that mediate NP-
a physiologic setting. signal of formulation LNP 1 (Fig. 7C). Taken cell interactions, highlighting the importance
Together, these data highlight the real- together, these findings suggest that SLC46A3 of considering both nanomaterials and cellu-
world relevance of the nanoPRISM screen- regulates cytosolic delivery of mRNA cargo by lar features in concert.
ing assay in general as well as the utility of way of LNP uptake. Expanding on this, we Although pooled screening is a powerful
SLC46A3 in particular as a potential biomarker. generated two additional LNPs, analogous to tool, there are several important limitations.
commercial formulations (table S5) (62–65). First, we primarily focused on lipid-based and
Solid lipid NP uptake and transfection are Although we observed lower transfection in polymeric NP formulations with translational
dependent on SLC46A3 expression LOXIMVI-SLC46A3 OE cells than in LOXIMVI- drug delivery potential. We recognize that
Given the recent translational success and vector control cells, these differences were not there are several additional categories of nano-
promising potential of nucleic acid–carrying statistically significant (P > 0.05). Neverthe- materials with wide-ranging properties, such
solid lipid NPs (LNPs) (60, 61), we sought less, the inverse relationship between SLC46A3 as inorganic systems, that can be useful for
to determine whether the relationship of expression and cell association in multiple both therapeutic and diagnostic applications
SLC46A3 expression extends to LNP associ- LNP formulations supports the relevance of (66, 67), and we believe that additional bio-
ation as well as transfection efficiency. We SLC46A3 as a predictive biomarker for lipid- markers that mediate the trafficking of in-
generated fluorescently (Cy5) labeled LNPs based NP formulations. organic NPs may be identified by using similar
that contained mRNA encoding green fluo- screening approaches. Second, the results of
rescent protein (GFP) (LNP 1) and incubated Discussion in vitro screens are often met with limited
these particles with engineered LOXIMVI cell This work represents high-throughput inter- success when translated in vivo because NP-
lines (tables S3 and S5). rogation of NP–cancer cell interactions through mediated delivery is dependent on many

A untreated
be possible to reanalyze our data in the future. We present a platform to study NP–cancer
control Especially for emerging fields such as pro- cell interactions simultaneously through the
SLC46A3 OE teomics and metabolomics, the opportunity use of pooled screening, genomics, and machine
5
Increasing to intersect NP delivery metrics with addi- learning algorithms. Application of this inte-
10
transfection tional datasets could add a new dimension grated platform should advance the rational
to our existing findings. design of nanocarriers.
4
Cy5 signal
10
One strength of our screening approach is
Increasing
the use of robust analytical tools, such as Materials and methods
LNP association
10
3 univariate analyses and random forest algo- Extended materials and methods are available
rithms, which enabled us to identify biomark- in the supplementary materials (29).
0
ers correlated with NP association. The robust
3 and quantitative manner in which we detected Base liposome synthesis
-10
3 4 5
EGFR hits for antibodies as well as antibody- A thin film was generated from a lipid mixture
0 10 10 10
GFP signal targeted NPs shows the utility of this platform composed of 31 mol % 1,2-distearoyl-sn-glycero-
for the development and optimization of tar- 3-phosphocholine (DSPC), 31 mol % cholesterol,
B C
control
SLC46A3 OE geted drug delivery platforms, including anti- 31 mol % 1,2-distearoyl-sn-glycero-3-phospho-
**
0.008
60
0.056 body-targeted NPs, and its potential to apply (1′-rac-glycerol) (DSPG), and 6 mol % 1,2-
**
0.008
LNP-associated fluorescence
to other targeted therapeutics, including ADCs. distearoyl-sn-glycero-3-phosphoethanolamine

Normalized transfection
200
40 0.095 This method of analysis could provide thera- (DSPE) (Avanti) and rehydrated to 2 mg/ml
peutic insights in the design of ADCs, specif- under heat (65°C) and sonication. The liposome
100 ically in evaluating the effects of conjugation suspension was extruded by using an Avestin
20
site or linker chemistry. LiposoFast LF-50 liposome extruder to a diam-
By clustering NP-specific biomarkers across eter of 50 to 100 nm. Liposomes were fluo-
0
0 formulations, we constructed interaction rescently labeled through N-hydroxysuccinimide
LNP 1 LNP 2 LNP 3
LNP 1 networks, identifying and connecting genes (NHS)–coupling of sulfo-cyanine NHS ester
Fig. 7. Solid lipid NP-cell association and transfec-
associated with NP binding, recognition, and dye to DSPE headgroups according to the dye
tion are SLC46A3-dependent, as determined
subcellular trafficking. This provides the sci- manufacturer (Lumiprobe) instructions. Lipid
with flow cytometry. (A) Contour plot of Cy5 signal
entific community with a blueprint for the film generation, rehydration, extrusion, and
and GFP signal indicating decreased LNP-cell associa-
fundamental study of cellular processes that dye labeling steps were similarly applied to all
tion and transfection efficacy in LOXIMVI cells over-
mediate NP engagement, with applications for liposome formulations unless noted otherwise.
expressing SLC46A3. (B) Quantification of LNP signal
both basic and translational research.
We identified expression of SLC46A3, a Tangential flow filtration (TFF)
revealed a significant change in LNP-cell association
across control and SLC46A3-overexpressing LOXIMVI
lysosomal transporter, to be a negative regu- To remove excess dye, crude NP solution was
cells (**P = 0.008, Mann-Whitney). LNP-associated
lator and potential biomarker for lipid-based connected to a Spectrum Labs KrosFlo II sys-
fluorescence is defined as median fluorescence
NP uptake and downstream functional effi- tem by using masterflex, Teflon-coated tubing.
intensity normalized to untreated cells. (C) LOXIMVI-
cacy. Although SLC46A3 has recently been D02-E100-05-N membranes were used to pu-
SLC46A3 OE cells exhibited lower transfection
implicated in hepatic copper homeostasis rify the particles until dye was no longer seen
efficiency than that of LOXIMVI-vector control cells
as well as sensitivity to ADCs in cancer cells in the permeate. Samples were run at flow rates
after dosing of three different LNP formulations
(54–56), its role in NP delivery was previously of 80 ml/min with size 16 tubing. Phosphate-
(Mann-Whitney). Normalized transfection is defined as
unexplored. We first validated SLC46A3 as buffered saline (PBS) was used as the exchange
median GFP intensity normalized to untreated cells.
a negative regulator of lipid-based NP uptake buffer for the first five washes followed by
in a panel of nonpooled cell lines, as well as milliQ water for the rest of the purification
engineered isogenic cell lines with modulated steps. After TFF, liposomes were characterized
SLC46A3 expression. Because all current FDA- by means of dynamic light scattering (DLS).
factors beyond the nano-cell interface (8). approved NPs for anticancer applications For layer-by-layer (LbL) synthesis, TFF was
However, the level of molecular characteriza- are liposomal formulations, there is notable used for purification after deposition of each
tion and statistical and computational power potential for this biomarker to be quickly im- polyelectrolyte layer, following the above pro-
afforded by annotated biological datasets, such plemented in clinical studies with existing, cedure. Instead of PBS, only milliQ water was
as the CCLE, is currently unrivaled. Therefore, approved formulations. To this end, we re- passed through the TFF for LbL NP purification.
existing in vivo screens cannot yet provide this capitulated our findings in an in vivo model
breadth or statistical power. Keeping transla- using an analog of an FDA-approved liposomal PLGA NP synthesis
tional barriers in mind is key to the successful NP formulation. PLGA (Sigma Aldrich) was dissolved at a con-
validation of candidate biomarkers, and for Moreover, we demonstrated that SLC46A3 centration of 10 mg/ml in acetone, and Cy5
this reason, we used multiple isogenic models has potential as a predictive biomarker beyond free acid dye (Lumiprobe) was dissolved at a
and tested a range of lipid-based NPs across liposomal NPs by investigating solid lipid NPs. concentration of 50 mg/ml in dimethyl sulfox-
in vitro and in vivo conditions. Third, an addi- Both LNP-cell association and mRNA transfec- ide (DMSO). 6 ml milliQ water were added to a
tional limitation of this screen is related to the tion were inversely correlated with SLC46A3 scintillation vial and stirred gently on a stir
availability of genomic datasets for each cell expression. These preliminary findings sug- plate; 2 ml dye were mixed with 1 ml PLGA
line tested because dataset completeness con- gest that SLC46A3 expression may serve as a solution and drawn up in a syringe with a
tributes to the power of detection for both predictive biomarker for functional delivery 27-gauge needle attached. The PLGA-Cy5
univariate and multivariate analyses. At the of nucleic acid cargo through lipid NPs. Our solution was slowly added to the water under
time of analysis, 10 feature sets were avail- findings support the continued exploration of constant stirring and left to stir 3 hours. An
able for the majority of cell lines in our pool. SLC46A3 as a potential biomarker for ther- additional 2 ml milliQ water were added the
However, as datasets expand over time, it will apeutic NP delivery. solution before purification by using TFF.

Synthesis of layer-by-layer NPs warm PBS and dissociated with 0.25% Trypsin- of studies were initially determined by use of
Liposomes and PLGA NPs were layered by EDTA before quenching with cell culture me- G*Power.
adding equal volumes of NP solution to poly- dium. Cells were placed on ice until analyzed For intratumoral studies, mice with estab-
electrolyte solution under sonication. Poly- by using a high-throughput analyzer. lished flank tumors were randomly assigned
electrolyte solutions for liposome layering, to either the 4- or 24-hour dosing cohort. After
with the exception of HA and alginate, were SLC46A3 overexpression: Viral transfection of injection, mice were imaged by using the
prepared in 50 mM hydroxyethylpiperazine LOXIMVI cells In Vivo Imaging System (IVIS) Spectrum
ethane sulfonic acid (HEPES) and 40 mM Lentiviral vectors were purchased from the whole-animal imaging device (PerkinElmer)
NaCl (pH 7.4). HA and alginate stocks were Broad Institute’s Genetic Perturbation Plat- using ex = 640/em = 700 nm to capture Cy5
prepared in 10 mM HEPES. All polyelectrolyte form (GPP), specifically ccsbBroad304_09945 signal. Immediately after imaging, mice were
solutions for PLGA NP layering were prepared (SLC46A3) and ccsbBroad304_99991 (Luciferase, humanely euthanized, and tumors were ex-
in water. Layered particles were incubated at vector control). LOXIMVI cells were trypsi- cised and imaged again with IVIS. For intra-
room temperature for 1 hour before being nized, counted, and resuspended to a concen- venous studies, NPs were administered to mice
purified by means of TFF and characterized tration of 1.36 × 106 cells/ml. A solution of 2X by using tail vein injections, each of three doses
by means of DLS. polybrene was added to the cell suspension so spaced 24 hours apart. Four hours after the
that the final concentration of polybrene was third and final injection, mice were humanely
Pooled PRISM cell dosing with NPs and 8 mg/ml. Cells were seeded into two six-well euthanized, and tumors were excised and
preparation for flow cytometry plates at 750,000 cells/well. Lentiviral vectors imaged with IVIS. Tumors were weighed, and
Cells were seeded at 200,000 cells/well in 0.5 ml were separately added to plates at six different their weights were recorded for normalization
RPMI-1640 medium supplemented with 10% doses: 0, 25, 50, 100, 200, and 400 ml. After, of tumor fluorescence by tumor mass.
fetal bovine serum (FBS) in a 12-well plate. 1 ml medium was added to each well, the cells
Cells were allowed to grow for 24 hours before were incubated overnight, and the medium Statistical analysis
treatment with NPs for 4 or 24 hours. After was changed at 17 hours after seeding. At All statistical analysis for nonpooled validation
incubation, cells were washed once with warm 48 hours after seeding, the cells were reseeded studies was performed by using GraphPad
PBS and dissociated with 0.25% Trypsin-EDTA. at 375,000 cells/well in 2 ml blasticidin con- PRISM 9. Detailed statistical information is
After 5 min at 37°C, the trypsin quenched with taining medium (final blasticidin concentra- provided for each figure in the associated cap-
cell culture medium. Cells were then trans- tion was 1 mg/ml). The selection progress was tion. Unless noted otherwise, for single com-
ferred to a FACS tube through a cell strainer monitored with flow cytometry (fig. S16). parisons (nonparametric), the Mann-Whitney
cap and placed on ice until sorting. test was used. For multiple comparison testing,
SLC46A3 permanent deactivation with the Kruskall-Wallis test was used to compare
Pooled PRISM cell dosing with antibodies CRISPR-Cas9 in T47D cells treatment groups with the parental control. The
Cells were washed and dissociated with StemPro SLC46A3-deactivated T47D cell lines were gen- datasets and code pertaining to nanoPRISM
Accutase. After incubation, cold FACS buffer erated through infection with lentiCRISPRv2- probabilistic model development and subse-
(PBS + 2% FBS) was added to each well, and Opti (Addgene 163126) vectors encoding Cas9 quent analyses (univariate and random forest)
cells were triturated and centrifuged. After and single-guide RNAs (sgRNAs) (68). The fol- are available on Zenodo (69–75).
spinning, cells were resuspended in FACS lowing oligonucleotides were used for sgRNA
REFERENCES AND NOTES
buffer at a concentration of 1 × 106 cells/ml. cloning and include cloning overhangs for liga-
1. J. Shi, P. W. Kantoff, R. Wooster, O. C. Farokhzad, Cancer
The cell solution was split into four groups: tion after BsmBI digest of lentiCRISPRv2-Opti nanomedicine: Progress, challenges and opportunities.
untreated control, (+) 15 ml 0.1 mg/ml Cy5- vector: sgGFP_F, caccGGGCGAGGAGCTGTT- Nat. Rev. Cancer 17, 20–37 (2017). doi: 10.1038/nrc.2016.108;
cetuximab, (+)15 ml 0.1 mg/ml Cy5-IgG, and (+) CACCG; sgGFP_R, aaacCGGTGAACAGCTC- pmid: 27834398
2. M. J. Mitchell et al., Engineering precision nanoparticles for
5 ml of EGFR-AF488 (used at undiluted stock CTCGCCC; sgSLC46A3_F, caccgAAAGCA- drug delivery. Nat. Rev. Drug Discov. 20, 101–124 (2021).
concentration provided by manufacturer, AGCTCCCCAAAATG; and sgSLC46A3_R, doi: 10.1038/s41573-020-0090-8; pmid: 33277608
InvivoGen). Samples were incubated in the aaacCATTTTGGGGAGCTTGCTTTc. 3. N. Boehnke, P. T. Hammond, Power in numbers: Harnessing
combinatorial and integrated screens to advance
dark at 4°C for 1 hour. Cells were then washed Clonal deactivation cell lines were isolated nanomedicine. JACS Au 2, 12–21 (2021). doi: 10.1021/
and resuspended in cold FACS buffer. through FACS, and biallelic frame-shifts were jacsau.1c00313; pmid: 35098219
confirmed with deep-sequencing [allele 1, 4. S. Tran, P. J. DeGiovanni, B. Piel, P. Rai, Cancer nanomedicine:
SLC46A3 validation studies A review of recent success in drug delivery. Clin. Transl. Med. 6, 44
–32 base pairs (bp) frameshift 501 reads; (2017). doi: 10.1186/s40169-017-0175-0; pmid: 29230567
Nonpooled screening allele 2, –10 bp frameshift; 477 reads). The T47D 5. S. Wilhelm et al., Analysis of nanoparticle delivery to tumours.
HCC1143 (RPMI-1640), HCC1395 (RPMI-1640), SLC46A3-deactivation line described has the Nat. Rev. Mater. 1, 16014 (2016). doi: 10.1038/
natrevmats.2016.14
HeLa (RPMI-1640), SW948 (RPMI-1640), mutant alleles c.442_453del and c.440_449del.
6. Y. H. Cheng, C. He, J. E. Riviere, N. A. Monteiro-Riviere, Z. Lin,
LOXIMVI (RPMI-1640), SJSA-1 (RPMI- Meta-analysis of nanoparticle delivery to tumors using a
1640), MCF7 [Eagle’s minimum essential Animal studies physiologically based pharmacokinetic modeling and
medium (EMEM)], DAOY (EMEM), MDA- All animal experiments were approved by the simulation approach. ACS Nano 14, 3075–3095 (2020).
doi: 10.1021/acsnano.9b08142; pmid: 32078303
MB-231 [Dulbecco’s modified Eagle's medium Massachusetts Institute of Technology Com- 7. Y. S. Youn, Y. H. Bae, Perspectives on the past, present, and
(DMEM)], CAOV3 (DMEM), T47D (RPMI- mittee on Animal Care (CAC; protocol num- future of cancer nanomedicine. Adv. Drug Deliv. Rev. 130, 3–11
1640), and HepG2 (DMEM) cells were seeded ber 0821-052-04) and were conducted under (2018). doi: 10.1016/j.addr.2018.05.008; pmid: 29778902
8. W. Poon, B. R. Kingston, B. Ouyang, W. Ngo, W. C. W. Chan,
individually at 10,000 cells/well in 100 ml the oversight of the Division of Comparative A framework for designing delivery systems. Nat. Nanotechnol.
medium, supplemented with 10% FBS and Medicine (DCM). Flank tumors of LOXIMVI- 15, 819–829 (2020). doi: 10.1038/s41565-020-0759-5;
1X Penicillin-Streptomycin. Cells were allowed vector control and LOXIMVI-SLC46A3 OE cells pmid: 32895522
9. W. Poon et al., Elimination pathways of nanoparticles. ACS
to grow overnight before treatment with NPs. were established with a subcutaneous injec- Nano 13, 5785–5798 (2019). doi: 10.1021/acsnano.9b01383;
Before dosing, all NP formulations were nor- tion of 0.5 × 106 to 1.0 × 106 cells as a 1:1 mix- pmid: 30990673
malized to a concentration of 50 mg/ml. Cells ture with MatriGel (Corning) and PBS to the 10. S. Correa et al., Tuning nanoparticle interactions with ovarian
cancer through layer-by-layer modification of surface
were dosed with 10 ml normalized NP solutions. right flank of NCr nude mice (5 to 7 weeks, chemistry. ACS Nano 14, 2224–2237 (2020). doi: 10.1021/
After incubation, cells were washed once with Taconic, NCRNU-F, NCRNU-M). Sample sizes acsnano.9b09213; pmid: 31971772

11. N. Boehnke et al., Theranostic layer-by-layer nanoparticles for Commun. 10, 2341 (2019). doi: 10.1038/s41467-019-10112-4; 56. K. Kinneer et al., SLC46A3 as a potential predictive biomarker
simultaneous tumor detection and gene silencing. Angew. pmid: 31138801 for antibody-drug conjugates bearing noncleavable linked
Chem. Int. Ed. 59, 2776–2783 (2020). doi: 10.1002/ 34. J. Barretina et al., The Cancer Cell Line Encyclopedia enables maytansinoid and pyrrolobenzodiazepine warheads. Clin.
anie.201911762; pmid: 31747099 predictive modelling of anticancer drug sensitivity. Nature Cancer Res. 24, 6570–6582 (2018). doi: 10.1158/1078-0432.
12. N. Boehnke, K. J. Dolph, V. M. Juarez, J. M. Lanoha, 483, 603–607 (2012). doi: 10.1038/nature11735; CCR-18-1300; pmid: 30131388
P. T. Hammond, Electrostatic conjugation of nanoparticle pmid: 22460905 57. Q. Zhao et al., Increased expression of SLC46A3 to oppose the
surfaces with functional peptide motifs. Bioconjug. Chem. 31, 35. M. Ghandi et al., Next-generation characterization of the progression of hepatocellular carcinoma and its effect on
2211–2219 (2020). doi: 10.1021/acs.bioconjchem.0c00384; Cancer Cell Line Encyclopedia. Nature 569, 503–508 (2019). sorafenib therapy. Biomed. Pharmacother. 114, 108864 (2019).
pmid: 32786506 doi: 10.1038/s41586-019-1186-3; pmid: 31068700 doi: 10.1016/j.biopha.2019.108864; pmid: 30981107
13. J. E. Dahlman et al., Barcoded nanoparticles for high 36. K. Tsuchikama, Z. An, Antibody-drug conjugates: Recent 58. G. Li et al., Mechanisms of acquired resistance to trastuzumab
throughput in vivo discovery of targeted therapeutics. Proc. advances in conjugation and linker chemistries. Protein Cell 9, emtansine in breast cancer cells. Mol. Cancer Ther. 17,
Natl. Acad. Sci. U.S.A. 114, 2060–2065 (2017). doi: 10.1073/ 33–46 (2018). doi: 10.1007/s13238-016-0323-0; 1441–1453 (2018). doi: 10.1158/1535-7163.MCT-17-0296;
pnas.1620874114; pmid: 28167778 pmid: 27743348 pmid: 29695635
14. B. Nogrady, How cancer genomics is transforming diagnosis 37. J. Rejman, V. Oberle, I. S. Zuhorn, D. Hoekstra, Size-dependent 59. C. K. Tsui et al., CRISPR-Cas9 screens identify regulators of
and treatment. Nature 579, S10–S11 (2020). doi: 10.1038/ internalization of particles via the pathways of clathrin- and antibody-drug conjugate toxicity. Nat. Chem. Biol. 15,
d41586-020-00845-4; pmid: 32214255 caveolae-mediated endocytosis. Biochem. J. 377, 159–169 949–958 (2019). doi: 10.1038/s41589-019-0342-2;
15. C. Yu et al., High-throughput identification of genotype-specific (2004). doi: 10.1042/bj20031253; pmid: 14505488 pmid: 31451760
cancer vulnerabilities in mixtures of barcoded tumor cell lines. 38. S. Behzadi et al., Cellular uptake of nanoparticles: Journey 60. X. Hou, T. Zaks, R. Langer, Y. Dong, Lipid nanoparticles for
Nat. Biotechnol. 34, 419–423 (2016). doi: 10.1038/nbt.3460; inside the cell. Chem. Soc. Rev. 46, 4218–4244 (2017). mRNA delivery. Nat. Rev. Mater. 6, 1078–1094 (2021).
pmid: 26928769 doi: 10.1039/C6CS00636A; pmid: 28585944 doi: 10.1038/s41578-021-00358-0; pmid: 34394960
16. S. M. Corsello et al., Discovering the anticancer potential of 39. A. Subramanian et al., Gene set enrichment analysis: 61. E. Samaridou, J. Heyes, P. Lutwyche, Lipid nanoparticles for
non-oncology drugs by systematic viability profiling. Nature A knowledge-based approach for interpreting genome-wide nucleic acid delivery: Current perspectives. Adv. Drug Deliv.
Cancer, (2020). expression profiles. Proc. Natl. Acad. Sci. U.S.A. 102, Rev. 154-155, 37–63 (2020). doi: 10.1016/j.addr.2020.06.002;
17. S. Correa, N. Boehnke, E. Deiss-Yehiely, P. T. Hammond, 15545–15550 (2005). doi: 10.1073/pnas.0506580102; pmid: 32526452
Solution conditions tune and optimize loading of therapeutic pmid: 16199517 62. M. Jayaraman et al., Maximizing the potency of siRNA lipid
polyelectrolytes into layer-by-layer functionalized liposomes. 40. A. Liberzon et al., Molecular signatures database (MSigDB) 3.0. nanoparticles for hepatic gene silencing in vivo. Angew. Chem.
ACS Nano 13, 5623–5634 (2019). doi: 10.1021/ Bioinformatics 27, 1739–1740 (2011). doi: 10.1093/ Int. Ed. 51, 8529–8533 (2012). doi: 10.1002/anie.201203263;
acsnano.9b00792; pmid: 30986034 bioinformatics/btr260; pmid: 21546393 pmid: 22782619
18. S. Correa et al., Highly scalable, closed-loop synthesis of drug- 41. A. Liberzon et al., The Molecular Signatures Database 63. K. A. Whitehead et al., Degradable lipid nanoparticles
loaded, layer-by-layer nanoparticles. Adv. Funct. Mater. 26, (MSigDB) hallmark gene set collection. Cell Syst. 1, 417–425 with predictable in vivo siRNA delivery activity. Nat.
991–1003 (2016). doi: 10.1002/adfm.201504385; (2015). doi: 10.1016/j.cels.2015.12.004; pmid: 26771021 Commun. 5, 4277 (2014). doi: 10.1038/ncomms5277;
pmid: 27134622 42. M. Dean, Y. Hamon, G. Chimini, The human ATP-binding pmid: 24969323
19. Z. J. Deng et al., Layer-by-layer nanoparticles for systemic cassette (ABC) transporter superfamily. J. Lipid Res. 42, 64. K. J. Kauffman et al., Optimization of lipid nanoparticle
codelivery of an anticancer drug and siRNA for potential 1007–1017 (2001). doi: 10.1016/S0022-2275(20)31588-1; formulations for mRNA delivery in vivo with fractional
triple-negative breast cancer treatment. ACS Nano 7, pmid: 11441126 factorial and definitive screening designs. Nano Lett. 15,
9571–9584 (2013). doi: 10.1021/nn4047925; pmid: 24144228 43. Y. Shamay et al., P-selectin is a nanotherapeutic delivery target 7300–7306 (2015). doi: 10.1021/acs.nanolett.5b02497;
20. S. W. Morton, Z. Poon, P. T. Hammond, The architecture and in the tumor microenvironment. Sci. Transl. Med. 10, 345ra87 pmid: 26469188
biological performance of drug-loaded LbL nanoparticles. (2018). pmid: 27358497 65. K. J. Hassett et al., Optimization of lipid nanoparticles for
Biomaterials 34, 5328–5335 (2013). doi: 10.1016/ 44. G. Saravanakumar, D. G. Jo, J. H. Park, Polysaccharide-based intramuscular administration of mRNA vaccines. Mol. Ther.
j.biomaterials.2013.03.059; pmid: 23618629 nanoparticles: A versatile platform for drug delivery and Nucleic Acids 15, 1–11 (2019). doi: 10.1016/j.omtn.2019.01.013;
21. G. Decher, Fuzzy nanoassemblies: Toward layered polymeric biomedical imaging. Curr. Med. Chem. 19, 3212–3229 (2012). pmid: 30785039
multicomposites. Science 277, 1232–1237 (1997). doi: 10.2174/092986712800784658; pmid: 22612705 66. Q. Q. Liu et al., Inorganic nanoparticles applied as functional
doi: 10.1126/science.277.5330.1232 45. J. Voigt, J. Christensen, V. P. Shastri, Differential uptake of therapeutics. Adv. Funct. Mater. 31, 2008171 (2021).
22. E. C. Dreaden et al., Tumor-targeted synergistic blockade of nanoparticles by endothelial cells through polyelectrolytes with 67. W. Paul, C. P. Sharma, Inorganic nanoparticles for targeted
MAPK and PI3K from a layer-by-layer nanoparticle. Clin. Cancer affinity for caveolae. Proc. Natl. Acad. Sci. U.S.A. 111, 2942–2947 drug delivery. Biointegration Med. Implant. Mater. 2020,
Res. 21, 4410–4419 (2015). doi: 10.1158/1078-0432.CCR-15- (2014). doi: 10.1073/pnas.1322356111; pmid: 24516167 334–373 (2020).
0013; pmid: 26034127 46. D. Szklarczyk et al., STRING v11: Protein-protein association 68. C. H. Adelmann et al., MFSD12 mediates the import of
23. E. C. Dreaden et al., Bimodal tumor-targeting from networks with increased coverage, supporting functional cysteine into melanosomes and lysosomes. Nature 588,
microenvironment responsive hyaluronan layer-by-layer (LbL) discovery in genome-wide experimental datasets. Nucleic Acids 699–704 (2020). doi: 10.1038/s41586-020-2937-x;
nanoparticles. ACS Nano 8, 8374–8382 (2014). doi: 10.1021/ Res. 47, D607–D613 (2019). doi: 10.1093/nar/gky1131; pmid: 33208952
nn502861t; pmid: 25100313 pmid: 30476243 69. N. Boehnke, J. P. Straehla, M. Kocak, M. Ronan, J. Roth, Data
24. O. P. Oommen, C. Duehrkop, B. Nilsson, J. Hilborn, 47. C. von Mering et al., STRING: A database of predicted functional and processing scripts for PRISM barcode sequencing data
O. P. Varghese, Multifunctional hyaluronic acid and chondroitin associations between proteins. Nucleic Acids Res. 31, 258–261 used in “Massively parallel pooled screening reveals genomic
sulfate nanoparticles: Impact of glycosaminoglycan (2003). doi: 10.1093/nar/gkg034; pmid: 12519996 determinants of nanoparticle-cell interactions”. Zenodo (2022);
presentation on receptor mediated cellular uptake and immune 48. B. Snel, G. Lehmann, P. Bork, M. A. Huynen, STRING: doi: 10.5281/zenodo.6642633
activation. ACS Appl. Mater. Interfaces 8, 20614–20624 (2016). A web-server to retrieve and display the repeatedly 70. M. Stephens, False discovery rates: A new deal. Biostatistics
doi: 10.1021/acsami.6b06823; pmid: 27468113 occurring neighbourhood of a gene. Nucleic Acids Res. 28, 18, 275–294 (2017). pmid: 27756721
25. J. S. Suk, Q. Xu, N. Kim, J. Hanes, L. M. Ensign, PEGylation as a 3442–3444 (2000). doi: 10.1093/nar/28.18.3442; 71. R Core Team, R: A language and envrionemnt for statistical
strategy for improving nanoparticle-based drug and gene pmid: 10982861 computing. R Foundation for Statistical Compouting;
delivery. Adv. Drug Deliv. Rev. 99 (Pt A), 28–51 (2016). 49. D. Martin et al., GOToolBox: Functional analysis of gene https://www.R-project.org.
doi: 10.1016/j.addr.2015.09.012; pmid: 26456916 datasets based on Gene Ontology. Genome Biol. 5, R101 (2004). 72. H. Wickham, ggplot2: Elegant graphics for data analysis. Use R,
26. E. Fröhlich, The role of surface charge in cellular uptake and doi: 10.1186/gb-2004-5-12-r101; pmid: 15575967 1-212 (2009).
cytotoxicity of medical nanoparticles. Int. J. Nanomedicine 50. M. Ashburner et al., Gene Ontology: Tool for the unification of 73. M. Kocak, A. Boghossian, Linear association function used in
7, 5577–5591 (2012). doi: 10.2147/IJN.S36111; pmid: 23144561 biology. Nat. Genet. 25, 25–29 (2000). doi: 10.1038/75556; the manuscript “Massively parallel pooled screening reveals
27. E. A. Berg, J. B. Fishman, Labeling antibodies using pmid: 10802651 genomic determinants of nanoparticle-cell interactions”.
N-hydroxysuccinimide (NHS)–fluorescein. Cold Spring Harb. 51. S. Carbon et al., The Gene Ontology resource: Enriching a GOld Zenodo (2022); doi: 10.5281/zenodo.6558445
Protoc. 2019, 229–231 (2019). pmid: 30824621 mine. Nucleic Acids Res. 49 (D1), D325–D334 (2021). 74. Y. Tang, M. Horikoshi, W. X. Li, ggfortify: Unified interface to
28. X. Jin et al., A metastasis map of human cancer cell lines. Nature
doi: 10.1093/nar/gkaa1113; pmid: 33290552 visualize statistical results of popular R packages. R J. 8,
588, 331–336 (2020). doi: 10.1038/s41586-020-2969-2;
52. L. Lin, S. W. Yee, R. B. Kim, K. M. Giacomini, SLC transporters as 474–485 (2016). doi: 10.32614/RJ-2016-060
pmid: 33299191
therapeutic targets: Emerging opportunities. Nat. Rev. Drug Discov. 75. M. Horikoshi, Y. Tang, ggfortify: Data visualization tools for
29. Materials and methods are available as supplementary materials.
14, 543–560 (2015). doi: 10.1038/nrd4626; pmid: 26111766 statistical analysis results(2016); https://CRAN.R-project.org/
30. J. A. Kim, C. Åberg, A. Salvati, K. A. Dawson, Role of cell cycle
53. A. Chapel et al., An extended proteome map of the lysosomal package=ggfortify.
on the cellular uptake and dilution of nanoparticles in a cell
population. Nat. Nanotechnol. 7, 62–68 (2011). doi: 10.1038/ membrane reveals novel potential transporters. Mol. Cell.
Proteomics 12, 1572–1588 (2013). doi: 10.1074/mcp. AC KNOWLED GME NTS
nnano.2011.191; pmid: 22056728
31. C. Åberg, J. A. Kim, A. Salvati, K. A. Dawson, Reply to ‘The M112.021980; pmid: 23436907 We thank the Koch Institute’s Robert A. Swanson (1969) Biotechnology
interface of nanoparticles with proliferating mammalian cells’. 54. J. H. Kim et al., Lysosomal SLC46A3 modulates hepatic Center for technical support, specifically the Flow Cytometry,
Nat. Nanotechnol. 12, 600–603 (2017). doi: 10.1038/ cytosolic copper homeostasis. Nat. Commun. 12, 290 (2021). High Throughput Sciences, Genomics Core, Microscopy, and
nnano.2017.139; pmid: 28681851 doi: 10.1038/s41467-020-20461-0; pmid: 33436590 Preclinical Modeling, Imaging and Testing cores; the Hope Babette
32. E. Panet et al., The interface of nanoparticles with proliferating 55. K. J. Hamblett et al., SLC46A3 is required to transport Tang (1983) Histology Facility; and the Peterson (1957)
mammalian cells. Nat. Nanotechnol. 12, 598–600 (2017). catabolites of noncleavable antibody maytansine conjugates Nanotechnology Materials Core Facility. We thank T. Golub and
doi: 10.1038/nnano.2017.140; pmid: 28681852 from the lysosome to the cytoplasm. Cancer Res. 75, A. Burgin for formative feedback and helpful discussion. We also
33. P. Rees, J. W. Wills, M. R. Brown, C. M. Barnes, H. D. Summers, 5329–5340 (2015). doi: 10.1158/0008-5472.CAN-15-1610; gratefully acknowledge T. Diefenbach and the Ragon Institute
The origin of heterogeneous nanoparticle uptake by cells. Nat. pmid: 26631267 Microscopy Core for assistance with imaging cytometry. The

LOXIMVI and T47D cell lines were gifts from the F. Gertler laboratory of Medical Engineering and Science. N.G.L. was supported materials availability: All data are available in the main text or the
(MIT), and the Jurkat cell line was a gift from the D. Sabatini by Cancer Research UK and the Brain Tumour Charity (C42454/ supplementary materials. License information: Copyright © 2022
laboratory (previously of the Whitehead Institute for Biomedical A28596) and a fellowship from the Ludwig Center at the Koch the authors, some rights reserved; exclusive licensee American
Research). We thank C. Straehla for help in figure design. Figures 1A Institute for Integrative Cancer Research. H.L. was supported by Association for the Advancement of Science. No claim to original US
and 6A were created in part by use of Biorender.com. Funding: This the Charles W. (1955) and Jennifer C. Johnson Cancer Research Fund government works. https://www.science.org/about/science-
work was supported in part by SPARC funding at The Broad and NIH (K08DK123414). Author contributions: Conceptualization: licenses-journal-article-reuse
Institute. This work was also supported by grants from the Koch N.B. and J.P.S. Methodology: N.B., J.P.S., and M.K. Formal Analysis:
Institute’s Marble Center for Cancer Nanomedicine and Frontier N.B., J.P.S., M.K., M.G.R., and M.R. Investigation: N.B., J.P.S., SUPPLEMENTARY MATERIALS
Research Program. This work was supported in part by the Koch H.C.S., M.G.R., D.R., N.N., A.G.B., and N.G.L. Visualization: N.B. and
science.org/doi/10.1126/science.abm5551
Institute Support (core) Grant P30-CA14051 from the National J.P.S. Funding acquisition: N.B., J.P.S., A.N.K., and P.T.H. Project
Materials and Methods
Cancer Institute. N.B. was supported by a US Department of administration: N.B., J.P.S., and M.R. Validation: N.B., J.P.S., H.C.S.,
Figs. S1 to S21
Defense Congressionally Directed Medical Research Programs C.H.A., R.R.C., J.H.C., and H.L. Supervision: J.A.R., A.N.K., and
Tables S1 to S5
Peer Reviewed Cancer Research Program Horizon Award (W81XWH- P.T.H. Writing – original draft: N.B. and J.P.S. Writing – review and
Reference (76)
19-1-0257) and the NIH-NCI (K99CA255844). J.P.S. was supported editing: N.B., J.P.S., H.C.S., M.K., M.G.R., M.R., C.H.A., R.R.C., N.N.,
MDAR Reproducibility Checklist
as a National Institutes of Health (NIH) grant T32 trainee A.G.B., N.G.L., J.H.C., H.L., J.A.R., A.N.K., and P.T.H. Competing
Data S1 to S2
(CA136432-08) and by the Helen Gurley Brown Presidential interests: N.B., J.P.S., A.N.K., and P.T.H. have submitted a patent
Initiative of Dana-Farber Cancer Institute. Fellowship support for application for the SLC46A3 biomarker discovery. A.N.K. is a founder
C.H.A. was from the NIH (NRSA F31 CA228241-01). R.R.C. is a and member of the board of 76Bio. P.T.H. is a member of the
fellow of the Parker B. Francis Foundation. N.N. was supported by science advisory board at Moderna; board member at Alector, Submitted 24 September 2021; resubmitted 14 March 2022
a grant from the Gates Foundation. Fellowship support for A.G.B. Advanced Chemotherapy Technologies, and Burroughs-Wellcome Accepted 16 June 2022
was from the NIH (F30 DK130564) and a Termeer Fellowship Fund; and board member and cofounder of LayerBio. Data and 10.1126/science.abm5551

Publish your research in the Science family of journals
The Science family of journals (Science, Science Advances, Science Immunology, Science
Robotics, Science Signaling, and Science Translational Medicine) are among the most highly-
regarded journals in the world for quality and selectivity. Our peer-reviewed journals are
committed to publishing cutting-edge research, incisive scientific commentary, and insights
on what’s important to the scientific world at the highest standards.
Submit your research today!

Learn more at Science.org/journals
RESEAR CH
◥ This mode contrasted with “extended-surface

RESEARCH ARTICLE SUMMARY side”-binding for heparin sugar ligands. uSTA-
derived restraints used in structural modeling
CORONAVIRUS suggested sialoside-glycan binding sites in a b
sheet–rich region of spike N-terminal domain
Pathogen-sugar interactions revealed by universal (NTD), distant from the receptor-binding do-
main (RBD) that binds ACE2 co-receptor and
saturation transfer analysis that has been identified as the site for other
sugar interactions. Consistent with this NTD
Charles J. Buchanan†, Ben Gaunt†, Peter J. Harrison, Yun Yang, Jiwei Liu, Aziz Khan, Andrew M. Giltrap, site being a previously unknown sialoside
Audrey Le Bas, Philip N. Ward, Kapil Gupta, Maud Dumoux, Tiong Kit Tan, Lisa Schimaski, sugar-binding pocket, uSTA-sialoside binding
Sergio Daga, Nicola Picchiotti, Margherita Baldassarri, Elisa Benetti, Chiara Fallerini, Francesca Fava, was minimally perturbed by antibodies that
Annarita Giliberti, Panagiotis I. Koukos, Matthew J. Davy, Abirami Lakshminarayanan, Xiaochao Xue, neutralize the ACE2-binding RBD domain.
Georgios Papadakis, Lachlan P. Deimel, Virgínia Casablancas-Antràs, Timothy D. W. Claridge, Strikingly, uSTA also shows that this sialoside
Alexandre M. J. J. Bonvin, Quentin J. Sattentau, Simone Furini, Marco Gori, Jiandong Huo, binding is disrupted in spike from multiple
Raymond J. Owens, Christiane Schaffitzel, Imre Berger, Alessandra Renieri, variants of concern (B1.1.7/alpha, B1.351/beta,
GEN-COVID Multicenter Study, James H. Naismith*, Andrew J. Baldwin*, Benjamin G. Davis* B.1.617.2/delta, and B.1.1.529/omicron) that
emerged later in the pandemic (September 2020
onward). Notably, these variants possess mul-
INTRODUCTION: The surface proteins found on NMR spectra (via a deconvolution algorithm) tiple hotspot mutations in the NTD. End-on
both pathogens and host cells mediate cell with Bloch-McConnell analysis of magnetization sialoside binding in a B-origin-lineage spike-
entry (and exit) and influence disease progres- transfer (as judged by this resonance signal in- NTD pocket was pinpointed by cryo-EM to a
sion and transmission. Both types of proteins tensity) to enable a structural, kinetic, and ther- previously unknown site that is created from
can bear host-generated posttranslational modynamic analysis of ligand binding. Such residues that are notably mutated or are in
modifications, such as glycosylation, that are quantification is beyond previously perceived regions where mutations occur in variants of
essential for function but can confound current limits of exchange rates, concentration, or system concern (e.g., His69, Val70, and Tyr145 in alpha
biophysical methods used for dissecting key and therefore represents a potentially universal and omicron). An analysis of beneficial genetic
interactions. Several human viruses (including saturation transfer analysis (uSTA) method. variances correlated with disease severity in
non-SARS coronaviruses) attach to host cell cohorts of patients from early 2020 suggests
surface N-linked glycans that include forms of RESULTS: In an automated workflow, uSTA can a model in which this site in the NTD of
sialic acid (sialosides). There remains, however, be applied to a range of even heavily modified B-origin-lineage SARS-CoV-2 (but not in later
conflicting evidence as to whether or how protein systems in a general manner to obtain variants) may have exploited a specific sialy-
SARS-associated coronaviruses might use such quantitative binding interaction parameters lated polylactosamine motif found on tetra-
a mechanism. In the absence of an appropriate (KD, kEx). uSTA proved critical in mapping antennary human N-linked glycoproteins,
biochemical assay, the ability to analyze the direct interactions between sialoside sugar known to be present in deeper human lung.
binding of such glycans to heavily modified ligands and relevant virus surface attachment
proteins and resolve this issue is limited. glycoproteins, including multiple variants of both CONCLUSION: Together, these results confirm a
severe acute respiratory syndrome coronavirus 2 distinctive sugar-binding mode mediated by the
RATIONALE: We developed and demonstrated (SARS-CoV-2) spike protein and influenza H1N1 unusual NTD of B-origin-lineage SARS-CoV-2
a quantitative extension of “saturation transfer” hemagglutinin protein. It was successful in spike protein that is lost in later variants. This
protein nuclear magnetic resonance (NMR) quantitating ligand NMR signals in spectral may implicate modulation of binding by SARS-
methods to a complete mathematical model of regions otherwise occluded by resonances from CoV-2 virus to human cell surface sugars as a
the magnetization transfer caused by inter- mobile protein glycans. In early-pandemic determinant of virulence and/or zoonosis. More
actions between protein and ligand. The de- (December 2019) B-origin-lineage SARS-CoV-2 generally, because cell surface glycans are widely
signed method couples objective resonance spike trimer, a clear “end-on” binding mode of relevant to biology and pathology, the uSTA
identification and intensity measurement in sialoside sugars to spike was revealed by uSTA. method can now provide ready, quantitative,
widespread analysis of complex, host-derived,
A Universal saturation transfer analysis (uSTA) NMR and posttranslationally modified proteins in
uSTA identifies a sugar-
Host their binding to putative ligands, which may
Binding pose
binding site and pose in cell

be relevant to disease, even in previously con-
▪
the unusual NTD of early-
pandemic SARS-CoV-2 founding complex systems.
spike. (A) This site and
NTD The list of author affiliations is available in the full article online.
pose, which were confirmed Sialic Sialic uSTA NMR cryo-EM
acid acid validation *Corresponding author. Email: [email protected]
by cryoÐelectron micros- (J.H.N.); [email protected] (A.J.B.);
copy, are heavily mutated B Variants lose binding [email protected] (B.G.D.)
†These authors contributed equally to this work.
in variants of concern. 1 Original KD 32 µm This is an open-access article distributed under the terms
(B) Analyses reveal a loss of of the Creative Commons Attribution license (https://
2 Alpha Mutations cluster
sialoside-spike binding, creativecommons.org/licenses/by/4.0/), which permits
near NTD binding Sialic unrestricted use, distribution, and reproduction in any
rationalized by clustering of 3 Beta site (cryo-EM)
Binding acid medium, provided the original work is properly cited.
mutations around the lost Cite this article as C. J. Buchanan et al., Science 377,
4 Delta Alpha*
binding site. *All alpha- Beta eabm3125 (2022). DOI: 10.1126/science.abm3125
5 Omicron
variant mutations are Delta
in >1 variant. Omicron READ THE FULL ARTICLE AT
uSTA sialic acid binding NTD >1 variant https://doi.org/10.1126/science.abm3125

RES EARCH
◥ in SARS-CoV-2, and yet neutralization by anti-

RESEARCH ARTICLE bodies against this domain suggests a poten-
tially important function in viral replication.
CORONAVIRUS The unresolved role of host cell-surface sialo-
sides for this pathogen has been noted as an
Pathogen-sugar interactions revealed by universal important open question (17). The hypothe-
sized roles for sugar interactions (20) in both
saturation transfer analysis virulence (21) and zoonosis (1) indicate that
there is an urgent need for precise, quantita-
Charles J. Buchanan1,2,3†, Ben Gaunt1†, Peter J. Harrison4,5, Yun Yang1,4, Jiwei Liu1, Aziz Khan1,2, tive, and robust methods for analysis.
Andrew M. Giltrap1,2, Audrey Le Bas1,4, Philip N. Ward4, Kapil Gupta6, Maud Dumoux1, Tiong Kit Tan7, In principle, magnetization transfer in pro-
Lisa Schimaski7, Sergio Daga8,9, Nicola Picchiotti10,11, Margherita Baldassarri8,9, Elisa Benetti9, tein nuclear magnetic resonance (NMR) spec-
Chiara Fallerini8,9, Francesca Fava8,9,12, Annarita Giliberti8,9, Panagiotis I. Koukos13, Matthew J. Davy1, troscopy could meet this need, as it can measure
Abirami Lakshminarayanan1,2, Xiaochao Xue2,14, Georgios Papadakis2, Lachlan P. Deimel14, ligand binding in its native state without the
Virgínia Casablancas-Antràs2,3, Timothy D. W. Claridge2, Alexandre M. J. J. Bonvin13, need for additional labeling or modification of
Quentin J. Sattentau14, Simone Furini9, Marco Gori10,15, Jiandong Huo1,4, Raymond J. Owens1,4, either ligand or protein (e.g., attachment to
Christiane Schaffitzel13, Imre Berger13, Alessandra Renieri8,9,12, GEN-COVID Multicenter Study, surface or sensor) (figs. S2 and S3; see text S1
James H. Naismith1,4*, Andrew J. Baldwin1,2,3*, Benjamin G. Davis1,2,16* for more details). Saturation transfer differ-
ence (STD) (22), which has been widely used
Many pathogens exploit host cell-surface glycans. However, precise analyses of glycan ligands binding to gauge qualitative ligand-protein interac-
with heavily modified pathogen proteins can be confounded by overlapping sugar signals and/or tions (23), detects the transfer of magnetization
compounded with known experimental constraints. Universal saturation transfer analysis (uSTA) while they are bound via “cross-relaxation.”
builds on existing nuclear magnetic resonance spectroscopy to provide an automated workflow for In reality, complex, highly modified protein
quantitating protein-ligand interactions. uSTA reveals that early-pandemic, B-origin-lineage severe acute systems have proven difficult to analyze in a
respiratory syndrome coronavirus 2 (SARS-CoV-2) spike trimer binds sialoside sugars in an “end-on” quantitative manner with current methods for
manner. uSTA-guided modeling and a high-resolution cryo–electron microscopy structure implicate the several reasons. First, mammalian proteins (or
spike N-terminal domain (NTD) and confirm end-on binding. This finding rationalizes the effect of those derived by pathogens from expression in
NTD mutations that abolish sugar binding in SARS-CoV-2 variants of concern. Together with genetic infected mammalian hosts) often bear large,
variance analyses in early pandemic patient cohorts, this binding implicates a sialylated polylactosamine highly mobile glycans. Critically, in the case of
motif found on tetraantennary N-linked glycoproteins deep in the human lung as potentially relevant glycoproteins such as SARS-CoV-2 spike that
to virulence and/or zoonosis. may themselves bind glycans, this leads to
contributions to protein NMR spectra that
S
may overlap with putative glycan ligand reso-
ialosides are present in glycans that are infected cell with its neuraminidase (NA or N) nances, thereby obscuring needed signal. Sec-
anchored to human cells, and they me- protein; HxNx variants of influenza with dif- ond, the NMR spectra of glycan ligands are
diate binding that is central to cell-cell ferent HA or NA protein types have a pro- themselves complex, comprising many over-
communication in human physiology and found effect on zoonosis and pathogenicity (1). lapped resonances as multiplets and limiting
that is at the heart of many host-pathogen The Middle East respiratory syndrome the accurate determination of signal intensities.
interactions. One of the most well-known [MERS (2)] virus, which is related to severe Finally, STD is commonly described as limited
examples is that of influenza virus, which binds acute respiratory syndrome coronavirus 1 and 2 to specific kinetic regimes and/or ligand-to-
to sialosides with its hemagglutinin (HA or H) (SARS-CoV-1 and -2), has been shown to ex- protein binding equilibrium positions (24). As
protein and cleaves off sialic acid from the ploit cell-surface sugar sialosides (2–6) as part a result, many regimes and systems have been
1 of an attachment strategy. Both SARS-CoV-1 considered inaccessible to STD.
Rosalind Franklin Institute, Harwell Science and Innovation
Campus, Oxford OX11 0FA, UK. 2Department of Chemistry, (7–9) and SARS-CoV-2 (10, 11) are known to Using a rigorous theoretical description,
University of Oxford, Oxford OX1 3TA, UK. 3Kavli Institute of gain entry to host cells through the use of coupled with a computational approach based
Nanoscience Discovery, University of Oxford, Oxford OX1 receptor-binding domains (RBDs) of their on a Bayesian deconvolution algorithm to ob-
3QU, UK. 4Division of Structural Biology, Wellcome Centre
for Human Genetics, University of Oxford, Headington, respective spike proteins that bind human jectively and accurately extract signal from all
Oxford OX3 7BN, UK. 5Diamond Light Source, Harwell cell-surface protein ACE2, but whether these observed resonances, we have undertaken an
Science and Innovation Campus, Oxfordshire, UK. 6Max viruses engage sialosides as part of the infec- optimized reformulation of the magnetization/
Planck Bristol Centre for Minimal Biology, University of
Bristol, Bristol, UK. 7MRC Human Immunology Unit, tion cycle has, despite predictions (6, 12), re- “saturation” transfer protocol (figs. S4, S5, S6,
Weatherall Institute of Molecular Medicine, University of mained unclear. Preliminary reports as to and S8). This approach reliably and quantita-
Oxford, John Radcliffe Hospital, Oxford, UK. 8Medical whether complex sialosides are or are not tively determines precise binding rates (kon,
Genetics, University of Siena, Siena, Italy. 9Med Biotech Hub
and Competence Center, Department of Medical
bound are contradictory and format-dependent koff , k ex), constants (KD ), and interaction
Biotechnologies, University of Siena, Siena, Italy. (13–15). Glycosaminoglycans on proteoglycans “maps” across a wide range of regimes (fig.
10
Department of Information Engineering and Mathematics, such as heparin have been identified as a S4), including systems previously thought to
University of Siena, Siena, Italy. 11Department of
Mathematics, University of Pavia, Pavia, Italy. 12Genetica
primary cooperative glycan attachment point be intractable.
Medica, Azienda Ospedaliera Universitaria Senese, Siena, (16, 17). Studies reporting sugar binding have
Italy. 13Bijvoet Centre for Biomolecular Research, Faculty of so far implicated binding sites in or close to Design of uSTA based on a comprehensive
Science, Utrecht University, Utrecht, Netherlands. 14Sir treatment of ligand-protein
the RBD of the spike protein. Surprisingly,
William Dunn School of Pathology, Oxford, UK. 15Maasai, I3S magnetization transfer
CNRS, Université Côte d’Azur, Nice, France. 16Department of the N-terminal domain (NTD, fig. S1), which
Pharmacology, University of Oxford, Oxford OX1 3QT, UK. has a putative glycan binding fold (10, 18, 19) While using existing STD methodology to study
*Corresponding author. Email: [email protected] and binds sialosides in other non-SARS co- the interaction between the SARS-CoV-2 spike
(J.H.N.); [email protected] (A.J.B.);
[email protected] (B.G.D.) ronaviruses (including MERS), has been less protein and sialosides, we noted several chal-
†These authors contributed equally to this work. explored. The NTD has no confirmed function lenges that resulted in the development of
Buchanan et al., Science 377, eabm3125 (2022) 22 July 2022 1 of 17

uSTA (see text S3 for more details). Our the- dard multidimensional approaches used for with x-ray crystal structures of BSA with other
oretical analyses (fig. S4) suggested that many resonance assignment (see Fig. 1, B and D, hydrophobic ligands, (33) it also revealed quan-
common assumptions or limits that are thought Fig. 2D, and Fig. 3A for examples; see also titative subtleties of this interaction at high
to govern the applicability of magnetization overlaps in all subsequent uSTA analyses and precision: Protein “grip” is felt more at the
transfer might in fact be circumvented, and table S7). distal edge of the indole moiety.
we set out to devise a complete treatment that 5) The uSTA software allowed the combina- Next, with indications of expanded capabil-
might accomplish this (figs. S5, S6, and S8). tion of intensity from scalar coupled multi- ity of uSTA in a benchmark system, we moved
This resulted in five specific methodological plets following a user-input assignment, to to first analyses of sugar-protein interactions.
changes that resulted in a more sensitive, ac- provide “per resonance” measures of satura- Sugar ligand trehalose (Tre) binds only weakly
curate, quantitative, and general method for tion transfer. These are provided as-is, and to trehalose repressor protein TreR and so
studying the interactions between biomole- also as h1/r6i interpolated “binding maps” that proves challenging in ligand-to-protein inter-
cules and ligands (summarized in figs. S5 and represent the interaction on nearby hetero- action analysis (34). Nonetheless, the uSTA
S8 and discussed in detail in text S4). atoms, thereby allowing ready visual inspec- workflow again successfully and rapidly deter-
1) We noted the discrepancy between KDs tion of the binding pose of a molecule. mined atom-specific transfer efficiencies with
determined using existing STD methods and high precision and resolution (fig. S3F). Atom-
those obtained using other biophysical meth- Testing of uSTA in model systems precise subtleties were revealed in this case as
ods (25). We performed a theoretical analysis The uSTA method (Fig. 1, A to C) was tested well: Hotspots of binding occur around OH-3/
using the Bloch-McConnell equations, a rigor- first in an archetypal, yet challenging, ligand- OH-4 and graduate to reduced binding around
ous formulation for studying the evolution of to-protein interaction (Fig. 1, D to G). Imple- both sugar rings, with only minimal binding
magnetization in exchanging systems that has mentation in an automated manner through of the primary OH-6 hydroxyl (fig. S3F).
been widely used to analyze chemical exchange software governing the uSTA workflow re- Once more, this uSTA-mapped P + L inter-
saturation transfer (26, 27), dark-state ex- duced artifacts arising from subjective, manual action proved consistent with prior x-ray crys-
change saturation transfer (28), and Carr-Purcell- analyses (fig. S6). The binding of L-tryptophan tal structures (35).
Meiboom-Gill (29) NMR data to describe protein (Trp) to bovine serum albumin (BSA, Fig. 1D)
motion. This analysis not only allowed us to is a long-standing benchmark (25) because Direct determination of ligand-protein KD
explain this discrepancy, but also enabled of the supposed role of hydrophobicity in using uSTA
fitting of data to give accurate kon, koff, and the plasticity of this interaction as well as a The precision of signal determination in uSTA
KD values for protein-ligand interactions that lack of corresponding fully determined, un- critically allowed variation of ligand/protein
were in excellent accord with alternative mea- ambiguous three-dimensional (3D; e.g., crystal) concentrations even down to low levels (see
surements (Fig. 1G and fig. S13); we also found structures. This is also a simpler amino acid– above), enabling direct determination of bind-
that the range of kon and koff in which the ex- protein interaction system (less-modified pro- ing constants in a manner not possible by
periment is applicable is far wider than prev- tein, small ligand) that classical NMR/STD classical methods. Following measurement of
iously recognized (fig. S4). methods are perceived (24, 32) to have already magnetization transfer between ligand and
2) In mammalian proteins, contributions delineated well. protein, variation with concentration (Fig. 1E)
from glycans on the surface of the protein As for a standard STD experiment, 1D 1H- was quantitatively analyzed using modified
could not be removed from the spectrum by NMR spectra were determined for both ligand Bloch-McConnell equations (36) (see Methods).
means of relaxation filters used in epitope and protein. In addition, mixed spectra con- These accounted for intrinsic relaxation, cross-
mapping (30) without compromising the sen- taining both protein and excess ligand (P + L) relaxation, and protein-ligand binding (Fig.
sitivity of the experiment. We addressed this were determined with and without excitation 1A) to directly provide measurements of equi-
instead by applying baseline subtraction using irradiation at frequencies corresponding to librium binding KD and associated kinetics
data obtained from a protein-only sample. prominent resonance within the protein but (kex). In the Trp/BSA system, this readily
3) The magnetization transfer, and hence far from any ligand (pulse “on”) or where the revealed KD = 38 ± 15 mM, kon = 1.6 (± 0.6) ×
the sensitivity of the experiment, will be center of the pulse was moved to avoid ligand 105 M–1 s–1, and koff = 6.0 ± 2.0 s–1 (Fig. 1G),
higher when the excitation frequency of the and protein (pulse “off,” labeled “1D” in the consistent with prior determinations of KD by
saturation pulse is close to a maximum in the figures). Deconvolved spectra for ligand deter- other solution-phase methods [KD = 30 ± 9 mM
protein NMR spectrum. If any ligand reso- mined in the presence of protein were matched by isothermal calorimetry (32)]. Note that this
nances are outside of the bandwidth of the with high accuracy by uSTA (Fig. 1E). More- direct method proved to be possible only be-
pulse, and if a “ligand-only” subtraction is ap- over, uSTA generated highly consistent binding cause of the ability of the uSTA method to
plied, the magnetization transfer can be max- “heatmaps” comprising atom-specific magneti- deconvolute a true signal with sufficient pre-
imized. With this condition, the response for zation transfer efficiencies (proton data map- cision, even at the lower concentrations used
a given protein-ligand system in fact becomes ped onto heteroatoms by taking a local 1/r6 and consequently lower signal (Fig. 1E). Thus,
invariant to the excitation frequency used average to enable visual comparison) that uSTA enabled atom-mapping and quantitation
(figs. S9 and S10). described the pose of ligand bound to pro- for ligand binding that were improved over
4) In complex molecules, such as sialosides, tein (Fig. 1F; see also figs. S8 and S11). These previous methods. Critically, these values were
NMR spectra are crowded and overlapped. To were determined over a range of ligand con- fully consistent with all observed NMR data
reliably obtain magnetization transfer mea- centrations even as low as 40 mM [Fig. 1, E(iii) and independently obtained measures of KD.
surements at all points in the ligand, we de- and F(iii)] where the ability of uSTA to extract
veloped a peak-picking algorithm based on accurate signal proved unprecedented and uSTA allows interrogation of designed
earlier work (31) that can automate the pro- critical to quantitation of binding (see below). crypticity in influenza HA virus attachment
cess, returning a list of peak locations and a Binding maps were strikingly consistent across Having validated the uSTA methodology, we
simulated NMR experiment that can be di- concentrations, indicating a single, consistent next used it to interrogate sugar binding by
rectly compared to the data. The locations of pose driven by the strongest interaction of viral attachment protein systems that have
the peaks are in excellent agreement with the protein with the heteroaromatic indole side proved typically intractable to classical meth-
locations for multiplets determined using stan- chain of Trp. This not only proved consistent ods. The hemagglutinin (HA) trimer of influenza

Fig. 1. Development of the uSTA method. (A to C) Schematic of the process from a sample containing protein only. This reveals contributions from individual
for uSTA that exploits comprehensive numerical analysis of relaxation and ligand multiplets originating from the ligand (yellow) and the protein-only baseline (black),
binding kinetics (A) using full and automatically quantified signal intensities allowing precise recapitulation of the sum (red). In (E), application of universal
in NMR spectra (B) and calculates per-resonance transfer efficiencies (C). In (B), deconvolution to STD spectra with varying concentrations of tryptophan allows
signal analysis determined the number of peaks that can give rise to the signal, uSTA using ligand resonances identified in (D). This in turn allows signal intensity
and returned simulated spectrum by convolving these with peak shape function. in the STD spectrum (P + L STD, light blue) to be determined with high precision.
The precise peak positions returned are in excellent agreement with the known Although signal-to-noise in the STD increases considerably with increasing ligand
positions of resonances identified using conventional means. Magnetization transfer concentration, the measured atom-specific transfer efficiencies as determined by
NMR experiments compare two 1D NMR spectra, where the second involves uSTA are consistent [(F), left, bar charts; right, transfer efficiency binding “maps”],
a specific saturation pulse that aims to “hit” the protein but “miss” the ligand in its showing that the primary contact between protein and ligand occurs on the distal edge
excitation. This is accomplished by acquiring the 1D spectrum with the saturation pulse (C-1, 2, 3, 4; N-7 and C-9 using the numbering shown) of the indole aromatic ring.
held off resonance at –35 ppm such that it will not excite protons in either ligand Application of the same uSTA workflow also allowed precise determination of
or protein [labeled “1D” in (B) and (C)]. The uSTA method requires these two spectra even weakly binding sugar ligand trehalose (Glc-a1, 1a-Glc) to E. coli trehalose
to be analyzed as described in (B), in pairs, one that contains the raw signal, and repressor TreR. Again, uSTA allows determination of transfer efficiencies with
the second that is the difference between the two. We define the “transfer efficiency” atom-specific precision (see fig. S3F). (G) Quantitative analysis of the STD build-up
as the fractional signal that has passed from the ligand to the protein. (D to curves using a modified set of Bloch-McConnell equations that account for binding
F) Application of uSTA to study the interaction between bovine serum albumin (BSA) and cross-relaxation allows us to determine thermodynamic and kinetic parameters
and L-tryptophan (Trp). In (D), the 1D 1H-NMR spectrum of the mixture at 200 mM that describe the BSA-Trp interaction, KD, kon, and koff. The values obtained are
Trp and 5 mM BSA (= P + L, blue) is dominated by ligand, yet the ligand (L) and protein indicated and are in excellent accord with those obtained by other methods (25, 32).
(P) can still be deconvolved by universal deconvolution, using a reference obtained Errors come from a bootstrapping procedure (see Methods).

Fig. 2. uSTA allows mapping of a designed cryptic sugar-binding site in sialosides in this synthetic variant. (B and C) Notably, wild-type and DRBS
H1N1 influenza hemagglutinin (HA). (A) HA presents on the surface of the viral variants of H1N1 HA in fact show a remarkably similar overall binding pattern
membrane and has been shown to bind with sialic acid surface glycans to (B) for the 2,3-sialo-trisaccharide 2 (focused through engagement with the sialoside)
mediate host cell entry. A designed (38) DRBS variant is generated by the but a significant intensity moderation (C) for the DRBS variant, indicative of a
creation of an N-linked glycosylation site via the creation of needed sequon NQT partial (but not complete) loss of binding consistent with design (38). (D) Raw
from wild-type NQR by the R205T point mutation in HA adjacent to the sialic spectral data demonstrate that atom-specific differences in intensity in the 1D versus
acid–binding site. In this way, disruption of sialic acid binding through designed the difference spectra can be discerned using uSTA. Note that atom numbering
“blocking” in HA-DRBS is intended to ablate the binding of HA wild-type to shown and used here is generated automatically by uSTA.
A virus is known to be essential for its ex- construct (10) of intact trimeric SARS-CoV-2 analyzed using uSTA, these revealed [Fig. 3, E
ploitation of sialoside binding (37); H1N1 has spike attachment protein (fig. S7) revealed and F(i, ii), table S7, and fig. S11] clear “end-on”
emerged as one of the most threatening var- extensive protein glycosylation with sufficient interactions by 1a as a ligand [Fig. 3F(i)], me-
iants in recent years. We took the H1 HA in both mobility to generate a strong 1H-NMR reso- diated primarily by the acetamide NHAc-5,
native form and a modified form, containing a nance in the region 3.4 to 4.0 ppm (Fig. 3A). but no reliably measurable interactions by
non-natural sequon specific for N-glycosylation Although lacking detail, these resonances 1b [Fig. 3F(ii)]. This detection of selective
that was previously designed (38) to block in- displayed chemical shifts consistent with the a-anomer interaction, despite the much greater
termolecular (in trans) sugar binding. This de- described mixed patterns of oligomannose, dominance of the b-anomer in solution, pro-
signed blocking in a so-called HA-DRBS variant hybrid, and complex N-glycosylation found on vided yet another demonstration of the power
(38) also notably creates an additional glycan SARS-CoV-2 spike after expression in human of the uSTA method, here operating in the
beyond the existing, potentially confounding, cells (40). As such, these mobile glycans on background of dominant alternative sugar
glycosylation background. It therefore provided SARS-CoV-2 spike contain sialoside glycan resi- (Fig. 3E and fig. S11). The a selectivity corre-
another test of uSTA’s ability to delineate rele- dues that not only confound analyses by clas- lates with the near-exclusive occurrence of sia-
vant sialoside ligand interactions in another sical NMR methods but are also potential losides on host cell surfaces as their a- but not
important pathogen protein (Fig. 2A). Despite competing, “internal” (in cis) ligands for any b-linked conjugates (see also below).
this intended blocking, the precision of uSTA putative attachment (in trans) interactions, as Having confirmed simple, selective mono-
was such that residual in trans binding of sialo- well as possible direct ligands for in trans saccharide a-sialoside binding, we explored
trisaccharide 2 could still be detected in HA- interactions in their own right (41). Therefore, extended a-sialoside oligosaccharide ligands
DRBS, albeit at a lower, modulated level [as their presence in the protein NMR analysis (compounds 2 and 3; Fig. 3, C and D) that
expected by design (38)]. Although H1 HA is presented clear confounding issues for typical would give further insight into the binding of
known to bind both sialo-trisaccharides 2 and classical STD analyses. As such, SARS-CoV-2 natural endogenous human cell-surface sugars
3, 2 is the less preferred (2,6- over 2,3-linked) spike represented a stringent and important as well as unnatural variants (compounds 4
ligand (39), and yet its binding could still be test of the uSTA method. to 6; Fig. 3, F and G) that could potentially
mapped (here, to a mode mediated primarily We used uSTA to evaluate a representative interrupt such binding. Sialosides are often
by the sialoside moiety). In this way, the sen- panel of both natural and site-specifically mod- found appended to galactosyl (Gal/GalNAc)
sitivity of uSTA to detect even lower sialoside ified unnatural sialosides as possible ligands residues in either a2,3-linked (2) or a2,6-
binding to relevant proteins was confirmed. of spike (Fig. 3 and figs. S3 and S12). Use of linked form (3). Both were tested (Fig. 3, C
classical methods provided an ambiguous as- and D) and exhibited “end-on” binding con-
uSTA reveals natural, cryptic sialoside binding sessment (fig. S8), but use of uSTA immedi- sistent with that seen for N-acetyl-neuraminic
by SARS-CoV-2 spike ately revealed binding and nonbinding sugar acid (1) alone, but with more extended bind-
We next probed putative, naturally cryptic ligands (Fig. 3, figs. S8, S10, and S11, and table ing surfaces (Fig. 3D), qualitatively suggesting
sialoside binding sites in SARS-CoV-2. Our S7). Initially, the simplest sialoside, N-acetyl- a stronger binding affinity (see below for quan-
analysis of the 1D protein 1H-NMR spectrum of neuraminic acid (1), was tested as a mixture of titative analysis). Common features of all sialo-
the purified prefusion-stabilized ectodomain its mutarotating anomers (1a ⇔ 1b). When side binding modes were observed: The NHAc-5

Fig. 3. uSTA reveals interaction acetamide of the terminal sialic acid (Sia) is a
of sialosides with SARS-CoV-2 binding hotspot in 1, 2, and 3 that drives the
spike protein. A panel of natural, “end-on” binding. Differences were also ob-
unnatural, and hybrid variant served: The a2,6-trisaccharide (3) displayed a
sialoside sugars 1Ð6 (see fig. S12) more extended binding face yet with less in-
was used to probe interaction tense binding hotspots (Fig. 3D) engaging ad-
between sialosides and spike. ditionally the side-chain glycerol moiety (C7–C9)
(A) The 1D 1H-NMR of SARS-CoV-2 of the terminal Sia acid as well as the OH-4
spike protein shows considerable C4 hydroxyl of the Gal residue. The interac-
signal in the glycan-associated tion with a2,3-trisaccharide (2) was tighter and
region despite protein size, indicative more specific to NHAc-5 of the Sia.
of mobile internal glycans in spike These interactions of the glycerol C7–C9
protein. This effectively masks tra- side chain detected by uSTA were probed fur-
ditional analyses, as without careful ther through construction (fig. S12) of unnatu-
subtraction of the proteinÕs contri- ral modified variants (4 to 6; Fig. 3, F and G).
butions to the spectrum (fig. S8), the Replacement of the OH-9 hydroxyl group of
ligand cannot be effectively studied. sialic acid with azide N3-9 in 4 [Fig. 3F(iii) and
(B) Application of the uSTA workflow table S7] was well tolerated, but larger changes
(fig. S6) to SARS-CoV-2 spike pro- (replacement with aromatic group biphenyl-
tein (shown in detail for 2). The uSTA carboxamide BPC-9) in 5 and 6 led to an ap-
process of ligand peak assignment parently abrupt shift in binding mode that was
and deconvolution → P + L peak instead dominated by the unnatural hydropho-
assignment and deconvolution → bic aromatic modification (Fig. 3G and Fig. 4C).
application to P + L STD yields As for native sugar 1, azide-modified sugar 4
precise atom-specific transfer effi- also interacted with spike in a stereochemi-
ciencies (fig. S6). Note how in (ii) cally specific manner with only the a-anomer
individual multiplet components, displaying interaction [Fig. 3F(iii)], despite
have been assigned (yellow); the dominance of the b-anomer in solution [Fig.
back-calculated deconvolved 3F(iv)]. uSTA allowed precise dissection of in-
spectrum (red) is an extremely close teraction contributions in these unusual hybrid
match for the raw data (purple). (natural-unnatural) sugar ligands that could
The spectrum is a complex super- not have been determined using classical meth-
position of the ligand spectrum ods (see text S5).
(and protein only yet uSTA again Using variable concentrations of the most
accurately deconvolves the potent natural ligand a2,3-trisaccharide 2 [6 mM
spectrum, revealing the contribution spike, 2 at 60 mM, 200 mM, 1 mM, and 2 mM
of protein-only (black) and the ligand excitation at 5.3 ppm] and variable concen-
peaks (yellow). Using these data, trations of spike protein, we used the uSTA
uSTA analysis of the STD spectrum method to directly determine solution-phase
pinpoints ligand peaks and signal affinities (Fig. 4A): KD = 32 ± 12 mM, kon =
intensities. Spectral atom numbering 6300 ± 2300 M–1 s–1, and koff = 0.20 ± 0.08 s–1.
shown and used here is generated We also probed binding in a different mode by
automatically by uSTA; all other measuring the affinity of spike to 2 when dis-
numbering in sugars follows carbo- played on a modified surface (fig. S13) using
hydrate nomenclature convention. surface plasmon resonance (SPR) analysis. The
(C and D) Application of the latter generated a corresponding KD = 23.7 ±
uSTA workflow (fig. S6) reveals 3.6 mM (kon = 1004 ± 290 M–1 s–1). Such similar
atom-specific binding modes to values for sialoside ligand in solution (by uSTA)
spike protein for both natural or when displayed at a solid-solution interface
(e.g., sialoside) trisaccharides (by SPR) suggested no substantial avidity gain
Siaa2,3Galb1,4Glc (2) and from display of multiple sugars on a surface.
Siaa2,6Galb1,4Glc (3). Comparison
of the uSTA method focused on the Structural insights from uSTA delineate
NHAc methyl resonance shows excellent agreement (C). The uSTA method allowed determination of binding binding to SARS-CoV-2 spike
surfaces for both trisaccharides 2 [D(i)] and 3 [D(ii)]. (E and F) STD spectrum (E) and mapped atom-specific transfer uSTA analyses consistently identified binding
efficiencies (F) for sialic acid (1) and 9-N3 azido variant 4. Both interconverting a and b anomeric forms could hotspots in sugars 1 to 4 providing the highest
be readily identified. Despite the dominance of the b form [94 %, E(i)], application of the uSTA method following transfer efficiencies in an atom-specific man-
assignment of resonances from the two forms allowed determination of binding surfaces simultaneously [E(ii), ner, particular the “end” NHAc-5 acetamide
F(i, ii)]. Spike shows strong binding preference for the a anomers [F(i, iii) versus F(ii, iv)] despite this strong methyl group of the tip sialic acid residue in
population difference. Binding surfaces were also highly similar to those of extended trisaccharides 2 and 3. all. A combination of uSTA with so-called high
(G) Using these intensities, atom-specific transfer efficiencies can be determined with high precision, shown here ambiguity–driven docking (HADDOCK) meth-
for hybrid sialoside 5. The details of both the unnatural BPC moiety and the natural sialic acid moiety can be mapped; ods (42, 43) was then used to probe likely re-
although the unnatural aromatic BPC dominates interaction, uSTA nonetheless delineates the subtleties of the gions in SARS-CoV-2 spike for this “end-on”
associated contributions from the natural sugar moiety in this ligand (see also figs. S5 and S6). binding mode via uSTA data-driven atomistic

Fig. 4. Quantitative uSTA analyses allow comparison and predictive restraints to be identified in natural (1Ð4) and hybrid (5, 6) sugars. Data are shown at
for protein ligand-binding prediction. (A) Quantitative analysis of the STD constant protein and ligand concentrations. With the BPC moiety present, the
build-up curves using a modified set of Bloch-McConnell equations that account build-up of magnetization occurs significantly faster than when not; various such
for binding and cross-relaxation allow us to determine thermodynamic and kinetic hybrid ligands give highly similar curves. By contrast, natural ligands have a much
parameters that describe the SARS-CoV-2•2 interaction, KD, kon, and koff. The slower build-up of magnetization. This, together with the absolute transfer
values obtained are indicated. Errors come from a bootstrapping procedure (see efficiencies being very different, and the overall pattern on the interaction map
Methods). The lower ligand concentrations yield data that are of lower sensitivity combine to reveal that the ligands are most likely binding via two different modes
than higher concentrations. The transfer efficiencies will be higher in this case, as and possibly locations on the protein. (D) Coupling uSTA with an integrative
more molecules are effectively involved in the binding. Thus, data at lower modeling approach such as HADDOCK (42, 43) allowed generation and, by
concentrations will in general have more scatter and higher transfer efficiencies. quantitative scoring against the experimental uSTA data, selection of models that
These data points are desirable for the analysis, as it is here where we expected provide atomistic insights into the binding of sugars to the SARS-CoV-2 spike
the greatest variation of transfer efficiency with ligand concentration. The analysis protein, as shown here by superposition of uSTA binding “map” onto modeled
is applied globally and so the uncertainties in the final fitted parameters from poses. uSTA mapping the interaction between SARS-CoV-2 spike [based on RCSB
the bootstrapping analysis (see Methods) provide a direct and confident measure 7c2l (19)] with ligands 1a, 2, and 3 identifies the NHAc-5 methyl group of the tip
of the goodness of fit. (B) Normalized uSTA transfer efficiencies of the NAc-5 sialic acid residue making the strongest interaction with the protein. By filtering
methyl protons can be determined for each ligand studied here. This allowed HADDOCK models against this information, we obtain structural models that
relative contributions to “end on” binding to be assessed via uSTA in a “mode- describe the interaction between ligand and protein (fig. S14). Most strikingly,
specific” manner. This confirmed strong a- over b-sialoside selectivity. Errors were we see the same pattern of interactions between protein and sialic acid moiety in
determined through a bootstrapping procedure where mixing times were sampled each case, where the NAc methyl pocket is described by a pocket in the spike
with replacement, allowing for the construction of histograms of values in the NTD. Although sequence and structural homology are low (fig. S1), MERS spike
various parameters that robustly reflect their fitting errors. (C) Normalized build- protein possesses a corresponding NHAc-binding pocket characterized by an
up curves for the most intense resonances allowed two distinct modes of binding aromatic (Phe39)–hydrogen-bonding (Asp36)–hydrophobic (Ile132) triad (5).

models. In each case, a cluster of likely poses uSTA to compare the relative potency of the ablated binding toward sialoside 2 as compared
emerged (Fig. 4D) for 1a, 2, and 3 (see fig. S14 sialoside binding identified here to previously to first-phase B-origin-lineage spike (Fig. 5D
for details) consistent with “end-on” binding identified (17) heparin binding motifs (Fig. 5, A and fig. S15).
where the acetamide NHAc-5 methyl group of and B). Heparin sugars 7 and 8 of similar size Finally, to explore the possible role of sia-
the sialic acid moiety was held by the unusual to natural sialosides 3 and 4 were selected so loside binding in relation to ACE2 binding, we
b sheet–rich region of the NTD of SARS-CoV-2 as to allow a near ligand-for-ligand compari- also used uSTA to probe the effects upon bind-
spike. Under the restraints of uSTA and homol- son based on similar potential binding surface ing of the addition of a known, potent neutral-
ogy, a glycan-binding pocket was delineated by areas. 7 and 8 also differed from each other izing antibody of ACE2-spike binding, C5 (Fig.
a triad of residues (Phe79, Thr259, and Leu18) me- only at a single glycan residue (residue 2) site 5, D and E, and fig. S16) (45, 46). Assessment of
diating aromatic, carbonyl hydrogen–bonding, to allow possible dissection of subtle contribu- binding to sialoside 2 in the presence and ab-
and hydrophobic interactions, respectively. tions to binding. Unlike the “end-on” binding sence of antibody at a concentration sufficient
However, the sequence and structural homol- seen for sialosides 3 and 4, uSTA revealed an to saturate the RBD led to only slight reduc-
ogy to prior (i.e., MERS) coronavirus spike pro- extended, nonlocalized binding interface for 7 tion in binding. Uniformly modulated atomic
teins in this predicted region was low; the and 8 consistent instead with “side-on” bind- transfer efficiencies and near-identical bind-
MERS spike protein uses a corresponding ing [Fig. 5, A(ii) and B(ii)]. ing maps (Fig. 5, E and F) were consistent with
NHAc-binding pocket characterized by an Next, we examined the possible evolution of a maintained sialoside-binding pocket with
aromatic (Phe39), hydrogen-bonding (Asp36), sialoside binding over lineages of SARS-CoV-2 undisrupted topology and mode of binding.
and hydrophobic (Ile132) triad to bind the (44). Four notable variants of concern—alpha/ Together, these findings allow us to con-
modified sugar 9-O-acetyl-sialic acid (5). B1.1.7, beta/B1.351, delta/B.1.617.2, and omicron/ clude that the sialoside binding observed with
SARS-CoV-2 glycan attachment mechanisms B.1.1.529—emerged in later phases of the pan- uSTA involves a previously unidentified “end-
have to date only identified a role for spike RBD demic. When these corresponding spike pro- on” mechanism/mode that operates in addi-
in binding rather than NTD (15, 17). We used tein variants were probed by uSTA, all displayed tion to and potentially cooperatively with
Fig. 5. Comparison of SARS-CoV-2 glycan attachment mechanisms and SARS-CoV-2 spike. (D) Ablated binding of the a-2,3-sialo-trisaccharide 2, as
variant evolution via uSTA suggests binding away from the RBD that measured by the transfer efficiency of the NHAc protons, is identified by
is lost. (A and B) Two heparin tetrasaccharides (7, 8) are shown by uSTA uSTA in the lineage variants of SARS-CoV-2 spike [see (C) for colors]. This
[A(iii), B(iii)] to bind B-origin-lineage SARS-CoV-2 spike protein (“original” proves consistent with mutations that appear in the sialoside-binding site
spike) in a “side-on” mode [A(ii), B(ii)]. Atom specific binding is shown of NTD identified in this study [see (C) and Fig. 6]. (E and F) uSTA of
in A(ii) and B(ii). Assignments shown in green use conventional glycan sialoside 2 with B-origin-lineage SARS-CoV-2 spike in the presence of the
numbering. (C) Substantial numbers of mutations arise in the NTD region potent RBD-neutralizing nanobody C5 [spike E(i), nanobody-plus-spike E(ii)]
identified by uSTA in the B.1.1.7/alpha (cyan), B.1.351/beta (orange), shows essentially similar binding patterns with uniformly modulated atomic
B.1.617.2/delta (dark blue) and B.1.1.529/omicron (green) lineage variants of transfer efficiencies (F).

ACE2 binding in SARS-CoV-2. The primary makes a hydrogen bond with the side chain of so-called poly-N-acetyl-lactosamine [polyLacNAc
sialoside glycan-binding site SARS-CoV-2 Tyr145, the N-acetyl group with Ser247, and the or (Gal-GlcNAc)n] chain-extension variants
spike is distinct from that of heparin (“end-on” carboxylate with Gln183, and there are hydro- found in tetraantennary N-linked glycoproteins
versus “side-on”), not in the RBD (not neutral- phobic interactions with Trp152. Lowering the (Fig. 7D), including those displaying sialyl-Gal-
ized by RBD-binding antibody), and found map threshold to 1.6s would be consistent with GlcNAc sialoside motifs (52, 53). Variants in
instead in an unusual NTD region that has a second pyranoside (e.g., galactoside) residue LGALS3BP were present in 9 of 114 a/pauci-
become altered in emergent variants (loss of (fig. S17C). At this contour level, the map clear- symptomatic subjects or mildly affected patients
binding in alpha and beta variants of concern). ly covers the axially configured carboxylate of (~8%) compared to 8 of the remaining 419
the sialic acid (fig. S17). The middle galactoside patients who required more intensive care:
Cryo-EM pinpoints the sialoside-binding site in residue of 9 positioned in this density would oxygen support, CPAP/BiPAP, or intubation
B-origin-lineage spike make contacts with Arg248 and Leu249. (<2%); none of the 69 most seriously affected
Structural analysis of the possible binding of Structural superposition of the NTD with patients (intubated) carried variants of LGALS3BP
sialosides has been hampered to date by the that of the MERS spike (RCSB 6NZK) shows (Fig. 7B). Intact LGALS3BP gene product Gal-
moderate resolution, typically less than 3 Å, that although the sialic acid–binding pockets 3-BP therefore appears correlated with more
of most SARS-CoV-2 spike structures. Cur- of both are on the outer surface, these pockets severe COVID-19 outcome. The other implicated
rently deposited cryo-EM–derived coulombic are 12 Å apart (as judged by the C2 atom of gene, B3GNT8, encodes a protein glycosyltrans-
maps show that the NTD of spike is often the respective sialic acids; Fig. 6D). In MERS spike, ferase, b-1,3-N-acetyl-glucosaminyltransferase-8
least well-resolved region. In our initial at- sialic acid is bound at the edge of the central (GlcNAcT8 or b3GnT8), that is responsible for
tempts with native protein, large stretches of b sheet, whereas in SARS-CoV-2 the sugar is the creation of the anchor point of poly-N-
amino acids within the NTD were not exper- bound at the center of the sheet; thus, the acetyl-lactosamine (polyLacNAc) in such tetra-
imentally located (45); the most disordered pockets use different elements of secondary antennary N-linked glycoproteins (Fig. 7D)
regions occur in the NTD regions that con- structure. Because of distinct changes in the (54). Again, rare variants in B3GNT8 were
tribute to the surface of the spike. A stabilized structure of the loops connecting the strands, present in 11 of 114 of a/pauci-symptomatic
closed mutant form (47) of the spike was ex- the sialic acid pocket from one protein is not subjects or mildly affected patients (~10%) com-
amined and gave improvements, but the cou- present in the other protein. Several regions of pared to 10 of the remaining 419 patients who
lombic map was still too weak and noisy to additional density were not fitted by the mod- required more intensive care (~2%) (Fig. 7C).
permit tracing of many of the loops in the el (fig. S18).
NTD. However, with a reported fatty acid– Discussion
bound form of the spike (48), which has shown Disclosure of sialoside trisaccharide as a ligand Experimentally, there are still few, if any, or-
prior improved definition of the NTD, we were for B-origin-lineage SARS-CoV-2 correlates with thogonal approaches to the useful surface dis-
able to collect a 2.3 Å dataset in the presence clinical genetic variation in early-phase pandemic play methods (e.g., “glycoarrays”) currently used
of the a2,3-sialo-trisaccharide 9. The map was A distinctive mode of sialoside binding by spike for readily surveying ligands that might be ex-
clear for almost the entire structure including confirms a potential attachment point for ploited by pathogens. Following validation in
the previously identified linoleic acid (fig. S17D); SARS-CoV-2 found commonly on cell surfaces model ligand-protein systems, uSTA provided
only 13 N-terminal residues and two loops (sialosides are attached both as glycolipid and a ready method for identifying sugar ligands
(residues 618 to 632 and 676 to 689) were not glycoprotein glycoconjugates), thus raising bound by pathogens, as well as their binding
located. Although the density is weaker at the the question of whether glycosylation function parameters and poses, even in posttransla-
outer surface of the NTD than at the core of the in humans affects infection by SARS-CoV-2 tionally modified (e.g., glycosylated) protein
structure (fig. S17B), the map was of sufficient and hence the presentation and pathology of systems.
quality to model N-glycosylation at site Asn149, COVID-19 disease. Analysis of whole-exome In an influenza virus HA protein variant de-
which is in a flexible region, and the fucosylation sequencing data of an early 2020 cohort of signed to abolish binding through competition
state of N-linked glycans at Asn165 (Fig. 6A). 533 COVID-19–positive patients (see table S1) by an added glycan site on HA (55), uSTA was
We observed density in a pocket at the sur- identified two glycan-associated genes within nonetheless able to unambiguously reveal and
face of the NTD lined by residues His69, Tyr145, the top five that were most influential upon “map” residual sialoside binding despite the
Trp152, Gln183, Leu249, and Thr259 (Fig. 2B and disease severity. Specifically, recursive feature presence of an added protein-linked glycan as
fig. S17). This density is absent in other spike elimination applied to a LASSO (least absolute “internal blocker.” This is a protein type that
structures of higher than 2.7 Å resolution (PDB shrinkage and selection operator)–based (49) has been well-studied in array formats (39);
IDs 7jji, 7a4n, 7dwy, 6x29, 6zge, 6xlu, 7n8h, logistic regression model identified LGALS3BP we showed here that, even with a “blocked” HA
6zb5, and 7lxy), even those (such as PDB IDs (fourth of >18,000 analyzed genes) and B3GNT8 (in a glycosylated state) and a non-preferred
7jji, 6zb5, and 6zge) that have a well-ordered (fifth of >18,000) (Fig. 7A and fig. S19). Variants 2,3-sialoside ligand, binding could still be
NTD. The density when contoured at 2.6s is in these two genes were beneficially associated mapped by uSTA.
fitted by an a-sialoside consistent with the ter- with less severe disease outcome (Fig. 7, B and In the spike trimer of the B-origin lineage of
minal residue of 9, with the distinctive gly- C; see also tables S2 to S6 for specific B3GNT8 SARS-CoV-2, despite the presence of mobile,
cerol and N-acetyl groups clear. To further and LGALS3BP genetic variants, B3GNT8 c2 protein-linked glycans, uSTA clearly revealed
strengthen our confidence in the identifica- five categories, B3GNT8 c2 2×2, LGALS3BP c2 sialoside binding and, through mapping, re-
tion of the sialic acid, we determined a native five categories, LGALS3BP c2 2×2, respectively). vealed that this binding is more potent when
(unsoaked) structure to 2.4 Å using the same LGALS3BP encodes for a secreted protein, the sialosyl moiety terminates galactosyl oligo-
batch of protein (fig. S17D). This structure galectin-3–binding protein (Gal-3-BP, also saccharides. This pose is in agreement with
showed no density in the sialic acid binding known as Mac-2-BP), that is a partner and our cryo-EM structures, which show that the
site supporting our assignment. The sialoside blocker of a specific member (Gal-3) of the NHAc-5 N-acetyl group at the sugar’s tip is
was therefore included in the refinement and galectin class of carbohydrate-binding proteins buried, a mode of binding we refer to as “end-
the thermal factors (108 Å2) were comparable (50). Galectins are soluble and are typically on.” Prior modeling was partly misled by use
to those for the adjacent protein residues (95 secreted and implicated in a wide range of of lower-resolution structures of spike, be-
to 108 Å2). In this position, the glycerol moiety cellular functions (51). Notably, Gal-3 binds the cause the NTD is highly disordered in these

initial structures. uSTA and cryo-EM also iden-

tified a second, differing mode of binding in
SARS-CoV-2 spike only by hybrid, aromatic A B
sugars (e.g., 5, 6, 9) driven by aromatic en-
gagement, but the physiological relevance of
this binding pocket is currently unclear. We
cannot exclude the presence of another sialic
acid–binding site (15), but there is no struc-
tural support for it.
The spike sialoside-binding site in the NTD
is also coincident or near to numerous muta-
tional and deletion hotspots in, for example,
alpha and omicron (His69, Val70, Tyr145) and
beta (b strand Leu241-Leu242-Ala243-Leu244) var-
iants of concern (Fig. 5C). Changes remove C
either key interacting residues (alpha/omicron
variant: residues 69, 70, 145) or perturb struc-
turally important residues (beta variant: b
strand) that form the pocket. The “end-on”
binding we observed is quite different to the
“side-on” binding observed for heparin (17).
Heparins are often bound in a non–sequence-
specific, charge-mediated manner, consistent with
such a “side-on” mode. The location of three
binding sites in the trimer, essentially at the
extreme edges of the spike (Fig. 6C), also im-
poses substantial geometric constraints for avid-
ity enhancement through multivalency (56, 57).
We also find a clear link between our data D
and genetic analyses of patients that correlate
with the severity of their disease. This asso-
ciation suggests potential roles in infection
and disease progression for cell-surface gly-
cans and the two glycan-associated genes that
we have identified. Despite their independent
identification here, both gene products inter-
act around a common glycan motif: the poly-
LacNAc chain-extension variants found in
tetraantennary N-linked glycoproteins. Con-
sistent with the sialoside ligands found here,
these glycoproteins contain Sia-Gal-GlcNAc E
motifs within N-linked polyLacNAc chains.
These motifs have recently been identified in
the deeper human lung (58).
These data lead us to suggest that B-origin-
lineage SARS-CoV-2 virus may have exploited
glycan-mediated attachment to host cells (Fig.
7D) using N-linked polyLacNAc chains as a
foothold. Reduction of Gal-3-BP function would Fig. 6. Cryo-EM analysis of sialoside binding in B-origin-lineage SARS-CoV-2 spike. (A) In the presence
allow its target, the lectin Gal-3, to bind more of sialoside 9 (fig. S18), a well-ordered structure with resolution (2.3 Å) sufficient to identify even
effectively to N-linked polyLacNAc chains, glycosylation states of N-linked glycans within spike was obtained. Here they reveal a paucimannosidic base
thereby competing with SARS-CoV-2 virus. bis-fucosylated chitobiose core structure GlcNAcb1,4(Fuca1,3-)(Fuca1,6-)GlcNAc–Asn. (B) The sialoside-
Similarly, loss of b3GnT8 function would ab- binding site in the NTD is bounded by His69, Tyr145, Trp152, Ser247, and Gln183. Gln183 controls stereochemical
late the production of foothold N-linked poly- recognition of a-sialosides by engaging COOH-1. Tyr145 engages the C7-C9 glycerol side chain of sialoside.
LacNAc chains, directly denying the virus a Ser247 engages the NHAc-5. Notably, Tyr145 and His69 are deleted in the alpha variant of SARS-CoV-2 that
foothold. We cannot exclude other possible loses its ability to bind sialoside (see Fig. 5). See also fig. S17 for coulombic maps. (C) Spike’s sialoside
mechanisms including, for example, the role binding site is found in a distinct region of the NTD (left, from side, right from above) that is at the “edge”
of N-linked polyLacNAc chains in T cell regu- of the spike (see text). (D) Superposition (right) of SARS-CoV-2-NTD (left, gray) with MERS-NTD (middle,
lation (59) or glycolipid ligands (15). This magenta) shows distinct sites 12 Å apart. (E) A comparison of the ligand-binding mode measured by uSTA
analysis of the influence of genetic variation NMR and a map calculated from the cryo-EM structure. For each hydrogen environment in the resolved
upon susceptibility to virus was confined to sialic acid, an array of distances to all protein hydrogens, r, was calculated. The interaction of each ligand
“first wave,” early-pandemic patients infected environment with the spike protein was defined as the summation of all 1/r6 environment-protein distances.
with B-origin-lineage SARS-CoV-2. Our dis- Values were interpolated in a 1/r6 manner onto heteroatoms following the same procedure according to the
covery here also that in B-lineage virus such described NMR methods, and the color bar was scaled from zero to the maximum value.

A B
D product of gene
Gal-3 LGALS3BP
= Gal-3-BP
putative sugar
ligand added by the
identified by uSTA product of gene
B3GNT8 =
host cell GlcNAcT8
surface
Sia
Gal
GlcNAc
Man
Fuc
Fig. 7. Analyses of early 2020 first-phase SARS-CoV-2 PCR-positive based logistic regression weightings after recursive feature elimination
patients reveals glycan-associated genes suggesting a model of glycan analysis of 533 SARS-CoV-2Ðpositive patients. Positive weights score
interaction consistent with uSTA observations of sialoside binding in susceptible response of gene variance to COVID-19 disease, whereas negative
B-origin-lineage SARS-CoV-2. (A) GEN-COVID workflow. Left: The GEN-COVID weights confer protective action through variance. Variation in glycan-
Multicenter Study cohort, of 533 SARS-CoV-2 PCR-positive subjects of associated genes B3GNT8 and LGALS3BP score second and third out of all
different severity from phase one of the pandemic, was used for rare variant (>18,000) genes as the most protective, respectively (highlighted red).
identification. Upper right: Whole-exome sequencing (WES) data were (C) Distribution of rare variants in B3GNT8 and LGALS3BP. Left: Rare beneficial
analyzed and binarized into 0 or 1 depending on the presence (1) or the mutations distributed along the Gal-3-BP protein product of LGALS3BP, divided into
absence (0) of variants in each gene. Lower right: LASSO logistic regression the SRCR (scavenger receptor cysteine-rich) domain (light blue) and the BACK
feature selection using a Boolean representation of WES data leads to the domain (light orange). Right: Rare beneficial mutations distributed along the
identification of final sets of features divided according to severity or bGlcNAcT8 protein product of B3GNT8 divided into the predicted transmembrane
mildness of disease, contributing to COVID-19 variability. See also (79, 80) for (TM) domain (light blue) and glycosyltransferase catalytic (GT) domain (light
further details of background methodology. (B) Histogram of the LASSO- orange), which catalyzes the transfer of polyLacNAc-initiating GlcNAc onto

tetraantennary N-linked glycoproteins [see also (D)]. The different colors of identified, B3GNT8 and LGALS3BP produce gene products bGlcNAcT8 and
the mutation bands (top to bottom) refer to the severity grading of the PCR- Gal-3-BP, respectively, that manipulate and/or engage with processes associated
positive patients who carried that specific mutation (red, Hospitalized intubated; with a common polyLacNAc-extended chain motif found on tetraantennary
orange, Hospitalized CPAP/BiPAP; pink, Hospitalized Oxygen Support; light N-linked glycoproteins. A model emerges in which any associated loss of
blue, Hospitalized w/o Oxygen Support; blue, Not hospitalized a/pauci- function from variance leads either to loss of polyLacNAc-extended chain (due to
symptomatic). (D) A proposed coherent model consistent with observation of loss of initiation by bGlcNAcT8) or enhanced sequestration of by Gal-3
implicated B3GNT8 and LGALS3BP genes and the identification of sialosides polyLacNAc-extended chain (which is antagonized by Gal-3-BP). Both would
as ligands for spike by uSTA and cryo-EM. Strikingly, although independently potentially lead to reduced access of virus spike to uSTA-identified motifs.
binding to certain sialosides was ablated in GTTCGTGTTCCTGGTGCTG-3′) and the reverse cells were transfected using ExpiFectamine293
later phases of the pandemic (after September primer (5′-GTCATTCAGCAAGCTTAAAAAGG- transfection reagent (ThermoFisher) according
2020) in variants of concern further highlights TAGAAAGTAATAC-3′), resulting in an aviTag/ to the manufacturer’s instructions. The cells
the dynamic role that sugar binding may play Bap sequence plus 6His in the 3′ terminus of were cultured at 37°C, 8% CO2, 125 rpm (25 mm
in virus evolution and may be linked, as has the construct. The template for B-lineage- throw) for 5 to 6 days before purification.
been previously suggested for H5N1 influenza origin (wild type) spike is previously described For the purification of wild type, alpha, beta,
A virus, to the “switching” of sugar-binding pre- (62). The templates for the alpha and beta delta, and omicron spike, the medium in which
ferences by pathogens during or after zoonotic spike are in the supplementary materials. the spike protein was secreted was supple-
transitions (1). The focused “end-on” binding of For B-lineage-origin, alpha, and beta spike, mented with 1× PBS buffer at pH 7.4 (1:1 v/v)
the N-acetyl group in the N-acetyl-sialosides, Expi293 cells (Thermofisher Scientific) were and 5 mM NiSO4. The pH was adjusted with
which are found as the biosynthetically exclusive used to express the Spike-Bap protein. The NaOH to pH 7.4 and filtered using a 0.8-mm
form of sialosides in humans (60), might have cells were cultured in Expi293 expression filter. The mixture was stirred at 150 rpm for
been a contributing factor in driving zoonosis. media (Thermofisher Scientific) and were 2 hours at room temperature. The spike pro-
Finally, our data also raise the question of transfected using PEI MAX 40kDa (Poly- tein was purified on an Akta Express system
why binding might be ablated in later variants science) if cells were >95% viable and had (GE Healthcare) using a 5-ml His trap FF GE
of SARS-CoV-2. Again by comparison with reached a density of 1.5 × 106 to 2 × 106 cells per Healthcare column in PBS, 40 mM imidazole,
influenza, which uses neuraminidases for the ml. Following transfection, cells were cultured pH 7.4, and eluted in PBS, 300 mM imidazole,
purpose of “release” when budding from a at 37°C and 5% CO2 at 120 rpm for 17 hours. pH 7.4. The protein was then injected onto
host cell (61), we speculate that in the absence Enhancers (6 mM valproic acid, 6.5 mM sodium either a Superdex 200 16/600 or 10/300 gel
of its own encoded neuraminidase, SARS- propionate, 50 mM glucose, all from Sigma) filtration column (GE Healthcare) in deuter-
CoV-2 must walk a tight balance between the were then added and protein was expressed ated PBS buffer, pH 7.4. The eluted protein
ability to bind human host glycans (poten- at 30°C for 5 days before purification. was concentrated using an Amicon Ultra-4
tially useful in a zoonotic leap) and cell-to- For delta and omicron spike, cDNA was 100kDa concentrator at 2000 rpm, 16°C (pre-
cell transmission (where release could become synthesized (IDT) as gBlock, flanked by KpnI washed multiple times with deuterated PBS)
rate-limiting). One answer to this problem and XhoI restriction sites based on the HexaPro to a concentration of roughly 1 mg/ml.
would be to ablate N-glycan binding via the spike sequence (63). HexaPro delta spike was
sialoside motif subsequent to a successful zoo- made based on the B-lineage-origin HexaPro Protein expression and purification: Influenza HA
notic leap. This solution also has the advan- spike with these additional mutations: T19R, Freestyle 293-F cells were cultured in Freestyle
tage of removing a potential site for antibody G142D, E156G, del157/158, L452R, T478K, D614G, expression media (Life Technologies) (37°C,
neutralization for an interaction that might P681R, D950N. HexaPro Omicron BA.1 spike 8% CO2, 115 rpm orbital shaking). Cells were
prove pivotal or critical in the context of zoo- is made based on the original Wuhan HexaPro transfected at a density of 109 cells/liter with
nosis as a potentially global driver of virus spike with these additional mutations: A67V, pre-incubated expression vector (300 mg/liter)
fitness. Our combined data and models may HV69-70 deletion, T95I, G142D, VYY143-145 and polyethyleneimine (PEI) MAX (Polysciences)
therefore support decades-old hypotheses (20) deletion, N211 deletion, L212I, ins214EPE, (900 mg/liter). Expression vectors encoded ter-
proposing the benefit of cryptic sugar binding G339D, S371L, S373P, S375F, K417N, N440K, minally His-tagged wild-type influenza A virus
by pathogens that may be “switched on and G446S, S477N, T478K, E484A, Q493R, G496S, (IAV) NC99 (H1N1) HA or a DRBS mutant
off” to drive fitness in a different manner (e.g., Q498R, N501Y, Y505H, T547K, D614G, H655Y, previously described (55). After 5 days, super-
in virulence or zoonosis) as needed. N679K, P681H, N764K, D796Y, F817P, N856K, natant was harvested and protein was purified
A892P, A899P, A942P, Q954H, N969K, L981F, via immobilized metal chromatography.
Methods K986P, V987P. Mutations for delta and omi-
Protein expression and purification: SARS-CoV-2 spike cron spike were guided by the following data- Protein expression and purification:
The templates for wild type, alpha, and beta bases: https://covariants.org/variants/21K. C5 anti-spike nanobody
spike were kindly provided by P. Supasa and Omicron and https://viralzone.expasy.org/ C5-Nanobody was purified as described (46).
G. Screaton (University of Oxford). The gene 9556. The cDNA fragment was digested, Purified C5 nanobody was then dialyzed into
encoding amino acids 1 to 1208 of the wild cleaned up, and ligated to the paH vector deuterated PBS buffer using 500 ml of Slide-A-
type, alpha, and beta SARS-CoV-2 spike glyco- backbone using T4 Ligase (the same backbone Lyzer cassette (3.5 kDa cutoff).
protein ectodomain [with mutations of RRAR → as the original HexaPro Spike (B-lineage-origin)
GSAS at residues 682–685 (the furin cleavage Addgene plasmid # 154754). Ligated plasmids Errors
site) and KV → PP at residues 986–987, as well were transformed in NEB DH5-alpha cells and The errors in the transfer efficiencies were
as inclusion of a T4 fibritin trimerization do- plated on agar with ampicillin. Colonies were estimated using a bootstrapping procedure.
main] was cloned into the pOPINTTGneo-BAP picked, cultured, and purified using the Qiagen Specifically sample STD spectra were assembled
vector using the forward primer (5′-GTCCAAG- HiSpeed MaxiPrep Kit and sent for Sanger through taking random combinations with re-
TTTATACTGAATTCCTCAAGCAGGCCACCAT- sequencing to confirm the identity. Expi293F placement of mixing times, and the analysis to

obtain the transfer efficiency was performed The total relaxation delay was set to 5 s, 1047.05, 2055.57, 2630.61, and 2979.65. The
on each. This process was repeated 100 times during which the saturation pulse was applied. saturation times used were 0.1 s, 0.5 s, 0.9 s,
to enable evaluation of the mean and standard The data were acquired in an interleaved fash- 2 s, 3 s, and 4 s.
deviation transfer efficiency for each residue. ion, with each individual excitation frequency Spectra were also acquired on a 600-MHz
Mean values correspond well with the value being repeated eight times (L4) until the total spectrometer with Bruker Avance III HD con-
from the original analysis, and so we take the desired number of scans was achieved. Again, sole and 5-mm TCI CryoProbe, running TopSpin
standard deviation as our estimate in un- the spectrum was centered on the water peak, 3.2.6, recorded in table S9, and a 950-MHz
certainty, which further is in accord with and the receiver gain was optimized. After spectrometer with Bruker Avance III HD con-
values obtained from independent repeated recording of the free induction decay (FID), sole and 5-mm TCI CryoProbe, running TopSpin
measurements. and prior to the recycle delay, a pair of water- 3.6.1, recorded in table S11. The 950-MHz spec-
selective pulses wee applied to destroy any trometer used a SampleJet sample changer.
Reagent sources unwanted magnetization. For all gradients Samples were stored at 15°C. The parameters
6′-Sialyllactose sodium salt and 3′-sialyllactose (excitation sculpting and spoil), the duration used for the STD experiments were the same as
sodium salt were purchased from Carbosynth was 3 ms using the smooth-square shape above, with the following varying by instrument:
and used directly (6′-sialyllactose sodium salt, SMSQ10.100. On the 600-MHz spectrometer, typical ac-
CAS-157574-76-0, 35890-39-2; 3′-sialyllactose In a typical experiment, two excitation fre- quisition parameters were sweep width of
sodium salt, CAS-128596-80-5, 35890-38-1). BSA quencies were required, one exciting protein, and 9615.39 Hz with typically 128 scans per tran-
and L-tryptophan were purchased from Sigma one exciting far from the protein (+20,000 Hz, sient (NS = 16 * L4 = 8), 32,768 complex points
Aldrich. Heparin sodium salt, from porcine +33 ppm from the carrier). A range of mixing in the direct dimension and two dummy scans,
intestinal mucosa, IU ≥ 100/mg was pur- times were acquired to allow us to carefully executed prior to data acquisition.
chased from Alfa Aesar. All other chemicals quantify the buildup curve to obtain KD values. On the 950-MHz spectrometer, typical ac-
were purchased from commercial suppliers A typical set of values used was 0.1 s, 0.3 s, 0.5 s, quisition parameters were sweep width of
(Alfa Aesar, Acros, Sigma Aldrich, Merck, 0.7 s, 0.9 s, 1.1 s, 1.3 s, 1.5 s, 1.7 s, 1.9 s, 2.0 s, 2.5 s, 15,243.90 Hz with typically 128 scans per
Carbosynth, Fisher, Fluorochem, VWR) and 3.0 s, 3.5 s, 4.0 s, and 5.0 s. transient (NS = 16 * L4 = 8), 32,768 complex
used as supplied, unless otherwise stated. See Off- and on-resonance spectra were acquired points in the direct dimension and 2 dummy
supplementary materials for syntheses of key for 16 saturation times, giving a total acquisi- scans, executed prior to data acquisition.
compounds. tion time of 8.7 hours.
The experiment was acquired as a pseudo-3D uSTA data analysis
Protein NMR experiments experiment, with each spectrum being acquired NMR spectra with a range of excitation fre-
All NMR experiments in table S8 were con- at a chosen set of excitation frequencies and quencies and mixing times were acquired on
ducted at 15°C on a Bruker AVANCE NEO 600 mixing times. Recycle delays were set to 10 s ligand-only, protein-only, and mixed protein/
MHz spectrometer with CPRHe-QR-1H/19F/ for BSA + tryptophan STDs, and were 5 s ligand samples (fig. S6).
13C/15N-5mm-Z helium-cooled cryoprobe. Sam- otherwise. To analyze an STD dataset, two projections
ples were stored in a Bruker SampleJet sample For STD 10- to 50-ms Gaussian experiments, were created by summing over all 1D spectra
loader while not in magnet, at 4°C. the saturation times used were every other and summing over all corresponding STD
1D 1H NMR spectra with w5 water sup- time from the default STD: 0.1 s, 0.5 s, 0.9 s, spectra. These two projections provide ex-
pression were acquired using the Bruker pulse 1.3 s, 1.7 s, 2 s, 3 s, 4 s. ceptionally high signal-to-noise, suitable for
sequence zggpw5, using the smooth square List1: For STD var freq 1, the on-resonance detailed analysis and reliable peak detection.
Bruker shape SMSQ.10.100 for the pulsed-field frequencies in Hz relative to an offset of The UnidecNMR algorithm was first executed
gradients. The spectrum was centered on the 2820.61 Hz are: 337.89, 422.36, 524.93, 736.11, on the 1D “pulse off” spectra to identify peak
water peak, and the receiver gain was ad- 914.10, 1276.12, 1336.46, 1380.21, 1458.64, 1556.69, positions and intensities. Having identified
justed. Typical acquisition parameters were 1693.96, 2494.93, 2597.50, 2790.58, 2930.86, possible peak positions, the algorithm then
sweep width of 9615.39 Hz, 16 scans per tran- 3362.27, 3663.95, 3986.75, 4099.88, 4326.15, analyzes the STD spectra but only allowing
sient (NS), with four dummy scans, 32,768 4484.53, 4703.25, 4896.33, 5824.01, 6006.53, resonances in places already identified in the
complex points (TD), and a recycle delay (d1) and 6208.65. The saturation times used were 1D spectrum. Both analyses are conducted using
of 1 s for a total acquisition time of 54 s. Ref- 2 s, 3 s, 4 s, and 5 s. the protein-only baselines for accurate effective
erence 1D spectra of protein-only samples List2: For STD var freq 2, the on-resonance subtraction of the protein baseline without the
were acquired similarly with 16,384 scans frequencies in Hz relative to an offset of need to use relaxation filters (fig. S8).
per transient with a total acquisition time of 2820.61 Hz are: –2399.99, –1979.99, –1530.00, The ligand-only spectra were analyzed sim-
12.5 hours. –1050.01, –330.021, 338.096, 1679.95, 1829.94, ilarly and in each case, excellent agreement
An STD experiment with excitation sculpted 1979.94, 2129.94, 2279.94, and 2579.93. The with the known assignments was obtained,
water suppression was developed from the saturation times used were 0.1 s, 0.5 s, 2 s, providing us with confidence in the algorithm.
Bruker pulse sequence stddiffesgp.2. The sat- and 5 s. The mixed protein/ligand spectrum was then
uration was achieved using a concatenated List3: For STD var freq 3, the on-resonance analyzed, which returned results very similar
series of 50-ms Gaussian-shaped pulses to frequencies in Hz relative to an offset of 2820.61 Hz to the ligand-only case. Contributions from the
achieve the desired total saturation time are: –2579.98, –2459.99, –2339.99, –2039.99, protein, although small, were typically evident
(d20). The shape of the pulses was specified –1488.00, –1120.03, –345.02, 311.97, 1079.96, in the spectra, justifying the explicit inclusion
by the Bruker shape file Gaus.1.1000, where 1379.95, 1679.95, 1979.94, 2279.94, and 2579.93. of the protein-only baseline during the analy-
the pulse is divided into 1000 steps and the The saturation times used were 0.1 s, 0.3 s, 0.5 s, sis. When analyzing the mixture, we included
standard deviation for the Gaussian shape and 0.9 s. the protein-only background as a peak shape
is 165 steps. The field of the pulse was set List4: For STD var freq 4, the on-resonance whose contribution to the spectrum could
to 200 Hz, which was calculated internally frequencies in Hz relative to an offset of be freely adjusted. In this way, the spectra of
through scaling the power of the high-power 2820.61 Hz are: –2461.55, –1973.77, –1518.69, protein/ligand mixtures could be accurately
90° pulse. –1270.72, –693.69, –274.80, 280.08, 808.02, and quickly deconvolved, with the identified

ligand resonances occurring in precisely the were quantitatively analyzed as described below this calculation cannot be routinely used to fit
positions expected from the ligand-only spectra. to obtain KD and koff rates. The values we obtain to experimental data.
The results from the previous steps were then performing this analysis on BSA/Trp closely It would be very desirable to extract quan-
used to analyze the STD spectra. As these have match those measured by ITC, and the values titative structural parameters, as well as chem-
much lower signal-to-noise, we fixed the ligand we measure for ligand 2 and Spike are in good ical properties such as interaction strengths
peak positions to be only those previously iden- agreement with those measured by SPR as de- and association/dissociation rates, directly from
tified. Otherwise the protocol performed as described in the text. STD data. In what follows, we develop a simple
scribed previously, where we used protein-only The coverage of protons over the ligands quantitative model for the STD experiment to
STD data to provide a baseline. studied here was variable; for example, there achieve this goal. We will treat the system as
These analyses allow us to define a “transfer are no protons on carboxyl groups. To enable comprising just two spins, one to represent the
efficiency,” which is simply the ratio of the sig- a complete surface to be rendered, the transfer ligand and one to represent the protein, and we
nal from a given multiplet in the STD spectrum efficiencies for each proton were calculated allow the two spins to exist either in isolation or
to the total expected in the raw 1D experiment. as described above, and the value is then trans- in a bound state. We can safely neglect scalar
To obtain “per atom” transfer efficiencies, sig- ferred to the adjacent heteroatom. For hetero- coupling and so we only need to allow the x, y,
nals from the various pre-assumed components atoms not connected to an observed proton, a and z basis operators for each spin, together
on the multiplets from each resonance were 1/r6 weighted average score was calculated. This with an identity operator to ensure that the
first summed before calculating the ratio. In the approach allows us to define a unique surface. system returns to thermal equilibrium at long
software, this is achieved by manually anno- Caution should be exercised when quantita- times. As such, our evolution matrix R will be
tating the initial peak list using information tively interpreting such surfaces where there a square matrix with 13 × 13 elements.
obtained from independent-assignment experi- are no direct measurements of the heteroatom. For the spin part, our model requires us to
ments (see figs. S21 and S22). In practice, raw unformatted FIDs are sub- consider the chemical shift of the ligand in the
Over the course of the project, it became mitted to the uSTA pipeline, and the various free and bound states, and the chemical shifts
clear that subtracting the transfer efficiencies steps described are performed largely auto- of the protein in the free and bound states. In
obtained from a ligand-only sample was an matically, where a user needs to manually practice however, it is sufficient to set the free
essential part of the method (figs. S9 and S10). adjust processing settings such as phasing and protein state on resonance with the pulse, and
Depending on the precise relationship among choosing which regions to focus on, iteratively the free ligand chemical shift is set to a value
the chemical shift of excitation, the location adjust the peak shape to get a good match be- that matches experiment.
of the ligand peaks, and the excitation profile tween the final reconvolved spectrum and the The longitudinal and transverse relaxation
of the Gaussian train, we observed small ap- raw data, and input manual atomic assign- rates are calculated for the free and bound
parent STD transfer in the ligand-only sample ments for each observed multiplet. The uSTA states using a simple model assuming in each
that cannot be attributed to ligand binding, pipeline then provides a user with a report state there are two dipole-coupled spins sepa-
arising from a small residual excitation of lig- that shows the results of the various stages rated by a distance R with similar Larmor fre-
and protons, followed by internal cross-relaxation. of analysis, and uses pymol to render the sur- quency. In addition, cross-relaxation between
It is likely that this excitation occurs at least faces. The final transfer efficiencies delivered ligand and protein is allowed only when the
in part via resonances of the ligand that are by the program can be combined with a folder two are bound. The relaxation rates are char-
exchange-broadened, such as OH protons, which containing a series of HADDOCK models to acterized by an effective distance and an effec-
are not directly observed in the spectrum. When provide final structural models (Fig. 5). tive correlation time,
exciting far from the protein, zero ligand ex-
citation is observed, as we would expect, but Quantitative analysis via uSTA 1
R1 ¼ K ½J ð0Þ þ 3J ðwÞ þ 6J ð2wÞ
when exciting close to the methyls, or in the In principle, a complete description of the 4
aromatic region, residual ligand excitation could saturation transfer experiment can be achieved
be detected in ligand-only samples (figs. S9 and via the Bloch-McConnell equations. If we can 1 5 9
R2 ¼ K J ð0Þ þ J ðwÞ þ 3J ð2wÞ
S10). Without the ligand-only correction, the set up a density matrix describing all the spins 4 2 2
uSTA surface may appear to be highly depen- in the system, their interactions, and their rates 1
dent on choice of excitation frequency. How- of chemical change in an evolution matrix R, s ¼ K ½6J ð2wÞ J ð0Þ
4
ever, with the ligand correction, the relative then we can follow the system with time ac-
uSTA profiles become invariant with exci- cording to: which are each parameterized in terms of an
tation frequency. In general, therefore, we interaction constant (depending on effective
rðt Þ ¼ rð0Þexpð Rt Þ
advise acquiring these routinely, and so the distance)
uSTA analysis assumes the presence of these The challenge comes from the number of spins 2
m0 ħg2H
data (figs. S6 and S8). The invariance of rela- that must be included and the need to accu- K¼
4pr3
tive transfer efficiency with excitation fre- rately describe all the interactions between them,
quency suggests that the internal evolution which will need to also include how these are and a spectral density function (depending
of magnetization within the protein during modulated by molecular motions in order to on effective distance)
saturation (likely on the micro-/millisecond get an accurate description of the relaxation 2 t
time scales) is much faster than the effective processes. This is illustrated by the CORCEMA J ðwÞ ¼
5 1 þ w2 t2
cross-relaxation rate between protein and method (64) that takes a static structure of a
ligand (occurring on the seconds time scale). protein/ligand complex and estimates STD The longitudinal and transverse relaxation rates
Having identified the relevant resonances transfers. The CORCEMA calculations per- R1 and R2 describe auto-relaxation of diagonal z
of interest and performed both a protein and formed to arrive at cross-relaxation rates as- elements and xy elements, respectively. The
residual ligand subtraction, we reanalyzed the sume the complex is rigid, which is often a cross-relaxation rates s describe cross-relaxation
spectra without first summing over the dif- poor approximation for a protein, and because and couple z elements between the ligand and
ferent mixing times, in order to develop the of the large number of spins involved, the protein in the bound state. We ensure that the
quantitative atom-specific build-up curves. These calculation is sufficiently intensive such that system returns to equilibrium at long times by

adding elements of the form R1M0 or sM0 cross relaxation. From tG and rIS(ligand) we For analysis of spike binding, a flow rate of
linking the identify element and the z matrix estimate R1 and R2 of the ligand; from tE and 10 ml/min was used at 16°C. Serial dilutions of
elements. Overall, the relaxation part of the rIS(protein) we obtain R1 and R2 of the protein; spike (0.19, 0.50, 1.36, and 3.68 mM) were in-
model is parameterized by two correlation times, and from tE and rIS(complex) we calculate the jected for 30 s association and 150 s dissocia-
one for the ligand and one for the protein/ cross-relaxation rate. These values are com- tion starting with the lowest concentration.
complex, and three distances, one for the ligand bined with a factor that accounts for the larger Buffer-only runs were carried out before in-
auto-relaxation rates, one for protein auto- number of spins present in the protein, “fac,” jection of spike and after the first two dilutions.
relaxation rates, and one for the protein/ligand and the on and off rates, to complete a set of BSA (3.03 mM in PBS) was used as a negative
separation. eight parameters that specify our model. The control, and a mouse serum in a 100-fold dilution
Finally, the chemical kinetics govern the distances should be considered “effective” was used as a positive control.
rates at which the spins can interconvert. We values that parameterize the relaxation rates,
will take a simple model where PL ⇆ P + L, although in principle it should be possible to Analysis of SPR data
whose dissociation constant is given by obtain physical insights from their interpreta- To analyze the SPR data, we assume an equi-
koff ½P ½L tion. The concentration-independent relaxa- librium of the form PL ⇆ P + L characterized
KD ¼ ¼ tion rates can be separated from the exchange by a dissociation constant
kon ½PL
rates by comparing the curves as a function of koff ½P½L
The free protein concentration can be deter- ligand and protein concentration. By treating KD ¼ ¼
kon ½PL
mined from knowledge of the KD and the the system as comprising two spins, we are
total ligand and protein concentrations: effectively assuming that the cross-relaxation To follow the kinetics of binding and dissoci-
within the protein is very efficient. In the STD ation, we assume that the SPR response G is
1
½P ¼ PTot LTot KD þ experiment, saturation pulses are applied for proportional to the bound complex G = k[PL],
2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi several seconds, which is sufficient for near- which leads to the following kinetic equation:
ðLTot þ KD PTot Þ2 þ 4PTot KD saturating spin diffusion within a protein. dG koff kon
Because of the complexity of the model, op- þ G ½P ½L ¼ 0
dt k k
from which the bound protein concentration timization of the parameters via a gradient
and the free and bound ligand concentrations descent method can get stuck in local minima. This can be solved when restrained by the
can be easily calculated. In practical applications, it is advisable to start total number of binding sites, Ltot = [L] + [PL].
The density matrix is initialized with the the optimization over a range of initial condi- Under conditions of constant flow, we assume
free and bound protein/ligand concentrations tions, particularly in the rates, to ensure that that the free protein concentration is constant,
assigned to the relevant z operators. It was the lowest possible c2 is achieved. which leads to the following:
found to be important to additionally include
a factor that accounts for the increased proton Thermostability assays Gon ¼
density within the protein. The saturation Thermal stability assays were performed using kkon Ltot ½P
f1 exp½ ðkoff þ kon ½P Þt g
pulse is then applied either as a concatenated a NanoTemper Prometheus NT.48 (Membrane koff þ kon ½P
series of Gaussian pulses whose duration and Protein Laboratory, Diamond Light Source). To
peak power in Hz needs to be specified, exactly 11 ml of 2 mM spike (deuterated PBS), 2 ml of And similarly, for dissociation where we take
matching the pulse shapes and durations used trisaccharide 2 (deuterated PBS) was titrated to the concentration of free protein to be zero:
in the experiment (see NMR methods above). give final concentrations of 0.1, 0.2, 0.4, 0.8, 1.6,
Build-up curves and transfer efficiencies can and 2.0 mM. Samples were then loaded into Goff ¼ kRe koff t
be easily simulated using this model and com- capillaries and heated from 15° to 95°C. Anal-
pared to data, and the various parameters can ysis was performed using PR.ThermControl The recovery of the chip was not complete after
be optimized to fit to the data. In total, the v2.3.1 software. each protein concentration and wash step, as
model is characterized by eight parameters: has been observed for shear-induced lectin-
KD, koff, the correlation times of the ligand SPR binding measurement assays ligand binding with glycans immobilized onto
and the protein, the three distances described All experiments were performed on a Biacore a chip surface (65). Nonetheless, the data were
above, and the proton density within the pro- T200 instrument. For the immobilization of well explained by a global analysis where the
tein. There is substantial correlation between SiaLac onto the sensor chip, a flow rate of on and off rates were held to be identical for
the effects of the various parameters, and care 10 ml/min was used in a buffer solution of HBP- each replicate, but the value of k was allowed
is needed using optimization to avoid local EP (0.01 M HEPES pH 7.4, 0.15 M NaCl, 3 mM to vary slightly between runs, and an additional
minima. By obtaining data at various protein EDTA, 0.005% v/v surfactant P20). A CM5 sen- constant was introduced to Goff to account for
and ligand concentrations, however, it is pos- sor chip (carboxymethylated dextran) was equil- incomplete recovery of the SPR signal following
sible to break this degeneracy and obtain well- ibrated with HBS-EP buffer at 20°C. The chip standard approaches. Concentration of spike
described values as in the text. was activated by injecting a mixture of N- was insufficient to get the plateau region of the
In practical terms, the initial rate of the build- hydroxysuccinimide (50 mM) and EDC-HCl binding, and so the specific time values taken
up curve is predominantly affected by the cross- (200 mM) for 10 min followed by a 2-min for the on rate affect the fitted values.
relaxation rate and the off rate, and the final wash step with buffer. Ethylenediamine (1 M
height of the build-up curve is mostly influ- in PBS) was then injected for 7 min followed by Modeling of the N-terminal domain of
enced by the proton density in the protein and a 2-min wash step followed by ethanolamine- SARS-CoV-2 with glycans
KD. Software to perform this analysis has been HCl (1 M, pH 8.5) for 10 min and then a further We modeled the structure of the NTD on Pro-
directly incorporated into the uSTA software. 1-min wash step. Finally, SiaLac-IME (5.6 mM tein Data Bank (PDB) entry 7c2l (19) because it
in PBS) reagent 10 was injected over 10 min provided much better coverage of the area of
Parameters fitted by the model and a final 2-min wash step was performed interest when compared to the majority of the
Overall the model is parameterized by a set (see fig. S13 and supplementary materials for templates available at PDB as of 15 July 2020.
of values that characterize the intrinsic and further details). The models were created with Modeller (66),

using the “automodel” protocol without refin- two datasets were 60 e/Å2 (5492 movies) and regression model. Specifically, for a set of n
ing the “loop.” We generated 10 models and 61 e/Å2 (8284 movies), respectively. samples {xi, yi} (i = 1, …, n), each of which
ranked them by their DOPE score (67), selecting Motion correction was performed by consists of p input features xi,k ∈ ci, k = 1, …, p
the top five for ensemble docking. MotionCor2 (69). The motion-corrected micro- and one output variable yi ∈ Y, these features
graphs were imported into cryoSPARC (70) assumed the meaning of genes, whereas the
Docking of 3′-sialyllactose to SARS-CoV-2 NTD and contrast transfer function values were samples were the patients involved in the
We docked 3′-sialyllactose to NTD with version estimated using Gctf (71) in cryoSPARC. Tem- study. The space c = c1 × c2 … × cp was
2.4 of the HADDOCK webserver (42, 43). The plates were produced by 2D classification from denoted “input space,” whereas the “hypothesis
binding site on NTD was defined by compar- 5492 micrographs with particles auto-picked space” was the space of all the possible func-
ison with PDB entry 6q06 (5), a complex of by Laplacian-of-Gaussian (LoG)–based algorithm tions f: c → Y mapping the inputs to the
MERS-CoV spike protein and 2,3-sialyl-N- in RELION 3.0 (72, 73). Particles were picked output. Given that the number of features
acetyl-lactosamine. The binding site could from all micrographs using Template picker (p) is substantially higher than the number
not be directly mapped because of confor- in cryoSPARC. Multiple rounds of 2D classi- of samples (n), LASSO regularization (49) has
mational differences between the NTDs of fication were carried out and the selected 2D the effect of shrinking the estimated coefficients
MERS-CoV and SARS-CoV-2, but by inspection classes (372,157 particles) were subjected to to zero, providing a feature selection method for
a region with similar properties (aromatics, 3D classification (Heterogeneous Refinement sparse solutions within the classification tasks.
methyl groups, and positively charged residues) in cryoSPARC) using six classes. One class was Feature selection methods based on such regu-
could be identified. We defined in HADDOCK predominant after 3D classification. Nonuniform larization structures (embedded methods) were
the sialic acid as “active” and residues 18, 19, 20, refinement (74) was performed for this class most applicable to our scope because they were
21, 22, 68, 76, 77, 78, 79, 244, 254, 255, 256, 258, (312,018 particles) with C1 and C3 symmetry, computationally tractable and strictly con-
and 259 of NTD as “passive,” meaning the sialic respectively, yielding a 2.27 Å map for C3 nected with the classification task of the ML
acid needs to make contact with at least one of symmetry and a 2.44 Å map for C1 symmetry. algorithm.
the NTD residues but there is no penalty if it See also fig. S23 and table S12 for cryo-EM data As the baseline algorithm for the embedded
doesn’t contact all of them, thus allowing the collection, refinement, and validation statistics. method, we adopted the logistic regression (LR)
compound to freely explore the binding pocket. model that is a state-of-the-art ML algorithm
Because only one restraint was used, we dis- Genetic analysis of clinical samples for binary classification tasks with probabilistic
abled the random removal of restraints. Follow- Variant calling: Reads were mapped to the hg19 interpretation. It models the log-odds of the
ing our small-molecule docking recommended reference genome by the Burrow-Wheeler posterior success probability of a binary var-
settings (68), we skipped the “hot” parts of aligner BWA. Variants calling was performed iable as the linear combination of the input:
the semi-flexible simulated annealing protocol according to the GATK4 best practice guide-
(“initiosteps” and “cool1_steps” set to 0) and lines. Namely, duplicates were first removed PrðY ¼ 1jX ¼ xÞ Xp
log ¼ b0 þ bk xk
also lowered the starting temperature of the by MarkDuplicates, and base qualities were 1 PrðY ¼ 1jX ¼ xÞ k¼1
last two substages to 500 and 300 K, respec- recalibrated using BaseRecalibration and
tively (“tadinit2_t” and “tadinit3_t” to 500 and ApplyBQSR. HaplotypeCaller was used to where x is the input vector, bk are the co-
300, respectively). Clustering was performed calculate Genomic VCF files for each sample, efficients of the regression, and X and Y are
based on “RMSD” with a distance cutoff of 2 Å, which were then used for multi-sample calling the random variables representing the input
and the scoring function was modified to by GenomicDBImport and GenotypeGVCF. In and the output, respectively. The loss function
order to improve the specificity-sensitivity bal- to be minimized is given by the binary cross-
ance, variants’ quality scores were calculated by entropy loss
HADDOCKscore ¼ 1:0∗EvdW þ 0:1∗Eelec
VariantRecalibrator and ApplyVQSR, and only
þ 1:0∗Edesol þ 0:1∗EAIR X
n
variants with estimated truth sensitivity above î ð1
½yi log y yi Þlogð1 ^ i Þ
y
99.9% were retained. Variants were annotated i¼1
All other settings were kept to their default by ANNOVAR.
values. Finally, the atom-specific transfer effi- Rare variant selection: Missense, splicing, where y ^ = Pr(Y = 1|X = x) is the predicted
ciencies determined by uSTA were used to and loss-of-function variants with a frequency target variable and y is the true label. As
filter cluster candidates. lower than 0.01 according to ExAC_NFE (Non already introduced, in order to enforce both
Finnish European ExAC Database) were con- the sparsity and the interpretability of the re-
Cryo-EM analysis sidered for further analyses. A score of 0 was sults, the model is trained with the additional
SARS-CoV-2 spike protein, generated and pu- assigned to each sample where the gene is not LASSO regularization term
rified as described (48), in 1.1 mg/ml was incu- mutated, and a score of 1 was assigned when
bated with 10 mM ethyl(triiodobenzamide) at least one variant is present on the gene. X
p
l jbk j
siallyllactoside overnight at 4°C. A 3.5-ml sam- The cohort was distributed as follows. Eth- k¼1
ple was applied to glow-discharged Quantifoil nicity: 504 white, 4 Black, 5 Asian, 16 Hispanic
gold R1.2/1.3 300-mesh grids and blotted for ethnicity, 4 patients for which this information In this way, the absolute value of the surviving
~3 s at 100% humidity and 6°C before vitrifi- was not available. Sex: 317 male, 216 female. weights of the LR algorithm was interpreted
cation in liquid ethane using Vitrobot (FEI). Age: minimum age 19 years, maximum age as the feature importance of the subset of most
Two datasets were collected on Titan Krios 99 years, mean age 62.5 years. relevant genes for the task. Because a feature-
equipped with a K2 direct electron detector ranking criterion can become suboptimal when
at the cryo-EM facility (OPIC) in the Division Gene prioritization by logistic regression the subset of removed features is large (75), we
of Structural Biology, University of Oxford. Discriminating genes in COVID-19 disease were applied recursive feature elimination (RFE)
Both datasets were collected by SerialEM at a interpreted in a framework of feature selection methodology. For each step of the procedure, we
magnification of 165,000× with a physical analysis using a customized feature selection fitted the model and removed the features with
pixel size of 0.82 Å per pixel. Defocus range approach based on the recursive feature elimi- smallest ranking criteria in a recursive manner
was –0.8 mm to –2.4 mm. Total doses for the nation algorithm applied to the LASSO logistic until a certain number of features was reached.

The fundamental hyperparameter of LR is the 13. A. N. Baker et al., The SARS-COV-2 Spike Protein Binds Sialic 35. U. Hars, R. Horlacher, W. Boos, W. Welte, K. Diederichs, Crystal
strength of the LASSO term tuned with a grid Acids and Enables Rapid Detection in a Lateral Flow Point structure of the effector-binding domain of the trehalose-
of Care Diagnostic Device. ACS Cent. Sci. 6, 2046–2052 repressor of Escherichia coli, a member of the LacI family, in
search procedure on the accuracy of the 10-fold (2020). doi: 10.1021/acscentsci.0c00855; pmid: 33269329 its complexes with inducer trehalose-6-phosphate and
cross-validation. The k-fold cross-validation pro- 14. W. Hao et al., Binding of the SARS-CoV-2 spike protein to noninducer trehalose. Protein Sci. 7, 2511–2521 (1998).
vided the partition of the dataset into k batches, glycans. Sci. Bull. 66, 1205–1214 (2021). doi: 10.1016/ doi: 10.1002/pro.5560071204; pmid: 9865945
then exploited k–1 batches for the training and

j.scib.2021.01.010; pmid: 33495714 36. H. M. McConnell, Reaction Rates by Nuclear Magnetic
15. L. Nguyen et al., Sialic acid-containing glycolipids mediate Resonance. J. Chem. Phys. 28, 430–431 (1958). doi: 10.1063/
the remaining test batch as a test, by repeating binding and viral entry of SARS-CoV-2. Nat. Chem. Biol. 18, 1.1744152
this procedure k times. In the grid search 81–90 (2022). doi: 10.1038/s41589-021-00924-1 37. J. E. Stencel-Baerenwald, K. Reiss, D. M. Reiter, T. Stehle,
16. W. Hao et al., Binding of the SARS-CoV-2 Spike Protein to T. S. Dermody, The sweet spot: Defining virus-sialic acid
method, a cross-validation procedure was car- Glycans. bioRxiv 2020.2005.2017.100537 (2020). doi: 10.1101/ interactions. Nat. Rev. Microbiol. 12, 739–749 (2014).
ried out for each value of the regularization 2020.05.17.100537 doi: 10.1038/nrmicro3346; pmid: 25263223
hyperparameter in the range [10–4, ..., 106]. 17. T. M. Clausen et al., SARS-CoV-2 Infection Depends on Cellular 38. D. Lingwood et al., Structural and genetic basis for
Heparan Sulfate and ACE2. Cell 183, 1043–1057.e15 (2020). development of broadly neutralizing influenza antibodies.
Specifically, the optimal regularization param-
doi: 10.1016/j.cell.2020.09.033; pmid: 32970989 Nature 489, 566–570 (2012). doi: 10.1038/nature11371;
eter is chosen by selecting the most parsimo- 18. N. Behloul, S. Baha, R. Shi, J. Meng, Role of the GTNGTKR pmid: 22932267
nious parameter whose cross-validation average motif in the N-terminal receptor-binding domain of the 39. H.-Y. Liao et al., Differential receptor binding affinities of
accuracy falls in the range of the best one along SARS-CoV-2 spike protein. Virus Res. 286, 198058 (2020). influenza hemagglutinins on glycan arrays. J. Am. Chem. Soc.
doi: 10.1016/j.virusres.2020.198058; pmid: 32531235 132, 14849–14856 (2010). doi: 10.1021/ja104657b;
with its standard deviation. During the fitting 19. X. Chi et al., A neutralizing human antibody binds to the pmid: 20882975
procedure, the class unbalancing was tackled by N-terminal domain of the Spike protein of SARS-CoV-2. 40. Y. Watanabe, J. D. Allen, D. Wrapp, J. S. McLellan, M. Crispin,
penalizing the misclassification of minority class Science 369, 650–655 (2020). doi: 10.1126/science.abc6952; Site-specific glycan analysis of the SARS-CoV-2 spike.
pmid: 32571838 Science 369, 330–333 (2020). doi: 10.1126/science.abb9983;
with a multiplicative factor inversely proportional 20. M. G. Rossmann, The canyon hypothesis. J. Biol. Chem. 264, pmid: 32366695
to the class frequencies. For the RFE, the number 14587–14590 (1989). doi: 10.1016/S0021-9258(18)63732-9; 41. M. P. Lenza et al., Structural Characterization of N-Linked
of included features at each step of the algorithm pmid: 2670920 Glycans in the Receptor Binding Domain of the SARS-CoV-2
21. Y. Watanabe et al., Vulnerabilities in coronavirus glycan shields Spike Protein and their Interactions with Human Lectins.
as well as the final number of features was fixed despite extensive glycosylation. Nat. Commun. 11, 2688 Angew. Chem. Int. Ed. 59, 23763–23771 (2020). doi: 10.1002/
at 100. All data preprocessing and the RFE pro- (2020). doi: 10.1038/s41467-020-16567-0; pmid: 32461612 anie.202011015; pmid: 32915505
cedure were coded in Python; the LR model 22. M. Mayer, B. Meyer, Characterization of Ligand Binding 42. C. Dominguez, R. Boelens, A. M. J. J. Bonvin, HADDOCK:
by Saturation Transfer Difference NMR Spectroscopy. Angew. A protein-protein docking approach based on biochemical or
was used, as included, in the scikit-learn module Chem. Int. Ed. 38, 1784–1788 (1999). doi: 10.1002/ biophysical information. J. Am. Chem. Soc. 125, 1731–1737
with the liblinear coordinate descent optimiza- (SICI)1521-3773(19990614)38:12<1784::AID-ANIE1784>3.0. (2003). doi: 10.1021/ja026939x; pmid: 12580598
tion algorithm. CO;2-Q; pmid: 29711196 43. G. C. P. van Zundert et al., The HADDOCK2.2 Web Server:
23. J. L. Wagstaff, S. L. Taylor, M. J. Howard, Recent developments User-Friendly Integrative Modeling of Biomolecular Complexes.
and applications of saturation transfer difference nuclear J. Mol. Biol. 428, 720–725 (2016). doi: 10.1016/
RE FE RENCES AND N OT ES magnetic resonance (STD NMR) spectroscopy. Mol. Biosyst. 9, j.jmb.2015.09.014; pmid: 26410586
571–577 (2013). doi: 10.1039/C2MB25395J; pmid: 23232937 44. A. Rambaut et al., A dynamic nomenclature proposal for
1. S. Yamada et al., Haemagglutinin mutations responsible for the
24. J. Angulo, P. M. Enríquez-Navas, P. M. Nieto, Ligand-receptor SARS-CoV-2 lineages to assist genomic epidemiology.
binding of H5N1 influenza A viruses to human-type receptors.
binding affinities from saturation transfer difference (STD) Nat. Microbiol. 5, 1403–1407 (2020). doi: 10.1038/s41564-
Nature 444, 378–382 (2006). doi: 10.1038/nature05264;
NMR spectroscopy: The binding isotherm of STD initial growth 020-0770-5; pmid: 32669681
pmid: 17108965
rates. Chem. Eur. J. 16, 7803–7812 (2010). doi: 10.1002/ 45. J. Huo et al., Neutralizing nanobodies bind SARS-CoV-2
2. W. Li et al., Identification of sialic acid-binding function for the
chem.200903528; pmid: 20496354 spike RBD and block interaction with ACE2. Nat. Struct. Mol.
Middle East respiratory syndrome coronavirus spike
25. R. H. McMenamy, J. L. Oncley, The specific binding of L- Biol. 27, 846–854 (2020). doi: 10.1038/s41594-020-0469-6;
glycoprotein. Proc. Natl. Acad. Sci. U.S.A. 114, E8508–E8517
tryptophan to serum albumin. J. Biol. Chem. 233, 1436–1447 pmid: 32661423
(2017). doi: 10.1073/pnas.1712592114; pmid: 28923942
(1958). doi: 10.1016/S0021-9258(18)49353-2; pmid: 13610854 46. J. Huo et al., A potent SARS-CoV-2 neutralising nanobody
3. C. Schwegmann-Wessels, G. Herrler, Sialic acids as receptor
26. P. Vallurupalli, G. Bouvignies, L. E. Kay, Studying “invisible”
determinants for coronaviruses. Glycoconj. J. 23, 51–58 shows therapeutic efficacy in the Syrian golden hamster model
excited protein states in slow exchange with a major state of COVID-19. Nat. Commun. 12, 5469 (2021). doi: 10.1038/
(2006). doi: 10.1007/s10719-006-5437-9; pmid: 16575522
conformation. J. Am. Chem. Soc. 134, 8148–8161 (2012). s41467-021-25480-z; pmid: 34552091
4. R. J. G. Hulswit et al., Human coronaviruses OC43 and HKU1
doi: 10.1021/ja3001419; pmid: 22554188 47. M. McCallum, A. C. Walls, J. E. Bowen, D. Corti, D. Veesler,
bind to 9-O-acetylated sialic acids via a conserved receptor-
27. T. Xie, T. Saleh, P. Rossi, C. G. Kalodimos, Conformational Structure-guided covalent stabilization of coronavirus spike
binding site in spike protein domain A. Proc. Natl. Acad.
states dynamically populated by a kinase determine its
Sci. U.S.A. 116, 2681–2690 (2019). doi: 10.1073/ glycoprotein trimers in the closed conformation. Nat. Struct.
function. Science 370, eabc2754 (2020). doi: 10.1126/science. Mol. Biol. 27, 942–949 (2020). doi: 10.1038/s41594-020-
pnas.1809667116; pmid: 30679277
abc2754; pmid: 33004676 0483-8; pmid: 32753755
5. Y.-J. Park et al., Structures of MERS-CoV spike glycoprotein
28. N. L. Fawzi, J. Ying, D. A. Torchia, G. M. Clore, Probing 48. C. Toelzer et al., Free fatty acid binding pocket in the
in complex with sialoside attachment receptors. Nat. Struct.
exchange kinetics and atomic resolution dynamics in locked structure of SARS-CoV-2 spike protein. Science 370,
Mol. Biol. 26, 1151–1157 (2019). doi: 10.1038/s41594-019-
high-molecular-weight complexes using dark-state exchange 725–730 (2020). doi: 10.1126/science.abd3255;
0334-7; pmid: 31792450
saturation transfer NMR spectroscopy. Nat. Protoc. 7, pmid: 32958580
6. E. Qing, M. Hantak, S. Perlman, T. Gallagher, Distinct Roles for
1523–1533 (2012). doi: 10.1038/nprot.2012.077; pmid: 22814391 49. R. Tibshirani, Regression Shrinkage and Selection Via the
Sialoside and Protein Receptors in Coronavirus Infection.
mBio 11, e02764-19 (2020). doi: 10.1128/mBio.02764-19; 29. G. Bouvignies et al., Solution structure of a minor and transiently Lasso. J. R. Stat. Soc. B 58, 267–288 (1996). doi: 10.1111/
pmid: 32047128 formed state of a T4 lysozyme mutant. Nature 477, 111–114 j.2517-6161.1996.tb02080.x
7. W. Li et al., Angiotensin-converting enzyme 2 is a functional (2011). doi: 10.1126/science.abc2754; pmid: 33004676 50. K. Koths, E. Taylor, R. Halenbeck, C. Casipit, A. Wang, Cloning
receptor for the SARS coronavirus. Nature 426, 450–454 30. M. Mayer, B. Meyer, Group epitope mapping by saturation and characterization of a human Mac-2-binding protein, a new
(2003). doi: 10.1038/nature02145; pmid: 14647384 transfer difference NMR to identify segments of a ligand in member of the superfamily defined by the macrophage
8. K. Kuba et al., A crucial role of angiotensin converting enzyme direct contact with a protein receptor. J. Am. Chem. Soc. 123, scavenger receptor cysteine-rich domain. J. Biol. Chem. 268,
2 (ACE2) in SARS coronavirus-induced lung injury. Nat. Med. 6108–6117 (2001). doi: 10.1021/ja0100120; pmid: 11414845 14245–14249 (1993). doi: 10.1016/S0021-9258(19)85233-X;
11, 875–879 (2005). doi: 10.1038/nm1267; pmid: 16007097 31. M. T. Marty et al., Bayesian deconvolution of mass and ion pmid: 8390986
9. F. Li, W. Li, M. Farzan, S. C. Harrison, Structure of SARS mobility spectra: From binary interactions to polydisperse 51. L. Johannes, R. Jacob, H. Leffler, Galectins at a glance.
coronavirus spike receptor-binding domain complexed with ensembles. Anal. Chem. 87, 4370–4376 (2015). doi: 10.1021/ J. Cell Sci. 131, jcs208884 (2018). doi: 10.1242/jcs.208884;
receptor. Science 309, 1864–1868 (2005). doi: 10.1126/ acs.analchem.5b00140; pmid: 25799115 pmid: 29717004
science.1116480; pmid: 16166518 32. L. Fielding, S. Rutherford, D. Fletcher, Determination of protein- 52. S. R. Stowell et al., Galectin-1, -2, and -3 exhibit differential
10. D. Wrapp et al., Cryo-EM structure of the 2019-nCoV spike in ligand binding affinity by NMR: Observations from serum recognition of sialylated glycans and blood group antigens.
the prefusion conformation. Science 367, 1260–1263 (2020). albumin model systems. Magn. Reson. Chem. 43, 463–470 J. Biol. Chem. 283, 10109–10123 (2008). doi: 10.1074/jbc.
doi: 10.1126/science.abb2507; pmid: 32075877 (2005). doi: 10.1002/mrc.1574; pmid: 15816062 M709545200; pmid: 18216021
11. R. Yan et al., Structural basis for the recognition of SARS-CoV-2 33. A. Bujacz, K. Zielinski, B. Sekula, Structural studies of bovine, 53. N. A. Kamili et al., Key regulators of galectin-glycan
by full-length human ACE2. Science 367, 1444–1448 (2020). equine, and leporine serum albumin complexes with naproxen. interactions. Proteomics 16, 3111–3125 (2016). doi: 10.1002/
doi: 10.1126/science.abb2762; pmid: 32132184 Proteins 82, 2199–2208 (2014). doi: 10.1002/prot.24583; pmic.201600116; pmid: 27582340
12. D. Morniroli, M. L. Giannì, A. Consales, C. Pietrasanta, F. Mosca, pmid: 24753230 54. A. Togayachi et al., b3GnT2 (B3GNT2), a major polylactosamine
Human Sialome and Coronavirus Disease-2019 (COVID-19) 34. I. Pérez-Victoria et al., Saturation transfer difference NMR synthase: Analysis of B3GNT2-deficient mice. Methods Enzymol.
Pandemic: An Understated Correlation? Front. Immunol. 11, reveals functionally essential kinetic differences for a sugar- 479, 185–204 (2010). doi: 0.1016/s0076-6879(10)79011-x
1480 (2020). doi: 10.3389/fimmu.2020.01480; binding repressor protein. Chem. Commun. 2009, 5862–5864 55. M. Kanekiyo et al., Self-assembling influenza nanoparticle
pmid: 32655580 (2009). doi: 10.1039/b913489a; pmid: 19787122 vaccines elicit broadly neutralizing H1N1 antibodies.

Nature 499, 102–106 (2013). doi: 10.1038/nature12202; Mach. Learn. 46, 389–422 (2002). doi: 10.1023/ funding the project “PAT-COVID: Host genetics and pathogenetic
pmid: 23698367 A:1012487302797 mechanisms of COVID-19” (ID n. 2020-2016_RIC_3). We thank EU
56. M. Mammen, S.-K. Choi, G. M. Whitesides, Polyvalent 76. C. Buchanan et al., Pathogen-sugar interactions revealed by project H2020-SC1-FA-DTS-2018-2020, entitled “International
Interactions in Biological Systems: Implications for Design and universal saturation transfer analysis. Zenodo, doi: 10.5281/ consortium for integrative genomics prediction (INTERVENE)” -
Use of Multivalent Ligands and Inhibitors. Angew. Chem. Int. zenodo.6299883 (2022).doi: 10.5281/zenodo.6299883 grant agreement 101016775. Author contributions: B.G., C.J.B.,
Ed. 37, 2754–2794 (1998). doi: 10.1002/(SICI)1521-3773 77. P. I. Koukos, A. M. J. J. Bonvin, Modelling of the binding of M.J.D., T.D.W.C., and A.J.B. ran NMR experiments. A.J.B. and C.J.B.
(19981102)37:20<2754::AID-ANIE2754>3.0.CO;2-3; various sialic acid-containing oligosaccharides to the NTD developed the underlying deconvolution algorithm, UnidecNMR.
pmid: 29711117 domain of the spike protein of SARS-CoV2. Zenodo, A.J.B. developed the modified Bloch-McConnell KD analysis method.
57. R. T. Lee, Y. C. Lee, Affinity enhancement by multivalent lectin- doi: 10.5281/zenodo.4271288 (2020). A.J.B. and C.J.B. performed the processing, uSTA applications,
carbohydrate interaction. Glycoconj. J. 17, 543–551 (2000). 78. C. J. Buchanan, A. J. Baldwin, Universal Saturation Transfer and surrounding analysis. C.J.B. and A.K. assigned key uSTA
doi: 10.1023/A:1011070425430; pmid: 11421347 Analysis software. Zenodo, doi: 10.5281/zenodo.6303964 spectra. P.H. and A.L.B. purified spike protein and prepared samples
58. N. Jia et al., The Human Lung Glycome Reveals Novel Glycan (2022). for NMR analysis; P.H. and A.L.B. conducted thermal denaturation
Ligands for Influenza A Virus. Sci. Rep. 10, 5320 (2020). 79. N. Picchiotti et al., Post-Mendelian genetic model in COVID-19. assays. A.K. synthesized 9-BPC-Neu5Ac, expressed and purified
doi: 10.1038/s41598-020-62074-z; pmid: 32210305 Cardiol. Cardiovasc. Med. 5, 673–694 (2021). doi: 10.26502/ NmCSS and Pd2,6ST, assigned proton peaks in 1H NMR spectra of
59. M. A. Toscano et al., Differential glycosylation of TH1, TH2 and fccm.92920232 sialosides, designed TOCSY NMR experiments, and performed the
TH-17 effector cells selectively regulates susceptibility to 80. C. Fallerini et al., Common, low-frequency, rare, and ultra-rare SPR experiments. A.M.G., A.K., and X.X. synthesized 5, 6, and 9. AMG
cell death. Nat. Immunol. 8, 825–834 (2007). doi: 10.1038/ coding variants contribute to COVID-19 severity. Hum. Genet. generated reagents for SPR studies. A.K., A.M.G., and A.J.B. analyzed
ni1482; pmid: 17589510 141, 147–173 (2022). doi: 10.1007/s00439-021-02397-7; SPR data. G.P. generated 7 and 8 and assigned their 1H NMR
60. P. Tangvoranuntakul et al., Human uptake and incorporation of pmid: 34889978 spectra. J.H. cloned spike-BAP; P.W., M.D., L.S., and T.K.T.
an immunogenic nonhuman dietary sialic acid. Proc. Natl. expressed spike protein. A.L. aided design of protein-ligand NMR
Acad. Sci. U.S.A. 100, 12045–12050 (2003). doi: 10.1073/ experiments, prepared samples, and contributed to data analysis.
ACKN OWLED GMEN TS
pnas.2131556100; pmid: 14523234 A.R. is coordinating the GEN-COVID Consortium and supervised
61. J. L. McAuley, B. P. Gilbertson, S. Trifkovic, L. E. Brown, We thank X. Chen (University of California, Davis) for providing the the genotype-phenotype correlation of glycosylation-associated
J. L. McKimm-Breschkin, Influenza Virus Neuraminidase Structure plasmid encoding Pd2,6ST; J. Mascola (NIH) for the plasmids genes. A.G. performed WES experiments. E.B. performed
and Functions. Front. Microbiol. 10, 39 (2019). doi: 10.3389/ encoding IAV NC99 (H1N1) HA variants; P. Supasa and G. Screaton WES alignment and joint call, which was supervised by S.F. N.P.
fmicb.2019.00039; pmid: 30761095 (University of Oxford) for templates for alpha and beta spike; performed logistic regression analysis, which was supervised by
62. J. Huo et al., Neutralization of SARS-CoV-2 by Destruction of M. Kutuzov and O. Dushek for useful discussions and assistance M.G. M.B. and F.F. collected and supervised the collection of
the Prefusion Spike. Cell Host Microbe 28, 445–454.e6 (2020). with SPR experiments; A. Quigley and N. Gamage [Wellcome Trust clinical data. C.F. and S.D. performed specific analysis on B3GNT8
doi: 10.1016/j.chom.2020.06.010; pmid: 32585135 Membrane Protein Lab, Diamond Light Source (20289/Z16/Z)] and LGALS3BP genes. L.P.D. and Q.J.S. performed HA expression
63. C.-L. Hsieh et al., Structure-based design of prefusion- for help with thermal stability assays; C. Redfield for assistance and assisted with experimental design. A.M.J.J.B. and P.I.K.
stabilized SARS-CoV-2 spikes. Science 369, 1501–1505 (2020). with high-field Bruker instrumentation; and L. Baker for insights into performed the structural modeling of the SARS-CoV-2 spike-sugar
doi: 10.1126/science.abd0826; pmid: 32703906 the structural models. Funding: Upgrades of 600-MHz and 950-MHz interactions. A.J.B. and A.M.J.J.B. scored these against the NMR
64. V. Jayalakshmi, N. R. Krishna, Complete relaxation and spectrometers were funded by the Wellcome Trust (grant ref: data to obtain structural models for the complex. V.C.-A., C.J.B.,
conformational exchange matrix (CORCEMA) analysis of 095872/Z/10/Z) and the Engineering and Physical Sciences A.J.B., and B.G.D. produced figures. J.H.N., Y.Y., and J.L. performed
intermolecular saturation transfer effects in reversibly forming Research Council (grant ref: EP/R029849/1), respectively, and by the EM experiments and analysis. I.B., C.S., and K.G. produced
ligand-receptor complexes. J. Magn. Reson. 155, 106–118 the University of Oxford Institutional Strategic Support Fund, the the protein used in EM. B.G.D., A.J.B., T.D.W.C., and J.H.N.
(2002). doi: 10.1006/jmre.2001.2499; pmid: 11945039 John Fell Fund, and the Edward Penley Abraham Cephalosporin supervised and designed experiments, analyzed data, and wrote
65. K. Nakamura et al., Immobilized glycosylated Fmoc-amino Fund. We thank the Wellcome Trust for support (20289/Z/16/Z the manuscript. All authors read and commented on the manuscript.
acid for SPR: Comparative studies of lectin-binding to linear and 100209/Z/12/Z). I.B. is an Investigator of the Wellcome Competing interests: The uSTA software is freely available to all
or biantennary diLacNAc structures. Carbohydr. Res. 382, Trust (106115/Z/14/Z). Cryo-EM time was funded by a philanthropic academic users. In the event that the uSTA is licensed by a
77–85 (2013). doi: 10.1016/j.carres.2013.10.003; gift to support COVID-19 research at the University of Oxford. The commercial party, C.J.B. and A.J.B. will then be afforded royalties in
pmid: 24211369 Chemistry theme at the Rosalind Franklin Institute is supported line with standard university practice. All authors declare that they
66. A. Sali, Comparative protein modeling by satisfaction of spatial by the EPSRC (V011359/1 (P)). This project has received funding have no other competing interests. Data and materials availability:
restraints. Mol. Med. Today 1, 270–277 (1995). doi: 10.1016/ from the European Research Council (ERC) under the European The plasmid for the C5 nanobody is available on Addgene
S1357-4310(95)91170-7; pmid: 9415161 Union’s Horizon 2020 research and innovation programme (plasmid #171925). Raw spectral data for Figs. 1 to 5 and the data
67. M. Y. Shen, A. Sali, Statistical potential for assessment and (grant agreement 101002859). A.M.J.J.B. and P.I.K. acknowledge used to plot Fig. 4, A to C have been deposited (76). Associated
prediction of protein structures. Protein Sci. 15, 2507–2524 financial support from the European Union Horizon 2020 projects spectral data are also shown in table S7. HADDOCK datasets
(2006). doi: 10.1110/ps.062416606; pmid: 17075131 BioExcel (823830) and EOSC-Hub (777536) and the IMI-CARE and results have been deposited (77). All associated clinical data
68. P. I. Koukos, L. C. Xue, A. M. J. J. Bonvin, Protein-ligand pose project (101005077). This study is part of the GEN-COVID have been supplied in tables S1 to S6, conducted under trial
and affinity prediction: Lessons from D3R Grand Challenge 3. Multicenter Study (https://sites.google.com/dbm.unisi.it/gen-covid), (NCT04549831, www.clinicaltrial.org). The uSTA software is free
J. Comput. Aided Mol. Des. 33, 83–91 (2019). doi: 10.1007/ the Italian multicenter study aimed at identifying the COVID-19 for academic use and has been deposited (78). The coordinates
s10822-018-0148-4; pmid: 30128928 host genetic bases. Specimens were provided by the COVID-19 and EM maps have been deposited with the RCSB: Spike_native_C1
69. S. Q. Zheng et al., MotionCor2: Anisotropic correction of beam- Biobank of Siena, which is part of the Genetic Biobank of (EMDB-14155), Spike_native_C3 (EMDB-14153) (PDB 7QUS).
induced motion for improved cryo-electron microscopy. Siena, member of BBMRI-IT, of Telethon Network of Genetic Spike_ sialic acid _C1 (EMDB-14154) and Spike_ sialic acid_C3
Nat. Methods 14, 331–332 (2017). doi: 10.1038/nmeth.4193; Biobanks (project GTB18001), of EuroBioBank, and of RDConnect. (EMDB-14152) (PDB 7QUR). License information: This work
pmid: 28250466 We thank the CINECA consortium for providing computational is licensed under a Creative Commons Attribution 4.0 International
70. A. Punjani, J. L. Rubinstein, D. J. Fleet, M. A. Brubaker, resources and the Network for Italian Genomes (NIG; www.nig. (CC BY 4.0) license, which permits unrestricted use, distribution,
cryoSPARC: Algorithms for rapid unsupervised cryo-EM cineca.it) for its support. We thank private donors for the support and reproduction in any medium, provided the original work is
structure determination. Nat. Methods 14, 290–296 (2017). provided to AR (Department of Medical Biotechnologies, University properly cited. To view a copy of this license, visit https://
doi: 10.1038/nmeth.4169; pmid: 28165473 of Siena) for the COVID-19 host genetics research project creativecommons.org/licenses/by/4.0/. This license does not
71. K. Zhang, Gctf: Real-time CTF determination and correction. (D.L n.18 of March 17, 2020). We also thank the COVID-19 Host apply to figures/photos/artwork or other content included
J. Struct. Biol. 193, 1–12 (2016). doi: 10.1016/j.jsb.2015.11.003; Genetics Initiative (www.covid19hg.org/), MIUR project in the article that is credited to a third party; obtain authorization
pmid: 26592709 “Dipartimenti di Eccellenza 2018–2020” to the Department of from the rights holder before using such material.
72. S. H. W. Scheres, RELION: Implementation of a Bayesian Medical Biotechnologies University of Siena, Italy, and “Bando
approach to cryo-EM structure determination. J. Struct. Biol. Ricerca COVID-19 Toscana” project to Azienda Ospedaliero–
180, 519–530 (2012). doi: 10.1016/j.jsb.2012.09.006; Universitaria Senese. We thank Intesa San Paolo for the 2020 SUPPLEMENTARY MATERIALS
pmid: 23000701 charity fund dedicated to the project N.B/2020/0119
science.org/doi/10.1126/science.abm3125
73. J. Zivanov et al., New tools for automated high-resolution cryo- “Identificazione delle basi genetiche determinanti la variabilità
Supplementary Methods
EM structure determination in RELION-3. eLife 7, e42166 clinica della risposta a COVID-19 nella popolazione italiana” and the
Supplementary Text
(2018). doi: 10.7554/eLife.42166; pmid: 30412051 Tuscany Region for funding within “Bando Ricerca COVID-19
Figs. S1 to S23
74. A. Punjani, H. Zhang, D. J. Fleet, Non-uniform refinement: Toscana” project supporting research at the Azienda Ospedaliero-
Tables S1 to S12
Adaptive regularization improves single-particle cryo-EM Universitaria Senese. The Italian Ministry of University and
References (81–102)
reconstruction. Nat. Methods 17, 1214–1221 (2020). Research for funding within the “Bando FISR 2020” in COVID-19 fo
doi: 10.1038/s41592-020-00990-8; pmid: 33257830 the project “Editing dell’RNA contro il Sars-CoV-2: hackerare il Submitted 16 September 2021; accepted 16 June 2022
75. I. Guyon, J. Weston, S. Barnhill, V. Vapnik, Gene Selection virus per identificare bersagli molecolari e attenuare l’infezione - Published online 23 June 2022
for Cancer Classification using Support Vector Machines. HACKTHECOV” and the Istituto Buddista Italiano Soka Gakkai for 10.1126/science.abm3125

Pushing the Boundaries of Knowledge
As AAAS’s first multidisciplinary, open access journal, Science Advances publishes
research that reflects the selectivity of high impact, innovative research you expect
from the Science family of journals, published in an open access format to serve
a vast and growing global audience. Check out the latest findings or learn how to
submit your research: ScienceAdvances.org
G O L D O P E N A C C E S S , D I G I TA L , A N D F R E E T O A L L R E A D E R S
RES EARCH
◥ nitrogen uptake and transport activity. More-

RESEARCH ARTICLE SUMMARY over, OsDREB1C overexpression led to more
efficient carbon and nitrogen allocation from
PLANT SCIENCE source to sink, thus boosting grain yield, par-
ticularly under low-nitrogen conditions. Addi-
A transcriptional regulator that boosts grain yields tionally, the OsDREB1C-OE plants flowered 13 to
19 days earlier and accumulated higher biomass
and shortens the growth duration of rice at the heading stage than WT plants under long-
day conditions.
Shaobo Wei†, Xia Li†, Zefu Lu, Hui Zhang, Xiangyuan Ye, Yujie Zhou, Jing Li, Yanyan Yan, OsDREB1C is localized in the nucleus and
Hongcui Pei, Fengying Duan, Danying Wang, Song Chen, Peng Wang, Chao Zhang, Lianguang Shang, the cytosol and functions as a transcriptional
Yue Zhou, Peng Yan, Ming Zhao, Jirong Huang, Ralph Bock, Qian Qian, Wenbin Zhou* activator that directly binds to cis elements in
the DNA, including dehydration-responsive ele-
ment (DRE)/C repeat (CRT), GCC, and G boxes.
INTRODUCTION: Rapid population growth, ris- these transcription factors for their respon- Chromatin immunoprecipitation sequencing
ing meat consumption, and the expanding use siveness to light and nitrogen supply in rice, and (ChIP-seq) and transcriptomic analyses identi-
of crops for nonfood and nonfeed purposes found that the gene Dehydration-Responsive fied a total of 9735 putative OsDREB1C-binding
increase the pressure on global food produc- Element-Binding Protein 1C (OsDREB1C), a sites at the genome-wide level. We discovered
tion. At the same time, the excessive use of member of the APETALA2/ethylene-responsive that five genes targeted by OsDREB1C [ribulose-
nitrogen fertilizer to enhance agricultural element binding factor (AP2/ERF) family, ex- l,5-bisphosphate carboxylase/oxygenase small
production poses serious threats to both human hibits properties expected of a regulator that subunit 3 (OsRBCS3), nitrate reductase 2
health and the environment. To achieve the can simultaneously modulate photosynthesis (OsNR2), nitrate transporter 2.4 (OsNRT2.4),
required yield increases and make agriculture and nitrogen utilization. nitrate transporter 1.1B (OsNRT1.1B), and
more sustainable, intensified breeding and flowering locus T-like 1 (OsFTL1)] are closely
genetic engineering efforts are needed to obtain RESULTS: OsDREB1C expression is induced in associated with photosynthesis, nitrogen utili-
new crop varieties with higher photosynthetic rice by both light and low-nitrogen status. We zation, and flowering, the key traits altered by
capacity and improved nitrogen use efficiency generated overexpression lines (OsDREB1C- OsDREB1C overexpression. ChIP-quantitative
(NUE). However, progress has been slow, largely OE) and knockout mutants (OsDREB1C-KO) polymerase chain reaction (ChIP-qPCR) and
due to the limited knowledge about regulator in rice, and conducted field trials in northern, DNA affinity purification sequencing (DAP-seq)
genes that potentially can coordinately optimize southeastern, and southern China from 2018 assays confirmed that OsDREB1C activates the
carbon assimilation and nitrogen utilization. to 2021. OsDREB1C-OE plants exhibited 41.3 transcription of these genes by binding to the
to 68.3% higher yield than wild-type (WT) plants promoter of OsRBCS3 and to exons of OsNR2,
RATIONALE: Transcription factors control diverse due to increased grain number per panicle, OsNRT2.4, OsNRT1.1B, and OsFTL1. By show-
biological processes by binding to the promoters elevated grain weight, and enhanced harvest ing that biomass and yield increases can also
(or intragenic regions) of target genes, and a index. We observed that light-induced growth be achieved by OsDREB1C overexpression in
number of transcription factors have been promotion of OsDREB1C-OE plants was ac- wheat and Arabidopsis, we have demonstrated
identified that control carbon fixation and companied by enhanced photosynthetic capac- that the mode of action and the biological
nitrogen assimilation. A previous comparative ity and concomitant increases in photosynthetic function of the transcription factor are evolu-
analysis of maize and rice leaf transcriptomes assimilates. In addition, 15N feeding experi- tionarily conserved.
and metabolomes revealed a set of 118 can- ments and field studies with different nitrogen
didate transcription factors that may act as fertilization regimes revealed that NUE was im- CONCLUSION: Overexpression of OsDREB1C
regulators of C4 photosynthesis. We screened proved in OsDREB1C-OE plants due to elevated not only boosts grain yields but also confers
higher NUE and early flowering. Our work
demonstrates that by genetically modulating
Transcription factors Rice: Wheat: the expression of a single transcriptional reg-
related to C4 yield increases yield increases ulator gene, substantial yield increases can be
photosynthesis 41.3 – 68.3% 17.2 – 22.6% achieved while the growth duration of the crop
is shortened. The existing natural allelic varia-
Light OsDREB1C Early
OsFTL1 tion in OsDREB1C, the highly conserved func-
flowering
Low nitrogen tion of the transcription factor in seed plants,
C
and the ease with which its expression can be
OsDREB1C
OsRBCS3 Enhanced altered by genetic engineering suggest that
photosynthesis
this gene could be the target of future crop im-
OsDREB1C OsDREB1C
provement strategies toward more efficient
▪
N
OsNR2
OsDREB1C
Improved and more sustainable food production.
OsNRT1.1B nitrogen use
OsDREB1C
efficiency
OsNRT2.4
C Carbon assimilated by photosynthesis The list of author affiliations is available in the full article online.
N Nitrogen taken up by roots *Corresponding author. Email: [email protected]
These authors contributed equally to this work.
Cite this article as S. Wei et al., Science377, eabi8455
OsDREB1C coordinates yield and growth duration. OsDREB1C was identified by its responsiveness to light and (2022). DOI: 10.1126/science.abi8455
low nitrogen in a screen of 118 transcription factors related to C4 photosynthesis. Transcriptional activation of
multiple downstream target genes by OsDREB1C confers enhanced photosynthesis, improved nitrogen utilization, READ THE FULL ARTICLE AT
and early flowering. Together, the activated genes cause substantial yield increases in rice and wheat. https://doi.org/10.1126/science.abi8455
386 22 JULY 2022 ¥ VOL 377 ISSUE 6604 science.org SCIENCE

RES EARCH
◥ Overexpression of OsDREB1C not only in-

RESEARCH ARTICLE creases rice yields, but also confers early grain
maturation because of higher rates of photo-
PLANT SCIENCE synthesis, improved nitrogen utilization, and
early flowering. The versatile functions of
A transcriptional regulator that boosts grain yields OsDREB1C are likely conferred by the tran-
scription factor acting near the top of the
and shortens the growth duration of rice hierarchy and coordinately targeting multiple
genes and pathways. Recognition of the con-
Shaobo Wei1†, Xia Li1†, Zefu Lu1, Hui Zhang2, Xiangyuan Ye1, Yujie Zhou1, Jing Li1, Yanyan Yan1, served dehydration-responsive element (DRE)/
Hongcui Pei1, Fengying Duan1, Danying Wang3, Song Chen3, Peng Wang4, Chao Zhang5, C repeat (CRT) motif present in these loci
Lianguang Shang5, Yue Zhou6, Peng Yan6, Ming Zhao1, Jirong Huang2, Ralph Bock7, facilitates fine-tuning of the intricate networks
Qian Qian1,3, Wenbin Zhou1* of carbon assimilation, nitrogen utilization, re-
source allocation, and induction of flowering.
Complex biological processes such as plant growth and development are often under the control Our results uncover OsDREB1C as a transcrip-
of transcription factors that regulate the expression of large sets of genes and activate subordinate tional regulator that promotes the expression
transcription factors in a cascade-like fashion. Here, by screening candidate photosynthesis-related of key genes in carbon and nitrogen metabo-
transcription factors in rice, we identified a DREB (Dehydration Responsive Element Binding) family lism and controls flowering pathways in crops,
member, OsDREB1C, in which expression is induced by both light and low nitrogen status. We show that thus providing a target for future crop improve-
OsDREB1C drives functionally diverse transcriptional programs determining photosynthetic capacity, ment strategies.
nitrogen utilization, and flowering time. Field trials with OsDREB1C-overexpressing rice revealed yield
increases of 41.3 to 68.3% and, in addition, shortened growth duration, improved nitrogen use efficiency, RESULTS
and promoted efficient resource allocation, thus providing a strategy toward achieving much-needed OsDREB1C boosts grain yield and harvest index
increases in agricultural productivity. To investigate how the coordination of photo-
synthesis and nitrogen utilization affects grain
G
yield in cereals, we conducted RNA-sequencing
lobally, >800 million people are suffering As the regulators of biological processes, (RNA-seq) analyses in rice plants grown under
from hunger and food insecurity (1). transcription factors control plant metabo- low- versus high-nitrogen conditions (18). We
By 2050, crop production needs to be lism, growth, and development by binding to examined the differential expression of the
increased by 50 to 70% to feed nearly the promoters (or intragenic regions) of target subset of transcription factors associated with
10 billion people despite the reduced genes (9, 10). An example of a transcription photosynthesis gene expression in maize (17).
availability of arable land on the planet (2, 3). factor in plant architecture is TB1 (Teosinte We detected nitrogen-regulated expression
Meeting this challenge will likely require the Branched 1) of maize, which limits branch for 13 of these transcription factors, with five
development of new breeding and genetic en- outgrowth and initiates the formation of fe- genes showing a greater than fourfold induc-
gineering strategies that optimize photosynthetic male inflorescences (11, 12). The transcription tion under low-nitrogen conditions (Fig. 1A).
capacity as well as water and nutrient use factor IPA1 (Ideal Plant Architecture 1) pro- Investigation of light-regulated mRNA accu-
efficiency (4). Growth and crop yield depend on motes rice yield by reducing unproductive mulation showed that one of the five genes,
carbon and nitrogen assimilation and photo- tillers and increasing grain number per panicle. Os06g0127100 (encoding a DREB-type tran-
synthate translocation from vegetative source Elevated IPA1 levels also enhance pathogen scription factor previously shown to be inducible
organs to sink tissues (5, 6). For example, nitro- immunity (13, 14). Another transcription factor, by abiotic stress), displayed a diurnal rather
gen uptake and transport must be coordinated HYR (HIGHER YIELD RICE), enhances the than a circadian expression profile, with expres-
with carbon fixation and the production of expression of photosynthesis genes and can sion increasing with the duration of light ex-
carbohydrates by photosynthesis. Therefore, increase rice yield under multiple stress con- posure (Fig. 1B and fig. S1).
research efforts have been devoted to iden- ditions (15). The rice transcription factor GRF4 To facilitate the functional analysis of this
tifying transcriptional regulators that control (GROWTH-REGULATING FACTOR4) co- apparently nitrogen-regulated and light-induced
the coordination between carbon assimilation ordinates nitrogen assimilation, carbon fix- transcription factor, we generated a series of
and nitrogen utilization (7, 8). ation, and growth (7). Often, binding motifs OsDREB1C-overexpressing lines (OsDREB1C-OE)
and functions of transcription factors are con- and OsDREB1C-knockout mutants (OsDREB1C-
1
served in monocot and dicot species (7, 16). KO) in the Oryza sativa cv. Nipponbare genetic
Institute of Crop Sciences, Chinese Academy of Agricultural
Sciences, Beijing 100081, China. 2Shanghai Key Laboratory Previous work has been directed at iden- background (fig. S2). Field tests of these plant
of Plant Molecular Sciences, College of Life Sciences, tifying key transcription factors that regulate lines in Beijing in 2018 revealed that OsDREB1C
Shanghai Normal University, Shanghai 200234, China. 3State photosynthesis, and nitrogen and carbon meta- overexpression led to increases in grain yield
Key Laboratory of Rice Biology, China National Rice
Research Institute, Chinese Academy of Agricultural
bolism, using comparative analysis of maize and per plant of 45.1 to 67.6% and in yield per plot
Sciences, Hangzhou 310006, China. 4CAS Center for rice leaf transcriptomes and metabolomes (17). A of 41.3 to 68.3% compared with wild-type (WT)
Excellence in Molecular Plant Sciences, Institute of Plant set of 118 transcription factors were considered plants (Fig. 1C). Conversely, OsDREB1C KO
Physiology and Ecology, Chinese Academy of Sciences,
Shanghai 200032, China. 5Lingnan Laboratory of Modern
as candidate regulators of photosynthesis and, resulted in yield decreases (from 16.1 to 29.1%
Agriculture, Genome Analysis Laboratory of the Ministry of especially, of favorable properties related to C4 in yield per plant and 13.8 to 27.8% in yield
Agriculture, Agricultural Genomics Institute at Shenzhen, photosynthesis (17). Here, we screened these per plot) compared with the WT (Fig. 1, D
Chinese Academy of Agricultural Sciences, Shenzhen 518124,
transcription factors for their responsiveness and E, and table S1). A detailed phenotypic
China. 6State Key Laboratory of Protein and Plant Gene
Research, School of Advanced Agricultural Sciences, Peking- to light and nitrogen supply in rice. We report analysis showed that the higher yield of the
Tsinghua Center for Life Sciences, Peking University, Beijing the identification of a transcription factor from OsDREB1C-OE lines was mainly attributable
100871, China. 7Max Planck Institute of Molecular Plant the DREB (Dehydration Responsive Element to an enhanced grain number per panicle and
Physiology, Am Mühlenberg, 14476 Potsdam-Golm, Germany.
*Corresponding author. Email: [email protected] Binding) family, OsDREB1C, that modulates an increased 1000-grain weight (Fig. 1F and fig.
These authors contributed equally to this work. both photosynthesis and nitrogen utilization. S3C), traits apparently resulting from increased
Wei et al., Science 377, eabi8455 (2022) 22 July 2022 1 of 10

A B 5 C D
Fold change 75
4
* OE2 WT KO1
Relative expression level

60
** ** **
Grain yield per plant (g)

5 10
3
Low N High N
*
** 45
** **
Os01g0705700 7.64 1.00 2 ** ** **
30
Os01g0108600 6.23 1.00
1
15
Os04g0301500 5.78 1.00
Os06g0127100 4.79 1.00 0 0
E1
1
E2
3
E5
2
T
11 pm
7 m
3 m
am
am
9 m
11 am
11 m
11 pm
am
7 m
8 m
am
pm
pm
KO
KO
KO
W
Os03g0741100 4.50 1.00
a
p
a
a
a
O
11
7
8
7
3
Time (h)
E F G 150 H
Os05g0579300 3.23 1.00
Os01g0191300 2.81 1.00 1.5 300 0.8

**
** * ** ** **
Grain number per panicle

Os01g0848400 2.12 1.00
** **
Grain yield per plot (kg)

** 0.6
Straw weight (g)

Os04g0546800 2.11 1.00 **
Harvest index
1.0 200 100
** ** **
**
Os01g0868000 2.06 1.00 ** **
** 0.4 ** **
Os09g0434500 2.06 1.00 * ** **
0.5 100 50
Os05g0537400 2.05 1.00 0.2
Os04g0636500 2.04 1.00

0.0 0 0 0.0
E1
E1
1
1
E1
E1
1
E2
E2
3
3
E2
E2
E5
E5
2
3
3
T
T
E5
E5
2
T
T
KO
KO
KO
KO
KO
KO
KO
KO
KO
KO
KO
KO
W
W
W
W
O
O
O
O
O
O
O
O
O
O
O
O
Fig. 1. OsDREB1C overexpression in transgenic plants boosts grain yield. the x-axis indicates the light period, and the black bar indicates the dark period. Data
(A) List of the top 13 genes up-regulated in response to nitrogen deprivation are presented as means ± SD (n = 3 biological replicates). *P < 0.05, **P < 0.01
(adjusted P < 0.05). The genes represent the overlap of previously reported RNA- compared with the first time point (11:00 p.m.), Student’s t test. (C) Phenotypes of
seq datasets (17) and an expression analysis of a subset of 118 rice transcription WT and transgenic rice plants grown in Beijing in 2018. (D to H) Yield-related
factors (16), and were sorted by the fold change in low versus normal nitrogen parameters including grain yield per plant (D), grain yield per plot (E), grain
supply. The color scale represents the log2-fold change of the FPKM (fragments per number per panicle (F), straw weight (G), and harvest index (H). The data were
kilobase of transcript per million mapped reads) ratio under low- versus high- obtained from the field experiment shown in (C). Box plots in (D) and (F) to
nitrogen conditions, with the FPKM value of each gene under high-nitrogen (H) show median (horizontal lines) and 10th to 90th percentiles, and outliers are
conditions set to 1.00. (B) qRT-PCR analysis of Os06 g0127100 expression in plotted as dots (n = 138 biological replicates). Data in (E) are presented as
10-day-old O. sativa cv. Nipponbare seedlings grown in soil in a growth chamber means ± SD (n = 3 plots, 44 plants within a plot). *P < 0.05, **P < 0.01
under long-day photoperiod (16 hours light/8 hours dark, 28°C). The white bar below compared with WT, Student’s t tests.
secondary branch number and grain length, that the most pronounced yield increases were the dark (figs. S4, C to F, and S6), although
width, thickness, and density (fig. S3, A and B detected in the long-day photoperiod and tem- the OsDREB1C-OE lines showed taller shoots
and D to K, respectively). The OsDREB1C-OE perate climate conditions of Beijing. Never- during the first 10 days, likely because of the
plants exhibited higher grain yield but re- theless, there was still a strong yield improvement larger grain size, which provides more re-
duced straw weight compared with WT plants for OsDREB1C-OE plants in the short-day and serves for initial growth (fig. S6). Overall, these
(Fig. 1G), thus leading to an increased harvest tropical conditions of Hainan, with yield in- data suggested a light-induced mechanism
index (the ratio of grain yield to aboveground creases in the range of 7.8 to 16% yield per of growth improvement. We next investigated
biomass; Fig. 1H) and raising the possibility plant and 12.0 to 37.0% yield per plot (table whether photosynthetic capacity is improved
that OsDREB1C controls resource allocation S4 and fig. S3N). Also, the harvest index of by the overexpression of OsDREB1C. Leaves of
between vegetative and reproductive tissues. OsDREB1C-OE plants (~0.62) was higher than OsDREB1C-OE plants contained higher levels
The harvest index of OsDREB1C-OE plants in WT (~0.54) and OsDREB1C-KO (~0.32) plants of photosynthetic pigments (chlorophylls and
was increased by 40.3 to 55.7%, whereas it was in Hainan (table S4). carotenoids) compared with WT plants, where-
decreased by 22.4 to 33.7% in OsDREB1C-KO as pigment levels were reduced in OsDREB1C-
plants (table S1). In addition, key grain quality OsDREB1C improves photosynthetic capacity KO plants (fig. S7A). Analysis of leaf mesophyll
traits were enhanced in OsDREB1C-OE plants, and nitrogen utilization cells revealed that both chloroplast number
suggesting that yield improvement does not To further explore the molecular basis of the and size were increased in OsDREB1C-OE plants
entail a quality penalty (table S2). yield enhancement conferred by OsDREB1C (fig. S7, B and C). Biochemical analysis of photo-
To assess the stability of the yield enhance- overexpression, a series of physiological mea- synthetic protein complexes by blue-native poly-
ment conferred by OsDREB1C, we conducted surements were conducted with hydroponi- acrylamide gel electrophoresis revealed elevated
field trials over several years and at three dif- cally grown WT and transgenic plants. In these levels of photosystem I (PSI) and PSII dimers,
ferent sites that represent very different envi- experiments, we observed faster growth of PSII-CP43 monomers, and light-harvesting com-
ronmental conditions (tables S1 and S3 to S5). OsDREB1C-OE plants already at the seedling plex II (LHCII) trimers in OsDREB1C-OE plants
Data for the Beijing field trial in 2019 showed stage, whereas the growth of OsDREB1C-KO (fig. S7D). Immunoblotting confirmed an in-
an even larger increase in grain yield for seedlings was retarded compared with WT creased abundance of PSI, PSII, cytochrome
OsDREB1C-OE plants than in 2018 and similar plants (fig. S4, A and B). In addition, we no- b6/f, ATP synthase, and LHC proteins in leaves
yield reductions in OsDREB1C-KO plants (table ticed that OsDREB1C-OE plants displayed of the overexpression lines (fig. S7E). OsDREB1C
S3). Having tested in temperate (Beijing; tables longer roots, probably related to their increased overexpression also led to enhanced amounts
S1 and S3 and fig. S3M), tropical (Hainan auxin content (fig. S5). We detected no growth of the large and small subunits of ribulose bis-
Province; table S4 and fig. S3N), and subtropical differences among the WT, overexpression, and phosphate carboxylase/oxygenase (RbcL and
(Zhejiang Province; table S5) locations, we noted KO plants when the seedlings were cultivated in RbcS, respectively) and ribulose-1,5-bisphosphate

Fig. 2. OsDREB1C overexpression promotes A 5 WT OE1 OE2 OE5 B 0.4 WT OE1 OE2 OE5
RubisCO content (g m-2)

KO1 KO2 KO3 KO1 KO2 KO3
photosynthetic capacity. (A and B) RuBisCO 4
* **
RubisCO activity
0.3
(µmol mg-1 s-1)

content (A) and RuBisCO activity (B) in the fourth
3 ** **
leaf of 3-week-old rice seedlings grown hydroponi-
0.2
*
**
cally. Data are presented as means ± SD (n = 2
6 biological replicates). (C) Light-response curve of
1 0.1
net photosynthesis fitted by the Farquhar–von
Caemmerer–Berry (FvCB) model and generated at 0 0.0
30°C and 400 ppm CO2 concentration in the field in WT OE1 OE2 OE5 KO1 KO2 KO3 Initial Total
Hainan in 2019. Data are presented as means ± SD
(n = 3 biological replicates). *P < 0.05 for OE1, C 40 WT OE1 OE2 OE5 D 75 WT OE1 OE2 OE5
KO1 KO2 KO3 KO1 KO2 KO3
OE2, and OE5 compared with WT, Student’s t test. *
(µmol CO2 m-2 s-1)

(µmol CO2 m-2 s-1)
* 60
* *
Photosynthesis
*
Photosynthesis
30
(D) CO2-response curve of net photosynthesis * * *
* 45
(A-Ci curve) fitted by the FvCB model and generated 20 *
at 1200 mmol m−2 s−1 photosynthetic photon flux 30
*
density and 30°C. Data are presented as means ± SD 10
15
(n = 3 biological replicates). *P < 0.05 for OE1, 0 0
OE2, and OE5 compared with WT, Student’s t test. 500 1000 1500 2000 200 400 600 800 1000 1200 1400
-15
(E) Diurnal changes in the photosynthesis rate of flag -10 Light intensity (µmol photons m s ) -2 -1
CO2 (µmol mol-1)
leaves measured with a LICOR-6400 XT instrument
at the heading stage (from 8:00 a.m. to 4:00 p.m.) E 40 WT OE1 OE2 OE5 F 0.8
in the field in Hainan. **P < 0.01 for OE1, OE2, KO1 KO2 KO3
**
Stomatal conductance
**
(µmol CO2 m-2 s-1)

and OE5 compared with WT, *P < 0.05 for OE1 at 30
** 0.6 **
(mol H2O m-2 s-1)

Photosynthesis
8:00 a.m., all Student’s t test. (F) Stomatal
* ** ** **
conductance of flag leaves at the heading stage in 20 0.4
Hainan. Measurements were performed with a
LICOR-6400 XT instrument at 30°C in ambient air 10 0.2
between 10:00 a.m. and 11:00 a.m. Data in (E) and
(F) are presented as presented as means ± SD (n = 6 0 0.0
biological replicates). (G and H) Maximum rates of 6 8 10 12 14 16 18 WT OE1 OE2 OE5 KO1 KO2 KO3
carboxylation (Vcmax) (G) and electron transport Time (h)
(Jmax) (H) in WT, OsDREB1C-OE, and OsDREB1C-KO
G 100 H 250
plants grown in the field in Hainan. Values were *
generated from the A-Ci curve and fitted by the FvCB ** ** *
**
Vcmax (µmol m-2 s-1)
80 200
Jmax (µmol m-2 s-1)

model. Data are presented as means ± SD (n = 3
60 150
biological replicates). All measurements in (C) to (H)
were performed with flag leaves of field-grown 40 100
rice plants at the heading stage. *P < 0.05,
20 50
**P < 0.01 compared with WT, Student’s t test.
0 0
WT OE1 OE2 OE5 KO1 KO2 KO3 WT OE1 OE2 OE5 KO1 KO2 KO3
carboxylase-oxygenase (RuBisCO) activase (fig. stomata. This is in agreement with the higher compared with WT and OsDREB1C-KO plants
S7E). Accordingly, both RuBisCO content and maximum rate of RuBisCO carboxylation and (Fig. 3, A and B). Additionally, OsDREB1C over-
activity were increased in OsDREB1C-OE plants the higher maximum rate of electron transport expression increased both nitrogen uptake by
(Fig. 2, A and B). derived from modeling of AÐCi curves (Fig. 2, roots and transport activity from roots to shoots
Next, we evaluated the role of OsDREB1C G and H). Analysis of key products of photo- (Fig. 3, C and D), thus resulting in elevated
in regulating photosynthetic capacity by synthetic metabolism showed that OsDREB1C- nitrogen content, nitrogen use efficiency (NUE),
investigating rice plants grown in paddy fields. OE leaves accumulated higher amounts of and protein abundance in leaves of field-grown
Consistent with the above-described results, starch, sucrose, and fructose, thus potentially plants (Fig. 3, E and F, and fig. S10). Consis-
key photosynthetic parameters, including diur- explaining the improved grain filling (fig. S9). tent with these findings, mature field-grown
nal changes, light-response curves, and CO2- Together, these results suggest that the observed OsDREB1C-OE plants had increased total
response curves of net photosynthesis, were growth and yield increases in OsDREB1C-OE nitrogen contents in above-ground organs (Fig.
improved in OsDREB1C-OE plants and com- rice plants result, at least in part, from en- 3G and fig. S11A), along with improved photo-
promised in OsDREB1C-KO plants (Fig. 2, C to hancement of photosynthetic capacity. synthetic NUE in a range of nitrogen supply
E). In addition, OsDREB1C-OE plants displayed To investigate whether OsDREB1C also in- conditions and a higher intrinsic water use
higher stomatal conductance while maintaining fluences nitrogen utilization, we conducted a efficiency upon low-level nitrogen application
15
a similar intercellular CO2 concentration (Fig. 2F N feeding experiment and monitored nitrogen (fig. S12), indicating higher efficiency of car-
and fig. S8), suggesting that the higher carbox- uptake and transport activity in hydroponically bon gain at lower nitrogen cost. Analysis of
ylation rates supported by higher RuBisCO grown seedlings. After 3 hours of incubation in carbon and nitrogen distribution showed that
content and activity prevent the buildup of higher a 15N-nitrate solution, OsDREB1C-OE seedlings the OsDREB1C-OE plants accumulated more
intercellular CO2 levels despite the opened had higher 15N contents in shoots and roots carbon and nitrogen in the grains, but less in

A 3 B 20
nitrogen transport to aerial organs, and (iii)
** ** promote resource allocation from leaves and
N (µmol h-1 g-1 shoot DW)

**
N (µmol h-1 g-1 root DW)

** shoots to grains.
15
2 ** **
OsDREB1C promotes flowering and shortens
10 growth duration
1 OsDREB1C-OE plants also exhibited a pro-
5
nounced early flowering phenotype in our
15
15
field trials (Fig. 4, A and B). Early flowering
0 0
WT OE1 OE2 OE5 KO1 KO2 KO3 WT OE1 OE2 OE5 KO1 KO2 KO3 was accompanying with higher biomass accumu-
lation at the heading stage (Fig. 4C). OsDREB1C-
C 10 D 1.2 OE plants flowered 13 to 19 days earlier than
Root-to-shoot transport activity

** the WT, whereas OsDREB1C-KO plants flowered
N-nitrate uptake activity
* *
(µmol h-1 g-1 root DW)
8 * 3 to 7 days later under the long-day conditions in

0.9
*
the field in Beijing (Fig. 4B). Overexpression of
6
0.6 OsDREB1C also had an effect on leaf maturation
4 and duration of the growth period (Fig. 4D).
0.3 However, these differences were less evident
2
under the short photoperiod in Hainan (table S4).
15
0 0.0 To further explore the influence of OsDREB1C

WT OE1 OE2 OE5 KO1 KO2 KO3 WT OE1 OE2 OE5 KO1 KO2 KO3 on flowering, the expression of key genes in-
volved in the induction of flowering was analyzed,
E F 100 kg ha-1 200 kg ha-1 including the two FT-like genes, Hd3a (OsFTL2)
4.5 150 and RFT1 (OsFTL3) (19, 20), and the downstream
*
** * ** NUE (Grain yield per kg N)
transcription factor gene OsMADS14. Expression
* of all three genes was up-regulated in OsDREB1C-
*
N content (%)
3.0 100 OE plants and down-regulated in OsDREB1C-KO

plants (Fig. 4, E to G). By contrast, expression
* ** of Hd1 and Ehd1, two genes upstream of
** **
1.5 50 * florigen, was not affected in OsDREB1C-OE
** ** and OsDREB1C-KO plants (Fig. 4, H and I),
indicating that OsDREB1C affects flowering
0.0 0 time by controlling a specific subset of flowering
WT OE1 OE2 OE5 KO1 KO2 KO3
O 1
3
E2
E5
3
E1
KO1
E5
2
T
E2
KO1
OT
regulators.
E
KO
KO
KO
KO
W
W
O
O
O
O
G 1.2 Leaf Shoot Grain H 120 Leaf Shoot Grain OsDREB1C enhances yield in an elite cultivar
N distribution per plant (g)
To further test whether OsDREB1C overexpres-

N distribution ratio (%)
0.9
*
90 * * sion can increase the yield of elite rice varieties,
* we transformed the p35S::OsDREB1C construct
* ** into Xiushui 134 (XS134), a high-yielding elite
0.6 60
** temperate japonica cultivar that is widely cul-
*
0.3 * 30 tivated in southern China. Transgenic XS134
rice plants (OsDREB1C-XSOE) exhibited in-
* * * *
0.0 0 creased height, longer panicles, higher grain
WT OE1 OE2 OE5 KO1 KO2 KO3 WT OE1 OE2 OE5 KO1 KO2 KO3 numbers per panicle, and higher grain yields
over 2 consecutive years, similar to those seen
Fig. 3. OsDREB1C overexpression increases nitrogen uptake and transport. (A and B) 15N content in
in OsDREB1C-OE plants in the Nipponbare
shoots (A) and roots (B) of 3-week-old WT, OsDREB1C-OE, and OsDREB1C-KO seedlings incubated with
background (Fig. 5, A to G, and figs. S14 and
0.5 mM K15NO3 for 3 hours. (C) 15N-nitrate uptake activity of roots. (D) 15N transport activity from roots to
S15). The grain yield per plot was increased
shoots. Data in (A) to (D) are presented as means ± SD (n = 5 biological replicates). (E) Nitrogen content
by 10.3 to 12.7% in 2020 and 30.1 to 41.6% in
of flag leaves of WT, OsDREB1C-OE, and OsDREB1C-KO plants at the heading stage grown in the field in Beijing
2021 in Hangzhou (fig. S14G and Fig. 5H),
in 2021. Data are presented as means ± SD (n > 5 biological replicates). (F) NUE of WT, OsDREB1C-OE, and
accompanied by an increased harvest index
OsDREB1C-KO plants grown with 100 or 200 kg ha−1 nitrogen supply in the field in Beijing in 2021. Box plot
of up to 10.5 and 15.7%, respectively (fig. S14H
shows median (line) and individual values (black dots) (n > 5 biological replicates). *P < 0.05, **P < 0.01
and Fig. 5I). Moreover, OsDREB1C-XSOE lines
compared with WT, Student’s t test. (G and H) Nitrogen distribution (G) and nitrogen distribution ratio (H) in
flowered 2 days earlier than the WT in Hangzhou
seeds, straw, and leaves of mature plants grown in the field in Beijing in 2019. Data are presented as means ± SD
(fig. S14I). Accordingly, OsDREB1C-XSOE lines
(n = 4 biological replicates). *P < 0.05, **P < 0.01 compared with WT, Student’s t test.
displayed 26.2 to 42.4% higher grain yields per
plant in Hainan (fig. S15), whereas no difference
in flowering time was observed because of the
their mature leaves, without substantial alter- compared with WT plants (fig. S11), leading short-day growth conditions.
ations in the carbon-to-nitrogen ratio (Fig. 3H to a further boost in grain yield, especially Next, we analyzed the OsDREB1C sequences
and fig. S11). These findings indicate more ef- under conditions of low nitrogen supply (fig. in 709 rice accessions, including 299 indica,
ficient resource allocation from source (leaves) S13). In conclusion, OsDREB1C appears to (i) 355 temperate japonica, 14 tropical japonica,
to sink (grains) in the overexpression plants stimulate nitrogen uptake by roots, (ii) enhance and 41 intermediate varieties (21). Three distinct

A B 150 C 80
WT OE1 OE2 OE5 KO1 KO2 KO3
** ** ** ** **
**
** ** **
Flowering time (DAS)

60
100
Biomass (g)
40
50
20
0 0
E1
1
T
E2
3
E5
E1
1
T
2
E2
3
E5
KO
W
KO
KO
KO
KO
W
KO
O
O
D OE1 OE2 OE5
E F G H I
WT 20 Hd3a 5 RFT1 5 OsMADS14 2.0 Hd1 2.0 Ehd1
KO1 KO2 KO3

60

4 4 **
** ** 15 * **
1.5 1.5
SPAD value
45 ** 3 3
30
10 * 1.0 1.0
** 2 2
** **
15 ** 5
1 1 *
0.5 0.5
**
* **
0 0 0 0 0.0 0.0
110 120 130 140 150
E2
1
T
E5
2
E2
1
T
E5
E2
E2
1
E5
2
E2
1
1
T
T
E5
E5
T
KO
KO
KO
KO
W
KO
KO
W
KO
KO
KO
KO
W
W
O
O
O
O
O
O
O
DAS (d)
Fig. 4. OsDREB1C overexpression leads to early flowering and shortens OE2, and OE5 at four time points and for KO1, KO2, and KO3 at 112 and 129 DAS
the overall growth period. (A) Growth of WT, OsDREB1C-OE, and OsDREB1C- compared with WT, all Student’s t test. (E to I) Relative gene expression
KO plants in natural long-day conditions in Beijing in 2019. Scale bar, 50 cm. levels of the flowering regulators Hd3a (E), RFT1 (F), OsMADS14 (G), Hd1 (H),
(B) Flowering time of field-grown plants. DAS, days after sowing. (C) Biomass and Ehd1 (I) in WT, OsDREB1C-OE, and OsDREB1C-KO plants. RNAs were
of rice plants at the heading stage of OsDREB1C-OE plants (105 DAS) grown extracted from leaves of field-grown plants before heading (~90 DAS)
in the field in Beijing in 2021. (D) Soil plant analysis development (SPAD) in the Beijing field in 2019. Data in (B) to (I) are presented as means ± SD
value of flag leaves at different development stages (112, 129, 139, and [n = 3 biological replicates, except for n =10 for (C)]. *P < 0.05, **P < 0.01
152 DAS) of plants grown in the field in Beijing in 2019. **P < 0.01 for OE1, compared with WT, Student’s t test.
haplotypes were identified (Hap. 1 to Hap. 3) one-hybrid assays verified the direct binding strongly enriched and included the two nitrogen
on the basis of nucleotide polymorphisms, of OsDREB1C to DRE/CRT (GCCGAC) as well transporter genes, OsNRT2.4 and OsNRT1.1B
most of which reside in the promoter regions as the GCC box (GCCGCC) and G box (CACGTG) (Fig. 6D). Other genes with functions in the
(fig. S16). cis elements in vitro (fig. S20A). Measurement nitrogen metabolic process were also present
of the transcription-stimulating activity in rice in the DEG set, including the nitrate reductase
Identification of target genes of OsDREB1C protoplasts demonstrated that OsDREB1C was gene OsNR2. When searching for flowering-
We next wanted to characterize the molecular able to activate transcription of the GUS re- related DEGs that could potentially explain
functions of OsDREB1C in more detail. Multiple porter gene (fig. S20, B and C). Taken together, the pronounced early-flowering phenotype,
amino acid sequence alignment revealed a these data suggest that OsDREB1C functions the gene OsFTL1 (FT-Like 1) was found. OsFTL1
conserved AP2 domain among all OsDREB1C as a transcriptional activator. is a homolog of the Arabidopsis FT gene that
homologs (figs. S17 and S18). Expression analysis To identify genome-wide binding sites of plays a central role in integrating signals from
showed that OsDREB1C was expressed ubiqui- OsDREB1C in vivo, we performed chromatin the different flowering pathways (23, 24).
tously in all rice tissues examined (root, stem, immunoprecipitation sequencing (ChIP-seq) Moreover, a key photosynthesis-related gene,
leaf, and panicle), but particularly strongly in experiments with rice protoplasts transiently OsRBCS3, encoding the RuBisCO small sub-
the root (fig. S19A). During the growth period, expressing the OsDREB1C-GFP fusion protein. unit and known to be transcriptionally acti-
OsDREB1C transcript levels peaked at the til- These analyses identified a total of 9735 puta- vated by light (25), was also up-regulated in
lering stage (fig. S19B). Transient expres- tive OsDREB1C-binding sites, of which 68% OsDREB1C-OE plants.
sion assays in rice protoplasts revealed that localized to genic regions and 32% to intergenic
OsDREB1C-GFP and YFP-OsDREB1C fusion regions (Fig. 6A). The core motif found to be OsDREB1C directly activates key pathway genes
proteins mainly localize to the nucleus, but a enriched in the OsDREB1C-binding regions To verify whether these five candidate genes
substantially weaker signal in the cytoplasm was DRE/CRT (GCCGAC) (Fig. 6B). We next (OsRBCS3, OsNR2, OsNRT2.4, OsNRT1.1B, and
was also discernable (fig. S19C). Very similar analyzed the differentially expressed genes OsFTL1) were targets of OsDREB1C, we per-
patterns of OsDREB1C-GFP subcellular localiza- (DEGs) between OsDREB1C-OE and WT plants formed ChIP–quantitative polymerase chain
tion were observed in Arabidopsis and Nicotiana (as determined by RNA-seq), and extracted reaction (qPCR) experiments using transgenic
benthamiana (fig. S19, D and E). the DEGs associated with OsDREB1C-binding plants expressing an OsDREB1C-GFP fusion
Sequence analysis with PlantPan3.0 sug- peaks. In this way, 345 up-regulated genes protein and DNA affinity purification sequenc-
gested that the OsDREB1C protein may bind were identified as putative OsDREB1C targets ing (DAP-seq) assays in vitro. The results re-
to the DRE/CRT (GCCGAC) motif that had (Fig. 6C). Gene ontology (GO) enrichment anal- vealed that OsDREB1C directly binds to the
been identified as a core cis-acting element ysis was then conducted to associate biological promoter of OsRBCS3 and to exons of OsNR2,
regulating gene expression in response to processes with those DEGs. Transmembrane OsNRT2.4, OsNRT1.1B, and OsFTL1 (Fig. 6E
drought, salt, and cold stresses (22). Yeast transport–related genes were found to be most and fig. S21A). Electrophoretic mobility shift

A B 100 C 30
** **
**
90
**
Plant height (cm)
Panicle number
20
80
70
10
60
50 0
1 34 E -8 E -9 -1 2
13
4 -8 -9 -12
XS134 XSOE-8 XSOE-9 XSOE-12
XS SO SO OE XS OE OE OE
X X XS XS XS XS
D 250 E 100 F 30
** **
** ** **
200 80
Seed setting rate (%)
1000-grain weight (g)

20
150 60
100 40
10
50 20
0 0 0
4 -8 -9 2 4 4 -8 -9
S1
3
OE OE E-1 S1
3
OE
-8
OE
-9
E-1
2 13 OE OE -12
X XS XS O X XS XS O XS XS XS OE
XS XS XS
G 50 H 1000 I 0.8
** ** ** ** **
40 800 **
Grain yield per plot (g)
0.6
**
Straw weight (g)
Harvest index
30 600
0.4
20 400
0.2
10 200
0 0 0.0
4 -8 34 -8 -9 -1 2 4 -8 -9 -12
S1
3
OE E-9 E-1
2
XS
1 OE OE OE 13 OE OE OE
X XS
O O XS XS XS XS XS XS
XS XS X S
Fig. 5. OsDREB1C confers yield gains in an elite rice germplasm. (A) Whole-plant phenotypes of mature Xiushui134 (XS134) and XS134-OsDREB1C-OE plants
(XSOE-8/9/12) grown in Hangzhou in 2020. Scale bar, 20 cm. (B to I) Yield parameters of XS134 and XS-OE plants grown in Hangzhou in 2021, including plant
height (B), panicle number (C), grain number per panicle (D), seed setting rate (E), 1000-grain weight (F), straw weight (G), grain yield per plot (H), and harvest
index (I). The box plots in panels (B), (C), and (G) show the median (horizontal line) and individual values (black dots) (n = 100 biological replicates). Data in
F(D) to (I) except (G) are presented as means ± SD (n = 6 plots). *P < 0.05, **P < 0.01 compared with XS134, StudentÕs t test.
assay (EMSA) confirmed that the DRE/CRT OE plants and decreased in OsDREB1C-KO time of OsFTL1-OE plants was drastically
elements are necessary for OsDREB1C binding plants (fig. S21B). Taken together, these results shortened, ranging from 45 to 47 days, whereas
(Fig. 6F and fig. S22). Moreover, luciferase- suggest that OsDREB1C can activate gene ex- that of WT plants was 116 to 118 days under the
based transient transactivation assays verified pression directly by binding to the promoter long-day photoperiod in Beijing (fig. S23, B and
that OsDREB1C activates the expression of of OsRBCS3 and to the exons of OsNR2, D). Under the short-day photoperiod in Hainan,
OsRBCS3, OsNR2, OsNRT2.4, OsNRT1.1B, and OsNRT2.4, OsNRT1.1B, and OsFTL1. OsFTL1-OE plants flowered 10 to 13 days earlier
OsFTL1 (Fig. 6G). Gene expression analyses To clarify whether OsFTL1 induction is re- (fig. S23C). These results are consistent with a
revealed that the mRNA levels of OsRBCS3, sponsible for the early-flowering phenotype of previous study on floral induction in transgenic
OsNR2, OsNRT2.4, OsNRT1.1B, and OsFTL1 OsDREB1C-OE plants, we generated OsFTL1 plants grown in culture vessels (26). OsFTL1-OE
correlated with OsDREB1C levels, in that the overexpression lines in the rice cultivar plants were dwarfed and exhibited reduced
expression levels were increased in OsDREB1C- Nipponbare (fig. S23, A and E). The heading grain yields (fig. S23F), probably because of their

Fig. 6. OsDREB1C induces A TTS

(12.24%; 1627)
E 30 OsRBCS3
1 kb
Input
transcription of photosynthesis,
30
nitrogen utilization, and flowering- Exon
(22.83%; 3034) ChIP
related genes. (A) Distribution of 90
Promoter-TSS
candidate OsDREB1C-binding regions (26.47%; 3518) DAP
Intergenic
across the rice genome as determined (31.98%; 4251) P1 P2 P3
by ChIP-seq. TSS, transcription start 80 OsNR2

1 kb
Input
site; TTS, transcription termination Intron
(6.49%; 862)
80
site. (B) Motif analysis using HOMER ChIP
to identify core motifs enriched within B p = 1e-482
GCCGACA
600
the experimentally determined (by AC C DAP
CT
TG
GT
A
ChIP-seq) OsDREB1C-binding regions. GAAT C
TG
A
G
AT
A
G
C
T G T A
TC
C
TG
P1 P2 P3
(C) Venn diagram showing the overlap Positive
GTCGGCGT
OsNRT2.4
between putative OsDREB1C target G
AC
T Input
30 1 kb
TG
ATCC
AG
A
genes identified by ChIP-seq and CA C T
A
C
GAT
CC
T
A A
G
T
A
G T
30
Reverse ChIP
differentially up-regulated genes in
OsDREB1C-OE1 relative to the WT as C OsDREB1C target genes
200
DAP
p = 1.361e-15
identified by RNA-seq. (D) GO enrich-
ment analysis of the overlapping gene P1 P2 P3
9390 60 OsNRT1.1B 1 kb
set in (B). (E) OsDREB1C preferentially Input
345
binds to the OsRBCS3 promoter and 512 60
to exons of OsNR2, OsNRT2.4, ChIP-Seq ChIP
RNA-Seq
OsNRT1.1B, and OsFTL1, as validated 600
DAP
by both ChIP-qPCR and DAP-seq. The D -log10(FDR) 0 5 10
diagrams in (E) depict the putative transmembrane transport P1 P2 P3
OsFTL1
promoter region of OsRBCS3 and nitrogen compound metabolic process Input
80 1 kb
exons of the OsNR2, OsNRT2.4,

primary metabolic process 80
OsNRT1.1B, and OsFTL1 genes. P1, P2, ChIP
response to stress
and P3 indicate primers used in the 60
regulation of flower development
ChIP-qPCR experiments for the five DAP
response to light stimulus
examined loci shown in fig. S21. There P1 P2 P3
are DRE/CRT elements within P3 of

OsRBCS3, P1 of OsNR2 and OsNRT1.1B, F OsRBCS3 OsNR2 OsNRT2.4 OsNRT1.1B OsFTL1
and P2 of OsNRT2.4 and OsFTL1. GST - +

-
- - - - +
-
- - - - +
-
- - - - +
-
- - - - + - - -
GST-OsDREB1C + + + + + + + + + + + + + + + + + - + + +
(F) EMSA data confirming that the Biotin-labeled probe - + + + + - + + + + - + + + + - + + + + - + + + +
- - - 50× 500× - - - 50× 500× - - - 50× 500× - - - 50× 500× - - - 50× 500×
GST-OsDREB1C protein binds to Unlabeled probe
promoters and exons containing Shifted band

DRE/CRT elements. (G) OsDREB1C
activates transcription from the
OsRBCS3 promoter and from OsNR2,
OsNRT1.1B, OsNRT2.4, and OsFTL1 Free probe
exon-luciferase fusion constructs in
transient transactivation assays. G 35S REN OsRBCS3 promoter LUC 35S REN 35S mini OsNR2/OsNRT2.4/OsNRT1.1B/OsFTL1 Exon LUC
Shown are relative ratios of the
transcriptional activities conferred by 35 OsRBCS3 10 OsNR2 16 OsNRT2.4 10 OsNRT1.1B 10 OsFTL1
OsDREB1C expression to the empty 28 ** 8 * ** 8 * 8
vector control. Asterisks indicate
12 **
LUC/REN
LUC/REN
LUC/REN
LUC/REN
LUC/REN
21 6 6 6
significant differences between the 8
14 4 4 4
control and OsDREB1C. LUC/REN, ratio
4
of firefly luciferase to Renilla luciferase 7 2 2 2
activity. Data are presented as 0

ol
0
ol
0
ol
0
ol
0
ol
ntr B1
C
ntr 1C ntr 1C ntr 1C ntr 1C
means ± SD (n ≥ 3 biological replicates). Co E Co EB Co EB Co EB Co EB
DR DR DR DR DR
Os O s O s Os Os
*P < 0.05, **P < 0.01 compared with
WT, StudentÕs t test.
shortened vegetative phase (and the insufficient lines showed that OsDREB1C overexpression in bigger leaves, and flowered up to 4 days earlier
buildup of resources for allocation to seeds). the wheat cultivar Fielder also improved photo- than the WT (Fig. 7, G to J). The biomass yield was
synthetic capacity, reduced the time to flowering increased by 14.2 to 35.8% in the OsDREB1C-OE
OsDREB1C effects in wheat and Arabidopsis by 3 to 6 days, and conferred increased grain Arabidopsis plants (Fig. 7, K and L).
To determine whether the function of DREB1C yield per plant by 17.2 to 22.6% in the field and
is conserved in other plant species, we generated by 18.6 to 23.5% in the greenhouse (Fig. 7, A to F, DISCUSSION
transgenic wheat and Arabidopsis plants over- and fig. S24). Likewise, transgenic Arabidopsis Transcription factors of the DREB subfamily
expressing OsDREB1C. Analysis of the transgenic lines overexpressing OsDREB1C had more and belong to the AP2/ERF family and have been

A B C D E F
Photosynthesis (µmol CO2 m-2 s-1)

70 30 80 50 18
** ** * *

* **
Grain yield per plant (g)

** ** *
1000-grain weight (g)

65 60
40 *
** ** 20 12
30
60 ** 40
20
10 6
55 20
10
50 0 0 0 0
r -5 -8 -9
lde OE OE OE lde
r -5 -8 -9 lde
r -5 -8 -9 r
lde OE-
5 -8 -9 lde
r -5 -8 -9 lde
r -5 -8 -9
Fie Ta Ta Ta Fie OE OE OE Fie OE OE OE Fie OE OE Fie OE aOE OE Fie OE aOE OE
Ta Ta Ta Ta Ta Ta Ta Ta Ta Ta T Ta Ta T Ta
G H 60 I 50
J 30
K 2.0
L 0.20
** **
**

** ** ** **
Rosette leaf number

40 ** ** ** 0.15
Fresh weight (g)

1.5
Dry weight (g)

40 ** 20 * ** *
30
** 1.0 0.10
20
20 10
0.5 0.05
10
0 0 0 0.0 0.00
l-0 E-10 1 2 l-0 E-10 1 2 l-0 E-10 1 2 l-0 E-10 1 12 l-0 E-10 1 2
Co E-1 OE-1 Co E-1 OE-1 Co E-1 OE-1 Co E-1 OE- Co E-1 OE-1
l-0 0 1 2 AtO AtO At AtO AtO At AtO AtO At AtO AtO At AtO AtO At
Co E-1 E-1 E-1
AtO AtO AtO
Fig. 7. OsDREB1C increases flowering, photosynthetic capacity, and and OsDREB1C-OE Arabidopsis (p35S::OsDREB1C-GFP, AtOE-10/11/12) plants
yields in wheat and Arabidopsis. (A) Early flowering phenotype of OsDREB1C- at the seedling and flowering stages. Note the early flowering of all three
OE field-grown wheat plants (pUBI::OsDREB1C, TaOE-5/8/9) compared with overexpression lines. Plants were grown in short-day conditions (8 hours
WT plants (cv. Fielder) at the booting stages. Scale bar, 10 cm. (B and C) Flowering light/16 hours dark) in a growth chamber for 2 weeks and then transferred to
time (B) and photosynthesis rate (C) of Fielder and OsDREB1C-OE wheat long-day conditions (16 hours light/8 hours dark). Scale bar, 4 cm.
plants at the heading stage grown in the field in Beijing in 2021. Data are (H) OsDREB1C expression levels in OsDREB1C-OE Arabidopsis plants. Data
presented as means ± SD (n > 5 biological replicates). (D to F) Grain number are presented as means ± SD (n = 3 biological replicates). (I to L) Flowering
per panicle (D), 1000-seed weight (E), and grain yield per plant (F) of Fielder time (I), rosette leaf number (J), fresh weight (K), and dry weight (L)
and OsDREB1C-OE wheat plants grown in the field in Beijing in 2021. Data of WT (Col-0) and OsDREB1C-OE Arabidopsis plants. Data are presented as
are presented as means ± SD (n > 30 biological replicates). *P < 0.05, **P < means ± SD (n = 10 biological replicates). *P < 0.05, **P < 0.01 compared
0.01 compared with Fielder, Student’s t test. (G) Phenotype of WT (Col-0) with Col-0, Student’s t test.
demonstrated to activate multiple downstream requirement for a high level of nitrogen fertilizer elevated yield, presumably through coordinated
genes in response to abiotic stresses such as to attain high yields. This is mainly because of regulation of several target genes of OsDREB1C,
drought, salinity, and freezing in various seed the low NUE of most major crop varieties (34). including amino acid and ammonium trans-
plants (27, 28). The DREB-binding DRE/CRT In addition to its negative environmental im- porters (figs. S26 and S27). OsDREB1C binds to
cis-element (GCCGAC motif) is present in the pact, excessive application of nitrogen fertilizer the exon regions rather than the promoters of
promoter of many stress-inducible genes (29). has other undesired effects, including delayed four of the five verified target genes. However,
However, the previous work has also revealed flowering, extended growth duration, and re- the preferential binding of transcription factors
a trade-off between growth and stress tolerance, duced yield potential (35). In this study, we to intragenic regions (exons and/or introns) of
in that constitutive overexpression of DREB1 have shown that by engineering the expres- target genes has been demonstrated in a num-
genes in Arabidopsis and rice, although con- sion of a transcription factor that controls and ber of previous studies (10, 36–38).
ferring improved stress tolerance, often leads coordinates photosynthesis, nitrogen utiliza- Currently, the relative rates of yield increase
to growth retardation and yield penalties tion, and flowering time without affecting achieved by plant breeding are declining and
(27, 29–31). In the course of this work, we have known genes involved in high yield and early have fallen below 1% per year for most cereal
shown that the inverse relationship between flowering (fig. S25), it is possible to achieve crops (39). In view of this trend and the need
stress tolerance and yield can be uncoupled enhanced growth and increased yields while to double the world’s food production by 2050
and even reversed. Overexpression of OsDREB1C at the same time improving the efficiency of despite reduced availability of arable land and
in rice plants resulted in substantial yield in- nitrogen utilization. Thus, our work demon- the challenges of climate change, the very large
creases of 41.3 to 68.3%, and these yield gains strates that three key agricultural traits can be yield increases achieved in the field by engineer-
were accompanied by a shortened vegetative improved simultaneously by OsDREB1C over- ing the expression of a single transcriptional
phase, in that the overexpression plants flowered expression: yield, NUE, and flowering time. regulator gene are unprecedented. Our findings
much earlier than the WT. Although “high- We provide several lines of physiological and suggest that after centuries of breeding for yield,
yielding” and “early-maturing” have long been molecular evidence in support of the assump- there is still potential for substantial leaps in the
seen as conflicting traits in crop breeding, a tion that OsDREB1C confers accelerated vegeta- yields of the world’s main staple crops.
report has shown that the long noncoding tive growth and biomass accumulation before In the present study, overexpression of
RNA Ef-cd shortens maturity duration with- heading by (i) enhancing photosynthetic capac- OsDREB1C was achieved using transgenic
out incurring a yield penalty (32). Similarly, ity through OsRBCS3; (ii) enhancing nitrogen technologies. Alternatively, genome-editing tech-
overexpression of the nitrate transporter gene uptake and transport through expression of nologies could be used to achieve OsDREB1C
OsNRT1.1A confers both high yield and early OsNRT1.1B, OsNRT2.4, and OsNR2; and (iii) overexpression, for example, by using base editors
maturation (33). promoting early flowering through OsFTL1. to introduce expression-enhancing point muta-
An unsolved problem pertinent to both We propose that efficient subsequent allocation tions into the promoter region of the OsDREB1C
agricultural productivity and the environmental of assimilated carbohydrates and nitrogen from gene (40, 41), thus creating transgene-free, high-
footprint of current agricultural practices is the leaves to the panicle further contributes to the yielding varieties. Also, the existing natural

variation of OsDREB1C in rice provides a ge- transgenic Arabidopsis plants expressing 13. Y. Jiao et al., Regulation of OsSPL14 by OsmiR156 defines ideal
netic resource that can be readily tapped. Fi- OsDREB1C-GFP. The binding ability of OsDREB1C plant architecture in rice. Nat. Genet. 42, 541–544 (2010).
doi: 10.1038/ng.591; pmid: 20495565
nally, the conserved function of OsDREB1C in to the predicted core promoter motifs was 14. J. Wang et al., A single transcription factor promotes both yield
seed plants offers the potential to substan- tested with yeast one-hybrid assays. The trans- and immunity in rice. Science 361, 1026–1028 (2018).
tially improve biomass and yield in other crops. activation capacity of OsDREB1C was eval- doi: 10.1126/science.aat7675; pmid: 30190406
15. M. M. Ambavaram et al., Coordinated regulation of
In summary, our data suggest OsDREB1C en- uated by transient transactivation assays. photosynthesis in rice increases yield and tolerance to
gineering as a generally applicable strategy to To identify the genome-wide binding sites of environmental stress. Nat. Commun. 5, 5302 (2014).
increase crop yields, with the added benefits of OsDREB1C, ChIP-seq experiments were per- doi: 10.1038/ncomms6302; pmid: 25358745
16. S. J. H. Kuijt et al., Interaction between the GROWTH-
shortening the growth duration and lowering formed with rice protoplasts transiently ex- REGULATING FACTOR and KNOTTED1-LIKE HOMEOBOX
the environmental footprint of agriculture. pressing the OsDREB1C-GFP fusion protein, and families of transcription factors. Plant Physiol. 164, 1952–1966
RNA-seq analysis with WT and OsDREB1C-OE (2014). doi: 10.1104/pp.113.222836; pmid: 24532604
Methods summary plants grown in the field. Up-regulated genes 17. L. Wang et al., Comparative analyses of C4 and C3
photosynthesis in developing leaves of maize and rice.
The candidate gene OsDREB1C was identified in OsDREB1C-OE plants associated with Nat. Biotechnol. 32, 1158–1165 (2014). doi: 10.1038/nbt.3019;
by a screen of 118 transcription factors that OsDREB1C-binding peaks were extracted as pmid: 25306245
are related to C4 photosynthesis, induced by putative OsDREB1C targets, and GO enrich- 18. W. Xin et al., An integrated analysis of the rice transcriptome
and metabolome reveals root growth regulation mechanisms in
light, and highly expressed under low-nitrogen ment analysis was conducted to identify the response to nitrogen availability. Int. J. Mol. Sci. 20, 5893
conditions. The function of OsDREB1C was associated biological processes. ChIP-qPCR using (2019). doi: 10.3390/ijms20235893; pmid: 31771277
investigated by transgenic overexpression and transgenic plants expressing an OsDREB1C-GFP 19. S. Kojima et al., Hd3a, a rice ortholog of the Arabidopsis FT
gene, promotes transition to flowering downstream of Hd1
CRISPR-Cas9Ðbased gene-editing approaches. fusion protein and DAP-seq using expressed under short-day conditions. Plant Cell Physiol. 43, 1096–1105
Agrobacterium-mediated transformation was HALOTag-OsDREB1C fusion proteins in vitro (2002). doi: 10.1093/pcp/pcf156; pmid: 12407188
used to introduce the constructed vectors into were performed to verify the binding of 20. R. Komiya, A. Ikegami, S. Tamaki, S. Yokoi, K. Shimamoto,
Hd3a and RFT1 are essential for flowering in rice. Development
rice (Nipponbare and Xiushui134) and wheat OsDREB1C to target genes and test for target 135, 767–774 (2008). doi: 10.1242/dev.008631;
(Fielder), and the floral dip method was used gene activation. Direct binding of OsDREB1C pmid: 18223202
to transform Arabidopsis (Col-0). Gene expres- to target sequences was examined by EMSA. 21. X. Li et al., Analysis of genetic architecture and favorable allele
usage of agronomic traits in a large collection of Chinese rice
sion levels in the overexpression and KO lines The transcriptional activation of targets genes accessions. Sci. China Life Sci. 63, 1688–1702 (2020).
were evaluated by quantitative reverse tran- by OsDREB1C was tested by dual luciferase doi: 10.1007/s11427-019-1682-6; pmid: 32303966
scription PCR (qRT-PCR). reporter assays and qRT-PCR. 22. Y. Sakuma et al., Functional analysis of an Arabidopsis
transcription factor, DREB2A, involved in drought-responsive
The growth rates of hydroponically grown Details for experimental procedures are
gene expression. Plant Cell 18, 1292–1309 (2006).
rice seedlings were monitored under standard provided in the supplementary materials. doi: 10.1105/tpc.105.035881; pmid: 16617101
growth conditions. Etiolated growth was as- 23. I. Kardailsky et al., Activation tagging of the floral inducer FT.
sessed in constant darkness. Photosynthesis Science 286, 1962–1965 (1999). doi: 10.1126/
RE FERENCES AND NOTES
science.286.5446.1962; pmid: 10583961
rates were measured with a LICOR-6400XT 1. Food and Agriculture Organization of the United Nations, 24. M. J. Yanovsky, S. A. Kay, Molecular basis of seasonal time
gas exchange system in field-grown plants. “Hunger and food insecurity” (FAO, 2020); https://www.fao. measurement in Arabidopsis. Nature 419, 308–312 (2002).
org/hunger/en/. doi: 10.1038/nature00996; pmid: 12239570
Accumulation levels of photosynthesis-related 2. D. Tilman, C. Balzer, J. Hill, B. L. Befort, Global food demand 25. M. A. Parry, A. J. Keys, P. J. Madgwick, A. E. Carmo-Silva,
proteins were assessed by immunoblotting. and the sustainable intensification of agriculture. Proc. Natl. P. J. Andralojc, Rubisco regulation: A role for inhibitors.
RuBisCO content was quantified by immuno- Acad. Sci. U.S.A. 108, 20260–20264 (2011). doi: 10.1073/ J. Exp. Bot. 59, 1569–1580 (2008). doi: 10.1093/jxb/ern084;
pnas.1116437108; pmid: 22106295 pmid: 18436543
blot assays, and RuBisCO activity was deter- 3. Food and Agriculture Organization of the United Nations, 26. T. Izawa et al., Phytochrome mediates the external light signal
mined as the rate of CO2 fixation on RuBP “The future of food and agriculture: Trends and challenges” to repress FT orthologs in photoperiodic flowering of rice.
using spectrophotometry. Sugar and starch con- (FAO, 2017); https://www.fao.org/3/i6583e/i6583e.pdf. Genes Dev. 16, 2006–2020 (2002). doi: 10.1101/gad.999202;
4. J. Bailey-Serres, J. E. Parker, E. A. Ainsworth, G. E. D. Oldroyd, pmid: 12154129
tents were measured using enzymatic diges- J. I. Schroeder, Genetic strategies for improving crop yields. 27. Y. Ito et al., Functional analysis of rice DREB1/CBF-type
tion methods. Nitrogen uptake and transport Nature 575, 109–118 (2019). doi: 10.1038/s41586-019-1679-0; transcription factors involved in cold-responsive gene
activity were evaluated with the 15N-labeling pmid: 31695205 expression in transgenic rice. Plant Cell Physiol. 47, 141–153
5. G. M. Coruzzi, L. Zhou, Carbon and nitrogen sensing and
assay. Carbon and nitrogen contents were (2006). doi: 10.1093/pcp/pci230; pmid: 16284406
signaling in plants: Emerging ‘matrix effects’. Curr. Opin. Plant 28. Q. Liu et al., Two transcription factors, DREB1 and DREB2, with
assessed with an IsoPrime 100 analyzer. NUE Biol. 4, 247–253 (2001). doi: 10.1016/S1369-5266(00)00168-0; an EREBP/AP2 DNA binding domain separate two cellular
was calculated as the ratio of grain yield to pmid: 11312136 signal transduction pathways in drought- and low-temperature-
applied nitrogen fertilizer. Phytohormone 6. A. Nunes-Nesi, A. R. Fernie, M. Stitt, Metabolic and signaling responsive gene expression, respectively, in Arabidopsis.
aspects underpinning the regulation of plant carbon nitrogen Plant Cell 10, 1391–1406 (1998). doi: 10.1105/tpc.10.8.1391;
contents were quantified using ultraperfor- interactions. Mol. Plant 3, 973–996 (2010). doi: 10.1093/mp/ pmid: 9707537
mance liquid chromatography quadropule ion ssq049; pmid: 20926550 29. J. G. Dubouzet et al., OsDREB genes in rice, Oryza sativa L.,
trap mass spectrometry. Agronomic and yield 7. S. Li et al., Modulating plant growth-metabolism coordination encode transcription activators that function in drought-,
for sustainable agriculture. Nature 560, 595–600 (2018). high-salt- and cold-responsive gene expression. Plant J. 33,
traits were assessed in field experiments with doi: 10.1038/s41586-018-0415-5; pmid: 30111841 751–763 (2003). doi: 10.1046/j.1365-313X.2003.01661.x;
randomized block design and performing 8. K. Wu et al., Enhanced sustainable green revolution yield via pmid: 12609047
three replicates for each experiment in Beijing, nitrogen-responsive chromatin modulation in rice. Science 367, 30. M. Kudo et al., A gene-stacking approach to overcome the
eaaz2046 (2020). doi: 10.1126/science.aaz2046; trade-off between drought stress tolerance and growth in
Hangzhou, and Hainan in four successive years pmid: 32029600 Arabidopsis. Plant J. 97, 240–256 (2019). doi: 10.1111/
(May 2018 to May 2022). 9. K. Kaufmann, C. A. Airoldi, Master regulatory transcription tpj.14110; pmid: 30285298
A sequence alignment of OsDREB1C ortho- factors in plant development: A blooming perspective. 31. Y. Shi, Y. Ding, S. Yang, Molecular regulation of CBF signaling
Methods Mol. Biol. 1830, 3–22 (2018). doi: 10.1007/978-1- in cold acclimation. Trends Plant Sci. 23, 623–637 (2018).
logs was constructed and phylogenetic anal- 4939-8657-6_1; pmid: 30043361 doi: 10.1016/j.tplants.2018.04.002; pmid: 29735429
ysis was performed using ESPript 3.0 and 10. S. J. Burgess et al., Genome-wide transcription factor binding 32. J. Fang et al., Ef-cd locus shortens rice maturity duration
MEGA 7, respectively. Genetic variation and in leaves from C3 and C4 grasses. Plant Cell 31, 2297–2314 without yield penalty. Proc. Natl. Acad. Sci. U.S.A. 116, 18717–18722
(2019). doi: 10.1105/tpc.19.00078; pmid: 31427470 (2019). doi: 10.1073/pnas.1815030116; pmid: 31451662
haplotype association analyses were performed 33. W. Wang et al., Expression of the nitrate transporter gene
11. P. Cubas, N. Lauter, J. Doebley, E. Coen, The TCP domain: A
by variant cell format tools using a general motif found in proteins regulating plant growth and OsNRT1.1A/OsNPF6.3 confers high yield and early maturation
linear model implemented in TASSEL 5.0. development. Plant J. 18, 215–222 (1999). doi: 10.1046/ in rice. Plant Cell 30, 638–651 (2018). doi: 10.1105/
j.1365-313X.1999.00444.x; pmid: 10363373 tpc.17.00809; pmid: 29475937
Subcellular localization of OsDREB1C was as-
12. J. Doebley, A. Stec, L. Hubbard, The evolution of apical 34. S. M. Swarbreck et al., A roadmap for lowering crop nitrogen
sessed by transient expression assays in rice dominance in maize. Nature 386, 485–488 (1997). requirement. Trends Plant Sci. 24, 892–904 (2019).
protoplasts and tobacco leaves and in stable doi: 10.1038/386485a0; pmid: 9087405 doi: 10.1016/j.tplants.2019.06.006; pmid: 31285127

35. H. Li, B. Hu, C. Chu, Nitrogen use efficiency in crops: Lessons ACKN OWLED GMEN TS data generated during this study have been deposited in the
from Arabidopsis and rice. J. Exp. Bot. 68, 2477–2488 (2017). We thank X. Li (Institute of Crop Sciences, CAAS) for help with National Center for Biotechnology Information (NCBI) BioProject
doi: 10.1093/jxb/erx101; pmid: 28419301 drawing the working model and X. Ye and K. Wang (Institute of database under accession number PRJNA724935 for RNA-seq,
36. Z. Dong et al., Ideal crop plant architecture is mediated by Crop Sciences, CAAS) for help with wheat transformation. PRJNA841272 for ChIP-seq, and PRJNA841281 for DAP-seq. All other
tassels replace upper ears1, a BTB/POZ ankyrin repeat gene Funding: This research was supported by the National Key data are available in the main text or the supplementary materials.
directly targeted by TEOSINTE BRANCHED1. Proc. Natl. Acad. Research and Development Program of China (grants Requests for materials should be addressed to W.Z. License
Sci. U.S.A. 114, E8656–E8664 (2017). doi: 10.1073/ 2016YFD0300100 and 2016YFD0300102). W.Z. was supported by information: Copyright © 2022 the authors, some rights
pnas.1714960114; pmid: 28973898 the Innovation Program of the Chinese Academy of Agricultural reserved; exclusive licensee American Association for the
37. E. González-Grandío et al., Abscisic acid signaling is controlled Sciences and the Elite Youth Program of the Chinese Academy of Advancement of Science. No claim to original US government works.
by a BRANCHED1/HD-ZIP I cascade in Arabidopsis axillary Agricultural Science. Author contributions: W.Z., S.W., and X.L. https://www.science.org/about/science-licenses-journal-article-reuse
buds. Proc. Natl. Acad. Sci. U.S.A. 114, E245–E254 (2017). conceived and designed the experiments. S.W. and X.L. performed
doi: 10.1073/pnas.1613199114; pmid: 28028241 most of the experiments. H.Z., X.L., and S.W. produced the SUPPLEMENTARY MATERIALS
38. X. Yang et al., Regulation of plant architecture by a new histone transgenic plants. X.Y., Y.Z., J.L., Y.Y., and F.D. characterized the
science.org/doi/10.1126/science.abi8455
acetyltransferase targeting gene bodies. Nat. Plants 6, 809–822 phenotypes of transgenic plants. D.W. and S.C. performed field
(2020). doi: 10.1038/s41477-020-0715-2; pmid: 32665652 experiments with Xiushui134 rice transgenic plants. Z.L., H.P., and
Figs. S1 to S27
39. R. A. T. Fischer, G. O. Edmeades, Breeding and cereal yield progress. S.W. analyzed DAP-seq and RNA-seq results. Y.Z. and P.Y.
Tables S1 to S10
Crop Sci. 50, S-85–S-98 (2010). doi: 10.2135/cropsci2009.10.0564 performed the ChIP-qPCR experiment. S.W. and Z.L. analyzed the
40. A. V. Anzalone et al., Search-and-replace genome editing without ChIP-seq results. L.S. and C.Z. performed haplotype analysis.
MDAR Reproducibility Checklist
double-strand breaks or donor DNA. Nature 576, 149–157 (2019). S.W., X.L., and W.Z. wrote the manuscript. R.B., J.H., W.P., Q.Q.,
doi: 10.1038/s41586-019-1711-4; pmid: 31634902 and M.Z. critically commented on and edited the manuscript. All
41. J. Tan, F. Zhang, D. Karcher, R. Bock, Expanding the genome- authors discussed and commented on the manuscript. Competing
targeting scope and the site selectivity of high-precision base interests: The authors declare no competing interests. Patent Submitted 10 May 2021; resubmitted 29 March 2022
editors. Nat. Commun. 11, 629 (2020). doi: 10.1038/s41467- applications related to this work have been submitted by W.Z., Accepted 26 May 2022
020-14465-z; pmid: 32005820 S.W., and X.L. Data and materials availability: Raw sequence 10.1126/science.abi8455

RESEAR CH
◥ Partially constrained hallucination using a

RESEARCH ARTICLES multiobjective loss function
To extend existing trRosetta-based design
PROTEIN DESIGN methods to scaffold functional sites (Fig. 1A),
we optimized amino acid sequences for fold-
Scaffolding protein functional sites using ing to a structure containing the desired func-
tional site using a composite loss function that
deep learning combines the previously used hallucination loss
with a motif reconstruction loss over the func-
Jue Wang1,2†, Sidney Lisanza1,2,3†, David Juergens1,2,4†, Doug Tischer1,2†, Joseph L. Watson1,2†, tional motif [rather than the entire structure,
Karla M. Castro5, Robert Ragotte1,2, Amijai Saragovi1,2, Lukas F. Milles1,2, Minkyung Baek1,2, as in (13)] (Fig. 1B; see materials and methods
Ivan Anishchenko1,2, Wei Yang1,2, Derrick R. Hicks1,2, Marc Expòsit1,2,4, Thomas Schlichthaerle1,2, in the supplementary materials). Although
Jung-Ho Chun1,2,3, Justas Dauparas1,2, Nathaniel Bennett1,2,4, Basile I. M. Wicky1,2, Andrew Muenks1,2, we succeeded in generating structures with
Frank DiMaio1,2, Bruno Correia5, Sergey Ovchinnikov6,7*, David Baker1,2,8* segments closely recapitulating functional sites,
Rosetta structure predictions suggested that
The binding and catalytic functions of proteins are generally mediated by a small number of the sequences poorly encoded the structures
functional residues held in place by the overall protein structure. Here, we describe deep learning (fig. S1A), and hence we used Rosetta design
approaches for scaffolding such functional sites without needing to prespecify the fold or secondary calculations to generate more-optimal se-
structure of the scaffold. The first approach, “constrained hallucination,” optimizes sequences quences (14). Several designs targeting pro-
such that their predicted structures contain the desired functional site. The second approach, grammed cell death ligand 1 (PD-L1) generated
“inpainting,” starts from the functional site and fills in additional sequence and structure to create a by constrained hallucination with binding
viable protein scaffold in a single forward pass through a specifically trained RoseTTAFold network. motifs derived from programmed cell death
We use these two methods to design candidate immunogens, receptor traps, metalloproteins, protein 1 (PD-1) (table S1) (15), followed by
enzymes, and protein-binding proteins and validate the designs using a combination of in silico Rosetta design, were found to have binding
and experimental tests. affinities in the mid-nanomolar range (fig. S1,
B to E). Although this experimental valida-
T
tion is encouraging, the requirement for se-
he biochemical functions of proteins are (8), these do not extend readily to more complex quence design using Rosetta is inconsistent
often carried out by a subset of residues sites composed of three or more chain seg- with the aim of jointly designing sequence
that constitute a functional site—for ex- ments, and the generated backbones are not and structure.
ample, an enzyme active site or a protein guaranteed to be designable (i.e., encodable by Following the development of RoseTTAFold
or small-molecule binding site—and hence some amino acid sequence). (RF) (16), we found that it performed better
the design of proteins with new functions An ideal method for functional de novo pro- than trRosetta in guiding protein design by
can be divided into two steps. The first step tein design would (i) embed the functional site functional site–constrained hallucination (fig.
is to identify functional site geometries and with minimal distortion in a designable scaf- S1G), likely reflecting the better overall mod-
amino acid identities that produce the desired fold protein; (ii) be applicable to arbitrary site eling of protein sequence-structure relation-
activity—for enzymes, this can be done using geometries, searching over all possible scaffold ships (16). Constrained hallucination with
quantum chemistry calculations (1–3), and for topologies and secondary structure compo- RoseTTAFold has the further advantages that,
protein binders, by fragment docking calcu- sitions for those optimal for harboring the because 3D coordinates are explicitly modeled
lations (4, 5). Alternatively, functional sites can specified site; and (iii) jointly generate back- (trRosetta only generates inter-residue distances
be extracted from a native protein having the bone structure and amino acid sequence. We and orientations), site recapitulation can be as-
desired activity (6, 7). Here, we focus on the previously demonstrated that the trRosetta sessed at the coordinate level and additional
second step: Given a functional site descrip- structure-prediction neural network (11) can problem-specific loss terms can be implemented
tion from any source, design an amino acid be used to generate new proteins by maxi- in coordinate space that assess interactions
sequence that folds up to a three-dimensional mizing the trRosetta output probability that with a target (fig. S2; materials and methods).
(3D) structure containing the site. Previous a sequence folds to some (unspecified) 3D struc-
methods can scaffold functional sites made ture during Monte Carlo sampling in sequence Generalized functional motif scaffolding by
up of one or two contiguous chain segments space (12). We refer to this process as “hal- missing information recovery
(6–10), but, with the exception of helical bundles lucination,” as it produces solutions that the While powerful and general, the constrained
network considers to be ideal proteins but hallucination approach is compute-intensive,
that do not correspond to any known natural as a forward and backward pass through the
1
Department of Biochemistry, University of Washington, protein; crystal and nuclear magnetic reso- network is required for each gradient descent
Seattle, WA 98105, USA. 2Institute for Protein Design,
University of Washington, Seattle, WA 98105, USA.
nance structures confirm that the hallucinated step during sequence optimization. In the train-
3
Graduate Program in Biological Physics, Structure and sequences fold to the hallucinated structures ing of recent versions of RoseTTAFold, a subset
Design, University of Washington, Seattle, WA 98105, USA. (12). trRosetta can also be used to design se- of positions in the input multiple sequence
4
Molecular Engineering Graduate Program, University of
Washington, Seattle, WA 98105, USA. 5Institute of
quences that fold into a target backbone struc- alignment are masked, and the network is
Bioengineering, École Polytechnique Fédérale de Lausanne, ture by carrying out sequence optimization trained to recover this missing sequence in-
CH-1015 Lausanne, Switzerland. 6FAS Division of Science, using a structure recapitulation loss function formation in addition to predicting structure.
Harvard University, Cambridge, MA 02138, USA. 7John
that rewards similarity of the predicted struc- This ability to recover both sequence and struc-
Harvard Distinguished Science Fellowship Program, Harvard
University, Cambridge, MA 02138, USA. 8Howard Hughes ture to the target structure (13). Given this tural information provides a second solution to
Medical Institute, University of Washington, Seattle, WA ability to design both sequence and struc- the functional site scaffolding problem: Given
98105, USA. ture, we reasoned that trRosetta could be a functional site description, a forward pass
*Corresponding author. Email: [email protected] (D.B.);
[email protected] (S.O.) adapted to tackle the functional site scaffold- through the network can be used to com-
These authors contributed equally to this work. ing problem. plete, or “inpaint,” both protein sequence and

RES EARCH | R E S E A R C H A R T I C L E S
A Epitope Viral Receptor

Active Sites
Protein-Protein Fig. 1. Methods for protein function design.
Presentation Traps Interactions (A) Applications of functional-site scaffolding.
(B and C) Design methods. (B) Constrained
hallucination. At each iteration, a sequence is
passed to the trRosetta or RoseTTAFold neural
network, which predicts 3D coordinates and
inter-residue distances and orientations (fig. S2).
The predictions are scored by a loss function
that rewards certainty of the predicted structure
B Method 1: Hallucination C Method 2: Inpainting
along with motif recapitulation and other task-
Predicted Final Partial Completed
Sequence Sequence
specific functions. MCMC, Markov chain Monte Carlo.
Structure Design Sequence
(C) Missing information recovery (“inpainting”).
RFjoint Partial sequence and/or structural information is
Neural network input into a modified RoseTTAFold network (called
MCMC or
Loss function
gradient
update - Hallucination RFjoint), and complete sequence and structure are
- Motif
- Problem-specific
output. (D) Protein design challenges formulated as
missing information recovery problems. Question
marks in column 1 indicate missing sequence
Desired Partial Completed
Motif Structure Structure information; gray cartoons in column 2, missing
structural information. (E) RFjoint can simultaneously
recover structure and sequence of a masked protein
D Partial information Design task F
region. 2KL8 was fed into RFjoint with a continuous
Structure (length 30) window of sequence and structure
prediction masked out, with the network tasked with predicting
the missing region of protein. Outputs (inpainted
Sequence region in gray) closely resemble the original protein
design (2KL8, left) and are confidently predicted by
AlphaFold (pLDDT/motif RMSD of models shown,
Loop design from left to right: 91.6/0.91, 92.0/0.69, and 90.4/
0.82). (F and G) Motif scaffolding benchmarking
data comparing RFjoint with constrained hallucina-
Functional tion. A set of 28 de novo designed proteins,
site design published since RoseTTAFold was trained, were
G used. For each protein, 20 random masks of length
30 were generated, and RFjoint and hallucination were
E 2KL8 L8 Inpaints
p
tasked with filling in the missing sequence and
structure to “scaffold” the unmasked “motif.” For
this mask length, RFjoint typically modestly outper-
forms hallucination, both in terms of the RMSD of
the unmasked protein (the “motif”) to the original
structure (F) and in AlphaFold confidence (pLDDT
in the replaced region) (G). Circles represent
average of 20 outputs for each of the bench-
marking proteins. Triangle represents 2KL8. Colors
in all panels: native functional motif, orange;
hallucinated/inpainted scaffold, gray; constrained motif, purple; binding partner, blue; nonmasked region, green; and masked region, light-gray dotted lines.
structure in a masked region of protein carried out further training on fixed-backbone native sequence recovery during training) and
(Fig. 1C; materials and methods). Here, the sequence design in addition to the standard structure prediction (fig. S4C). We call this net-
design challenge is formulated as an infor- fixed-sequence structure prediction task to work RFjoint and use it to generate all inpainted
mation recovery problem, analogous to the avoid model degradation (fig. S3; materials designs below unless otherwise noted.
completion of a sentence given its first few and methods). This model, denoted RFimplicit, To evaluate in silico the quality of designs
words using language models (17) or the was able to recover small, contiguous regions generated by our methods, we use the AlphaFold
completion of corrupted images using in- missing both sequence and structure (fig. S3). (AF) protein structure prediction network (21),
painting (18). A wide variety of protein struc- Encouraged by this result, we trained a model which has high accuracy on de novo designed
ture prediction and design challenges can be explicitly on inpainting segments with missing proteins (22) (fig. S7A). RF and AF have dif-
similarly formulated as missing information sequence and structure given the surrounding ferent architectures and were trained inde-
recovery problems (Fig. 1D). Although protein protein context, in addition to sequence design pendently, and hence AF predictions can be
inpainting has been explored before (19, 20), and structure prediction tasks (fig. S4A; mate- regarded as a partially orthogonal in silico test
in this study we approach it using the power of rials and methods and algorithm S1). The re- of whether RF-designed sequences fold into
a pretrained structure-prediction network. sulting model was able to inpaint missing the intended structures, analogous to tradi-
We began from a RoseTTAFold (RF) mod- regions with high fidelity (Fig. 1E and fig. S4) tional ab initio folding (13, 23). We used AF
el trained for structure prediction (16) and and performed well at sequence design (32% to compare the ability of hallucination and

RESE ARCH | R E S E A R C H A R T I C L E S
Fig. 2. Design of epitope A RSV-F B ACE2

scaffolds and receptor traps.
(in silico)
(A) Design of proteins scaffolding
immunogenic epitopes on RSV Site II
protein F (site II: PDB ID 3IXT chain P Site V ace2_76
rsvf-ii_141
residues 254 to 277; site V: PDB ID
5TPN chain A residues 163 to 181).
Comparisons of the RF hallucinated
models to AF2 structure predictions
from the design sequence are in rsvf-ii_158 ace2_1157
fig. S9; here, because of space
constraints, we show only the AF2
model (the two are very close in
all cases). Here and in the following
figures, we assess the extent rsvf-v_hal_1 rsvf-v_hal_2 rsvf-v_hal_3 ace2_1007
of success in designing sequences
that fold to structures harboring C D 20
1.0 RSV-F
(103 deg cm2 dmol–1)

the desired motif through two
RUmax, normalized
0.8 rsvfv_hal_1
metrics computed on the AF2 0
rsvfv_hal_2
predictions: prediction confidence 0.6
MRE
(AF pLDDT) and the accuracy 0.4 rsvfv_hal_1 S20R rsvfv_hal_1
of recapitulation of the original rsvfv_hal_2 S19R rsvfv_hal_2
0.2 rsvfv_hal_3
rsvfv_hal_3 S17R
scaffolded motif (motif AF-RMSD).
0.0 rsvfv_hal_2 L18R
For RSV-F designs, these metrics rsvfv_hal_3 L16R 200 210 220 230 240 250
3.2 10 32 100 316 1000 3162
are rsvf_ii_141 (85.0, 0.53 Å), [hRSV90] (nM) Wavelength (nm)
rsvf_ii_158 (82.9, 0.51 Å), rsvf_ii_171
(88.4, 0.69 Å), rsvfv_hal_1 (82, 0.7 Å), rsvfv_hal_2 (88, 0.64 Å), and rsvfv_hal_3 surface plasmon resonance signal (response units) of purified RSV-F
(86, 0.65 Å). (B) Design of COVID-19 receptor trap based on ACE2 interface epitope scaffolds and point mutants at various concentrations of hRSV90
helix (PDB ID 6VW1 chain A residues 24 to 42). Design metrics: ace2_76 (89.1, antibody, with sigmoid fits. RSV-F refers to purified trimeric native F protein.
0.55 Å), ace2_1157 (80.4, 0.47 Å), and ace2_1007 (83.3, 0.57 Å). Colors: Kd values are as follows: RSV-F: 24 nM; rsvfv_hal_1: 0.9 mM; rsvfv_hal_2:
native protein scaffold, light yellow; native functional motif, orange; hallucinated 1.0 mM; rsvfv_hal_3: 1.3 mM. (D) Mean residue ellipticity (MRE) versus
scaffold, gray; hallucinated motif, purple; and binding partner, blue. See wavelength, from CD spectroscopy, for the three RSV-F site V hallucinations
table S2 for additional metrics on each design. (C) Normalized maximum with binding activity.
inpainting to rebuild missing protein regions critically, we assessed the activities of the backbone RMSD in a variety of folds [Fig. 2A
(Fig. 1, F and G, and fig. S5). Inpainting yielded designs experimentally (with the exception of and fig. S9; structures and sequences for all
solutions with more accurately predicted fixed those labeled “in silico” in Figs. 2 to 5). designs below are given in data S1 and S2 and
regions (“AF-RMSD”; Fig. 1G and fig. S5B) and differ considerably from native proteins (table
structures overall more confidently predicted Designing immunogen candidates and S2); RF hallucinated models and AF structure
from their amino acid sequences (“AF pLDDT”; receptor traps predictions are shown in figs. S9, S11, and S17;
Fig. 1F and fig. S5A) and required only 1 to 10 s The goal of immunogen design is to scaffold only the AF model is shown in the main figures].
per design on an NVIDIA RTX 2080 graphics a native epitope recognized by a neutralizing Inpainting also generated scaffolds for RSV-F
processing unit (hallucination requires 5 to antibody as accurately as possible in order to site V, with comparable quality but less diver-
20 min per design). However, hallucination elicit antibodies binding the native protein sity than the hallucinations (fig. S8).
gave better results when the missing region upon immunization. Additional interactions We expressed 37 hallucinated RSV-F site V
was large (fig. S5) and generated greater struc- with the antibody are undesirable because the scaffolds with high AF pLDDT and low motif
tural diversity (fig. S8; and see below). aim is to elicit antibodies recognizing only the AF-RMSD in Escherichia coli and found that
In the following sections, we highlight the original antigen, and hence for hallucination, three bound the neutralizing antibody hRSV90
power of the constrained hallucination and we add a repulsive loss term to penalize in- (27) with a dissociation constant (Kd) of 0.9
inpainting methods by designing proteins teractions with the antibody beyond those to 1.3 mM (Fig. 2C and fig. S11; materials and
containing a wide range of functional motifs present in the scaffolded epitope (fig. S2; sup- methods and supplementary text). The Kd for
(Figs. 2 to 5 and table S1). For almost all plementary text). As a test case, we focused on the RSVF trimer is lower (23 nM), but the
problems, we obtained designs that are closely respiratory syncytial virus F protein (RSV-F), interface is larger, encompassing both sites II
recapitulated by AF with overall and motif which has several antigenic epitopes for which and V (27). Mutation of either of two key epi-
(functional site) root mean square deviation structures with neutralizing antibodies have tope residues reduced or abolished binding of
(RMSD) of typically <2 and <1 Å, respectively, been determined (7, 9, 10). We scaffolded the designs, suggesting that they bind the target
with high model confidence [predicted local RSV-F site II, a 24-residue helix-loop-helix through the scaffolded motif (Fig. 2C and fig.
distance difference test (pLDDT) > 80; table motif that had previously been grafted suc- S11A), and circular dichroism (CD) spectra were
S2]; such recapitulation suggests that the cessfully onto a three-helix bundle (7), as well consistent with the designed scaffold structures
designed sequences encode the designed as RSV-F site V, a 19-residue helix-loop-strand for both the original hallucinations (Fig. 2D)
structures [although it should be noted that motif that has not yet been scaffolded success- and the epitope mutants (fig. S11C). Four of
AF has limited ability to predict protein sta- fully (27). We were able to hallucinate designs the inpainted designs bound hRSV90 by yeast
bility (24) or mutational effects (25, 26)]. More recapitulating both epitopes to sub-angstrom display but were poorly expressed in E. coli

Fig. 3. Design of metal binding. (A) Scaffolding of di-iron binding site from E. coli of dife_inp_1 in the presence and absence of Co2+ are both consistent
cytochrome b1 (PDB ID 1BCF chain A residues 18 to 25, 27 to 54, 94 to 97, with the predicted helical structure. (E) Temperature dependence of
and 123 to 130) using inpainting. Colors: native protein scaffold, light yellow; dife_inp_1 CD signal in the presence and absence of Co2+. Coordination of Co2+
native functional motif, orange; hallucinated scaffold, gray; hallucinated in the core stabilizes the protein. Protein concentration: 6.7 mM; Co2+
motif, purple; and bound metal, blue. (B) Absorbance spectra of dife_inp_1 concentration: 53.3 mM. (F) Inpainted design EFhand_inp_1 scaffolding the
(or mutant) in the presence (or absence) of an eight-fold molar excess of double EF-hand motif with input motif residues in purple, input nonmotif
Co2+. Peaks at 520, 555, and 600 nm, consistent with Co2+ binding to residues in green, and overlaid with the native motif from PDB ID 1PRW
the scaffolded motif (32). In the mutant, the six coordinating residues (orange). (G) CD spectra of EFhand_inp_1 incubated with and without CaCl2
[side chains shown in (A)] are mutated to alanine (E16A, E55A, H58A, E89A, suggest stabilization of the protein upon binding calcium. (H) Tryptophan-
H92A, E115A). Protein concentration: 200 mM. (C) dife_inp_1 Co2+ titration enhanced terbium fluorescence spectra of EFhand_inp_1 suggests that
(protein concentration: 200 mM). Quantification of the absorbance at 550 nm, the design binds terbium (57). Terbium binding signal is competed by 1 mM
using a predicted extinction coefficient of 155 for Co2+ binding the motif CaCl2 (red). Design metrics (AF pLDDT, motif AF-RMSD): dife_inp_1
(32), is consistent with both binding sites being recapitulated. (D) CD spectra (92, 0.65 Å) and EFhand_inp1 (84, 0.7 Å).

A Carbonic Anhydrase II (in silico) B Fig. 4. In silico design of enzyme active sites.
(A and B) Hallucinations using backbone description
of site using RF. (C and D) Hallucination using
side-chain description of site using AF2 augmented
with trRosetta (materials and methods). (A) Carbonic
anhydrase II active site (PDB ID 5YUI chain A
residues 62 to 65, 93 to 97, and 118 to 120).
(B) D5-3-ketosteroid isomerase active site (PDB ID
hcA_1 hcA_2 1QJG chain A residues 14, 38, and 99). Colors: native
protein scaffold, light yellow; native functional
C KSI (in silico) D motif, orange; hallucinated scaffold, gray; hallucinated
motif, purple; and bound metal, blue. [(B) and (D)]
Zoomed-in view of designed active sites. Design
metrics (AF pLDDT, motif AF-RMSD): hcA_1 (73,
1.04 Å), hcA_2 (71, 0.62 Å), KSI_1 (84, 0.30 Å Cb),
and KSI_2 (72, 0.53 Å Cb).
KSI_1 KSI_2
A PD-1 / PD-L1 B C pdl1_inp_1

Fig. 5. Design of protein-binding proteins. Designs
containing target-binding interfaces built around
Max BLI signal (nm)

native-complexÐderived binding motifs. Targets are
0.15
in blue, native scaffolds in yellow or pink, native motifs
in orange, designed scaffolds in gray, and designed
0.10
motifs in purple. (A) Crystal structure of HAC PD-1 in
complex with PD-L1. (B) Inpainted PD-L1 binder
0.05
superimposed on PD-1 interface motif. (C) BLI binding
signal versus PD-L1 concentration. Kd = 326 nM.
pdl1_inp_1 101 102 103 (D) Crystal structure of previously designed TrkA
[PD-L1] (nM)
D F minibinder in complex with TrkA, superimposed
TrkA on TrkA receptor dimer. (E) Hallucinated bivalent TrkA
trkA_hal_1 binder. Protein topology diagrams are on the right.
Max BLI signal (nm)
WT
0.4 (F) BLI binding signal versus TrkA concentration;
F27R mutations at both scaffolded binding sites reduce
0.3
F68R TrkA binding. (G) Hallucinated Mdm2 binder designs
0.2
superimposed on native p53 helix in complex with
E F27R F68R
0.1 Mdm2 (see also fig. S17, D and E). New binding
0 interactions (hallucinated residues within 5 Å of the
101 102 103 104 target) are in green. (Inset) Overlay of mdm2_hal_1
[Binder] (nM)
and native p53 helix showing key side chains
trkA_hal_1 for binding.
G p53 / Mdm2 (in silico)
mdm2_hal_1 mdm2_hal_2 mdm2_hal_3
(fig. S11, C to E). Overall, the designs provide a the native receptor to avoid opportunities for face with sub-angstrom accuracy (Fig. 2B and
diverse set of promising starting points for fur- viral escape. As a test case, we scaffolded the fig. S9C).
ther RSV-F epitope-based vaccine development. helix of human angiotensin-converting enzyme 2
We next applied hallucination to the in silico (hACE2) interacting with the receptor binding Designing metal-coordinating proteins
design of receptor traps that neutralize viruses domain of severe acute respiratory syndrome Di-iron sites are important in biological sys-
by mimicking their natural binding targets and coronavirus 2 (SARS-CoV-2) spike protein (28). tems for iron storage (29) and can mediate
thus are inherently robust against mutational The hallucinated hACE2 mimetics have a di- catalysis (30, 31). We were able to recapitulate
escape. We again augmented the loss function verse set of helical topologies, and AF structure the di-iron site from E. coli bacterioferritin,
with a penalty on interactions beyond those in predictions recapitulate the binding inter- composed of four parallel helical segments, to
SCIENCE science.org 22 JULY 2022 ¥ VOL 377 ISSUE 6604 391

sub-angstrom AF-RMSD using both inpainting methods; a side chain–predicting version of followed by inpainting (materials and meth-
(Fig. 3, A to E, and fig. S13) and hallucination RF was not available at the time) but found it ods) to scaffold them on a single chain (Fig. 5,
(fig. S12; the hallucinations were not tested difficult to obtain accurate side-chain place- D and E). A design predicted to be well struc-
owing to buried polar residues; supplemen- ment; the landscape may be too rugged with tured (AF pLDDT > 80) and interact with
tary text). The designs had diverse helix con- the high-resolution side chain–based loss (sup- TrkA (inter-PAE < 10 Å) was expressed, pu-
nectivities and low structural similarity to the plementary text). Better results were obtained rified, and found to bind TrkA, as assessed
parent [figs. S13B and S12; template modeling with a two-stage approach using, first, both AF by biolayer interferometry (BLI) (Fig. 5F). A
(TM)–score 0.55 to 0.71 to PDB ID 1BCF_A]. We and trRosetta (to smoothen the loss landscape) double mutant that knocked out both de-
chose 96 inpainted designs to test experimentally and a description of the active site at the back- signed binding sites abolished TrkA binding,
and found that 76 had soluble expression, at bone level, followed by a second all-atom AF- whereas single mutants knocking out either
least eight (see supplementary text) had a only stage once the overall backbone was one of the binding sites maintained partial
spectroscopic shift indicative of Co2+ binding roughly in place. This yielded multiple plausi- binding (Fig. 5F and fig. S16), suggesting that
(a proxy for iron binding) (32, 33), and three ble solutions with nearly exact matches to the the protein binds two molecules of TrkA, as
(dife_inp_1, dife_inp_2, and dife_inp_3; Fig. 3B catalytic side-chain geometry (Fig. 4, C and D, designed.
and fig. S13E) had CD spectra consistent with and fig. S9E). In silico validation with a held- RoseTTAFold is able to predict the structures
the designed fold (Fig. 3D and fig. S13F) and out AF model (materials and methods) reca- of protein complexes (40), and we hypothesized
were stabilized by metal binding (Fig. 3E and pitulated the designed active sites. The use of that it could generate additional binding inter-
fig. S13G). Mutation of the metal binding resi- stage-specific loss functions illustrates the actions between hallucinated or inpainted
dues abolished binding (Fig. 3B and fig. S13E), ready customizability of the hallucination binder and a target beyond the scaffolded
and titration analysis of dife_inp_1 suggested approach to specific design challenges without motif. We used a “two-chain” hallucination
that both metal binding sites were successfully network retraining. protocol (fig. S17; materials and methods) to
scaffolded (Fig. 3C). design binders to the Mdm2 oncogene by scaf-
We next scaffolded the calcium-binding Designing protein-binding proteins folding the native N-terminal helix of the tumor
EF-hand motif (34), a 12-residue loop flanked To design binders to the cancer checkpoint suppressor protein p53 and obtained diverse
by helices. Both constrained hallucination and protein PD-L1, we scaffolded two discontig- designs with AF inter-PAE < 7 Å, target-aligned
inpainting readily generated scaffolds recapit- uous segments of the interfacial b sheet from binder RMSD < 5 Å, binder pLDDT > 85, and
ulating either one or two EF-hand motifs to a high-affinity mutant of PD-1 (Fig. 5A; mate- spatial aggregation propensity (SAP) score < 35
within 1.0 Å AF-RMSD of the native motif rials and methods) (15). Inpainting yielded (fig. S17, D and E); three examples are shown
(Fig. 3F; fig. S14, A and B; and table S2). We designs with not only good AF predictions of in Fig. 5G.
chose 20 hallucinations and 55 inpaints to the binder monomer (AF pLDDT > 80, motif The above approaches to protein-binder de-
display on yeast and screen for calcium binding AF-RMSD < 1.4 Å) but also of the complex sign require starting from a previously known
using tryptophan-enhanced terbium fluores- between the binder and PD-L1, with an inter- binding motif, but hallucination should in
cence (35). Six hallucinations and four in- chain predicted alignment error (inter-PAE) of principle be able to generate de novo inter-
paintings had fluorescence consistent with <10 Å (materials and methods). In contrast to faces as well. To test this, we used two-chain
ion binding [fig. S14A; materials and methods; our initial efforts with trRosetta hallucination hallucination to optimize 12-residue peptides
one of these proteins (EFhand_inp_2) was de- (fig. S1; supplementary text), it was not nec- for binding to 12 targets starting from ran-
signed using RFimplicit (supplementary text)]. essary to redesign the inpainted sequences dom sequences, minimizing an interchain
The top hit from yeast, the inpainted EFhand_ using Rosetta. Of 31 designs selected for ex- entropy loss (fig. S17H). Most of the halluci-
inp_1, purified from E. coli as a monomer (fig. perimental testing, one design, pdl1_inp_1, nated peptides bound at native protein inter-
S14C), had the expected CD spectrum (Fig. 3G) bound PD-L1 with a Kd of 326 nM (Fig. 5, B action sites (fig. S18A); the remainder bound
and a clear terbium binding signal (Fig. 3H) that and C), worse than high-affinity consensus in hydrophobic grooves resembling protein
was eliminated by CaCl2 competition (Fig. 3H). (HAC) PD-1 (Kd = 110 pM) (37) but better binding sites (fig. S18B). We used the same
than wild-type PD-1 (Kd = 3.9 mM) (37). The procedure to generate 55- to 80-residue bind-
In silico design of enzyme active sites pdl1_inp_1 design expressed as a monomer ers against TrkA and PDL-1 without starting
We next sought to scaffold the active site of (fig. S15E), was thermostable, and had a CD motif information and obtained designs pre-
carbonic anhydrase II, which catalyzes the spectrum consistent with that of a mixed a-b dicted by AF to complex with the target, at the
interconversion of carbon dioxide and bicar- fold (fig. S15F). Unlike native PD-1, which has native ligand binding site, with a target-aligned
bonate and has recently been of interest for an immunoglobulin family b-sandwich fold, binder RMSD < 5 Å and an inter-PAE < 10 Å
carbon sequestration (31–33). The active site pdl1_inp_1 has two helices buttressing the (fig. S17, F and G).
consists of three Zn2+-coordinating histidines interfacial b sheet, as well as an additional Unlike classical protein design pipelines,
on two strands and a threonine on a loop, fifth inpainted strand extending the interface which treat backbone generation and sequence
which orients the CO2 (table S1). Despite the (fig. S15, A and B). The closest Protein Data design as two separate problems, our methods
complexity of the irregular, discontinuous Bank (PDB) (38) hit had a TM-score of 0.61, simultaneously generate both sequence and
three-segment site, hallucination was able and the closest Basic Local Alignment Search structure, taking advantage of the ability of
to generate designs with sub-angstrom motif Tool (BLAST) NR hit had a sequence iden- RoseTTAFold to reason over and jointly opti-
AF-RMSDs with correct His placement for Zn2+ tity of 25.4%. mize both data types. This results in excellent
coordination (Fig. 4A and fig. S9D); these are We next used our methods to design ligands performance in both generating protein back-
less than 100 residues in size, considerably engaging multiple receptor binding sites. bones with a geometry capable of hosting a
smaller than the 261-residue native protein. The nerve growth factor (NGF) receptor TrkA desired site and sequences that strongly en-
We next scaffolded the catalytic side chains dimerizes upon ligand binding (39), and start- code these backbones. Our hallucinated and
of D5-3-ketosteroid isomerase (KSI) (table S1) ing from the TrkA-NGF crystal structure, we inpainted backbones accommodate all of the
involved in steroid hormone biosynthesis (36). positioned helical segments derived from tested functional sites much more accurately
We attempted to use gradient descent by two copies of a previously designed TrkA than any naturally occurring protein in the
backpropagation through AF (materials and binding protein (4) and used hallucination PDB or AF predictions database (fig. S20 and

table S3; supplementary text) (41), and our its key strength, the ability to use arbitrary loss 44. S. Biswas, G. Khimulya, E. C. Alley, K. M. Esvelt, G. M. Church,
designed structures are predicted more con- functions tailored to specific problems and Nat. Methods 18, 389–396 (2021).
45. D. Repecka et al., Nat. Mach. Intell. 3, 324–333 (2021).
fidently from their (single) sequences than design any length sequence without retraining. 46. J.-E. Shin et al., Nat. Commun. 12, 2403 (2021).
most native proteins with known crystal struc- The ability of our inpainting approach to ex- 47. Z. Wu, K. E. Johnston, F. H. Arnold, K. K. Yang, Curr. Opin.
tures and are on par with structurally vali- pand from a given functional site to generate a Chem. Biol. 65, 18–27 (2021).
48. N. Anand et al., Nat. Commun. 13, 746 (2022).
dated de novo designed proteins (fig. S7, A coherent sequence-structure pair should find 49. A. Madani et al., bioRxiv 2021.07.18.452833 [Preprint]
and B). The hallucination and inpainting wide application in protein design because of (2021); https://doi.org/10.1101/2021.07.18.452833.
approaches are complementary: Hallucina- its speed and generality. The two approaches 50. S. Ovchinnikov, P.-S. Huang, Curr. Opin. Chem. Biol. 65,
136–144 (2021).
tion can generate diverse scaffolds for mini- individually, and the combination of the two, 51. N. Anand, R. Eguchi, P.-S. Huang, “Fully differentiable full-atom
malist functional sites but is computationally should increase in power as more-accurate pro- protein backbone generation,” Seventh International
expensive because it requires a forward and tein structure, interface, and small-molecule Conference on Learning Representations (ICLR 2019),
New Orleans, Louisiana, 6 to 9 May 2019.
backward pass through the neural network binding prediction networks are developed.
52. R. R. Eguchi, C. A. Choe, P.-S. Huang, PLOS Comput. Biol. 18,
to calculate gradients for each optimization e1010271 (2022).
step (materials and methods), whereas in- RE FERENCES AND NOTES 53. Z. Lin, T. Sercu, Y. LeCun, A. Rives, “Deep generative models
painting usually requires larger input motifs 1. D. Röthlisberger et al., Nature 453, 190–195 (2008). create new and diverse protein structures,” 35th Conference
2. J. B. Siegel et al., Science 319, 1387–1391 (2008). on Neural Information Processing Systems (NeurIPS 2021),
but is much less compute-intensive and out- 6 to 14 December 2021.
3. J. B. Siegel et al., Science 329, 309–313 (2010).
performs the hallucination method when 4. L. Cao et al., Nature 605, 551–560 (2022). 54. M. Jendrusch, J. O. Korbel, S. K. Sadiq, bioRxiv 2021.10.11.
more starting information is provided. This 5. A. Chevalier et al., Nature 550, 74–79 (2017). 463937 [Preprint] (2021); https://doi.org/10.1101/2021.
difference in performance can be understood 6. E. Procko et al., Cell 157, 1644–1656 (2014). 10.11.463937.
7. B. E. Correia et al., Nature 507, 201–206 (2014). 55. L. Moffat, J. G. Greener, D. T. Jones, bioRxiv 2021.08.24.457549
by considering the manifold in sequence- 8. D.-A. Silva et al., Nature 565, 186–191 (2019). [Preprint] (2021); https://doi.org/10.1101/2021.08.24.457549.
structure space corresponding to folded pro- 9. F. Sesterhenn et al., Science 368, eaay5051 (2020). 56. L. Moffat, S. M. Kandathil, D. T. Jones, bioRxiv 2022.01.27.478087
teins. The inpainting approach can be viewed 10. C. Yang et al., Nat. Chem. Biol. 17, 492–500 (2021). [Preprint] (2022); https://doi.org/10.1101/2022.01.27.478087.
11. J. Yang et al., Proc. Natl. Acad. Sci. U.S.A. 117, 1496–1503 (2020). 57. L. Li et al., J. Phys. Chem. C 112, 12219–12224 (2008).
as projecting an incomplete input sequence- 58. J. Wang et al., RFDesign: Protein hallucination and inpainting
12. I. Anishchenko et al., Nature 600, 547–552 (2021).
structure pair onto the subset of the mani- 13. C. Norn et al., Proc. Natl. Acad. Sci. U.S.A. 118, e2017228118 (2021). with RosettaFold, version 2, Zenodo (2022); https://doi.org/
fold of folded proteins (as represented by 14. D. Tischer et al., bioRxiv 2020.11.29.402743 [Preprint] (2020); 10.5281/zenodo.6808038.
RoseTTAFold) containing the functional site— https://doi.org/10.1101/2020.11.29.402743.

15. R. Pascolutti et al., Structure 24, 1719–1728 (2016). AC KNOWLED GME NTS
if insufficient starting information is provided, 16. M. Baek et al., Science 373, 871–876 (2021). We thank L. Goldschmidt and K. VanWormer, respectively, for
this projection is not well determined, but with 17. J. Devlin, M.-W. Chang, K. Lee, K. Toutanova, arXiv:1810.04805 maintaining the computational and wet lab resources at the
sufficient information, it produces protein-like [cs.CL] (2019). Institute for Protein Design; C. Norn for general discussions about
18. R. A. Yeh et al., arXiv:1607.07539 [cs.CV] (2017). trRosetta; B. Coventry for advice on interface design; C. Goverde
solutions, updating sequence and structure in- 19. Z. Li, S. P. Nguyen, D. Xu, Y. Shang, Proc. Int. Conf. Tools. Artif. for advice on RSV-F epitopes and motif grafting methods; T. Yu,
formation simultaneously. The loss function Intell. 29, 1085–1091 (2017). G. R. Lee, L. An, and X. Wang for advice on flow cytometry; R. Dong
used in the hallucination approach is con- 20. N. Anand, P. Huang, in Advances in Neural Information and V. Muhunthan for exploratory analyses; N. Hiranuma for
Processing Systems 31, S. Bengio et al., Eds. (Curran exploratory RoseTTAFold training sessions; B. Trippe for feedback
structed with the goal that minima lie in the Associates, Inc., 2018), pp. 7494–7505. on the manuscript; S. Pellock for expertise on enzyme design;
protein manifold, but there will likely not be a 21. J. Jumper et al., Nature 596, 583–589 (2021). A. Fitzgibbon for conceptual discussions on training RoseTTAFold;
perfect correspondence, and hence stochastic 22. R. Chowdhury et al., bioRxiv 2021.08.02.454840 [Preprint] and C. Garcia for providing biotinylated TrkA. Funding: We
(2021); https://doi.org/10.1101/2021.08.02.454840. thank Microsoft for support and for providing Azure computing
optimization of the loss function in sequence 23. K. T. Simons, R. Bonneau, I. Ruczinski, D. Baker, Proteins 37 resources. This work was supported with funds provided by
space may not produce solutions that are as (suppl. 3), 171–176 (1999). the Audacious Project at the Institute for Protein Design (D.B. and
protein-like as those from the inpainting 24. T.-E. Kim et al., bioRxiv 2021.12.17.472837 [Preprint] (2021); A.S.); a Microsoft gift (M.B. and J.D.); Eric and Wendy Schmidt
https://doi.org/10.1101/2021.12.17.472837. by recommendation of the Schmidt Futures (D.J.); the DARPA
approach.
25. M. A. Pak et al., bioRxiv 2021.09.19.460937 [Preprint] (2021); Synergistic Discovery and Design project HR001117S0003
https://doi.org/10.1101/2021.09.19.460937. contract FA8750-17-C-0219 (D.B. and W.Y.); the DARPA Harnessing
Conclusion 26. G. R. Buel, K. J. Walters, Nat. Struct. Mol. Biol. 29, 1–2 (2022). Enzymatic Activity for Lifesaving Remedies project HR001120S0052
The approaches for scaffolding functional sites 27. J. J. Mousa, N. Kose, P. Matta, P. Gilchuk, J. E. Crowe Jr., contract HR0011-21-2-0012 (N.B.); the Washington Research
Nat. Microbiol. 2, 16271 (2017). Foundation (J.W.); the Open Philanthropy Project Improving Protein
presented here require no inputs other than 28. T. W. Linsky et al., Science 370, 1208–1214 (2020). Design Fund (D.B. and D.T.); Amgen (S.L.); the Human Frontier
the structure and sequence of the desired 29. F. Frolow, A. J. Kalb, J. Yariv, Nat. Struct. Biol. 1, 453–460 (1994). Science Program Cross Disciplinary Fellowship (LT000395/2020-C)
functional site and, unlike previous methods, 30. A. Lombardi, F. Pirro, O. Maglio, M. Chino, W. F. DeGrado, and EMBO Non-Stipendiary Fellowship (ALTF 1047-2019) (L.F.M.);
Acc. Chem. Res. 52, 1148–1159 (2019). the EMBO Fellowship (ALTF 191-2021) (T.S.); European Molecular
do not require specifying the secondary struc- 31. J. R. Calhoun et al., Biopolymers 80, 264–278 (2005). Biology Organization Grant (ALTF 139-2018) (B.I.M.W.); the “la
ture or topology of the scaffold and can simul- 32. A. M. Keech et al., J. Biol. Chem. 272, 422–429 (1997). Caixa” Foundation (M.E.); the National Institute of Allergy and
taneously generate both sequence and structure. 33. E. N. G. Marsh, W. F. DeGrado, Proc. Natl. Acad. Sci. U.S.A. 99, Infectious Diseases (NIAID) Federal Contract HHSN272201700059C
5150–5154 (2002). (I.A.), NIH grant DP5OD026389 (S.O.); the National Science
Despite a recent surge of interest in using 34. M. Yáñez, J. Gil-Longo, M. Campos-Toimil, in Calcium Foundation MCB 2032259 (S.O.); the Howard Hughes Medical
machine learning to design protein sequences Signaling, Md. S. Islam, Ed., vol. 740 of Advances in Institute (D.B., R.R., and K.M.C.), the National Institute on Aging
(42–49), the design of protein structure is rela- Experimental Medicine and Biology (Springer Netherlands, grant 5U19AG065156 (D.B., J.L.W., D.R.H., and M.E.); the National
2012), pp. 461–482. Cancer Institute grant R01CA240339 (D.B. and J.-H.C.); Swiss
tively underexplored, likely because of the 35. S. J. Caldwell et al., Proc. Natl. Acad. Sci. U.S.A. 117, National Science Foundation (K.M.C. and B.C.); Swiss National
difficulty of efficiently representing and learn- 30362–30369 (2020). Center of Competence for Molecular Systems Engineering (K.M.C.
ing structure (50). Generative adversarial net- 36. H.-S. Cho et al., J. Biol. Chem. 274, 32863–32868 (1999). and B.C.); Swiss National Center of Competence in Chemical
37. R. L. Maute et al., Proc. Natl. Acad. Sci. U.S.A. 112, Biology (K.M.C. and B.C.); and European Research Council grant
works and variational autoencoders have been
E6506–E6514 (2015). 716058 (K.M.C. and B.C.). Author contributions: Designed the
used to generate protein backbones for spe- 38. H. M. Berman et al., Nucleic Acids Res. 28, 235–242 (2000). research: J.W., S.L., D.J., D.T., J.L.W., S.O., and D.B. Developed
cific fold families (51–53), whereas our ap- 39. C. Wiesmann, M. H. Ultsch, S. H. Bass, A. M. de Vos, Nature the motif-constrained hallucination method: J.W., D.T., S.L.,
proach leverages the training of RoseTTAFold 401, 184–188 (1999). I.A., and S.O. Contributed code and ideas for hallucination: M.B.
40. I. R. Humphreys et al., Science 374, eabm4805 (2021). and J.D. Generated designs using hallucination: J.W., S.L., D.T.,
on the entire PDB to generate an almost un- 41. K. Tunyasuvunakool et al., Nature 596, 590–596 (2021). and S.O. Developed the inpainting method: D.J. and J.L.W.
limited diversity of new structures and enable 42. J. Ingraham, V. K. Garg, R. Barzilay, T. Jaakkola, “Generative Contributed code and ideas for inpainting: M.B., J.W., S.L., and
the scaffolding of any desired constellation models for graph-based protein design,” 33rd Conference on D.T. Generated designs using inpainting: D.J., J.L.W., and A.S.
of functional residues. Our “activation max- Neural Information Processing Systems (NeurIPS 2019), Analyzed data: J.W., S.L., D.J., D.T., J.L.W., and M.E. Trained neural
Vancouver, Canada, 8 to 14 December 2019. networks: D.J., J.L.W., and M.B. Performed RSV-F experiments:
imization” hallucination approach extends 43. A. Strokach, D. Becerra, C. Corbi-Verge, A. Perez-Riba, K.M.C., R.R., L.F.M., and J.W. Performed di-iron experiments: J.L.W.
related work in this area (54–56) by leveraging P. M. Kim, Cell Syst. 11, 402–411.e4 (2020). and D.J. Performed EF-hand experiments: A.S. and J.L.W.

Performed PD-L1 experiments: W.Y., D.R.H., J.W., S.L., and D.J. authors, some rights reserved; exclusive licensee American Tables S1 to S5
Contributed reagents and technical expertise: T.S., J.-H.C., L.F.M., Association for the Advancement of Science. No claim to original Algorithm S1
N.B., B.I.M.W., B.C., A.M., and F.D. Wrote the manuscript: J.W., US government works. https://www.science.org/about/science- References (59Ð86)
D.J., J.L.W., S.L., D.T., S.O., and D.B. Competing interests: The licenses-journal-article-reuse MDAR Reproducibility Checklist
authors declare that they have no competing interests. Data and
SUPPLEMENTARY MATERIALS Data S1 and S2
materials availability: Code and neural network weights are
available at https://github.com/RosettaCommons/RFDesign science.org/doi/10.1126/science.abn2100
and https://github.com/sokrypton/ColabDesign and archived Materials and Methods
at Zenodo (58). Plasmids of designed proteins are available Supplementary Text Submitted 11 November 2021; accepted 24 June 2022
upon request. License information: Copyright © 2022 the Figs. S1 to S21 10.1126/science.abn2100
SURFACE CHEMISTRY important quantum effect on surface reactions;

for example, in H atom recombination, only
Quantum effects in thermal reaction rates at one out of four electron spin combinations
yields a stable H2 molecule. However, the spin-
metal surfaces degeneracy of reactants and products has, to
our knowledge, never been included in calcu-
Dmitriy Borodin1,2*, Nils Hertl1,2, G. Barratt Park1,2,3, Michael Schwarzer1, Jan Fingerhut1, lations of reaction rates at metal surfaces.
Yingqi Wang4, Junxiang Zuo4, Florian Nitz1, Georgios Skoulatakis2, Alexander Kandratsenka2, This paper reports kinetic data for H atom
Daniel J. Auerbach2, Dirk Schwarzer2, Hua Guo4, Theofanis N. Kitsopoulos1,2,5,6, Alec M. Wodtke1,2* recombination on both the Pt(111) and Pt(332)
surfaces obtained with velocity-resolved kinetics
There is wide interest in developing accurate theories for predicting rates of chemical reactions (VRK), which was previously used only to study
that occur at metal surfaces, especially for applications in industrial catalysis. Conventional methods first-order and pseudo–first-order reactions on
contain many approximations that lack experimental validation. In practice, there are few reactions model catalysts (15, 16). For this work, we have
where sufficiently accurate experimental data exist to even allow meaningful comparisons to theory. extended VRK to the measurement of rate
Here, we present experimentally derived thermal rate constants for hydrogen atom recombination constants for second-order reactions by mea-
on platinum single-crystal surfaces, which are accurate enough to test established theoretical suring the absolute reactant flux, which, when
approximations. A quantum rate model is also presented, making possible a direct evaluation of combined with known sticking probabilities
the accuracy of commonly used approximations to adsorbate entropy. We find that neglecting the (17, 18), provides accurate initial concentra-
wave nature of adsorbed hydrogen atoms and their electronic spin degeneracy leads to a 10× to tions [H]0 and eliminates the main source of
1000× overestimation of the rate constant for temperatures relevant to heterogeneous catalysis. error found in previous work.
These quantum effects are also found to be important for nanoparticle catalysts. To understand the kinetics more deeply, we
also constructed a quantum rate model (QRM)
E
that accurately reproduced experimental rate
normous effort has gone into developing leading to H2 formation. Being perhaps the constants over 12 orders of magnitude for
predictive theories of thermal reaction simplest reaction for theoretical modeling and temperatures between 250 and 1000 K with no
rates (1), with one goal being accurate omnipresent as an elementary step in indus- adjustable parameters. Comparison to a cor-
kinetic models of heterogeneous catal- trial catalysis [e.g., hydrogenation of unsaturated responding classical rate model (CRM) revealed
ysis, an industrial cornerstone of modern fats (6), ammonia synthesis (7), and electro- how large and crucially important quantum
society (2). Modeling real catalytic reactors chemical hydrogen production (8)], it is an effects are; the classical reaction rate constants
presents technical problems because they often obvious starting point for the development of were ~20 times larger than quantum rate
involve networks of reactions (3, 4), compli- accurate rate theories in surface chemistry. constants even at 1000 K, with an increasing
cating meaningful comparisons to experiment Unfortunately, large uncertainties in the ex- deviation at lower temperatures. For reactions
that could test a theory’s assumptions. A pos- perimentally derived second-order rate con- at stepped surfaces, the errors were even higher.
sible solution is to compare experiment and stants arise because of difficulties in obtaining This dramatic quantum reduction of the reac-
theory using simplified model systems that accurate initial concentrations (9). If these and tion rate resulted from both the delocalization of
involve only a single elementary reaction. other experimental problems could be over- the adsorbed H* nuclei as well as the influence
Unfortunately, even this comparison is seldom come, this reaction would provide an ideal of electron spin degeneracy.
achieved because accurate measurements of system for benchmarking rate theory, espe-
elementary reaction rates are rare in surface cially for testing approximate treatments of Results
chemistry (5). quantum effects. The experiments are described in detail in
Illustrative of these problems is the thermal From the study of gas-phase reactions, exact the supplementary materials (SM). Briefly, a
recombination of H atoms on transition metals, treatments of nuclear quantum effects are often pulsed molecular beam with a controlled mix-
considered to be unnecessary above ~500 K (10), ture of H2 and D2 illuminated either a Pt(111)
1
Institute for Physical Chemistry, University of Göttingen, and, because most catalytic reactors operate or Pt(332) crystal facet, with step densities
Tammannstraße 6, 37077 Göttingen, Germany. 2Department
of Dynamics at Surfaces, Max Planck Institute for
at increased temperatures, one might conclude of 0.1 to 0.6% and 16.7%, respectively. The
Multidisciplinary Sciences, am Faßberg 11, 37077 Göttingen, that a classical approximation (11) or approx- transient rates of HD formation were then
Germany. 3Department of Chemistry and Biochemistry, imate ad hoc quantum treatments, like har- recorded using VRK, where pulsed laser-
Texas Tech University, Lubbock, TX 79409-1061, USA. monic transition-state theory (hTST) (12, 13),
4 ionization, time-of-flight mass spectrometry
Department of Chemistry and Chemical Biology, University
of New Mexico, Albuquerque, NM 87131, USA. 5Department would be sufficient to model surface chemis- reports the product’s mass-to-charge ratio (m/Z)
of Chemistry, University of Crete, 71003 Heraklion, Greece. try. But the need to go beyond hTST has been and its density as a function of delay between
6
Institute of Electronic Structure and Laser, FORTH, 71110 pointed out recently (14) and new methods the pulsed molecular and laser beams. Because
Heraklion, Greece.
*Corresponding author. Email: [email protected] (D.B.); were reported, although they also lack valida- the ions were detected with slice imaging
[email protected] (A.M.W.) tion from experiment. Electron spin is another (19, 20) yielding product velocity, we could

accurately compute the transient product

flux as a function of reaction time at the
surface. Initial reactant concentrations are
needed to obtain second-order rate constants
(SM section S2a). These values were obtained
from the absolute flux profiles of the incident
molecular beams (SM section S2b) and known
sticking coefficients (17, 18) (SM section S3).
Finally, VRK data obtained at m/Z = 2, 3, and 4
led to isotopic branching fractions (SM section
S4), from which we obtained isotope-specific
rate constants.
The QRM developed in this work is an exact
formulation of a thermal rate constant. It
yields accurate isotope-specific thermal rates
as long as one has accurate isotope-specific
D E
thermal sticking probabilities S0H2 ;HD;D2 ðT Þ,
adsorption energies E0H2 ;HD;D2 , and reactant
QH ;D and product QH2 ;HD;D2 partition functions:
k H 2 ðT Þ ¼
sffiffiffiffiffiffiffiffiffiffiffiffiffiffi
kB T QH2 =V E0H2
S0H2 ðT Þ exp
2pmH2 ðQH =AÞ2 kB T
ð1Þ
where T is temperature, kB is the Boltzmann

constant, V is reference volume, and A is refer- Fig. 1. VRK of H atom recombination on Pt(111) and Pt(332). Measured HD formation rates for Pt(111)
ence area for the partition function evaluation. (○) and Pt(332) (+) are compared with the results of the QRM (dashed and solid lines). The temperature
The QRM rate constant for H2 desorption by dependence and the transient rate of the measurements are quantitatively captured by the model for both
the recombination of two adsorbed H atoms, facets. The shaded regions of the top three panels indicate 2s uncertainty, mainly associated with the
kH2 ðT Þ, given by Eq. 1, is derived from the absolute reactant flux measurement (~30%) and the dissociative adsorption energies. The excellent
principle of detailed balance provided in SM agreement between VRK and QRM is achieved without adjustable parameters. a.u., arbitrary units.
section S5a. Expressions for kHD ðT Þ and kD2 ðT Þ
were easily obtained by analogy. Accurate values
D E
Figure 1 presents the experimentally ob- temperature programmed desorption (TPD)
for E0H2 ;HD;D2 and S0H2 ;HD;D2 ðT Þ can be ob-
tained HD formation rates for reactions on for T < 400 K (23–26) and molecular beam
tained from prior experiments (SM sections S3 Pt(111) and Pt(332) and compares them with a relaxation spectrometry (MBRS) for T > 400 K
and S6). These measured quantities allowed simulation of the experiment. The simulations (27, 28). The uncertainty in the previously
us to avoid errors associated with the theoret- used rate constants from the QRM for all three reported rate constants spans three orders of
ical determination of the thermal dissociative isotopologs (fig. S9) and accounted for the magnitude. We note that previous work studied
adsorption rates and density functional theory temporal profile of the dosing pulse and its different isotopic recombination reactions
(DFT) calculations of adsorption energies, spatial inhomogeneity, f (t, r) (Fig. 2A), where (fig. S8); however, given the small isotope effect
which can be highly dependent on the choice t is time and r is radial distance, as well as found in the present study (Fig. 2B and fig. S9),
of exchange-correlation functional (21). In ad- reactant diffusion. This aspect of the data these differences between experiments cannot
dition, the partition function for the hydrogen analysis goes beyond past work and is essential explain the large range of reported values.
molecule in the gas phase QH2 is well known. because the rates of second-order reactions The VRK measurements clearly distinguish
The adsorbate partition functions QH ;D are are sensitive to surface concentration dis- the accuracy of two previous MBRS measure-
crucial inputs to the QRM and were computed tributions and gradients. The full diffusion- ments that fall within the uncertainty range
with a quantum potential energy sampling reaction model is described in SM section S7 of (27) but differ by two orders of magnitude
(QPES) method, where the nuclear part of the and accounts for well-known diffusion effects from rate constants reported in (28).
partition function is obtained by a direct state on surface reaction rates discussed in previous A hallmark of a fundamentally correct
count. States and energies were obtained by work (22). Figure 2B shows that the isotope model is its ability to reproduce accurate ex-
solving the nuclear Schrödinger equation with effect at these temperatures is small and well perimental data over a broad temperature
DFT interaction potentials computed with two described by the QRM. Inspection of Figs. range. The QRM uses a fundamentally correct
different functionals and assuming a static Pt 1 and 2B clearly shows that the QRM, which ab initio adsorbate partition function that leads
surface (SM section S1b). This procedure was has no adjustable parameters, reproduces ex- to excellent agreement with experiment over
performed for H interacting with both Pt(111) perimental data for reactions on both Pt(111) a large temperature range. The performance
and Pt(332). We found that QH is weakly depen- and Pt(332). of the QRM for Pt(111) at temperatures between
dent on the choice of DFT functional (SM sec- Figure 3 shows the VRK-derived H* recom- 650 and 950 K is demonstrated by comparison
tion S5e). The electronic contribution to QH , bination rate constants (black circles) compared to VRK-derived rate constants, whereas low-
which accounts for the twofold spin degeneracy with those of previous work (light red trapezoids) temperature comparisons rely on TPD. Un-
of the H-Pt system, was explicitly included. for reactions on Pt(111). Previous studies used certainties in the TPD-derived rate constants

A B
Fig. 2. Calibration of the molecular beam and isotopic branching. (A) The range. The inset shows the temporal profile of the molecular beam pulse.
space-dependent H2 and D2 dosing profiles used to determine the absolute initial (B) Isotopic branching fraction from VRK experiments (symbols) and QRM (lines).
concentration of H* and D*. These results were obtained from laser-based The agreement shows that QRM correctly predicts the isotope effect. The error
calibration of the molecular beam flux and are required to accurately determine bars and the gray-shaded region reflect 2s uncertainty in the experiment and model,
the recombination rate constants. The shaded regions indicate the 2s uncertainty respectively. Note that some symbols have been shifted by ±5 K for clarity.
A B
Fig. 3. Rate constants for H atom recombination on Pt(111). (A) Light red the modeled TPD spectra reflects the uncertainty of the experimental H2
trapezoids show the temperature-range and rate-constant uncertainties of chemisorption energy. The ability of the QRM rate constants to quantitatively
previous work (23Ð28). Shown are experimental results from this work (○) with reproduce experimental data demonstrates the importance of both nuclear
2s error bars compared with the results of the QRM (black solid line), hTST and electronic quantum effects. (B) Comparison of the approximate
(green dotted line), CRM (blue dash-dotted line), and QRM neglecting electron predictions of three rate models to QRM rate constants. Neglecting spin
spin (black dashed line). The inset at the bottom left shows an expanded degeneracy, using a fully classical approximation or a commonly adopted
view. The inset at the top right compares TPD spectra (broad gray lines) from approximate quantum model both introduce large errors even at high
(29) with the predictions of the QRM model, QRM neglecting spin, the CPES temperatures. Similar errors are seen for recombination rates on the stepped
model, and the hTST model for three initial H* coverages of 0.1, 0.2, Pt(332) surface (see fig. S14). See fig. S12 for a detailed decomposition of the
and 0.3 ML. The gray-shaded region and the horizontal error bar on one of errors observed from hTST and adsorbate entropy approximations.
arise from questionable approximations used TPD spectra from (29), where the influence lines of the QRM are in excellent agreement
to derive rate constants from the data, neglect of steps was carefully identified and removed with the TPD spectra [broad gray lines, from
of the coverage dependence of adsorption (SM section S8). Here, we also accounted for (29)] for three initial coverages.
energies (26), dubious estimations of prefac- the previously reported coverage dependence
tors (25), and neglect of the influence of steps of the adsorption energy (24, 30) (fig. S11 and Discussion
(23). To make the most meaningful compar- SM section S6). The comparison is shown in The aforementioned comparisons to kinetics
ison, we used the QRM to directly simulate the top-right inset of Fig. 3A. The solid black experiments carried out between 250 and 950 K

A B
Fig. 4. The influence of steps on H atom recombination on Pt. (A) Rate from CPES (blue dash-dotted line) for H* bound to Pt nanoparticles from (11).
constants derived from VRK experiments (symbols) for H atom recombination on Also shown are QPES entropies for H* bound to Pt(111) (solid black line) and
Pt(111) and Pt(332) are compared with QRM predictions (solid lines). The 2s Pt(332) (solid red line) that were obtained in this work (see SM section S9). The
uncertainty of the rate constants of Pt(332) QRM is shown as a red-shaded nuclear quantum effect contribution is 12 J mol−1 K−1, and the contribution of
region. The inset is a magnification of the area enclosed by the dotted rectangle. electron spin is 6 J mol−1 K−1. The comparison suggests that the nanoparticle-
(B) Entropies obtained experimentally at 598 K (symbols with 2s error bars) and size dependence of the H* entropy is determined by the concentration of steps.
demonstrated the validity of the QRM rate as a green dotted line in Fig. 3. This approxi- the CRM are similar for reactions on Pt(332)
constants over 12 orders of magnitude and for mation overestimates the experimental reac- (fig. S14).
H atom coverages up to 0.3 monolayer (ML). tion rate constant by two to three orders of A major source of error in the CPES method
Within the context of the principle of detailed magnitude at all temperatures between 200 arises from the classical description of the
balance as implemented in the QRM, this and 1200 K. The major source of errors in hTST adsorbate’s in-plane motion. This can be under-
coherent picture demonstrates the quantitative arise from the harmonic simplifications made stood by considering that the in-plane zero-point
consistency of previously reported sticking co- to the H-Pt interaction potential (resulting in energy of H* on Pt(111) (58 meV) is almost
efficients and binding energies with the kinetics errors of a factor 5 to 25) and the neglect of equal to the classical diffusion barrier (60 meV)
measurements of this work. The agreement over recrossing corrections to TST (with errors of (see fig. S13). Thus, classical and quantum de-
such a wide range of rates provides confidence a factor 5 to 10) (see fig. S12A for details). scriptions of H* motion on the surface lead
in the QRM rate constants, making H recombi- The next, more sophisticated level of rate to very different results. CPES excludes H* from
nation on Pt(111) a reliable benchmark for ap- theory uses the complete potential energy classically forbidden regions of space, whereas
proximate rate theories in surface chemistry. sampling (CPES) method to characterize en- quantum mechanically, there is a substantial
It is worth noting that the QRM as imple- tropy associated with the in-plane degrees of probability to populate these regions. Further-
mented in this work is semiempirical because freedom of H*. CPES is considered by many to more, CPES does not account for the uncer-
it relies on experimental values of thermal provide the most accurate adsorbate partition tainty principle, which prevents localization
sticking coefficients and adsorption energies. function (31), and it has been applied to char- of H* at the classical energy minimum at low
However, it also provides a path to an ab initio acterize H interaction at metals (11). It ac- temperature. The surface area explored by the
theory of thermal reaction rates, if these quan- counts for anharmonicity by using a semiclassical H atom is underestimated by CPES and thus
tities can be accurately calculated from first partition function computed from the adsorbate so too is the adsorbate entropy. This results in
principles. potential energy surface, which may be obtained an overestimate of the corresponding rate con-
The framework of the QRM allows us to with DFT (11, 14, 31). To evaluate this approach, stant. Our results underscore the importance of
critically test the quality of predictions based we modified the QRM, replacing the QPES quantum delocalization and help explain why
on approximations that are commonly used by the CPES adsorbate partition function but the deviations of CRM become more severe at
in kinetic modeling of heterogeneous catal- retaining the other parameters in Eq. 1. This low temperatures. Quantum delocalization is
ysis. The results of this analysis are shown in substitution serves to illustrate the classical also the reason why hTST fails. All quantum
Fig. 3. The most widely used model for rate counterpart of the QRM, which we hereafter states above the ground state exhibit prob-
constants (hTST) introduces quantum effects denote as the CRM. The rate constants predicted ability maxima at positions far from the po-
in an approximate way, where nuclear parti- by the CRM are shown as blue dash-dotted lines tential energy minimum (fig. S13).
tion functions are computed assuming separable in Fig. 3. The CRM performed better than hTST We may investigate other sources of error in
motion of contributing degrees of freedom but nevertheless overestimated the rate constant the CRM by using the QPES partition function
that can each be approximated as a harmonic by a factor of 20, even at temperatures as high as but neglecting electron spin. Figure 3A shows
oscillator. By definition, recrossing correc- 1000 K. The error is more than 100-fold at 300 K, rate constants predicted on this basis, and in
tions are not included in hTST (12, 13). The a temperature typical for electrochemical appli- Fig. 3B, one can see that neglect of electron
hTST rate constants, calculated by placing the cations. Although our detailed analysis is focused spin degeneracy led to a 4× overestimate of
dividing surface far above the surface, are shown on Pt(111), the errors introduced by hTST and the rate constant at all temperatures. This

result can be understood intuitively if we con- smaller nanoparticles exhibit higher step con- 12. S. Bhandari, S. Rangarajan, C. T. Maravelias, J. A. Dumesic,
sider that when two H* atoms attempt to react, centrations. Figure 4B shows measured H* M. Mavrikakis, ACS Catal. 10, 4112–4126 (2020).
13. M. Jørgensen, H. Grönbeck, ACS Catal. 6, 6730–6738 (2016).
they must approach one another in one of four entropies for Pt nanoparticles of various sizes 14. A. Bajpai, P. Mehta, K. Frey, A. M. Lehmer, W. F. Schneider, ACS
degenerate states with either parallel (triplet) reproduced from (11)—the entropy increases Catal. 8, 1945–1954 (2018).
or antiparallel (singlet) spins. Only the singlet with the nanoparticle size, consistent with ob- 15. D. Borodin et al., J. Am. Chem. Soc. 143, 18305–18316
(2021).
state correlates with the formation of gas- servations presented above that Pt steps reduce 16. J. Neugebohren et al., Nature 558, 280–283 (2018).
phase singlet H2 products; hence, including spin H* entropy. Also shown are CPES entropies 17. R. van Lent et al., Science 363, 155–157 (2019).
degeneracy reduces the reaction rate by a factor reported in (11), which fail to describe entropies 18. A. C. Luntz, J. K. Brown, M. D. Williams, J. Chem. Phys. 93,
5240–5246 (1990).
of four. This result should not come as a surprise derived from experiment. Notably, the entropies 19. K. Tonokura, T. Suzuki, Chem. Phys. Lett. 224, 1–6 (1994).
to those familiar with rate calculations for gas- found using the QPES method for H* bound 20. C. R. Gebhardt, T. P. Rakitzis, P. C. Samartzis, V. Ladopoulos,
phase reactions where spin degeneracy is rou- to Pt(111) (see Fig. 4B) are in good agreement T. N. Kitsopoulos, Rev. Sci. Instrum. 72, 3848–3853 (2001).
21. G. J. Kroes, Phys. Chem. Chem. Phys. 23, 8962–9048
tinely included (1). Nonetheless, to the best of with entropies for the largest particle sizes.
(2021).
our knowledge, this is the first demonstration Note that surfaces of large nanoparticles are 22. D. R. Olander, J. Colloid Interface Sci. 58, 169–183 (1977).
that the rates of thermal reactions at metal primarily composed of the (111) facets (11). 23. K. Christmann, G. Ertl, T. Pignet, Surf. Sci. 54, 365–392
surfaces depend on adsorbate spin. Figure 4B also shows QPES entropies for H* (1976).
24. B. Poelsema, K. Lenz, G. Comsa, J. Phys. Condens. Matter 22,
The QRM developed in this work also de- on the (332) facet, which compare well to ex- 304006 (2010).
scribes the reaction rate on stepped surfaces. perimentally obtained entropies on small nano- 25. P. Samson, A. Nesbitt, B. E. Koel, A. Hodgson, J. Chem. Phys.
In Fig. 4A, we compare the predicted rate con- particles. This comparison further supports our 109, 3255–3264 (1998).
26. M. J. van der Niet, A. den Dunnen, L. B. Juurlink, M. T. Koper,
stants of the QRM for Pt(111) and Pt(332) sur- hypothesis that the H* entropy decreases as J. Chem. Phys. 132, 174705 (2010).
faces to those derived from VRK experiments. nanoparticle size decreases and the relative 27. G. E. Gdowski, J. A. Fair, R. J. Madix, Surf. Sci. 127, 541–554
For details of the QRM treatment of the re- importance of step defects increases. This re- (1983).
28. M. Salmerón, R. J. Gale, G. A. Somorjai, J. Chem. Phys. 70,
action on Pt(332), see SM sections S1b and S6. sult points out the importance of quantum 2807–2818 (1979).
Experiment shows that near 700 K, the rate effects even for the description of thermody- 29. S. K. Jo, Surf. Sci. 635, 99–107 (2015).
constants for reaction on the (332) facet is namic state functions in advanced catalytic 30. B. J. J. Koeleman, S. T. de Zwart, A. L. Boers, B. Poelsema,
L. K. Verhey, Nucl. Instrum. Methods Phys. Res. 218, 225–229
larger than that on the (111) facet, an effect that materials. (1983).
is quantitatively captured by the QRM. This 31. M. Jørgensen, H. Grönbeck, J. Phys. Chem. C 121, 7199–7207
result may appear surprising, because the H Conclusion (2017).
32. B. Poelsema, G. Mechtersheimer, G. Comsa, Surf. Sci. 111,
atom’s binding energy is larger at steps than As this work has shown, H* recombination on
519–544 (1981).
at terraces (30, 32). A naïve view of Eq. 1 suggests Pt surfaces exhibits large quantum effects even 33. D. Borodin, Hydrogen atom recombination on Pt(111) and
that this leads to a lower rate constant. However, at increased temperatures relevant to cataly- Pt(332). Zenodo (2022); doi:10.5281/zenodo.6565004.
careful analysis of the thermally populated sis. These quantum effects in the reaction rates
AC KNOWLED GME NTS
quantum states used in the QPES partition and in the thermodynamic properties of the
We thank J. C. Tully for helpful discussions. Funding: D.B. and M.S.
functions showed that at these temperatures, adsorbed H atoms arise in part from the H atom’s thank the BENCh graduate school, funded by the DFG
H atoms on the (332) facet tend to remain lo- light mass, where a careful treatment of its wave (389479699/GRK2455). T.N.K., G.S., A.K., M.S., and J.F.
calized near step sites (fig. S15). This fact re- properties is required to obtain accurate results. acknowledge support from the European Research Council under
the European Union’s Horizon 2020 research and innovation
duces their in-plane translational entropy and Such nuclear quantum effects will diminish in program (grant agreement no. 833404). Y.W., J.Z., and H.G.
leads to an increase in the rate constant because importance for heavier adsorbates. However, acknowledge the US National Science Foundation (grant. no. CHE-
the effect of entropy is larger than that produced the effect of spin degeneracy demonstrated 1951328), and H.G. thanks the Alexander von Humboldt Foundation
for a Humboldt Research Award. The calculations were partially
by a larger step binding energy. here will remain of general importance for a performed at the Center for Advanced Research Computing
This observation reflects how changing tem- host of reactions of heavier species involved in (CARC) at the University of New Mexico and at the National Energy
perature alters the relative influence of energy real-world catalysis. At present, it is not easily Research Scientific Computing (NERSC) Center. Author
contributions: D.B., M.S., and J.F. conducted the transient kinetics
and entropy on the rate constant. In past work, possible to determine the lowest-energy spin experiments. Flux calibration procedures were developed by D.B.,
similarities in TPD spectra of H2 desorbing state for metal surfaces with DFT. Developing G.B.P., M.S., F.N., D.J.A., and T.N.K. D.B. and D.S. developed the
from Pt(111) and a B-type stepped Pt surface theoretical and experimental methods that are reaction-diffusion analysis. D.B. and A.M.W. developed the
quantum rate model. N.H., Y.W., J.Z., and H.G. conducted DFT
at T ~ 350 K were taken as evidence for a lack able to probe the general influence of spin on calculations. D.B., N.H., A.K., and H.G. analyzed DFT calculations
of preferential step binding (26). Inspection reaction rates presents the next challenge on the and developed methods for description of nuclear partition
of QRM rate constants in Fig. 4A reveals that way toward fully predictive surface chemistry at functions. M.S., J.F., G.S., T.N.K., D.J.A., D.S., and A.K. participated
in discussion of the results. D.B., D.J.A., H.G., and A.M.W. wrote
at 350 K, the similarity in desorption rate metal catalysts.
the manuscript and the supporting material. All authors
constants arises from compensation between contributed to the reviews of the manuscript and the supporting
energetic and entropic contributions (see SM RE FERENCES AND NOTES material. Competing interests: None declared. Data and
section S8 for details). Our work supports material availability: All data needed to evaluate the conclusions
1. J. L. Bao, D. G. Truhlar, Chem. Soc. Rev. 46, 7548–7596
in the paper are present in the paper or the supplementary
conclusions derived from He-atom and ion- (2017).
materials and are publicly available in the Zenodo repository
2. F. Studt, Front. Catal. 1, 658965 (2021).
scattering experiments that H binds more 3. A. H. Motagamwala, J. A. Dumesic, Chem. Rev. 121, 1049–1076
(33). License information: Copyright © 2022 the authors, some
strongly to B-type steps (30, 32). Only at much (2021).
rights reserved; exclusive licensee American Association for the
Advancement of Science. No claim to original US government
lower temperatures does the energetic prefer- 4. J. E. Sutton, W. Guo, M. A. Katsoulakis, D. G. Vlachos, Nat.
works. https://www.science.org/about/science-licenses-journal-
Chem. 8, 331–337 (2016).
ence for H binding at steps cause the rate 5. G. B. Park et al., Nat. Rev. Chem. 3, 723–732 (2019).
article-reuse
constant on the stepped surface to drop below 6. P. Sabatier, Catalysis in Organic Chemistry (D. Van Nostrand
that on (111) terraces. Company, 1922). SUPPLEMENTARY MATERIALS
7. G. Ertl, Angew. Chem. Int. Ed. 47, 3524–3535 (2008).
Because the QRM approach developed in science.org/doi/10.1126/science.abq1414
8. A. Godula-Jopek, Hydrogen Production by Electrolysis (Wiley-VCH,
this work provides an accurate determination Materials and Methods
2015).
Supplementary Text
of rate constants on stepped surfaces, we expect 9. M. P. D’Evelyn, R. J. Madix, Surf. Sci. Rep. 3, 413–495 (1983).
Figs. S1 to S15
that QPES entropies used in the QRM would 10. Y. V. Suleimanov, F. J. Aoiz, H. Guo, J. Phys. Chem. A 120,
References (34Ð71)
8488–8502 (2016).
help us understand experiments performed 11. M. García-Diéguez, D. D. Hibbitts, E. Iglesia, J. Phys. Chem. C Submitted 21 March 2022; accepted 16 June 2022
on size-selected Pt nanoparticles (11), because 123, 8447–8462 (2019). 10.1126/science.abq1414

EVOLUTION born mice raised in common conditions (fig.

S2 and table S1), which suggests a strong genetic
A chromosomal inversion contributes to divergence component to these ecotype-defining traits.
in multiple traits between deer mouse ecotypes A large inversion is associated with tail length
and coat color
Emily R. Hager1†‡, Olivia S. Harringmeyer1†, T. Brock Wooldridge1, Shunn Theingi1, Using an unbiased forward-genetic approach,
Jacob T. Gable1, Sade McFadden1, Beverly Neugeboren1, Kyle M. Turner1§, we identified genomic regions linked to ecotype
Jeffrey D. Jensen2, Hopi E. Hoekstra1* differences in morphology. We intercrossed
forest and prairie mice in the laboratory to gen-
How locally adapted ecotypes are established and maintained within a species is a long-standing erate 555 second-generation (F2) hybrids (forest
question in evolutionary biology. Using forest and prairie ecotypes of deer mice (Peromyscus female × prairie male, n = 203 F2s; prairie
maniculatus), we characterized the genetic basis of variation in two defining traits—tail length and female × forest male, n = 352 F2s) and per-
coat color—and discovered a 41-megabase chromosomal inversion linked to both. The inversion formed quantitative trait locus (QTL) mapping
frequency is 90% in the dark, long-tailed forest ecotype; decreases across a habitat transition; for each trait (12) (Fig. 2, fig. S3, and table S2).
and is absent from the light, short-tailed prairie ecotype. We implicate divergent selection in We identified five regions associated with tail
maintaining the inversion at frequencies observed in the wild, despite high levels of gene flow, length variation [total percent variance ex-
and explore fitness benefits that arise from suppressed recombination within the inversion. plained (PVE): 27%; individual PVE: 2.6 to 12.1%].
We uncover a key role for a large, previously uncharacterized inversion in the evolution and Only one region, on chromosome 15, was strong-
maintenance of classic mammalian ecotypes. ly and significantly associated with coat color
variation (PVE, dorsal hue: 40.0%; PVE, flank
W
hue: 45.6%). Each QTL exhibited incomplete
ide-ranging species that occupy diverse One of the most abundant and widespread dominance, and the forest allele was always
habitats often evolve distinct ecotypes— mammals in North America is the deer mouse associated with forest traits—longer tails or
intraspecific forms that differ in her- (Peromyscus maniculatus), which is continu- redder coats. The one significant QTL for coat
itable traits relevant to their local ously distributed across diverse habitats from color overlapped with the largest-effect locus
environments (1). Ecotypes frequently the Arctic Circle to central Mexico. In the early associated with tail length (95% Bayesian
differ in multiple locally adaptive phenotypes 1900s, a taxonomic revision of this species credible intervals: dorsal hue = 0.4 to 40.5 Mb;
(2), and although ecotypes sometimes show described two distinct ecotypes: a forest and a flank hue = 0.4 to 39.4 Mb; tail length = 0.4
partial reproductive isolation (2), many expe- prairie form (7). Several features distinguish to 41.5 Mb). Thus, a single region on chro-
rience substantial intraspecific gene flow (3). the semiarboreal forest mice that occupy dark- mosome 15 was strongly associated with
This raises an important question: How are soil habitats from their more terrestrial prairie ecotype differences in both tail length and
differences in multiple traits maintained be- counterparts that occupy light substrates. coat color.
tween ecotypes when migration acts as a Most notably, forest mice typically have longer The QTL peak on chromosome 15 exhibited
homogenizing force? tails and darker coats than those of prairie a consistently strong association with both
One explanation is that natural selection mice (7–9), with large differences in these morphological traits across half the chromo-
keeps each locus associated with locally adaptive traits maintained between ecotypes despite some (Fig. 3A). This pattern reflects reduced
trait variation at migration-selection equilib- evidence for gene flow (10, 11). This consistent recombination between forest and prairie
rium (4). However, in cases of high migration, divergence in multiple traits provides an op- alleles in the laboratory cross: Only 2 of 1110
this requires strong selection acting on many portunity to test the mechanisms that estab- F2 chromosomes were recombinant in this
independent alleles. Linkage disequilibrium lish and maintain ecotypes. region (Fig. 3B). We also found consistently
can play an important role by allowing linked elevated FST (proportion of the total genetic
loci, each with potentially weaker selective ef- Forest and prairie mice differ in multiple traits variance explained by population structure)
fects, to establish and be maintained together To study divergence between the forest (Fig. 3C) and high linkage disequilibrium (Fig.
(5), which can lead to concentrated genetic and prairie ecotypes, we selected two focal 3D) across this genetic region in wild popu-
architectures of ecotype-specific traits (6). Char- populations—one from a coastal temperate lations relative to the rest of the chromosome
acterizing the genetic basis of the full set of rainforest (P. m. rubidus, referred to hereafter (whole-genome resequencing: n = 15 forest,
ecotypic differences and the role of migration, as the forest ecotype) and one from an arid n = 15 prairie). Together, these data are con-
selection, and recombination in maintaining sagebrush steppe habitat (P. m. gambelii, ref- sistent with reduced recombination across half
these differences is thus critical to understand- erred to as the prairie ecotype) in the north- of chromosome 15 in both laboratory and wild
ing local adaptation specifically and biological western US—separated by ~500 km (Fig. 1A). populations.
diversification more generally. After establishing laboratory colonies from This pattern of suppressed recombination
wild-caught mice, we measured both the could be produced by a large genomic rear-
wild-caught mice and their laboratory-reared rangement (or a set of rearrangements). To
1
descendants for four traits previously reported determine the nature of any structural varia-
Department of Molecular and Cellular Biology, Department
of Organismic and Evolutionary Biology, Museum of
to distinguish forest and prairie ecotypes (7–9): tion on chromosome 15, we used PacBio long-
Comparative Zoology, and Howard Hughes Medical tail, hindfoot, and ear lengths as well as coat read sequencing (n = 1 forest, n = 1 prairie) (12).
Institute, Harvard University, Cambridge, MA 02138, USA.
2
color (brightness, hue, and saturation across We generated independent de novo assemblies
School of Life Sciences, Arizona State University, Tempe,
three body regions). We also measured body for each individual and mapped the resulting
AZ 85287, USA.
*Corresponding author. Email: [email protected] length and weight. We found that forest mice contigs to the reference genome for P. m. bairdii
†These authors contributed equally to this work. ‡Present had longer tails; longer hind feet; and darker, (12). In the forest individual, one contig mapped
address: Department of Biomedical Engineering, Boston redder coats compared with prairie mice near the center of the chromosome (from 41.19
University, Boston, MA 02215, USA. §Present address: Centre
for Teaching Support & Innovation, University of Toronto, (Fig. 1, B and C; fig. S1; and table S1). These to 40.94 Mb) and then split and mapped in
Toronto, ON M5S 3H1, Canada. phenotypic differences persisted in laboratory- reverse orientation to the beginning of the

A forest B Tail C Hue

Body Tail + red ***
length length
40
90 ***
ns ***
P. m. rubidus
degrees
length (mm)
~500 km 60 45
prairie
30 forest DF V
50
forest
0 prairie + yellow prairie

P. m. gambelii dorsal flank
Fig. 1. Forest and prairie mice differ in tail length and pigmentation. bar, 1 cm. (C) Coat color (hue) values for the dorsal and flank regions of wild-
(A) Map shows the approximate range of forest (green) and prairie (brown) deer caught adult mice (n = 16 forest and 20 prairie). Boxplots indicate the median
mouse ecotypes in North America. Collection sites of wild-caught forest (P. m. (center white line) and the 25th and 75th percentiles (box extents); whiskers
rubidus, green) and prairie (P. m. gambelii, brown) ecotypes from western and show largest or smallest value within 1.5 times the interquartile range. Black dots
eastern Oregon, USA, respectively, are shown. Photos illustrate representative show individual data points. (Inset) Dorsal (D), flank (F), and ventral (V) regions
habitat; pink flags indicate trap lines. (B) Body length (left; not including the tail) from a representative forest and prairie mouse. ns = P > 0.05; ***P < 0.001
and tail length (right) for wild-caught adult mice (n = 38 forest and 32 prairie). (WelchÕs t test, two-sided). Original photography in (B) and (C) is copyrighted
Lines connect body and tail measurements for the same individual. Means are by the President and Fellows of Harvard College (photo credit: Museum of
shown in bold. (Inset) Image of a representative tail from each ecotype. Scale Comparative Zoology, Harvard University).
chromosome (from 0 to 5 Mb). By contrast, in in laboratory mice (three with nonsynonymous The inversion changes substantially in fre-
the prairie individual, a single contig mapped substitutions and four with associated pigment quency across the habitat transition, from 90%
continuously to the reference genome in this phenotypes as well; table S4). These 19 genes in the forest population to absent in the prairie
region (37 to 41.3 Mb) (Fig. 3E). Because we are thus strong candidates for contributing to population (Fig. 4E). This frequency difference
found no other forest-specific rearrangements tail length and coat color variation. of the inversion is extreme: It is greater than
in this region (fig. S4), we determined that the allele frequency difference at the maximally
chromosome 15 harbors a simple 41-Mb in- Inversion frequency and divergence in differentiated single-nucleotide polymorphism
version. Using putative centromere-associated wild populations (SNP) in 99.92% of blocks with similar levels of
sequences in Peromyscus (12), we determined To investigate whether the inversion and as- linkage disequilibrium (12) (Fig. 4F). Moreover,
that the inversion is paracentric, with the sociated traits (longer tails and redder coats) similar to the changes in phenotype, the
centromere located outside of the inversion may be favored in forested habitats, we col- transition in inversion frequency occurs over
(Fig. 3G). lected deer mice across a sharp habitat transi- only a short distance: Inversion frequency
Inversions may affect phenotypes directly tion between the focal forest and prairie sites decreases from 100 to 62.5% in the 50-km
through the effects of their breakpoints or and estimated habitat type and mean soil hue Cascades region and then drops further within
indirectly by carrying causal mutations (13). at each capture site (n = 136 mice from 22 sites, the next 100 km (i.e., inversion frequency drops
Using the long-read sequencing data, we supplemented by 12 additional museum speci- from 100 to 4% over less than one-third of
localized the inversion breakpoint to base pair mens from two sites; figs. S6 and S7). We found the total transect distance; Fig. 4E). The sharp
resolution (Fig. 3F and fig. S5). The breakpoint that much of the transition in both habitat change in inversion frequency across the envi-
falls within an intron of a long intergenic type and soil hue occurs in a narrow region ronmental transect, and its extreme forest-
noncoding RNA (lincRNA), and an additional across the Cascade mountain range (Fig. 4, A prairie allele frequency difference, suggest
four annotated genes (two lincRNAs and two and B), and the phenotypic clines estimated that the inversion may be favored in forested
protein-coding genes) occur within 200 kb of using either all adult wild-caught individuals habitat.
the breakpoint. Although the breakpoint may or only those from the Cascades region both The inversion also strongly contributes to
disrupt their expression patterns, these genes identified sharp transitions in coat color and genetic differentiation between the forest and
have no known functions associated with tail length that colocalize with this environ- prairie ecotypes by carrying many highly dif-
either pigmentation or skeletal phenotypes mental transition (Fig. 4, C and D). Specifically, ferentiated SNPs. For example, FST between
(table S3). An additional 149 protein-coding mean hue changes by 3.2° (63% of the forest- the forest and prairie ecotypes in the inversion
genes are located within the inversion, of which prairie difference), and mean tail length changes region is high compared with the genome-
29 contain at least one fixed nonsynonymous by 13 mm (47% of the forest-prairie difference) wide average (inversion region: mean FST =
mutation between the inversion and reference across the 50-km Cascades region; tail length 0.376; genome-wide, excluding inversion region:
alleles. Ten of the genes within the inversion changes by an additional 4 mm within the next mean FST = 0.071; fig. S8). The strong genetic
(four with nonsynonymous substitutions) are 100 km, coincident with continued changes divergence between the inversion and reference
associated with pigmentation phenotypes when in forestation (Fig. 4). Together, the strong haplotypes is reflected in maximum likelihood–
disrupted in laboratory mice, and 13 are as- correlation between phenotype and habitat is based trees built from the region of chromosome
sociated with tail or long-bone length phenotypes consistent with local adaptation. 15 that contains the inversion (affected region:

A
15 tail length
10
LOD
5
0
70
flank hue
dorsal hue
50
LOD
30
10
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 X
chromosome
B
Tail length (resid. body) Dorsal hue Flank hue
20 PVE: 12% 35 PVE: 40% PVE: 46%
a: 2.7 mm a: -2.0° a: -1.9°
d/a: 0.19 d/a: 0.26 40 d/a: 0.30
millimeters
degrees
degrees
40
0
45
45
-20
50
f/f f/p p/p f/f f/p p/p f/f f/p p/p
n= 144 279 119 142 279 120 142 279 120
genotype at chromosome 15: 20 Mb
Fig. 2. A region on chromosome 15 is strongly associated with both tail with significant QTL peaks. For tail length analysis, body length was included
length and coat color. (A) Statistical association [log of the odds (LOD) as an additive covariate. (B) Tail length (left; shown after taking the residual
score] of ancestry with tail length (top; blue) and dorsal and flank hue (bottom; against body length in the hybrids), dorsal hue (center), and flank hue (right) of
dorsal, dark red; flank, light red) in laboratory-reared F2 hybrids (tail, n = 542; F2 hybrids, binned by genotype at 20 Mb on chromosome 15 (f/f, homozygous
hue, n = 541). Physical distance (in base pairs) is shown on the x axis; axis labels forest; f/p, heterozygous; p/p, homozygous prairie) (sample sizes are given
indicate the center of each chromosome. Dotted lines indicate the genome- below the x axes). Points and error bars show means ± standard deviations.
wide significance threshold (a = 0.05) based on permutation tests, and shaded PVE, percent of the variance explained by genotype; a, additive effect of one
rectangles indicate the 95% Bayesian credible intervals for all chromosomes forest allele; d/a, absolute value of the dominance ratio.
0 to 40.9 Mb) and the rest of the chromosome migration: initial migration rates of 8.3 × 10−7 mice have mixed forest and prairie ancestry
(unaffected region: 40.9 to 79 Mb). In the unaf- [prairie-to-forest, 95% confidence interval (CI) = genome-wide (fig. S10).
fected region, forest and prairie mice cluster 3.7 × 10−9 to 1.8 × 10−6] and 3.6 × 10−6 (forest- These high migration estimates coupled
by ecotype, with limited divergence between to-prairie, 95% CI = 1.1 × 10−8 to 4.5 × 10−6) after with the large, habitat-associated differences
the groups (Fig. 4G). By contrast, in the affected a forest-prairie population split 2.2 million in inversion frequency may indicate a history
region, mice cluster into two highly distinct generations ago (95% CI = 1.1 to 5.5 million of natural selection. To test this hypothesis, we
groups on the basis of genotypes at the in- generations) (Fig. 5A and fig. S9). Because simulated the spread of the inversion under
version (Fig. 4H). This pattern suggests that the estimated effective population sizes (Ne) our demographic model using SLiM (12). We
the inversion harbors a high density of sites are large (prairie Ne = 1.9 × 106 to 4.3 × 106; found that divergent selection was the most
that are divergent between ecotypes. forest Ne = 1.8 × 105 to 1.2 × 106), the effective likely scenario to explain both the high fre-
number of migrants per generation (Nem) is quency of the inversion in the forest and its
Evolutionary history of the inversion consistently high over time: Nem = 3.5 (prairie- low frequency in the prairie (fig. S11). Using
To explore the evolutionary history of the in- to-forest) and Nem = 0.6 (forest-to-prairie), with approximate Bayesian computation, we esti-
version, we first estimated a best-fitting dem- a recent shift to Nem > 10 in both directions mated selection coefficients (s) for the inversion
ographic model for the forest and prairie ~30,000 generations ago (Fig. 5A), consistent of 3.3 × 10−4 (95% CI = 9.2 × 10−5 to 1.6 × 10−3)
populations using neutral sites across the with high levels of gene flow (15). High migra- in the forest population and −4.1 × 10−3 (95%
genome to avoid the confounding effects of tion levels between forest and prairie ecotypes CI = −9.3 × 10−3 to −7.1 × 10−4) in the prairie
background selection (12, 14). The data were are further supported by genomic data from population (Fig. 5B). These values suggest that
best fit by a model with a long history of high the Cascades region: We found that the Cascades the observed distribution of the inversion in

A Trait association E forest long-read contig

70 flank hue
dorsal hue
tail length
50
LOD
30
chromosome 15
10 (P. man. reference)
B Recomb. events prairie long-read contig
F * forest
count
20
breakpoint
* prairie
0 28 kb 52 kb
C Allele freq. diverg. *

*
0.6
0.4
FST
*
0.2
*
breakpoint
0.0
0 40 80
position (Mb)
D G reference (prairie) allele centromere

Linkage diseq.
inverted (forest) allele

R2 1
affected region unaffected region
0.5
0 0 20 40 60 80
position (Mb)
chromosome 15 chromosome 15
Fig. 3. Chromosomal region associated with tail length and coat color is a sequencing for one forest (top) and one prairie (bottom) mouse. Only contigs
large inversion. Across chromosome 15, data are from F2 hybrids [(A) and that span the inversion breakpoint are shown. The region of chromosome
(B)] and wild-caught mice [(C) and (D), (n = 15 forest and 15 prairie)]. (A) LOD 15 affected by the inversion is highlighted (purple). (F) (Top) Alignment
score for tail length (blue), dorsal hue (dark red), and flank hue (light red). between regions of the forest and prairie contigs surrounding the breakpoint
(B) Number of recombination breakpoint events, binned in 1-Mb windows. (C) FST (black, alignment quality; green, forest contig; brown, prairie contig). Large
between forest and prairie mice estimated in 10-kb windows with a step prairie insertion near the breakpoint is a transposon. (Bottom) Base pairÐlevel
size of 1 kb (light gray dots). Dark gray line shows data smoothed with a alignment around the breakpoint (gray, mismatch). (G) Model of the inverted
moving average over 500 windows. (D) Linkage disequilibrium across forest (green) and reference (tan) alleles. The inversion spans 0 to 40.9 Mb (affected
and prairie mice. Heatmap shows R2 (squared correlation) computed between region, purple) and excludes 40.9 to 79 Mb (unaffected region, gray), with
genotypes at thinned SNPs (12). (E) Contigs assembled from long-read predicted centromere location shown in black.
the wild is best explained by both positive ations or 50,000 to 128,000 years, assuming Although it is formally possible that the in-
selection in the forest and negative selection three generations per year), which suggests version carries only a single mutation that alone
in the prairie, a conclusion robust to the un- that the inversion predates the modern habitat confers a strong enough benefit (s ≥ 3 × 10−4)
certainty in the model parameter estimates distribution (16) (Fig. 5C). Together, these re- to explain its current distribution, an alternative
(fig. S12) and to variation in the timing of the sults suggest that the inversion was most hypothesis is that the inversion carries two or
introduction of the inversion after the forest- likely established in the forest population more beneficial mutations (e.g., one mutation
prairie split (fig. S13). We also used simu- under strong divergent selection over the last that contributes to tail length and a second
lations to assess the minimum age of the ~250,000 generations. to color variation), each with smaller selection
inversion required to achieve its divergence Our estimates of forest-prairie migration coefficients. In this scenario, theory predicts
from the reference allele (12): We estimated rates and selection on the inversion allowed that the inversion could confer a fitness ad-
the inversion to be at least 247,000 genera- us to explore possible fitness effects from the vantage in the forest beyond the individual
tions old (95% CI = 149,000 to 384,000 gener- inversion’s suppression of recombination. mutations it carries by reducing the migration

A F Forest-prairie
soil hue Habitat and soil color allele frequency diff.
40 45 50 a b c d e f x105
Oregon inv.
number of SNPs
9
habitat:
6
forest non-forest Elevation
km 3
douglas fir sagebrush
silver fir 1 0
0 0 0.25 0.5 0.75 1
mixed conifer -200 -100 0 100 200 300 allele frequency difference
ponderosa transect dist. (km)
B C G chr 15: unaffected
Dorsal hue
Sites hab. soil
a 35
30
40
hue (degrees)
35
45
-20 0 20
40
b
45
-200 -100 0 100 200 300
transect dist. (km)
D H chr 15: affected
Tail length
c
90
110
80
70
length (mm)
90
-20 0 20
70
-200 -100 0 100 200 300

transect dist. (km)
e E Chr 15 inv. allele freq.

1
1
0.5
0.75
frequency
0
0.5 -20 0 20
f
0.25
scale: 0.01
0
ecotype inv. genotype
-200 -100 0 100 200 300 forest inv/inv
transect dist. (km) prairie ref/ref
Fig. 4. Associations between genotype, phenotype, and environment in CIs. Insets show best-fit clines using only data from the central Cascades (hue,
wild mice. (A) Elevation and habitat characteristics (top row indicates majority n = 90; tail, n = 97; genotype, n = 136). (F) Allele frequency differences for the
habitat category, and bottom row indicates mean soil hue) at sites across an maximally differentiated SNP between forest and prairie mice in 200-bp windows
environmental transect. Letters indicate sites shown in (B). Soil hue and habitat across the genome (12). The inversion forest-prairie allele frequency difference
category were estimated within 1 km of each site. (Map) Sampled sites across (90%) is shown in black. (G and H) Maximum likelihood trees for unaffected (G)
Oregon. Transect distance refers to the east-west distance from the highest- (40.9 to 79 Mb) and affected (H) (0 to 40.9 Mb) regions of chromosome 15,
elevation site, and dotted lines in (C), (D), and (E) indicate distance = 0. (B) Photos shown on the same scale. Branch colors indicate ecotype (green, forest; brown,
of capture sites from each habitat type, with habitat and soil classification prairie), and dots indicate inversion genotype (tan, homozygous reference,
as in (A). (C to E) Best-fit clines for dorsal hue (C) (n = 143), tail length (D) n = 15; green, homozygous inversion, n = 14; heterozygous mouse excluded, n = 1).
(n = 180), and inversion genotype (E) (n = 178) fit to the full dataset, with 95% Red arrows highlight the forest mouse homozygous for the reference allele.

Fig. 5. Evolutionary history of the inversion. A Demographic B Forest +sinversion

(A) Best-fit demographic model. Ne, effective model
population size; m, migration rate. (B) Posterior
probability distributions for the selection 2
coefficient associated with the inversion in the
density
forest (top, green) and prairie (bottom, brown)
1
populations, when the inversion is introduced
Divergence time
150,000 generations ago (for additional ~2.2e6 generations
introduction times, see fig. S13). The estimated
Population size 0
selection coefficient is positive in forest and −6 −5 −4 −3 −2
104
negative in prairie. (C) Posterior probability 105 log10(+forest sinversion)
distribution for the age of the inversion. 106 Prairie -sinversion
(D) Estimated fitness effects of suppressed Nem ~ 3.53
recombination within the inversion. Two beneficial Migration rate shift
loci (A and B) were introduced into the forest ~3e4 generations
Nem ~ 0.64 2
population on the inversion or on a standard
density
Nem ~ 12.14
haplotype, varying the ratio of the selection
coefficients for A (sA) and B (sB), with sA + sB kept 1
Nem ~ 11.07
constant at 3 × 10−4. bp, base pairs. Bar height
shows the difference in final mean fitness of 0
the forest population between the inversion and
Forest Prairie −6 −5 −4 −3 −2
standard haplotype scenarios. Asterisks indicate a log10(−prairie sinversion)
significant difference in mean fitness (P < 0.05)
C Age of inversion D Effects of suppressed recombination
computed with permutation tests. (Left) Two
fitnessinversion- fitnessstandard
fitnessinversion- fitnessstandard
beneficial loci at varying distances apart, without 4e-6 Ratio sA/sB
* * 4e-6 * * *
* *
deleterious mutations. (Right) Two beneficial 6e-6 0.01 *
0.1 *
loci separated by 100 kb, with deleterious 2e-6
density
3e-6
0.5 * * *
mutations introduced according to distributions 4e-6 1.0
* *
2e-6 0
of fitness effects (DFE): f0: 100% of mutations *
*
neutral (2Ns = 0, where N indicates population * Ratio s /s
A B
2e-6 1e-6 **
-2e-6 0.1
size and s indicates selection coefficient); f1: *
1.0
50% of mutations neutral (2Ns = 0), 50% weakly 0
0 -4e-6
deleterious (−10 < 2Ns < −1); f2: 33% of mutations 0 2e5 4e5 6e5 1e2 1e3 1e4 1e5 1e6 1e7 f0 f1 f2 f3
neutral (2Ns = 0), 33% weakly deleterious age (generations) distance between A & B (bp) DFE
(−10 < 2Ns < −1), 33% moderately deleterious
(−100 < 2Ns < −10); f4: 25% of mutations neutral
(2Ns = 0), 25% weakly deleterious (−10 < 2Ns < −1), 25% moderately deleterious (−100 < 2Ns < −10), 25% strongly deleterious (−1000 < 2Ns < −100).
load suffered by each mutation (5, 17, 18). in (14)] into the two–beneficial locus simu- (8, 9, 23): Long tails have repeatedly evolved in
To investigate this possibility, we used our lations. With weakly or moderately deleteri- association with forest habitat in deer mice (20)
estimates of migration, selection, and recom- ous mutations, the inversion maintained its and across mammals (24), and forest mice are
bination to simulate the spread of two bene- selective advantage over the standard haplotype better climbers (23), with tail length differences
ficial mutations in the forest population either in the forest (Fig. 5D and fig. S16). Only when between the ecotypes likely sufficient to affect
within an inversion or on a freely recombining strongly deleterious mutations were introduced climbing performance (25). Coat color is subject
(standard) haplotype, varying the distance did the inversion accumulate a substantial to pressure from visually hunting predators (19),
between the mutations (12). We found that if mutational load, which results in the inversion and many mammals, including deer mice, evolve
the two mutations are at least 10 kb apart being disadvantageous relative to the standard coats to match local soil color (9, 26). By sam-
(which is likely, given the inversion size of 41 Mb) haplotype in the forest (Fig. 5D and fig. S16). pling along an environmental transect, we found
and the selection coefficient for the weaker locus Thus, our results suggest that, under a wide evidence that each of these traits is closely as-
is at least 10% of that of the stronger locus range of conditions, if this inversion carries sociated with habitat (forestation for tail length
[which is possible, given independent evi- two or more beneficial mutations, its suppres- and soil hue for coat color), which further suggests
dence for selection acting on coat color and sion of recombination likely confers an additional that these traits are involved in local adaptation.
tail length—e.g., (19, 20)], the beneficial muta- selective advantage in the forest population High migration rates between the forest and
tions are more likely to establish and be by linking adaptive alleles in the face of high prairie ecotypes, as we estimated in this work,
maintained at higher frequencies in the forest migration rates. makes the strong ecotypic divergence in mul-
when carried by the inversion than on the tiple traits puzzling. By characterizing the
standard haplotype (Fig. 5D and figs. S14 and Discussion genetic architecture of tail length and coat
S15). We also explored possible costs associ- In 1909, Wilfred Osgood described several color variation, we help resolve how differ-
ated with the inversion suppressing recombina- morphological differences—including tail length ences in these traits are maintained between
tion (i.e., mutational load accumulation) (21, 22) and coat color—that distinguish forest and ecotypes: Namely, we discover a previously
by introducing deleterious mutations according prairie ecotypes of P. maniculatus (7). Long unknown inversion, involving half a chromosome,
to four fitness-effect distributions [as described tails are thought to be beneficial for arboreality that has a large effect on both ecotype-defining

traits and in the expected direction (i.e., it is inversions is limited [(53), but see (54)]. Our 52. I. Sanchez-Donoso et al., Curr. Biol. 32, 462–469.e6
associated with long tails and reddish fur in results thus underscore the important and (2022).
53. G. Dobigny, J. Britton-Davidian, T. J. Robinson, Biol. Rev. 92,
forest mice). Because recombination between perhaps widespread role of inversions in local 1–21 (2017).
the inversion and the noninverted prairie haplo- adaptation, including in mammals, and highlight 54. H. Stefansson et al., Nat. Genet. 37, 129–137 (2005).
type is suppressed in heterozygotes, the in- how selection acting on inversion polymor- 55. E. R. Hager, O. S. Harringmeyer, oharring/chr15_inversion:
v1.0.0, version 1.0.0, Zenodo (2022); https://doi.org/10.5281/
version ensures that longer tail length and phisms may maintain intraspecific divergence zenodo.6354918.
redder coat color alleles are coinherited in the in multiple traits in the wild.
forest, despite high levels of gene flow (except AC KNOWLED GME NTS
in the unlikely scenario that only a single We thank C. Lewarch for assistance in the field; S. Niemi,
RE FERENCES AND NOTES M. Streisfeld, S. Stankowski, G. Binford, S. Bishop, K. Saunders,
pleiotropic mutation within the inversion af- S. Finch, and T. Schaller for help with field logistics;
1. G. Turesson, Hereditas 3, 100–113 (1922).
fects both traits). The role of this inversion in K. Pritchett-Corning and S. Griggs-Collette for help establishing
2. D. B. Lowry, Biol. J. Linn. Soc. 106, 241–257 (2012).
phenotypically differentiating these ecotypes breeding colonies; Harvard’s Office of Animal Resources for
3. A. Tigano, V. L. Friesen, Mol. Ecol. 25, 2144–2164
animal care; and M. Omura, J. Chupasko, M. Mullon, and
is consistent with theoretical predictions and (2016).
J. Mewherter for help preparing and accessioning museum
empirical examples of concentrated genetic 4. J. B. S. Haldane, Math. Proc. Camb. Philos. Soc. 26, 220–230
specimens. G. J. Kenagy, D. S. Yang, and E. Kingsley provided
(1930).
architectures arising under local adaptation 5. R. Bürger, A. Akerman, Theor. Popul. Biol. 80, 272–288
advice at the start of the project; P. Audano provided advice on
with gene flow (6, 27, 28). (2011).
long-read sequencing and analysis; and N. Edelman, A. Kautt,
E. Kingsley, and three anonymous reviewers provided helpful
Our modeling implicates divergent selection 6. S. Yeaman, M. C. Whitlock, Evolution 65, 1897–1911
feedback on the manuscript. The University of Washington
(2011).
in maintaining the inversion at high frequency 7. W. H. Osgood, N. Am. Fauna 28, 1–285 (1909).
Burke Museum provided specimens used in this study. Funding:
in the forest ecotype and absent from the prairie This work was partially funded by Putnam Expedition grants
8. W. F. Blair, Evolution 4, 253–275 (1950).
from the Museum of Comparative Zoology (MCZ) to E.R.H., a
ecotype. The inversion’s selective effects are 9. L. R. Dice, Am. Nat. 74, 212–221 (1940).
MCZ grant-in-aid of student research to S.T., and the MCZ
10. S. W. Calhoun, I. F. Greenbaum, K. P. Fuxa, J. Mammal. 69,
likely driven by its strong association with tail 34–45 (1988).
Chapman Fellowship for the Study of Vertebrate Locomotion
length and coat color (explaining 12 and 40% to E.R.H. and J.T.G. as well as funding from the American
11. D. S. Yang, G. Kenagy, Ecol. Evol. 1, 26–36 (2011).
Society of Mammalogists grants-in-aid of research to E.R.H.
of the trait variances, respectively), although it 12. Materials and methods are available as supplementary
and O.S.H., the Harvard College Research Program to S.T.,
materials.
is possible other traits are involved. Although 13. R. Villoutreix et al., Mol. Ecol. 30, 2738–2755 (2021).
and a Society for the Study of Evolution R. C. Lewontin Early
inversions can have phenotypic effects be- Award to O.S.H. E.R.H. was supported by an NIH training
14. P. Johri et al., Mol. Biol. Evol. 38, 2986–3003 (2021).
grant to Harvard’s Molecules, Cells, and Organisms graduate
cause of their breakpoints disrupting genes 15. S. Wright, Genetics 16, 97–159 (1931).
program (NIH NIGMS T32GM007598) and by the Theodore
16. G. Hewitt, Nature 405, 907–913 (2000).
or gene expression (13), the inversion’s break- H. Ashford Fellowship. O.S.H. was supported by a National
17. M. Kirkpatrick, N. Barton, Genetics 173, 419–434 (2006).
point does not occur in or near candidate Science Foundation (NSF) Graduate Research Fellowship, a
18. B. Charlesworth, N. H. Barton, Genetics 208, 377–382
Harvard Quantitative Biology Student Fellowship (DMS
genes for tail length and coat color variation. (2018).
1764269), and the Molecular Biophysics training grant (NIH
Alternatively, inversions may influence phe- 19. S. N. Vignieri, J. G. Larson, H. E. Hoekstra, Evolution 64,
NIGMS T32GM008313). J.D.J. was funded by the National
2153–2158 (2010).
notypes through the mutations they carry: The 20. E. P. Kingsley, K. M. Kozak, S. P. Pfeifer, D. S. Yang,
Institutes of Health (1R35GM139383-01). H.E.H. is an
investigator of the Howard Hughes Medical Institute. Author
inversion is highly differentiated from the ref- H. E. Hoekstra, Evolution 71, 261–273 (2017).
contributions: E.R.H. and H.E.H. initially conceived of the
erence haplotype, thus harboring many muta- 21. E. L. Berdan, A. Blanckaert, R. K. Butlin, C. Bank, PLOS Genet.
project. E.R.H., J.T.G., and K.M.T. planned and conducted the
17, e1009411 (2021).
tions that may influence tail length and/or coat 22. B. Charlesworth, J. D. Jensen, Annu. Rev. Ecol. Evol. Syst. 52,
field collections. E.R.H. conducted the QTL mapping experiment
color. We expect that more than one mutation and analyzed the forest-prairie phenotypes; E.R.H. and
177–197 (2021).
K.M.T. generated genetic data for the cross; and E.R.H.,
contributes to the inversion’s selective benefit 23. B. E. Horner, Contrib. Lab. Vertebr. Biol. 61, 1–85
J.T.G., S.T., S.M., B.N., and K.M.T. generated wild and laboratory
(1954).
in the forest, given the size of the inversion 24. S. T. Mincer, G. A. Russo, Proc. R. Soc. B. 287, 20192885
phenotype data. O.S.H. analyzed genetic data for the F2 cross,
(41 Mb), its large selection coefficient in the E.R.H. and O.S.H. performed QTL mapping analyses, T.B.W.
(2020).
analyzed long-read sequence data, O.S.H. and T.B.W. analyzed
forest (s ≈ 3 × 10−4, or Ns ≈ 120), and its as- 25. E. R. Hager, H. E. Hoekstra, Integr. Comp. Biol. 61, 385–397
wild-caught forest-prairie genetic data, O.S.H. analyzed transect
(2021).
sociation with two largely developmentally 26. R. D. H. Barrett et al., Science 363, 499–504 (2019).
genetic data, and E.R.H. and O.S.H. performed cline analyses.
distinct traits. If this is the case, the inversion’s T.B.W. performed demographic simulation, model-fitting, and
27. R. Faria, K. Johannesson, R. K. Butlin, A. M. Westram,
inference, and O.S.H. performed selection simulations,
suppression of recombination likely provides Trends Ecol. Evol. 34, 239–248 (2019).
model-fitting, and inference, both with input from J.D.J. E.R.H.,
28. M. J. Thompson, C. D. Jiggins, Heredity 113, 1–8 (2014).
an additional benefit (beyond the individual O.S.H., T.B.W., J.D.J., and H.E.H. wrote the manuscript, with
29. A. H. Sturtevant, Proc. Natl. Acad. Sci. U.S.A. 7, 235–237
effects of its mutations) in the forest popula- input from all authors. Competing interests: The authors
(1921).
declare no competing interests. Data and materials
tion, as long as strongly deleterious mutations 30. Z. Fang et al., Genetics 191, 883–894 (2012).
availability: Associated data are available as supplementary
are uncommon. This finding—that recombi- 31. K. Huang, R. L. Andrew, G. L. Owens, K. L. Ostevik,
files data S1 to S8 and on the NCBI Sequence Read Archive
L. H. Rieseberg, Mol. Ecol. 29, 2535–2549 (2020).
nation suppression is likely beneficial in this 32. D. B. Lowry, J. H. Willis, PLOS Biol. 8, e1000500
(nos. PRJNA687993, PRJNA688305, and PRJNA816517).
Associated scripts are available on Zenodo (55). License
system—provides empirical support for the (2010).
information: Copyright © 2022 the authors, some rights
local adaptation hypothesis, which posits that 33. C.-R. Lee et al., Nat. Ecol. Evol. 1, 0119 (2017).
reserved; exclusive licensee American Association for the
34. M. Joron et al., Nature 477, 203–206 (2011).
inversions are beneficial in the face of gene flow 35. K. Kunte et al., Nature 507, 229–232 (2014).
Advancement of Science. No claim to original US government
because they increase linkage disequilibrium works. https://www.science.org/about/science-licenses-
36. Z. Yan et al., Nat. Ecol. Evol. 4, 240–249 (2020).
journal-article-reuse
between adaptive alleles (5, 17, 18). 37. C. Mérot et al., Mol. Biol. Evol. 38, 3953–3971 (2021).
38. C. Cheng et al., Genetics 190, 1417–1432 (2012).
One hundred years after Alfred Sturtevant 39. J. L. Feder, J. B. Roethele, K. Filchak, J. Niedbalski,
first provided evidence of chromosomal inver- J. Romero-Severson, Genetics 163, 939–953 (2003). SUPPLEMENTARY MATERIALS
sions in laboratory stocks of Drosophila (29) 40. M. Kapun, T. Flatt, Mol. Ecol. 28, 1263–1282 (2019).
science.org/doi/10.1126/science.abg0718
41. E. L. Koch et al., Evol. Lett. 5, 196–213 (2021).
and, separately, forest-prairie ecotypes were first 42. D. Lindtke et al., Mol. Ecol. 26, 6189–6205 (2017).
described in wild populations of Peromyscus 43. T. Dobzhansky, Evolution 1, 1–16 (1947).
Figs. S1 to S21
Tables S1 to S6
(7), we found that a large chromosomal inver- 44. A. Brelsford et al., Curr. Biol. 30, 304–311.e4 (2020).
45. D. Ayala, R. F. Guerrero, M. Kirkpatrick, Evolution 67, 946–958
sion is key to ecotype divergence in this classic MDAR Reproducibility Checklist
(2013).
system. Inversions have been identified in as- Data S1 to S8
46. F. C. Jones et al., Nature 484, 55–61 (2012).
sociation with divergent ecotypes in diverse 47. M. Matschiner et al., Nat. Ecol. Evol. 6, 469–481 (2022).
species, including plants (30–33), invertebrates 48. C. Küpper et al., Nat. Genet. 48, 79–83 (2016).
49. H. B. Thorneycroft, Science 154, 1571–1572 (1966). Submitted 14 December 2020; resubmitted 26 November 2021
(34–45), fish (46, 47), and birds (48–52). In 50. S. Lamichhaney et al., Nat. Genet. 48, 84–88 (2016). Accepted 24 June 2022
mammals, however, evidence for ecotype-defining 51. E. R. Funk et al., Nat. Commun. 12, 6833 (2021). 10.1126/science.abg0718

RES EARCH
◥ In particular, the molar ratios of olefin to

REPORTS paraffin (o/p) in these by-products were very
high. For example, the o/p ratio of C6 mole-
CATALYSIS cules is 14.7, which is favorable for separation
and purification of desired products (13). A
Physical mixing of a catalyst and a hydrophobic higher gas hourly space velocity (GHSV) of
3600 ml gCoMnC–1 hour–1 decreased the CO
polymer promotes CO hydrogenation conversion to 15.4% on the CoMnC catalyst,
but the CoMnC/PDVB still showed a remark-
through dehydration ably high CO conversion of 50.7% (table S1).
Under a GHSV of 7200 ml gCoMnC–1 hour–1,
Wei Fang1†, Chengtao Wang1,2†, Zhiqiang Liu3†, Liang Wang1*, Lu Liu1, Hangjie Li2, Shaodan Xu4, the CoMnC catalyst exhibited poor CO con-
Anmin Zheng3*, Xuedi Qin2, Lujie Liu1, Feng-Shou Xiao1,5* version of 5.5%, whereas the CoMnC/PDVB
still showed CO conversion of 19.4%. In this
In many reactions restricted by water, selective removal of water from the reaction system is critical case, the space-time productivity of light ole-
and usually requires a membrane reactor. We found that a simple physical mixture of hydrophobic fins on the CoMnC/PDVB reached as high
poly(divinylbenzene) with cobalt-manganese carbide could modulate a local environment of catalysts as 7.1 mmol gCoMnC–1 hour–1, which exceeded
for rapidly shipping water product in syngas conversion. We were able to shift the water-sorption equilibrium the rate for CoMnC (2.1 mmol gCoMnC–1 hour–1,
on the catalyst surface, leading to a greater proportion of free surface that in turn raised the rate of carbon basis) by a factor of 3.4 (fig. S4). In
syngas conversion by nearly a factor of 2. The carbon monoxide conversion reached 63.5%, and 71.4% addition, a CO2 selectivity of 46.0 to 48.5% was
of the hydrocarbon products were light olefins at 250°C, outperforming poly(divinylbenzene)-free obtained in these cases, similar to that in
catalyst under equivalent reaction conditions. The physically mixed CoMn carbide/poly(divinylbenzene) general Fischer-Tropsch synthesis to olefins
catalyst was durable in the continuous test for 120 hours. (8, 14) and OX-ZEO (oxide-zeolite) reaction
processes for converting syngas to olefins
(15, 16). Adjusting the H2/CO ratio to 3.6 and
S
elective and rapid removal of water ically mixed the hydrophobic promoter with introducing a small amount of CO2 in the
product from a reaction system has been the catalyst, unlike previous chemical mod- syngas feed reduced the CO2 selectivity in the
a highly desirable pathway toward boost- ifications; in the syngas conversion to olefins products to 23.7%, giving a one-pass yield of
ing catalytic performance in reactions that with a cobalt-manganese carbide (CoMnC) light olefins at 28.0% (CO2 included in calcu-
are restricted by water thermodynami- catalyst, we achieved an increase in light olefin lating the olefin yield; table S4).
cally and/or kinetically (1, 2). Membrane reactors productivity by a factor of up to 3.4 by mix- We compared the performance of CoMnC/
designed to include water-conduction nano- ing the catalyst with the promoter. Mechanis- PDVB to different catalysts tested previously
channels could shift the reaction equilibrium tic studies revealed that the water molecules in syngas conversion to light olefins. The data in
(3), but preparation of defect-free membranes rapidly desorbed from the catalyst surface after fig. S5 show the comparison of the selectivity-
at a large scale is challenging. Chemical hydro- they formed from CO hydrogenation, which conversion results reported for various cata-
phobilization of the catalyst surface could avoided the competitive adsorption with CO lysts in their stable period during the reaction
substantially contribute to reactions by ac- reactant, a crucial step in the reaction process. (table S5). The general Fischer-Tropsch synthesis
celerating water diffusion (4–7), but in many The CoMnC catalyst was prepared via co- to olefin processes produced olefins having a
cases the chemical interactions might change precipitation and carbonization procedures wide carbon number distribution within the
the structure of the catalyst surface or even (fig. S1) (8, 9). In the syngas conversion range C1–C20 (17, 18). The selectivity of 53.0%
block the active sites by hindering access of under the given reaction conditions (H2/CO of to light olefins, with a CO conversion of 80.0%
reactant molecules. 2, 1800 ml gCoMnC–1 hour–1, 0.1 MPa, 250°C), over Fe-based catalysts, has been reported as
Enabling rapid water diffusion from an un- the CoMnC catalyst showed a CO conversion one of the most efficient processes (14), requir-
changed catalyst surface is an attractive alter- of 32.2%, with selectivity for light olefins ing a high reaction temperature of 340°C.
native. By promoting rapid desorption of water (C2= to C4=) of 60.8% (fig. S2 and table S1; the The OX-ZEO process (15), which combines the
molecules once they are formed on the catalyst CO2 product was excluded in calculating the cascade reactions of CO hydrogenation over
surface (2), the sorption equilibrium of water selectivity). In our initial attempt, we physically metal oxide and C-C coupling of methanol or
is shifted, as described by *H2O ⇌ * + H2O mixed the CoMnC catalyst with a nonporous ketene intermediates over zeolite, showed
(* denotes the catalyst surface sites). We phys- poly(divinylbenzene) (PDVB) (water-droplet superior selectivity to light olefins but yielded
contact angle 145°, irregular morphology, low CO conversion (e.g., 17.0% over the ZnCrOx/
1
Key Lab of Biomass Chemical Engineering of Ministry of surface area <5 m2/g; table S2). This hydro- SAPO-34 catalyst) at even higher temperatures
Education, College of Chemical and Biological Engineering, phobic polymer has a chemically inert sur- (400°C). Relative to these processes, the reaction
Zhejiang University, Hangzhou 310027, China. 2Key Lab of
Applied Chemistry of Zhejiang Province, Department of
face (fig. S3) and good thermal stability (10). over the CoMnC/PDVB catalyst proceeded at
Chemistry, Zhejiang University, Hangzhou 310028, China. The mixed catalyst, denoted CoMnC/PDVB, lower reaction temperatures with high selec-
3
National Center for Magnetic Resonance in Wuhan, State had substantially improved CO conversion tivity to light olefins and efficiently suppressed
Key Laboratory of Magnetic Resonance and Atomic and
Molecular Physics and Mathematics, Wuhan Institute of
(63.5%) and selectivity to light olefins (71.4%) methane formation. The CoMnC/PDVB catalyst
Physics and Mathematics, Innovation Academy for Precision (Fig. 1) relative to the CoMnC catalyst under also exhibited enhanced catalytic performance
Measurement Science and Technology, Chinese Academy of equivalent reaction conditions. With regard relative to the bare CoMn catalyst, which has
Sciences, Wuhan 430071, China. 4College of Materials and
to the C5+ by-products over CoMnC/PDVB, been regarded as a superior catalyst for low-
Environmental Engineering, Hangzhou Dianzi University,
Hangzhou 310018, China. 5Beijing Advanced Innovation the proportions of pentene and hexene were temperature syngas conversion to light olefins (8).
Center for Soft Matter Science and Engineering, Beijing 50.3% and 26.0%, respectively (table S3); these The catalytic performance of CoMnC/PDVB
University of Chemical Technology, Beijing 100029, China. are desired products for the production of high- catalyst was influenced by the manner of mixing
*Corresponding author. Email: [email protected] (L.W.);
[email protected] (A.Z.); [email protected] (F.-S.X.) performance polymers and valuable chemicals of the CoMnC and PDVB components (Fig. 2, A
These authors contributed equally to this work. (11, 12). and B, table S6, and fig. S6). Compared with a

RES EARCH | R E P O R T S
Fig. 1. Catalytic data in light olefin

production from syngas. (A and
B) CO conversion (A) and hydrocarbon
distribution (B) over the CoMnC
and CoMnC/PDVB catalysts. Reaction
conditions: 1.0 g of CoMnC catalyst
or mixture containing 1.0 g of CoMnC
and 1.0 g of PDVB, H2/CO/Ar at
64/32/4, 1800 ml gCoMnC–1 hour–1,
0.1 MPa, 250°C. The error bounds
were estimated by repeating the
experiment more than six times. Inset
in (A): Water-droplet contact angles
(CAs) of CoMnC and CoMnC/PDVB.
*Reaction using a feed gas of H2/CO/
Ar/CO2 at 68/19/3/10.
Fig. 2. Effect of different physical mixture

methods and catalyst durability. (A) Syngas
conversion performance of CoMnC/PDVB produced
using different mixing techniques. Reaction
conditions: 1.0 g of CoMnC and 1.0 g of PDVB,
H2/CO at 2, 1800 ml gCoMnC–1 hour–1, 0.1 MPa,
250°C. (B) Photographs of the catalyst beds
with different mixing techniques. (C) Durability
of CoMnC/PVDB catalyst in the syngas conversion
to light olefins. Reaction conditions: 1.0 g of
CoMnC catalyst physically mixed with 1.0 g
of PDVB (powder-mixing), H2/CO at 2,
1800 ml gCoMnC–1 hour–1, 0.1 MPa, 250°C.
powder mixture of the CoMnC and PDVB, the of 29.1% and lower olefin selectivity of 61.2% sion was constant at steady state with an aver-
granule mixture, which was prepared by gran- were similar to those of bare CoMnC catalyst. age value of ~64.7% (Fig. 2C). Even after reaction
ulating the CoMnC and PDVB components This result indicated that PDVB is inert for the for 120 hours, the CO conversion was well
and then mixing them together (40 to 60 mesh), reaction and that the promotion with PDVB maintained at 62.8% with stable CoMnC and
showed lower CO conversion of 53.7% and required physical mixing with the CoMnC PDVB components, confirming the good dura-
similar C2-C4 olefin selectivity (66.4%). In a catalyst. bility of the CoMnC/PDVB. In this process, the
dual-bed reactor, the PDVB was packed below The CoMnC/PDVB catalyst was used in a selectivity of light olefins was also constant
the CoMnC catalyst bed and separated by a continuous reaction test to evaluate durability. at 70.0%, with an average productivity at
layer of inert quartz sand. The CO conversion After activation for ~15 hours, the CO conver- 5.9 mmol gCoMnC–1 hour–1. During the test, the

selectivities for undesired methane and C2–C4 than those on the bare CoMnC, indicating that resistance in syngas conversion (25–27). Follow-
alkanes remained lower than 5.0% and 7.0%, the physically mixed PDVB indeed boosted ing this route, we also modified the CoMn
respectively. The selectivity for C5+ products the activity. catalyst using tetraethyl orthosilicate and
was ~18.0%, and 92.0% of those are C5–C8 It might be expected that the PDVB could dimethyl diethyloxysilane, but this resulted
olefins (19, 20). participate in the CoMnC carbonization that in lower CO conversion over the CoMn@Si
To understand the promotion of PDVB in leads to the distinguishable catalytic perfor- and CoMn@Si-c catalysts relative to the bare
the syngas conversion, we used temperature- mances (22, 24). We excluded this hypothesis by CoMnC catalyst (table S1). XRD patterns of
programmed surface reaction mass spectrom- characterizing the CoMnC phase change as the used catalysts showed the presence of oxide
etry (TPSR-MS) by feeding syngas to the a function of reaction time, which showed phases and suggested that hindered carbon-
catalysts. The CoMnC catalyst has been reported negligible difference in the x-ray diffraction ization led to formation of the active CoMn
to have superior activity for hydrogen activa- (XRD) patterns of the CoMnC with and without carbides (fig. S13) (27).
tion and cleavage of the C-O bond (8, 21–23), PDVB during the reaction periods (fig. S7); this We also varied the amount of PDVB in the
and these steps can be identified in TPSR-MS result was also supported by the TEM charac- catalyst bed. PDVB/CoMnC weight ratios of
tests. The dependences of propylene, methane, terization (fig. S8). After removal of the PDVB 0.5, 1.0, and 1.5 resulted in distinguishable
and water signals (m/z = 42, 16, and 18, re- component from the used CoMnC/PDVB cata- CO conversions at 56.4%, 63.5%, and 70.5%,
spectively) on reaction temperatures (Fig. 3A) lyst, the resulting CoMnC component exhibited respectively (table S7). The sensitivity of CO
show that on the CoMnC catalyst, the water performance comparable to that of the as- conversion to PDVB content indicated that it
and methane signals appeared at 166° and prepared CoMnC (table S1). These results fur- plays a crucial role in catalysis. Although PDVB
203°C, which we assigned to the C-O cleavage ther indicate that the physical mixture with PDVB in the physical mixture did not change the
and hydrogenation reaction. At 217°C, the did not change the catalyst structure, and they catalyst structure, it might optimize the water
propylene signal began to appear because the are consistent with the high stability of PDVB diffusion because of its hydrophobicity. Thus,
C-C coupling step occurred after C-O dissoci- at the reaction temperature (figs. S9 to S12). we studied the role of added water in CO
ation. Similar signals also appeared on the The previous strategies of chemically mod- conversion over the CoMnC catalyst (Fig. 3B).
CoMnC/PDVB catalysts, but the signals of ifying the catalyst surface with hydrophobic Considering that the water production rate
propylene and water were obviously stronger organosilanes were developed to improve water was 8.8 to 23.5 mg gCoMn–1 hour–1 in the
CoMnC/PDVB-catalyzed syngas conversion
(calculated according to the oxygen balance
in the reaction system and the amount of
collected water product after reaction; fig. S14),
we added water at this rate to investigate
its influence on the CO conversion. The CoMnC
catalyst showed CO conversion of 33.5% at
the beginning of the reaction without water
injection, which then decreased to ~8.2% (aver-
age value) after water injection with a feed
rate of ~10.5 mg gCoMn–1 hour–1. Interestingly,
addition of PDVB to the catalyst efficiently
minimized the negative effect of water, which
exhibited only a relatively slight decrease in
CO conversion to ~57.4% under the equivalent
water feed. The activity of the CoMnC catalyst
was continuously reduced as more water was
added to the feed gas (fig. S15), and the CO
conversion dropped to 4.7% with water feed
rate of ~20.0 mg gCoMn–1 hour–1. These data
confirmed the water-restricted feature of the
CoMnC-catalyzed syngas conversion. In con-
trast, the profile of CO conversion as a func-
tion of water concentration was flatter over
the CoMnC/PDVB catalyst. This hydrophobic
material may have helped the water product
to rapidly desorb after it was formed on the
CoMnC surface and also hindered its read-
sorption (fig. S16), which would free up active
sites for the continuous conversion of more
CO molecules.
Fig. 3. PDVB-optimized water sorption. (A) The dependences of water, methane, and propylene signals This hypothesis was supported by a pulse
(m/z at 18, 16, and 42) on temperature-programmed surface reaction by feeding syngas to the CoMnC and CoMnC/ experiment to explore the CO adsorption on
PDVB catalysts. (B) Data showing the influence of water on syngas conversion with CoMnC and CoMnC/PDVB. the catalyst surface with and without water
Reaction conditions: 1.0 g of CoMnC or 1.0 g of CoMnC physically mixed with 1.0 g of PDVB (powder-mixing), H2/CO injection (Fig. 3C). The CoMnC and CoMnC/
at 2, 1800 ml gCoMnC–1 hour–1, 0.1 MPa, 250°C. The water feed rate was ~10.5 mg gCoMn–1 hour–1. (C) Transient PDVB catalysts were localized within the flow-
response curves obtained during pulses of 10% CO/He (5 ml/min) into pure He flow (30 ml/min) at 250°C over the ing He atmosphere at 250°C, with periodic
CoMnC and CoMnC/PDVB catalysts. CO flowed through the water at 40°C for introducing water to the catalysts. pulsing of the CO or mixture of CO and water.
(D and E) CO desorption in situ FTIR spectra of anhydrous CoMnC (D) and water-pretreated CoMnC catalysts (E). In the test without water, the CO signals on

both catalysts were extremely weak compared

with that in the blank run without catalysts,
indicating that efficient CO adsorption oc-
curred on both CoMnC and CoMnC/PDVB
catalysts. In these cases, the CO signals were
similar, revealing the negligible effect of PDVB
on CO sorption, and the CO adsorption domi-
nantly occurred on the CoMnC. When water
was fed into the reaction (~7 vol% in CO), the
CO pulse peaks on the CoMnC catalyst were
markedly increased, with the intensities sim-
ilar to that in the blank run, revealing the
hindered CO adsorption by competition with
water.
Under the equivalent test, the CO pulse peaks
were still weak on the CoMnC/PDVB catalyst,
similar to the result of the water-free test, con-
firming the negligible influence of water on
CO sorption in the presence of PDVB. A similar
phenomenon was observed in the equivalent
tests by changing the water feed amount (fig.
S17). On the basis of these results, we conclude
that the water on the catalyst surface could
hinder the CO adsorption, while the PDVB ef-
ficiently shifted the sorption equilibrium de-
scribed by *H2O ⇌ * + H2O by accelerating
the rapid water desorption and hindering its
readsorption. These features led to a higher
proportion of free catalyst surface for con-
tinuous conversion of CO, boosting the syngas
conversion to olefins (fig. S18).
In situ Fourier transform infrared (FTIR)
spectra characterizing the CO hydrogenation
on the CoMnC and CoMnC/PDVB catalysts
are shown in fig. S19. Introducing CO and hy-
drogen to the CoMnC catalyst led to the for-
mation of obvious bands at 2717 to 2955 cm–1,
1320 to 1527 cm–1, and ~1039 cm–1, which were Fig. 4. Theoretical simulation. (A and B) Models showing the water diffusion within regions surrounded
assigned to the olefin species from CO hydro- by hydrophilic and hydrophobic surfaces. (C) Mean square displacement (MSD) and diffusion coefficient
genation and C-C coupling (28–30). The broad (Ds) showing the water diffusion efficiency at 250°C. (D) SEM image of the CoMnC/PDVB granule.
signals at ~1600 cm–1 and 3400 to 3600 cm–1 (E) Scheme showing the escape of water from the CoMnC surface through the region surrounded by
appeared and continuously increased with re- PDVB. (F and G) Models showing the water escape through different regions. The regions I are hydrophilic
action time because of the water product ad- and the regions III are hydrophilic and hydrophobic, respectively. (H) The number of water molecules that
sorbed on the catalyst surface (31–34). On escaped from region III from an initial state with 100 water molecules on region I as a function of time.
the CoMnC/PDVB catalyst, the signals of olefin
products were solely observed with an extremely study (figs. S20 and S21). According to previous on water diffusion by a theoretical simula-
weak water signal, suggesting the rapid desorp- studies, less water on the catalyst surface might tion, in which we explored water molecule
tion of water once it is formed on the catalyst. reduce the concentration of carbon-based inter- diffusion in a region surrounded by hydro-
The effect of water on CO adsorption was mediates and enhance the hydrogenation activ- philic or hydrophobic surfaces (Fig. 4, A and
further explored (Fig. 3, D and E). Introducing ity (35, 36), which could explain the improved B). Water diffusion efficiency was quantified
CO to the CoMnC catalyst led to signals of selectivity of short hydrocarbons and slightly by the diffusion coefficient Ds. The hydrophilic
chemically adsorbed CO on the CoMnC sur- reduced o/p ratios on the CoMnC/PDVB cat- surface interacted with water molecules to
face (22, 33). When water was co-fed with CO, alyst relative to bare CoMnC. slow down the transportation (Fig. 4C), giv-
the signals of chemically adsorbed CO were In addition, the olefin products adsorbing ing Ds = 2.2 × 10–7 m2/s. In the region sur-
almost undetectable with only the bands of on the catalyst surface might also hinder the rounded by hydrophobic surfaces, the water
gaseous CO (34), confirming that CO adsorp- CO conversion to some extent, as confirmed diffusion was accelerated with obviously
tion was hindered under competition with by catalysis studies with ethylene in the syngas higher Ds = 4.7 × 10–7 m2/s (Fig. 4C), given
water. This phenomenon might explain the feed (fig. S22). However, further tests suggested the relatively weak interaction between water
negative effect of water on the syngas con- a negligible effect of PDVB on olefin sorption and the hydrophobic surface (37). The theo-
version by the competitive adsorption, in good (fig. S23). Thus, PDVB accelerated syngas con- retical simulation therefore gives a qualitative
agreement with the results of pulse experi- version in the catalytic tests; this was primarily trend of the diffusion of water molecules along
ments. In contrast, the catalyst containing attributed to its hydrophobicity in removing the different surface.
PDVB exhibited comparable CO adsorption water rather than olefins. We further inves- In our work, the CoMnC catalyst and PDVB
with and without water injection in a FTIR tigated the influence of promoter wettability were physically mixed and randomly distributed

in the catalyst granules. As observed in the scan- improved to 88.7% and 92.4% over the CoMnC/ REFERENCES AND NOTES
ning electron microscope (SEM) image of an nanoporous PDVB catalysts, but selectivity 1. M. Wolf, N. Fischer, M. Claeys, Nat. Catal. 3, 962–965 (2020).
actual CoMnC/PDVB catalyst granule (Fig. for light olefins was at 53.4% and 37.6% (table 2. H. Li et al., Science 367, 667–671 (2020).
3. S. M. Hashim, A. R. Mohamed, S. Bhatia, Rev. Chem. Eng. 27,
4D), the CoMnC and PDVB were packed tight- S1). Relative to CoMnC/PDVB, the CoMnC/nano- 157–178 (2011).
ly and disordered, with intergranular distances porous PDVB catalysts showed higher selec- 4. Y. Xu et al., Science 371, 610–613 (2021).
ranging from nanometers to micrometers tivity for the heavier olefins (C5 to C8 selectivity 5. S. Crossley, J. Faria, M. Shen, D. E. Resasco, Science 327,
68–72 (2010).
(Fig. 4E). We studied how the hydrophobic of 33.7% and 43.7%, respectively). This phenom- 6. H. Yang, T. Zhou, W. Zhang, Angew. Chem. Int. Ed. 52,
promoter affects the escape of water that is enon might be the result of the high adsorp- 7455–7459 (2013).
produced on the relatively hydrophilic surface tion capacity of nanoporous PDVB for the 7. Z. Jin et al., Science 367, 193–197 (2020).
8. L. Zhong et al., Nature 538, 84–87 (2016).
(e.g., the CoMnC surface). In Fig. 4, F and G, olefin products (fig. S32), which prolonged 9. R. Yang et al., J. Energy Chem. 32, 118–123 (2019).
we show models simulating the water escape the retention time of olefin products in the 10. Y. Zhang et al., Nano Today 4, 135–142 (2009).
from the solid surface (region I) through the catalyst bed to benefit chain growth. These 11. P. Ramírez de la Piscina, J. L. G. Fierro, G. Muller, J. Sales,
N. Homs, Catal. Lett. 14, 45–49 (1992).
region surrounded by hydrophobic or hydro- data confirmed that product distribution could 12. W. Fan, P. Wu, T. Tatsumi, J. Catal. 256, 62–73 (2008).
philic surfaces (region III), respectively. Nota- be adjusted by changing the nanoporosity of 13. R. Yang, R. Gao, Z. Qian, Y. Wang, Separ. Purif. Tech. 230,
bly, these regions were separated from each the PDVB promoter. 115884 (2020).
14. H. M. Torres Galvis et al., Science 335, 835–838 (2012).
other (region II) for simulating the physical Similar trends were also observed in the
15. F. Jiao et al., Science 351, 1065–1068 (2016).
mixing with intergranular distances between reaction with the CoMnC catalyst mixed with 16. K. Cheng et al., Angew. Chem. Int. Ed. 55, 4725–4728
the CoMnC and PDVB granules in the catalyst methyl group–modified silica with a hydro- (2016).
(figs. S24 and S25). phobic surface (SiO2-Me, figs. S33 to S35), which 17. S. Soled, E. Iglesia, R. A. Fiato, Catal. Lett. 7, 271–280
(1990).
The number of escaped water molecules as a showed CO conversion at 74.7% and lower se- 18. M. D. Shroff et al., J. Catal. 156, 185–207 (1995).
function of diffusion time in the systems with lectivity for C2–C4 olefins at 45.5% (table S1). 19. T. Ren, M. Patel, K. Blok, Energy 31, 425–451 (2006).
hydrophobic and hydrophilic promoters, from Higher selectivity for C5–C8 products and lower 20. R. Snel, Catal. Rev., Sci. Eng. 29, 361–445 (1987).
21. H. Wang et al., J. Am. Chem. Soc. 135, 4149–4158 (2013).
an initial state with 100 water molecules on o/p ratios were obtained than with the reaction 22. Z. Li et al., ACS Catal. 7, 3622–3631 (2017).
region I (Fig. 4H), indicated that more water over CoMnC/PDVB. When graphite, a hydro- 23. F. Jiang et al., Catal. Today 369, 158–166 (2021).
molecules escaped through the hydrophobic phobic carbon material, was mixed with CoMnC 24. Y. Yang et al., Appl. Catal. A 549, 179–187 (2018).
25. R. Xie et al., Catal. Commun. 12, 589–592 (2011).
channel than from the hydrophilic channel (CoMnC/Gra, fig. S36), the CO conversion was 26. M. Ojeda et al., Langmuir 22, 3131–3137 (2006).
under the equivalent conditions. For exam- 53.0% with 67.3% selectivity for light olefins 27. X. Yu et al., Appl. Catal. B 232, 420–428 (2018).
ple, after 500 ps, ~32% of the initial water (table S1). Given that the graphite is earth- 28. C. J. Weststrate, J. W. Niemantsverdriet, ACS Catal. 8,
10826–10835 (2018).
molecules escaped from the model with the abundant and extremely cheap, our strategy 29. C. J. Weststrate, I. M. Ciobîcă, J. van de Loosdrecht,
hydrophobic channel, whereas only 13% escaped for shifting water-mediated sorption equilib- J. W. Niemantsverdriet, J. Phys. Chem. C 120, 29210–29224
from the model with the hydrophilic channel. rium could be implemented simply by mixing (2016).
30. A. A. Efremov, A. A. Davydov, React. Kinet. Catal. Lett. 15,
The influence of water concentration on dif- hydrophobic graphite with the current catalysts.
327–331 (1980).
fusion rate was also simulated by regulating In addition, when relatively hydrophilic 31. L. Marchese, J. Chen, J. M. Thomas, S. Coluccia, A. Zecchina,
the number of water molecules in region I of materials (such as a mixture of PDVB, SiO2, J. Phys. Chem. 98, 13350–13356 (1994).
the initial state (e.g., 25, 50, and 100 water and hydrophilic polymers; figs. S37 to S40) 32. K. Tanaka, J. M. White, J. Phys. Chem. 86, 4708–4714 (1982).
33. D. J. C. Yates, J. Phys. Chem. 65, 746–753 (1961).
molecules; fig. S26). The results showed that were used in the CoMnC-catalyzed syngas 34. C.-J. Jia et al., J. Am. Chem. Soc. 133, 11279–11288 (2011).
increasing the concentration of water mole- conversion, the CO conversion was markedly 35. A. K. Dalai, B. H. Davis, Appl. Catal. A 348, 1–15 (2008).
cules on the hydrophobic model surface could reduced (tables S8 to S10). For example, poly- 36. C. J. Bertole, C. A. Mims, G. Kiss, J. Catal. 210, 84–96
(2002).
accelerate the diffusion rate of water mole- styrene (PS, fig. S40), which has composi- 37. K. Liu, X. Yao, L. Jiang, Chem. Soc. Rev. 39, 3240–3255 (2010).
cules, which helps to explain the rapid water tion similar to PDVB but is more hydrophilic,
diffusion in the syngas conversion reaction failed to promote the CoMnC-catalyzed syngas AC KNOWLED GME NTS
with continuously produced water molecules. conversion, showing CO conversion at 10.7% We thank F. Chen for help in SEM characterization. Funding:
Supported by the National Key Research and Development
The models showed that the hydrophobic pro- with methane selectivity of 37.8% and C2–C4 Program of China (2021YFA1500404), National Natural Science
moter physically regulated the catalyst by accel- olefin selectivity of 37.8% (table S10). These Foundation of China (U21B20101, 21932006, 22032005, 22102143,
erating the water diffusion, in good agreement data show the importance of a hydrophobic and 22125304), China Postdoctoral Science Foundation
(2021M702803), and National Postdoctoral Program for Innovative
with the experimental results. promoter. Talents (BX20200291). Author contributions: W.F. and C.W.
PDVB-promoted water sorption can also be Our approach could be used to upgrade performed the catalyst preparation, characterization, catalytic tests,
directly observed through a model experiment industrially catalytic processes without mod- and data analysis. Z.Q.L. and A.Z. performed the theorical calculation
and wrote the corresponding part. L.L., H.L., X.Q., S.X., and L.J.L.
of CuSO4·5H2O dehydration, because of its color ifying the catalysts themselves. In addition, the participated in the catalyst characterization and discussion. L.W. and
change from blue to white upon dehydration. strategy is conceptually different from catalyst F.-S.X. designed the study, analyzed the data, and wrote the paper.
After mixing a small amount of PDVB to the hydrophilization with organosilanes, described Competing interests: The authors declare that there is no conflict of
interest. Data and materials availability: All data are available in the
CuSO4·5H2O (0.75 wt% of PDVB in the mix- as a chemical modification route, where the
manuscript or the supplementary materials. License information:
ture), the color change was obviously accelerated, conversion was not obviously improved in Copyright © 2022 the authors, some rights reserved; exclusive
as confirmed by the photographs in fig. S27. syngas conversion but the CO2 selectivity was licensee American Association for the Advancement of Science. No
This result again confirms that PDVB promoted reduced by hindering the undesired water-gas claim to original US government works. www.science.org/about/
science-licenses-journal-article-reuse
water desorption and hindered readsorption shift (4). This difference might result from the
when physically mixed. distinguishable distances between the active SUPPLEMENTARY MATERIALS
We prepared physical mixtures of CoMnC site and the hydrophobic surface for these science.org/doi/10.1126/science.abo0356
catalyst with different materials whose wetta- different systems. Considering that many Materials and Methods
bility was distinguishable. When nanopores hydrogenation reactions are strongly affected Figs. S1 to S40
Tables S1 to S10
were introduced to the PDVB (two nanopo- by water, the physical regulation method using References (38–57)
rous PDVB materials with distinguishable a promoter with desired wettability could guide
Submitted 9 January 2022; resubmitted 18 April 2022
surface areas at 488.2 and 623.3 m2/g; table S2 the design of more efficient catalysts in the Accepted 24 May 2022
and figs. S28 to S31), the CO conversion further future. 10.1126/science.abo0356

ORGANIC CHEMISTRY vious total syntheses in which these strategic key

steps were largely independent (17). To this end,
A concise synthesis of tetrodotoxin we established a linear precursor that contained
all the oxygen functionalities of the TTX skele-
David B. Konrad1†‡, Klaus-Peter Rühmann2‡, Hiroyasu Ando3, Belinda E. Hetzler2, Nina Strassner4, ton, which would then be conjoined by a ring-
Kendall N. Houk4, Bryan S. Matsuura2*‡, Dirk Trauner2*¤ forming reaction. This would be followed by
installation of the a-tertiary amine through
Tetrodotoxin (TTX) is a neurotoxic natural product that is an indispensable probe in neuroscience, a C–C bond formation, to introduce the C2 frag-
biosynthetic and ecological enigma, and a celebrated target of synthetic chemistry. Here, we present a ment that would eventually be incorporated into
stereoselective synthesis of TTX that proceeds in 22 steps from a glucose derivative. The central cyclohexane the dioxa-adamantane. Finally, we needed to
ring of TTX and its a-tertiary amine moiety were established by the intramolecular 1,3-dipolar cycloaddition develop a method to oxidize and lactonize this
of a nitrile oxide, followed by alkynyl addition to the resultant isoxazoline. A ruthenium-catalyzed C2 fragment in a highly efficient manner.
hydroxylactonization set the stage for the formation of the dioxa-adamantane core. Installation of the Our ultimate retrosynthetic analysis is sum-
guanidine, oxidation of a primary alcohol, and a late-stage epimerization gave a mixture of TTX and marized in Fig. 1. Although not all compounds
anhydro-TTX. This synthetic approach could give ready access to biologically active derivatives. shown therein were defined in such detail at
the outset of our study, it captures the essence
T
of our synthetic plan. We reasoned that we
etrodotoxin (TTX) is a neurotoxic natural The first total synthesis of TTX, in racemic could trace TTX back from an oxidation of an
product that has inspired and empow- form, by Kishi and Fukuyama in 1972 stands alkynyl isoxazolidine of type 1, which would
ered chemists and biologists for more as a landmark achievement in organic synthe- stem from bicyclic isoxazoline 2, the product
than a century (1–3). As a selective blocker sis that, at the time, seemed hard to surpass of an intramolecular 1,3-dipolar cycloaddition.
of voltage-gated sodium channels, it has (6). After a pause of more than 30 years, Isobe Nitromethane would serve as a key linchpin
played a crucial role in the elucidation of the and co-workers published the first asymmetric in the assembly of 2, reacting first in an in-
action potential, and it is still routinely used to synthesis in 2003 (7). This was followed shortly termolecular Henry reaction with aldehyde
silence excitable cells in neural systems. Its by the Hinman and Du Bois’ asymmetric syn- 3, followed by a dehydration to generate a re-
isolation from widely differing species, such as thesis (2003) (8), a second Isobe approach active nitrile oxide intermediate that would
pufferfish, starfish, sea snails, octopi, toads, and (2004) (9, 10), and a racemic and two asym- close the central cyclohexane ring within a
newts, has prompted intense investigations into metric syntheses by Sato’s group (2005, 2008 (3+2) cycloaddition. We have previously devel-
its true biological producers, its biosynthesis, and 2010, respectively) (11–13). In 2017 and oped an asymmetric synthesis of unsaturated
and its ecological role. It is now clear that TTX 2020, Fukuyama, Yokoshima, and colleagues aldehydes similar to 3 via a Kiyooka aldol re-
is synthesized by bacteria and accumulated by revisited the molecule and published two dis- action and used it toward kweichowenol A, a
metazoan hosts as a defense against predators tinct asymmetric routes to TTX (14, 15). In polyoxygenated cyclohexene isolated from the
(4). Its toxicology and therapeutic utility in addition, several studies have been published plant Uvaria kweichowensis (25). Although an
humans have been studied for decades and that intercept late-stage intermediates of the analogous route gave the aldehyde 3 in suffi-
are still a topic of ongoing research (5). previous syntheses e.g., by the Alonso (2010) cient quantities to proceed with the synthesis
As a synthetic target, TTX has been cele- (16, 17), Ciufolini (2015) (18–20), and Hudlicky of TTX, we found it more practical and eco-
brated for the sheer intellectual challenge it (2018) (21) groups. Other approaches toward nomical to start from the glucose-derived
provides and for the opportunity to demon- the molecule have been outlined (3, 22, 23). building block 4. All the carbons of glucose
strate methodological and strategic advances. An analysis of previous syntheses revealed and two of its stereocenters would be retained
Its simple carbon framework, consisting of a several common features that motivated us to over the course of the synthesis, making this
cyclohexane ring with C1 and C2 side chains, pursue a distinct strategy: (i) The cyclohexane an attractive starting material.
stands in stark contrast to the dense network core was either incorporated in the starting Previously known exo-methylene building
of polar functional groups that adorn it. Two material, or was formed early, and then oxy- block 5 was synthesized in three steps on a
hydroxy groups in a syn relationship engage a gens were added using epoxidations, dihydrox- decagram scale from commercially available
carboxylate as an ortho acid to form the sig- ylations, or allylic oxidations of strategically glucose derivative 4 and was also used by Sato
nature dioxa-adamantane core of TTX, which placed alkenes. (ii) The a-tertiary amine on and colleagues in their approach to TTX (see
is fused to a cyclic guanidine via an a-tertiary C8a was established via C–N bond formation, supplementary materials) (13). Regioselective
amine. One primary, two secondary, and a ter- which must overcome considerable steric hin- reductive cleavage of the benzylidene acetal
tiary hydroxy group, as well as a hemiaminal, drance. Several methods have been implemented placed a benzyl ether at C5, yielding 6 (Fig. 2).
contribute further to the structural complexity to address this challenge, such as intramolecular A subsequent dihydroxylation then installed
of the molecule, which features four rings and nitrogen transfer (sigmatropic rearrangement, the tertiary alcohol with the correct absolute
nine contiguous stereocenters. aza-conjugate addition, nitrene insertion) and configuration at C6, as well as the C11 primary
intermolecular SN1-type nucleophilic substitu- alcohol of TTX, providing 7 in excellent yield
1
Department of Chemistry, Ludwig-Maximilians-Universität tion. (iii) The dioxa-adamantane was always and as a single diastereomer (26). Protection
München, Butenandtstr. 5-13, 81377 Munich, Germany. 2Department formed spontaneously with careful orchestra- of the vicinal diol as the acetonide, followed by
of Chemistry, New York University, 100 Washington Square East,
New York, NY 10003, USA. 3Graduate School of Infection Control
tion of the sequence to adjust the oxidation an Appel reaction, gave the primary iodide
Sciences and Kitasato Institute for Life Sciences, Kitasato state of C10 and set the labile C9 stereocenter. (iv) 8. Under conditions developed by Soengas
University, Tokyo, Japan. 4Department of Chemistry and Every synthesis introduces the guanidine at a and Silva, 8 underwent a reductive cleavage
Biochemistry, University of California, Los Angeles, CA 90095, USA.
late stage (7, 24) and uses protecting groups upon treatment with tert-butyl lithium at
*Corresponding author. Email: [email protected] (B.S.M.);
[email protected] (D.T.) amenable to global deprotection in the final step. low temperature, to yield a d,e-unsaturated
†Present address: Department of Pharmacy, Ludwig-Maximilians- Our synthetic analysis was guided by an at- aldehyde (12, Fig. 3), which engaged in a Henry
Universität München, Butenandtstr. 5-13, 81377 Munich, Germany. tempt to link the formation of the cyclohexane reaction in situ upon addition of nitromethane.
‡These authors contributed equally to this work.
§Present address: Department of Chemistry, University of Pennsylvania, core with the establishment of the a-tertiary This afforded nitro alcohols 9a,b as a separa-
Philadelphia, PA 19104-6323, USA. amine as closely as possible, in contrast to pre- ble 1:1 mixture of diastereomers (27).

TMS group upon workup, isoxazolidine alkyne

NH2
OH 10 19 was isolated as a single diastereomer, which
HO OH HO 9 HN O
NH HN NH possessed all the skeletal carbons of TTX and
8a
N HO 9
4a
4 HO
H O O H O OH
five out of its nine stereocenters with the correct
NH2 8 H
O
10 H 5
7 5
absolute configuration. The presence of a free
OH O H RO OR
O H O O hydroxy group in 10, which presumably coor-
OH O dinates to the nucleophile after deprotonation,
tetrodotoxin (TTX) OH
was found to be crucial for the success of the
1 addition to the oxime ether moiety. Isoxazoline
18, by contrast, gave very low conversion. The
O
Me3Si N O
lithiated TMS-acetylide was highly effective,
N O Me whereas more functionalized C2 synthons were
HO 8 O H found to be unreactive.
MeO O
H O Our next goal was the oxidative elaboration
6
RO OR RO OR 6 of the alkyne to introduce the stereogenic C9
O O HO O Ph
11 alcohol and the C10 lactone that would ulti-
O O Me3Si OH
11 mately engage in the formation of the signature
2 3 MgCl 4 ortho acid of TTX. Our original strategy was to
Fig. 1. Retrosynthetic analysis and synthetic design. form the requisite hydroxylactone via a gold- or
silver-catalyzed 5-endo-dig hydroetherification
of the C8 hydroxy group of 19 onto the alkyne
terminus (Fig. 4A). Oxidation of the resultant
With nitro alcohols 9a,b in hand, we were ane core of TTX with the correct configuration at dihydrofuran to a hydroxylactone, followed by
ready to attempt the key nitrile oxide cyclo- C8, albeit with the wrong configuration at C4a. isomerization, via a translactonization during
addition to form the cyclohexane core of tetro- Both the diastereoselectivity of the oxa- the final deprotection, would unveil the central
dotoxin. Treatment of diastereomer 9a with Michael addition and the mechanism of the ortho acid. Unfortunately, extensive experimen-
phenyl isocyanate and catalytic amounts of 1,3-dipolar cycloaddition were further explored tation proved that this approach is exceptionally
triethylamine triggered a dehydrative nitrile with quantum mechanical density functional the- difficult to execute, requiring us to reexamine
oxide cycloaddition, yielding isoxazoline 10 in ory calculations (see supplementary materials). our sequence of bond-forming events. The logical
moderate yield as a single diastereomer (28). The addition of PMB alkoxide, which resulted in solution to this conundrum would be to engage
An x-ray crystallographic structure of a by- the stereoselective formation of 15, was shown to the alkyne terminus in a bridge-forming hydro-
product, N-phenyl carbamate 11, confirmed be favored by at least 3.4 kcal/mol (DG273K) rela- etherification with either the C5 or C7 hydroxy
the configurations of the C8 and C4a stereo- tive to the other five possible transition states. In group, followed by oxidation to the desired hy-
centers. Although the C4a stereocenter was the major transition state, the carbon chain droxylactone, which would obviate the need for a
inverted with respect to TTX, we reasoned that adopts an anti-configuration with respect to translactonization step. However, this strategic
this isoxazoline diastereomer would be more the forming bond, as the Felkin-Anh model pro- shift was not without risk, as such bridge-forming
accessible to nucleophiles for the formation poses, and the alkoxy group is syn to the double 6-endo-dig cyclizations have scarce precedent.
of the hindered a-tertiary amine (29). Expos- bond, an “inside alkoxy” arrangement shown to To this end, we needed to establish a protect-
ing diastereomer 9b to the same reaction be favored for related cycloadditions (31). ing group scheme that would also be compat-
conditions only resulted in decomposition. The 1,3-dipolar cycloaddition can proceed ible with the subsequent introduction of the
The poor diastereoselectivity of the Henry re- through two possible pathways, initiated either guanidine moiety and the final steps. Expo-
action and low yields of the cycloaddition se- by elimination of the tert-butyl carbonate (Boc- sure of 19 to Boc2O gave 20, which features
verely hampered material throughput. Moving OH), followed by nitrile oxide cycloaddition, or both a tert-butyl carbonate and carbamate
forward, we needed to develop a strategy that by cycloaddition of the Boc-nitronate, followed moiety. The next critical deprotection step
would selectively install the stereocenter at C8— by Boc-OH elimination. Calculations show that required the simultaneous cleavage of the C5
ideally with a protected hydroxyl—which is crit- an elimination of Boc-OH proceeds quickly with and C7 benzyl ethers in the presence of the
ical in the ensuing cycloaddition. We reasoned a barrier of only 9.9 kcal/mol (DG298K), where- alkyne and isoxazolidine groups, which are
that this could be accomplished via a dehydra- as both cycloadditions occur with barriers of sensitive to hydrogenation conditions, as well
tion to the corresponding nitroalkene followed ~23 kcal/mol. Thus, the nitrile oxide pathway is as three acid-sensitive protecting groups. After
by a conjugate addition with an appropriate O- strongly favored (see supplementary materials considerable experimentation, we found that
nucleophile. This was realized by subjecting 8 to for details). The computational investigation fur- the benzyl ethers could be cleanly removed
the same reductive fragmentation and Henry re- ther highlights a clear preference for the experi- by using Pieber and Seeberger’s recently in-
action sequence, followed by in situ dehydration, mentally observed stereoisomer at C4a with a troduced chromoselective photochemical de-
which yielded nitroalkene 14 exclusively as the free energy difference of 1.4 kcal/mol. Although benzylation (33), a singularly effective protocol
E-isomer. 14 underwent an oxa-Michael addi- both transition states (shown in supplementary that afforded diol 21 in excellent yield. The
tion upon treatment with the lithiated alkoxide of materials) adopt chair conformations, the one Boc-group on the C8-alcohol and the use of a
p-anisyl alcohol, presumably resulting in nitro- favoring the observed product avoids a 1,3- 525-nm green light-emitting diode as the irra-
nate anion 15. This intermediate could be inter- diaxial interaction of the alkoxy groups. diation source were key to this reaction’s suc-
cepted with Boc-anhydride (via 16) to trigger the After deprotection of the p-methoxybenzyl cess. The use of either the unprotected alcohol
formation of the transient nitrile oxide 17 and (PMB) group with ceric ammonium nitrate (CAN), or higher-energy blue light resulted in com-
subsequent 1,3-dipolar cycloaddition, affording 10 was subjected to lithiated trimethylsilyl (TMS)– plete substrate decomposition. Following the
isoxazoline 18 in high yield and on a decagram acetylide which, in the presence of BF3·OEt2, debenzylation, a selective methanolysis of the
scale (30). This reaction cascade was exquisitely underwent addition from the convex face of the carbonate gave a triol, which could be protected
diastereoselective, affording the central cyclohex- bicyclic isoxazoline (32). After cleavage of the to bis-acetonide 22. At this stage, we elected to

5. OsO4 (1.4 mol %)

quinuclidine (1.4 mol %)
MeSO2NH2 (1.0 equiv) 6. PTSA (4 mol%) MeO O
K2CO3 (3.0 equiv) MeO O 2,2-dimethoxypropane I
MeO O MeO O OH
O 4. DIBAL (3.0 equiv) OH K3Fe(CN)6 (3.0 equiv) then
6 BnO OBn
BnO OBn O
BnO O Ph toluene, -10 to 0 °C BnO OBn 1:1 t-BuOH/H2O 11 OH imidazole (10.0 equiv)
OH PPh3 (5.0 equiv) O
(96%) (98%) I2 (4.5 equiv), DCM
5 6 7 8
(3 steps from (91%)
commercial starting
material 4)
incorrect
1:1 d.r. configuration
7. t-BuLi (2.0 equiv) PhHN O
NO2 N O N O
then R2 PhNCO (4.0 equiv)
MeNO2 (10.0 equiv) Et3N (0.2 equiv) HO 4a O
R1 8 8
9a H + H ≡
Et2O, –78 °C BnO OBn toluene, r.t. BnO OBn BnO OBn
O O O
(89%, 1:1 d.r.) O (35%) O O
9a: R1 = OH, R2 = H 10 11
9b: R1 = H, R2 = OH
Fig. 2. Opening sequence and initial attempts to form the carbocyclic core. DIBAL, diisobutylaluminium hydride; PTSA, p-toluene sulfonic acid; DCM,
dichloromethane; PhNCO, phenylisocyanate; r.t., room temperature.
7. t-BuLi (2.0 equiv)
then O NO2 NO2 NO2
MeO O O H
I MeNO2 (10.0 equiv) N HO MsO 8
Et2O, –78 °C to r.t. O MsCl, Et3N elimination
BnO OBn BnO OBn
O O BnO OBn BnO OBn BnO OBn
then O O O
O MsCl (4.0 equiv) O
O O O
Et3N (3.0 equiv)
8 12 9a/b 13
(79%) 14
correct
configuration
N O
8. PMBOH (2.05 equiv) Bn O Bn O Bn O PMBO 4a
n-BuLi (2.00 equiv) O Boc2O O
PMBO O PMBO O PMBO O 8 H
then O O O (3+2) CA
N N -CO2
O 8 OBn -t-BuO O OBn OBn BnO OBn
Boc2O (3.5 equiv) -t-BuOH O
DMAP (0.1 equiv) H
t-BuO O
H O N O
(86%) 15 16 17
18
Ln 10
N O 10. TMSCCLi (4.2 equiv) HN O
Li Bn Li α-tertiary
9. CAN (4.0 equiv) HO BF3·OEt 2 (1.05 equiv) 8a
O O amine HO
NaHCO3 (2.0 equiv) H
H BnO O then TMS O O formation 8 H
7 5
O O
4:1 MeCN/H2O, r.t. BnO OBn OBn TBAF (6.0 equiv) OBn BnO OBn
O O O O
N NaH2PO4 (3.15 equiv) N
(93%) O H2O (25 vol% of THF) O
BF3
10 (70% + 26% SM) 19
Fig. 3. Development of a diastereoselective route to the cyclohexane core and installation of the a-tertiary amine. MsCl, methanesulfonyl chloride; PMBOH,
p-methoxybenzyl alcohol; Boc2O, di-tert-butyl dicarbonate; DMAP, 4-dimethylaminopyridine; (3+2) CA, (3+2) cycloaddition; CAN, ceric ammonium nitrate; TMSCCLi,
lithium trimethylsilylacetylide; TBAF, tetra-n-butylammonium fluoride; THF: tetrahydrofuran.
reduce the sensitive N-Boc isoxazolidine. By using this via a 6-endo-dig cyclization with gold- or scale (36). Pushing this finding even further, we
SmI2, the N–O bond was cleaved in excellent silver-based p-acid catalysts, but this approach postulated that the resultant dihydropyran could
yield, resulting in a primary alcohol, which was was thwarted by the substrate’s propensity to be converted to the key hydroxylactone 25 by
then protected as its silyl ether 23 (Fig. 4A). undergo an undesired 5-exo-dig cyclization. transforming the cycloisomerization catalyst
The stage was now set for the next key step of Our solution to this problem was inspired into an oxidant. This hypothesis was supported
our synthesis, the conversion of alkyne 23 to by reports from Trost (34) and McDonald (35) by Blechert and co-workers, who demonstrated
hydroxylactone 25. To take full advantage of the on the catalytic generation of metallo-vinylidene that ring-closing metathesis could be coupled
steric environment provided by the proximal carbenes, which would render the alkyne ter- with olefin dihydroxylation on simple bis-alkenes
acetonide, we aimed at forming the C10-O5 bond minus (C10) electrophilic. We found that 23 by using Ru-based metathesis catalysts (37).
first to yield a dihydropyran that could then be could be converted to bridged dihydropyran However, we would need to identify conditions
oxidized to the hydroxylactone in a stereose- 24, with CpRu(PPh3)2Cl as a cycloisomerization that would further oxidize the hypothetical diol
lective manner. Initially, we attempted to achieve catalyst, in nearly quantitative yield on a 300-mg to the hydroxylactone, without overoxidation of

the desired product (e.g., oxidative cleavage or 25 with almost complete diastereoselectivity ting the C9 stereocenter in the processes, and
ketolactone formation). This was achieved by (Fig. 4B). We believe the chemoselectivity of the had the added benefit of simplifying the pu-
the addition of Oxone and a cosolvent mixture second oxidation is due to the increased hydricity rification, because the cycloisomerization cat-
to the reaction (38). Presumably, under these con- of the C10–H bond of the hemiacetal interme- alyst was copolar with 24 on silica (39).
ditions, the catalyst is oxidized to RuO4, which, in diate. This single reaction combines the C10-O5 Having found a satisfying solution for the
turn, oxidized 24 to the desired hydroxylactone bond formation with two oxidation events, set- hydroxylactone problem, we decided to install
Fig. 4. Continuation of the synthesis and proposed mechanism for the key hydroxylactonization. (A) Synthesis. (B) Proposed mechanism. DDQ, 2,3-dichloro-5,6-dicyano-1,4-
benzoquinone; TBSCl, tert-butyldimethylsilyl chloride; DMF, dimethylformamide; Cp, cyclopentadienyl.
O O O
18. TMSOTf (4.0 equiv) 20. TMSCl (2.5 equiv)
OH Et3N (5.0 equiv) OTMS 21. CrO3 Py2 OTMS
2,6-lutidine (8.0 equiv) O O O
NBoc
DCE, 85 °C O O then O O (6.0 equiv) O O 22. 3:1 TFA/H 2O
25 NH2 N H NH
19. TBAF (3.0 equiv) NBoc H NHBoc DCM, r.t. r.t.
O (1.1 equiv) O O NBoc
THF, r.t. O O O N
OH MeS NHBoc OTMS HO (93 %, 2 steps)
Boc
(76 %) HgCl2 (2.0 equiv)
26 27 28
DMF, r.t.
(92%)
O H
O O
OH O O
O OH O H
O O OH
O O H NH2 O O
O O H NH O HO H NH2
O HO O H NH2
H NH N N
N N
O
O NH2 NH 4a HN N + N
N HO HO
correct H O H OH
HO H
OH H
H configuration
H (–)-TTX 1:1.4
29a 29b 29c 4,9-anhydro TTX (30)
Fig. 5. Completion of the synthesis. TMSOTf, trimethylsilyl triflate; DCE, 1,2-dichloroethane; Py, pyridine; TFA, trifluoroacetic acid.

the guanidine moiety next. Removal of the cycloaddition for the construction of high- 16. F. Cagide‐Fagín, R. Alonso, Eur. J. Org. Chem. 2010, 6741–6747
N-Boc protecting group with trimethylsilyl ly substituted cyclitols. For this purpose, the (2010).
17. H. Lago-Santomé, R. Meana-Pañeda, R. Alonso, J. Org. Chem.
triflate, followed by cleavage of the silyl ethers, humble C1 synthon nitromethane performed 79, 4300–4305 (2014).
one of which was transient, gave amino diol spectacularly—involved in no fewer than six 18. B. A. Mendelsohn, M. A. Ciufolini, Org. Lett. 11, 4736–4739
26. This compound was protected in situ as bond-forming and -breaking events—relaying (2009).
19. J. Chau, S. Xu, M. A. Ciufolini, J. Org. Chem. 78, 11901–11910 (2013).
the bis-trimethyl silyl ether and then under- an oxygen to C4, and embedding its nitro- 20. S. Xu, M. A. Ciufolini, Org. Lett. 17, 2424–2427 (2015).
went clean guanylation under Kishi’s condi- gen and carbon into the a-tertiary amine 21. D. Baidilov et al., Angew. Chem. Int. Ed. 57, 10994–10998 (2018).
tions to afford compound 27, which features over the course of the synthesis. Although the 22. K. Nishikawa et al., Org. Lett. 23, 1703–1708 (2021).
23. J. G. Robins, J. S. Johnson, Org. Lett. 24, 559–563 (2022).
all the atoms of TTX (Fig. 5). bicyclic isoxazoline 10 possessed the incor- 24. R. J. Bergeron, J. S. McManis, J. Org. Chem. 52, 1700–1703
In the last phase of our synthesis, we had to rect stereocenter at C4a, necessitating a late- (1987).
overcome two final obstacles: the oxidation stage epimerization, it guided the subsequent 25. D. B. Konrad, B. Kicin, D. Trauner, Synlett 30, 383–386 (2019).
26. J. Eames et al., J. Chem. Soc., Perkin Trans. 1 1999, 1095–1104
of the primary silyl ether and the epimeri- acetylide addition from the convex face of
(1999).
zation of the C4a stereocenter. A footnote in the molecule to install the desired configura- 27. R. G. Soengas, A. M. S. Silva, Eur. J. Org. Chem. 2013,
Kishi’s seminal publication suggested the tion of the C8a a-tertiary amine. Although 5022–5027 (2013).
latter presented a potential liability owing to this late-stage epimerization carried some 28. T. Mukaiyama, T. Hoshino, J. Am. Chem. Soc. 82, 5339–5342 (1960).
29. A. Hager, N. Vrielink, D. Hager, J. Lefranc, D. Trauner, Nat. Prod.
the propensity of the C5-O bond to undergo strategic risk, we had reason to believe that it Rep. 33, 491–522 (2016).
facile elimination, and all prior approaches would succeed based on the coherence of the 30. Y. Basel, A. Hassner, Synthesis 1997, 309–312 (1997).
had set this stereocenter earlier in their re- dioxa-adamantane core, which would render 31. K. N. Houk et al., J. Am. Chem. Soc. 106, 3880–3882 (1984).
32. S. Diethelm, E. M. Carreira, J. Am. Chem. Soc. 137, 6084–6096
spective syntheses (6). Nevertheless, we con- b-eliminations or retro-aldol reactions revers- (2015).
tinued by treating bis-trimethyl silyl ether ible. The alkynylation of 10 emphasizes the 33. C. Cavedon et al., Org. Lett. 23, 514–518 (2021).
27 with Collins reagent, which effected selec- utility of oxime ethers as a-tertiary amine pre- 34. B. M. Trost, Y. H. Rhee, J. Am. Chem. Soc. 121, 11680–11683 (1999).
35. F. E. McDonald, K. S. Reddy, Y. Díaz, J. Am. Chem. Soc. 122,
tive deprotection and oxidation of the pri- cursors but also highlights the need for more 4304–4309 (2000).
mary alcohol, resulting in 28 as a mixture of methodological development in this area. The 36. M. J. Zacuto, D. Tomita, Z. Pirzada, F. Xu, Org. Lett. 12,
hemiaminal isomers. This transformation Ru-catalyzed oxidative lactonization of alkyne 684–687 (2010).
37. S. Beligny, S. Eibauer, S. Maechling, S. Blechert, Angew. Chem.
could not have been performed with a more 23 represents a notable advance in establish- Int. Ed. 45, 1900–1903 (2006).
stable tert-butyl dimethyl silyl ether in place, ing the C9 and C10 hydroxylactone and should 38. B. Plietker, J. Org. Chem. 68, 7123–7125 (2003).
that is, with a guanidinylated derivative of have future applications in the synthesis of 39. D. W. Knight, I. R. Morgan, A. J. Proctor, Tetrahedron Lett. 51,
638–640 (2010).
compound 25. Crude 28 was then dissolved in other natural products such as the ginkolides 40. K. Tsuda et al., Chem. Pharm. Bull. 12, 1357–1374 (1964).
25% aqueous trifluoroacetic acid and stirred and quassinoids (42, 43). Furthermore, this 41. M. Yotsu-Yamashita, J.-H. Jang, Y. Cho, K. Konoki, Forensic
at room temperature overnight. Notably, this synthesis served as a proving ground for the Toxicol. 29, 61–64 (2011).
42. K. Nakanishi, Bioorg. Med. Chem. 13, 4987–5000 (2005).
effected the desired deprotections, epimeriza- chromoselective photochemical debenzylation 43. I. J. Curcino Vieira, R. Braz-Filho, Stud. Nat. Prod. Chem. 33,
tion, and cyclizations and gave a 1:1.4 mixture and recognizes the strategic value of imple- 433–492 (2006).
44. D. B. Konrad et al., Supplementary NMR-package for
of TTX and 4,9-anhydro TTX (30) in good menting new technology and methods in highly
“A Concise Synthesis of Tetrodotoxin,” Zenodo (2022).
yield. We observed the formation of 30 in un- complex settings. https://doi.org/10.5281/zenodo.6629561.
usually high proportions, which could be ex- Taken together, these strategic decisions re-
plained by the guanidine participating in the sulted in one of the shortest and the most AC KNOWLED GME NTS
We thank I. Žamarija, A. Novak, C. Wanzke, K. Schwärzer, E. Miller,
epimerization process. Intramolecular con- efficient syntheses of tetrodotoxin to date, ac-
R. Bechtel, N. Kurrle, and B. Kicin for experimental assistance and
densation of the N3 nitrogen with the C4a al- complishing this goal in 22 total steps and 11% P. Mayer for x-ray analyses. Funding: We gratefully acknowledge
dehyde would form a highly stabilized iminium overall yield from commercially available start- financial support by the NSF (CHE-1900154). D.B.K. thanks the
cation 29a, which could undergo elimina- ing materials. Our route is scalable and can Friedrich-Ebert-Stiftung for a doctoral scholarship. K.P.R. and B.E.H.
are supported by an NYU MacCracken Fellowship. H.A. was
tion to enamine 29b. Protonation to the more be adapted to the production of other scarce supported by the JSPS Overseas Challenge Program for Young
thermodynamically favored iminium 29c tetrodotoxin derivatives to better understand Researchers. Author contributions: D.T. conceived and directed
situates the electrophilic C4 in close proximity their biosynthesis and chemical ecology. It is the project. B.S.M., D.B.K., K.P.R., and D.T. conceptualized and
designed the synthetic strategy. B.S.M., D.B.K., K.P.R., H.A., and
to the C9 hydroxy group, which can attack at also amenable for the procurement of TTX B.E.H. performed and analyzed the synthetic chemistry experiments.
a rate that is kinetically competitive with derivatives that could serve as next-generation N.S. and K.N.H. performed computational studies. B.S.M., D.B.K.,
solvolysis. We believe that the elimination and analgesics. K.P.R., B.E.H., and D.T. prepared the manuscript. B.S.M., D.B.K., and
D.T. acquired funding for the project. Competing interests: The
tautomerization reactions are likely driven authors declare that they have no competing interests. Data and
by unfavorable syn-pentane interactions be- materials availability: Crystallographic data for compounds
RE FERENCES AND NOTES
tween the C6 oxygen and the C4 iminium. 11 (CCDC 2117035) and 14 (CCDC 2117036) are available free of
1. J. Chau, M. A. Ciufolini, Mar. Drugs 9, 2046–2074 (2011). charge from the Cambridge Crystallographic Data Centre. Zipped
However, given the multitude of transforma- 2. T. Nishikawa, M. Isobe, Chem. Rec. 13, 286–302 (2013). free induction decay files for the NMR spectra of compounds 24, 25,
tions that are taking place in this final step, 3. M. Makarova, L. Rycek, J. Hajicek, D. Baidilov, T. Hudlicky, 26, 27, S4, S5, and TTX have been deposited in Zenodo (44).
it is difficult, if not impossible, to pinpoint the Angew. Chem. Int. Ed. 58, 18338–18387 (2019). License information: Copyright © 2022 the authors, some rights
4. C. T. Hanifin, Mar. Drugs 8, 577–593 (2010). reserved; exclusive licensee American Association for the
exact sequence of events, which could take 5. E. G. Moczydlowski, Toxicon 63, 165–183 (2013). Advancement of Science. No claim to original US government works.
place in parallel and converge on TTX. TTX 6. Y. Kishi et al., J. Am. Chem. Soc. 94, 9219–9221 (1972). http://www.science.org/about/science-licenses-journal-article-reuse
and 30 are known to be in equilibrium with 7. N. Ohyabu, T. Nishikawa, M. Isobe, J. Am. Chem. Soc. 125,
8798–8805 (2003). SUPPLEMENTARY MATERIALS
one another, favoring TTX, and they can be 8. A. Hinman, J. Du Bois, J. Am. Chem. Soc. 125, 11510–11511 (2003).
readily interconverted (7, 40). Indeed, upon 9. T. Nishikawa, D. Urabe, M. Isobe, Angew. Chem. Int. Ed. 43,
science.org/doi/10.1126/science.abn0571
heating of this mixture for 3 days as a solution 4782–4785 (2004).
NMR spectra
10. D. Urabe, T. Nishikawa, M. Isobe, Chem. Asian J. 1, 125–135 (2006).
in 5% d3-AcOD–95% D2O to 60°C, a 2.9:1 ratio X-ray data
11. K. Sato et al., J. Org. Chem. 70, 7496–7504 (2005).
of TTX and anhydro-TTX (30) was obtained Calculations
12. K. Sato et al., J. Org. Chem. 73, 1234–1242 (2008).
Figs. S1 to S12
(see supplementary materials). TTX and 30 13. S. Akai et al., Bull. Chem. Soc. Jpn. 83, 279–287 (2010).
Tables S1 to S5
have been separated on an analytical scale (41). 14. T. Maehara, K. Motoyama, T. Toma, S. Yokoshima,
T. Fukuyama, Angew. Chem. Int. Ed. 56, 1549–1552 (2017).
Taking stock of our strategic disconnections, 15. K. Murakami, T. Toma, T. Fukuyama, S. Yokoshima, Submitted 31 October 2021; accepted 15 June 2022
our route showcases the power of the Huisgen Angew. Chem. Int. Ed. 59, 6253–6257 (2020). 10.1126/science.abn0571

BIOGEOGRAPHY also have narrower thermoneutral zones than

their temperate counterparts (6Ð9, but see
Interspecific competition limits bird speciesÕ ranges 10, 11). Janzen’s hypothesis has taken on new
relevance in the era of climate change as it
in tropical mountains predicts that, because tropical species’ distribu-
tions are limited by physiological adaptation
Benjamin G. Freeman1,2*, Matthew Strimas-Mackey3, Eliot T. Miller3 to temperature, they will have disproportion-
ately strong distributional responses to warm-
Species’ geographic ranges are limited by climate and species interactions. Climate is the prevailing ing temperatures. This geographic prediction
explanation for why species live only within narrow elevational ranges in megadiverse biodiverse tropical is met: Tropical montane species are track-
mountains, but competition can also restrict species’ elevational ranges. We test contrasting predictions ing temperature increases with upslope range
of these hypotheses by conducting a global comparative test of birds’ elevational range sizes within shifts much more closely than temperate mon-
31 montane regions, using more than 4.4 million citizen science records from eBird to define species’ tane species (12).
elevational ranges in each region. We find strong support that competition, not climate, is the leading However, there is an alternative explanation
driver of narrow elevational ranges. These results highlight the importance of species interactions for narrow elevational ranges in the tropics
in shaping species’ ranges in tropical mountains, Earth’s hottest biodiversity hotspots. that emphasizes the importance of species
interactions rather than climate. At the same
S
time that Janzen put forth his pioneering ideas
pecies that live on tropical mountains The dominant explanation for narrow eleva- about temperature seasonality, other research-
usually occur in narrow elevational tional ranges in the tropics is that they are the ers were arguing that interspecific competition
ranges, whereas species in temperate result of physiological adaptation to the low- could limit tropical montane species’ eleva-
mountains tend to have broader ele- temperature seasonality of tropical climates. tional ranges (13, 14). In this view, historical
vational ranges (1, 2). This pattern is Temperatures range from hot in the lowlands and ecological factors explain why large and
important for determining global patterns of to cold in the highlands but are relatively con- topographically complex tropical montane re-
biodiversity because the notable species rich- stant at any single elevation, unlike temperate gions have accumulated exceptional biodiver-
ness observed in many tropical mountains is mountains which experience seasonal temper- sity over long time scales (15, 16). The buildup
due to nearly complete species turnover be- ature fluctuations. Janzen (4) was the first to of high regional species richness is hypothe-
tween low and high elevations (1–3). It is not describe how stable temperatures in the tropics sized to result in intense interspecific compe-
known, however, whether elevational ranges could shape species’ elevational ranges. He tition that constrains species to narrow ranges
in the tropics are more constrained by climate hypothesized that tropical species experience despite their ability to live in a broader array
or species interactions. selection to physiologically adapt to the par- of environments. That is, species have narrow
ticular thermal conditions they experience, re- realized niches rather than narrow fundamen-
sulting in the evolution of narrow fundamental tal niches. The strongest evidence presented
niches that manifest as restricted elevational for the interspecific competition hypothesis has
1
Biodiversity Research Centre, University of British Columbia, ranges (4, 5). In support of this hypothesis, been case examples of “natural experiments”
Vancouver, BC V6T 1Z4, Canada. 2Department of Zoology, tropical species have thermal tolerances that that compared species’ elevational ranges in
University of British Columbia, Vancouver, BC, Canada.
3
Cornell Lab of Ornithology, Ithaca, New York, USA, 14850. tend to match the temperature conditions they montane regions where they were sympatric
*Corresponding author. Email: [email protected] experience within their elevational zone, and versus allopatric with a closely related species.
1 17. Pyrenees
18 18. European Alps
2 4
3 17 19. Western Himalaya
1. Northern Rockies 5
2. Cascades 20 21 20. Eastern Himalaya
3. Siskiyous 19 21. Japanese Alps
6 22
4. Northeastern North America 8 22. Taiwan
9 7 23
5. Appalachians 23. Luzon
24
6. Cuba 10 11 25 26 27 24. Western Ghats
12 25. Sri Lanka
7. Jamaica 13 28
8. Hispaniola 26. Peninsular Malaysia
9. Mesoamerica 14 29 27. Mindanao
10. Southern Central America 15 28. Northern Sulawesi
1 1. Northern Tropical Andes 29. Northeastern Madagascar
12. Chocó 30 30. Australian Alps
13. Central Tropical Andes 16 31 31. Southern Alps
14. Southern Tropical Andes
15. Atlantic Forest
16. Southern Temperate Andes
Fig. 1. Map of 31 montane regions included in this study. We included montane regions that had elevational gradients of ≥1400 m, natural forest vegetation along
the entire elevational gradient, and sufficient fine-scale distributional data from eBird to define speciesÕ regional elevational ranges (mean incidence records per region
= 235,438; mean incidence records per 100 m elevational band = 8426).

In many instances such species were reported the globe (Fig. 1). We defined species’ eleva- tropical regions varied by a factor of 20 in
to be “elevational replacements” in sympatry tional distributions within each region using their species richness such that the correla-
(i.e., inhabiting narrow nonoverlapping ele- 4.4 million fine-scale locality records from tion between temperature seasonality and re-
vational ranges) whereas in allopatry they eBird, a global citizen science project (17); all gional species richness was relatively weak
were reported to live within expanded eleva- told, our dataset contains elevational ranges [fig. S2; Spearman’s r = −0.46 (95% CI =
tional ranges. These patterns were inter- for 5397 unique species-by-region combina- −0.76 to -0.16, P = 0.0090)]. This allowed us
preted as indicating competitive release in tions (see table S1 for information on regions to statistically disentangle temperature sea-
allopatry and competitive exclusion in sym- and fig. S1 for illustration of how we defined sonality from regional species richness.
patry (13, 14). However, there have been no species’ elevational ranges within regions). We tested contrasting predictions of Janzen’s
general tests of the interspecific competition Regions ranged in latitude from 43°S to 52°N hypothesis and the interspecific competition
hypothesis. and in species richness from 23 to 618 species; hypothesis. Janzen’s hypothesis predicts that,
We provide a global test of contrasting hy- species richness is defined as the total num- all else equal, elevational range sizes are nar-
potheses to explain why tropical species have ber of forest bird species within a given re- rower in regions with reduced temperature
narrow elevational ranges: Janzen’s hypoth- gion in the eBird dataset. As expected, the seasonality whereas the interspecific compe-
esis, which emphasizes abiotic controls on regions with the highest species richness were tition hypothesis predicts that, all else equal,
biodiversity, and the interspecific competition all located in the tropics and had low temper- elevational range sizes are narrower when
hypothesis, which emphasizes biotic controls ature seasonality [absolute latitude and tem- regional species richness is high. These two
on biodiversity. We conducted a comparative perature seasonality were tightly correlated; hypotheses are not mutually exclusive. We
analysis of forest bird species’ elevational Spearman’s r = 0.93; 95% confidence intervals therefore used the relative explanatory power
ranges within 31 montane regions across (CI) = 0.86 to 0.99; P << 0.001]. However, of one predictor variable versus another to
A Predictions B Data
Janzen’s hypothesis Competition hypothesis
Temperature Species Temperature Species Temperature Species

seasonality richness seasonality richness seasonality -0.36 richness
Negative 0.037 -0.88

Positive
Mean elevational range size Mean elevational range size Mean elevational range size
C D
2000 2000
(m)size (m)
Partial residual elevational range size (m)
elevational range size (m)
range
1500 1500
range size
residual elevational
1000 1000
Partialelevational
500 500
0 0
0 2500 5000 7500 10000 0 200 400 600
temperature seasonality
Temperature (bio4)
seasonality regional
Regionalspecies
speciesrichness
richness
Fig. 2. Predictions and data for the contrasting hypotheses examined in analysis with standardized variables and standard errors: regional species richness =
this study. (A) Janzen’s hypothesis emphasizes climatic controls on species’ −0.88 ± 0.17, temperature seasonality = 0.037 ± 0.12; multiple regression parameter
elevational range sizes whereas the interspecific competition hypothesis estimates and standard errors: regional species richness = −1.39 ± 0.37,
emphasizes biotic controls on species’ elevational range sizes. We used data from temperature seasonality = 0.023 ± 0.17). Trendlines in (C) and (D) illustrate
31 regions to test these nonÐmutually exclusive predictions by (B) path analysis expected values with 95% CIs shown in gray shading; points show partial residuals
and (C and D) multiple regression. Elevational range size is better predicted from a multiple regression that included regional species richness, temperature
by regional species richness than temperature seasonality in both analyses (path seasonality, and methodological covariates as predictor variables.

assess the relative importance of the two pos- seasonality. In a path analysis, regional species species richness was much stronger (Fig. 2B;
sible mechanisms. All models were multivariate richness was negatively associated with eleva- regional species richness parameter estimate
and included mountain height, sampling com- tional range size (−0.88 ± 0.17; parameter esti- and standard error = −1.39 ± 0.37, t = −3.74,
pleteness, and climate change velocity as addi- mate and standard error) whereas temperature P = 0.00097; Cohen’s f 2 = 0.56) than evidence
tional predictor variables. seasonality was unrelated to elevational range for an effect of temperature seasonality (Fig. 2C;
We found regional species richness to be a size (0.037 ± 0.12; Fig. 2B, fig. S3, and tables S2 temperature seasonality parameter estimate
better predictor of mean elevational range size to S4). Similarly, in a multiple regression the and standard error = 0.023 ± 0.017, t = 1.31,
of species within a region than temperature evidence for an effect resulting from regional P = 0.20; Cohen’s f 2 = 0.069; see table S3).
A B C All species D Shared species
Elevational range size (m)

richness
species richness
Elevation (m)
elevation (m)
Chocó
Latitude
slope
Latitude
Amazon
Species
slope
Choco Amazon Choco Amazon Choco Amazon

Longitude
Longitude Chocó Amazon Chocó Amazon Chocó Amazon
Fig. 3. A case example of how increased regional species richness is associated with narrower elevational ranges. (A) At the equator, the tropical Andes are
divided into two biogeographic regions: the western Chocó slope and the eastern Amazon slope. (B) Species richness of forest birds is high in both regions but
is higher on the Amazon slope. Elevational range sizes are larger in the Chocó when considering both (C) all species in our analysis (n = 618 species in the Amazon and
n = 498 species in the Chocó) and (D) the 278 shared species that live in both regions. Effect sizes for (C) and (D) are Cohen’s d = 0.61 (95% CI = 0.48 to 0.73) and
Cohen’s d = 0.43 (95% CI = 0.26 to 0.60), respectively.
A Sympatric region B Allopatric region C 1.00

D
Proportion showing competitive release
Proportion showing competitive release

1.00
No competitive Competitive
release release
0.75 0.75
0.50 0.50
Upper Upper Upper
species species species
0.25 0.25
Lower
species 0.00 0.00
no yes lower upper

Territorial behavior Allopatric region
Fig. 4. Pairwise competition between elevational replacements is competitive release in environmental space; inferences of competitive release in
one mechanism by which higher regional species richness creates stronger elevation versus environmental space were tightly correlated [table S4]). (C)
interspecific competition that limits speciesÕ elevational ranges. (A) Eleva- Competitive release was more likely to occur when species defended territories,
tional replacements are pairs of closely related species that “replace” one another consistent with the hypothesis that behavioral competition limits elevational
along mountain slopes in sympatric regions. (B) We tested for competitive release in ranges in sympatry (P = 0.023). However, competitive release was not more likely
allopatric regions, where one species of elevational replacement is not present (A). when the upper species lived in the allopatric region (D), in contrast to the
We asked whether species expanded their distributions in allopatry (with effect longstanding idea that competition limits warm range limits more so than cold
sizes Hedge’s g > 0.20) to inhabit elevations and positions within multivariate range limits (P = 1; see figs. S7 to S58 for detailed results of each comparison).
environmental space that in the sympatric region are inhabited by their Effect sizes for (C) and (D) are Cramer’s V = 0.34 (95% CI: 0.086 to 0.59) and
replacement (sample size = 52 comparisons, see fig. S3 for illustration of Cramer’s V = 0.013 (95% CI: 0.0014 to 0.31).

Results were similar in an alternative phylo- ecological factors contribute to the observed Field studies measuring interspecific terri-
genetic mixed model that analyzed each specific patterns. toriality can test this interpretation. By con-
species-by-site combination and incorporated To shed light on how greater regional spe- trast, competitive release in allopatry was not
phylogenetic nonindependence between species richness generates increased interspecific associated with the relative position of species’
cies and spatial nonindependence between competition that results in narrow elevational elevational ranges [Fig. 4D, Fisher’s exact test,
regions (table S5). Limited sampling can lead ranges, we examined the mechanism that orig- P = 1; Cramer’s V = 0.013 (95% CI: 0.0014 to
to the systematic underestimation of species’ inally inspired the interspecific competition 0.31); see figs. S7 to S58 for results of individual
elevational range sizes in lowland and foothill hypothesis: pairwise competition between natural experiments].
tropical regions where diversity is highest (18), “elevational replacements”, defined as closely Tropical mountains are home to the greatest
but our analysis of sampling completeness related species that inhabit different eleva- concentration of terrestrial biodiversity on
finds patterns opposite of that expected if tional zones in sympatry. Elevational replace- Earth because species live only within narrow
sampling bias affects our results (fig. S4). ments are common in a wide range of tropical elevational zones, creating high species turn-
The greater importance of regional species taxa, but have been best studied in tropical over along mountain slopes. The prevailing
richness compared with temperature season- birds (13, 14). We followed previous research- hypothesis for why tropical species live in
ality holds in a species-level analysis. Species ers by analyzing natural experiments, compar- narrow elevational ranges is that they have
that live within multiple regions in our dataset ing species’ ranges in locations where a given evolved physiological adaptations to specific
tended to have smaller elevational ranges in species lives in sympatry with a putative com- thermal conditions, ultimately leading to the
more diverse regions but did not tend to have petitor with locations where the species lives buildup of high species richness in tropical
smaller elevational ranges in regions with lower in allopatry (19–21). However, we wanted to test mountains. We present evidence that overturns
temperature seasonality. For 176 species found for general patterns, which required moving this explanation—our analysis of a global data-
in five or more regions, the average correlation beyond the handful of previously examined set of millions of citizen science data records
between elevational range size and regional spe- case examples that likely suffer from serious reveals that the narrow elevational ranges of
cies richness was negative [mean Spearman’s r = ascertainment bias. We therefore conducted tropical birds are driven more by species in-
−0.17 (95% CI = −0.23 to −0.095), degrees of free- a comparative analysis of all 52 natural ex- teractions than by the direct effects of climate.
dom (df) = 175, t = −4.58, P = 0.0000087], but the periments in Neotropical birds that met pre- Whether the patterns we demonstrate gener-
average correlation between elevational range defined criteria, using eBird data to define alize to other taxa, particularly ectotherms, is a
size and temperature seasonality was close to species’ elevational distributions in sympatric key unanswered question. Regardless, our re-
zero [mean Spearman’s r = -0.050 (95% CI −0.13 and allopatric regions. sults suggest a new interpretation for why
to 0.027), df = 175, t = −1.29, P = 0.22]. We found evidence for competitive release tropical montane birds (and potentially other
We illustrate these patterns by highlighting in nearly half of natural experiments: in 23 of tropical taxa) are shifting noticeably upslope
two regions on opposite slopes of the Andes at 52 cases, species in allopatry inhabited eleva- (12): warming likely causes these upslope
the equator: the western (Chocó) versus the tional zones and positions within environmental shifts indirectly, by altering the outcomes
eastern (amazonian) slope (Fig. 3). These ad- space that in sympatry are occupied by their of species interactions, rather than directly
jacent slopes are at the same latitude and have close relative (elevational release inferred when through physiological stress. In this view, it is
similar climates and elevational relief, but are Hedges’ g > 0.20; Fig. 4B, figs. S5 and S6, and biodiversity itself that makes Earth’s hottest
biogeographically distinct and differ in spe- table S6). We then tested whether these eleva- biodiversity hotspots disproportionately respon-
cies richness. The Chocó slope is a center of tional shifts in allopatry had consequences for sive to climate change.
endemism and highly biodiverse but has fewer elevational range sizes in allopatry. For cases in
species of forest birds than the amazonian which we inferred competitive release, elevational REFERENCES AND NOTES
slope (in our dataset, 498 versus 618 species of ranges expanded in allopatry [fig. S5B; mean 1. C. M. McCain, Ecol. Lett. 12, 550–560 (2009).
forest-dwelling birds; Fig. 3B). Consistent with elevational range expansion = 330 m (95% CI: 157 2. C. D. Cadena et al., Proc. Biol. Sci. 279, 194–201 (2012).
predictions of the interspecific competition to 504 m)], but there was no change in elevational 3. I. Quintero, W. Jetz, Nature 555, 246–250 (2018).
4. D. H. Janzen, Am. Nat. 101, 233–249 (1967).
hypothesis, elevational ranges are narrower range size between sympatry and allopatry for 5. C. K. Ghalambor, R. B. Huey, P. R. Martin, J. J. Tewksbury,
on the amazonian slope than the Chocó slope cases in which we did not infer competitive re- G. Wang, Integr. Comp. Biol. 46, 5–17 (2006).
[Fig. 3C; average elevational range sizes = lease [mean elevational range expansion = 2 m 6. N. R. Polato et al., Proc. Natl. Acad. Sci. U.S.A. 115,
12471–12476 (2018).
1114 m versus 1475 m, respectively; df = 971.9, (95% CI: −178 to 187 m)]. Thus, elevational re- 7. J. M. Sunday, A. E. Bates, N. K. Dulvy, Nat. Clim. Chang. 2,
t = 9.91, P < 2.2−16; Cohen’s d = 0.61 (95% CI = placements often—but certainly not always— 686–690 (2012).
0.48 to 0.73)]. Again, this pattern is replicated have narrow elevational ranges in sympatry that 8. A. A. Shah et al., Funct. Ecol. 31, 2118–2127 (2017).
9. B. A. Gill et al., Proc. Biol. Sci. 283, 20160553 (2016).
in a species-level analysis. For 278 shared spe- appear to be limited by pairwise competition. 10. G. A. Londoño, M. A. Chappell, J. E. Jankowski, S. K. Robinson,
cies found in both regions, elevational ranges Why do some elevational replacements show Funct. Ecol. 31, 204–215 (2017).
were more narrow on average on the amazo- competitive release in allopatry whereas others 11. H. S. Pollock, J. D. Brawn, Z. A. Cheviron, Funct. Ecol. 35,
93–104 (2021).
nian slope than on the Chocó slope [Fig. 3D; do not? We tested two hypotheses: (i) behav-
12. B. G. Freeman, Y. Song, K. J. Feeley, K. Zhu, Ecol. Lett. 24,
1354 m versus 1602 m; df = 277, t = −6.39, P = ioral interactions are a key mechanism by which 1697–1708 (2021).
7.1−10; Cohen’s d = 0.43 (95% CI = 0.26 to 0.60)]. competition restricts elevational ranges (22–26) 13. J. M. Diamond, Science 179, 759–769 (1973).
The comparison of the Chocó versus Amazon and (ii) competition restricts species’ warm 14. J. Terborgh, J. S. Weske, Ecology 56, 562–576 (1975).
15. C. Rahbek et al., Science 365, 1108–1113 (2019).
slopes illustrates our broader result, that re- range limits more than their cool range limits 16. D. Schluter, M. W. Pennell, Nature 546, 48–55 (2017).
gional species richness predicts forest bird (27–30). Consistent with the hypothesis that 17. B. L. Sullivan et al., Biol. Conserv. 142, 2282–2292
species’ elevational range sizes within and behavior is an important mechanism limiting (2009).
18. R. K. Colwell, G. C. Hurtt, Am. Nat. 144, 570–595 (1994).
between regions. We interpret this as indicat- ranges in elevational replacements, competi- 19. R. P. Anderson, M. Gomez-Laverde, A. T. Peterson, Glob. Ecol.
ing that general patterns of elevational spe- tive release in allopatry is 2.4 times more likely Biogeogr. 11, 131–141 (2002).
cialization in birds are better explained by the when species defend territories (18 out of 20. E. E. Gutiérrez, R. A. Boria, R. P. Anderson, Ecography 37,
741–753 (2014).
interspecific competition hypothesis than by 31 cases) than when they do not [5 out of 21. C. D. Cadena, B. A. Loiselle, Ecography 30, 491–504
Janzen’s hypothesis, while also acknowledg- 21 cases; Fig. 4C, Fisher’s exact test, P = 0.023; (2007).
ing that multiple historical, evolutionary, and Cramer’s V = 0.34 (95% CI: 0.086 to 0.59)]. 22. B. G. Freeman, Glob. Ecol. Biogeogr. 29, 171–181 (2020).

23. J. E. Jankowski, S. K. Robinson, D. J. Levey, Ecology 91, ACKN OWLED GMEN TS reserved; exclusive licensee American Association for the
1877–1884 (2010). S. Freeman, M. Pennell, L. Rowe, D. Schluter, O. Robinson, Advancement of Science. No claim to original US government
24. B. Pasch, B. M. Bolker, S. M. Phelps, Am. Nat. 182, E161–E173 R. Colwell, and an anonymous reviewer provided comments that works. https://www.sciencemag.org/about/science-licenses-
(2013). greatly improved the manuscript. We are indebted to eBird journal-article-reuse
25. B. G. Freeman, J. A. Tobias, D. Schluter, Ecography 42, users across the globe, as well as the Cornell Lab of Ornithology
1832–1840 (2019). team that developed and maintains this phenomenal project. SUPPLEMENTARY MATERIALS
26. D. L. Altshuler, Am. Nat. 167, 216–229 (2006). Funding: This work was funded by the following: Natural
science.org/doi/10.1126/science.abl7242
27. C. Darwin, On the Origin of Species by Means of Natural Sciences and Engineering Council 379958 and National Science
Selection (Murray, 1859). Foundation 1523695 (to B.G.F.). Author contributions:
Figs. S1 to S58
28. R. H. MacArthur, Geographical Ecology: Patterns in the Conceptualization: B.G.F., E.T.M., and M.S.-M. Analysis: B.G.F.,
Tables S1 to S6
Distribution of Species (Princeton Univ. Press, 1972). M.S.-M., and E.T.M. Writing: B.G.F., E.T.M., and M.S.-M.
References (31Ð47)
29. A. M. Louthan, D. F. Doak, A. L. Angert, Trends Ecol. Evol. 30, Competing interests: Authors declare that they have no
780–792 (2015). competing interests. Data and materials availability: Data Submitted 31 July 2021; resubmitted 2 January 2022
30. A. Paquette, A. L. Hargreaves, Ecol. Lett. 24, 2427–2438 and code are available at: 10.5281/zenodo.6450245. License Accepted 7 June 2022
(2021). information: Copyright © 2022 the authors, some rights 10.1126/science.abl7242
CORONAVIRUS vidual amino acid mutations in SARS-CoV-2

variant RBDs.
Shifting mutational constraints in the SARS-CoV-2 We constructed comprehensive site-saturation
mutagenesis libraries in the ancestral Wuhan-Hu-1
receptor-binding domain during viral evolution RBD (201 residues) and RBDs from four vari-
ants: Alpha (N501Y), Beta (K417N+E484K+N501Y),
Tyler N. Starr1*†, Allison J. Greaney1,2,3†, William W. Hannon1,4, Andrea N. Loes1,5, Kevin Hauser6, Delta (L452R+T478K), and Eta (E484K). We
Josh R. Dillen6, Elena Ferri6, Ariana Ghez Farrell1, Bernadeta Dadonaite1, Matthew McCallum7, cloned these mutant libraries into a yeast-
Kenneth A. Matreyek8, Davide Corti9, David Veesler5,7, Gyorgy Snell6, Jesse D. Bloom1,2,5* surface display platform and determined the
impact of every amino acid mutation on ACE2
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has evolved variants with binding affinity and yeast surface-expression
substitutions in the spike receptor-binding domain (RBD) that affect its affinity for angiotensin- levels by means of fluorescence-activated cell
converting enzyme 2 (ACE2) receptor and recognition by antibodies. These substitutions could sorting (FACS) and high-throughput sequenc-
also shape future evolution by modulating the effects of mutations at other sites—a phenomenon ing (figs. S1 and S2 and data S1) (2). The effect
called epistasis. To investigate this possibility, we performed deep mutational scans to measure of each mutation on ACE2 binding is shown in
the effects on ACE2 binding of all single–amino acid mutations in the Wuhan-Hu-1, Alpha, Beta, Fig. 1, and an interactive version of this figure
Delta, and Eta variant RBDs. Some substitutions, most prominently Asn501→Tyr (N501Y), cause is available at https://jbloomlab.github.io/SARS-
epistatic shifts in the effects of mutations at other sites. These epistatic shifts shape subsequent CoV-2-RBD_DMS_variants/RBD-heatmaps. We
evolutionary change—for example, enabling many of the antibody-escape substitutions in the used monomeric ACE2 ectodomain to mea-
Omicron RBD. These epistatic shifts occur despite high conservation of the overall RBD structure. sure 1:1 binding affinities, which provide more
Our data shed light on RBD sequence-function relationships and facilitate interpretation of ongoing granularity to reveal affinity-enhancing effects
SARS-CoV-2 evolution. compared with our previous measurements
using the natively dimeric ACE2 ligand, where
T
some mutational effects are masked by avidity
he severe acute respiratory syndrome For example, we identified the N501Y mutation (fig. S1F) (2). Mutant effects on ACE2 binding
coronavirus 2 (SARS-CoV-2) spike receptor- as enhancing ACE2-binding affinity before the and protein expression in yeast-displayed RBD
binding domain (RBD) has evolved emergence of this consequential mutation in have been shown to closely correlate with ACE2
rapidly since the virus emerged (1). We the Alpha variant (3). (Single-letter abbrevia- binding and protein expression in the context
previously used deep mutational scan- tions for the amino acid residues are as follows: of full spike trimers displayed on mammalian
ning to experimentally measure the impact A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, cells (9, 10).
of all single–amino acid mutations on the His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; We identified sites where the impacts of
angiotensin-converting enzyme 2 (ACE2)– Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; mutations differ between RBD variants (Fig. 2
binding affinity of the ancestral Wuhan-Hu-1 and Y, Tyr. In the mutants, other amino acids and figs. S3 and S4), reflecting epistasis among
RBD (2). These measurements have helped in- were substituted at certain locations; for ex- the substitutions that distinguish SARS-CoV-2
form surveillance of SARS-CoV-2 evolution. ample, N501Y indicates that asparagine at variants and other mutations across the RBD.
position 501 was replaced by tyrosine.) These epistatic shifts in mutational effects
1
However, as proteins evolve, the impacts of on ACE2 binding are primarily attributable to
Basic Sciences Division, Fred Hutchinson Cancer Research
Center, Seattle, WA 98109, USA. 2Department of Genome
individual amino acid mutations can shift, a the N501Y mutation: The effects of mutations
Sciences, University of Washington, Seattle, WA 98109, USA. phenomenon known as epistasis (4). For ex- in the Delta (L452R+T478K) and Eta (E484K)
3
Medical Scientist Training Program, University of Washington, ample, the same N501Y mutation that enhances RBDs are similar to those in the ancestral
Seattle, WA 98109, USA. 4Molecular and Cellular Biology
SARS-CoV-2 binding to ACE2 severely impairs Wuhan-Hu-1 RBD, and the differences in
Graduate Program, University of Washington, Seattle, WA
98109, USA. 5Howard Hughes Medical Institute, Seattle, WA ACE2 binding by SARS-CoV-1 and other diver- the Beta (K417N+E484K+N501Y) RBD largely
98109, USA. 6Vir Biotechnology, San Francisco, CA 94158, gent sarbecoviruses (5). Furthermore, N501Y recapitulate those in the Alpha RBD that con-
USA. 7Department of Biochemistry, University of Washington, epistatically enabled other affinity-enhancing tain N501Y alone (Fig. 2, A and B). One excep-
Seattle, WA 98195, USA. 8Department of Pathology, Case
Western Reserve University School of Medicine, Cleveland, OH mutations that emerged in the Omicron var- tion is a distinct epistatic shift in the effects of
44106, USA. 9Humabs BioMed SA, a subsidiary of Vir iant of SARS-CoV-2 (6–8). To more system- mutations to serine or threonine at site 419
Biotechnology, 6500 Bellinzona, Switzerland. atically understand how epistasis shifts the in the Beta RBD that introduce an N-linked
*Corresponding author. Email: [email protected] (T.N.S.);
[email protected] (J.D.B.) effects of mutations, we performed deep muta- glycosylation motif when an asparagine is
†These authors contributed equally to this work. tional scans to measure the impacts of all indi- present through the K417N mutation (fig. S3D).

Wuhan-Hu-1
RK
KR
x x x x x x x x x
x x x x x x x x x x
HD
DH
x
x x x x x x x x x
E x x x x x x E
QN NQ
x x x x x x
x x x x x x x x x x x x x x x x x x
ST TS
x x x x x x x x x x x x x x x x
x x x x x x x x x x x x
YW WY
x x x x x x x x x x x x x x x
x x
FA A FI
x x x x x x x x x x x x x x
IL x
x
x
x x x
x x x
x
x
x x x x
x x
x x x x L
MV x x x x x x x x x x x x x x x x VM
GP PG
x x x x x x x x x x x
C x x x x x x x x C
Alpha (N501Y)
RK
KR
x x x x x x x x x
x x x x x x x x x x
HD
DH
x
x x x x x x x x x
E x x x x x x E
QN NQ
x x x x x x
x x x x x x x x x x x x x x x x x
ST TS
YW WY
x x
FA A FI
IL x
x
x
x x x
x x x
x
x
x x x x
x x
x x x x L
GP PG
C x x x x x x x x C log10(KD)
Beta (K417N+E484K+N501Y)
RK 2
KR
x x x x x x x x x
x x x x x x x x x x
affinity
higher
HD
DH
x
x x x x x x x x x
E x x x x x E
QN NQ
x x x x x x
0
ST TS
YW WY
x x
FA A FI
affinity
lower
IL x
x
x
x x x
x x x
x
x
x x x x
x x
x x x x L -2
GP PG
C x x x x x x x x C
-4
Delta (L452R+T478K)
RK
KR
x x x x x x x x x x
x x x x x x x x x x x n.d.
HD
DH
x
x x x x x x x x x
E x x x x x x E
QN NQ
x x x x x x
ST TS
YW WY
x x
FA A FI
IL x
x
x
x x x
x x x
x
x
x x x
x x
x x x x L
GP PG
C x x x x x x x x C
Eta (E484K)
RK
KR
x x x x x x x x x
HD
DH
x
x x x x x x x x x
E x x x x x E
QN NQ
x x x x x x
ST TS
YW WY
x x
FA A FI
IL x
x
x
x x x
x x x
x
x
x x x x
x x
x x x x L
GP PG
C x x x x x x x x C
331
335
340
345
350
355
360
365
370
375
380
385
390
395
400
405
410
415
420
425
430
435
440
445
450
455
460
465
470
475
480
485
490
495
500
505
510
515
520
525
530
RBD site
Fig. 1. Deep mutational scanning maps of ACE2-binding affinity for all acid in each variant is indicated with an “x”, and gray squares indicate
singleÐamino acid mutations in five SARS-CoV-2 RBD variants. The impact missing mutations in each library. An interactive version of this map is at
on ACE2 receptor-binding affinity [Dlog10(Kd), where Kd is the dissociation https://jbloomlab.github.io/SARS-CoV-2-RBD_DMS_variants/RBD-heatmaps,
constant] of every single–amino acid mutation in SARS-CoV-2 RBDs, as and raw data are in data S1. The effects of mutations on RBD surface
determined with high-throughput titration assays (fig. S1). The wild-type amino expression are in fig. S2.
The RBD sites that exhibit notable epistatic Hu-1), so that the double mutant has a 387-fold viral particles, suggesting that the remaining
shifts because of N501Y fall into three struc- increased binding affinity. The Q498R+N501Y Omicron RBD mutations are deleterious with-
tural groups (Fig. 2B). The largest shift in double mutation was first discovered in di- out buffering by Q498R+N501Y (Fig. 3C and
mutational effects is at the direct N501-contact rected evolution studies (6) and is present in fig. S6, A and B).
residue Q498 (Fig. 2C), together with further the RBD of the Omicron BA.1 and BA.2 vari- There is also evolutionary relevance of the
epistatic shifts at sites 491 to 496 composing ants (8). The epistasis between these two mu- epistasis of N501Y with mutations on the 446–
the central b strand of the ACE2 contact tations is crucial for enabling the Omicron RBD 449 loop, which composes the epitope for an
surface (Fig. 2B and fig. S3A). A second cluster to bind ACE2 with high affinity despite having important class of human antibodies (15, 16).
of sites exhibiting epistatic shifts in the pres- a large number of mutations (11–13). Specif- Although mutations to G446 escape this class
ence of N501Y include 446, 447, and 449, which ically, the set of mutations in the Omicron of antibodies in the Wuhan-Hu-1 RBD (16, 17),
do not directly contact N501 but are spatially RBD are predicted to strongly impair ACE2 these mutations incur stronger ACE2-binding
adjacent to residue 498 (Fig. 2, B and C, and affinity on the basis of their summed single- deficits in the N501Y background (figs. S3B and
fig. S3B). A third group of sites that epistati- mutant effects in Wuhan-Hu-1 (Fig. 3B, left), S5D). Conversely, mutations to Y449 strongly
cally shift because of N501Y includes residue but their summed single-mutant effects in decrease ACE2-binding affinity in the Wuhan-
R403 (Fig. 2C), together with several residues the Beta background (which has N501Y) is Hu-1 RBD but are better tolerated when accom-
(505, 506, and 406) that structurally link site about zero (Fig. 3B, right), which is consistent panied by N501Y (Figs. 2C and 3D). Mutations
501 to site 403 (fig. S3C). with the actual affinity of the Omicron RBD to Y449 can escape monoclonal antibodies
Some of these epistatic shifts are of clear for ACE2. Therefore, the affinity buffer con- (fig. S6, C to E) (15, 18) and reduce neutraliza-
relevance during the evolution of SARS-CoV-2. ferred by the epistatic Q498R+N501Y pair tion by polyclonal sera (19, 20) and have been
One of the strongest epistatic shifts is the enables the Omicron spike to tolerate other described in several variants that also contain
potentiation of Q498R by N501Y (Figs. 2C and mutations that decrease ACE2 binding (Fig. 3B N501Y, including the C.1.2, A.29, and B.1.640
3A). Although Q498R alone weakly reduces and fig. S5A) but contribute to antibody escape lineages (19, 21).
ACE2 affinity in the Wuhan-Hu-1 RBD, it con- (fig. S5, B and C) (14). Consistent with these To more systematically examine how epi-
fers a 25-fold enhancement in affinity when affinity measurements, introducing R498Q and static shifts caused by N501Y affect patterns
present in conjunction with N501Y (which Y501N reversions into the Omicron BA.1 spike of sequence variation during SARS-CoV-2 evo-
itself improves binding 15-fold in Wuhan- reduces cell entry by spike-pseudotyped lenti- lution, we counted the occurrence of substitutions

on a global SARS-CoV-2 phylogeny (22). Sub- affected patterns of mutation accumulation in curred disproportionately on Y501 genomes
stitutions more often occurred in backgrounds prior SARS-CoV-2 evolution, and our data until its predominance in Omicron lineages.
that contain the amino acid at site 501 with enable identification of mutations such as We hypothesize that the strong affinity gain
which they had more favorable epistasis with those at site Y449 whose evolutionary rele- caused by the Q498R+N501Y double mutant
respect to ACE2 affinity (Fig. 3E). Therefore, vance may grow if N501Y variants continue (Fig. 3A) is not directly advantageous itself but
epistatic shifts caused by N501Y have directly to predominate. Q498R had not previously oc- rather becomes beneficial in Omicron because
A
Epistatic shift in mutational effects vs. Wuhan-Hu-1
498
0.4 Alpha
449 Beta
403 Delta
Eta
0.2
446, 491–
447 496 sites of
419 505,
506 strong
406 antibody
escape
0.0
331
335
340
345
350
355
360
365
370
375
380
385
390
395
400
405
410
415
420
425
430
435
440
445
450
455
460
465
470
475
480
485
490
495
500
505
510
515
520
525
530
RBD site
B Alpha Beta Delta Eta

N501Y K417N+E484K+N501Y L452R+T478K E484K
T478K
Epistatic shift in E484K
mutational effects vs. Y449 G446 Y449 G446
E484K
Wuhan-Hu-1 S494 S494 L452R
max
Q498
8 Q498
8
N501Y N501Y
K417N
Y505 Y505
R403 R403
A419
0
C
site 498 site 449 site 403 site 484
Binding affinity (-log10(KD)),
12
12
12
12
epistatic shift: 0.48 epistatic shift: 0.33 epistatic shift: 0.29 epistatic shift: 0.01
R
10
10
10
10
S
Q
Beta RBD
AQ Y A M KR T
S
QRK
E
W K R G NC G
K TM
H LIVW
PA
G NH
S TQ DH
E NA
C N SF Y D EM C F T MD
C
P
V P VHI
W
8
8
8
8
L
I
L FY
E
G Y L
6
6
6
6
VI
P D W F
6 8 10 12 6 8 10 12 6 8 10 12 6 8 10 12
Binding affinity (-log10(KD)), Wuhan-Hu-1 RBD
Fig. 2. Epistatic shifts in mutational effects across RBD variants. (A) The that are mutated in each RBD variant. (C) Mutation-level plots of epistatic shifts
shift in mutational effects on ACE2 binding at each RBD site between the at sites of interest. Each scatter plot shows the measured affinity of all 20 amino
indicated variant and Wuhan-Hu-1. An interactive version of this plot is at acids in the Beta versus Wuhan-Hu-1 RBD. Red dashed lines indicate the parental
https://jbloomlab.github.io/SARS-CoV-2-RBD_DMS_variants/epistatic-shifts. RBD affinities, and the gray dashed line indicates the additive (nonepistatic)
The epistatic shift is calculated as the Jensen-Shannon divergence in the set of expectation. Epistatic shifts can reflect idiosyncratic mutation-specific shifts (such
Boltzmann-weighted affinities for all amino acids at each site. Gray shading indicates as site 498) or global changes in mutational sensitivity at a site (such as site 449).
sites of strong antibody escape based on prior deep mutational scanning of the Site 484 does not have a substantial epistatic shift and is shown for comparison.
Wuhan-Hu-1 RBD (11). (B) Ribbon diagram of the Wuhan-Hu-1 RBD structure Scatterplots of additional sites of interest are available in fig. S3. Epistatic shifts in
(PDB 6M0J) colored according to epistatic shifts. Labeled spheres indicate residues mutational effects on RBD expression are available in fig. S4.

Fig. 3. Functional and evolutionary relevance of A B Single mutation effects Single mutation effects
epistatic interactions. (A) Double-mutant cycle N501Y+ measured in Wuhan-Hu-1 measured in Beta
Q498R
Binding affinity (-log10(KD))

diagram illustrating the positive epistasis interaction
Cumulative additive binding relative

T478K G339D
11
4 4 N440K S373P
G496S
S477N
to Wuhan-Hu-1 ( log10(KD))
between N501Y and Q498R. Asterisk indicates S371L
Q493R
Q498R E484A
expected double-mutant binding affinity assuming N501Y 2
G339D
T478K
S373P
Q498R 2
S375P
10
N440K E484A Y505H
additivity. (B) Affinity-buffering of Omicron BA.1 S477N
N501Y
S371L
Q493R Measured N501Y Measured
additive G446S K417N
mutations. Each diagram shows the cumulative Wuhan- Omicron Omicron
S375P affinity affinity
Hu-1 0 0 G446S
9
Wuhan-Hu-1 Wuhan-Hu-1
addition of individually measured effects on ACE2- K417N
binding affinity [Dlog10(Kd)] for each single-RBD Q498R

G496S
-2 -2
substitution in Omicron BA.1 as measured in 0 1 2 Y505H
Number mutations
the Wuhan-Hu-1 (left) or Beta (right) RBDs. Mutation Mutations ordered by effect on affinity
effect is calculated in the labeled direction even
C 100 D E
Change in mutation effect on binding

N501Y
when the reference state in a background differs; for _ _ Q498K
10
0.66× 3
N501Y/ R403G
due to N501Y ( log10(KD))

0.27×
_ _
Entry titer (RLU/pg p24)

Y449H
Binding affinity (-log10(KD))

example, N501Y in the Beta background is the 0.16×
Wuhan- G496R Y449D
Hu-1 2 Y449S Y449N log (K )
opposite-sign effect of the measured Y501N muta-
9
10 D
Q498R Y449H
higher affinity
1 G504V 1
tion. The red line indicates the Wuhan-Hu-1 affinity, G447V G447S
with Y501
1 N448K
0
8
and asterisks indicate the actual affinity of the -1
0
Omicron BA.1 RBD relative to Wuhan-Hu-1 as
higher affinity
with N501
0.01 additive
_
7
measured in (12) (fig. S5, A to C). (C) Efficiency of -1 P499T
S494L
G446V Q498H
entry of Omicron BA.1 (or reversion mutant) spike- Y449H
none
Omicron
BA.1
+R498Q
+Y501N
+R498Q+
Y501N
1/32 1/4 2 16
pseudotyped lentivirus on a human embryonic 0 1 2 more frequent more frequent
Number mutations with N501 with Y501
kidney (HEK)–293T cell line expressing low levels of Ratio of substitution count on
Spike gene Y501 versus N501 genomes
ACE2 (fig. S6, A and B). Labels indicate fold
decrease in geometric mean (red bar) of biological
triplicate measurements. (D) Double-mutant cycle illustrating positive epistasis between N501Y and Y449H. (E) Impact of epistasis on SARS-CoV-2 sequence
evolution. Plot illustrates the change in a mutation’s effect between Alpha (N501Y) versus Wuhan-Hu-1 deep mutational scanning data, versus the ratio in number of
observed occurrences of the substitution in genomes containing N501 versus Y501 in a global SARS-CoV-2 phylogeny as of 25 May 2022 (22). We are counting
substitution occurrence as an event on the phylogeny independent of the number of offspring of a node and not the raw number of sequenced genomes with which a
mutation is observed. A pseudocount was added to all substitution counts to enable ratio comparisons, and substitutions that were observed less than two times
in total are excluded. Color scale reinforces the DDlog10(Kd) metric on the y axis. Labeled mutations are those with jDDlog10 ðKd Þj > 0:9. The vertical line at x ~ 0.6 marks
equal relative occurrence on Y501 versus N501 genomes given the larger number of substitutions that had been observed on N501 genomes.
Fig. 4. Epistatic shifts are not accompanied by A B 0.5

large structural perturbations. (A) Global Wuhan-Hu-1 498
effects versus Wuhan-Hu-1

498
Epistatic shift in mutation

alignment of the Wuhan-Hu-1 (PDB 6M0J) and Beta Beta 0.4
496 449 446 449 449

(PDB 7EKG) RBD backbones. Key sites are labeled.
0.3
(B) Correlation between the extent of epistatic 403 403
shift in mutational effects at a site and its structural 498 0.2 447 496 447 496
403
perturbation in Beta versus Wuhan-Hu-1 RBDs 501 494
494 493
0.1
(backbone Ca or all-atom average displacement
from aligned x-ray crystal structures) (figs. S7 and 0.0
0 1 2 3 0 2 4 6
S8). (C) Molecular dynamics simulation of RBD 381- C All-atom average
variants bound to ACE2. Volumetric maps (top) show 387
Sitewise structural displacement versus Wuhan-Hu-1 (Å)
the 3D space occupied by key residues over the
course of simulation. Cartoon diagrams (bottom) C
illustrate the fraction of simulation frames in which Wuhan-Hu-1 Beta Wuhan-Hu-1+Q498R Omicron
a salt bridge (black arrow) or polar or nonpolar (Q498+N501) (Q498+Y501) (R498+N501) (R498+Y501)
(gray arrow) contact is formed between residue pairs
D38
(fig. S9C). Equivalent diagrams for Omicron+Y501N D38 D38 D38
E37 E37 E37
are provided in fig. S9A (R498-N501 for comparison E37
with Wuhan-Hu-1+Q498R), and apo ACE2 is provided
in fig. S9B. Histograms of contact distances over
R498 R498
the course of the simulations are provided in fig. S9C. Q498 K353 Q498 K353
K353
N501
N501 K353 Y501
Y501
E35 E35 E35 E35
87% 87% 80% 55%
79% 81% 79% 75%

0% 0% 34%
Q493 D38 Q493 7% D38 Q493 D38 R493 D38
K31 K31 K31 K31

39% 0% 70%
0% 0% 0% 18% 13%
67% E37 49% E37 64% E37

Q498 Q498 88% E37 R498 R498
5%
94% 0%
0% 0% 7% 29% 38% 46%
92% 16%
N501 K353 Y501 K353 N501 K353 Y501 K353

54% 47% 5% 16%

it can buffer other beneficial antibody-escape Omicron, the Q498R and N501Y combination production, J. C. Nix for x-ray data collection, and T. I. Croll
mutations as described above. pose K353ACE2 in a stable rotamer that main- for help with structure refinement. Use of the Stanford
Synchrotron Radiation Lightsource, SLAC National Accelerator
Other common combinations of mutations tains the D38ACE2 salt bridge and reanimates Laboratory, is supported by the US Department of Energy
are not involved in specific epistatic interac- the E37ACE2 salt bridge present in the apo ACE2 (DOE), Office of Science, Office of Basic Energy Sciences under
tions. For example, substitutions at sites 417, structure (fig. S9B) while adding a new minor contract DE-AC02-76SF00515. The SSRL Structural Molecular
Biology Program is supported by the DOE Office of Biological
484, and 501 arose together in the Beta and salt bridge contact between R498 and D38ACE2. and Environmental Research and by the NIH NIGMS (including
Gamma variants. Early studies disagree on This complex epistatic reconfiguration of a P41GM103393). The contents of this publication are solely
whether there is epistasis among these muta- polar contact network illustrates how the dy- the responsibility of the authors and do not necessarily
represent the official views of NIGMS or NIH. Funding: This
tions with respect to ACE2 binding (6, 23, 24), namic basis of RBD:ACE2 interaction leads to project has been funded in part with federal funds from
but our data demonstrate strict additivity (figs. dynamic evolutionary variability. the NIAID-NIH under contracts 75N93021C00015,
S3E and S5E). The co-occurrence of mutations Overall, SARS-CoV-2 has explored a diverse HHSN272201400006C, and R01AI1417097 (to J.D.B.) and
DP1AI158186 and HHSN272201700059C (to D.V.). Funding
at these three sites in SARS-CoV-2 variants may set of mutations during its evolution in hu-
was also provided by a Pew Biomedical Scholars Award (to
instead reflect antigenic selection for E484 and mans. Our results show how this ongoing evo- D.V.), an Investigators in the Pathogenesis of Infectious
K417 mutants [which escape different classes of lution is itself shaping potential future routes Disease Awards from the Burroughs Wellcome Fund (to D.V.),
neutralizing antibodies (15)], whereas N501Y of change by shifting the effects of key mu- Fast Grants (to D.V.), and the Natural Sciences and Engineering
Research Council of Canada (to M.M.). T.N.S. is an HHMI
might globally compensate for the affinity- tations on receptor-binding affinity. Other Fellow of the Damon Runyon Cancer Research Foundation. D.V.
decreasing effect of K417 mutations. These ex- human coronaviruses have proven adept at and J.D.B. are investigators of the Howard Hughes Medical
amples illustrate how N501Y can enable viral escaping from antibody immunity (27) be- Institute. Author contributions: T.N.S., A.J.G., and J.D.B.
designed the study. T.N.S. and A.J.G. performed deep
evolution through specific epistatic modulation cause they can undergo extensive evolutionary mutational scanning experiments. T.N.S. analyzed epistasis in
(such as Y449 mutations) as well as nonspecific remodeling of the amino acid sequence of the deep mutational scanning data. W.W.H. created interactive
affinity-buffering (such as K417N). their receptor-binding domain while retain- data visualizations. T.N.S. and A.N.L. performed pseudotyped
lentiviral entry and neutralization experiments, with reagents
To examine the structural basis for epistatic ing high receptor affinity (28, 29). Our work and assistance from A.G.F., B.D., K.A.M., and D.C.;
shifts in mutational effects, we examined ACE2- provides large-scale sequence-function maps T.N.S. and W.W.H. analyzed the SARS-CoV-2 evolutionary data.
bound RBD crystal structures of the Wuhan- that help to understand how a similar process E.F., J.R.D., M.M., D.V., and G.S. determined the Beta x-ray
crystal structure. K.H. and G.S. performed and analyzed
Hu-1 and Beta RBDs (25, 26), including a newly may play out for SARS-CoV-2. molecular dynamics simulation. T.N.S., A.J.G., and J.D.B. wrote
determined crystal structure of the ACE2-bound the initial draft, and all authors edited the final version.
Beta RBD (plus antibodies S304 and S309) at RE FERENCES AND NOTES Competing interests: J.D.B. has consulted for Moderna on
viral evolution and epidemiology and consults for Apriori Bio on
2.45 Å resolution (table S1). These compar- 1. K. Tao et al., Nat. Rev. Genet. 22, 757–773 (2021).
deep mutational scanning. T.N.S., A.J.G., A.N.L., and J.D.B.
isons do not reveal clear structural perturba- 2. T. N. Starr et al., Cell 182, 1295–1310.e20 (2020).
receive a share of intellectual property revenue as inventors
3. Y. Liu et al., Nature 602, 294–299 (2022).
tions that explain epistatic shifts between the 4. T. N. Starr, J. W. Thornton, Protein Sci. 25, 1204–1218
on Fred Hutchinson Cancer Research Center–optioned
Wuhan-Hu-1 and Beta RBDs; residues with technology and patents related to deep mutational scanning of
(2016).
viral proteins. K.H., E.F., J.R.D., D.C., and G.S. are employees of
large epistatic shifts between backgrounds 5. T. N. Starr et al., Nature 603, 913–918 (2022).
Vir Biotechnology and may hold shares in Vir Biotechnology.
6. J. Zahradník et al., Nat. Microbiol. 6, 1188–1198 (2021).
show extents of variation between Wuhan- 7. N. Bate et al., bioRxiv 473975 [Preprint] (2021);
Data and materials availability: Site saturation mutagenesis
Hu-1 and Beta structures similar to those that libraries are available from Addgene (catalog 1000000182 to
doi: 10.1101/2021.12.23.473975.
1000000186). Raw sequencing data are on the NCBI SRA
these residues show within replicate structures 8. R. Viana et al., Nature 603, 679–686 (2022).
under BioProject PRJNA770094, BioSamples SAMN25941479
9. K. Javanmardi et al., Mol. Cell 81, 5099–5111.e8 (2021).
of Wuhan-Hu-1 or Beta itself (fig. S7). More 10. K. K. Chan, T. J. C. Tan, K. K. Narayanan, E. Procko, Sci. Adv. 7,
and SAMN22208699 (PacBio sequencing), and SAMN25944367
broadly, there is minimal change between (Illumina barcode sequencing). All code and data at various
eabf1738 (2021).
stages of processing is available together with summaries
Wuhan-Hu-1 and Beta RBD backbones (Fig. 4A 11. E. Cameroni et al., Nature 602, 664–670 (2022).
notebooks detailing the computational pipeline at https://
12. B. Meng et al., Nature 603, 706–714 (2022).
and fig. S8A), and we did not detect any cor- github.com/jbloomlab/SARS-CoV-2-RBD_DMS_variants. Final
13. M. McCallum et al., Science 375, 864–868 (2022).
relation between structural displacement of mutant deep mutational scanning phenotypes are available on
14. A. J. Greaney, T. N. Starr, J. D. Bloom, bioRxiv 471236
GitHub (https://github.com/jbloomlab/SARS-CoV-2-RBD_DMS_
backbone or side chain atoms in variant RBD [Preprint] (2021), doi:10.1101/2021.12.04.471236.
variants/blob/main/results/final_variant_scores/final_variant_
structures and epistatic shifts in mutational 15. A. J. Greaney et al., Nat. Commun. 12, 4196 (2021).
scores.csv) and data S1, and interactive visualizations of key
16. A. J. Greaney et al., PLOS Pathog. 18, e1010248 (2022).
effects (Fig. 4B and fig. S8, B to E). These ob- 17. F. Schmidt et al., Nature 600, 512–516 (2021).
data are available at https://jbloomlab.github.io/SARS-CoV-2-
RBD_DMS_variants. Github files are also archived at Zenodo
servations indicate that epistatic shifts in 18. T. N. Starr et al., Nature 597, 97–102 (2021).
(30). Coordinates for the SARS-CoV-2 Beta RBD coordinated
mutant effects occur despite conservation of 19. C. Scheepers et al., Nat. Commun. 13, 1976 (2021).
with ACE2, S309, and S304 have been deposited in the Protein
20. T. Tada et al., Cell Rep. 38, 110237 (2022).
the global static RBD structure. 21. P. Colson et al., bioRxiv 21268174 [Preprint] (2021);
Data Bank (PDB) with accession code 8DF5. License
To explore the cause of epistasis between information: This work is licensed under a Creative Commons
doi: 10.1101/2021.12.24.21268174.
Attribution 4.0 International (CC BY 4.0) license, which
Q498R and N501Y (Fig. 3A), we performed 22. J. McBroome et al., Mol. Biol. Evol. 38, 5819–5824 (2021).
permits unrestricted use, distribution, and reproduction in any
23. C. Laffeber, K. de Koning, R. Kanaar, J. H. G. Lebbink, J. Mol.
molecular dynamics simulations of the Wuhan- Biol. 433, 167058 (2021).
medium, provided the original work is properly cited. To view a
Hu-1 (Q498-N501), Beta (Q498-Y501), and Omi- copy of this license, visit https://creativecommons.org/
24. M. Yuan et al., Science 373, 818–823 (2021).
licenses/by/4.0/. This license does not apply to figures/
cron (R498-Y501) RBDs bound to ACE2 (14, 25), 25. J. Lan et al., Nature 581, 215–220 (2020).
photos/artwork or other content included in the article that is
26. P. Han et al., Nat. Commun. 12, 6103 (2021).
in addition to in silico mutated complexes of 27. R. Eguia et al., PLOS Pathog. 17, e1009453 (2020).
credited to a third party; obtain authorization from the rights
Wuhan-Hu-1+Q498R and Omicron+Y501N holder before using such material.
28. A. H. M. Wong et al., Nat. Commun. 8, 1735 (2017).
(Fig. 4C and fig. S9). The Wuhan-Hu-1 struc- 29. Z. Li et al., eLife 8, e51230 (2019).
30. T. Starr, A. J. Greaney, W. Hannon, J. Bloom, jbloomlab/
ture features a stable polar contact network SUPPLEMENTARY MATERIALS
SARS-CoV-2-RBD_DMS_variants: published version. Zenodo
between ACE2 residues D38 and K353 and (2022). science.org/doi/10.1126/science.abo7896
RBD residue Q498. The affinity-enhancing Materials and Methods
ACKN OWLED GMEN TS Figs. S1 to S9
N501Y substitution present in Beta repositions Table S1
We thank the Genomics and Flow Cytometry core facilities at
K353ACE2 in an orientation that reinforces References (31Ð62)
Fred Hutchinson Cancer Research Center, K. Munson at the
the D38ACE2 salt bridge but disrupts all Q498 MDAR Reproducibility Checklist
University of Washington PacBio Sequencing Services, and the
Data S1
contacts. By contrast, the affinity-decreasing Fred Hutchinson Scientific Computing group supported by
ORIP grant S10OD028685. We thank R. Eguia for experimental
Q498R mutation alone improves the coordi-
assistance. We thank S. Weber and the library synthesis
nation between residue 498 and D38ACE2 but team at Twist Bioscience for library construction. We thank Submitted 25 February 2022; accepted 23 June 2022
leaves K353ACE2 incompletely satisfied. In P. Hernandez and N. Czudnochowski for support with protein 10.1126/science.abo7896

OPTICS Here, we explore the radiation emitted by

a radiation source embedded in a PTC. We
Amplified emission and lasing in photonic formulate the quantum theory describing the
emission of light by atoms in an excited state
time crystals and the classical theory of radiating dipoles
embedded in PTCs, and show that radiation
Mark Lyubarov1,2,3, Yaakov Lumer1,2, Alex Dikopoltsev1,2, Eran Lustig1,2, is always exponentially amplified when as-
Yonatan Sharabi1,2, Mordechai Segev1,2,4* sociated with the momentum gap and its
linewidth becomes narrower with time. This
Photonic time crystals (PTCs), materials with a dielectric permittivity that is modulated periodically effect allows us to propose nonresonant tun-
in time, offer new concepts in light manipulation. We study theoretically the emission of light able PTC lasers which draw their energy from
from a radiation source placed inside a PTC and find that radiation corresponding to the the modulation.
momentum bandgap is exponentially amplified, whether initiated by a macroscopic source, Our model consists of a PTC with a source
an atom, or vacuum fluctuations, drawing the amplification energy from the modulation. The of radiation inside (Fig. 1A). First, we consider
radiation linewidth becomes narrower with time, eventually becoming monochromatic in the middle an empty PTC medium (no radiation source)
of the bandgap, which enables us to propose the concept of nonresonant tunable PTC laser. and derive the eigenmodes, and then add
Finally, we find that the spontaneous decay rate of an atom embedded in a PTC vanishes at the band an arbitrary radiation source. Starting with
edge because of the low density of photonic states. Maxwell equations with e = e(t), m = 1, we can
write the wave equation for the magnetic field
P
as follows:
hotonic time crystals (PTCs) are dielectric the modulation. PTCs bear some relation to
media with a refractive index that ex- optical parametric amplifiers, but the latter @t ½eðt Þ@t þ c2 k2 H k ¼ 0 ð1Þ
periences large, ultrafast periodic variations are resonant phenomena: the frequency of the
in time (1–5). Generally, a wave propagat- pump is equal to the sum of the frequencies where we use a Fourier transform in space
ing in a medium undergoing an abrupt of the signal and idler and phase matching because the system is homogeneous and k
change in the refractive index experiences guarantees conservation of momentum, so is a good quantum number. Physically, this
time reflection and time refraction. The time only a specific wave is amplified. In contra- means that the eigenmodes are shaped as
reflection is especially interesting because distinction, PTCs display a significant momen- plane waves, defined by their wave number k.
causality imposes that the wave reflected from tum gap in which every wave is amplified. For For each k, this equation has two Floquet
the temporal interface propagates backward a detailed comparison of PTCs and optical eigenmodes:
in space rather than in time (6). Periodic mod- parametric amplifiers, see (7). 1;2
ulation of the refractive index makes these Apart from a momentum band structure, Hk1;2 ðt Þ ¼ Hk0 ðt Þeiwk t
ð2Þ
time reflections and time refractions inter- the abrupt temporal modulation of the per-
fere giving rise to bands and bandgaps in the mittivity also opens up new possibilities such where wk1;2 are Floquet quasifrequencies and
momentum (1, 3, 4). The dispersion relation of as a frequency conversion (2), photon pair Hk0(t) is a periodic function in time, con-
PTCs seems analogous to spatial photonic creation (8–12), topological temporal edge structed from harmonics of the modulation
crystals (SPCs), in which the refractive index states (5), antireflection temporal coatings period T. We assume that e is real (i.e., the
is periodic in space. However, despite the sim- (13), extreme energy transformations (14), medium is lossless), so if H(t) is an eigenmode,
ilarity, there are fundamental differences: SPCs interaction with free electrons (15), and am- i.e., solution of (Eq. 1), then so is H*(t), which
are stationary in time so energy conservation plified localization in temporally disordered means that w1k ¼ w2k ¼ wk . Solving for the
governs most processes, whereas in PTCs, en- media (16). Experimentally, time refraction dispersion relation, we find that the dispersion
ergy is not conserved and causality dictates has already been observed in photonics (17), curve forms a band structure (Fig. 1C). In the
the dynamics in the system. Conversely, waves whereas time reflection has thus far only been bands, the frequency wk is real and the two
propagating in SPCs exchange momentum observed with water waves (18), acoustic waves modes are oscillating at the same frequency,
with the spatial lattice, whereas in spatially (19), and elastic waves (20). This is because whereas in the gaps, wk has an imaginary part,
homogeneous PTCs, momentum is conserved. of the highly demanding requirements for ob- with one mode exponentially growing with
The most important feature of PTCs is the serving time reflections: The refractive index time and the other exponentially decaying. To
existence of a bandgap in momentum, because change should act as a “wall,” analagous to explore the response of the PTC to the exci-
the modes associated with this gap have two a spatial interface causing Fresnel reflection. tation, we add to Eq. (1) a radiation source as-
solutions in which the mode amplitude grows For light in the near infrared, the modulation sociated with a temporally dependent current
or decays exponentially with time, and both should be at few femtosecond rates with an density j(r,t):
solutions are physical. The exponential growth absolute permittivity change of De > 0.1, which
of the gap modes is nonresonant; it occurs for is difficult to realize in experimental conditions. @t ½eðt Þ@t þ c2 k2 H k ðt Þ ¼ 4pickñj k ðt Þ ð3Þ
all wave vectors associated with the momen- However, recent progress with epsilon-near-
tum gaps, which offers an avenue for ampli- zero materials (21–24) brings these ideas close where jk(t) is a Fourier k component of cur-
fication of radiation by drawing energy from to experimental realization (25). rent j(r,t). For a point dipole, we assume
The existence of momentum bands and gaps jðr; tÞ ¼ d 0 dðrÞeiwt q(t), where q(t) is a Heavi-
in a PTC raise fundamental questions about the side step function denoting that the current is
1
Physics Department, Technion – Israel Institute of turned on at t = 0. Physically, the field Hk(t)
emission of light by a radiation source embed-
Technology, Haifa 32000, Israel. 2Solid State Institute,
Technion – Israel Institute of Technology, Haifa 32000, ded in a PTC. An analogous study has led to the is the response of the medium to this current.
Israel. 3Physics and Engineering Department, ITMO discovery of the inhibition of spontaneous We can express it in a general form through
University, St. Petersburg 197101, Russia. 4Department of emission in the bandgap of SPCs (26), which Green’s function as follows:
Electrical and Computer Engineering, Technion – Israel
Institute of Technology, Haifa 32000, Israel. has had major consequences, such as threshold-
*Corresponding author. Email: [email protected] less lasing (27, 28). H k ðt Þ ¼ 4ipc∫∞ ∞ Gk ðt; t′Þk j k ðt′Þdt′ ð4Þ

Fig. 1. Emission by a point

dipole embedded in a PTC.
(A) Sketch of the PTC, with
permittivity varying as e(t) = eav +
D/2·cos(Wt), W = 2p/T with a
dipole antenna inside. The dipole
radiation is exponentially amplified
with time. (B) Exponential
growth of electromagnetic energy
associated with the dipole
emission for different dipole fre-
quencies w0 and modulation
amplitudes D. (C) Complex dis-
persion relation (band structure)
of the PTC for eav = 2, D = 1. The
values of wk at the bandgap
around kg are complex, indicating
exponentially growing and decaying eigenmodes. (D) Power spectrum of dipole emission versus wave number as it evolves with time. Initially, the point dipole with
frequency w0 excites all eigenmodes. k0 is a wave number of the mode resonantly excited by the dipole with frequency w0: wk(k0) = w0. The emission linewidth initially
occupies all bands and gaps, located at kg, but eventually, after a short time, the radiation in the gap becomes dominant and narrower with time, reflecting the
stronger emission at midgap. In each moment in time, the spectrum is normalized by the total radiation power. The horizontal axes in (C) and (D) coincide.
where fields in Fig. 1, B and D, is described in detail This analysis explains the exponentially
in section 4 of (7). Initially, the point dipole growing dipole emission in a PTC. The dipole
@t ðeðt Þ@t Þ þ c2 k2 Gk ðt; t′Þ ¼ dðt t′Þ ð5Þ with frequency w0 excites all the eigenmodes excites the gap modes, which, once excited,
with proper wave number k0 such that wk(k0) = grow exponentially regardless of the dipole,
and then express this Green’s function through w0. This is because these waves lie on the dis- even when mismatched. The key issue here
the eigenmodes from Eq. (2): persion curve and thus are perfectly phase is that a point dipole excites modes with all
matched. However, within a few oscillation k, jk ≠ 0 ∀k, including the exponentially
Gk ðt; t′Þ ¼ cycles, the gap modes start to dominate even growing gap modes. Thus, any point source
8 if k0 does not belong to the gap. These modes in a PTC results in a exponentially growing
< 0; t < t 0 are not phase matched with the dipole fre- emission, even when the excitation is a single
H k ðt′ÞH k ðt Þ H 1 k ðt′ÞH 2 k ðt Þ
2 1
quency, but they nevertheless grow exponen- flash in time. The emission from this flash will
: ′ ;
eðt ÞðH 2 k ðt′Þ@t′ H 1 k ðt′Þ H 1 k ðt′Þ@t 0 H 2 k ðt′ÞÞ tially in time, which overshadows any phase grow exponentially, drawing energy from the
matching. modulation.
t > t′ ð6Þ We can understand the exponentially grow- Next, we quantize our model. First, we write
ing response in a PTC through Fig. 2, which the electromagnetic field Hamiltonian in a PTC
Green’s function, Gk(t,t′), represents the re- shows the difference between excited gap
sponse of the medium at time t to a single modes in SPC and in PTC, where the excitation
homogeneous “flash” at time t′. The detailed in the SPC is by a point source in real space
derivation of Eq. (6) is provided in (7). A closer and the excitation in the PTC is by a flash in
look at Eqs. 4 to 6 reveals that, in the momentum time. The solution of Eq. (5) should be ex-
bandgap where Im(wk) ≠ 0, the medium re- pressed through two eigenmodes on either
sponds with exponentially growing emission side of the excitation point and stitched with
even to the slightest flash of radiation emitted two stitching conditions. The physical con-
from the current source. This seemingly coun- straints in both cases reveal which contribu-
terintuitive feature is a consequence of the tions are unphysical and should be removed.
lack of energy conservation in the medium. In In the case of the SPC (Fig. 2A), the solution
fact, the energy deposited into the exponen- must obey energy conservation, so only eva-
tially growing gap modes comes not from the nescent waves are allowed on either side of
source but rather from the external modula- the excitation point in space. Therefore, the
tion of the medium. The exponentially grow- response to the excitation at a frequency in the Fig. 2. Excitation of gap modes in SPC and PTC.
ing dipole emission is shown in Fig. 1B for gap of a SPC are evanescent waves. Conversely, (A) The one-dimensional SPC is excited by a point
various dipole frequencies and permittivity in the PTC (Fig. 2B), two of the four modes are source at position x0 and emits at a given
profiles. The growth rate barely depends on the propagating back in time and therefore can- frequency within the photonic bandgap. The
frequency of the dipole but strongly depends not be excited because they are restricted by source can couple only to the spatially evanescent
on the amplitude of the permittivity modu- causality. Thus, Green’s function must be ex- part on either side of x0 because of energy
lation. The larger the modulation, the sooner pressed with two forward-propagating waves conservation. (B) In the PTC, the source is a flash
the growth takes place and the steeper it is. in time, one of which is exponentially decaying at t0 and can excite only the parts of the modes
The energy spectrum (k) of the dipole emis- and the other exponentially growing, which is that evolve forward in time, as dictated by
sion and its evolution with time are depicted allowed because there is no energy conserva- causality. One of these two modes is exponentially
in Fig. 1D. The numerical simulation of the tion in PTCs. growing in time.

of atomic transition. Moreover, even if the

two-level atom is not in resonance with the
momentum-gap, i.e., w0 ≠ W/2, the emission
into gap modes eventually governs the dy-
namics of the atom.
It is now natural to ask if there are any
circumstances under which we can still talk
about spontaneous emission (in the usual sense
of being induced by quantum fluctuations with
no photons around) in a PTC and what the
physical consequences of this might be. This
question can be answered partially by address-
ing the Floquet modes associated with the
band, ignoring the influence of the gap modes.
This assumption can be justified if the decay
Fig. 3. Spontaneous emission rate into the Floquet modes associated with the first band. time of the atom is shorter than the inverse
(A) Spontaneous emission rate in the first band of the PTC. Starting at a low wave number, the emission rate growth rate of the number of photons in the
increases and then reaches a maximum, but then declines and goes to zero at the band edge kb, where the gap modes tsp < 1/Im(wk) or if the PTC with
band structure is curved. (B) Dispersion in the first band of the PTC. The slope of the bandstructure becomes the embedded atom is placed in a resonator
more vertical closer to the bandgap, indicating the vanishing density of states. with all resonator eigenmodes residing inside
the PTC bands (rather than in the gaps). In
as follows: with the classical case, Nk in the gap eigen- this case, we show in (7) that the spontaneous
ð
modes should grow exponentially, which is im- emission rate is:
nðt Þ
X ck þ nnðrt Þ possible with Hermitian Hamiltonians such as
Hf ¼ ℏ
nr
a†k ak þ a† k a k þ V X 2 k2m
nðt Þ 2 the one in Eq. (7). The absence of eigenstates g¼ Vm
m fi
ð11Þ
k ℏ2 p @wf
Þ
in the gap brings complexity in studying the @k k¼k
nðt Þ nr m
nðt Þ dynamics of the excited atom, interacting with
þ a†k a† k
nr
ak a k ð7Þ the radiation field described below, but the where
2
pffiffiffiffiffiffiffiffi exponential growth of the number of pho-
1 T
where nðt Þ ¼ eðt Þ is the time-varying re- tons in the momentum gap and the classical/ Vfim ¼ ∫ ϕ ðt Þ Hint ðt Þjϕi ðt ÞieimWt dt ð12Þ
T 0 f
fractive index, nr is the mean value of refrac- semiclassical intuition allow us to make some
tive index obtained by averaging through one safe statements on the dynamics in this unusual is the coupling constant between the initial
modulation cycle, and a†k ðak Þ are the creation quantum system. and the final Floquet eigenstates through Hint
(annihilation) operators for mode with the wave To describe the emission from excited atoms and km : wf ðkm Þ ¼ w0 þ mW is the wave num-
vector k. This Hamiltonian is derived in (7) in PTC, we add the atomic and the interaction ber of the mode corresponding to mth har-
following the quantization procedure de- parts to the Hamiltonian (Eq. 7): monic of the atomic transition (30). Analyzing
scribed in (29). It follows our intuition gained
H ¼ Hf þ Ha þ Hint ð8Þ the dynamics of the emission rate g(k) with-
in the classical case: It is time dependent in the band, we observe that there are two
throughXn(t) and it conserves momentum, Ha ¼ ℏw0 sz ð9Þ
competing contributions: the closer to the
Hf ; ka†k ak ¼ 0, but it does not con- band edge the larger the Vfim , because for
k modes in the vicinity of the gap the oscil-
serve the number of photons. The Hamiltonian X ℏgk lations are larger, whereas at
Hint ¼ a þ a†k ðsþ þ s Þ ð10Þ the
1 band edge,
(Eq. 7) allows describing the dynamics of the k eðt Þ k the density of states, r¼k2 @w is smaller.
@k
free field for each photon pair {k, –k} separately. Fig. 3A shows that at the band edge, the rate
The resulting dynamics agrees with the classical where we assume a two-level atom and dipole of spontaneous emission vanishes because the
case: For modes with k associated with the band interaction. We first analyze what happens density of states goes down to zero. The low
of the PTC, the expectation value of the number of with an initially excited atom interacting density of states is apparent from the vertical
photons, Nk ðt Þ ¼ hyðt Þ a†k ak þ a† k a k yðt Þi, with the vacuum field. In the case of a static slope of the dispersion near the band edge
oscillates near some constant value, whereas medium, the result is the exponential decay (Fig. 3B). The implication is intriguing: Even
if k belongs to the PTC bandgap, Nk grows of the atom from the excited state to the though the Floquet modes have larger oscil-
exponentially with time at the same rate as in ground state, known as spontaneous emission. lations closer to the band edge, which natu-
the classical case. The periodic variation of n(t) In a PTC, no analytic solution is feasible, rally increases the strength of the light-matter
allows introducing the Floquet eigenmodes because the number of photons in the initially interaction, the emission rate at the edge goes
jyk ðt Þi ¼ e iwk t jϕk ðt Þi of the Hamiltonian empty gap modes grows exponentially regard- to zero because there are no states to radiate
(Eq. 7), with wk being the Floquet eigen- less of the atom. This means that the atom into. Thus, an “atom” or a nano-antenna with
frequency. Let us first list the main features emission into these modes cannot be clearly directional emission at the band edge would stay
of the quantum Floquet eigenmodes, the de- divided into spontaneous and stimulated emis- in the excited state forever, unable to relax to the
tailed analysis of which is provided in (7). In sions: The rate of transitions grows with time as ground state through spontaneous emission.
the bands, wk coincides with the Floquet fre- a consequence of the photons already created In one-dimensional PTCs, the presence of a
quency calculated in the classical analysis, and by the PTC. In addition to stimulated emission, gap in the momentum alters the light-matter
the Floquet eigenstates experience weak os- stimulated absorption also takes place, which interactions in a profound way, bringing to
cillations in the number of photons. Con- results in complex dynamics of the atom. As in question foundational issues such as the meaning
versely, in the bandgap, the eigenstates of the the classical case, the growth in the number of spontaneous and induced emission in such
Hamiltonian cannot exist: By correspondence of photons barely depends on the frequency media and the lifetime of an atom in excited

states. The exponential growth of energy in contributions: All authors contributed substantially to all relevant SUPPLEMENTARY MATERIALS
the modes associated with the PTC gap and aspects of this research. Competing interests: The authors science.org/doi/10.1126/science.abo3324
declare no competing interests. Data availability: All data are Materials and Methods
the nonmonotonous growth rate raise the available in the main text or the supplementary materials. License Figs. S1 to S3
exciting idea of PTC lasers extracting their information: Copyright © 2022 the authors, some rights reserved; References (31–35)
energy from the modulation. The simplest exclusive licensee American Association for the Advancement of
Science. No claim to original US government works. https://www. Submitted 28 January 2022; accepted 27 May 2022
setting for such a laser is to construct a res- science.org/about/science-licenses-journal-article-reuse 10.1126/science.abo3324
onator by placing mirrors on either side of
the dielectric medium with its permittivity
modulated in time. The cavity length should
be much larger than the wavelength of the CORONAVIRUS
waves of interest, such that momentum con-
servation applies despite the finite size of the Pathogenicity, transmissibility, and fitness of
resonator. Cavities with shorter lengths can
also exhibit momentum gaps but require SARS-CoV-2 Omicron in Syrian hamsters
additional treatment of the spatial modes.
Because the amplification of the waves asso- Shuofeng Yuan1†, Zi-Wei Ye1†, Ronghui Liang1†, Kaiming Tang1†, Anna Jinxia Zhang1†, Gang Lu2,3†,
ciated with the gap modes attains a maximum Chon Phin Ong4†, Vincent Kwok-Man Poon1,5, Chris Chung-Sing Chan1,5, Bobo Wing-Yee Mok1,
at midgap, any saturation mechanism will Zhenzhi Qin1, Yubin Xie1, Allen Wing-Ho Chu1, Wan-Mui Chan1, Jonathan Daniel Ip1, Haoran Sun6,
eventually result in stable monochromatic Jessica Oi-Ling Tsang1,5, Terrence Tsz-Tai Yuen1, Kenn Ka-Heng Chik1,5, Chris Chun-Yiu Chan1,
emission. Thus, controllable periodic change Jian-Piao Cai1, Cuiting Luo1, Lu Lu1,5, Cyril Chik-Yan Yip6, Hin Chu1,5,7, Kelvin Kai-Wang To1,5,6,7,8,
of the permittivity can give rise to coherent Honglin Chen1,5,6,8, Dong-Yan Jin4,8‡, Kwok-Yung Yuen1,3,5,6,7,8‡, Jasper Fuk-Woo Chan1,3,5,6,7,8‡¤*
radiation from an almost arbitrary source and,
under some conditions, the emission can be The in vivo pathogenicity, transmissibility, and fitness of the severe acute respiratory syndrome
shaped into pulses by designing the modulation. coronavirus 2 (SARS-CoV-2) Omicron (B.1.1.529) variant are not well understood. We compared these
virological attributes of this new variant of concern (VOC) with those of the Delta (B.1.617.2) variant
RE FE RENCES AND N OT ES in a Syrian hamster model of COVID-19. Omicron-infected hamsters lost significantly less body weight
1. F. Biancalana, A. Amann, A. V. Uskov, E. P. O’Reilly, Phys. Rev.
and exhibited reduced clinical scores, respiratory tract viral burdens, cytokine and chemokine
E Stat. Nonlin. Soft Matter Phys. 75, 046607 (2007).
2. J. R. Zurita-Sánchez, P. Halevi, J. C. Cervantes-Gonzalez, Phys. dysregulation, and lung damage than Delta-infected hamsters. Both variants were highly transmissible
Rev. A 79, 053821 (2009). through contact transmission. In noncontact transmission studies Omicron demonstrated similar or
3. J. R. Reyes-Ayona, P. Halevi, Appl. Phys. Lett. 107, 074101 (2015). higher transmissibility than Delta. Delta outcompeted Omicron without selection pressure, but this
4. A. M. Shaltout, J. Fang, A. V. Kildishev, V. M. Shalaev, “Photonic
time-crystals and momentum band-gaps,” paper presented scenario changed once immune selection pressure with neutralizing antibodies—active against Delta but
at the 2016 Conference on Lasers and Electro-Optics (CLEO): poorly active against Omicron—was introduced. Next-generation vaccines and antivirals effective
Science and Innovations, San Jose, CA, 5–10 June 2016. against this new VOC are therefore urgently needed.
5. E. Lustig, Y. Sharabi, M. Segev, Optica 5, 1390–1395 (2018).
6. J. T. Mendonça, P. K. Shukla, Phys. Scr. 65, 160–163 (2002).
S
7. See the supplementary materials.
8. A. B. Shvartsburg, Phys. Uspekhi 48, 797–823 (2005). evere acute respiratory syndrome coro- its enhanced transmissibility and moderate
9. M. Uhlmann, G. Plunien, R. Schützhold, G. Soff, Phys. Rev. Lett.
navirus 2 (SARS-CoV-2) was first iden- level of antibody resistance has subsequently
93, 193601 (2004).
10. J. T. Mendonca, G. Brodin, M. Marklund, Phys. Lett. A 372, tified in late 2019 and quickly developed replaced the Alpha variant since mid-2021.
5621–5624 (2008). into the most important global health The Omicron (B.1.1.529) variant, first iden-
11. J. Sloan, N. Rivera, J. D. Joannopoulos, M. Soljačić, Phys. Rev. challenge in recent decades (1–3). Despite tified in South Africa in November 2021, has
Lett. 127, 053603 (2021).
12. F. Belgiorno et al., Phys. Rev. Lett. 105, 203901 (2010). rapid generation of vaccines and antivirals and now affected at least 149 countries (7, 8). This
13. V. Pacheco-Peña, N. Engheta, Optica 7, 323–331 (2020). global implementation of various nonpharma- new VOC has a notably high number of muta-
14. H. Li, S. Yin, E. Galiffi, A. Alù, Phys. Rev. Lett. 127, 153903 (2021). ceutical public health measures, the coronavirus tions (>30) at the spike, which significantly
15. A. Dikopoltsev et al., Proc. Natl. Acad. Sci. U.S.A. 119,
e2119705119 (2022).
disease 2019 (COVID-19) pandemic continues reduces the neutralizing activity of vaccine-
16. Y. Sharabi, E. Lustig, M. Segev, Phys. Rev. Lett. 126, 163902 (2021). nearly 2 years after the emergence of more var- induced serum antibodies as well as therapeu-
17. G. Lerosey et al., Phys. Rev. Lett. 92, 193904 (2004). iants of concern (VOC) exhibiting enhanced tic monoclonal antibodies (9–15). Preliminary
18. V. Bacot, M. Labousse, A. Eddi, M. Fink, E. Fort, Nat. Phys. 12,
972–977 (2016).
immunoevasiveness and/or transmissibility (4). analyses of the severity of infections caused
19. A. Derode, P. Roux, M. Fink, Phys. Rev. Lett. 75, 4206–4209 (1995). The Alpha (B.1.1.7) variant emerged in mid-2020 by Omicron compared with previous variants
20. C. Draeger, M. Fink, Phys. Rev. Lett. 79, 407–410 (1997). and quickly outcompeted the Beta (B.1.351) as determined by hospitalization rates have
21. N. Kinsey et al., Optica 2, 616–622 (2015).
22. L. Caspani et al., Phys. Rev. Lett. 116, 233901 (2016).
variant (5, 6). The Delta (B.1.617.2) variant with been inconclusive, with some showing reduced
23. M. Z. Alam, I. De Leon, R. W. Boyd, Science 352, 795–797
(2016).
1
24. O. Reshef, I. De Leon, M. Z. Alam, R. W. Boyd, Nat. Rev. Mater. State Key Laboratory of Emerging Infectious Diseases, Carol Yu Centre for Infection, Department of Microbiology, School of
4, 535–551 (2019). Clinical Medicine, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Pokfulam, Hong Kong Special Administrative
25. E. Lustig et al., “Towards photonic time-crystals: Observation Region, China. 2Key Laboratory of Tropical Translational Medicine of Ministry of Education, Hainan Medical University, Haikou,
of a femtosecond time-boundary in the refractive index,” Hainan, China. 3Academician Workstation of Hainan Province, Hainan Medical University-The University of Hong Kong Joint
paper presented at the 2021 Conference on Lasers and Laboratory of Tropical Infectious Diseases, Hainan Medical University, Haikou, Hainan, China; and The University of Hong Kong,
Electro-Optics (CLEO): QELS–Fundamental Science, Pokfulam, Hong Kong Special Administrative Region, China. 4School of Biomedical Sciences, Li Ka Shing Faculty of Medicine,
San Jose, CA, 9–14 May 2021. The University of Hong Kong, Pokfulam, Hong Kong Special Administrative Region, China. 5Centre for Virology, Vaccinology and
26. E. Yablonovitch, Phys. Rev. Lett. 58, 2059–2062 (1987). Therapeutics, Hong Kong Science and Technology Park, Hong Kong Special Administrative Region, China. 6Department
27. O. Painter et al., Science 284, 1819–1821 (1999). of Infectious Disease and Microbiology, The University of Hong Kong-Shenzhen Hospital, Shenzhen, Guangdong Province,
28. S. Noda, A. Chutinan, M. Imada, Nature 407, 608–610 (2000). China. 7Department of Microbiology, Queen Mary Hospital, Pokfulam, Hong Kong Special Administrative Region, China.
29. W. Vogel, D.-G. Welsch, Quantum Optics (Wiley, 2006). 8
Guangzhou Laboratory, Guangdong Province, China.
30. T. Bilitewski, N. R. Cooper, Phys. Rev. A 91, 033601 (2015). *Corresponding author. Email: [email protected]
†These authors contributed equally to this work. ‡These authors contributed equally to this work. §Present address: State Key Laboratory of
ACKN OW LEDG MEN TS Emerging Infectious Diseases, Carol Yu Centre for Infection, Department of Microbiology, Li Ka Shing Faculty of Medicine, The University of
Funding: The initial stages of this research were funded by a grant Hong Kong, Pokfulam, Hong Kong Special Administrative Region, China; and Department of Infectious Disease and Microbiology, The
from the US Air Force Office of Scientific Research. Author University of Hong Kong-Shenzhen Hospital, Shenzhen, Guangdong Province, China.

Fig. 1. Pathogenicity of Omicron and Delta in the Syrian hamster model analysis in (D to F) (n = 5 including 3 male and 2 female hamsters per variant
of COVID-19. (A) Scheme of the pathogenicity study comparing infections per time point). (D) Respiratory tract tissue infectious virus titers and
caused by Omicron and Delta in Syrian hamsters. At 0 dpi each hamster was (E) viral loads (Student’s t test). The dotted line in (D) represents the limit
intranasally inoculated with SARS-CoV-2 (n = 15 for each variant). In the of detection of the plaque assay (100 PFU/g). (F) Lung cytokine and
first independent experiment, the hamsters (n = 5 for each variant) were kept chemokine gene expression profiles (Student’s t test). Values on the y axis
alive for body weight and clinical score monitoring in (B and C). (B) Body represent the changes in Omicron- or Delta-infected relative to mock-infected
weight changes (n = 5 male hamsters per variant, Student’s t test). (C) samples (drawn in log10 scale). Black dots indicate the mock-infected
Clinical scores. A score of 1 was given for each of the following clinical signs: group (n = 5 per time point, including 3 male hamsters as indicated by
lethargy, ruffled fur, hunchback posture, and rapid breathing (n = 5 male black dots and 2 female hamsters as indicated by white dots). Data represent
hamsters per variant, two-tailed Mann-Whitney U test). In the second mean ± standard deviation. *P < 0.05, **P < 0.01, ***P < 0.001, ****P <
independent experiment, the hamsters were sacrificed at 2, 4, and 7 dpi for 0.0001. n.s., not significant. Actb, beta-actin; I.N., intranasal.

A Co-housing Sacrifice Sacrifice B 10 Index hamsters C

Omicron or for 4 hrs index for naïve for
Delta virus before viral load viral load
challenge separation detection detection
n.s.
Day 0 Day 1 Day 2 Day 3
Delta (6/6) Omicron(6/6)
(PFU/ml)
8
Index Round 1
Round 2
10
Isocage
6
Log
Delta
Omicron
4
Nasal wash
Non-contact Sacrifice Sacrifice
DMSO naïve1.25µM
D Omicron or
Delta virus
transmission
for 6 hrs before
index for
viral load
for
viral load
E F Delta (24/36) Omicron (30/36)
challenge detection detection
2.5µM
separation
10 Index hamsters
Day 0 Day 1 Day 2 Day 3
Index Naïve n.s. Round 1

Air ow
(PFU/ml)
8
Air ow
10
6
Log
Isolator Delta
Round 2
Air ow Omicron
4
Nasal wash
Fig. 2. Contact and noncontact transmission of Omicron and Delta among noncontact transmission study. (E) Nasal wash infectious virus titers in the intranasally
Syrian hamsters. (A) Scheme of the contact transmission study. (B) The nasal SARS-CoV-2–challenged index hamsters at 1 dpi. Data represent mean ± standard
wash infectious virus titers in the intranasally SARS-CoV-2–challenged index deviation of the pooled results of two independent experiments (n = 3 animals
hamsters at 1 dpi were determined by plaque assay to ensure successful infection of per group per experiment, Student’s t test). (F) Positive rates of infection among
animals. Data represent mean ± standard deviation of the pooled results of two the naïve hamsters after exposure to either Omicron or Delta (n = 18 animals
independent experiments (n = 3 animals per group per experiment, Student’s t test). per group per experiment, Student’s t test). Brown hamsters indicate those that
(C) Positive rates of infection among the naïve hamsters after exposure to either were not infected and red hamsters indicate those that were infected, as in (C). In
Omicron or Delta in two independent experiments (n = 3 animals per group per both contact and noncontact transmission studies, hamsters with Ct value ≤40
experiment). Red hamsters indicate those that were infected. (D) Scheme of the in either nasal turbinates or lungs were considered infected. n.s., not significant.
hospitalization rates and others showing a lack 1B). Furthermore, their clinical scores were proteins (fig. S2) between 2 and 7 dpi. At 7 dpi,
of significant difference (16). What is more significantly lower than those of the Delta- the dysregulated inflammatory cytokine and
apparent from early epidemiological data is infected hamsters (Fig. 1C). Early after infec- chemokine response was almost completely
that Omicron is spreading rapidly even in tion [2 days post infection (dpi)], the viral loads normalized in the Omicron-infected hamsters.
populations with high uptake rates of two- and infectious virus titers of the two variants The antibody response against the variant-
dose COVID-19 vaccinations (17, 18). How- in the nasal turbinates and trachea were sim- specific spike receptor-binding domain (RBD)
ever, whether this is a result of the intrinsic ilar but the lung viral loads and virus titers were of the Omicron-infected hamsters was also sig-
transmissibility of Omicron or other extrinsic significantly lower in the Omicron- than Delta- nificantly lower than that of the Delta-infected
environmental and social factors is unknown. infected hamsters (Fig. 1, D and E). During the hamsters (fig. S3).
At present, the in vivo pathogenicity, trans- acute (4 dpi) and regenerative (7 dpi) phases of The lung sections of the Omicron-infected
missibility, and fitness of Omicron is poorly infection, the viral burden of Omicron became hamsters collected at 2 dpi showed alveolar
understood. We investigated these virological consistently lower than that of Delta throughout wall congestion whereas Delta-infected hamsters
attributes of Omicron by comparing them the upper and lower respiratory tract (Fig. 1, exhibited more severe and diffuse peribron-
with those of Delta in a Syrian hamster model D and E). At 7 dpi, the viral titers in the trachea chiolar and alveolar inflammatory infiltrates
of COVID-19, which closely simulates non- and lungs were already below the detection (fig. S4). At 4 dpi, both groups of hamsters
lethal human disease and has been widely limit [<100 plaque-forming units (PFU)/g] in the exhibited bronchiolar epithelial destruction
used to study various aspects of SARS-CoV-2 Omicron-infected hamsters. Earlier clearance and peribronchiolar and perivascular inflam-
infection biology (19–25). of virus shedding in oral swabs (fig. S1A) and matory infiltrates. However, the pathological
We first compared the clinical signs, viral feces (fig. S1B) were observed in the Omicron- changes in the Delta-infected hamsters were
burden, and cytokine and chemokine profiles infected hamsters. Consistent with these find- more diffuse than the Omicron-infected ham-
of Omicron and Delta in our hamsters (Fig. 1A), ings, the Omicron-infected hamsters generally sters. At 7 dpi, the lung sections of the Omicron-
finding that Omicron-infected animals showed expressed lower levels of inflammatory cyto- infected hamsters appeared mostly normal
limited reductions in body weight (<5%) (Fig. kine and chemokine genes (Fig. 1F) and/or whereas those of the Delta-infected hamsters

Co-housing
for 4 hours
before
separation
Co-housing
for 4 hours
before
separation
Fig. 3. Comparative in vivo fitness of Omicron and Delta in Syrian hamsters. serum samples (n = 6) collected at day 100 after vaccination against authentic
Schemes of the in vivo competition models with (A) nonvaccinated and Omicron and Delta using a microneutralization assay. ID50, 50% inhibitory dose.
(B) vaccinated index hamsters. (C) The hamster serum samples (n = 6) were (E) The Omicron-to-Delta ratios in the nasal turbinate, trachea, and lung of
collected at the indicated days following vaccination for detection of antibody against the nonvaccinated and vaccinated index hamsters, and those of the (F) naïve
wild-type SARS-CoV-2 spike receptor-binding domain (RBD) (HKU-001a strain, hamsters exposed to the nonvaccinated or vaccinated index hamsters (Student’s
GenBank accession number: MT230904) with an enzyme-linked immunosorbent t test). Data represent mean ± standard deviation of the results of n = 6 biological
assay (Student’s t test). (D) The neutralizing activity of the vaccinated hamster replicates. **P < 0.01, ***P < 0.001, and ****P < 0.0001.
continued to exhibit blood vessel congestion The experiment was repeated twice. All index than that of Delta in both rounds of exper-
and alveolar wall inflammatory infiltration. hamsters had similar nasal wash virus titers iments if the individual numbers of RT-PCR–
This suggested that lung damage was resolved at 1 dpi (Fig. 2B). All 12 naïve hamsters were positive naïve hamsters were counted.
more quickly in the Omicron-infected ham- found to be infected 2 days after exposure (Fig. To investigate why Omicron emerged as the
sters. The histological scores were significantly 2C), indicating that both variants are highly dominant circulating SARS-CoV-2 variant, we
lower in the Omicron-infected hamsters be- transmissible through close contact. The mean compared its fitness with that of Delta. Con-
tween 2 and 7 dpi. Additionally, viral nucleocap- virus titer in the lungs of Delta-infected naïve sistent with our recent preliminary findings
sid proteins were more abundantly expressed hamsters was significantly higher than that at an early time point, Delta consistently ex-
in the lung sections of the Delta-infected than of Omicron-infected naïve hamsters (fig. S8). hibited a significant fitness advantage over
Omicron-infected hamsters throughout 2 to Next, we randomly grouped 42 hamsters into Omicron for up to 72 hours post infection
7 dpi (fig. S5). Similarly, the Omicron-infected six groups of index and naïve hamsters (1:6 in vitro (fig. S9, A and B) (26). However, this
hamsters generally showed less severe histo- ratio) in our established noncontact transmis- scenario changed when selection pressure by
pathological changes (fig. S6) and less abun- sion system; we then repeated the experiment vaccinated sera containing antibodies with
dant viral nucleocapsid protein expression twice (total n = 84) (Fig. 2D) (20). The ham- reduced anti-Omicron but preserved anti-
(fig. S7) in their nasal turbinates than the sters were sacrificed at 2 dpi (index) or 2 days Delta neutralizing activity was present (fig.
Delta-infected hamsters from 4 to 7 dpi. after exposure to index (naïve). All index ham- S9C), with Omicron significantly outcompet-
Thus, Omicron exhibits attenuated patho- sters were successfully infected with similarly ing Delta (fig. S9, D to F). We next validated
genicity in Syrian hamsters compared with high nasal wash virus titers at 1 dpi (Fig. 2E). our in vitro findings with in vivo competi-
Delta. All 12 cages, including 30 out of 36 (83.3%) tion models. We included both nonvaccinated
Another key question we sought to answer Omicron-exposed and 24 out of 36 (66.7%) (Fig. 3A) and vaccinated (Fig. 3B) index ham-
was the comparative transmissibility of Omi- Delta-exposed naïve hamsters became infected sters and intranasally challenged them with
cron and Delta in vivo. To this end, we first (Fig. 2F). Although the sample size was under- the two variants (1:1 ratio). The vaccinated
cohoused six index SARS-CoV-2–challenged powered to reach statistical significance (P = index hamsters were observed 100 days after
hamsters (n = 3 for each variant) with six naïve 0.173, chi-square test), the transmission rate vaccination with an inactivated SARS-CoV-2
hamsters for 4 hours in a 1:1 ratio (Fig. 2A). of Omicron was consistently ~10 to 20% higher vaccine and showed waning serum antibody

responses in comparison with the peak activity The transmissibility of Omicron is a key than Delta, Omicron undoubtedly still causes
at 28 days postvaccination (Fig. 3C). Their factor in optimizing public health control obvious disease in infected hosts. Taking into
serum-neutralizing antibody activity against measures and predicting the evolution of the consideration Omicron’s high transmissibil-
Omicron was markedly lower than Delta (Fig. pandemic. Recent epidemiological studies ity, our findings highlight the urgent need for
3D). In the nonvaccinated index hamsters, have suggested that Omicron may be spreading next-generation COVID-19 vaccines and broad-
Delta significantly outcompeted Omicron. By even faster (up to 4.2 times) than Delta in its spectrum therapeutics, as well as improve upon
contrast, Omicron exhibited a marked fitness early stage (32). The estimated effective repro- nonpharmaceutical measures to reduce acute
advantage over Delta in the vaccinated index ductive number (Rt) of Omicron in South Africa and chronic disease burden (Long Covid) on
hamsters (Fig. 3E and fig. S10). Delta similarly and the UK is 2.5 to 3.7, with a doubling time of the general public and healthcare facilities.
outcompeted Omicron in naïve hamsters that 3 days (33). Our head-to-head comparison showed
were exposed to nonvaccinated index hamsters. that Omicron exhibits similar or higher trans- REFERENCES AND NOTES
By contrast, when naïve hamsters were ex- missibility than Delta through both contact and 1. J. Cohen, D. Normile, Science 367, 234–235 (2020).
2. P. Zhou et al., Nature 579, 270–273 (2020).
posed to vaccinated index hamsters the repli- noncontact transmission despite generally having 3. J. F. Chan et al., Lancet 395, 514–523 (2020).
cation advantage of Delta was significantly lower respiratory tract viral loads. Other factors 4. K. K. To et al., Emerg. Microbes Infect. 10, 507–535
diminished (Fig. 3F). Thus, Delta shows a such as the efficiency of the variants in entering (2021).
5. N. G. Davies et al., Science 372, eabg3055 (2021).
fitness advantage over Omicron in the absence cells and their ability to remain as infectious 6. L. Ulrich et al., Nature 602, 307–313 (2022).
of selection pressure. Under immune selection particles in aerosols or on inanimate surfaces for 7. World Health Organization, Classification of Omicron
pressure, however, Omicron becomes the domi- prolonged periods should be investigated (34). (B.1.1.529): SARS-CoV-2 Variant of Concern (WHO, 2021);
https://www.who.int/news/item/26-11-2021-classification-of-
nant variant causing infection. To provide insight on the replacement of omicron-(b.1.1.529)-sars-cov-2-variant-of-concern.
Novel SARS-CoV-2 variants will continue to Delta by Omicron as the dominant SARS-CoV-2 8. World Health Organization, Enhancing response to Omicron
emerge as long as the virus maintains its wide variant, we compared their fitness in cell cul- SARS-CoV-2 variant: Technical brief and priority actions
for Member States (WHO, 2022); https://www.who.int/docs/
circulation among humans and nonhuman ture and hamster models. Notably, the rapidly default-source/coronaviruse/2022-01-07-global-technical-
mammals. Although it has recently become disseminating Omicron was consistently out- brief-and-priority-action-on-omicron—corr2.pdf?sfvrsn=
clear that Omicron exhibits immune evasion competed by Delta in vitro and in nonvaccinated 918b09d_20.
9. L. Lu et al., Clin Infect Dis. ciab1041 (2021).
to most existing anti–SARS-CoV-2 therapeutic hamsters, which may be the result of its un- 10. L. Liu et al., Nature 602, 676–681 (2022).
monoclonal antibodies and vaccine-induced usually high number of genetic mutations. 11. S. Cele et al., Nature 602, 654–656 (2022).
neutralizing antibodies (9–15), understanding Omicron exhibits a significant fitness advan- 12. Y. Cao et al., Nature 602, 657–663 (2022).
13. D. Planas et al., Nature 602, 671–675 (2022).
of the in vivo pathogenicity, transmissibility, tage over Delta under selection pressure in vitro 14. E. Cameroni et al., Nature 602, 664–670 (2022).
and fitness of this VOC remains incomplete. In in the presence of vaccinated serum and in ham- 15. J. Ai et al., Emerg. Microbes Infect. 11, 337–343 (2022).
the pathogenicity study we demonstrated that sters with waning serum-neutralizing antibody 16. H. Ledford, Nature 600, 577–578 (2021).
17. L. Espenhain et al., Euro Surveill. 26, 2101146 (2021).
although the viral load and infectious virus levels. These findings help explain why Omicron
18. L. T. Brandal et al., Euro Surveill. 26, 2101147 (2021).
titer of the two variants were similar in the has outcompeted Delta and has become the 19. J. F. Chan et al., Clin. Infect. Dis. 71, 2428–2446
nasal turbinates and trachea, Omicron is sig- predominant SARS-CoV-2 strain especially (2020).
nificantly less replicative in the lungs even at in populations with high rates of previous 20. J. F. Chan et al., Clin. Infect. Dis. 71, 2139–2149 (2020).
21. J. F. Chan et al., Clin Infect Dis. ciab817 (2021).
the early post infection stage (2 dpi). Omicron infection and/or vaccination with first gen- 22. C. Li et al., Clin. Infect. Dis. ciac142 (2022).
also consistently induces less cytokine and eration COVID-19 vaccines eliciting subop- 23. A. J. Zhang et al., Clin. Infect. Dis. 73, e503–e512
chemokine dysregulation and tissue damage timal anti-Omicron neutralizing antibody (2021).
24. G. D. de Melo et al., Sci. Transl. Med. 13, eabf8396
in the lungs. In human lung–derived Calu-3 responses. However, our findings should be (2021).
cells, Omicron shows reduced replication com- interpreted carefully and should not be con- 25. J. A. Plante et al., Nature 592, 116–121 (2021).
pared with Delta and the D614G strain in sidered as evidence against COVID-19 vac- 26. H. Zhao et al., Emerg. Microbes Infect. 11, 277–283
(2022).
pseudovirus and/or live virus assays (26, 27). cination. By contrast, our findings in hamsters 27. C. Zeng et al., bioRxiv 2021.12.16.472934 [Preprint] (2021);
Moreover, the Omicron spike exhibits reduced with waning serum-neutralizing antibody doi: 10.1101/2021.12.16.472934v1.
receptor binding and fusogenicity as well as >3 months after vaccination are supportive of 28. K. P. Y. Hui et al., Nature 603, 715–720 (2022).
29. H. Shuai et al., Nature 603, 693–699 (2022).
S1 subunit shedding in vitro (26, 27). A recent booster vaccines because recent data have
30. P. J. Halfmann et al., Nature 603, 687–692 (2022).
study using an ex vivo lung organ culture model shown that antibody neutralization is mostly 31. Imperial College London, Report 50 - Hospitalisation risk for
showed that Omicron exhibits enhanced repli- restored by mRNA vaccine booster doses (35). Omicron cases in England (ICL, 2021); https://www.imperial.ac.
cation compared with Delta in the bronchi (28). Our study has certain limitations. The trans- uk/mrc-global-infectious-disease-analysis/covid-19/report-50-
severity-omicron/.
This differs from the findings of the present mission rate of SARS-CoV-2 may vary according 32. H. Nishiura et al., J. Clin. Med. 11, 30 (2021).
study and recently reported animal model to different durations of exposure and diag- 33. World Health Organization, Enhancing readiness for Omicron
data which show that Omicron is generally nostic criteria applied. We selected 6 hours of (B.1.1.529): Technical brief and priority actions for member
less replicative than Delta throughout the noncontact transmission to simulate the real- states (WHO, 2021); https://www.who.int/docs/default-
source/coronaviruse/20211217-global-technical-brief-and-
upper and lower respiratory tract (29, 30). life scenarios of RT-PCR testing after being priority-action-on-omicron_latest-2.pdf?sfvrsn=bdd8297c_
This apparent discrepancy may be caused exposed to an infected index patient within the 9&download=true.
by different study models and conditions. same facility for a routine business day and on 34. R. Hirose et al., bioRxiv 2022.01.18.476607v2 [Preprint] (2022).
35. B. J. Gardner, A. M. Kilpatrick, medRxiv [Preprint]
Our in vivo findings help explain the obser- medium-haul flights. It would be worthwhile to
2021.12.10.21267594v2 (2021).
vations in early epidemiological studies that compare the transmissibility of Omicron and
report lower rates of hospitalization caused Delta after different durations of exposure; it AC KNOWLED GME NTS
by Omicron compared with Delta (16, 31). may also be important to investigate the path- We thank the staff at the Centre for Comparative Medicine Research
of The University of Hong Kong for their facilitation of this study.
The shorter duration of virus shedding in oral ogenicity and transmissibility of Omicron in
Funding: This study was partly supported by funding from the
swabs and feces of Omicron-infected ham- additional animal models as each model has Health and Medical Research Fund (20190572 and COVID1903010,
sters may also have implications for infection its own advantages and disadvantages in re- Projects 6, 7, and 15), the Food and Health Bureau, The
Government of the Hong Kong Special Administrative Region
control of Omicron-infected patients should capitulating human disease.
(to S.Y., H. Chen., and J.F.-W.C.); the Collaborative Research Fund
the same viral shedding pattern be confirmed In summary, the present study shows that (C7060-21G) (to J.F.-W.C.) and Theme-Based Research Scheme
in humans. despite comparatively lower pathogenicity (T11-709/21-N) (to D.-Y.J.), the Research Grants Council of the

Hong Kong Special Administrative Region; Health@InnoHK, SEMICONDUCTORS

Innovation and Technology Commission, the Government of the
Hong Kong Special Administrative Region (to K.-Y.Y.); the
Consultancy Service for Enhancing Laboratory Surveillance of High ambipolar mobility in cubic boron arsenide
Emerging Infectious Diseases and Research Capability on
Antimicrobial Resistance for Department of Health of the
Hong Kong Special Administrative Region Government (to K.-Y.Y.
revealed by transient reflectivity microscopy
and J.F.-W.C.); the National Program on Key Research Project
of China (2020YFA0707500 and 2020YFA0707504) (to J.F.-W.C.); Shuai Yue1,2,3,4†, Fei Tian5,6†, Xinyu Sui1,4, Mohammadjavad Mohebinia7, Xianxin Wu1,4, Tian Tong2,
Sanming Project of Medicine in Shenzhen, China (SZSM201911014)
Zhiming Wang3, Bo Wu8, Qing Zhang9, Zhifeng Ren5*, Jiming Bao2,5,7*, Xinfeng Liu1,4*
(to K.-Y.Y. and J.F.-W.C.); the High Level-Hospital Program,
Health Commission of Guangdong Province, China (to K.-Y.Y. and
J.F.-W.C.); Emergency Collaborative Project (EKPG22-01) of Semiconducting cubic boron arsenide (c-BAs) has been predicted to have carrier mobility of 1400 square
Guangzhou Laboratory (to K.K.-W.T., H. Chen, D.-Y.J., K.-Y.Y., and centimeters per volt-second for electrons and 2100 square centimeters per volt-second for holes at
J.-F.W.C.); Emergency COVID-19 Project (2021YFC0866100),
Major Projects on Public Security, National Key Research and room temperature. Using pump-probe transient reflectivity microscopy, we monitored the diffusion of
Development Program (to K.K.-W.T., H. Chen, D.-Y.J., K.-Y.Y., photoexcited carriers in single-crystal c-BAs to obtain their mobility. With near-bandgap 600-nanometer
and J.-F.W.C.); the Major Science and Technology Program pump pulses, we found a high ambipolar mobility of 1550 ± 120 square centimeters per volt-second,
of Hainan Province (ZDKJ202003) (to G.L.); the research project
of Hainan Academician Innovation Platform (YSPTZX202004) in good agreement with theoretical prediction. Additional experiments with 400-nanometer pumps
(to G.L., K.-Y.Y., and J.F-.W.C.); the Hainan Talent Development on the same spot revealed a mobility of >3000 square centimeters per volt-second, which we
Project (SRC200003) (to G.L.); the University of Hong Kong attribute to hot electrons. The observation of high carrier mobility, in conjunction with high thermal
Outstanding Young Researcher Award (to J.-F.W.C.); and the
University of Hong Kong Research Output Prize (Li Ka Shing Faculty
conductivity, enables an enormous number of device applications for c-BAs in high-performance
of Medicine) (to J.-F.W.C.); and donations from the Shaw Foundation electronics and optoelectronics.
Hong Kong, Richard Yu and Carol Yu, Michael Seak-Kan Tong,
I
May Tam Mak Mei Yin, Lee Wan Keung Charity Foundation Limited,
Providence Foundation Limited (in memory of the late Lui Hac-Minh),
Hong Kong Sanatorium and Hospital, Hui Ming, Hui Hoy and
n 2018, the predicted high room-temperature transient reflectivity microscopy, we observed
Chow Sin Lan Charity Fund Limited, The Chen Wai Wai Vivien thermal conductivity (k) of cubic boron an ambipolar mobility of ~1550 cm2 V−1 s−1
Foundation Limited, Chan Yin Chuen Memorial Charitable arsenide (c-BAs), >1300 W m−1 K−1, was and obtained a >3000 cm2 V−1 s−1 mobility
Foundation, Marina Man-Wai Lee, the Hong Kong Hainan
Commercial Association South China Microbiology Research Fund,
experimentally demonstrated (1–3). At for photoexcited hot carriers. We used photo-
the Jessie and George Ho Charitable Foundation, Perfect Shape about the same time, c-BAs was also luminescence and Raman spectroscopy to
Medical Limited, Kai Chong Tong, Tse Kam Ming Laurence, predicted to have high carrier mobility val- probe the relative level of p-type doping and
Foo Oi Foundation Limited, Betty Hing-Chu Lee, Ping Cham So, and
ues of 1400 cm2 V−1 s−1 for electrons and found that a high hole concentration will
Lo Ying Shek Chi Wai Foundation. The funding sources had no
role in the study design, data collection, analysis, interpretation, or 2100 cm2 V−1 s−1 for holes (4). A higher hole substantially reduce the ambipolar mobility.
writing of the report. Author contributions: S.Y. and J.F.-W.C. mobility of >3000 cm2 V−1 s−1 was later predicted We grew c-BAs single crystals using the
conceived and designed the study. S.Y., Z.W.Y., R.L., K.T., A.Z., under a small 1% strain (5). Such a high carrier same seeded chemical vapor transport tech-
G.L., C.P.O., V.K.-M.P., C.C.-S.C., B.W.-Y.M., Z.Q., Y.X., A.W.-H.C.,
W.-M.C., J.D.I., H.S., J.O.-L.T., T.T.-T.Y., K.K.-H.C., C.C.Y.C., J.-P.C., mobility is due to a weak electron–phonon nique reported previously (3, 9). These crystals
C.L., C.C.-Y.Y., L.L., K.K.-W.T., H. Chen, H. Chu, and J.F.-W.C. interaction and small effective mass (4–7). Like typically appear as slabs with (111) top and
designed and/or performed experiments. S.Y., G.L., K.K.-W.T., H. Chen, those predicting the thermal conductivity of bottom surfaces. We used scanning electron
D.-Y.J., K.-Y.Y., and J.F.-W.C. acquired funding. S.Y., D.-Y.J., K.-Y.Y.,
and J.F.-W.C supervised the study. S.Y. and J.F.-W.C. wrote the c-BAs, these calculations were based on non- microscopy to image a corner facet (111) of
manuscript, and all authors reviewed and edited the paper. defective c-BAs with high crystal quality and a an as-grown c-BAs slab that we labeled sam-
Competing interests: J.F.-W.C. has received travel grants from very low impurity level (4, 5). The simulta- ple 1 (Fig. 1A). This facet is one of the eight
Pfizer Corporation Hong Kong and Astellas Pharma Hong Kong
Corporation Limited and was an invited speaker for Gilead
neous high thermal conductivity and carrier equivalent (111) surfaces, and we chose it for
Sciences Hong Kong Limited and Luminex Corporation. K.K.-W.T., mobility makes c-BAs a promising material mobility measurement because of its relatively
H. Chen, and K.-Y.Y. have collaboration with Sinovac Biotech Ltd. for many applications in electronics and opto- high quality, which can be seen from sharp
and China National Pharmaceutical Group Co., Ltd. (Sinopharm).
H. Chen and K.-Y.Y. have patent applications on intranasal
electronics. Despite this potential, the high (0.02°) characteristic peaks in the x-ray dif-
vaccines. The other authors declare no competing interests. mobility has not been experimentally verified (8). fraction (XRD) pattern (Fig. 1B and inset), a
Data and materials availability: Complete sequences of In this study, using ultrafast spatial-temporal narrow (0.6 cm−1) longitudinal optical (LO)
the SARS-CoV-2 Omicron (hCoV-19/Hong_Kong/HKU-211129-001/
phonon peak at 700 cm−1 in the Raman spectrum
2021; EPI_ISL_6841980) and Delta (hCoV-19/Hong Kong/HKU-
210804-001/2021; EPI_ISL_3221329) variants are available
1
Chinese Academy of Sciences (CAS) Key Laboratory of (Fig. 1C and inset) (1, 2), and the characteristic
through GISAID. All other data are provided in the manuscript or Standardization and Measurement for Nanotechnology, bandgap photoluminescence (PL) peak at
supplementary materials. License information: This work is National Center for Nanoscience and Technology, Beijing
100190, China. 2Department of Electrical and Computer 720 nm in the PL spectrum (Fig. 1D) (10),
licensed under a Creative Commons Attribution 4.0 International
(CC BY 4.0) license, which permits unrestricted use, distribution, Engineering and Texas Center for Superconductivity at the indicating high-quality crystal lattices, a low
and reproduction in any medium, provided the original work is University of Houston (TcSUH), University of Houston, mass disorder (11), and a low defect density,
properly cited. To view a copy of this license, visit https:// Houston, TX 77204, USA. 3Institute of Fundamental and
Frontier Sciences, University of Electronic Science and respectively (10). PL mapping shown in the
creativecommons.org/licenses/by/4.0/. This license does not
apply to figures/photos/artwork or other content included in the Technology of China, Chengdu, Sichuan 610054, China. inset of Fig. 1D also indicates the uniform crys-
4
article that is credited to a third party; obtain authorization School of Nanoscience and Technology, University of tal quality on the (111) surface (10). We per-
from the rights holder before using such material. Chinese Academy of Sciences, Beijing 100049, China.
5
Department of Physics and Texas Center for
formed all measurements at room temperature
Superconductivity at the University of Houston (TcSUH), and further characterized sample 1 and a sec-
University of Houston, Houston, TX 77204, USA. 6School of ond sample, labeled sample 2 (12) (fig. S1).
SUPPLEMENTARY MATERIALS Materials Science and Engineering, Sun Yat-sen University,
science.org/doi/10.1126/science.abn8939 Guangzhou, Guangdong 510006, China. 7Materials Science
The Hall effect is the most common tech-
Materials and Methods and Engineering Program, University of Houston, Houston, nique used to measure carrier mobility, but it
Figs. S1 to S10 TX 77204, USA. 8Guangdong Provincial Key Laboratory of requires four electrical contacts on a relatively
References (36–45) Optical Information Materials and Technology and Institute
large and uniform sample. To accommodate
MDAR Reproducibility Checklist of Electronic Paper Displays, South China Academy of
Advanced Optoelectronics, South China Normal University, the requirements of mobility measurement
Guangzhou 510006, China. 9School of Materials Science and in a small sample size or in inhomogeneous
Engineering, Peking University, Beijing 100871 China. materials, ultrafast pump-probe techniques
Submitted 28 December 2021; accepted 17 June 2022 *Corresponding author. Email: [email protected] (Z.R.);
Published online 23 June 2022 [email protected] (J.B.); [email protected] (X.L.) have been used to perform noncontact mea-
10.1126/science.abn8939 †These authors contributed equally to this work. surements with high spatial resolution (13–17).

Fig. 1. Characterizations of a c-BAs single crystal

(sample 1) on a corner (111) facet. (A) Scanning
electron microscopy image. Scale bar: 100 mm.
(B) X-ray diffraction pattern. (Inset) Magnified view
of the (111) peak. (C) Raman spectrum excited
by a 532-nm laser. (Inset) High-resolution
spectrum of the LO phonon. (D) Photoluminescence
spectrum excited by a 593-nm laser. (Inset)
photoluminescence mapping from the region marked
by a red rectangle in (A). Scale bar: 10 mm.
a.u., arbitrary units.
Fig. 2. Pump-probe transient reflectivity micros-

copy, carrier dynamics, and diffusion in sample 1.
(A) Schematic illustration of the experimental
setup. CMOS, complementary metal-oxide
semiconductor. (B) Evolution of a 2D transient
reflectivity microscopy image from a spot on
sample 1. Scale bar: 1 mm. (C) Typical transient
reflectivity dynamics (photoexcited carrier density
of 5 × 1018 cm−3). (D) Spatial profile (dots) and
Gaussian fit at 0.5 ps time delay from (B)
(fig. S4). (E) Evolution of variance of Gaussian
distributions extracted from Gaussian fitting
in (D). The corresponding mobility is included.
Because of our relatively thick samples, we and fig. S2). We subsequently obtained an mh are the electron and hole mobility values,
used reflectivity rather than transmission. We ambipolar mobility from the diffusion coef- respectively. Because c-BAs has an electronic
focused a femtosecond pump pulse on c-BAs ficient, D, through the Einstein relation, D/kBT = band structure similar to that of silicon, with
to photoexcite electrons and holes and monitored m/e, where kB is the Boltzmann constant, T is an indirect bandgap in the range of 1.82 to
the diffusion of excited carriers in space and the temperature, m is the mobility, and e is 2.02 eV (6, 7, 10, 18), we chose a 600-nm pump
time with a time-delayed probe pulse defocused the elementary charge. Ambipolar mobility pulse and an 800-nm probe pulse to avoid
on a larger area (6 mm in diameter) (12) (Fig. 2A is given by ma = 2memh/(me + mh), where me and the generation of hot carriers. Two-dimensional

Fig. 3. Carrier diffusion on a cross-sectional

surface of sample 2. (A) PL spectra of six locations
on a cross-sectional surface with increasing
distance from the edge. PL of the spot at 0 mm was
taken from the (111) surface around the edge.
(Inset) Optical image of the sidewall. Dashed circle
indicates location for pump-probe measurements
in (C) to (E). (B) Raman spectra of three of the
six locations shown in (A). (Inset) Magnified view of
the phonon line in the spectra of the five sidewall
locations. (C and D) Spatial profiles (dots) and
Gaussian fits (curves) of photoexcited carriers at
initial concentrations of 4.3 × 1018 cm−3 and
8.6 × 1018 cm−3, respectively, from a location
indicated by the dashed circle in (A). (E) Variance
and ambipolar mobility values from (C), (D),
and fig. S6.
Fig. 4. Transient reflectivity microscopy and

carrier diffusion measured using a 400-nm
pump and a 585- or 530-nm probe. (A) Repre-
sentative pump-probe transient reflectivity curve
from sample 1. The probe wavelength is 585 nm.
(B and C) Spatial profiles (dots) and Gaussian fits
(curves) of transient reflectivity from a spot in
sample 1 measured using 585- and 530-nm probes,
respectively. (D) Evolution of the variances of carrier
density distributions and carrier mobility from (B),
(C), and fig. S10. (E and F) Variance and ambipolar
mobility results, respectively, for sample 2 at
six locations corresponding to those shown in
Fig. 3, A and B.
(2D) diffusion images in Fig. 2B show the ex- The spread of distributions in Fig. 2B re- ficient, D, can be calculated from the slope
pansion of carriers over 10 ps, and a representa- flects diffusion of photoexcited electrons and using the equation s2t ¼ s20 þ aDt, where a is
tive time-resolved reflectivity as a function of the holes in space and time, and they can be well a constant depending on the dimensions of
time delay between the pump and the probe is fit by Gaussian functions (Fig. 2D). The change the system and detection configuration (15).
shown in Fig. 2C. A sudden negative differential in the variance s2 of carrier distributions is We chose an a of 2 for our experiment be-
reflectivity indicates a dominant electronic plotted in Fig. 2E. The linear increase in the cause of the much larger laser penetration
contribution, because reflectivity increases with variance with increasing time delay is a sig- (excitation) depth (60 mm at 600 nm) com-
lattice temperature (12, 19) (fig. S3). nature of diffusion, and the diffusion coef- pared with the thin top layer sampled by the

probe beam [20 nm at 800 nm, given by l/4pn prediction (20). To obtain the diffusion coef- REFERENCES AND NOTES
(13, 15–17, 19), where n is the refractive index of ficient of the carriers in sample 1, we used a 1. J. S. Kang, M. Li, H. Wu, H. Nguyen, Y. Hu, Science 361,
c-BAs (18)]. From the slope of the curve shown simpler method by varying the relative dis- 575–578 (2018).
2. S. Li et al., Science 361, 579–581 (2018).
in Fig. 2E and the Einstein relation, D/kBT = placement between focused pump and probe 3. F. Tian et al., Science 361, 582–585 (2018).
m/e, we obtained an ambipolar diffusion coef- beams along one direction (12) (figs. S7 and 4. T.-H. Liu et al., Phys. Rev. B 98, 081203 (2018).
ficient of ~39 cm2 s−1 and an ambipolar mobility S8). We plotted the resulting spatial profiles 5. K. Bushick, S. Chae, Z. Deng, J. T. Heron, E. Kioupakis, NPJ
Comput. Mater. 6, 3 (2020).
of 1550 ± 120 cm2 V−1 s−1, close to the predicted of the reflectivity after 1 ps for probe wave- 6. J. L. Lyons et al., Appl. Phys. Lett. 113, 251902 (2018).
value (4). lengths of 530 and 585 nm (Fig. 4, B and C) 7. K. Bushick, K. Mengle, N. Sanders, E. Kioupakis, Appl. Phys.
Given that the properties of c-BAs are not (12) and obtained an ambipolar diffusion Lett. 114, 022101 (2019).
8. X. Chen et al., Chem. Mater. 33, 6974–6982 (2021).
uniform even within a single crystal, especially coefficient of 80 cm2 s−1 and an ambipolar 9. F. Tian et al., Appl. Phys. Lett. 112, 031903 (2018).
in the direction perpendicular to (111) surfaces mobility of ~3200 cm2 V−1 s−1 (Fig. 4D). 10. S. Yue et al., Mater. Today Phys. 13, 100194 (2020).
(3, 10), we tested a cross-sectional surface of a Mobility of ~3600 cm2 V−1 s−1 was obtained 11. A. Rai, S. Li, H. L. Wu, B. Lv, D. G. Cahill, Phys. Rev. Mater. 5,
013603 (2021).
relatively thin (30-mm-thick) crystal labeled from the same spot as that shown in Fig. 2 12. See supplementary materials.
sample 2 (12) (fig. S1). An optical image of the for sample 1. These values are much larger 13. Y. Wan et al., Nat. Chem. 7, 785–792 (2015).
sample 2 sidewall is shown in the inset of than the predicted ambipolar mobility of 14. M. M. Gabriel et al., Nano Lett. 13, 1336–1340 (2013).
15. N. S. Ginsberg, W. A. Tisdale, Annu. Rev. Phys. Chem. 71, 1–30
Fig. 3A. We obtained PL spectra from several 1680 cm2 V−1 s−1 (4).
(2020).
spots at different distances from the edge (Fig. Using the same 400-nm pump, we also 16. R. Wang et al., Phys. Rev. B 86, 045406 (2012).
3A). The PL intensity increases with decreasing measured the ambipolar mobility of sample 2 17. L. Yuan et al., Nat. Mater. 19, 617–623 (2020).
distance from the edge and exhibits a noticeable at six locations corresponding to those shown 18. B. Song et al., Appl. Phys. Lett. 116, 141903 (2020).
19. A. J. Sabbah, D. M. Riffe, Phys. Rev. B 66, 165217
jump upon reaching the (111) surface, which in Fig. 3, A and B. The evolution of variance (2002).
agrees with our previous finding of a drastic of carrier distribution at these spots is shown 20. S. Sadasivam, M. K. Y. Chan, P. Darancet, Phys. Rev. Lett. 119,
change in PL from one surface to the opposite in Fig. 4, E and F, and fig. S11. The differences 136602 (2017).
21. Z. Y. Tian et al., Phys. Rev. B 105, 174306 (2022).
surface of a single crystal slab (10). Corresponding in the initial values of the variances at 1 ps are 22. A. Block et al., Sci. Adv. 5, eaav8965 (2019).
Raman spectra from the same locations are due to the different spot sizes of the pump 23. B. A. Ruzicka et al., Phys. Rev. B 82, 195414 (2010).
shown in Fig. 3B and its inset. Similar to the and probe beams in each measurement. The 24. E. Najafi, V. Ivanov, A. Zewail, M. Bernardi, Nat. Commun. 8,
15177 (2017).
PL results, the Raman spectrum of the (111) mobility clearly changes drastically across the 25. S. Y. Lim et al., IEEE J. Photovolt. 3, 649–655 (2013).
surface differs substantially from those of the sidewall, with the highest mobility (5200 ±
sidewall. We chose a spot ~11 mm from the 600 cm2 V−1 s−1) observed at a depth of 9.9 mm. AC KNOWLED GME NTS
edge (Fig. 3A, dashed circle in inset) and used Although local strain could result in such prom- We thank X. Bo and L. Yang for help with high-resolution Raman
and XRD measurements. We thank J. Ding and X. Qiu for help
three pump fluences to create different carrier inent carrier mobility enhancement (5), we did with the AFM measurement. We thank J. Zhao for helpful
densities, the reflectivity distributions of which not see any noticeable Raman shift among discussion. Funding: The work performed in China was
are shown in Fig. 3, C and D, and fig. S6 (12). these locations (Fig. 3B). We thus attribute the supported by the Strategic Priority Research Program of the
Chinese Academy of Sciences (XDB36000000); the Ministry of
We plotted the evolution of the variances high ambipolar mobility to photoexcited hot Science and Technology (2017YFA0205004); the National
and obtained an ambipolar mobility of carriers, which exhibit high carrier diffusion Natural Science Foundation of China (22173025, 22073022,
~1300 cm2 V−1 s−1 (Fig. 3E), indicating the coefficient and mobility values (20, 22–24). 11874130, and 52172171); the CAS Instrument Development
Project (Y950291); and the DNL Cooperation Fund, CAS
negligible effect of carrier density on the The position-dependent mobility on the
(DNL202016). The work performed at the University of Houston
mobility of sample 2 owing to nonlinear effects sidewall of sample 2 reveals that p-type doping (UH) is supported by the Office of Naval Research under
such as Auger recombination. in c-BAs can substantially reduce its mobility. Multidisciplinary University Research Initiative grant N00014-16-
The high carrier mobility of c-BAs is enabled Heavy p-type doping on the (111) surface can 1-2436, the Welch Foundation (E-1728), and a UH Small
Equipment Grant (000182016). Author contributions: X.L.,
by its distinctive weak electron–phonon inter- be seen from the Fano line shape of the LO J.B., and Z.R. conceived of the project. S.Y. performed the
action and its phonon–phonon scattering, phonon at 700 cm−1 and the higher background Raman, PL, transient reflectivity, and transient reflectivity
which should also enable the generation of level around 1000 cm−1 (Fig. 3B) (2, 8). This microscopy experiments. F.T. grew the crystal samples. X.S. and
S.Y. built the transient reflectivity setup. S.Y. built the transient
high-mobility hot carriers (20). To prove this, gradually increased doping level toward the reflectivity microscopy setup. M.M. performed the carrier
we used a 400-nm pulse as a pump and (111) surface is further supported by the cor- diffusion and laser heating simulation. X.W. performed the high-
selected a particular band (585 or 530 nm) responding increased PL intensity (10, 25). resolution Raman experiment. T.T. performed the temperature-
dependent reflectivity experiment. B.W. performed the carrier
with an optical filter from a white light P-type doping will result in reduced carrier diffusion theoretical analysis. J.B., S.Y., Z.R., X.L., Z.W., and
continuum beam as a probe pulse (12) (fig. mobility owing to the presence of ionized Q.Z. wrote the paper. All authors contributed to the discussion
S7). A typical transient reflectivity curve of dopants (these dopants are already activated) of the results and writing of the manuscript. Competing
interests: The authors declare that they have no competing
a probe (585 nm) from sample 1 is shown in and a lower electron mobility than hole mobil- interests. The c-BAs crystals were grown by the method
Fig. 4A. In contrast to the single exponen- ity, because minority carriers will dominate the disclosed in US patent publication 20210269318 and a new
tial decay previously observed when excited by carrier dynamics. The latter is supported by our patent filing on high mobility. Data and materials availability:
All data are available in the main text or the supplementary
a 600-nm pump (Fig. 2C), the dynamics of observation of a higher ambipolar mobility in
materials. License information: Copyright © 2022 the authors,
photoexcited carriers excited by the 400-nm p-type silicon than in undoped silicon (12) (figs. some rights reserved; exclusive licensee American Association
pump consist of three exponential decays: a S12 and S13). Clearly, the enhanced PL intensity for the Advancement of Science. No claim to original US
fast exponential decay with a ~1-ps lifetime, observed in the c-BAs samples in the current government works. https://www.science.org/about/science-
licenses-journal-article-reuse
a slow decay of ~20 ps, and an even slower study indicates that p-type doping has only
decay on the order of 1 ns (21). These decays introduced shallow acceptors rather than non-
SUPPLEMENTARY MATERIALS
correspond to rapid relaxation of high-energy radiative deep levels (10, 25). Because hot
photoexcited carriers, further relaxation of carriers can also be generated by electrical Materials and Methods
carriers to the conduction and valence band injection and low-intensity light, both hot Supplementary Text
edges, and a combination of lattice heating carriers and fully relaxed carriers can be used Figs. S1 to S13
References (26, 27)
and recombination and trapping of elec- for high-speed optoelectronic devices and high-
trons and holes at the band edges, respectively efficiency solar cells in conjunction with the Submitted 1 December 2021; accepted 16 June 2022
(20, 21), in good agreement with the theoretical high mobility of the band-edge carriers. 10.1126/science.abn4727

SEMICONDUCTORS ments and (ii) the decoupling between ther-

mal and electrical properties have not been
High ambipolar mobility in cubic boron arsenide identified.
We used an optical transient grating (TG)
Jungwoo Shin1†, Geethal Amila Gamage2†, Zhiwei Ding1†, Ke Chen1, Fei Tian2, Xin Qian1, method to measure electrical mobility and
Jiawei Zhou1, Hwijong Lee3, Jianshi Zhou3, Li Shi3, Thanh Nguyen4, Fei Han4, Mingda Li4, thermal conductivity on the same spot of c-BAs
David Broido5, Aaron Schmidt1, Zhifeng Ren2*, Gang Chen1* single crystals. Our experiments confirm that
c-BAs has simultaneous high thermal con-
Semiconductors with high thermal conductivity and electron-hole mobility are of great importance for ductivity and high electron and hole mobili-
electronic and photonic devices as well as for fundamental studies. Among the ultrahighÐthermal ties. Using ab initio calculations, we show that
conductivity materials, cubic boron arsenide (c-BAs) is predicted to exhibit simultaneously high electron ionized impurities strongly scatter charge car-
and hole mobilities of >1000 centimeters squared per volt per second. Using the optical transient riers, whereas neutral impurities are mainly
grating technique, we experimentally measured thermal conductivity of 1200 watts per meter per kelvin responsible for the thermal conductivity re-
and ambipolar mobility of 1600 centimeters squared per volt per second at the same locations on duction. These findings establish c-BAs as the
c-BAs samples at room temperature despite spatial variations. Ab initio calculations show that lowering only known semiconductor with this combi-
ionized and neutral impurity concentrations is key to achieving high mobility and high thermal nation of desirable properties and place it
conductivity, respectively. The high ambipolar mobilities combined with the ultrahigh thermal among the ideal materials for next-generation
conductivity make c-BAs a promising candidate for next-generation electronics. microelectronics applications.
We prepared c-BAs samples using multistep
T
chemical vapor transport with varying condi-
he performance of microelectronic and now been demonstrated experimentally (9–11), tions (18) (figs. S1 and S2). We used scanning
optoelectronic devices benefits from semi- with measured c-BAs thermal conductivities electron microscopy (SEM) to image a c-BAs
conductors with simultaneously high in the range of kRT = 1000 to 1300 Wm−1K−1, single crystal with a thickness of ~20 mm (Fig. 1,
electron and hole mobilities and high identifying c-BAs as the most thermally con- A and B) and confirmed the cubic structure
thermal conductivity (1, 2). However, mo- ductive semiconductor other than diamond. with x-ray diffraction (XRD) (Fig. 1C), in agree-
bility and thermal conductivity measurements First-principles calculations have also pre- ment with the literature (19).
have thus far identified no such materials. dicted that c-BAs should have simultaneously We used photoluminescence (PL) and Ra-
Two of the most widely used semiconductors, high RT electron and hole mobilities of me = man spectroscopies to identify the nonuni-
silicon and gallium arsenide (GaAs), for exam- 1400 cm2V−1s−1 and mh = 2100 cm2V−1s−1, respec- form impurity distribution in c-BAs (17, 20).
ple, have high room temperature (RT) elec- tively (12). The major reason for such high We measured the PL spectrum (Fig. 1D) and
tron mobilities of me = 1400 cm2V−1s−1 and electron and hole mobilities is the high energy performed two-dimensional (2D) PL mapping
8500 cm2V−1s−1, respectively. However, their and low occupation of polar optical phonons of c-BAs crystals (Fig. 1E). Local bright spots
corresponding RT hole mobilities (mh = in c-BAs, which give rise to weak carrier scat- indicate the spatial differences in charge car-
450 cm2V −1 s−1 for Si and 400 cm2V−1s−1 for tering. This feature distinguishes c-BAs from rier density and recombination dynamics. We
GaAs) and thermal conductivities (kRT = other III-V semiconductors, which have high also measured the Raman spectrum (Fig. 1F)
140 Wm−1K−1 for Si and 45 Wm−1K−1 for GaAs) electron mobility but much lower hole mobility, and performed 2D Raman background scat-
are lower than desired. Although graphene where me/mh > 10 to ~100 (13, 14), except for AlSb tering intensity (IBG) mapping (Fig. 1G). The
has high electron and hole mobilities and a (me = 200 cm2V−1s−1 and mh = 400 cm2V−1s−1). strong Raman peak at ~700 cm−1 is associated
high in-plane thermal conductivity, the cross- Despite the promising theoretical predic- with the longitudinal optical (LO) mode of
plane heat conduction is low (3, 4). Diamond tions, experimental measurements have not c-BAs at the zone center. The full width at
has the highest RT thermal conductivity and found high mobilities in BAs. Similar to the half maximum of the LO peak and IBG can be
excellent electron and hole mobilities; how- history of the development of other III-V semi- attributed to mass disorder resulting from
ever, its large bandgap of 5.4 eV hinders its conductors (15), the initial quality of c-BAs impurities, responsible for large k variation
effective doping and utilization as a semi- crystals has been limited by large and non- (11, 21).
conductor material (5). Recently, first-principles uniform defect concentrations. Because tradi- We used the TG technique (22–24) (Fig. 2A)
calculations have predicted that cubic boron tional bulk transport measurement methods to simultaneously measure electrical and ther-
arsenide (c-BAs) should have exceptionally high can only obtain the defect-limited behaviors mal transport on multiple spots (Fig. 1, circles
RT thermal conductivity of ~1400 Wm−1K−1, instead of the intrinsic properties, the high de- a to d). Two femtosecond laser pulses (pump)
10 times as high as that of Si. This high value fect densities in c-BAs crystals have prevented with wavevectors k1 and k2 create sinusoidal
stems from its unusual phonon dispersions and such measurements from assessing the va- optical interference on the c-BAs samples, ex-
chemical bonding properties that promote lidity of the predicted high mobilities. Fur- citing electron-hole pairs accordingly (fig. S3).
simultaneously weak three-phonon and four- thermore, previous studies have shown that A third laser pulse (k3; probe) arrives at the
phonon scattering (6–8). This prediction has thermal conductivity and electronic mobil- sample spot after delay time t, which is sub-
ity do not seem to have a strong relationship sequently diffracted to the direction of k1 − k2 +
1
with each other. Kim et al. measured kRT = k3 and mixed with a fourth pulse (k4) for het-
Department of Mechanical Engineering, Massachusetts
Institute of Technology, Cambridge, MA 02139, USA.
186 Wm−1K−1 and estimated mh = 400 cm2V−1s−1 erodyne detection. As the photoexcited car-
2
Department of Physics and Texas Center for Superconductivity, of a c-BAs microrod sample (16). Chen et al. mea- riers undergo diffusion and recombination,
University of Houston, Houston, TX 77204, USA. 3Materials sured kRT = 920 Wm−1K−1 and mh = 22 cm2V−1s−1 the corresponding diffraction signal decays
Science and Engineering Program, The University of Texas at
of millimeter-scale c-BAs crystals (17). The ob- with t. We show the calculated time-dependent
Austin, Austin, TX 78712, USA. 4Department of Nuclear
Science and Engineering, Massachusetts Institute of tained mobilities are much lower than the electron-hole profile in c-BAs in Fig. 2B and
Technology, Cambridge, MA 02139, USA. 5Department of calculated mobility and do not show a clear figs. S4 and S5.
Physics, Boston College, Chestnut Hill, MA 02467, USA. correlation with the measured thermal con- Diffusion and recombination of photoex-
*Corresponding author. Email: [email protected] (Z.R.);
[email protected] (G.C.) ductivity. The origins of (i) the discrepancy cited carriers result in a fast exponential decay
These authors contributed equally to this work. between ab initio calculations and experi- in the TG signal (t < 1 ns), followed by a slower

Fig. 1. Optical characterization of c-BAs single crystals. (A) Optical photograph. (B) SEM image. (C) XRD. a.u., arbitrary units; deg, degrees. (D and E) A typical
PL spectrum (D) and 2D PL intensity mapping (E) integrated over 100-nm spectrum range for each spot. The dashed circles show TG measurement spots (a to d). cps,
counts per second. (F and G) A typical Raman spectrum (F) and 2D mapping of background Raman scattering intensity (G) integrated over 100 cm−1 for each spot.
Fig. 2. Thermal and electron transport measurements. (A) Schematic illustration of TG experiments. (B) Calculated time-dependent electron-hole pair density in
c-BAs. CB, conduction band; VB, valence band; Eg, bandgap. (C) TG signal for c-BAs. Thermal conductivity is calculated from exponential fitting (red line). (D) Wavelength-
dependent electrical decay rate Ge and TG peak amplitude. (E) TG signal with varying diffraction grating periods q. (F) Electrical decay rate (Ge) and thermal decay
rate (Gth) versus q2. Error bars show experimental uncertainties.

thermal decay (t > 1 ns) with an opposite sign neutral impurities. Consequently, the thermal edly with charged impurities from 1016 cm−3,
(Fig. 2C). The short and long time decays are conductivity reduction from ionized impurities regardless of the mass of the impurity.
used to calculate charge carrier mobility and is smaller than that caused by the un-ionized We elucidated the different effects of neu-
thermal conductivity on the same spot, re- impurities, especially when the substituted im- tral and charged impurities on k and ma (Fig.
spectively (see fig. S6 for details). Thermal purity has a similar mass to that of the host 3D). Neutral impurities more strongly suppress
conductivity is directly calculated from the atom—i.e., Ge–As and Cþ B. k because of stronger bond perturbations com-
exponential fitting of the long time decay The bond perturbation and Coulomb poten- pared with those in charged impurities (27).
(red line). The electrical decay is sensitive to tial of impurities modify electron and hole Charged impurities predominantly contribute
the wavelength of the pump pulses. We use an transport dynamics in c-BAs differently. Build- to ma reduction regardless of their mass as a
optical parametric amplifier (OPA) to match ing on recent developments in computing for- result of Coulombic scattering. Charged im-
the wavelength of the pump beam with the mation energies for charged impurities (29), purities with masses similar to that of the host
bandgap (2.02 eV) of c-BAs to avoid excitation we used ab initio calculations to study the ef- atom would exhibit kRT above 1000 W m−1 K−1,
of high-energy electrons that can lead to hot fect of group IV impurities on the RT ma of c-BAs even at a high impurity density of 1019 cm−3,
electrons and holes with different scattering (Fig. 3B). We show electron-phonon scattering and ma is significantly reduced to below
dynamics and mobilities (25). We also deter- and long- and short-range defect scattering for 400 cm2V−1s−1 at a moderate level of 1018 cm−3.
mined the wavelength-dependent electrical holes in c-BAs with Si–As (see fig. S10 for details) We can also highlight the contrasting trends
decay rate Ge and the lock-in amplifier ampli- (Fig. 3C). Long-range Coulombic interaction in k and ma with neutral and charged impu-
tude of the TG peak (Fig. 2D). TG decays much with charged impurities is found to be the dom- rities from batches 0 to IV (Fig. 4A and table
faster at shorter wavelengths (l < 500 nm) and inant scattering mechanism near the band S1) (18). Solid and dashed lines in Fig. 4 show
reaches a plateau near the bandgap (l ~ 600 nm) edge. The lack of a Coulomb potential for neu- the trajectories of the calculated ma and k with
followed by signal loss for photon energy be- tral impurities results in a weaker carrier scat- neutral Si0As and charged Si–As from 1016 to
low the bandgap (l > 650 nm) (fig. S7). The tering, causing ma to not decrease until the 1020 cm−3, respectively. Scattered points are
slopes of electrical decay Ge and thermal decay concentration approaches 1018 cm−3, where the the measured ma and k values of samples from
Gth versus q2 (Fig. 2, E and F) are equivalent to electron-neutral impurity scattering starts to different batches, labeled with different colors.
the ambipolar diffusivity Da and thermal dif- show an effect. However, ma decreases mark- All measured data fit into the area between
fusivity Dth of c-BAs. Da is subsequently con-
verted to ambipolar mobility ma = eDa/kBT =
2memh/(me + mh), which is dominated by the low
mobility carrier, where kB is the Boltzmann
constant, e is the elementary charge, and T is
temperature.
We measured a wide variation of the RT
k and ma for spots a to d (a: 920 Wm−1K−1 and
731 cm2V−1s−1; b: 1132 Wm−1K−1 and 1482
cm2V−1s−1; c: 163 Wm−1K−1 and 331 cm2V−1s−1;
d: 211 Wm−1K−1 and 328 cm2V−1s−1). This large
spatial variation of thermal and electrical prop-
erties can be attributed to corresponding var-
iations in impurity density. A higher impurity
density lowers PL intensity and increases IBG.
To corroborate this trend, we intentionally
doped c-BAs with C (batch IV) and mea-
sured k = 200 to 953 Wm−1K−1 and ma = 195 to
416 cm2V−1s−1 along with large variation in IBG
and low PL intensity (figs. S8 and S9).
Common impurities in c-BAs are group IV
elements, such as C and Si. These impurities
can serve as electron acceptors in c-BAs be-
cause of low formation energies (26). Space
charges created by ionized impurities intro-
duce distortions in the local bonding envi-
ronment, driving distinct phonon scattering
mechanisms. The k of c-BAs can be calculated
by solving the phonon Boltzmann transport
equation, including three- and four-phonon
scattering and phonon-scattering by neutral
(solid lines) and charged (dashed lines) group
IV impurities on B or As sites (27, 28) (Fig. 3A). Fig. 3. Theoretical calculation of the impurity effects on thermal conductivity and mobility. (A and
Our calculated k decreases with increasing B) Calculated thermal conductivity (A) and ambipolar mobility (B) with neutral (solid lines) and charged
mass difference between the impurity and host (dashed lines) group IV impurities. Open circles are mh values of bulk samples measured by electrical
atoms. Upon impurity ionization, the num- probes (fig. S12). (C) Calculated electron-phonon and short- and long-range impurity scattering rates for
ber of valence electrons of the impurity (IV)
matches that of B or As (III or V), resulting in holes. Zero of energy is at the valence band maximum. SiÐAs ¼ 1018 cm 3 . (D) Thermal conductivity (solid
weaker bond perturbations than those from the lines) and mobility (dashed lines) differences between charged and neutral impurities.

Fig. 4. Ambipolar mobility and

thermal conductivity of c-BAs.
(A) Measured mobility and
thermal conductivity of c-BAs
from different batches (batches
0, I, II, III, and IV). See table S1
for details. The solid and dashed
lines show the calculated ma and k
with varying concentrations of neu-
tral Si0As and charged Si–As , respec-
tively. Typical uncertainties for ma
and k are 11%. (B) Temperature-
dependent ambipolar mobility
of c-BAs (III-a and III-b). The
solid and dashed lines show
calculated ma of pristine c-BAs
and Si, respectively (32).
the trajectory curves. Among the high-quality challenges in thermal management for next- 29. C. Freysoldt et al., Rev. Mod. Phys. 86, 253–305 (2014).
c-BAs batch (III), we measure ma = 1600 ± generation electronics. 30. F. Tian et al., Appl. Phys. Lett. 114, 131903 (2019).
31. X. Chen et al., Phys. Rev. Appl. 11, 064070 (2019).
170 cm2V−1s−1 and k = 1200 ± 130 Wm−1K−1. 32. N. D. Arora, J. R. Hauser, D. J. Roulston, IEEE Trans. Electron Dev.
We measured the temperature-dependent ma RE FERENCES AND NOTES 29, 292–295 (1982).
of two different spots (III-a and III-b) of 1. X. Qian, J. Zhou, G. Chen, Nat. Mater. 20, 1188–1202 (2021).
2. G. Chen, Nat. Rev. Phys. 3, 555–569 (2021). AC KNOWLED GME NTS
high-quality samples (fig. S11). Our measured
3. K. S. Novoselov et al., Science 306, 666–669 (2004). Funding: This work was supported by the Office of Naval Research
ma for III-a shows good agreement with 4. A. A. Balandin et al., Nano Lett. 8, 902–907 (2008). under Multidisciplinary University Research Initiative grant
calculations (Fig. 4B). Hall measurements of 5. C. J. H. Wort, R. S. Balmer, Mater. Today 11, 22–28 (2008). N00014-16-1-2436. This work made use of the MRSEC Shared
the bulk samples provide mh and carrier con- 6. L. Lindsay, D. A. Broido, T. L. Reinecke, Phys. Rev. Lett. 111, Experimental Facilities at MIT, supported by the National Science
025901 (2013). Foundation under award no. DMR-1419807. Author contributions:
centration p averaged over the entire sample 7. D. A. Broido, L. Lindsay, T. L. Reinecke, Phys. Rev. B 88, Conceptualization: J.S., G.A.G., K.C., Z.R., and G.C. Materials
with spatially varied impurity concentration. 214303 (2013). preparation: G.A.G. and F.T. TG experiments: J.S. and K.C.
The measured bulk mh plotted in Fig. 3B (see 8. T. L. Feng, L. Lindsay, X. L. Ruan, Phys. Rev. B 96, 161201 (2017). TG modeling: J.S., K.C., X.Q., and G.C. Time-domain thermoreflectance,
9. J. S. Kang, M. Li, H. Wu, H. Nguyen, Y. Hu, Science 361, frequency-domain thermoreflectance, PL, and Raman measurements:
fig. S12 for details) is limited by the average J.S. Hall measurements: H.L., Jian.Z., L.S., T.N., F.H., and M.L.
575–578 (2018).
impurity concentrations rather than local 10. F. Tian et al., Science 361, 582–585 (2018). Mobility calculations: Z.D. and Jiaw.Z. Data analysis: J.S., G.A.G.,
spots with low impurities. 11. S. Li et al., Science 361, 579–581 (2018). Z.D., K.C., D.B., Z.R., and G.C. Funding and project administration:
12. T. H. Liu et al., Phys. Rev. B 98, 081203 (2018). Z.R. and G.C. Supervision: Z.R. and G.C. Writing – original draft:
The high–spatial resolution TG measure-
13. D. L. Rode, Phys. Rev. B 3, 3287–3299 (1971). J.S., G.A.G., Z.D., Z.R., and G.C. Review & editing: all authors.
ments provide clear evidence of simultane- 14. A. Nainani, B. R. Bennett, J. B. Boos, M. G. Ancona, Competing interests: US patent publication no. 20210269318 has
ously high electron and hole mobilities in c-BAs K. C. Saraswat, J. Appl. Phys. 111, 103706 (2012). been granted to the method for growing c-BAs crystals presented
and demonstrate that through the elimination 15. J. I. Pankove, T. D. Moustakas, Semicond. Semimet. 50, 1–10 in this work. A patent application on the high mobility of Bas,
(1997). entitled “Ultra-high ambipolar mobility in cubic arsenide,” has been
of defects and impurities, c-BAs could exhibit 16. J. Kim et al., Appl. Phys. Lett. 108, 201905 (2016). filed with the US Patent Office. The authors declare no other
both high thermal conductivity and high elec- 17. X. Chen et al., Chem. Mater. 33, 6974–6982 (2021). competing interests. Data and materials availability: All data
tron and hole mobilities. Additionally, the 18. Materials and methods are available as supplementary are available in the main text and supplementary materials. License
materials online. information: Copyright © 2022 the authors, some rights reserved;
observed weak correlation between the local 19. J. A. Perri, S. Laplaca, B. Post, Acta Cryst. 11, 310 (1958). exclusive licensee American Association for the Advancement of
thermal conductivity and mobility is caused 20. S. Yue et al., Mater. Today Phys. 13, 100194 (2020). Science. No claim to original US government works. https://www.
by the different effects that neutral and 21. A. Rai, S. Li, H. L. Wu, B. Lv, D. G. Cahill, Phys. Rev. Mater. 5, science.org/about/science-licenses-journal-article-reuse
013603 (2021).
ionized impurities have on these quan- 22. A. A. Maznev, T. F. Crimmins, K. A. Nelson, Opt. Lett. 23,
tities. This notable combination of electronic 1378–1380 (1998). SUPPLEMENTARY MATERIALS
and thermal properties, along with a ther- 23. A. A. Maznev, K. A. Nelson, J. A. Rogers, Opt. Lett. 23,
1319–1321 (1998).
mal expansion coefficient and lattice constant Materials and Methods
24. S. Huberman et al., Science 364, 375–379 (2019).
that are closely matched to common semi- Supplementary Text
25. K. Chen et al., Carbon 107, 233–239 (2016).
Figs. S1 to S12
conductors such as Si and GaAs (30, 31), make 26. J. L. Lyons et al., Appl. Phys. Lett. 113, 251902 (2018).
Table S1
c-BAs a promising material for integrating 27. M. Fava et al., Npj Comput. Mater. 7, 54 (2021).
28. M. Fava et al., How dopants limit the ultrahigh thermal conductivity
with current and future semiconductor manu- of boron arsenide: A first principles study, version 1, Zenodo (2021); Submitted 1 December 2021; accepted 16 June 2022
facturing processes and addressing the grand https://doi.org/10.5281/zenodo.4453192. 10.1126/science.abn4290

online @sciencecareers.org
METABOLIC DISORDERS AND CARDIOVASCULAR HEALTH
As part of a broad institutional initiative in integrative biomedical sciences (IBio), Wayne State University is recruiting for faculty positions (tenured or
tenure-track, open rank) in the integrative biology of cardiometabolic health and disease. The recruitment seeks to complement existing IBio strengths
in cell biology, lipid metabolism, and cardiometabolic physiology. Of particular interest are investigators addressing molecular and cellular processes
that impact energy metabolism, metabolic signaling, and cardiovascular function.
Successful applicants will benefit from a highly collaborative research environment working with colleagues throughout the University. Candidates
must have a Ph.D., M.D., or relevant equivalent degree with evidence of peer recognition in the field, a commitment to excellence in research,
education, and training, and the ability to engage with broader integrative science themes for the purpose of achieving transformative and translational
research gains. Applicants are expected to have already established extramural research funding or be on a clear path to secure and sustain extramural
funding in support of their research programs. Faculty recruits will integrate with departments consistent with their areas of expertise and interests and
engage in all aspects of our academic mission including research, education, and service.
Competitive recruitment packages are available with salary and faculty rank based on qualifications. Applicants should apply to posting #046049
through the Online Hiring System (jobs.wayne.edu/hr). Applications will be accepted until positions are filled, but for full consideration this fall,
application materials should be submitted by September 1, 2022. Applications should include a curriculum vitae and a brief cover letter to the Assistant
Vice President for Integrative Biosciences indicating the applicant’s potential for synergy in the targeted thematic areas.
Wayne State University is a premier, public, urban research university located in the heart of Detroit where students from all backgrounds are offered
a rich, high quality education. Our deep-rooted commitment to excellence, collaboration, integrity, diversity and inclusion creates exceptional
educational opportunities preparing students for success in a diverse, global society. Wayne State University encourages applications from women,
people of color, and other underrepresented people. WSU is an affirmative action/equal opportunity employer.
Founded in 1868, Wayne State University offers a range of academic programs through 13 schools and colleges to nearly 28,000 students. The campus
in Midtown Detroit comprises 100 buildings on over 200 acres and includes the School of Medicine, the College of Engineering, the Eugene
Applebaum College of Pharmacy and Health Sciences, and the College of Nursing. The University is home to the Perinatology Research Branch of
the National Institutes of Health, the Karmanos Cancer Center (a National Cancer Institute-designated comprehensive cancer center), and a National
Institute of Environmental Health Sciences Core Center, CURES (Center for Urban Responses to Environmental Stressors).
JEFFERSON SCIENCE FELLOWSHIPS
Call for Applications

Established by the Secretary of State in 2003, Jefferson Science Fellowships
engage the American academic science, technology, engineering, and medical
communities in U.S. foreign policy and international development.
Science Careers helps you advance Administered by the National Academies of Sciences, Engineering,
your career. Learn how ! and Medicine, these fellowships are open to tenured, or similarly ranked,
faculty from U.S. institutions of higher learning who are U.S. citizens.
 Register for a free online account on ScienceCareers.org. After successfully obtaining a security clearance, selected Fellows spend
one year on assignment at the U.S. Department of State or the U.S. Agency
 Search hundreds of job postings and find your perfect job. for International Development serving as advisers on issues of foreign policy
 Sign up to receive e-mail alerts about job postings that and international development.
match your criteria.
 Upload your resume into our database and connect The application deadline is in October. To learn more and to apply,
with employers. visit www.nas.edu/jsf.
 Watch one of our many webinars on different career topics
such as job searching, networking, and more.
Visit ScienceCareers.org
today — all resources are free
www.twitter.com/NASEMJefferson
SCIENCECAREERS.ORG www.facebook.com/NASEM.JeffersonScienceFellowship
0722Recruitment_RC.indd 441 7/18/22 11:29 AM

WORKING LIFE
By Sophia X. Pfister
From weakness comes strength
I
t was 3 a.m. I was exhausted from taking care of my 3-month-old baby, but I couldn’t sleep. As I
tried to recall the topics of the five conference calls on my calendar for the morning, I again had
the haunting thought that I wasn’t good enough for my job—a director position I started shortly
before my baby was born. I imagined I would make mistakes in my presentations and my team
would lose respect for me. Tormented by these thoughts, I reached for a book from the pile on my
bedside table to distract myself. By chance I grabbed the Bible, which I had been too busy to read
since my baby was born. As I opened it to a random page and happened on the verse “For when I am
weak, then I am strong,” tears filled my eyes, and I could breathe again.
My upbringing gave me an acknowledge what I didn’t know

“achiever” personality. From child- and ask honest questions to learn
hood class president to prestigious from others. And when I got over-
university degrees to a leadership whelmed by my long to-do list, I
position in a large company, I was learned to accept my limits, iden-
regarded as a “star.” People see me tify the most important tasks, and
as confident, ambitious, compe- trust my team by delegating.
tent, and energetic. But I always Accepting my weakness also
feared seeming imperfect in the helped me find a path to more
eyes of others. I worked as hard as authentic leadership. I previously
I could to make up for my flaws. put my own and others’ feelings
But after becoming a new mom in a box, thinking that discussing
and starting a new job, I was un- them would distract from our pro-
able to excel no matter how hard ductivity, and instead focused on
I worked. The job required me to data, timelines, and deliverables.
attend meetings with almost no But after my own crisis, I began
break between 7 a.m. and 5 p.m., to pay more attention to my team
pushing my own work tasks late
into the night. I used a breast
“I shared my struggles members’ emotional well-being.
I shared my struggles as a new
pump under the table during
meetings and frequently forgot to
as a new mom and … my team mom and my fear of not finding
the best direction for the team.
eat. Mental and physical exhaus- members opened up.” In response, my team members
tion from back-to-back meetings opened up to me about the chal-
and lack of sleep made it difficult to think deeply and cre- lenges they were facing. These conversations helped build
atively about science. I wanted to offer useful comments trust, loyalty, and team morale.
in meetings, but my thoughts often became muddled, at I also became less judgmental when I had to give critical
times leaving me tongue-tied midsentence. I became so feedback to team members. Previously, I saw it as a persua-
anxious about my long to-do list that I could not calm sion contest to convince them to stop doing things their
down to tackle a single task. When a team member left for way and adopt the “right way,” and I dreaded doing it. But
a new job, I blamed myself. On top of it all, I developed now, I first seek to understand the motivation behind their
postpartum depression but was too ashamed to tell my behavior. This enables me to deliver feedback with the aim
doctor. It was the lowest point of my life, and I could no of helping each individual become their best self.
longer deny my weakness. Now, I am grateful for my weaknesses, as they make me
After the 3 a.m. epiphany, I wrote the Bible verse on humbler. They taught me that true strength, in life and in
ILLUSTRATION: ROBERT NEUBECKER

a sticky note and put it on the corner of my computer leadership, does not rely on authority and power, but on
as a reminder. I read it to myself as I transitioned from compassion, honesty, and kindness. j
one meeting to the next, and it began to transform my
approach to work. I realized that instead of focusing on Sophia X. Pfister is the director of research science at Varian, a Siemens
trying to make “clever” comments in meetings—and feel- Healthineers company in Palo Alto, California. Send your career story to
ing stressed that I couldn’t come up with any—I could [email protected].

CALL FOR PA P ERS
Space: Science & Technology is an online-only, Open Access journal published in affiliation with Be ijing Institute
of Technology (BIT) and distributed by the American Association for the Advancement of Science (AAAS).
BIT cooperates with China Academy of Space Technology (CAST) in managing the journal. The journal promotes
the exploration and research of space worldwide, to lead the rapid integration and technological
breakthroughs of interdisciplinary sciences in the space field, and to build a high-level academic platform
for discussion, cooperation, technological progress and information dissemination among professional
researchers, engineers, scientists and scholars.
Submit your research to Space: Science & Technology today!

Learn more at spj.sciencemag.org/space
The Science Partner Journal (SPJ) program was established by the American Association for the Advancement of Science (AAAS), the
nonprofit p ublisher o f t he Science f amily o f j ournals. T he S PJ p rogram f eatures h igh-quality, o nline-only, O pen A ccess p ublications p roduced
in collaboration with international research institutions, foundations, funders and societies. Through these collaborations, AAAS furthers its
mission to communicate science broadly and for the benefit o f a ll p eople b y p roviding t op-tier i nternational r esearch o rganizations w ith t he
technology, visibility, and publishing expertise that AAAS is uniquely positioned to offer as the world’s largest general science membership society.
Visit us at spj.sciencemag.org
@SpaceSciTech @SPJournals
A R T I C L E P R O C E S S I N G C H A R G E S WA I V E D U N T I L J U LY 2 0 2 3
0722Product.indd 443 7/14/22 7:45 AM

science.org/journal/sciimmunol
READY TO
PUT THE
SPOTLIGHT
ON YOUR
RESEARCH?
Submit your research:
cts.ScienceMag.org
Twitter: @SciImmunology
Facebook: @ScienceImmunology
0722Product.indd 444 7/14/22 7:45 AM

Science - 22 July 2022

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Science - 22 July 2022

Uploaded by

Copyright:

Available Formats

Misconduct suspicions cloud Lasing from time A screening approach to improve

Scan the code and start exploring

Brought to you by the Science/AAAS Custom Publishing Office.

Posters Podcasts Sponsored Advertorials Webinars

0722Product.indd 346 7/14/22 7:45 AM

366 In the glare of the Sun

368 To make a mirrorless laser

369 Improving catalysis by moving water

364 370 The quest for more food

354 Consortium seeks to expand

Fast-spreading subvariants are A continuation of the current 23-year-long

SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 347

442 Working Life

ILLUSTRATOIN: IAN HAYDON, INSTITUTE FOR PROTEIN DESIGN

348 22 JULY 2022 • VOL 377 ISSUE 6604 science.org SCIENCE

Confronting 21st-century monkeypox

Published online 19 July 2022; 10.1126/science.add9651

SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 349

Pandemic contributes to big drop in childhood vaccinations

PHOTO: SUNIL GHOSH/HINDUSTAN TIMES/GETTY IMAGES

350 22 JULY 2022 • VOL 377 ISSUE 6604 science.org SCIENCE

Native Hawaiians gain voice in managing Mauna Kea

SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 351

A nurse prepares a COVID-19 vaccine in Guwahati,

winding down. “Those systems are being

path remains unpredictable supposed to have disappeared from the

352 22 JULY 2022 • VOL 377 ISSUE 6604 science.org SCIENCE

SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 353

CREDITS: (PHOTO) CHINATOPIX/AP IMAGES; (GRAPHIC) V. ALTOUNIAN/SCIENCE

354 22 JULY 2022 • VOL 377 ISSUE 6604 science.org SCIENCE

Only 19,370 human genes are known EUROPE

SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 355

and targets to be met,” he says. “If you are INVASIVE SPECIES

PHOTO: MACROSCOPIC SOLUTIONS/SCIENCE SOURCE

356 22 JULY 2022 • VOL 377 ISSUE 6604 science.org SCIENCE

you’re [still] getting these high numbers …

Half of Americans anticipate a willingness to engage in violence—homicides

SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 357

358 22 JULY 2022 • VOL 377 ISSUE 6604 science.org SCIENCE

SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 359

360 22 JULY 2022 • VOL 377 ISSUE 6604 science.org SCIENCE

Possibly duplicated bands

SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 361

362 22 JULY 2022 • VOL 377 ISSUE 6604 science.org SCIENCE

SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 363

PHOTO: STEPHEN BELCHER/MINDEN PICTURES

364 22 JULY 2022 • VOL 377 ISSUE 6604 science.org SCIENCE

concern for deleterious mutations, versity or prevent the fixation of del-

allowing a small population an op- eterious alleles by facilitating gene

The vaquita is a small porpoise 2 Although applications of gene editing

in gillnets for shrimp and finfish anthropogenic and natural changes

SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 365

366 22 JULY 2022 • VOL 377 ISSUE 6604 science.org SCIENCE

SCIENCE science.org 22 JULY 2022 • VOL 377 ISSUE 6604 367

LASERS tal, Lyubarov et al. provide detailed classical

To make a mirrorless laser side such a crystal. When stimulated with a

368 22 JULY 2022 • VOL 377 ISSUE 6604 science.org SCIENCE

Improving catalysis by moving water

22 JULY 2022 • VOL 377 ISSUE 6604 369

such as chemically modifying the catalyst PLANT BIOLOGY

Maximum photosynthetic rate (µmol CO2 m−2s−1)

potential of this process to produce valuable

REFERENCES AND NOTES 50