23 FEBRUARY 2024, VOL 383 ISSUE 6685 Science

Citation preview

EDI TO R I A L

Research needs for loss and damage

science.org SCIENCE

,

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

y

798

y g

10.1126/science.adn4697

y

“…the fund’s design…represents a breakthrough of collaboration…”

g

also scarce on linkages between adaptation and L&D, such as the need to incorporate adaptation strategies when recovering from incurred damages. Further work should address L&D as part of climate-resilient development—a framework that currently only considers mitigation and adaptation options to support sustainable development. Addressing L&D is a critical component of climate action and should be overtly included in climate-resilient development, particularly in the developing world. Comprehensive methodologies are needed to assess the costs of L&D, both already incurred and projected as average global temperatures continue to rise. These inputs will help to clarify the capitalization needs of the L&D fund and inform ongoing negotiations on the broader new collective quantified goal on climate finance by demonstrating evidencebased approaches to determining the scale of finance needed. Scientific inputs are needed to support the Santiago Network in facilitating access to knowledge, resources, and technical assistance to comprehensively address climate risks. This includes assessing current and future L&D as well as comprehensive risk management approaches at regional, national, and local levels. Reports from technical assistance already provided show that there is a need for scientific input to help address data gaps and methodological challenges, including in the context of isolating climate change as a driver of human mobility. Scientific inputs relevant for L&D must consider attribution science, which assesses the causes of observed changes in the climate system. Regional disparities in data and research infrastructure must be addressed to allow for more equitable attribution research and expertise in developing countries. Until these challenges are addressed, the role of attribution science is likely to continue to be an area of contention. Throughout COP28, findings from the Intergovernmental Panel on Climate Change (IPCC), including on L&D, were frequently referenced, including in text on L&D in the GST. COP28 also invited the IPCC to examine how to best align its work with the next rounds of the GST. The IPCC considered this at its latest meeting in January, highlighting the importance of the scientific community in informing L&D policy and climate action. –Adelle Thomas

p

Adelle Thomas is a senior scientist at Climate Analytics, New York, NY, USA. adelle.thomas@ climateanalytics.org

T

he 28th Conference of the Parties (COP28) of the United Nations (UN) climate negotiations concluded in December against a backdrop of shattered climate records, extreme weather, and climate events in 2023 that resulted in substantial loss of life, impacts on health and wellbeing, climate-related mobility, food and water insecurity, and other socioeconomic effects worldwide. In this context, the importance of addressing loss and damage (L&D)—the negative impacts of climate change that persist despite mitigation and adaptation efforts—was paramount. With countries having agreed at the opening plenary to operationalize the L&D fund that was established in 2022 at COP27, inputs from the scientific community will be essential as the global community now seeks to bring the fund online. Pledges to the L&D fund amounted to just over US$700 million in the days following the agreement. Although the fund’s design deviates from the initial visions of developing nations, it represents a breakthrough of collaboration in the contentious and decades-long negotiations on L&D. In the end, contributions were agreed to be voluntary from both developed and developing countries, and the scale of finance needed was not included in the final decision text. The fund will provide support for responding to L&D both economic (e.g., rebuilding infrastructure) and noneconomic (e.g., addressing impacts to human health), including funding that is complementary to humanitarian actions taken immediately after an extreme weather event; funding for intermediate or long-term recovery, reconstruction, or rehabilitation; and funding for actions that address slow-onset events (e.g., sea level rise). Other aspects of L&D were also part of COP28. There was agreement on the UN hosting the Santiago Network, created at COP25 in 2019 to provide technical assistance for developing countries to respond to L&D. Parties endorsed recommendations from the Executive Committee of the Warsaw International Mechanism for Loss and Damage and the development of technical guides on relevant topics. L&D was included in the final text of the first Global Stocktake (GST)—a process that takes place every 5 years to assess Parties’ progress toward meeting the goals of the Paris Agreement. To move forward, research is needed on successes and challenges in responding to L&D. Research is

Passion is not misconduct

U

something bigger than himself. Kennedy “gave me the philosophical outlook to persevere,” he said. Mann lamented, “I would hate to think that young scientists would have to make a choice to withstand attack if they want to study politically contentious topics.” He also sees assaults on science as advancing the conservatives’ agenda. “They want to make doing science in these areas so toxic that young people won’t want to go into it,” he said. The hockey stick graph stands as published because the data were peer-reviewed and addressed questions from experts. Accusations of misconduct were addressed by running established processes outlined by federal agencies, journals, and universities. The paper has not been retracted, and calling it fraud thus met the standards for defamation. A lesson of the verdict is that despite its flaws, the process for vetting, publishing, and curating scientific findings in the literature is still the best game in town. Although some free speech advocates warn that the verdict will have a chilling effect on the criticism of scientific findings, perhaps the verdict can be viewed more optimistically as appropriately directing matters of opinion to blogs and opinion columns while matters of scientific disagreement are handled in the literature of scientific record. Another lesson is that although some may prefer scientists to be more restrained and circumspect than Mann, expressing passionate views does not justify false accusations from critics. Opinions are protected speech, and both scientists and their critics should be welcome to express them. There is a debate in the scientific community about whether scientists undermine their credibility by being outspoken. Suppressing one’s humanity harms one’s credibility even more. What’s important is that scientists are dispassionate in their research publications, not on social media or in opinion pages. Kwame Anthony Appiah, an advocate for the judicious usefulness of appearing neutral, told me that he views scientists as having “offstage” and “onstage” personas, and that the “onstage” version is present in the literature and at conferences. The Mann verdict validates Appiah’s analysis. Human passion varies over a range. But scientific consensus is reached rationally and through established processes. –H. Holden Thorp

H. Holden Thorp Editor-in-Chief, Science journals. [email protected]

y y g y ,

PHOTO: CAMERON DAVIDSON

g

“…vetting… scientific findings in the literature is still the best game in town.”

Published online 13 February 2024; 10.1126/science.ado6275

SCIENCE science.org

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

p

niversity of Pennsylvania climate scientist Michael Mann was awarded more than $1 million in a lawsuit against bloggers who accused him of scientific misconduct in inflammatory terms, likening his treatment of data to what a noted child molester did to children. The verdict suggests that there are limits to which scientists working on politically sensitive topics can be falsely attacked. But the case also says something profound about the difference between matters of opinion and scientific interpretations that can be worked out through normal academic processes. Although Mann has expressed strong—and even intemperate—emotions and words in political discourse, the finding of the District of Columbia Superior Court boiled down to the fact that it is not an opinion that determines when scientific misconduct occurs but rather, misconduct can be established using known processes. When Mann published his famous “hockey stick paper” in Nature in 1998, he began receiving attacks because the paper reported a sharp and alarming increase in global temperature after the Industrial Revolution. Numerous reviews of the paper’s findings were subsequently conducted by the National Science Foundation, Pennsylvania State University (where Mann was at the time), and the National Academy of Sciences. Although a correction was added to the paper to clarify a data matter, the evaluations all concluded that the paper stood. The jury therefore found that accusing Mann of fraud without sufficient evidence was defamatory—the bloggers’ allegation that Mann committed scientific misconduct was false. “It’s perfectly legitimate to criticize scientific findings,” said legal scholar Michael Gerrard, “but this verdict is a strong signal that individual scientists shouldn’t be accused of serious misconduct without strong evidence.” Further, the jury concluded that the bloggers’ statements were written with “maliciousness, spite, ill will, vengeance or deliberate intent to harm.” Mann told me that he’s aware that he is passionately outspoken but said the hockey stick paper evoked relentless berating and condemnation. It’s perfectly human to react when attacked. Mann noted that former Editor-in-Chief of Science Donald Kennedy—who wrote fiery editorials about climate change—helped him to understand that the denunciations were about

799

NEWS IN BRIEF

“

Getting research security right is more complex and has taken much longer than I would have liked.

”

White House science adviser Arati Prabhakar explaining at a congressional hearing why her office is still working after 3 years to write tougher standards for federally funded scientists.

Edited by Jeffrey Brainard

Cell therapy fights solid tumors | Although cell therapies that help the immune system destroy tumors have revolutionized treatment for blood cancers, success in solid tumors such as those of the brain, colon, and pancreas has lagged, in part because it’s hard to kill these malignant cells without also harming healthy tissue. But last week, the U.S. Food and Drug Administration (FDA) approved the first cell therapy for a solid tumor after a clinical trial showed that 31% of patients with an advanced case of melanoma, an aggressive skin cancer, were helped. The treatment, called Amtagvi and made by the company Iovance Biotherapeutics, harvests immune cells called T cells from a patient’s tumor, expands their numbers in the lab, and reinfuses them into the patient. Because these cells, called tumor-infiltrating lymphocytes, are already primed to attack the patient’s cancer, they don’t appear to harm healthy cells, although the treatment regimen has toxic side effects. FDA fasttracked its approval, and Iovance says it will now conduct a larger trial to confirm that it’s safe and effective. BIOMEDICINE

p g y

U.N. to mull satellite swarms HUMAN RIGHTS

800

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

Deportation threat raises stress | The threat of increased deportations steadily increased the psychological distress of Latinos in the United S O C I O L O GY

science.org SCIENCE

,

A

genetics journal has retracted 18 papers from China over concerns that study participants, many of them from ethnic minorities, were not fully informed about the study objectives and unable to choose freely whether to participate. The papers, some of them co-authored by scientists affiliated with China’s security services, relied in part on genetic samples collected from members of the Uyghur, Tibetan, Tujia, and Yi minorities, all of whom are considered vulnerable to state oppression. The authors, all but a few based at Chinese institutions, disagreed with the retractions or did not respond to queries from the journal, Molecular Genetics & Genomic Medicine. The mass retraction is a victory for bioinformatician Yves Moreau of KU Leuven, who notified the journal of the problems in 2021 and whose work had previously led to 11 similar retractions by other journals. Moreau says about 70 papers are still under investigation.

y

Ethical issues sink China genetics papers

| The United Nations agreed last week to investigate the impact of teeming satellite fleets on astronomy. Since SpaceX launched its first batch of Starlink satellites in 2019, astronomers have campaigned for regulation, and the International Astronomical Union and other organizations hope the U.N.’s involvement will lead to international standards and norms that member nations would enforce. More than 5000 Starlinks are already in orbit; they and other satellites leave bright trails in astronomical images, and their communications downlinks can blind radio telescopes. The scientific and technical subcommittee of the U.N.’s Committee on the Peaceful Uses of Outer Space plans to examine the topic for the next 5 years. ASTRONOMY

y g

A Uyghur man walks past a closed mosque in Kashgar, in China’s northwestern Xinjiang region.

CONSERVATION

U.S. protects iconic marine snail p

needs protection because populations are sparse, scattered, and heavily overfished. NOAA’s listing decision, issued on 14 February, has no immediate impact but authorizes the agency to ban imports of conch meat, a prospect that many researchers hope will spur other nations to tighten fishing regulations. According to NOAA, many Caribbean nations lack the regulations or resources for enforcement.

g

T

he queen conch, a large marine snail known for its showy shell and delectable flesh, has been added to the U.S. government’s list of species threatened with extinction. Scientists at the National Oceanic and Atmospheric Administration (NOAA) said the species (Strombus gigas, pictured), which is native to the Caribbean region and grows up to 35 centimeters long,

y

SCIENCE science.org

PHOTO: SHANE GROSS/NPL/MINDEN PICTURES

| An international nonprofit launched last week plans to develop technological and policy guardrails to reduce the risk that terrorists create weapons using modern biotechnology. The International Biosecurity and Biosafety Initiative for Science (IBBIS) was spearheaded by the Nuclear Threat Initiative, a nonpartisan think tank for arms control. IBBIS will focus on helping researchers, companies, and government agencies manage tools and information that might be used accidentally or deliberately to create deadly toxins and pathogens. The

BIOSECURITY

Ancient Down syndrome glimpsed | Drawing on a DNA database of nearly 10,000 ancient individuals buried around the world, geneticists have identified six infants with Down syndrome, or trisomy 21, in which people have an extra 21st chromosome. A seventh infant had trisomy 18, or Edwards syndrome, the first time that condition has been identified in the archaeological record. Reporting this week in Nature Communications, the research team said details of the seven individuals’ graves— which date from the late Stone Age to medieval times—suggest they were buried with respect. In one site from Iron Age Spain, a girl with Down syndrome was buried with offerings including shell beads and goats. In this first large-scale study of ancient Down syndrome, the ancient DNA ARCHAEOGENETICS

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

801

,

| A small clinical trial has shown that the blockbuster glucagon-like peptide-1 (GLP-1) drugs, already proven against obesity and type 2 diabetes, may reduce cravings in opiate addiction. At the annual meeting of AAAS (publisher of Science) in Denver last week, neuroscientist Patricia Grigson of Pennsylvania State University reported unpublished data from 20 people undergoing treatment at A D D I CT I O N

New group targets bioterror tools

group’s first project is to distribute free software designed to help companies that synthesize DNA screen their orders for potentially dangerous sequences. IBBIS hopes its software will allay concerns from some potential users wary about existing screening tools funded by the U.S. government. The group also expects to help screen reports, manuscripts, and artificial intelligence products to scrub methods for making bioterror agents.

y

Obesity drug cuts opiate craving

a residential rehabilitation center. All had recently experienced opiate withdrawal, and many were on buprenorphine, an existing treatment. Ten received liraglutide, an early form of the drugs that mimic GLP-1; ten other volunteers got a placebo. Over 3 weeks, self-reported craving was 40% lower in people getting liraglutide compared with the placebo, and many of those treated reported no cravings at all. The trial was limited by its small size and because some participants dropped out because of gastric distress, a side effect of GLP-1 drugs. Grigson plans to follow up with a larger trial using semaglutide, a newer form of the drug.

y g

States during the past decade, a large, nationally representative study has found. More than 770,000 adults took the same health survey annually between 2011 and 2018, a period when many federal and state officials sought to increase deportations, which have disproportionately affected Latino people. By analyzing clues to anxiety and depression in the survey responses, a research team concluded that the threat of deportation increased the psychological distress of Latinos by 10% during the 8 years of the study period. Although U.S.-born Latinos are not at risk of being deported, the study found their distress showed the largest increase, most likely because of fear of discrimination or that a loved one would be deported, the researchers suggest this week in the Proceedings of the National Academy of Sciences.

NE WS | I N B R I E F

provided conclusive evidence of the condition, which is difficult to identify based on skeletal indicators alone.

NASA to launch UV telescope | NASA last week approved the building of a new space observatory, the Ultraviolet Explorer (UVEX), to provide some of the capabilities of the aging Hubble Space Telescope. UV light is key to understanding hot young and old stars, supernovae, starburst galaxies, and the gas between them. But UV, like x-rays and most of the infrared spectrum, is blocked by Earth’s atmosphere and can be viewed only from space. At a cost of $300 million plus launch, NASA expects to send UVEX ASTRONOMY

into orbit in 2030. The agency currently relies heavily for UV observations on the 33-year-old Hubble telescope, which is long past its design life. UVEX will conduct a 2-year, all-sky survey but can also zoom in on fleeting events such as merging neutron stars, which emit light across the spectrum along with gravitational waves.

FDA OKs frostbite treatment | The U.S. Food and Drug Administration last week approved an injection that treats severe frostbite in adults, lowering the risk of finger or toe amputation. The drug, iloprost, is marketed by Eicos Sciences as Aurlumyn. Physicians have struggled to effectively DRUG DEVELOPMENT

| Carbon emissions from secondhand trucks and buses in developing countries have risen more than 30% since 2000, according to a new report from the U.N. Environment Programme. The analysis finds that exports of old, heavyduty vehicles (HDVs) worsen other kinds of air pollution in importing countries and make roads more dangerous. In the first attempt to quantify the trade and its impacts, researchers estimated that used HDVs from the European Union, Japan, and South Korea contribute more than half of the exports. (Data from North America and China were unavailable.) The report recommends exporting countries establish safety and pollution standards and importing nations tighten such rules. C L I M AT E P O L I C Y

science.org SCIENCE

,

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

Used trucks pollute, U.N. warns y

802

y g

D

uke University announced last week it will close its widely used, 100-year-old herbarium, sparking protests from scientists at the university and elsewhere. Home to 825,000 specimens, the herbarium hosts one of the largest collections of algae, fungi, lichens, and mosses in the United States, with a focus on the Southeast, the nation’s biodiversity hot spot. It is also considered a leader in making images and associated data about its specimens available in online databases. But the herbarium’s facilities are outdated and keeping it open would require at least $25 million, Susan Alberts, Duke’s dean of natural sciences, wrote in an email to Science. The decision is “a tragic mistake,” says Michael Donoghue, an evolutionary biologist at Yale University. Other scientists say the holdings are of growing importance for studying challenges such as invasive species, emerging infectious diseases, and climate change’s effects on biodiversity. Many small herbaria and a few large ones have closed during the past 30 years. A 2020 report by the National Academies of Sciences, Engineering, and Medicine called for developing long-term financial sustainability for these collections.

y

Herbarium closure draws protests

| Scientists have linked DNA evidence collected from a cemetery of enslaved people at a former iron smelting plant in Maryland to a modern-day descendant, in one of the first such findings. Last year, Agnes Jackson, 86, who is African American, took a DNA test to learn more about her family history, which like many Black people’s is difficult to trace beyond the mid–19th century because their ancestors were enslaved. A researcher team compared her sample and others submitted to the 23andMe commercial ancestry testing service with DNA collected from skeletal remains in the Catoctin Furnace cemetery, about 50 kilometers north of Washington, D.C. Last week, company geneticist Éadaoin Harney told Jackson and her daughters during a ceremony at the cemetery that the analysis shows they are related to a girl who died at age 2 or 3. The company revealed the link because Jackson had given her permission ahead of the analysis and had a historical connection to the furnace—her great-great-grandfather was an enslaved ironworker there. G E N E A L O GY

g

BIODIVERSITY

DNA connects identifies ancestor

p

Kathleen Pryer, director of Duke University’s herbarium, holding a fern named after singer Lady Gaga.

treat severe frostbite, which stops blood flow in the skin and underlying tissue of the extremities. Standard therapy has included blood thinners not approved to treat frostbite. Iloprost, originally approved in 2004 to treat pulmonary arterial hypertension, acts by dilating blood vessels and preventing clotting. In a clinical trial published in 2011, none of 16 participants with severe frostbite who received only iloprost had a bone-scan result indicating they would need amputation, versus nine of 14 who received no iloprost.

IN DEP TH

A new company called Safer Human Medicine intends to build multiple monkey enclosures on an 80-hectare property in rural Georgia.

p

ANIMAL RESEARCH

Giant monkey facility could ease U.S. shortage

B

A facility proposed by Safer Human Medicine could house far more monkeys than other U.S.-based commercial dealers have in their inventories, according to 2023 federal inspection reports.

0

10

20

30

Safer Human Medicine Inotiv Worldwide Primates Inc. Alpha Genesis Inc. Charles River Laboratories Primate Products Inc.

The proposed facility will house up to 30,000 monkeys.

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

803

,

Number of monkeys (thousands)

y

SCIENCE science.org

Monkey business

search monkeys used annually in the U.S., stopped exporting them in 2020 because of the COVID-19 pandemic. China hasn’t resumed exports; instead it is breeding monkeys for its own researchers and competing with the U.S. to buy monkeys from other countries, according to the report. Since then, “The price of monkeys has gone through the roof,” says JoAnne Flynn, a microbiologist at the University of Pittsburgh who uses macaques to study tuberculosis. She says she budgeted $7000 per animal in her National Institutes of Health (NIH) grants a few years ago, but they now cost about $20,000. In its report, NASEM recommended that the U.S. reduce its dependence on foreign monkeys in part by expanding domestic breeding programs. Many researchers obtain animals from seven National Primate Research Centers (NPRCs) operated in partnership with NIH, along with a number of private companies (see chart, left). In its 2024 budget request, NIH asked for $30 million to either expand NPRCs or construct new facilities. It’s not yet clear whether this will be approved in the final federal budget. In a statement, NIH said the money would be used to improve existing facilities, some of which have existed for more than 60 years, for example by upgrading housing and scientific equipment. The new funds could support expansions, but the agency says it isn’t currently planning to construct new facilities.

y g

ainbridge, a rural town in southwestern Georgia with a population of 14,000, could soon become home to 30,000 additional residents: cynomolgus macaques. A new company called Safer Human Medicine (SHM) has announced plans to build an 80-hectare facility that would sell monkeys to universities, contract research organizations, and pharmaceutical companies that perform research on the animals. The breeding facility would dwarf others in the United States and could ease a serious shortage of monkeys for research. But it faces pushback from the local community and concerns from animal welfare groups about the company’s leadership and its goals. SHM, which was founded in 2023 by former executives from major research animal suppliers, first announced its plans at a Bainbridge community meeting in December 2023. The company hopes to break ground this summer and acquire its first animals by the end of the year, although it will be many years before the facility could reach its full capacity. Unlike most other research animal suppliers, SHM doesn’t plan to do any research of its own and will only breed animals. “We had the idea that it would be a really good thing for research if someone could develop a domestic capability mak-

ing primates available to U.S. researchers,” SHM President David Johst told Science. “We want to be those folks.” The need is real. A 2023 report from the National Academies of Sciences, Engineering, and Medicine (NASEM) found that U.S. researchers had been forced to change or abandon projects because of difficulty obtaining monkeys. China, which supplied more than half of the 70,000 re-

y

CREDITS: (ILLUSTRATION) SAFER HUMAN MEDICINE; (GRAPHIC) M. HERSHER/SCIENCE; (DATA) USDA ANIMAL & PLANT HEALTH INSPECTION SERVICE

By Sara Reardon

g

Breeder would be the country’s largest; locals and animal welfare advocates are concerned

NE WS | I N D E P T H

Brain and immune differences found in small patient group By Jennifer Couzin-Frankel

T

y ,

science.org SCIENCE

y g

he deepest dive yet into myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) offers a complex view of this mysterious condition. ME/CFS produces crushing exhaustion, particularly after even mild exertion. Many patients struggle to find care, or doctors who believe their symptoms aren’t imagined. The new work, published this week in Nature Communications, affirms that ME/CFS is unquestionably biologically rooted, says Avindra Nath, clinical director of the U.S. National Institute of Neurological Disorders and Stroke, who led the study. It revealed brain activity differences, along with immune and other abnormalities, in 17 people with ME/ CFS compared with 21 healthy controls. The paper “does bring together the complexity” of the condition, says Jo Cambridge, an immunologist at University College London (UCL) who has found potentially related signs of immune dysfunction in ME/CFS patients. “It highlights all the things you’ve got to take into account when you’re investigating this disease.” But the time it took to collect these threads of evidence speaks to the challenge of studying ME/CFS systematically. Begun in 2016, the study included more than 75 scientists and cost more than $8 million. Advocates hope next steps will move more quickly. “This has to be the beginning of building a bigger research program,” says Brian Vastag, a former science journalist disabled by ME/CFS, who lobbied for the study to launch and was enrolled in it himself. Recruitment was an early challenge. Because infections seem to trigger the condition, Nath and his team only enrolled patients whose symptoms surfaced after a viral or bacterial illness. And they excluded anyone with another medical condition that might confound the results. They interviewed

y

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

Chronic fatigue study shows syndrome’s complexity

g

804

BIOMEDICINE

p

shuttered facility in Virginia. Envigo settled the case in July 2022 and surrendered 4000 beagles to the Humane Society. Inotiv, which bought Envigo in 2021, is also under federal investigation over a smuggling scheme in which one of its suppliers allegedly illegally imported wild monkeys from Cambodia while claiming they were bred in captivity. And SHM COO Kurt Derfler worked at animal supplier Charles River Laboratories until 2023, leaving just after the U.S. government subpoenaed the company over potential involvement in the Cambodian smuggling scheme. In a statement, Charles River said it is cooperating with the investigation and any concerns about its conduct will be found to be “without merit.” Inotiv did not respond to a request for comment. Johst, who was formerly general counsel at Charles River, says SHM is a different company. “Before we are judged guilty, I want to have the opportunity for us to show that we can run a high-quality animal facility,” he says. “We are going to pace our growth in a way that makes sure that we keep animal welfare at the forefront.” SHM has also faced pushback from Bainbridge residents, who have expressed concerns about the smell, waste, and potential for animals escaping. In an open letter sent to the community in January, SHM described the “doublebarrier” that will enclose the pens and other measures to address these concerns. Yet after a commuSafer Human Medicine will breed cynomolgus macaques, which nity meeting last month organized are primarily used by pharmaceutical companies for research by People for the Ethical Treatment and drug testing. of Animals (PETA), the local county’s board of commissioners “voted to disavow the Safer Human Medicine projBut animal welfare groups and some researchers want to see primate use reduced, ect,” according to a 14 February statement. not sustained. “The proposal to have this That could include rescinding tax breaks many monkeys bred by anybody is alarming,” that the county had offered SHM, although says Kathleen Conlee, vice president for aniJohst says he is confident that the deal the mal research issues at the Humane Society of company struck with the city and county the United States. “We shouldn’t be investing will hold up to any legal challenges. in this way.” She fears that making this many Kenneth Ramos, a medical geneticist monkeys available would fuel the expansion at Texas A&M University and lead author of primate research instead of encouraging of the NASEM report, says SHM’s facility researchers to look for alternatives. could help with the primate shortage but Conlee also worries about SHM’s leadthat he is waiting to hear more about the ership, which consists of former execucompany’s time frame and other plans. tives from breeding companies that have “Well-coordinated breeding centers could recently faced legal troubles related to be a step in the right direction,” he says. animal welfare. SHM’s CEO, Jim Harkness, Flynn certainly hopes so. The monkey was serving as chief operating officer shortage, she says, is “just making life very (COO) at animal supplier Envigo (now difficult for those of us doing research.” j Inotiv) when the U.S. Department of Agriculture cited the company for misSara Reardon is a science journalist based in treating beagles bred for research at a nowBozeman, Montana. With its $396 million facility, SHM hopes to make a further dent in the shortage. “We intend to source very good quality animals, get them into the U.S., and then start to develop some colonies,” Johst says. He adds that SHM plans to import monkeys from several Asian countries but would not specify which. For now, he says, they will only breed cynomolgus macaques, which are in particularly short supply and are typically used by pharmaceutical companies. (Academic researchers tend to use rhesus macaques, which are primarily bred in the U.S.) The facility’s startup costs are backed by finance firms, which Johst declines to name. SHM is “agnostic” about which companies or institutions it will sell to, Johst says, but it hopes to help small biotech companies that struggle to afford primates. “We may fill a nice niche there,” he says.

By Paul Voosen

R

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

805

,

oughly 1.5 million years ago, Earth went through a radical climatic shift. The planet had already been slipping in and out of ice ages every 40,000 years, provoked by wobbles in its orbit. But then, something flipped. The ice ages began to grow stronger and longer, with durations of 100,000 years, and overall, the planet grew cooler. And nothing about Earth’s orbit could explain it. The cause of this Mid-Pleistocene Transition (MPT), as it’s known, has been a major mystery for decades. A new compilation of global temperatures covering the past 4.5 million years, published on p. 884, points a finger at a familiar molecule: carbon dioxide. It suggests that a strengthening of an ocean pump in the waters around Antarctica sucked carbon dioxide out of the air and sent it plunging to the abyss, cooling the planet and intensifying the ice ages. The study even suggests the climate, then and now, could be more sensitive to carbon dioxide than modelers expect. “The power of the [carbon dioxide] control knob on the climate system really comes out of this work,” says Ruth Mottram, a climate scientist at the Danish Meteorological Institute. This view of the MPT also represents a blow for the other leading theory for the transition: that the ice sheets were themselves responsible. According to this “regolith hypothesis,” once the glaciers in the Northern Hemisphere scoured away enough soil to form on stable bedrock, they could grow taller before collapsing under their own weight, extending the length of ice ages. By reflecting more light into space and altering weather patterns, these bigger ice sheets could cool the planet more. But Peter Clark, a glaciologist at Oregon State University who pioneered the regolith hypothesis and is co– lead author of the new study, says the work

y

PHOTO: BETH MAZUR

Southern Ocean “pump” sparked deep freezes, new temperature record shows

y g

SCIENCE science.org

Sharp shift in ice age rhythm pinned to carbon dioxide

y

Myalgic encephalomyelitis/chronic fatigue syndrome study participant Brian Vastag (right) with study leader Avindra Nath.

GEOCHEMISTRY

g

ern University, who studies and treats ME/ CFS, says her big takeaway from the study “is that this is a disease of the brain. … The inability to sustain energy was coming from” that organ. But why this signaling persists remains a puzzle. (The researchers couldn’t find evidence of lingering pathogens, but such testing can be challenging.) The team uncovered other biological differences between the two groups, many of which have been reported in prior studies. Among other finds, ME/CFS patients had elevated heart rates and their blood pressure took longer to normalize after exertion, both signs of autonomic nervous system abnormalities. A protein called PD-1 was also elevated in T cells from patients’ cerebrospinal fluid, suggesting these immune cells were trying to fight something off, even if no invader could be detected. Nath and his colleagues suspect the immune dysfunction may lead to the brain differences. Drugs called checkpoint inhibitors, designed to help the immune system fight cancer, might support exhausted T cells, he suggests. He hopes for more insights into ME/CFS from a new wave of studies of Long Covid; the two conditions share symptoms though their biological overlap remains uncertain. Vastag now wonders whether the study he supported years ago was the best approach, given how long it took to complete. Vastag’s partner Beth Mazur also had ME/ CFS for many years, and died by suicide in December 2023. “She ran out of hope that she would ever get better,” he says. “They really need to follow up with whatever findings they have,” he adds. “The patient population is desperate.” j

p

and analyzed medical records for more than 200 people; 27 were invited to the National Institutes of Health (NIH) for a marathon of medical tests. But several of those turned out to have underlying diseases that had been overlooked, including cancer, pulmonary embolism, and a muscle ailment called myositis that causes weakness and fatigue. The surprise diagnoses suggest that people told they have ME/CFS often receive inadequate care— something many patients already know. “It’s a huge indictment of our medical system that they had to join a research study to find out what was wrong with them,” says Mike VanElzakker, a neuroscientist at Massachusetts General Hospital. The participants spent at least a week at NIH for “a megaworkup” that included skin and muscle biopsies, spinal taps, and energy expenditure measurements. “We took every biopsy and body fluid we could think of,” Nath says. Compared with healthy volunteers, people with ME/CFS showed no differences in muscle tissue that could explain postexertional fatigue. Cognitive and psychiatric tests were also similar across the groups. But functional magnetic resonance imaging revealed decreased activity in brain regions that, among other functions, help control movement. The researchers suggest brain signals may flash stop signs to prevent physical activity—similar to how a bout with illness forces rest. “When we have a bad flu, [we] can’t get out of bed,” says Jonathan Edwards, a rheumatologist at UCL. “It’s a central signaling problem” in the brain, he says. “There’s nothing wrong with your muscles.” Nancy Klimas, director of the Institute for Neuro-Immune Medicine at Nova Southeast-

NE WS | I N D E P T H

y g y ,

science.org SCIENCE

y

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

g

806

p

have struggled to disentangle the two eftrends within the new temperature rerepresents the beginning of the end for his fects. “There are a lot of assumptions,” says cord revealed that the planet was cooling idea. “The indications are that the ice sheets Jeremy Shakun, a paleoclimatologist at everywhere, in sync—a signature of carbon had little influence, or no influence, essenBoston College and co–lead author of the dioxide’s influence, which mixes throughtially, on global temperature,” he says. study. “We wanted something more direct.” out the atmosphere. This global influence Instead, Clark and his colleagues now arThe researchers began to compile pubis also seen during the MPT, when temgue that the linchpin of the climate about lished records based on other temperature peratures began to drop faster, eventually 1.5 million years ago was the Southern proxies. One relies on alkenones, a lipid settling to an average of some 4°C below Ocean, and in particular an expanse of sea produced in the ocean only by two closely preindustrial levels. ice extending off Antarctica. Their global related algal species; the number of douThe record also shows up to 3°C more temperature record shows a gradual decline ble carbon bonds in the molecule varies cooling across the 4.5 million years than that steepens at this time. Because the rewith temperature. Another compares the climate models suggest should have hapcord also holds regional trends, it showed ratio of magnesium and calcium in forapened. That’s alarming, because researchers that the Southern Ocean had cooled enough minifera; at higher temperatures, more trying to predict how hot global warming by then for sea ice to form more regularly magnesium is incorporated. And the last will make the planet look to the past to on vaster scales than it has in modern times. simply compares the abundances of difunderstand climate “sensitivity”—how temAs sea ice grows, it sucks freshwater ferent microbial species that appear to be peratures changed for a given injection or out of the ocean, leaving behind saltier, sensitive to temperature changes. withdrawal of carbon dioxide. The new dense water that plunges to the ocean bottemperature recontom, taking dissolved struction suggests clicarbon dioxide along mate models might be with it and cooling the underestimating the planet. In a feedback, influence of the gas, the cooling would at least in their projeclead to more sea ice tions of the past. “If formation, a strengththis data is to be beening of this ocean lieved, the climate syspump—the engine of tem is more sensitive a global conveyor belt than we’ve built into of ocean currents— these simulations,” and an intensification Chalk says. “We need of the ice age. Each ice to know which way to age would ultimately come down on that.” end as traditionally Ongoing efforts to envisioned: with colextend ice core relapsing northern ice cords beyond 1 milsheets that slide into lion years will help the sea. The sudden refine this picture, injection of freshadds Bärbel Hönisch, water into the oceans a geochemist at Cowould disrupt the lumbia University. global conveyor belt. An expansion of Antarctic sea ice 1.5 million years ago may have triggered longer and stronger ice ages. But this new record, Carbon dioxide would with its regional intriescape, the planet Assembling the temperature “stack” cacies, will likely be a foundation for much would warm, and the sea ice would melt. “It’s a solid mechanism for starting off required Shakun to gather 128 of these future work. “It’s a pretty beautiful record,” this MPT cascade,” says Thomas Chalk, published records. “That first little step is she says. Chalk says those research topa geochemist at the European Centre for endless,” he says. “Half the work.” He had ics could include questions that feed into Research and Education in Environmental to standardize each record and line up their climate forecasts, such as how weather Geosciences. Although the study is not the ages. He and his colleagues then used modpatterns could change over different confirst to propose that Antarctic sea ice had a ern climate models to predict past global tinents, or how clouds respond to warming role in the MPT, the temperature record fills temperatures given the ocean temperatures patterns in different oceans. a gap for the field, which chronically underthe proxies recorded. They validated the reRegardless, the new temperature record estimates the importance of the Southern sults against modern climate records and should close the book on an “ice-sheet Ocean, Mottram says. certain well-studied points in time, such as first” view of the Pleistocene, says Timothy To compile the record, the researchers the last ice age and the warm period before Herbert, a paleoclimatologist at Brown Unihad to look beyond the records from cores the Pleistocene. versity. “The dance of global temperature is drilled from polar ice sheets, because they The resulting temperature curves are primarily not called by ice, but more likely only go back about 800,000 years. Oxygen profoundly different from other published by [carbon dioxide].” And the record casts a isotopes in foraminifera, tiny sea creatures records and model results. When the spotlight on the crucial role of the Southern whose fossils are found in sediment cores 40,000-year ice ages first began, 2.7 milOcean, where sea ice is currently vanishing worldwide, hold longer records. But these lion years ago, the long-term cooling trend fast, with implications researchers are only isotopes don’t just capture ocean temperadidn’t budge—a sign that the reflectivity of beginning to puzzle through. “It’s the valve ture; they also reflect the amount of freshthe northern ice sheets wasn’t an imporon deep carbon storage,” he says. And today, water taken up in ice sheets. Researchers tant factor in temperatures. The regional that valve may be closing. j

PUBLISHING

Researchers buy citations to inflate metrics New study reveals “extreme” gaming tactics some could be using to improve their standing By Katie Langin

I

,

807

y

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

y g

That’s when they decided to create the fabricated Google Scholar profile and see whether they could purchase citations themselves. They asked ChatGPT to write 20 research articles on the topic of fake news, embedding many self-citations—references to papers written by the same fictional author—mimicking a practice some researchers use to boost their citation numbers. Then, they posted the articles on various preprint servers. Google Scholar detected those articles in its scan of the academic literature and populated the fictional author profile, listing those preprints as publications and giving the researcher credit for 380 self-citations contained within them. From there, it was relatively easy to purchase additional citations. Using the name of the fictional scientist, the research team contacted the vendor through WhatsApp and purchased the “50 citations package.” Within 40 days, five papers were published that each included 10 citations to the fake news researcher’s work. Four of the five appeared in a single chemistry journal. “This made no sense, as our fictional researcher’s papers were not even remotely related to chemistry,” Rahwan notes. The study suggests some researchers are using tactics similar to those employed by the NYU Abu Dhabi team to boost their citation

y

Naoki Masuda, University at Buffalo

g

SCIENCE science.org

“The game never ends.”

rankings. “The evidence they show in this paper is pretty strong,” says Naoki Masuda, an applied mathematician at the University at Buffalo who has studied anomalous citations. The authors can’t say how widespread these problems are. “We only focused on the outrageous cases,” Rahwan says. But they did see signs that other papers published by the same chemistry journal may have included citations that were purchased: Eleven other (real) scientists had received at least 10 citations from a single paper published in that journal. Bernhard Sabel, a neuropsychologist at Otto von Guericke University Magdeburg who has studied paper mills that sell authorship on scientific papers, says the academic community should be “very concerned” about these kinds of manipulations. “The problem is big, and it has been growing fast in the last 10-15 years,” Sabel adds. Google Scholar and other databases “have the muscles and should take action to clean up this mess of fraud citations,” he says. (Google didn’t respond to a request for comment on the platform’s methods and whether the company is aware of the problems identified in the study.) Byrne notes that the problem of erroneous citations isn’t restricted to Google Scholar, but she was surprised by some of the oddities of the platform identified by the new study. For instance, the research team found that citations persisted on Google Scholar even after the citing article had been removed from a preprint server. “This is clearly a loophole that needs to be closed,” she says. To identify potentially problematic citation patterns, the authors of the study proposed a new metric called the citation concentration index, or c2-index, which reflects the number of papers that cite the researcher at least that many times. Whereas a high h-index can indicate a productive, reputable researcher, a high c2-index would mean that impression should be taken with a grain of salt. In one real-life case, for example, a scientist had a c2-index of 45, meaning 45 papers had cited that scientist 45 times each. “Whatever can be done to reduce the problem is a welcome step,” Sabel says. But others worry that even if such an index is implemented, bad actors will adjust and stay a step ahead. “The fraudsters … can read this and relatively easily invent methods to conceal their behavior,” Masuda says. “So, the game never ends.” j

p

n 2023, a new Google Scholar profile appeared online featuring a researcher no one had ever heard of. Within a few months, the scientist, an expert in fake news, was listed by the scholarly database as their field’s 36th most cited researcher. This unknown researcher had an h-index of 19—meaning they’d published 19 academic articles that had been cited at least 19 times each. It was an impressive burst onto the academic publishing scene. But none of it was legitimate. The researcher and their institution were fictional, created by researchers at New York University (NYU) Abu Dhabi who were probing shady publishing practices. The publications were written by ChatGPT. And the citation numbers were bogus: Some came from the author excessively citing their own “work,” while 50 others had been purchased for $300 from a vendor offering a “citations booster service.” “The capacity to purchase citations in bulk is a new and worrying development,” says Jennifer Byrne, a cancer researcher at the University of Sydney who has studied problematic publications in the biomedical literature. In academia, a researcher’s h-index and the number of citations they’ve garnered are often used for hiring and promotion decisions. And the fabricated profile, which was part of a study posted as a preprint last week on arXiv, demonstrates the “extreme” tactics that can be employed to manipulate them, adds Byrne, who was not involved in the work. (The researchers declined to name the vendor to avoid giving them more business.) The study got started when Yasir Zaki, a computer scientist at NYU Abu Dhabi, and his colleagues noticed troubling patterns among real researchers. After combing through the Google Scholar profiles of more than 1.6 million scientists and looking at authors with at least 10 publications and 200 citations, the team identified 1016 scientists who had secured a 10-fold increase in citations over a single year. “You know something is off when a scientist experiences a sudden and massive spike in their citations,” Zaki says. The team flagged 114 scientists who had received more than 18 citations from a single paper—a suspicious sign, Zaki says, “since it is rare for even established scientists to have more than a handful of citations stemming from the same source.” In one particularly egregious case, 90% of the

references in a single paper cited one scientist’s publications. “It was … published in a journal that the suspicious scientist is an editor of,” Zaki says. Many of the citations associated with the 114 suspicious scientists came from low quality publications, the researchers say— including preprints, which aren’t subject to peer review. Some of the citing publications didn’t even reference the researcher’s work in the main text; the citation had simply been added to the reference list at the end. The team also noticed that one of the authors had received many citations from papers hosted by a single account on Research Gate, a social networking site for scientists. “To our shock, that account was openly advertising a citation-purchasing service!” says study author Talal Rahwan, a computer scientist at NYU Abu Dhabi.

NE WS | I N D E P T H

ASTRONOMY

Stellar remains of 1987 supernova found at last NASA’s JWST telescope spots fluorescing gas inflamed by surviving neutron star By Daniel Clery

y g y ,

science.org SCIENCE

y

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

g

808

p

The breakthrough came when Claes Fransson of Stockholm University and his colleagues took advantage of JWST’s sharp hen a nearby star exploded in eyesight and its ability to split light into a 1987, it created the first supernova spectrum. They knew photons streaming visible to the naked eye in 4 cenaway from a hot neutron star would catch turies and became one of the most up with and ionize certain elements in the intensely studied objects in space. blast cloud. As these atoms regained their Now, after more than 35 years of electrons, they would fluoresce at precise searching, researchers have finally discovoptical and infrared wavelengths. ered the cinder left behind. Using NASA’s new NASA’s Hubble Space Telescope failed to giant space telescope JWST, astronomers spot this signature in optical light. NASA’s spotted glowing gas at the center of the Spitzer Space Telescope saw some infrared blast that can only have been energized by emission but couldn’t pinpoint the source. something hot and compact inside it, they On 16 July 2022, just days after it began obreport on p. 898. They believe a neutron serving, JWST was pointed at SN 1987A. Its star, all that remains of the shattered star, large 6.5-meter mirror gives it is responsible. sharp vision and its spectral For once the cliché is apsensors extend far into the propriate: “This is the smokinfrared, where many of the ing gun,” says astronomer emission lines lurk. “JWST’s Emanuele Greco of the Gifield of view is just perfect for useppe S. Vaiana AstronomiSN 1987A,” Fransson says. cal Observatory of Palermo. The researchers found fluo“It’s a breakthrough in the rescing argon and sulfur in sense of the information we the middle of the supernova will get about such an extreme remnant. The argon “was so and young object.” strong you could hardly miss Most supernovae go off it,” Fransson says. when a large star, at least The team also measured eight times the mass of the the speed of the fluorescing Sun, abruptly runs out of its gases and concluded that the fusion fuel. Without radiamaterial was blasted out at a tion pressure to hold up the slower pace from a layer just star, the core collapses and outside the original star’s the outer layers come crashNASA’s JWST space telescope found fluorescing argon and sulfur gas at the center core—where argon and sulfur ing down. They bounce off the of SN 1987A’s expanding shell of debris, a sure sign of a hot neutron star within. core and, getting a boost from would be expected. Produca flash of tiny neutrino partiing such fluorescence requires cles, blast most of the star’s mass into space. Large Millimeter/submillimeter Array, a high-energy photons and the only viable The core survives, and for the most massive radio observatory in Chile. They showed a source, the team concludes, would be light stars, the result is a black hole. In others blob of warm material in the center of the from a hot neutron star. the collapse forms a solid ball of neutrons, supernova ejecta. But the researchers could The exact nature of the neutron star and packing a couple of solar masses into a citynot say whether it was warmed by the dethis radiation is not yet known, Fransson size object. A teaspoon of the stuff would cay of radioactive elements in the blast or says. It could be directly emitting x-rays weigh 1 billion tons. high-energy light emitted by a smoldering from a surface several million degrees hot. The last supernova in the Milky Way neutron star. “It was suggestive, but not Or the collapsed core could have formed occurred in 1604 and was recorded by conclusive,” she says. a pulsar, a fast-spinning neutron star with Johannes Kepler just a few years before Then, in 2021, Greco and his colleagues intense magnetic fields that can whip up Galileo Galilei first turned a telescope toreported finding archive images from two particles and cause them to emit ultraward the sky. Then, on 23 February 1987, NASA x-ray telescopes that showed x-rays violet light that’s energetic enough to trigcame the explosion of Sanduleak −69 202, from magnetically trapped particles near the ger the fluorescence. a blue supergiant star of 20 solar masses center of the supernova. But they couldn’t As years pass and the dusty supernova some 168,000 light-years from Earth in the tell whether those particles were trapped by remnant disperses, it will be easier for obLarge Magellanic Cloud, a nearby satellite the magnetic field lines of a neutron star or servers to study this newly minted neutron galaxy. Telescopes across the globe swivthose of a shock wave farther out. “We could star. “It’s fresh, right out of the oven,” Greco eled to watch the fireworks. Much of what not identify the central object,” he says. says. “It’s unique.” j

W

astronomers now know about core-collapse supernovae comes from this event, known as SN 1987A. “It has contributed a lot,” says Mikako Matsuura of Cardiff University. Theorists predicted SN 1987A left behind a neutron star, based in part on a 10-secondlong burst of neutrinos detected a few hours before the visible light show; a collapse into a black hole would have produced a shorter burst of the particles. But where was the neutron star? Observers tried to penetrate the cloud of expelled material without much success. Over the past 5 years, some clues emerged. In 2019, Matsuura and colleagues published images of SN 1987A taken with the Atacama

PRECISION MEDICINE

NIH megastudy analyzes first 250,000 genomes All of Us finds new DNA variants and refines genetic risk scores in diverse groups By Jocelyn Kaiser

y ,

GRAPHIC: D. AN-PHAM/SCIENCE

y g

809

y

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

g

SCIENCE science.org

p

of Us data have used less-detailed genome scans of participants to search for common variants that slightly raise disease risks. giant research database with genetic Tallying hundreds or thousands of these and health data of people across the markers to produce “polygenic risk scores” United States has hit its stride. This can identify people at high risk of say, prosweek, the All of Us study published an tate cancer or asthma. One paper, in Nature, analysis of genomes and linked health uses data from All of Us and other biobanks information from nearly 250,000 to firm up evidence that polygenic risk participants whose diversity may be greater scores for type 2 diabetes cluster into groups than that in any comparable biobank. The depending on associated complications— National Institutes of Health (NIH)-funded some people have a signature linked to heart effort, which plans to quadruple that genome disease, whereas others have scores linked number ultimately, has already found hunto kidney disease, suggesting they may need dreds of millions of never-reported genetic different treatments for their diabetes. The variants that have the potential to reveal new other paper, in Nature Medicine, draws on links between genes, health, and disease. All of Us data for 10 diseases to develop poly“This is a hugely important paper because genic risk scores that work for people with of the diversity that’s there and the resources African, Asian, and Hispanic anit’s going to provide for recestry. This “will help level the searchers,” says human geplaying field” so not just white neticist Rick Kittles of the Banking on diversity patients can learn their risks, Morehouse School of MediNearly half of the 526,000 people who are fully enrolled in the All of Us health and says geneticist Alicia Martin of cine, who wasn’t involved genomics study are nonwhite, and about 80% are disabled, older than age 65, sexual the Broad Institute. in the analysis, published or gender minorities, or from other groups routinely left out of biomedical research. Some of the more than in Nature. The effort is al7000 researchers using All of Us ready paying off: Two other Race and ethnicity are looking beyond genomics to papers also published this explore questions about health week showcase how All of Black White Hispanic care disparities. For example, the Us data can help research(18.83%) (54.38%) (15.8%) biobank’s detailed but anonyers develop genetic scores mized survey data on the 10% of for predicting disease risks in Skip or PNA* Other Asian More than one participants who report belongdiverse groups. (1.9%) (1.75%) (3.38%) race/ethnicity (3.98%) ing to sexual and gender minoriLaunched in 2018, All of ties is “an incredibly rich data Us is NIH’s answer to similar *Prefer not to answer source” that will enable studies projects in countries such as U.S. and international researchers, who of these often “unseen groups,” says social scithe United Kingdom, Iceland, and Japan. The gold standard is the 500,000-participant UK can usually get approval within 1 day to entist Stephanie Cook of New York University. Biobank with its high-quality data available access All of Us data using its cloud-based She is using it to explore whether homophobic to any qualified researcher. But 94% of its portal, have already begun to explore fresh discrimination and other stressors can raise participants are of European ancestry, limitlinks between disease and known variants. LGBTQ individuals’ heart disease risk. ing the relevance of findings to other groups. For example, a study last month used All of All of Us leaders note that the project All of Us aims to recruit 1 million people Us data to show that rare variants in a musalso stands apart from other biobanks bewilling to share their DNA, blood samples, cle gene that have been tied to heart disease cause participants can opt to receive inand electronic health records (EHRs). Diin Europeans pose a similar risk to people sights from the program. About half of the versity in many forms is a key goal: Of the of African ancestry. Identifying carriers of 51,000 people who were offered results back 526,000 participants who have completed these genes in the clinic could help them agreed to find out whether they carry risk all enrollment steps, 47% are racial and take steps to reduce their risk. genes for 59 diseases such as breast cancer ethnic minorities and about 80% are people Most of the 275 million new variants are and heart disease, Bick’s team reports. Bick over age 65, or who earn a low income, are rare, and many could prove harmless, which says one future goal is to study “how returndisabled, or are from other groups that “hiswould allow clinicians to rule them out when ing this medical information to participants torically have been left out of large-scale searching for the gene behind a child’s dischanges their health trajectories.” scientific studies,” says Alexander Bick, a ease, for instance. Other variants may cause The project has signed up 767,000 pargeneticist with the All of Us data center at functional changes in a gene’s protein that ticipants as of this week and hopes to reach Vanderbilt University Medical Center. in turn trigger disease and could offer fresh 1 million by the end of 2026. “It’s still early That rich diversity in the project’s first targets for drug development. days,” Martin says. But, “I’m really excited to 245,388 whole genomes revealed more The other new studies that draw on All see what comes from All of Us.” j

A

than 1 billion genetic variants, more than 275 million of which were novel. The findings expands the catalog of variants found by the UK Biobank and other whole genome databases by roughly 30%, says Bick, corresponding author on the Nature paper. All of Us data also reinforce wellestablished links between about 3700 variants and 117 diseases, confirming the quality of the new data. Unlike the UK Biobank, which draws on information from the country’s National Health Service, All of Us collects and folds together data from some 50 different types of EHRs used by various public and private health care systems. “I think we’ve been able to successfully show that we can do this in the United States,” All of Us CEO Josh Denny says.

NE WS

p g y y g

FEATURES

y

A SPIKING FEVER

,

Long neglected, Lassa fever is surging in West Africa. Researchers want to know why

S 810

itting on a bench outside the Irrua Specialist Teaching Hospital (ISTH) in Edo state in southwestern Nigeria in September 2023, Muhammed Luqman Dagana recounted his ordeal earlier in the year with Lassa fever, a deadly hemorrhagic disease of West Africa. At first the 33-yearold wasn’t alarmed—his fever, head-

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

By Leslie Roberts, in Irrua, Nigeria, and Kenema, Sierra Leone; Photography by Apochi Owoicho Reporting for this story was supported by the Pulitzer Center.

ache, body aches, and cough were innocuous enough. A doctor at his local clinic gave him antibiotics for typhoid fever and antimalarial drugs. But his symptoms persisted, so he tried another clinic. Again, the diagnosis was malaria and typhoid. Dagana continued to decline, his fever fluctuating wildly. When a third clinic referred him to the general hospital here, he science.org SCIENCE

N E WS

In September 2023, 13-year-old Ngozu Ofozor was preparing to go home after spending 2 weeks in the Lassa Fever Isolation Ward at the Irrua Specialist Teaching Hospital in Edo state in Nigeria.

y ,

811

y g

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

y

village of Lassa, in Borno state in northeastern Nigeria, were stricken by a mysterious fe-

g

SCIENCE science.org

IN 1969, two missionary nurses in the remote

brile disease. Both died. When a third nurse who cared for the second case, Penny Pinneo, fell ill with a spiking fever, chills, an ulcerated throat, and profound weakness, she was evacuated to New York Presbyterian Hospital. After a devastating and prolonged illness, she slowly recovered. Virologist Jordi Casals-Ariet and his team at the Yale University Arbovirus Research Unit analyzed blood samples from the nurses in search of what they suspected was a new virus. They isolated a single-stranded RNA virus in what became known as the Arenavirus family and named it for the village where it was discovered. But Casals-Ariet became infected while handling the samples and almost died, saved only by a risky transfusion of antibodies from Pinneo’s blood. When Casals-Ariet’s technician, Juan Roman, also became infected and died, Yale immediately halted work with the live virus and transferred it to the “hot” lab at what is now the U.S. Centers for Disease Control and Prevention (CDC) in Atlanta. The United States now classifies the Lassa virus as a biosafety level-4 pathogen, so dangerous that the live virus can only be studied in the highest security laboratories. In 1972, during a Lassa fever outbreak in Eastern Province in Sierra Leone, Thomas Monath, then a virus hunter at CDC, and his colleagues set out to find the animal reservoir. Similar viruses were known to be spread by small vertebrates, so they trapped 641 animals and removed tissue and blood samples. Bats, shrews, monkeys, and a single turtle turned up nothing. But Monath eventually isolated the virus from 14 specimens of multimammate mouse, Mastomys natalensis, so-called because of its long rows of nipples for nursing big litters. Ubiquitous across subSaharan Africa, M. natalensis, often referred to as the common African rat, is a denizen of poor villages, scurrying in and out of houses for food—especially during the dry season when farmers burn the fields after harvest, driving out the mice. The virus couldn’t have picked a better host. Prolific breeders, females bear 10 to 12 pups every 45 days, and an infected female passes the virus to her offspring. They aren’t sickened by the virus, but they are thought to carry and shed it for the rest of their lives. “That’s the definition of a good reservoir,” says rodent ecologist Elisabeth Fichet-Calvet of the Bernhard Nocht Institute for Tropical Medicine (BNITM), who has spent 2 decades studying the complex interplay of the Lassa virus and its host. Since 2016, several other rodent species in Nigeria, Guinea, Ghana, and Benin have been shown to harbor the virus. But M. natalensis remains the main driver of deadly outbreaks.

p

was barely able to walk and was having trouble breathing. Ten days after his symptoms began, he was finally diagnosed with Lassa fever and immediately transferred to the hospital’s high-containment isolation ward. By then Dagana was in and out of consciousness, with acute renal failure. The experience was “terrible, horrible,” said Dagana, who is married and has a young son. “People were suffering around me. I could hear them crying out.” A member of his community was brought in unconscious; 20 minutes later he was dead. “I was so frightened.” He spent more than 1 month in the hospital and underwent five rounds of dialysis. As of September 2023, his kidney function was still not back to normal. And yet he’s one of the lucky ones. Although about 80% of Lassa virus infections are mild or asymptomatic, 15%

to 20% of cases of severe disease are fatal—a number that can reach 70% in some places and in some epidemics. “That’s up there with Ebola,” says virologist Robert Garry of the Tulane University School of Medicine. Lassa fever kills far more people than Ebola does—10,000 or more a year, some researchers suspect, although no one knows for sure. But except for a small cadre of dedicated researchers, it has been largely ignored. Identified only half a century ago, the rodent-borne disease, which can more rarely be transmitted between people via body fluids, affects the rural poor, who live far from any health center. Most cases go undiagnosed and unreported, and many people die in their villages. “It is the most neglected of neglected diseases,” says Sylvanus Okogbenin, an obstetrician who chairs a working group at ISTH’s Institute of Viral and Emergent Pathogens Control and Research. A record-shattering epidemic in 2018 in Nigeria, the hardest hit country, put Lassa fever on the map, prompting both the World Health Organization (WHO) and the Nigerian government to declare a public health emergency. Confirmed cases, which numbered roughly 25 to 100 in previous outbreak years, reached 633, and 171 people died, including 45 health care workers. Cases in Nigeria have continued to climb each year since then. Infections still surge in the dry season, which runs from about December to May, but since 2018 have been occurring year-round. Most worrisome, cases are popping up outside the four known endemic countries: Nigeria, Sierra Leone, Guinea, and Liberia. As a result Lassa fever has vaulted up on the global health agenda, earning a spot on WHO’s list of notorious pathogens that have epidemic or pandemic potential and urgently require countermeasures. Several groups are working on badly needed diagnostics and drugs. The nonprofit Coalition for Epidemic Preparedness Innovations (CEPI) has named Lassa fever one of its priority diseases, investing in vaccine development and funding the largest ever study of Lassa’s true burden. And researchers in Africa and abroad are scrambling to learn why it is on the march—whether because of better surveillance, some change in the virus or in the rodents that carry it, a growing population, or a changing environment. “With climate change, what we saw 5 years ago doesn’t hold anymore,” says Gabrielle Breugelmans, director of epidemiology and data science at CEPI. “The hot spots of today are not the hot spots of tomorrow.”

Survivors’ stories Many of those who recovered from Lassa fever after treatment at the Irrua Specialist Teaching Hospital in Nigeria have dealt with life-changing consequences of the infection. Some are shunned by their communities, and many have permanent hearing loss.

Dora Eki Ikhazuangbe was teaching at a secondary school when she contracted Lassa fever in 1992, but gradual hearing loss forced her to retire early. At 74, she still wants to work. “If you give me a job, I will do it,” she says.

p g

Elizabeth Otoibhili was initially reluctant to go to the hospital when she learned she had Lassa fever, fearing her neighbors would shun her if they knew she had the disease. Only after consulting her pastor did she agree to seek treatment.

y

Muhammed Luqman Dagana was thought to have malaria or typhoid and was treated at three different clinics in 2023 before he was diagnosed with Lassa fever at the Irrua hospital. By then, he was near death. He has almost fully recovered. y g y ,

Wilfred Igberaese contracted Lassa fever while he was in school and lost hearing in his right ear. Since then, he has been unable to find a job. “I cannot communicate,” he says, and he cannot afford hearing aids. His wife is supporting him. “She is trying so hard.”

Bridget Ienlenaye, an intensive care nurse, has no idea how she contracted Lassa fever in 2022. But cases are common among health care workers, who often don’t know to protect themselves because they can’t tell from the vague early symptoms that their patients are infected.

N E WS | F E AT U R E S

the hospital too late for supportive care, and some even die in the ambulance. “Every day is key when you are talking about survival,” Okogbenin says. As the disease progresses, patients develop a sore throat, chest pain, abdominal pain, diarrhea, and vomiting. Once they begin bleeding from various orifices, the prognosis is grim. Seizures, confusion, and shock also portend a fatal outcome. Patients often die of renal failure, septic shock, or other complications about 14 to 21 days after the onset of symptoms. Lassa fever is especially dangerous during pregnancy and is usually fatal for the developing fetus.

Lagos and at BNITM. But the results would take weeks to come back, and in the meantime, patients would die. With the help of international partners, the hospital opened its own PCR lab in 2008. The turnaround time is now less than a day. But limited surveillance and testing continue to obscure the incidence of infection and disease across West Africa. One oft-cited figure of 100,000 to 300,000 infections a year and up to 5000 deaths was extrapolated from a small study in the 1980s in Sierra Leone, Breugelmans says. She and other researchers suspect the burden is far higher. Adetifa estimates that 2 million new infections occur

A region at risk Lassa fever has long been concentrated in endemic areas in Nigeria, Guinea, Liberia, and Sierra Leone. But in recent years, cases of the deadly hemorrhagic disease have been popping up in other parts of West Africa. With climate change and population growth, the virus is expected to extend its reach. Endemic areas

Small, regular outbreaks

p

In contrast to Ebola, where one spillover from bat to human can set off a chain of humanto-human transmission, the Lassa virus is usually contracted by direct contact with an infected rodent—young boys hunt and eat them—or its saliva, urine, and droppings. But the virus also spreads person to person, mostly when unsuspecting health workers without personal protective equipment (PPE) are exposed to an infected person’s blood or bodily fluids. “We lose a lot of specialized health care workers,” says Ifedayo Adetifa, director of the Nigeria Centre for Disease Control and Prevention (NCDC). Family clusters occur, too. In 2022, for example, a man returned to the United Kingdom from a trip to Mali and was hospitalized with what was later confirmed to be Lassa fever. He passed the virus on to his wife and child, and the child died. Infectious virus can persist in the body and semen for months after recovery, posing a long-term transmission risk.

Isolated cases

Mali Burkina Faso Benin

Guinea Sierra Leone Kenema Government Hospital

Ivory Coast

Nigeria

Ghana Togo

Liberia Atlantic Ocean

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

813

,

each year, with 300,000 to 500,000 clinical cases and 10,000 deaths. A recent study in PLOS Computational Biology ups the annual death toll to 18,000. The CEPI-funded Enable study, launched in 2020, aims to firm up the numbers. Researchers have been following about 23,000 people over 2 years in Nigeria, Sierra Leone, Benin, Guinea, and Liberia. Community health workers check in with each household every 2 weeks to see whether anyone has an acute febrile illness. If they test positive for Lassa fever, they are immediately isolated and hospitalized. So far, the study has found 39 cases—most of them at a study site in Edo—and five deaths. Many of those infected also had malaria. Enable is also periodically testing a subset of participants for asymptomatic infections.

y

No specific drugs exist, but IV fluids, oxygen, blood transfusions, and prophylactic antibiotics to prevent secondary infections can help. So can the antiviral drug ribavirin, although solid data are lacking. Many survivors are left with temporary or permanent hearing loss, which can rob them of jobs and livelihoods. The financial burden of treatment can be overwhelming, says Esther Okogbenin, a psychiatrist and researcher at ISTH who is married to Sylvanus Okogbenin. Stigma is intense, and survivors are often shunned by their communities, she says. Many suffer from post-traumatic stress disorder and depression. Only around 2003 did ISTH gain the ability to diagnose Lassa fever—by sending patient samples to collaborators with polymerase chain reaction (PCR) labs at the University of

Irrua Specialist Teaching Hospital

y g

SCIENCE science.org

y

went largely unrecognized in West Africa, even as communities in Nigeria and the other endemic countries began to document outbreaks of acute febrile illness. In January 1987, physician Felix Okogbo saw a rash of patients at his clinic in Ekpoma, Nigeria, near Irrua. Cases declined in March, then came back with a vengeance the following January. So he called the illness January fever. Then in January 1989, a woman in Ekpoma died of what was diagnosed as malaria, and her husband died soon after. After their two sons came home to bury them, both succumbed to the mysterious illness. One had already returned home to Chicago, where physicians were startled to find Lassa virus antigens during a postmortem. Okogbo himself fell ill in 1990, then lost his pregnant wife to a hemorrhagic illness the following year, during an outbreak that also killed 24 students at the university where she taught. Virologists came to investigate, and one sent a sample of Okogbo’s blood to CDC, where testing revealed antibodies to the Lassa virus. “That’s when I connected the dots,” says Okogbo, who joined ISTH in 1999. People blamed witchcraft for the deaths, Okogbo says. “I told them it is not witchcraft, it is a virus.” Even now, few health care workers suspect the disease when a patient arrives with fever and malaise, as Dagana learned last year. There are no cheap, rapid diagnostic tests that can be used in community clinics as exist for malaria. Only when a patient doesn’t respond to antimalarials or antibiotics, or health workers get sick, does Lassa come to mind—if it does at all. Patients often reach

g

CREDITS: (GRAPHIC) D. AN-PHAM/SCIENCE; (DATA) WORLD HEALTH ORGANIZATION; (PHOTOS, OPPOSITE PAGE) APOCHI OWOICHO

LONG AFTER THE DISEASE was identified, it

NE WS | F E AT U R E S

Preliminary results from the first 12 months, released in October 2023, suggest they are common: In Nigeria and in Liberia, 48% and 39% of participants, respectively, have antibodies to the virus in their blood. These findings will help researchers select sites for future vaccine trials, ideally conducted in communities with plenty of circulating virus and minimal immunity from previous exposure, which means a high risk of infection and illness.

science.org SCIENCE

,

nome as well. In 2007, as a postdoc at the Broad Institute, computational geneticist Pardis Sabeti developed an innovative method to mine the human genome for signals of recent, positive natural selection. The single strongest signal in the human population turned up in the Yoruba people in Nigeria: an increased frequency of a variant in a gene called LARGE. The gene encodes an enzyme that modifies a surface receptor to which the Lassa virus binds to gain entry into cells. Sabeti and colleagues suspected the variant may confer protection against Lassa fever, but didn’t have enough sequence data to know for sure.

y

DEADLY VIRUSES leave marks on the human ge-

y g

“I still see the scars at the community level,” Grant says, “especially among those whose family came into the hospital and never left.”

y

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

Felix Okogbo, Irrua Specialist Teaching Hospital

g

814

“I told them it is not witchcraft, it is a virus.”

“I had gone to medical school, and we had never learned about Lassa fever,” Sabeti says. Yet here it was, linked to the strongest signal of selective pressure on the human genome. “I thought, this is the most interesting virus ever, and hardly anyone is studying it.” So she and her group jumped in. They teamed up with molecular biologist Christian Happi, who now directs the African Centre of Excellence for Genomics of Infectious Diseases at Redeemer’s University in Nigeria, and together they set up collaborations with Lassa researchers at ISTH and KGH, training African scientists and providing research equipment and medical supplies. Already, Garry had established a training and research partnership with KGH, and virologist Stephan Günther at BNITM was beginning his long-term collaboration with researchers at ISTH. From those partnerships came the deepest dive yet into Lassa virus genomics and evolution. With a large team of African, European, and U.S. researchers, Sabeti and Happi analyzed blood samples collected earlier from Lassa fever patients at ISTH and KGH and sequenced 183 viral genomes, as well as 11 more from M. natalensis field samples. Their analysis of genetic variation among the viral sequences, published in Cell in 2015, confirmed previous suggestions that the virus has ancient origins, likely emerging in present-day Nigeria more than 1000 years ago, traveling westward to Guinea and Liberia about 400 years ago, and reaching Sierra Leone about 150 years ago. It is also one of the most genetically diverse viruses known—an order of magnitude more diverse than the Ebolavirus, for example. Seven distinct lineages have been identified so far, and “the genetic diversity continues to rise,” Garry says. “When it pops up in a new place, it has changed.” Each lineage is largely confined to a specific geographic area, suggesting the virus has been happily evolving and mutating in rodent populations that do not mix. “If you look at different viral lineages, they are very defined by river lines,” Sabeti says. She and others suspect some lineages may have evolved to become more virulent than others, which could help explain the severity of certain outbreaks. “A virus with that diversity has lots of ways it can go,” Sabeti says. Now, in new work published 17 years after Sabeti’s initial observation, she, Happi, and colleagues report that the LARGE variant is associated with a reduced likelihood of getting Lassa fever in a Nigerian cohort, but not a Sierra Leonean one. Their new study, published this month in Nature Microbiology,

p

“WHEN THE PAVED road ends, you know you are entering a Lassa community,” says Lansana Kanneh, a field surveillance officer in the Viral Hemorrhagic Fever Program at Kenema Government Hospital (KGH). It’s mid-September, and Kanneh is driving down a pitted dirt road, past women washing clothes in a muddy creek, to the remote village of Ngeihun, in the lower Bambara Chiefdom in Sierra Leone’s Eastern Province. This is M. natalensis territory, Kanneh says. Around Ngeihun, the fields and lush vegetation provide an ideal habitat for rodents. The flimsy housing here, made of crumbling mud brick with thatch or rusted tin roofs, does little to keep out the mice and the virus they carry. Kanneh points out the gaps under the front doors, the holes in the walls, the food stored in open containers. The small, windowless bedrooms are so dark that a rodent can happily spend its entire day in there, he says. Gutters are often clogged with garbage, and without running water, people have difficulty keeping dishes clean. “The houses where Lassa is are a fivestar hotel for a rodent,” says Donald Grant, director of the KGH program. In some villages around the small city of Kenema, up to 80% of the population has antibodies that show they have been exposed to the virus. Every time a Lassa fever case is detected, the outreach team from KGH visits the affected village within 24 hours. With the chief’s permission, they hold a community meeting with videos explaining the signs and symptoms of the disease, how it is transmitted, and the importance of proper food storage. They monitor the patient’s caregivers for the next 21 days. Then the ecology team sets live rodent traps in the house and around the community, baited with nut paste, ground cassava, oats, and dried fish. The following morning, clad in full PPE, the team removes the traps and any rodents they contain, and resets them. Most of the time they catch M. natalensis, says Ibrahim Abu Zacharia of the ecology team, and about two of every five animals are infected. Kenema has long been one of the hottest hot spots for Lassa fever in this part of West

Africa, but recently cases have been declining. KGH is now seeing about 50 cases a year, as opposed to 150 or 200 before 2013, Garry says. This could be evidence of more widespread immunity or a sign that community education about keeping out rodents has paid off. But Garry and others here fear the decline is illusory. A string of hardships in the region has bred mistrust of the health care system and the government, perhaps keeping people from seeking care. The Blood Diamond Wars, a devastating civil conflict lasting from 1991 to 2002, forced CDC to abandon its long-term Lassa fever research and control project there. Then in 2013, the largest ever Ebola outbreak hit Liberia, Guinea, and Sierra Leone, with Kenema at the epicenter. More than 11,000 people died, including 4000 Sierra Leoneans. Rumors swirled that Western health care workers had brought in the virus to kill them.

Jars at the Irrua Specialist Teaching Hospital hold preserved historical specimens of Mastomys, including M. natalensis (top), the main host of the Lassa virus. At the hospital, Grace Okonofua (bottom) processes patient blood samples.

,

PHOTOS: APOCHI OWOICHO

y

815

y g

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

y

SCIENCE science.org

g

THE EXPLOSIVE 2018 outbreak in Nigeria caught the hospital in Irrua off-guard. The Lassa Fever Isolation Ward was overflowing, Sylvanus Okogbenin recalls. “When the ward was full, we went to tents. When the tents were full, we moved into buildings that were not finished.” To see whether the virus had recently mutated to become more virulent or more transmissible person to person, researchers at the Broad Institute and Redeemer’s quickly began to sequence viral genomes from patients, feeding the information in real time

to NCDC. They found no evidence of a new, more virulent strain, just the same hodgepodge as from previous years, Happi says. And the data were consistent with multiple spillovers from rodents to humans, not increased human-to-human transmission. And yet Lassa fever is clearly on the rise—and on the move: In the past few years, some states in northern Nigeria have reported cases for the first time. Mali, Togo, and Benin have all reported their first cases in the past decade. If the virus hasn’t changed, what has? “Basically we don’t know fully,” says Danny Asogun, a public health physician at ISTH. Perhaps an increase in the number of rodents or the levels of virus they carry plays a role. Population growth, migration, crowding, and encroachment on wild lands are bringing people into closer contact with rodents,

p

also sheds light on the long-standing mystery of why some people develop severe disease and some mild. They identified two human genetic variants associated with Lassa fever fatality, although more work is needed to nail down their possible roles.

Asogun says. And new hosts, especially the Guinea multimammate mouse M. erythroleucus, which is also found in Nigeria, could be contributing, Fichet-Calvet says. Researchers have long puzzled over why Lassa fever cases are concentrated in specific hot spots in just four endemic countries when M. natalensis is all over sub-Saharan Africa. If they can figure out what determines this fragmented distribution, they may be better able to predict what other areas are at risk. Explanations abound. Perhaps only certain subspecies of M. natalensis carry the Lassa virus. Or in some regions the rodents may be coinfected with other arenaviruses that outcompete Lassa. But most attention has focused on ecological factors that favor virus circulation, perhaps by influencing its survival outside the host, or rodent behavior. Rainfall and to a lesser extent temperature seem to be key, Fichet-Calvet and David Rogers of the University of Oxford proposed in a 2009 study. Recent modeling by Raphaëlle Klitting and a team of international collaborators suggests current viral hot spots have a very particular combination of annual rainfall, annual temperature, and type of land cover. All of those conditions could change as the climate does, so the researchers modeled the possible distribution of the Lassa-friendly habitats in the coming decades under three climate change scenarios, adding other factors such as the conversion of forests to agriculture and population growth. By 2070, under a “moderate” climate change scenario, areas suitable for Lassa virus circulation could cover most of the region between Guinea and Nigeria. Susceptible regions could also appear for the first time in parts of Central and East Africa, Klitting’s team reported in 2022 in Nature Communications. In the worst case—that is, if the virus settles in all suitable habitats—by 2070 up to 700 million people could be at risk of Lassa fever, up from 92 million today. “This is not going to be a problem that goes away,” says Klitting, who is now head of genomics at the National Reference Center for Arboviruses in Marseille, France. She suspects the virus is likely to move slowly, based on its past dispersal patterns. Then again, “We’ve seen Lassa virus move long distances in the past,” she says, citing its westward migration from Nigeria hundreds of years ago. And she notes that Ebola, considered a disease of Central Africa, suddenly emerged in West Africa in late

NE WS | F E AT U R E S

p

Ibrahim Abu Zacharia, a member of the Lassa fever ecology team at Kenema Government Hospital in Sierra Leone, collects a live rodent trap from a home.

2013, with devastating effects. “We don’t know how often these big jumps happen,” she says, “but it could happen.”

“The houses where Lassa is are a five-star hotel for a rodent.”

science.org SCIENCE

,

Leslie Roberts is a science journalist in Washington, D.C.

y

Donald Grant, Kenema Government Hospital

for Lassa fever in Nigeria, with 9155 suspected cases, 1270 confirmed, and 227 deaths. ISTH was better prepared than in 2018, but it still had to open an overflow ward. The power supply was unstable, and clean linens and even soap were sometimes in short supply. Already, case counts suggest this year will be even worse. No one expects the Lassa virus to spark a global pandemic—unlike respiratory viruses such as SARS-CoV-2 or influenza that easily spread from one person to another. But for those who study the disease and treat its victims, the steady march of the disease is alarming. “Lassa fever is already in all of West Africa,” Sylvanus Okogbenin says. “That’s pandemic enough, really.” j

y g

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

LAST YEAR WAS THE WORST on record

y

816

Okogbenin and others have found that when the disease is still mild, supportive care can save both mother and fetus.  The perinatal mortality rate, which was 80% to 95% in 2018, has dropped to 50% to 65%, he says. But at 17.9%, the overall Lassa fever fatality rate in Nigeria “is still unacceptably high,” Adetifa says. “We want to bring it to the single digits.” In Kenema, the rate is still 70%. Garry, Grant, and others think lineage IV, which circulates there, may be more virulent than the others. Late diagnosis could also be a factor, Garry says, or simply quality of care. Sierra Leone is one of the poorest countries in the world, and the government hospital cannot afford dialysis machines.

g

WHEN IT COMES to countermeasures, “I think we have turned a corner,” Garry says. “When I started working on Lassa fever 20 years ago, there was nothing you could do.” Now, several groups are developing affordable, rapid diagnostic tests for use in rural communities. A new small molecule drug, Favipiravir, which interferes with viral genome replication and is already licensed in Japan to treat influenza, has just finished a phase 2 trial for Lassa fever at ISTH. “It’s a huge, huge development, but it’s still early stages,” says Günther, a partner on the study. “Maybe in 5 to 7 years we might have several drugs with better efficacy than ribavirin.” Meanwhile, CEPI is supporting the development of three vaccine candidates, each of which uses a viral vector to deliver a Lassa virus protein that triggers an immune response. CEPI aims to have one of these vaccines licensed in 5 to 10 years, Breugelmans says. Two are in phase 1 trials in healthy volunteers, and one may soon enter a larger phase 2 trial in several endemic countries. CEPI is also providing up to $40 million to support early work on a messenger RNA vaccine. Günther has a “very good feeling” about vaccines but cautions that “everything takes more time than you think in low-resource settings.” Access and affordability are issues. “Can we get vaccines and drugs to people who need them most?” Garry asks. Establishing vaccine manufacturing capabilities in Africa is part of the solution, Happi adds.

Complicating all these efforts is the genetic diversity of the virus. In the best case, diagnostics, drugs, and vaccines would work against all Lassa virus strains. So far vaccinemakers have focused on lineage IV, prevalent in Sierra Leone, Guinea, and Liberia, with the hope that the shots will be crossprotective. But only large-scale trials will tell. Even without a specific drug, Nigeria has been making steady progress in reducing the death rate, thanks in part to better supportive care. ISTH has purchased dialysis machines, introduced an improved, lower dose regimen for ribavirin, and taken a new approach to treating pregnant women. Previous guidelines called for terminating the pregnancy to save the mother’s life, but Sylvanus

PERSPECTIVES SOLAR CELLS

Tandem modules get better An all-perovskite tandem solar module breaks 24% efficiency By Cong Chen and Dewei Zhao

y y g

RE FE RE NC ES AN D N OT ES

M. A. Green et al., Prog. Photovolt. Res. Appl. 31, 3 (2023). Q. Jiang et al., Science 378, 1295 (2022). R. Lin et al., Nature 620, 994 (2023). H. Gao et al., Science 383, 855 (2024). T. Bu et al., Science 372, 1327 (2021). C. Liu et al., Science 382, 810 (2023). X. Y. Chin et al., Science 381, 59 (2023). R. He et al., Nature 618, 80 (2023). J. Zhu et al., Nat. Energy 8, 714 (2023). K. Xiao et al., Science 376, 762 (2022). M. V. Khenkin et al., Nat. Energy 5, 35 (2020). S.-H. Turren-Cruz, A. Hagfeldt, M. Saliba, Science 362, 449 (2018). 13. C. Wang et al., Nat. Energy 7, 744 (2022). 14. B. Abdollahi Nejand et al., Nat. Energy 7, 620 (2022). 15. M. Jošt, L. Kegelmann, L. Korte, S. Albrecht, Adv. Energy Mater. 10, 1904102 (2020).

AC KN OW LED G M E N TS

The authors receive support from the National Natural Science Foundation of China (nos. 52303347 and 62174112), the Fundamental Research Funds for the Central Universities (no. YJ2021157), and the Engineering Featured Team Fund of Sichuan University (no. 2020SCUNG102).

10.1126/science.adn7930 23 FEBRUARY 2024 • VOL 383 ISSUE 6685

829

,

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.

y

SCIENCE science.org

solvent during the coating of the perovskite onto the substrate. Gao et al. thus extended the window of time for a uniform largearea film of perovskite to form during the coating process—from 10 to 100 s—before crystallization was observed. A light-absorbing perovskite film typically has two contacts in a device—its top and bottom surfaces (buried interface). The quality of the interface is equally as important as homogeneity of the perovskite film (6–9). When coating the perovskite film onto a substrate, the top-to-bottom crystallization in the film easily induces defects at the buried interface as the solvent evaporates. This causes charge recombination—an energy-wasting process. Gao et al. observed an increase in the distribution of the AAH salt from the top to the bottom of the perovskite layer. The AAH+ ion at the bottom passivated the interfacial defects and reduced charge trapping, thus minimizing recombination loss at the buried interface. Ultimately, the synergistic effect of a uniform high-quality perovskite film and a passivated buried interface facilitated the performance improvement of Sn-Pb low-bandgap perovskite bottom subcells. Previous work on large-area widebandgap perovskite top subcells revealed that the compositional engineering of perovskites could affect the homogeneity

g

College of Materials Science and Engineering and Engineering Research Center of Alternative Energy Materials and Devices, Ministry of Education, Sichuan University, Chengdu, China. Email: [email protected]

“The approach may offer a viable route for commercialization.”

p

S

olar cells play a vital role in transforming energy needs. Single-junction solar cells made of metal halide perovskite materials have shown great potential, achieving about 26% efficiency in converting sunlight into electricity (1). This makes them attractive for photoelectronic applications, such as lasers, light detectors, and lightemitting diodes. Breaking this efficiency limit of single-junction solar cells could be achieved with tandem solar cells, where thermalization loss is reduced and solar spectrum utilization range is broadened (2). Small-area all-perovskite tandem cells can reach an efficiency of >28% (3). However, developing large-area (~20 cm2) allperovskite tandem modules is challenging because of difficulties in controlling homogeneous crystallization when scaling up perovskites. On page 855 of this issue, Gao et al. (4) report large-area all-perovskite tandem modules that achieve >24% efficiency. The approach may offer a viable route for commercialization. Tunable bandgaps of perovskites ensure the feasibility of all-perovskite tandem solar cells that consist of a wide-bandgap (~1.70 to 1.85 eV) perovskite top subcell and a low-bandgap (~1.2 to 1.3 eV) perovskite bottom subcell. A major obstacle in developing large-area all-perovskite tandem cells is inefficiency, attributed to inferior and heterogeneous crystallization of both wide-bandgap and low-bandgap perovskite layers. In general, a wet perovskite layer needs to be uniformly applied as a coating onto a conductive substrate. Coating the perovskite onto a large area requires more time than for a small area. This increase in time allows nonuniform crystallization of perovskite phases to spread. Therefore, it is necessary to develop strategies that extend the window of time for a homogeneous perovskite film to form. The slow nucleation of a wet perovskite film and the slow volatilization of sol-

vent can prolong the window of time for perovskite crystallization through scalable solution-processed coating technology (5). Gao et al. added a type of zwitterionic salt, aminoacetamide hydrochloride (AAH), to a mixed tin (Sn)–lead (Pb) metal halide perovskite precursor to trigger interactions between the constituents, which included lead iodide, tin iodide, formamidinium iodide, and solvents (dimethylformamide and dimethyl sulfoxide). In particular, coordination of the AAH salt and dimethylformamide effectively slowed the release of

of crystallization and facilitate the formation of large-area high-quality perovskite films in the coating process (10). Gao et al. noted that the large-area all-perovskite tandem solar module provided a certified efficiency of 24.5% with an aperture area of 20.25 cm2, exceeding that (22.4%) of a single-junction perovskite solar module at the same scale. This high efficiency offers a broad design space for perovskite technology. However, module reliability is a point of concern for future commercialization. Gao et al. observed that methylaminefree (formamidine and cesium) perovskite tandem modules with encapsulation had better operational, damp-heat, and thermal cycling stability compared with those based on formamidine and methylamine low-bandgap perovskite bottom subcells (11). Methylamine in perovskites is prone to volatilization under harsh stresses, such as heat and illumination, triggering perovskite decomposition (12). This should guide stability improvement of allperovskite tandem modules. The findings of Gao et al. support a focus on optimizing additive engineering and coating techniques for large-scale uniform high-quality perovskite films (13, 14). Further improving the efficiency of a largearea all-perovskite tandem solar module to the predicted practical limit of ~34% (15) is an important next goal. Also, minimizing the annual degradation of perovskite materials in the actual outdoor environment is a crucial factor in determining the commercial capability for all-perovskite tandem modules. Indeed, there is much room for improvement compared with the 25-year lifetime of mainstream crystalline silicon photovoltaics. j

I NS I GHTS | P E R S P E C T I V E S

COMPUTER SCIENCE

Overcoming the noise in neural computing Circuit strategies can enable noisy analog hardware to achieve high precision By James B. Aimone1 and Sapan Agarwal2

I

Sandia National Laboratories, Albuquerque, NM, USA. Sandia National Laboratories, Livermore, NM, USA. Email: [email protected]; [email protected] 2

y g

Challenges in computer architecture paradigms

,

Static data

Z. Al-Aly, Y. Xie, B. Bowe, Nature 594, 259 (2021). H. E. Davis et al., Nat. Rev. Microbiol. 21, 133 (2023). S. R. Stein et al., Nature 612, 758 (2022). A. C. Wong et al., Cell 186, 4851 (2023). J. Klein et al., Nature 623, 139 (2023). Y. Xie, T. Choi, Z. Al-Aly, JAMA Intern. Med. 183, 554 (2023). C. T. Bramante et al., Lancet Infect. Dis. 23, 1119 (2023). B. Bowe, Y. Xie, Z. Al-Aly, Nat. Med. 28, 2398 (2022). B. Bowe, Y. Xie, Z. Al-Aly, Nat. Med. 29, 2347 (2023). L. Spinney, Nature 602, 383 (2022). D. M. Altmann et al., Nat. Rev. Immunol. 23, 618 (2023). E. J. Topol, A. Iwasaki, Sci. Immunol. 7, eadd9947 (2022). S. Iketani et al., Nature 613, 558 (2023). K. S. Levine et al., Neuron 111, 1086 (2023). K. Bjornevik et al., Science 375, 296 (2022).

AC K N OW L E D G M E N TS

Z.A.-A. carried out consulting (uncompensated) for Pfizer.

Von Neumann

Input data

Computer architecture Memory

Output data

Instructions

Errors

Scaling challenges Increased communication distance through bottleneck

Processor Resistive crossbars

Increased noise and analog read errors

Spiking neuromorphic

Algorithm challenges to ensure stability and efficiency

10.1126/science.adl0867

832

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

y

Von Neumann designs separate processing from memory, creating a bottleneck in communication that is problematic with scaling. Analog crossbar architectures perform linear algebra operations, but as the number of crossbars increases, multiple noise sources affect computation precision. Spiking neuromorphic systems distribute computation over a large network but require new algorithms.

R E F E R E N C ES A N D N OT ES

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

y

1

g

n the past 20 years, microelectronics technology has seen a tapering off of Moore’s law (the exponential growth of transistor density on computer chips) and an end to Dennard scaling (wherein smaller transistors use correspondingly less power). To create computing systems that are more powerful, faster, and energy efficient for artificial intelligence (AI) and numerical computing applications, new approaches to computing are needed. Attaining these new capabilities not only requires exploration of new materials and devices, but as reported by Song et al. on page 903 of this issue (1), it also necessitates innovative circuits and new mathematical approaches. Multicore computer architectures, which allow more than one process to operate at the same time, and a system-on-chip approach, which integrates all processes on a single computer chip, have supported massive applications such as large language models and other AI technologies, but at

great cost (2, 3). Scaling applications to larger sizes in conventional architectures requires information to move greater distances between memory and logic. Moreover, this movement has an energy cost. For example, for a single bit to move from one end of a chip to the other is on the order of a picojoule—10,000 times more than the energy to switch a transistor (4). The separation of memory and logic in a traditional computer design (called the von Neumann architecture) enables programmability, precision, and scaling to larger programs and datasets. New architectures must trade off something to gain efficiency because it is physically impossible to place enough memory near general-purpose computing to have the same benefits of a traditional design. Two common approaches to overcome the von Neumann bottleneck—the slow movement of data between the memory and processor—make different trade-offs. Digital brain-like “spiking” architectures (neuromorphic computing) use low-precision but highly reliable event-driven communication, and analog crossbars use dense multibit memories but are limited by the noise inherent to

p

appetite for COVID-19 boosters, strategies to improve uptake (e.g., pairing it with the annual influenza vaccine) may be effective. Development of more durable, variant-proof vaccines that are not vulnerable to evasion by the ever-mutating virus needs to be accelerated (12). Nasally or orally administered vaccines that induce strong mucosal immunity to block infection and transmission should be pursued, and there are preliminary supportive data from clinical trials (10). It is also necessary to broaden the pipeline of SARS-CoV-2 antivirals, especially because of rising resistance (13). Governments and funding agencies should support a comprehensive portfolio of research in infection-associated chronic illnesses. Although SARS-CoV-2 can cause chronic disease, it is not alone. Influenza virus, Epstein-Barr virus, Ebola virus, polio virus, and many others also have long-term health effects (14, 15). Myalgic encephalomyelitis/chronic fatigue syndrome is also likely triggered by infection. However, how infectious agents cause chronic disease is not fully understood, and research in this area has been disproportionately underfunded relative to the burden of long-term disability and disease caused by infections. An interdisciplinary research agenda is essential to address knowledge gaps about such illnesses. This effort will not only provide insights into infection-associated chronic illnesses similar to Long Covid but also inform strategies to reduce the burden of chronic health loss from future pandemics. The world has already paid an extraordinarily heavy price for the biggest pandemic of the 21st century so far. In this crisis lies an urgent need to address the challenge of Long Covid. There is also a historic opportunity— that must not be squandered—to unlock a deeper understanding of infection-associated chronic illnesses and optimize preparedness for future pandemics. The world must rise to the occasion and address these challenges; the health and well-being of current and future generations depend on this. j

science.org SCIENCE

10.1126/science.adn8545

O

ver the past decade, planting trees has been widely promoted as a naturebased solution to remove carbon dioxide (CO2) from the atmosphere and thereby help to achieve the warming targets of the 2015 Paris Agreement (1). Previous studies (2–5) have compared the CO2 removal potential of planting trees, often against another land-based carbon mitigation option—bioenergy with carbon capture and storage (BECCS). On page 860 of this issue, Weber et al. (6) report their investigation of the effects of extensive forest expansion on Earth’s energy balance using Earth system models (ESMs) with a more complete representation of the climate and Earth system. They find that changes in albedo (surface reflectivity) and atmospheric composition from forest expansion could offset up to one-third of the climate mitigation effects of the CO2 sequestered. This suggests that the benefits of forestation may be overestimated. Forests cover 31% of the global land area but are not distributed uniformly (7). The Bonn Challenge, the New York Declaration on Forests, and the UN Decade on Ecosystem Restoration share a common ambition to restore 350 Mha of degraded and deforested lands by 2030 (~2.7% of the global land area). Such forest expansion (referred to as forestation) covers afforestation (planting trees where there were none recently), reforestation, and avoided deforestation. The other widely considered land-based carbon mitigation option is BECCS, where crops (e.g., miscanthus) or short-rotation coppice species (e.g., willow) are grown for energy production, with the capture and long-term storage of the CO2 produced during their combustion. The carbon mitigation potential for afforestation and reforestation is up to 10 billion tonnes (Gt) of CO2 per year, whereas that of BECCS is estimated to be 0.4 to 11.3 Gt of CO2 per year (8). Notably, the upper limits of these estimates are about a quarter of the total anthropogenic CO2 emissions in 2022 [40.7 ± 3.2 Gt of CO2 per year (9)]. A simplified climate model, implementing land-use scenarios that achieve either the 23 FEBRUARY 2024 • VOL 383 ISSUE 6685

833

,

J.B.A. and S.A. receive support from the US Department of Energy (DOE) Office of Science and Sandia National Laboratories Laboratory Directed Research and Development program. The US government retains a nonexclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this article or allow others to do so, for US government purposes. The views expressed here do not necessarily represent the views of DOE or the US government.

By Garry Hayman

y

ACK NOWL ED G M EN TS

Planting trees has diverse effects on composition and climate mitigation

y g

1. W. Song et al., Science 383, 903 (2024). 2. P. Henderson et al., J. Mach. Learn. Res. 21, 10039 (2020). 3. D. Patterson et al., arXiv:2104.10350 [cs.LG] (2021). 4. I. IRDS, “International Roadmap for Devices and Systems, 2022 White Paper: More Moore” (2022); https://irds.ieee.org/editions/2022/more-moore. 5. J. B. Aimone et al., Neuromorph. Comput. Eng. 2, 032003 (2022). 6. I. Hubara et al., J. Mach. Learn. Res. 18, 1 (2018). 7. J. D. Smith et al., Nat. Electron. 5, 102 (2022). 8. B. Feinberg et al., in 2021 IEEE International Symposium on High-Performance Computer Architecture (HPCA) (IEEE, 2021), pp. 761–774. 9. T. P. Xiao et al., IEEE Circuits Syst. Mag. 22, 26 (2021). 10. S. Misra et al., Adv. Mater. 35, 2204569 (2023).

Forestation is not an easy fix

y

R EFER ENCES AN D N OT ES

CLIMATOLOGY

g

SCIENCE science.org

have one bit of precision, but the collection of many independent bits can achieve extremely high precision. Similarly, analog noise limits any crossbar linear algebra computation to a relatively low level of precision. Song et al. found that they could dedicate subsequent crossbars to specifically address the residual error (the difference between desired and realized precision) and thus allow the desired precision to be reached overall. This is not altogether different from how numerical computing works in general, with successive relatively easy calculations approaching a desired solution that cannot be directly solved itself. The programming error addressed by Song et al. is just one source of error in analog systems. Read noise, drift, temperature variations, parasitic resistances, and analog circuit mismatches all limit the precision of the overall computation (9). Addressing these other challenges or linking this technique to different types of hardware accelerators will be critical for expanding this preconditioning approach to the other numerical steps in numerical applications. Seemingly inconsequential operations that are trivial in digital logic, such as generating a random number, often do not prove simple in emerging computing. Therefore, expanding the diversity of what may be compatible within a system, including analog crossbars, will be essential. For instance, can the noise of analog devices replace expensive pseudo-random number generators (10)? Future computing systems that efficiently combine different classes of architectures (see the figure) with a flexible approach to addressing precision concerns should reveal new opportunities for advancing numerical methods. j

p

analog computing. Spiking algorithms, which a single bit (a “1”) of information moves in a brain-like fashion, analogous to information transmission between neurons, show performance advantages at large scales on Monte Carlo simulations and graph analytics (5). However, these benefits are restricted to numerical methods that are well-suited for exploring many simple models in parallel, such as simulating the random walks of a stock price, and tasks with high-dimensional and complex data, such as social networks. Alternatively, analog crossbars implement dense linear algebra computations by efficiently performing matrix vector multiplication using the physics of the devices. Here, an analog memory element or programmable resistor replaces a large multiplier that would require thousands of transistors within a single device. Linear algebra is an important part of many existing algorithms, so using analog computing does not require rewriting algorithms as in spiking systems, but moving from digital logic to analog computing introduces a new drawback in the accumulation of noise. Nevertheless, many AI algorithms are resilient to noise, allowing them to operate on low-precision analog hardware (6). Apart from AI, the reliance on highprecision arithmetic is a major challenge for emerging computing. Although there has been success in leveraging new mathematical approaches to solve problems with spiking systems [such as using a stochastic differential equation instead of a partial differential equation (7)], mixed-precision approaches can be used to overcome precision concerns in analog crossbar approaches. For example, analog crossbars can solve complex linear systems that typically require high precision by using the hardware to implement preconditioners. Such preconditioners can make otherwise hard numerical computing problems more simple for higher-precision central processing unit solvers to find a precise overall solution with fewer digital operations (8). Song et al. demonstrated high-quality solutions to numerical solvers (computer programs that use numerical schemes) by developing a method to reduce programming error, thereby allowing for more accurate analog hardware. The authors overcame remaining analog errors by using an analog preconditioner as part of a faulttolerant mixed-precision algorithm to solve the overall numerical system. Although device and circuit noise limit the ability of a single crossbar to achieve the desired numerical precision, the authors recognized that multiple crossbars could be used to accomplish the same thing. In digital computing, precision is an emergent concept. Single transistors only

I NS I GHTS | P E R S P E C T I V E S

its vulnerability to other stresses or disturbances, such as fire, ozone damage, pests, and disease (12). There is an urgent need to quantify these impacts before large-scale forestation is undertaken. Planting trees has other environmental benefits, including increasing biodiversity and the provision of food and forestry products. To be effective for climate mitigation, biodiversity, and other ecosystem services, the forestation targets can only be met with practical guidelines and tools, institutional alignment, accountability mechanisms, and a robust evidence base (13). Local communities must also be engaged (14). j R E F E R E N C ES A N D N OT ES

y

1. United Nations Framework Convention on Climate Change, “Adoption of the Paris Agreement, FCCC/ CP/2015/L.9/Rev. 1” (2015). 2. A. B. Harper et al., Nat. Commun. 9, 2938 (2018). 3. G. D. Hayman et al., Earth Syst. Dyn. 12, 513 (2021). 4. I. Melnikova et al., Commun. Earth Environ. 4, 230 (2023). 5. Y. Cheng et al., Proc. Natl. Acad. Sci. U.S.A. 121, e2306775121 (2024). 6. J. Weber et al., Science 383, 860 (2024). 7. Food and Agriculture Organization of the United Nations, The State of the World’s Forests 2022: Forest Pathways for Green Recovery and Building Inclusive, Resilient and Sustainable Economies (2022). 8. Intergovernmental Panel on Climate Change, Climate Change and Land: An IPCC Special Report on Climate Change, Desertification, Land Degradation, Sustainable Land Management, Food Security, and Greenhouse Gas Fluxes in Terrestrial Ecosystems (2019). 9. P. Friedlingstein et al., Earth Syst. Sci. Data 15, 5301 (2023). 10. J. Laothawornkitkul et al., New Phytol. 183, 27 (2009). 11. R. Grote, U. Niinemets, Plant Biol. 10, 8 (2008). 12. M. Bustamante et al., Glob. Sustain. 10.1017/ sus.2023.25 (2023). 13. R. Chazdon, P. Brancalion, Science 365, 24 (2019). 14. K. D. Holl, Science 355, 455 (2017).

g

AC K N OW L E D G M E N TS

G.H. thanks D. Lawrence (National Center for Atmospheric Research) for useful discussions. G.H. also acknowledges funding from the UK Natural Environment Research Council for the TerraFIRMA project (NE/W004895/1). 10.1126/science.adn7026

y g

UK Centre for Ecology & Hydrology, Wallingford, UK. Email: [email protected]

the figure). Using two leading ESMs, Weber et al. investigated the effects of extensive forest expansion in suitable regions for two contrasting future atmospheric composition states: a warming scenario, with high amounts of deforestation relative to 2015, and a sustainable scenario, in which 40% forest expansion is achieved. They found that forestation decreases the albedo and, through BVOC emissions, increases aerosol formation and the concentrations of greenhouse gases (methane and ozone), causing changes in Earth’s energy balance. These changes occur through absorption at Earth’s surface (albedo) and scattering (by aerosol particles) of incoming solar radiation and through the absorption of outgoing thermal radiation from Earth’s surface (by ozone and methane). For the warming scenario, the overall change in Earth’s energy balance reduced the climate benefits of CO2 uptake by up to a third. The offset is smaller (14 to 18%) when forestation occurs alongside the reduction of emissions of CO2 and other atmospheric pollutants, as defined in the sustainable scenario. The study of Weber et al. reveals a possible overestimation of the climate benefit of CO2 sequestration by forestation owing to the release of BVOCs and changes in albedo. However, several other land surface feedbacks are not included in the ESMs that were used—e.g., the effects of wildfires on vegetation and of ground-level ozone on vegetation and crop productivity [and hence CO2 uptake (3)]. Planting trees or growing bioenergy crops are long-term climate mitigation solutions. But how resilient will forests (and bioenergy crops) be to future climate hazards and compound climate events? It takes time for vegetation to recover, for example, after a severe hot-dry period, and this may increase

p

2.0°C or 1.5°C warming targets of the Paris Agreement, was used to determine whether forestation or BECCS is the preferred mitigation option from a carbon-cycle perspective (2, 3). Forest-based mitigation was found to be more effective than BECCS for atmospheric CO2 removal, especially if bioenergy crops replace ecosystems with high carbon content. Furthermore, the effectiveness of BECCS was strongly dependent on assumptions about the choice of bioenergy crop, its biomass and use efficiency, and the fossil fuel emissions that are offset by using bioenergy production. More recent studies (4, 5), using more complex ESMs, investigated the impact of large-scale expansions of forests or bioenergy crops on the exchanges of energy and water between the land surface and atmosphere, arising, for example, from changes in surface albedo, evaporation, and plant transpiration. These analyses found differences in the amount of CO2 sequestered and in the surface climate both locally and regionally, with clear implications for the suitability of forestation and bioenergy crops in different regions of the world. Alongside the impact on global carbon budgets and regional energy and water cycles, vegetation—and trees in particular— are sources of biogenic volatile organic compounds (BVOCs). BVOCs are produced by plants for growth, development, reproduction, defense, and communication (10). There are both biological and physicochemical controls on BVOC emissions (11). Through their contribution to the production of tropospheric ozone and aerosol particles, BVOCs indirectly affect Earth’s energy balance (see

How forests and bioenergy crops affect the Earth system y

There are many coupled land and atmospheric processes, such as albedo, photosynthesis, and the release of BVOCs by vegetation, that have been included in recent ESMs. ESMs are used to investigate the effects of forests and bioenergy crops on the surface climate and carbon cycle. Although forestation increases CO2 uptake, this can be partially offset by attendant changes in albedo and atmospheric composition.

,

Aerosol OH

Solar radiation

NO3 CCN

O3

Transpiration

Transpiration

BVOCs Respiration

Precipitation

Condensation, evaporation

CO2

Forest

Photosynthesis

Respir piration Sensible heat

Longwave radiation

CO2

Photosynthesis

Bioene Bio enerrg ener energ g gyy cr cro ops op ps

Albedo

BVOCs, biogenic volatile organic compounds; CCN, cloud condensation nucleus; CO2, carbon dioxide; ESM, Earth system model; NO3, nitrate; O3, ozone; OH, hydroxyl radical.

834

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

science.org SCIENCE

INSIGHTS P OLICY FORUM

p g y y g y ,

EDUCATION

Humane genomics education can reduce racism Moving instruction “beyond Mendel” can counter inaccurate, essentialist views By Brian M. Donovan1, Monica Weindling1, Jamie Amemiya2, Brae Salazar1, Dennis Lee1, Awais Syed1, Molly Stuhlsatz1, Jeffrey Snowden1

F 818

or as long as the concept of race has existed, racial prejudice has been justified on hereditary grounds (1, 2). Justifications of prejudice often misappropriate the work of Mendel, who first expounded a scientific model of 23 FEBRUARY 2024 • VOL 383 ISSUE 6685

inheritance by breeding peas (3). Today, our understanding of inheritance has moved far beyond Mendel, and insights from genomics refute the prejudiced idea that racial inequality is determined by genes (1). Even so, many believe that inequality is genetic because they

are biased by an inaccurate conception of race called “genetic essentialism” (1, 2, 4). We present data from a randomized trial to argue that if teachers move genetics instruction beyond Mendel and toward more complex genomics concepts—what we call “humane genomics education”—they can protect students from believing in unscientific notions of genetic essentialism and support their sciscience.org SCIENCE

,

819

y

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

y g

SCIENCE science.org

y

BSCS Science Learning, Colorado Springs, CO, USA. Department of Psychology, Occidental College, Los Angeles, CA, USA. Email: [email protected]; [email protected] 2

If “races” are defined as geographic populations, then the essentialist assumption that there is little to no genetic variation among individuals of the same race—and thus that most genetic variation is between races—is wrong. (E) Also, social disparities between races involve differences in complex traits. Because complex traits are multifactorial and influenced by interactions between genes and environments (12), it is not scientifically accurate to claim that racial inequality is determined by genetic variation alone (1). (F) Indeed, humans inherit their genomes along with their environments, and scientists have not yet developed convincing or ethical methods to disentangle gene-environment covariance (1). Because racial differences in a trait can be environmentally determined even when intragroup differences in that same trait are genetically influenced, there is good reason to be skeptical of the claim that racial disparities are genetic: Historical and present-day racism have made the environments of racialized populations different (1). Teaching students genomics concepts D to F for the purpose of refuting genetic essentialism is what we call “humane genomics education.” RCTs conducted in the District of Columbia, Colorado, California, and Massachusetts already demonstrate that teaching 8th- to 12th-grade students about concept D can cause a reduction in students’ genetic essentialist beliefs about race (1). This reduction is driven by a change in how students perceive interracial gene variability (1). Furthermore, students with a greater understanding of multifactorial inheritance (i.e., concept E) are more likely to disbelieve genetic essentialism after learning concept D (1) because they are more likely to develop the perception that races are not that genetically different. Therefore, the connection between genetics instruction and belief in genetic essentialism of race depends on how instruction affects students’ perceptions of genetic variation between races (1). When students develop an accurate view of such variation (concept D) their belief in genetic essentialism declines (and vice versa). That said, those previous RCTs have shortcomings that undermine their educational importance. They have not demonstrated that reductions in genetic essentialism are accompanied by increases in genomics knowledge (concepts D to F), which undermines the claim that genomics education is responsible for these effects. Also, previous RCTs have not demonstrated whether students adopt a more accurate, social constructionist view of race (2). Additionally, the results of previous studies may have been biased toward a reduc-

g

1

GENETICS EDUCATION AND ESSENTIALISM Around the world, students receive a basic genetics education that focuses on singlegene inheritance (1). In a basic genetics education, students learn (A) Mendel’s laws of heredity and (B) how different versions of a gene (i.e., alleles) are inherited across generations through probabilistic mechanisms that are easily modeled by a Punnett square. Teachers also introduce students to (C) the DNA molecule and to the molecular processes (i.e., the central dogma) that link genotypes to phenotypes. Students often learn how mechanisms A to C operate in humans by exploring the racial prevalence of monogenic diseases such as sickle cell anemia (SCA) and cystic fibrosis (CF) (1). All genetics education standards in the United States follow this story (1). The problem is that the basic genetics education that the US public receives is a risk factor for the development of genetic essentialism during adolescence (1). This claim is based on several studies [see (1)], particularly findings from randomized control trials (RCTs) that show that learning about monogenic diseases can cause greater belief in genetic essentialism of race (1). For example, contrary to what genetic essentialism predicts, there are no gene variants—including SCA and CF alleles— that most individuals in one race possess and no individuals in another race possess. Yet instruction on SCA and CF genotypes in Black and white populations leads students to develop this perception (1). This perception then facilitates an increase in the belief that genes determine racial disparities (e.g., in educational attainment) (1). Because basic genetics education does not discuss patterns of racial similarity in the human genome (1), and because it does not discuss the multifactorial basis of complex human traits (1), students are never exposed to information that explicitly counters genetic essentialist views about race. At best, basic genetics instruction fails to challenge genetic essentialist beliefs about race and, at worst, it could unintentionally lead students to construct them (1). We contend that genetics education needs to move beyond Mendel and toward the complexity of human genomics if it is to prevent the development of genetic essentialism (1). To understand why, consider the following three genomics concepts (D to F): (D) Roughly 0.1% of the human genome varies between individuals, and when population geneticists partition this variable DNA, they find that most variation occurs within geographic populations (~95%) and much of the genetic variation that occurs across such populations (~5%) consists of common alleles that vary in frequency (10).

p

entifically accurate understanding of race as a social construction. Genetic essentialism is a form of psychological essentialism, which is an early-developing bias in humans (4). Psychological essentialism is observable across human cultures and refers to the belief that members of a social category share an unobservable and internal essence that determines their traits (4). People who endorse genetic essentialism believe that such essences are genetic (4), which leads them to believe that same-race individuals are genetically homogeneous, that races are nonoverlapping genetic groups, and that most racial differences are therefore determined by genes (4). Essentialist beliefs are socially dangerous and a biological misconception (1, 2, 4). For example, genetic essentialist beliefs about race facilitate intergroup hostility (5), support for eugenic policies (6), discrimination (4), and disinterest in cross-racial friendships (7). Psychological essentialism also inhibits biology learning because it involves misunderstandings of intraspecific genetic variation (1), such as the erroneous belief that human races are like dog breeds (8). For example, US high school students and adults inaccurately estimate that 37% of human genetic variation exists between racial groups (9), which is similar to the proportion of genetic variation across dog breeds (~27 to 33%) (8), but is 7.4 times more than the genetic variation that exists across human continental populations (~5%) (10). Two alternative conceptions of race are colorblindness and constructionism. People who believe in the former contend that racial discrimination is no longer a problem or that it can be ignored because race is not socially important, or real (2). By contrast, constructionism contends that race is a social concept and that racial disparities are caused by prejudice, discrimination, and institutional racism (2, 11). There is consensus in the social and biological sciences that colorblindness, like essentialism, is scientifically flawed (2, 11), whereas there is clear evidence that race is socially constructed and that historical and present-day racism are major causes of racial disparities (1, 2, 11). There is cross-cultural variation in the development of racial conceptualization (1, 2), and the proportion of US children who believe that each race possesses a genetic essence increases with age (1, 2). Thus, exposure to genetic ideas through informal (e.g., media) or formal (e.g., school) learning is hypothesized to affect the development of genetic essentialism (2, 4).

I NS I GHTS | P O L I C Y F O RU M

Complete case analysis of treatment effects for the crossover trial Means and Bonferroni-adjusted 95% confidence intervals from a three-level mixed-effects regression are shown for classes that first received humane genomics instruction followed by basic genetics ( ) or basic genetics followed by humane genomics ( ). In the second half of the crossover trial, classrooms that received humane genomics instruction (relative to basic) increased more in genomics knowledge (x2 = 12.01, p = 0.0005), environmental attributions (x2 = 55.14, p < 0.0001), and belief in social constructionism (x2 = 32.2, p < 0.0001) and decreased more in belief in racial discreteness (x2 = 56.42, p < 0.0001), genetic attributions (x2 = 111.92, p < 0.0001), and belief in genetic essentialism (x2 = 59.73, p < 0.0001). g 1 Mendelian genetics knowledge

Genomic knowledge 2 Genomics

6 1.6

0.5

Logit score

00 0.0 -0.75 –0.5 -1.25 10 –1.0 –11 75 55 –1.5

–1.55

4 1.4

–2.55

1.2

0.0

1.0

–3.55 –4.55

0.8

6 0.6

–0.5

Start

Halfway

End

n 5 Environmental attributions

2 05 –2.0 2.25 –2.5

–5.55 Start

Halfway

End

6 6 Gene Genetic essentialism

Start

Halfway

End

7 Racial acial ccolorblindness

Start

–0.2

9 –0.9

1.7

0.5

–0.4

0 –1.0

1.5

–1.1

–0.5

–0.8

–1.3

–1.0

–1.0

4 –1.4

Start

Halfway

End

Start

End

Start

0.9 Halfway

struction relative to basic genetics instruction (He). To investigate the validity of the results, we examine whether educational effects on genetic essentialism are biased by socially desirable reporting. Then, we answer the exploratory research question (RQ): Do adolescents adopt social constructionist or racially colorblind views after humane genomics instruction? We close by estimating the generalizability, clinical significance, feasibility, replicability, and scalability of the humane genomics approach.

Start

Halfway

End

science.org SCIENCE

,

mechanisms A to C outlined in the section on basic genetics. Then, these classrooms received humane genomics instruction (for 3 weeks) that targeted concepts D to F outlined in the section on humane genomics. The other half of each teacher’s classrooms (n = 26) received humane genomics first and then basic genetics instruction, using identical instructional materials for the same duration. Critically, the humane genomics condition did not explicitly define or use the terms social constructionism, racial colorblindness, or genetic essentialism. Nor did it expose students to the anthropological and sociological arguments in favor of social constructionism. These decisions ensure that treatment effects on racial conceptualization cannot be attributed to teaching to the test. Also, when teachers covered monogenic diseases (e.g., SCA) during basic genetics, they were asked not to discuss racial differences because prior research predicts that such discussions increase belief in genetic essentialism (1). This ensures that any treatment effect on genetic essentialism is not biased by an active comparison condition. We asked students to respond to validated instruments that were administered before instruction began, after instruction concluded during the first half of the crossover trial, and at the end of the study (table S2). We measured students’ (a) basic genetics knowledge, (b) knowledge of genom-

y

EXPERIMENTAL METHODS All study details, data, and code related to our preregistered randomized trial (13) are provided [see supplementary materials (SM)]. Between December 2019 and May 2022, we recruited 15 teachers (n = 14 high school, n = 1 middle school) and 1063 biology students from six US states (Colorado, Illinois, Indiana, Kansas, New Jersey, and Massachusetts; see fig. S1). Participating teachers (see table S1) received 40 hours of professional development to learn how to implement the humane genomics intervention and how to align their Mendelian and molecular genetics curricula with basic genetics (i.e., mechanisms A to C). We randomized classrooms to receive six weeks of instruction in two different orders (figs. S1 and S2). Half of each teacher’s classrooms (n = 25) received basic genetics instruction first (for 3 weeks) that targeted

End

y g

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

Halfway

1.1

y

tion in genetic essentialism. For example, the interventions used in previous studies have been implemented by the teachers who helped to design them. Relatedly, previous RCTs have not assessed whether student self-reports of essentialist beliefs are affected by social desirability bias. Finally, it is still not known whether humane genomics instruction affects belief in genetic essentialism when it is implemented along with basic genetics instruction (i.e., mechanisms A to C) (1). This is a conundrum for educators who are concerned about genetic essentialism and who are obligated to offer a basic genetics education because genetics standards are focused mostly on single-gene inheritance (1). Our cluster-randomized crossover trial overcomes these shortcomings and is the first to explore how basic genetics versus humane genomics instruction affects racial conceptualization. We hypothesize (H) that when adolescents participate in humane genomics instruction, they will grow more in their knowledge of genomics relative to those receiving basic genetics instruction (Ha). If this occurs, then (relative to basic genetics instruction) humane genomics instruction should lead adolescents to perceive less genetic discreteness between racial groups (Hb) and attribute complex human traits less to genes (Hc) and more to the environment (Hd). If so, then belief in genetic essentialism should be lower among students who receive humane genomics in820

1.3

–1.2

g

–0.6

End

S 8 Social constructionism

1.0

0.0

Halfway

p

Logit score

Gene attributions 4 Genetic -0.25

R 3 Racial discreteness

ics, (c) belief in the genetic discreteness of racial groups, (d) genetic attributions for complex human traits, (e) environmental attributions for complex human traits, (f ) belief in racial genetic essentialism, (g) belief in social constructionism, (h) colorblind racial beliefs, and (i) emotional response to instruction. Constructs b to f target the humane genomics hypotheses (Ha to He). Constructs g to h assess the RQ. Construct i is used to assess the feasibility of implementing humane genomics in classrooms, and construct a is used to assess the quality of instruction in the basic genetics condition. Importantly, our experiment can estimate internally valid treatment effects that are well powered, unbiased by missing data, and replicable with different models (tables S3 to S9).

y y g y ,

Scalability and feasibility Issues related to scalability and feasibility must be explored before claims of generalizability are believable. First, given concerns about discussing race in schools, it is important to establish that humane genomics instruction is emotionally safe for students. Second, the amount of teacher professional

g

SCIENCE science.org

Generalizability We conducted a set of exploratory analyses to identify factors that potentially moderated the treatment effect on genetic essentialism, and then used this information in The Generalizer software (14), which uses propensity score matching to make inferences about the population of schools to which our sample of schools generalizes. We also conducted a random-effects metaanalysis to assess between-state variation in the treatment effect on genetic essentialism and assessed the clinical significance of this effect by estimating the number needed to treat. We estimate that the reduction in genetic essentialism caused by humane genomics instruction has high generalizability to high schools in 20 states and medium generalizability to high schools in another 19 states (table S10). These claims are further corroborated by the fact that there was no between-state variation in the treatment effect on genetic essentialism (Q = 7.21, p = 0.21). The estimated number needed to treat is 14 (95% CI [8, 131]). Therefore, in the states where these results are highly generalizable, we predict that two biology students will change their belief in genetic essentialism when a highly trained biology teacher implements our humane genomics curriculum with fidelity in a class of 30 high schoolers.

development it takes to achieve the present results is large. Third, the crossover trial cannot tell us which genomics concepts (i.e., concepts D to F) are most important to learn to reduce essentialism. The humane genomics intervention used in the crossover trial appears to be emotionally safe because it did not cause students to experience greater frustration, anxiety, or confusion compared with the basic genetics instruction. Indeed, students of color reported significantly lower frustration, anxiety, and confusion while learning humane genomics as compared with basic genetics (see SM). There was no self-reported emotional effect for self-identified white students (see SM). To address concerns about scalability, we describe results of an additional preregistered person-randomized trial with n = 1001 undergraduates in the University of California system that used 60-minute online lessons, each targeting different combinations of humane genomics concepts (i.e., concepts D to F). We replicated all of the crossover trial results for Hb to He (table S11), although we did not measure genomics knowledge, racial colorblindness, social constructionism, or emotional responses. This replication study (table S11) suggests that the genetic essentialism result that was observed in our crossover trial is driven more directly by learning about patterns of human genetic variation within and between US census races (concept D) rather than multifactorial inheritance (concepts E and F), and it is not moderated by cultural values that highly correlate with undergraduates’ political ideologies (see tables S12 and S13). These findings suggest that humane genomics instruction can be scaled in a relatively cost-effective, emotionally safe, and time-efficient manner through an online platform that helps students understand patterns of genetic variation within and between US census races.

p

RESULTS To estimate the treatment effect of humane genomics instruction in the first half of the crossover trial, we used a two-level random intercept model with Bayesian multiple imputation for monotone missing data (see SM). To test whether these effects were reproducible in the second half of the crossover trial, we estimated a three-level random effects regression using complete case data of all surveys (see the figure). The results of the first model fully supported each component of the humane genomics hypothesis. Relative to basic genetics, classrooms that received humane genomics instruction had greater knowledge of genomics {b = 0.50, SE = 0.07, t = 7.08, p < 0.001, 95% confidence interval (CI) [0.36, 0.64], R2level-2 = 2.33%; see the figure, panel 2, “halfway”} and less belief in genetic essentialism (b = −0.24, SE = 0.03, t = −6.91, p < 0.001, 95% CI [−0.31, −0.17], R2level-2 = 11.53%; see the figure, panel 6, “halfway”). Humane genomics classrooms also had less belief in racial discreteness (b = −2.57, SE = 0.236, t = −10.89, p < 0.001, 95% CI [−3.04, −2.11], R2level-2 = 35.88%; see the figure, panel 3, “halfway”) and lower genetic attributions for complex human traits (b = −1.39, SE = 0.14, t = −10.17, p < 0.001, 95% CI [−1.65, −1.12], R2level-2 = 49.5%; see the figure, panel 4, “halfway”). Furthermore, humane genomics classrooms had greater environmental attributions (b = 1.28, SE = 0.12, t = 10.60, p < 0.001, 95% CI [1.04, 1.53], R2level-2 = 84%; see the figure, panel 5, “halfway”). All effects were reproduced in the second half of the crossover trial (see the figure). We were careful to avoid the possibility that students in the humane genomics condition were aware that genetic essentialist beliefs are socially unacceptable and thus underreported them (i.e., social desirability bias). First, the curriculum and instruction

involved in the humane genomics condition did not define or use the term genetic essentialism. Second, we estimated the presence and magnitude of social desirability bias on the baseline survey in the crossover trial using a method called a list experiment. This method revealed no evidence of socially desirable reporting on the genetic essentialism instrument (see SM). We then explored whether students gravitated toward racial colorblindness or social constructionism (RQ). Whereas there was no effect of genetics instruction on racial colorblindness (95% CI [−0.06, 0.10]; see the figure, panel 7), there was a positive effect of humane genomics instruction on belief in social constructionism after the first (b = 0.35, SE = 0.06, t = 5.69, p < 0.001, 95% CI [0.23, 0.47], R2level-2 = 9.89%; see the figure, panel 8, “halfway”) and second rounds of instruction (see the figure, panel 8).

DISCUSSION If teachers move beyond Mendel to instruct students about the complexities of contemporary genomics concepts for the purpose of refuting genetic essentialism, they can help students understand that racial disparities are not unreal, unimportant, or the product of genes. Rather, they are socially constructed. Basic genetics instruction, by contrast, yielded none of these benefits to students, even though the teachers in our study produced gains in students’ knowledge of basic genetic concepts (see the figure, panel 1). Previous studies (1) suggest that basic genetics education can increase belief in 23 FEBRUARY 2024 • VOL 383 ISSUE 6685

821

I NS I GHTS | P O L I C Y F O RU M

E

ssentialism is the lay assumption that categories of living things have underlying, unobservable “essences” (1, 2). When applied to sex and gender, this assumption has a range of negative consequences, including stereotyping and discrimination (1, 2). We investigated a potentially powerful—but so far unexamined—sociocultural influence on the development of essentialism about sex and gender: what high school biology textbooks teach adolescents about these topics. We content-analyzed six of the most widely used textbooks in the United States and found that they depart from established scientific knowledge about sex and gender, instead portraying these categories in a manner consistent with essentialism. Three basic assumptions undergird the essentialist view of sex and gender (1): (i) there is little to no variation in traits or behaviors within a sex or gender group; (ii) differences between sexes or genders are discrete—the groups do not overlap substantially in traits; and (iii) internal factors such as genes are the best explanation for all forms of variation within and between sex or gender groups. Scientific research on sex and gender is inconsistent with these assumptions (3, 4), yet they are commonly held. For example, substantial portions of US adults (ø40 to 70%) attribute gender differences in traits and behaviors to genetic causes (5). Theories suggest that essentialism of sex and gender develops through an interaction between general conceptual biases, such as an explanatory bias toward inher-

1

BSCS Science Learning, Colorado Springs, CO, USA. Department of Psychology, New York University, New York, NY, USA. 3Department of Curriculum and Instruction, University of Texas at Austin, Austin, TX, USA. Email: [email protected]; [email protected]; [email protected] 2

SU P P L E M E N TA RY M AT E R I A LS

science.org/doi/10.1126/science.adi7895 10.1126/science.adi7895

science.org SCIENCE

,

This material is based on work supported by the US National Science Foundation (NSF) under grant ECR-1660985 awarded to B.M.D. Any opinions, findings, and conclusions or recommendations expressed in these materials are those of the authors and do not necessarily reflect the views of the NSF. The authors are grateful to N. Rosenberg, A. Cimpian, G. Radick, and R. Wedow for helpful comments on previous versions of this manuscript; to J. Cimpian, R. Wedow, and J. Spybrook for consultation on the randomization method and statistical analysis; to E. Banilower for data on US biology teacher characteristics; and to A. Morning and J. Flanagan for feedback on the humane genomics curriculum. Data and code are available at OSF (13).

By Brian M. Donovan1, Awais Syed1, Sophie H. Arnold2, Dennis Lee1, Monica Weindling1, Molly A. M. Stuhlsatz1, Catherine Riegle-Crumb3, Andrei Cimpian2

y

AC K N OW L E D G M E N TS

Several widely used high school biology texts depart from established science

y g

1. B. M. Donovan, HGG Adv. 3, 100058 (2021). 2. A. Morning, The Nature of Race: How Scientists Think and Teach about Human Difference (Univ. California Press, 2011). 3. J. P. Jackson Jr., D. J. Depew, Darwinism, Democracy, and Race: American Anthropology and Evolutionary Biology in the Twentieth Century (Routledge, 2017). 4. I. Dar-Nimrod, S. J. Heine, Psychol. Bull. 137, 800 (2011). 5. S. Y. Kimel, R. Huesmann, J. R. Kunst, E. Halperin, Pers. Soc. Psychol. Bull. 42, 688 (2016). 6. B. Y. Cheung, A. Schmalor, S. J. Heine, PLOS ONE 16, e0257954 (2021). 7. M. J. Williams, J. L. Eberhardt, J. Pers. Soc. Psychol. 94, 1033 (2008). 8. H. L. Norton, E. E. Quillen, A. W. Bigham, L. N. Pearson, H. Dunsworth, Evolution 12, 17 (2019). 9. B. M. Donovan et al., Sci. Educ. 103, 529 (2019). 10. N. A. Rosenberg, Hum. Biol. 83, 659 (2011). 11. Committee on the Use of Race, Ethnicity, and Ancestry as Population Descriptors in Genomics Research et al., Using Population Descriptors in Genetics and Genomics Research: A New Framework for an Evolving Field (National Academies Press, 2023). 12. T. F. C. Mackay, E. A. Stone, J. F. Ayroles, Nat. Rev. Genet. 10, 565 (2009). 13. B. M. Donovan et al., Moving beyond Mendel. OSF (2023); https://osf.io/qekm3/?view_only=c7000ddf22 2b4fdf82fe183d01d4da83. 14. E. Tipton, K. Miller, The Generalizer (2022); https:// thegeneralizer.org/. 15. C. D. M. Royal et al., Adv. Genet. 2, e10037 (2021).

Sex and gender essentialism in textbooks

y

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

R E F E R E N C ES A N D N OT ES

EDUCATION

g

822

cell disease is controlled by several genes and gene-environment interactions (15). These genomic facts are not discussed in biology textbooks (1), even though such facts offer a more accurate view of SCA (15). We predict that teaching about SCA through an educational framework rooted in genomics will not reinforce essentialist views, and we think this is the most coherent way to link humane genomics instruction with the content of the basic genetics curriculum. Coherent learning experiences that are implemented repeatedly can create enduring changes in how people view the world. Several humane genomics learning experiences spread over many years of biology instruction will be needed to reduce the prevalence of genetic essentialist beliefs. Our study demonstrates that if biology instructors move beyond Mendel and toward a more humane genomics education, they can sow the seeds for a more genetically literate and less racially prejudiced society. j

p

genetic essentialism when students learn about the racial prevalence of monogenic diseases. The fact that we controlled for this factor likely explains why basic genetics instruction did not affect genetic essentialism beliefs. Thus, the reduction in genetic essentialism is attributable only to humane genomics instruction. Importantly, as students begin to disbelieve genetic essentialism, they also appear to gravitate toward constructionism. This is important because our intervention did not teach students the definition of constructionism. Therefore, this result is not a consequence of teaching to the test. Instead, it appears that students decided that constructionism was more plausible than colorblindness after learning about race and genomics (i.e., concepts D to F). We contend that the ideal instructional sequence to reduce genetic essentialism is to introduce students to the models of Mendelian genetics (i.e., mechanisms A to C) and then move beyond these models and highlight their limitations using a humane genomics curriculum (see SM). This prediction is most likely to be accurate when highly trained biology teachers implement this instructional sequence in a high school located in one of the states where our results have high generalizability (table S10). Whether our results generalize to other states or countries cannot be determined with these data. Also, unless educational standards are reformed to include concepts D to F, it will be difficult to reproduce these results at scale because teachers will not have the curricula, professional development, or institutional support to implement humane genomics instruction. Until then, humane genomics instruction will likely need to be delivered to teachers and students online through short lessons to achieve scale. Our results suggest that this will be effective. That said, one limitation in how our study modeled business-as-usual biology instruction is that students were not taught about racial differences in the prevalence of monogenic diseases during basic genetics instruction. Such disparities are a component of the basic genetics curriculum (1). How this information in the basic genetics curriculum interacts with humane genomics instruction is unknown. Even so, the decision to omit race from a discussion of the monogenic diseases is better understood as a strength rather than a flaw. Indeed, diseases such as SCA are incorrectly described and racialized by the basic genetics curriculum (1). For example, SCA occurs across racial groups (15), and although sickle cell trait is caused by variation in a single gene, the severity of sickle

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

823

,

Variation within sex/gender groups People who hold essentialist beliefs tend to believe that individuals within a sex/

y

Sex and gender We initially attempted to make a systematic distinction between terms that pertained to sex (such as “Y chromosome”) and terms that pertained to gender (such as “men”). However, it quickly became clear that this would not be feasible. The information available in the textbooks was not sufficient to make this distinction (see SM). This blurring of the linguistic boundary between sex and gender suggests in and of itself that textbooks may conflate the two. Going forward, we use the term “sex/gender” when appropriate to describe our results. Despite the imprecision in how sex and gender terminology was used, we were able to code whether sex and gender were explicitly differentiated in a paragraph. Of the 362 paragraphs coded, none differentiated between sex and gender in any way. Thus, textbooks inappropriately conflate between a biological phenomenon (sex) and a sociocultural phenomenon (gender).

y g

such as menstruation, which only occurs among egg-producing females. Altogether, nearly all trait variation that exists within and between human sexes is not what essentialism predicts, and neither is the causal source of this variation (that is, there are no genetic “essences”). The same arguments apply to gender, perhaps even more forcefully, because gender is a socially constructed lay interpretation of the biological phenomenon of sex (3, 4). Individuals who identify as women or girls are often expected to adopt a set of socially and culturally prescribed activities, abilities, and interests that distinguish them from individuals who identify as men or boys (3, 4). Thus, differences in complex traits (such as activities, abilities, and interests) between individuals who identify as different genders have no biological basis and are instead explained by sociocultural factors (4). Notably, despite social expectations for distinct gender attributes, complex traits vary substantially and continuously within each gender and have distributions that are highly overlapping across genders (4). Thus, the predictions of essentialism are incorrect about gender as well. Although sex (a biological phenomenon) and gender (a sociocultural phenomenon) are carefully distinguished among biologists who study these phenomena (3), this distinction is often absent in public discourse, where sex and gender are typically conflated. If biology textbooks also

TEXTBOOK ANALYSIS We aimed to identify a sample of high school biology textbooks that are widely used across the United States. A challenge is that textbook manufacturers do not publish statistics on how many schools or students use their textbooks. Thus, we instead used information found on the websites of state and county boards of education to identify textbooks adopted in at least two of the following states: California, Texas, New York, and Florida. Because these states are the most populous in the United States, they strongly influence which textbooks are adopted throughout the United States (13). This process [see supplementary materials (SM)] identified six textbooks published between 2009 and 2016. We estimate that collectively, these textbooks are used by 66% of introductory high school biology classes across the United States (14). Out of a total of 216 chapters across the six textbooks, 10 chapters fit the criteria for inclusion: They were genetics chapters that discussed sex or gender. Next, we identified 362 paragraphs in these 10 chapters that included mention of sex or gender terminology. Our analyses focused on these paragraphs (see the box and SM).

y

SCIENCE science.org

“...essentialism is … a lay view… it should have no place in the biology curriculum.”

conflate the two phenomena, they would be giving voice and lending authority to an uninformed lay view that is out of step with well-established scientific knowledge.

g

THE SCIENCE OF SEX AND GENDER Sexual reproduction generates new allelic combinations within a species (3). Sex determination is the process by which an organism develops a particular sex—the ability to produce a particular type of gamete, along with any associated phenotypic traits. This process is tremendously variable across species. In some species (such as cichlid fish), an individual’s sex can be determined by the temperature of their physical surroundings and can reverse. Some species have more than two sexes (for example, some fungi have thousands); others have more than two sex chromosomes (for example, the platypus has 10) or sex chromosomes other than X and Y (for example, birds have Z and W sex chromosomes). Considerable complexity is present among humans as well. Because so many biological mechanisms are involved in sex determination, biologists now understand human sexes to be complex phenotypes that emerge through interactions between genes, including the sex-determining SRY gene on the Y chromosome and a range of hormone-encoding genes (such as SOX9, SF1, and WNT4), and the environment inside and outside the body (for example, through epigenetic regulation) (3). As a result of this complexity, human sex variation is not strictly dichotomous at the biological level; rather, it is best described as a somewhat continuous, bimodal distribution (3). This biological variation in-

tersects with the cultural practices of medical clinicians to influence sex assignment (3), often in ways that reduce the underlying biological complexity to a simpler binary: females and males. However, many intersex humans exist who blur the hard lines between males and females (3). Further blurring these lines, most phenotypic traits associated with sex categories are the products of complex gene-environment interactions that give rise to vast continuous variability among individuals of the same sex and, conversely, substantial overlap between individuals of different sexes. For example, human sexes have similar patterns of gene expression and heritability for many complex traits (11), including brain morphology (12), physical traits (3), and disease severity (3). Notable exceptions to this pattern of between-sex similarity are traits

p

ent causes (2, 6) and the social and cultural input to which children are exposed (2, 7). We investigated high school biology textbooks as a sociocultural source of essentialist ideas about sex and gender, one that may be particularly important given textbooks’ presumed scientific authority and considering that the vast majority of high school science classrooms in the United States have designated textbooks that are used on a regular basis to guide instruction (8). In addition, studies have linked biology textbooks to the development of essentialism about racial groups (9). Thus, examining whether biology textbooks discuss sex and gender in ways that are consistent with essentialism is a valid means of improving our understanding of how essentialism develops during adolescence. Of course, science textbooks often discuss ideas that were considered accurate earlier in the history of science and are now known to be incomplete or partially inaccurate, such as the Bohr model of the atom. Yet essentialism is not a scientific model (10); it is a lay view (1, 2) that is at odds with the scientific consensus on sex and gender (3, 4). Thus, it should have no place in the biology curriculum.

I NS I GHTS | P O L I C Y F O RU M

Variation between sex/gender groups People who hold essentialist beliefs tend to believe that sex/gender groups are mutually exclusive—that members of one group are entirely dissimilar from those in another group (1, 2). By contrast, the reality is that most traits overlap considerably across sex/gender groups (3, 4). We coded for how often sex/gender groups were described as (i) categorically different or (ii) similar or overlapping (see table S3). Sixteen percent of paragraphs described categorical differences between sex/gender groups [b = 0.16, 95% CI (0.10, 0.22)]. By contrast, only 11% of paragraphs described similarities or overlaps across sex/gender groups [b = 0.11, 95% CI (0.06, 0.16)]. The difference between these code proportions was not statistically significant [b = 0.05, 95% CI (−0.01, 0.11)]. Yet because sex/gender groups overlap considerably on most complex traits (3, 4), even this seemingly balanced presentation of similarities and categorical differences is more consistent with essentialism than with the scientific consensus on sex and gender. An ancillary analysis also revealed that textbooks never mentioned the phenomenon of intersex. By doing so, textbooks again departed from scientific reality and

Internal versus external explanations People who hold essentialist beliefs tend to believe that variation within and between sex/gender groups is the product of a single internal factor (such as a gene or a hormone) or multiple internal factors (such as alleles) that are inherited together through a sex chromosome (1, 2). By contrast, the reality is that most differences between humans, especially in complex traits and behaviors, cannot be reduced to internal factors alone (3, 4). We coded the explanations in the textbooks as (i) internal (appealing to factors originating inside the body, such as genes) versus (ii) external (appealing to factors originating from outside the body, such as the environment) (see table S3). Internal explanations were given in 12% of paragraphs [b = 0.12, 95% CI (0.06, 0.20)]. External explanations were given in only 1% of paragraphs [b = 0.01, 95% CI (0.003, 0.02)]. This difference was statistically significant [b = 0.11, 95% CI (0.05, 0.19)].  In an ancillary analysis, we coded for the number of causes mentioned in an explanation: single versus multiple. We found no difference in the prevalence of singleand multiple-cause explanations (see SM).

y

presented an exaggerated picture of sex/ gender differences, which is consistent with essentialism.

g

individuals within such groups are relatively homogeneous, which is an idea that is consistent with essentialism.  

p

gender group are uniform (no variation) or differ by type (little variation) (1, 2). By contrast, the reality is that most traits are multifactorial and thus vary continuously within sex/gender groups (3, 4). We coded for how often members of the same sex/ gender group were described as (i) uniform (not differing at all), (ii) differing by type (for example, women with blue eyes versus brown eyes), or (iii) differing in a continuous way (see table S3). Twelve percent of paragraphs described individuals of a single sex/gender group as uniform [b = 0.12, 95% CI (0.08, 0.17)] (see SM for analytic strategy). In addition, 10% of paragraphs described individuals of a single sex/gender group as differing by type [b = 0.10, 95% CI (0.05, 0.16)]. By contrast, descriptions of continuous variation within a sex/gender group occurred in only 3% of paragraphs [b = 0.03, 95% CI (0.01, 0.05)]. There was no significant difference between the code proportions indicating uniformity and differences by type [b = 0.02, 95% CI (−0.03, 0.06)]. Both of these codes were more common than the code indicating continuous variation [b = −0.10, 95% CI (−0.15, −0.04), and b = −0.08, 95% CI (−0.14, −0.02), respectively]. These findings indicate that textbooks underemphasize the vast amount of continuous variability within sex/gender groups that has been documented by biological research and instead convey that

Sample textbook passages that illustrate essentialist codes EXAMPLE FROM TEXTBOOK

“You might be surprised to learn that the ability to roll your tongue is an inherited trait. Being able to roll the tongue is likely to be dominant over not being able to roll it. Characteristics such as tongue rolling are called discrete, because they tend to fall into distinct, or discrete, categories. Either you can roll your tongue or you cannot. Either you are male or female. Either you have dimples or you do not.”

1. Between group variation: Discreteness

Characteristics such as tongue rolling vary continuously. For example, some individuals can roll their tongues more than others. Likewise, human sexual development is also somewhat continuous. For example, while genetic males and females are the most common sexes, scientists estimate that at least 2 out of every 10,000 humans are intersex.

“With X-linked traits, males are more likely to be affected. Males only have one copy of the X chromosome. This means that if they have just one copy of the allele for a disorder, they are affected by that disorder. Females, on the other hand, have two copies of the X chromosome. If they have one copy of the allele for the disorder, females are only carriers of the disorder, not affected by it.”

1. Between group variation: Discreteness

Because of selective X inactivation (the process in which one X chromosome in females is silenced), human XX females can carry X-linked disorders without showing any symptoms. However, several other biological mechanisms can still cause female carriers to show symptoms of X-linked disorders. The percentage of female carriers who show symptoms varies dramatically across disorders for this reason, but the symptoms are generally mild. By contrast, X-linked disorders tend to affect a high percentage of human XY males (there is high penetrance) because they do not have a “spare” X chromosome with a functional copy of the affected gene. In human males, the severity of the symptoms of X-linked disorders varies.

y

EXAMPLE OF SCIENTIFICALLY ACCURATE EXPLANATION

,

2. Within group variation: Uniformity 3. Within group variation: Differences by type 4. Explanation type: Internal 5. Number of explanations: Single

These examples are drawn from the textbooks in the research sample. We do not pinpoint the specific textbook that each quote comes from because the textbooks were remarkably similar in the aggregate with respect to how they discussed sex and gender. For a full list of codes, see table S3.

824

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

y g

CODE(S) ASSIGNED

science.org SCIENCE

RE FE REN C ES AN D N OT ES

y g

1. I. Dar-Nimrod, S. J. Heine, Psychol. Bull. 137, 800 (2011). 2. S. A. Gelman, The Essential Child: Origins of Essentialism in Everyday Thought (Oxford Univ. Press, 2003). 3. E. A. Khramtsova, L. K. Davis, B. E. Stranger, Nat. Rev. Genet. 20, 173 (2019). 4. T. E. S. Charlesworth, M. R. Banaji, J. Neurosci. 39, 7228 (2019). 5. E. R. Cole, T. E. Jayaratne, L. A. Cecchi, M. Feldbaum, E. M. Petty, Sex Roles 57, 211 (2007). 6. A. Cimpian, E. Salomon, Behav. Brain Sci. 37, 461 (2014). 7. M. Rhodes, T. M. Mandalaywala, Wiley Interdiscip. Rev. Cogn. Sci. 8, e1437 (2017). 8. E. R. Banilower et al., Report of the 2018 NSSME (Horizon Research, 2018), p. 442. 9. B. M. Donovan, Hum. Genet. Genomics Adv. 3, 1 (2021). 10. B. M. Donovan, M. Stuhlsatz, D. C. Edelson, Z. Buck Bracey, Sci. Educ. 4, 103 (2019). 11. S. M. Smith et al., Nat. Neurosci. 24, 737 (2021). 12. D. Joel et al., Proc. Natl. Acad. Sci. U.S.A. 112, 15468 (2015). 13. M. G. Watt, IARTEM E-Journal 2, 38 (2009). 14. B. M. Donovan et al., High school biology textbooks in the US convey inaccurate, essentialist messages about gender. OSF (2023);https://osf.io/deqj3/?view_only=6 052aedf895c4b8bbeb0f95f1d8d1e68. 15. V. L. Snyder, F. S. Broadway, J. Res. Sci. Teach. 41, 617 (2004).

y

AC KN OW LE DG M E N TS

SU PP L EM E N TARY M AT E RIALS

science.org/doi/10.1126/science.adi1188

10.1126/science.adi1188 23 FEBRUARY 2024 • VOL 383 ISSUE 6685

825

,

This material is based on work supported by the US National Science Foundation (NSF) under grants DRL-1956152, DRL1956119, and DRL-1956167 awarded to B.M.D., C.R.-C., and A.C., respectively. Any opinions, findings, and conclusions or recommendations expressed in these materials are those of the authors and do not necessarily reflect the views of the NSF. The authors are grateful to A. Brubaker and J. Flanagan for their feedback on this research; to S. Gelman, T. Schmader, M. Neiman, and S. A. Taylor for helpful comments on previous versions of this manuscript; and to E. Banilower for help with the generalizability analysis. Several authors are employed at a science education research and development organization (BSCS Science Learning) that has published several widely used high school biology textbooks, including ones that are included in the current research and training samples. All data and code can be found at OSF (14).

y

SCIENCE science.org

We highlight several aspects of current textbooks that could be revised. First, textbooks can define and differentiate the concepts of sex and gender, just as biologists do (3). Then, when discussing sex, textbooks can explain that sex determination in many species (humans included) is complex, involving multiple genetic, hormonal, and environmental factors (3). For example, including the phenomenon of intersex would highlight the complexity of human sex variation. Textbooks can also communicate that there is a tremendous amount of (continuous) variability within individuals of the same sex or gender and that individuals belonging to different sex or gender groups overlap substantially on complex traits. Last, and perhaps most crucially, textbooks can communicate that the traits that are stereotypically associated with a sex or gender group cannot be explained by genes alone—the story is far more complicated. j

g

DISCUSSION When describing sex/gender groups as uniform, or as composed of different types, biology textbooks are expressing essentialist views that are inconsistent with scientific reality: It is continuous variation that is the norm within sex and gender groups. When describing between-group variation, biology textbooks discuss differences and similarities at similar rates. In actuality, sex and gender groups overlap substantially on most complex traits (3, 4). Rather than reflecting this reality, textbooks paint a picture that is consistent with the essentialist notion that sex and gender groups are discrete. When explaining the patterns of variation within and between sex/gender groups, textbooks are far more likely to use internal factors than external factors. On the one hand, this may seem unsurprising: The texts in our sample were communicating the science of genetics, and genes are located inside our bodies. On the other hand, most variation within and between sex and gender groups is not reducible to genes alone (3, 4). For example, most human sex differences involve complex phenotypic traits, which are best explained through multifactorial models that take into account environmental factors and gene-environment interactions (3). In addition, because gender roles are socially constructed, environmental factors rather than genes account for gender differences (for example, activities, abilities, and

interests). Rather than communicate this complexity, the textbooks instead suggested that variation in one or more genes inherited through the sex chromosomes was the only viable explanation for variation within and between sex/gender groups—an essentialist perspective. One limitation of our study is that we did not search for sex and gender terms outside of genetics chapters. We may have thus underidentified messages that are inconsistent with essentialism about sex and gender. However, qualitative studies that have analyzed the nongenetics chapters of biology textbooks by using the lenses of feminist and queer theory—which were developed to uncover and counter gender essentialism—do not support this optimistic view (15). It is noteworthy that the four US states from which we sampled the textbooks differ in political orientation. Textbook content is strongly influenced by cultural and political struggles, especially when it comes to dimensions such as sex and gender. Yet the variability in code frequencies across our sample of textbooks was close to zero (see SM), meaning that our textbooks were strongly isomorphic in how they discussed sex/gender. This reinforces the claim that our statistical inferences are generalizable to the broader population of US biology textbooks. However, our data cannot speak to whether the same patterns would be observed in other countries’ textbooks. Biology education has long been criticized for presenting an oversimplified view of genetic inheritance (9). A salient example of this is the failure to communicate that genetics is more complicated than the inheritance patterns that Gregor Mendel produced using 19th-century pea varieties (9). In fact, Mendelian traits that segregate in an either/or way because of variation in a single gene are the exception to the rule when it comes to the panoply of ways in which humans differ from one another (3). The present results highlight another important way in which biology education falls short and one with implications for young people’s fundamental beliefs about sex and gender. Prior research has found that passages from business-as-usual biology textbooks can reinforce gender essentialism in students (10). More optimistically, the present results also suggest how textbooks could be changed to avoid these undesirable consequences. Because textbook content strongly influences what is taught in science classrooms (8), changing how textbooks describe and explain sex and gender variation could yield a broad decrease in gender essentialist beliefs among US adolescents.

p

Alternative interpretation? We interpreted the evidence above as suggesting that textbooks convey essentialist messages about sex/gender. However, perhaps most of the essentialist descriptions that we identified in our coding (such as within-group uniformity) concerned the small set of traits involved in human sex differentiation, such as reproductive organs. At a somewhat superficial first glance that ignores the biological complexity of human sex differentiation, this subset of traits are relatively uniform within a sex, discrete between sexes, and explainable by internal causes. Our claim that textbooks are conveying unscientific views would be weakened if the descriptions we labeled “essentialist” were applied predominantly to this subset of traits. To investigate this alternative interpretation, we tested whether essentialist (versus anti-essentialist) descriptions were more strongly associated with the traits that are integral to sex differentiation versus other (such as cognitive or behavioral) traits. We found no evidence that this was the case (see SM), which is consistent with our claim that textbooks are conveying essentialist messages about sex and gender.

I NS I GHTS | P O L I C Y F O RU M

EDUCATION

The sociopolitical in human genetics education Education must go beyond only countering essentialist and deterministic views of genetics By R. G. Duncan1, R. Krishnamoorthy2, U. Harms3, M. Haskel-Ittah4, K. Kampourakis5, N. Gericke6, M. Hammann7, M. JimenezAleixandre8, R. H. Nehm9, M. J. Reiss10, A. Yarden4

flawed, or that it cannot generate knowledge that is credible and valuable. Science has always been a sociopolitical and valueladen endeavor; this is a feature, not a bug. Yet this reality in no way legitimizes the arguments of science deniers who misconstrue the sociopolitical nature of science as partisan and biased and use this misrepresentation to dismiss broad scientific consensus. When one ignores the reality that science is entangled with and inseparable from the sociopolitical context in which it is done, one ends up with naïve and problematic views of science that can lead to its misrepresentation as noncredible.

science.org SCIENCE

,

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

y

826

y g

Department of Learning and Teaching, Rutgers University, New Brunswick, NJ, USA. 2College of Education, Pennsylvania State University, State College, PA, USA. 3Department of Biology Education, IPN–Leibniz Institute for Science and Mathematics Education, Kiel, Germany. 4Department of Science Teaching, Weizmann Institute of Science, Rehovot, Israel. 5Department of Biology, Section of Biology and IUFE, University of Geneva, Geneva, Switzerland. 6Centre of Science, Mathematics, Engineering Education Research, Karlstad University, Karlstad, Sweden. 7 Centre for Biology Education, University of Münster, Münster, Germany. 8Department of Applied Learning, Universidade de Santiago de Compostela, Santiago de Compostela, Spain. 9 Department of Ecology and Evolution, Stony Brook University, Stony Brook, NY, USA. 10Department of Curriculum, Pedagogy and Assessment, University College London, London, UK. Email: [email protected]

y

1

g

BEYOND COMPLEX GENETICS The current instructional overemphasis on Mendelian inheritance and the central dogma contribute to problematic views that are deterministic (genes exclusively dictate phenotypes) and essentialist (groups are homogeneous and inherently different from each other). We build on existing efforts in genetics education to shift instruction to more complex and accurate models of gene–environment interactions (multifactorial inheritance) and human trait variation within and across populations (5). Such complex knowledge of genetics, and the explicit countering of race as biological, can lessen genetic deterministic and essentialist beliefs (5). However, though absolutely necessary, a shift to more complex genetics is not enough, because the sociopolitical nature of the environment and how genetic populations are defined remain implicit. This results in incomplete scientific understandings that will not, by themselves, counter scientific racism. Our contention here is that successful genetic education has to be antiracist, it cannot be race-neutral. Therefore, a core learning objective for human genetics education should be understanding that neither the environment nor scientists’ definitions of genetic populations are neutral but rather that they are shaped by the historical, social, and political contexts in which they exist. Below, we unpack the

p

B

iological and genetics research has established that racial categories have no biological or genetic basis. Social sciences have established that race is a socially and politically constructed categorization that undergirds racism, which has devastating consequences for racially oppressed groups. Continued use of race as a proxy for describing human genetic variation is therefore troubling and has received attention in the scientific community and the public arena (1, 2). There is also growing recognition of the sociopolitical nature of human genetics and genomics research itself and the sociopolitical ramifications of its findings (2). To address these problematic aspects of human genetics research and the public’s perceptions of its findings, we call on scientists to openly engage sociopolitical factors in their genetics education efforts. Even for those not involved in equity and justice work, acknowledging the sociopolitical matters for the generation and communication of robust, nuanced, and productive scientific knowledge. The methods of conducting genetics research and its outcomes are steeped in, and influenced by, power and privilege dynamics in broader society. The kinds of questions asked, biological differences sought, and how populations are defined and examined are all informed by the respective dominant culture (often Eurocentric, white, economically privileged, masculine, and heteronormative) and its predominant ways of knowing and being (3). Findings from human genetics and genomics research subsequently play into existing sociopolitical dynamics by providing support for claims about putative differences between groups and the prevalence of particular traits in particular groups (3). Historically, such research has been used

in support of eugenic movements to legitimize forced sterilization and genocides. Yet it would be a mistake to assume that such research is merely a discredited past relic, a stain on the otherwise objective and rational track record of genetic research. Rather, it was mainstream work conducted by prominent researchers and supported by major professional societies. The reality is that some modern human genetics is still informed by the same racist logic (4). A recent report by the National Academies of Sciences, Engineering, and Medicine (NASEM) (1) has urged scientists to scrutinize and justify their use of “race, ethnicity, and genetic ancestry” as population descriptors in genetics and genomics research. That report calls for a shift in scientific practice to engage with the sociopolitical nature of genetics research in terms of how populations are described and how environments are measured. It also calls for more education related to this shift in practice. Education is an important lever in any approach to this problem, and practicing scientists need to be involved, given their profound impact in educating current and future scientists and teachers. Scientists advise on curriculum development initiatives, engage K-12 students through outreach, and, of particular importance, teach prospective scientists and teachers in their courses. By arguing that genetics educators should engage with the “sociopolitical,” we are referring to historical and ongoing power differentials between individuals and groups that result in the distribution, often inequitable, of resources and privilege. Contrary to common views that rigorous science is value-free and impervious to the sociopolitical context, it is well established that science is and has always been affected by the dominant culture and its values. To explicitly acknowledge this reality in science teaching, including genetics education, is imperative; not doing so is misleading and has the potential to backfire. To be clear, just because science is inherently sociopolitical does not mean that it is

sociopolitical nature of the environment and genetic populations, and why it is important to emphasize the sociopolitical.

ILLUSTRATION: BORIS SV/GETTY IMAGES

,

827

y

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

y g

Emphasize the sociopolitical context of the environment In the context of human gene–environment interactions, instruction should address how physical, economic, and social environments can be damaging to people, particularly from minoritized and disadvantaged backgrounds, and are a cause of health dis-

y

SCIENCE science.org

RECOMMENDATIONS FOR EDUCATION Why is it important for students to know about the sociopolitical nature of genetic populations and the environment in multifactorial models of trait variation? Is it not enough to counter genetic essentialism? We offer several reasons. First, if one wishes to dismantle racism (and other systems of oppression) in science and society, then one needs to understand the ways in which such oppression is woven

g

Genetic populations Similarly, how geneticists construe human genetic populations and how these are operationalized in terms of sampling and stratification are also sociopolitical. Genetic distinctions between human populations are not natural; they are the consequences of categorizations developed by geneticists for the purposes of their research and the questions they pursue (3, 8). The search for genetic differences among populations, even when not done using explicit racial categories, can still yield findings that are problematic in that they can make social hierarchies appear “natural”—for example, caste differences in India (9). Yet, genetic studies can also promote notions of genetic unity and similarity across culturally distinct social groups. This was the case with a study by the Indian Genome Variation Consortium, which showed greater genetic

similarities between the rival and culturally distinct Hindu and Muslim groups living in Kashmir compared with their counterparts elsewhere in India (8). Regardless of whether research highlights genetic differences or similarities across social groupings, these endeavors are sociopolitical in that they are situated in, and therefore informed by, historical and ongoing dynamics of social hierarchies. Ruha Benjamin (8) describes other prominent cases that exemplify the non-neutrality of genomics research and argues that these efforts ultimately serve to unify or differentiate the population, often as part of a larger national “branding” process.

p

Environment The environment is not neutral; rather, it is experienced differently by different people. For example, in the US, people of color show higher allostatic load (cumulative burden of chronic stress) compared with white people even when adjusted for age and poverty rates; these differences amount to aging almost a decade faster because of discrimination and ongoing oppression (6). The COVID-19 pandemic brought into sharp focus persistent racial and ethnic inequities in disease risk and outcome. Moreover, what might be considered the “same” environment, such as seeing a police car in your rearview mirror, can trigger different physiological responses for different people, with particularly stressful ones for racially minoritized individuals who often experience discrimination (6). Differences in environments are not benign. Rather, the racial health disparities in the US largely stem from living in hostile environmental conditions both physically and socially. That these environments are oppressive is not happenstance; it is largely by design—through the historical and ongoing efforts of white supremacy (7). Thus, systemic racism is at play in shaping both the environment and the physiological responses to it within individuals and across generations.

into the fabric of genetics research and disrupt and counteract these practices early and often through education. To achieve this aim, it is not enough to counter genetic essentialism. The understanding that race is not genetic (or biological) does not automatically translate into an understanding that race is a social construct, or that it can, and does, shape our biology. Moreover, knowing that race is a social construct does not automatically explain racial disparities in health or any other arena because it ignores the systemic nature of racism and the resulting inequities. Solely countering beliefs in race-based genetic differences and focusing on the similarities between racial groups obscures the real and devastating differences in the well-being of minoritized racial groups. This can lead to racial “color blindness” of a genetic flavor that sees everyone as the same and turns a blind eye to the impact of racism on people’s biology. Second, knowing that the environment interacts with our genes positions the environment as an important consideration but does not problematize it as resulting from socially constructed power hierarchies with real and tangible (physical, social, economic) manifestations that are decidedly not neutral or benign. Sociopolitical awareness allows students to understand that there is no genetic basis for race on the one hand, but that racism has profound biological impacts on the other. Ignoring the political nature of how we define both genetic groups and the environment leads to unnuanced and impoverished understandings of the very phenomena that scientists aim to study. Third, students should come to value their social responsibility toward the moral and ethical ramifications (intended or not) of genetics research that they sponsor and consume. As a society, we underwrite and are responsible for the knowledge we develop and how we use it. Toward these ends, we provide three recommendations for secondary and postsecondary genetics education, grounded in genetics education research (5, 10).

I NS I GHTS | P O L I C Y F O RU M

parities. Moreover, instruction should emphasize that these environments and their detrimental outcomes did not come about by chance but are manifestations of past and ongoing systemic racism and its underlying power and profit motives. Current treatments of the environments in genetics education often ignore the sociopolitical context and position the environment as largely neutral background conditions.

R E F E R E N C ES A N D N OT ES

science.org SCIENCE

,

10.1126/science.adi8227

y

1. National Academies of Sciences, Engineering, and Medicine, “Using population descriptors in genetics and genomics research: A new framework for an evolving field” (National Academies Press, 2023); https://doi. org/10.17226/26902. 2. D. O. Martschenko, M. Smith, Nat. Genet. 53, 255 (2021). 3. J. A. Hamilton, B. Subramaniam, A. Willey, Fem. Stud. 43, 612 (2017). 4. A. C. F. Lewis et al., Science 376, 250 (2022). 5. M. Haskel-Ittah, A. Yarden, Eds., Genetics Education: Current Challenges and Possible Solutions (Springer, 2021). 6. D. R. Williams, J. A. Lawrence, B. A. Davis, Annu. Rev. Public Health 40, 105 (2019). 7. C. Anderson, White Rage: The Unspoken Truth of Our Racial Divide (Bloomsbury, 2017). 8. R. Benjamin, Policy Soc. 28, 341 (2009). 9. Y. Egorova, Genomics Soc. Policy 6, 39 (2010). 10. B. Donovan, R. H. Nehm, Sci. Educ. 29, 1451 (2020). 11. R. Yehuda, Sci. Am. 327, 50 (2022). 12. J. Doucet-Battle, Sweetness in the Blood (Univ. of Minnesota Press, 2021). 13. D. Morales-Doyle, Sci. Educ. 101, 1060 (2017). 14. E. Tan, A. Calabrese Barton, A. Benavides, Sci. Educ. 103, 1011 (2019). 15. J. L. Graves, A. H. Goodman, Racism, Not Race: Answers to Frequently Asked Questions (Columbia Univ. Press, 2022).

y g

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

y

828

g

Scrutinize the sociopolitical categorization of human populations Instruction should encourage students to critically scrutinize how human populations are defined and bounded in genetic studies; they can identify the ways in which such categorization may serve political ends (i.e., who benefits and who loses from these boundaries). Similarly, instruction should foster awareness of the potential ramifications of findings from comparisons of social groups in terms of naturalizing these social categories (e.g., comparisons based on race contribute to making race seem a “natural” biological construct). Research continues to use categorization schemes that unnecessarily racialize traits biologically or genetically; e.g., race is still viewed as a risk factor for type 2 diabetes in ways that can amplify a biological or genetic basis and downplay an environmental one (12). Although this may be done with good intention (to help minoritized communities), it can, and often does, naturalize group differences as biological and promote deficit views of racialized communities and their cultures (12). There are clearly challenges in implementing these recommendations. Both long- and short-term strategies involving multiple leverage points will be needed. In the long term, we see instructional materials as a powerful lever bolstered by a commitment to fund and implement them across education systems. Designing such materials will require the work of interdisciplinary teams of geneticists, science educators, and sociologists who share an equity

studies (STS) courses, and we would welcome requiring such courses for any science or education major. However, we know that knowledge and learning are situated (context and discipline dependent), and, therefore, to be relevant and useful, it is important for ideas to be learned in the discipline of their use rather than in a separate STS course. The second suggestion is to engage students in reflection on the undergirding assumptions and aims of common research approaches in genetics (e.g., genome-wide association studies)—questioning who decides how to define populations, which populations are included or excluded in studies, who benefits from the knowledge generated by these approaches, and who might be harmed by them (intentionally or not). Such questions can help students become more attuned to the intended and unintended, oppressive or liberating, consequences of research. We acknowledge that engaging with the educational recommendations we propose, especially given social polarization, is challenging and potentially risky (but change invariably involves risk and struggle). Yet we believe that engaging with these recommendations is worth the risk. Moreover, this risk can be lessened if those with power, such as professional societies in science, join the struggle with their voices, resources, and backing. The NASEM report (1) has recentered these issues, and we call on other organizations to follow suit. Through a commitment to critical and caring education, we can change minds and structures toward more just futures. j

p

Entangle environment and biology To grasp the biological effects of oppressive environments, students need to understand how the environment is entangled with the mind and body—how is it, exactly, that the environment “gets under our skin”? Instruction, therefore, needs to open up the black box of the mechanisms involved in gene regulation and epigenetic changes in response to environmental conditions. Students should come to understand that, in a sense, and to a certain extent, we can inherit effects of environments, especially traumatic ones, experienced by prior generations (11).

focus. There are powerful exemplars of curricula at the high school level that engage students with ambitious science, its sociopolitical dimensions, and a focus on social justice (13, 14). There is a growing number of excellent books (15) and online resources for anti-racist genetics and biology education—for example, the LabXchange’s “Racism as a Public Health Crisis” curriculum, and the Fred Hutchinson Cancer Center’s materials on “Race, Racism, and Genetics.” These resources include supports for teachers in creating brave and safe spaces for discussions about race and genetics. Funding and committed support of national and professional science and science education organizations will also be instrumental for these efforts. In the short term, we see scientists’ role in the education of future scientists and teachers as one powerful lever for change. Undergraduate coursework in biology and genetics, often taught by faculty in those departments, is a space where we can begin “sowing the seeds” of sociopolitical awareness in genetics. We offer two practical suggestions that science faculty can take to support such learning. The first suggestion is to bring in the historical context of key discoveries in genetics in ways that highlight how the sociopolitical milieu at the time afforded and constrained the theories and methods used and their social consequences. For example, the monumental accomplishment of the Human Genome Project is often taught in genetics courses. What tends to be missing from this account is the broader sociopolitical context that shaped this project and its legacy. The prevailing, and rather essentialist and reductionist (it all boils down to molecules), view at the time of DNA as a blueprint for human capacities, powered this Herculean project. Although this project, and those that followed, showed the overwhelming genetic similarity between people, it nonetheless contributed to a theoretical and methodological “science of difference” and ushered in an ongoing flurry of research efforts to locate socially relevant genetic differences within and between groups. The search for genetic differences has risks and benefits; instruction should address both and draw on emerging best practices in the field (1). In this sense, the Human Genome Project was developed in, and sustained by, a sociopolitical context that upheld (and still upholds) value-laden group differences. This understanding can help students see the inter-relatedness of the historical, social, and political context of genetics science and society. Arguably, some of these ideas are taught in science and technology

By revealing the Higgs boson with detectors such as the Compact Muon Solenoid (shown), scientists proved that the Higgs field exists.

PHYSICS

By Don Lincoln

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

835

,

SCIENCE science.org

10.1126/science.adn9364

y

PHOTO: AFP VIA GETTY IMAGES

A

y g

the subatomic world quite differently. It begins with the Universe being filled with a nalogies are a favorite technique series of fields. Where these fields are quiused to explain complex physics escent, space is relatively empty; however, topics to nonscientist audiences; where these fields experience vibrations, however, they are often imperfect, particles can be found. Indeed, these quanleading to unwelcome misconceptized vibrations are particles. Electrons are tions. Theoretical physiquantized vibrations of electron cist Matt Strassler calls such fields, and quarks—found inside efforts “phibs”—physics fibs. protons and neutrons—are vibraMotivated by a common phib tions of quark fields. used by many scientists (includAccording to modern physics, ing myself ) to explain how the there exists a Higgs field, which Universe-filling Higgs field can interacts with the other fields give mass to subatomic particles, and changes their properties. It Strassler sets out to offer a more is only through the interaction of Waves in an accurate description in his new the Higgs field and, for example, Impossible Sea: book, Waves in an Impossible Sea. the electron field that the familHow Everyday Life His approach is very ambitious, reiar electron can exist. Emerges from the Cosmic Ocean quiring that he find ways to clearly If this all seems confusing, Matt Strassler present one of the most complex that’s because it is. Yet Strassler Basic Books, 2024. and mathematical theories in all explains it all. In a deft series of 384 pp. of science: quantum field theory. chapters, he introduces key conPeople often imagine atoms as miniacepts necessary to understand how quanture solar systems, with electron “planets” tum field theory works and the role the orbiting a nuclear “sun.” Even those with Higgs field plays in our modern world. He some advanced training think of atoms as begins with the principle of relativity and nuclei surrounded by a probabilistic cloud explains the crucial role this principle plays of electrons. Quantum field theory models in our understanding of the laws of nature. He then turns his attention to waves and The reviewer is at Fermi National Accelerator Laboratory, reminds the reader of the features of waves US Department of Energy, Batavia, IL 60510, USA. Email: [email protected] that are critical to quantum theories.

y

A theoretical physicist offers an intuitive primer on the nature of matter and energy

g

Explaining quantum field theory

p

B O OKS et al .

Next, Strassler blends together the ideas of waves and relativity and begins to build an intuitive understanding of quantum field theory. To my knowledge, no other book presents this modern understanding of the nature of matter and energy at a level accessible to the motivated nonscientist. For years, Strassler has written a blog that demystifies complex physics topics (of which I am a longtime fan), and it shows in this book. This is not to say that Waves in an Impossible Sea is an easy read. Although the writing style itself is engaging and conversational, the topics Strassler discusses are not—especially if one is encountering them for the first time. As he himself concedes, it is likely that some passages will require several readings. Like all books, this one has both high and low points. The fact that Strassler has tackled this subject at all is a great boon to the curious nonscientist. A second excellent contribution is that he introduces a topic that I have not seen explained in another comparable book: the distinction between a field and the medium that transmits it. I also intend to adopt some of his colorful descriptions, such as his reference to light as having a “nightmare property.” (If you are chasing it, you can never catch it; and if you are running away from it, you can never escape.) On the downside, there were some places where I thought Strassler belabored a few easier points and breezed through some of the more difficult ones. His discussion of Galilean relativity and introduction to waves could have been trimmed, for example, and the revelation that the Higgs field does more than give mass to fundamental particles, it can generate a very weak force that behaves similarly to gravity over short distances, might have received more attention. I also found Strassler’s frequent use of endnotes terribly distracting. Having to constantly flip between the main text and the end of the book grew tiresome, and I would have much preferred these points to have been set as footnotes or incorporated into the body of the text. However, these are quibbles in an otherwise excellent book. Overall, I found Waves in an Impossible Sea to be a unique and valuable addition to the library of anyone who wants to understand both the modern world of quantum field theory and the role the Higgs field plays in shaping the laws of nature. I recommend it strongly. j

I N SI G H TS | B O O K S

ASSESSMENT

Metrics and their misapplications A pair of anthropologists probe the perils of overquantification and explore ways to rebuild trust By Jonathan Wai

Audit Culture: How Indicators and Rankings Are Reshaping the World Cris Shore and Susan Wright Pluto Press, 2024. 256 pp.

y g y ,

science.org SCIENCE

y

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

g

836

p

administrators are often incentivized to ensure that audits assess how well their institutions meet ideas espoused in mission and s Hungarian neuroscientist Peter vision statements, rather than focusing on Érdi has noted, “Human beings are student education and faculty scholarship. competitive. We want to know who is The authors also detail how the pursuit the strongest, who is the richest, and of being a “world-class” institution often who is the cleverest of all” (1). Howmeans seeking to rise in global university ever, quantification and measurement rankings, a metric that also has bearing tools are often applied inappropriately, as on the evaluation of university leaders. anthropologists Cris Shore and Susan Wright Although I agree with much that the audocument in their new book, Audit Culture, thors write here, there is no mention of how which examines how auditing works at varimany of these metrics align reasonably well ous scales, from the individual to the global, with an institution’s level of scholarship (2) and across various sectors, from health, to or that university rankings have shown coneducation, to the military, and to industry. siderable stability across the As the authors explain in the past century (3). introduction, “This book is a In the book’s final chapter, warning about the dangers Shore and Wright make sugof audit culture and how it is gestions for how auditors and shrinking the space for meanevaluators can reclaim acingful public discussion, reduccountability and trust. These ing public accountability to a include reforming the audit set of bureaucratic templates, industry as a whole but also and hollowing out democracy.” five general guidelines. The Shore and Wright explain first of these—“narratives, not how rankings and metrics just numbers”—emphasizes the have taken over so many doimportance of having metmains, moving into areas that rics serve qualitative evaluare not necessarily amenable ation instead of the reverse. to quantification and those “Encouragement, not punishthat may lose an important ments” urges evaluators to dimension of meaning when provide support for underperranked or measured. Here, the formers rather than assigning authors describe how stateblame. The authors maintain funded schools in England were micromanaged by numthat we must stop thinking of bers and how school rankings organizations and people in Domains such as education suffer when reduced to test scores and rankings. became blunt tools misused only economic terms and that by parents to pick schools for their children. Shore and Wright also dive into how rebuilding trust with those being audited Readers learn the history of accountancy; metrics have come to dominate health care will be critical. And finally, they argue that how relationships formed between auditing and how they have been used to exert manaudits should return to their original purfirms, the government, and other entities; agerial control at a distance. They explore pose as a type of public hearing, rather than and how conflicts of interest have built up the convoluted history of the continuous a practice that uses transparency as a form over time. Shore and Wright argue that “a reform of the UK health system, explaining of control. j major problem with the current model of how the government’s attempts to maintain RE FE REN CES A ND N OT ES auditing is that it generates conflicts of inaccountability and regulation resulted in “a 1. P. Érdi, Ranking: The Unwritten Rules of the Social Game terest and blurs the boundaries between the proliferation of new agencies and systems We All Play (Oxford Univ. Press, 2019). private audit firms and the companies they designed to measure, monitor, rank and 2. D. Wang, A.-L. Barabási, The Science of Science (Cambridge Univ. Press, 2021). are supposed to audit.” manage the different components in an in3. J. Wai, “The undergraduate institutions with the most In chapter 4, the authors ask: “If ‘what is creasingly fragmented landscape.” Nobel Prize winners,” Forbes, 8 October 2020; https:// measured is likely to become what matters’… Many academics can sense that metrics www.forbes.com/sites/jonathanwai/2020/10/08/ the-undergraduate-institutions-with-the-most-nobeland quantification have increasingly taken prize-winners/?sh=3608ff4118cd. over how scholars are hired and evaluated, The reviewer is at the Department of Education Reform but measurement also pervades the growand Department of Psychology, University of Arkansas, ing academic administrative sector. Here, 10.1126/science.adn4674 Fayetteville, AR 72701, USA. Email: [email protected]

A

then those who do the measuring hold a position of exceptional power in society. But whose interests and agendas do these measurements serve?” A focus on the Programme for International Student Assessment (PISA) reveals that although many ranking systems were started by a solo entrepreneur (in this case, German mathematician Andreas Schleicher), they are sustained by soft power generated through networks of advocates. PISA scores, for example, are tied to economic gains, encouraging policy-makers to think of education in terms of return on investment, but “PISA also gains impact through its capacity to name, shame or glorify.”

The water in Iran’s Gorgan Bay flows from the Caspian Sea through a narrow channel.

LET TERS

Edited by Jennifer Sills

RE F ER E NC ES AND NOTES

PHOTO: NASA

Building on Iran’s Gorgan Bay restoration Gorgan Bay, the largest gulf in the Caspian Sea, is a valuable environmental, economic, and social complex. The bay and its contiguous areas were declared a Wetland of International Importance by the Ramsar SCIENCE science.org

1. Ramsar Sites Information Service, “Miankaleh Peninsula, Gorgan Bay and Lapoo-Zaghmarz Ab-bandan” (1975); https://rsis.ramsar.org/ris/36. 2. UNESCO Biosphere Reserves, “Miankaleh Biosphere Reserve, Islamic Republic of Iran” (1976); https:// en.unesco.org/biosphere/aspac/miankaleh. 3. “Gorgan Bay revival should gather pace: VP,” Tehran Times (2022). 4. F. Bakhtiari, “Gorgan Bay at risk of disappearance within a decade,” Tehran Times (2020). 5. Convention on Biological Diversity, “Marine and coastal biodiversity” (2018); https://www.cbd.int/doc/c/c65d/ ea29/94cabf22e3a9ff7828e0ece8/sbstta-22-07-add1en.pdf. 6. BirdLife International, “Important bird areas factsheet: Miankaleh Peninsula and Gorgan Bay” (2020). 7. A. AghaKouchak et al., J. Great Lakes Res. 41, 307 (2015). 8. “Fulfilling the promise of revitalizing Gorgan Bay to preserve the environment, create jobs, and promote tourism,” Iranian Students’ News Agency (2024); https://www.isna.ir/news/1402093021553 [in Farsi]. 9. “45-cm increase in the water level of Gorgan Bay/ reconciliation of migratory birds with the bay,” Iranian Students’ News Agency (2024); https://www.isna.ir/ news/1402102920533 [in Farsi]. 10. J. Liu et al., Chin. Sci. Bull. 66, 1014 (2021). 11. United Nations, “Transforming our world: the 2030 Agenda for Sustainable Development” (UN General Assembly, A/RES/70/1, 2015); https://www.refworld. org/docid/57b6e3e44.html. 12. W. Jenkins et al., BioScience 71, 1157 (2021).

10.1126/science.adn9742 23 FEBRUARY 2024 • VOL 383 ISSUE 6685

837

,

10.1126/science.ado5974

RE FE REN C ES AN D N OT ES

y

1. S. Vaitiekėnas et al., Science 367, eaav3392 (2020). 2. S. Frolov, V. Mourik, “Post-publication review of ‘Flux-tuned topological superconductivity in full-shell nanowires’ Vaitiekenas et al. Science 2020,” Zenodo (2022); https://doi.org/10.5281/zenodo.6344446. 3. S. Guéron, P. Hakonen, A. MacDonald, “Science paper ‘Flux-induced topological superconductivity in full-shell nanowires’, report from expert panel,” Zenodo (2024); https://doi.org/10.5281/zenodo.10647079. 4. H. H. Thorp, Science 373, 500 (2021).

of Civil Engineering, College of Engineering, Australian University-Kuwait, Kuwait City, Kuwait. 2Water, Energy, and Environmental Engineering Research Unit, Faculty of Technology, University of Oulu, Oulu, Finland. 3Department of Geography and Geology, University of Turku, FI-20014 Turun Yliopisto, Turku, Finland. 4Henan Provincial Key Lab of Hydrosphere and Watershed Water Security, North China University of Water Resources and Electric Power, Zhengzhou, Henan 450011, China. 5Regional Climate Group, Department of Earth Sciences, University of Gothenburg, Gothenburg, Sweden. *Corresponding author. Email: [email protected]

y g

Editor-in-Chief, Science

1Department

y

H. Holden Thorp

Masoud Irannezhad1,2*, Amin Sadeqi3, Junguo Liu4, Deliang Chen5

g

On 27 March 2020, Science published the Research Article “Flux-induced topological superconductivity in full-shell nanowires” by S. Vaitiek˙ enas et al. (1). Pursuant to reader concerns about whether the authors selected data for publication that were not representative of the entirety of the data collected in association with the project (2), Science contacted the authors’ institution (University of Copenhagen), which convened an independent expert panel to conduct an investigation. The panel has now released a detailed report (3), and we are alerting readers to this report while we await a formal decision on the matter from the Danish Committee on Research Misconduct. This statement replaces an earlier Editorial Expression of Concern published on 30 July 2021 (4).

p

Editor’s note

Convention in 1975 (1) and designated a United Nations Educational, Scientific and Cultural Organization (UNESCO) Biosphere Reserve in 1976 (2). However, between 2017 and 2021, about 35% of this once-thriving bay dried up (3) as a result of a substantial decline in water level, high sediment accumulation, dense growth of seagrass, closure of canals, decreased precipitation, and increased evapotranspiration (4). Iran, which took action to reverse the trend (4), should now focus on long-term protection of both the bay and the Caspian Sea. A combination of wetlands and dunes, Gorgan Bay serves as the habitat for a wide variety of flora and fauna, including 5 Critically Endangered sturgeon (5); 288 native and migratory bird species, 15 of which are at heightened risk (6); and the Caspian seal, the smallest seal on Earth (5). Gorgan Bay desiccation increases the extinction risk of these species and could lead to higher water salinity and acidity, habitat loss, soil degradation, food chain collapse, and salt storms (4), reminiscent of the catastrophic death of the Aral Sea (7). In 2022, Iran’s Ports and Maritime Organization dredged the 11-km-long Ashuradeh canal connecting Gorgan Bay to the Caspian Sea. On the sides of the canal, sediment traps were installed, and on the dikes, Salicornia, a halophytic vegetable, was cultivated to prevent micro-dust storms (8). Only 4 weeks after the canal reopened in 2023, a volume of about 1.5 billion m3 of water rushed from the Caspian Sea into Gorgan Bay, raising its water level by 40 cm (8). This increase refilled about 400 m of dried shores along the Caspian Sea, revived about 3000 ha of Gorgan Bay’s desiccated area, and decreased the water salinity from 48 to 27 parts per thousand (8). The continued success of Gorgan Bay’s restoration depends on the health of the Caspian Sea. If the sea’s water level continues to decrease, Iran would have to pump

water to the bay instead of relying on gravity (9). To protect the bay long term, Iran should take steps to protect both the bay and the sea, including establishing a comprehensive monitoring system, reforming water governance, developing contingency plans for extreme climate events such as drought or inundation, effectively managing land use and integrated water resources, and enacting legal protections (8, 10). Protecting Gorgan Bay will benefit humans by boosting the economy (8), but Iran must also monitor the environmental effects of revitalized tourism and marine recreation. To align with the 2030 United Nations Sustainable Development Goals (SDGs) (11), the restoration of Gorgan Bay must ensure both ecosystem and human rights to water (12).

I N SI G H TS | L E T T E R S

Ship sturgeon extinction risk in the Caspian Sea

R E F E R E N C ES A N D N OT ES

1. E. J. Hilton, L. Grande, J. Paleontol. 80, 672 (2006). 2. IUCN, “Sturgeon more critically endangered than any other group of species” (2010). 3. V. J. Birstein et al., Sturgeon Biodiversity and Conservation (Springer, 1997). 4. J. Holčik, Ed., The Freshwater Fishes of Europe: General Introduction to Fishes Acipenseriformes (AULA-Verlag GmbH, vol. 1, part II, 1989), pp. 206–226. 5. R. Safari et al., Iran. J. Fish. Sci. 7, 229 (2008). 6. J. Freyhof et al., “Acipenser nudiventris” (The IUCN Red List of Threatened Species, 2022). 7. “Sturgeons,” CITES; https://cites.org/eng/prog/ sturgeon.php. 8. “Implementation of Resolution Conf. 8.9 (Rev.): Acipenseriformes,” CITES (2000), p. 44; https://cites. org/sites/default/files/eng/com/ac/16/16-07-2.pdf. 9. L. Zhang et al., Proc. Natl. Acad. Sci. U.S.A. 120, e2217386120 (2023). 10. R. Stone, Science 309, 1806 (2005). 11. “Sturgeon fishing ban in Caspian Sea extended,” Financial Tribune (2019). 12. Statistical Yearbook of Iranian Fisheries (Iran Fisheries Planning and Budget Bureau, Tehran, 2023). 10.1126/science.adn8556

Moslem Sharifinia Shrimp Research Center, Iranian Fisheries Sciences Research Institute, Agricultural Research, Education and Extension Organization, Bushehr 75169–89177, Iran. Email: [email protected]

y ,

Aquaculture is often presented as a viable alternative to wild fish capture (1). However, current aquaculture practices threaten wild fish populations, marine ecosystems, and the environment. Replacing the fish meal and fish oil used in most aquaculture facilities with feed made from insects would substantially improve the sustainability of farmed fish. Although aquaculture promotes food security and economic growth, it does not benefit wild fish populations (1, 2). Between 1970 and 2014, the growth of aquaculture production did not lead to a reduction in the capture of wild fish (1). By contrast, aquaculture contributes to increased capture by using fish meal and fish oil that contain wild fish. Nearly two pounds of wild fish

y g

Improve aquaculture with insect meal

y

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

Fisheries Department, Faculty of Natural Resources, University of Guilan, Sowmeh Sara, Guilan, Iran. Email: [email protected]

g

838

Bahram Falahatkar

are required for every pound of farmed fish raised on processed fish meal (2). Aquaculture’s demand for wild fish, especially forage species, contributes to increased overfishing and ecosystem depletion (1, 3). Fish farming also displaces and pollutes ecosystems. Aquaculture facilities often replace natural habitats such as wetlands and mangroves, which serve as crucial nurseries for wild fish populations (4). The processes used to produce feed, including transportation, emit greenhouse gases that contribute to climate change (5). Waste such as chemical-laden wastewater can pollute and degrade local marine ecosystems (6). Replacing fish-based aquafeed could improve aquaculture’s sustainability, and insects are the most promising alternative. Compared with plant-based proteins, insects have a high protein content and require fewer resources to produce (7–9). Switching to insect meal would reduce the demand for wild fish, which would ease overfishing pressures and mitigate habitat destruction. Incorporating insect meal could also address the ecosystem degradation and pollution caused by aquaculture. Because of the small size of insects, they can be cultured on a small scale, reducing the need for extensive land usage and thus lessening the impact on natural habitats (7, 10, 11). The process of growing insects and converting them into feed produces fewer greenhouse gases (8, 12) and less nutrient-rich effluent (9) than produced by using fish, which would mitigate climate change and cause less water pollution. If the aquaculture industry transitioned to insect meal, farmed fish could become a sustainable alternative to wild-caught fish, benefiting ecosystems and biodiversity in addition to improving global food security.

p

After 225 million years of existence (1), 85% of the world’s sturgeons are nearing extinction (2). The sturgeon is particularly vulnerable to anthropogenic activities because it is large, moves slowly, migrates between saltwater and freshwater, and reproduces late and infrequently (3). The ship sturgeon (Acipenser nudiventris) distribution used to include the Caspian Sea, Black Sea, Aral Sea, and Azov Sea (4), but there are now just two isolated populations in the Caspian Sea (5). In 2019, the ship sturgeon was listed as Endangered by the International Union for Conservation of Nature (IUCN) (6) and as Critically Endangered by the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES) (7). Iran has taken steps to restore the ship sturgeon population but cannot succeed without regional and international cooperation. The decline of the sturgeon has been driven by overfishing and poaching, compromised quality of and connectivity between habitats and spawning grounds, and pollution (3). Global warming may also pose a threat. After the collapse of the Soviet Union in 1991, decreased regulations and increased poaching devastated sturgeon stocks (3). Migration disruptions have also been threatening the ship sturgeon for decades. Migrations to the Kura River, the species’ main spawning grounds, ranged from 66 to 112 fish between 1983 and 1987 but dropped to only 1 to 4 fish in the late 1980s (8). In addition, the construction of hydropower facilities and dams on other rivers continues, further preventing sturgeon migration (9). Iran has taken steps to restore the ship sturgeon population. The country banned the export of ship sturgeon meat and caviar from the Caspian Sea in 2002 (10) and banned fishing for sturgeon in 2011 (11). Iran has released more than 800,000 juvenile ship sturgeon in the southern basin of the Caspian Sea between 2012 and 2022 (12). However, the reduced connectivity between the rivers and sea in recent years has hindered the fish’s ability to migrate and reproduce, undermining the long-term benefits of efforts to replenish the population. Additional efforts are needed to prevent the ship sturgeon’s extinction. Because sturgeon migrate, they need access to habitats in regions extending beyond Iran’s jurisdiction. The countries sharing the Caspian Sea—Iran, Russia, Azerbaijan, Kazakhstan, and Turkmenistan—should work together to conserve and restore the species.

Coordinated policies should include the establishment and enforcement of fishing and poaching bans, with harsher consequences for catching juvenile and immature fish; better protection of river spawning beds and improved connectivity between the rivers and sea; regulations to mitigate pollution; and an analysis of the effects of global warming on spawning. Captive breeding programs should continue, and juveniles should be released in protected areas. Finally, the five countries should use news media and public education to inform public opinion about the protection of sturgeon in the Caspian Sea.

R E F E R E N C ES A N D N OT ES

1. S. B. Longo et al., Conserv. Bio. 33, 832 (2019). 2. R. L. Naylor et al., Nature 405, 1017 (2000). 3. M. Troell et al., Proc. Natl. Acad. Sci. U.S.A. 111, 13257 (2014). 4. M. McSherry et al., Front. For. Glob. Change 6, 1094306 (2023). 5. D. H. F. Robb et al., “Greenhouse gas emissions from aquaculture: A life cycle assessment of three Asian systems,” Food and Agriculture Organization Fisheries and Aquaculture technical paper no. 609 (2017). 6. S. R. Bush et al., Science 341, 1067 (2013). 7. K. J. Hawkey et al., Annu. Rev. Anim. Biosci. 9, 333 (2021). 8. M. J. Adegbeye et al., J. Clean. Prod. 242, 118319 (2020). 9. A. Van Huis, D. G. Oonincx, Agron. Sustain. Dev. 37, 1 (2017). 10. P. E. Kee et al., Environ. Res. 221, 115284 (2023). 11. F. A. Madau et al., Sustainability 12, 5418 (2020). 12. S. Smetana et al., Anim. Front. 13, 112 (2023). 10.1126/science.ado0380

science.org SCIENCE

RESEARCH IN S CIENCE JOU R NA L S Edited by Michael Funk

ARCTIC ECOLOGY

p

The advancing forest front

SCIENCE science.org

PHOTO: THE WASHINGTON POST VIA GETTY IMAGES

Science p. 849, 10.1126/science.adl6163

ELECTRON MICROSCOPY

Pushing ptychography resolution Aberration correction in electron microscopy is usually performed with complex and expensive lens optics, but ptychography offers an alternative method in which convergent-beam diffraction patterns are collected at different probe positions and used to determine the image partial computationally. Nguyen et al. show that higher resolution

can be achieved by extending data collection to electrons with large momentum scattering and by accounting for the partial coherence of the probe. They demonstrated resolution up to 0.44 angstrom in an uncorrected microscope for twisted tungsten diselenide bilayer samples. —PDS Science p. 865, 10.1126/science.adl2029

EVOLUTION

Title et al. looked at the evolution of the group and used a large number of natural history diet observations to explore their role in ecosystems. They found that a pulse of evolutionary innovation that occurred at the origin of snakes more than 150 million years ago led to an expansion of diet in squamate reptiles, which had cascading impacts on ecosystems that persist today. —SNV Science p. 918, 10.1126/science.adh2449

Snake shift Snakes are important members of today’s ecosystems and, together with lizards, make up one-third of the vertebrate biota. Their iconic status largely comes from their morphology and role as impressive and varied predators.

PREBIOTIC CHEMISTRY

Out of a toxic soup, a molecule of life Pantetheine is an essential chemical moiety within enzymes

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

841

,

The securamine and securine alkaloids, which are biosynthesized by small marine invertebrates, stand out for their dense core of fused heterocycles and both chlorine and bromine substituents. Alexander et al. report chemical syntheses of eight members of this natural product family. Key to the approach is the initial assembly of a macrocycle that undergoes a remarkably selective cascade of internal modifications upon exposure to anhydrous hydrochloric acid. Finely tuned oxidative photocyclizations of

advanced intermediates then deliver the various differentiated final products. —JSY

y

Acid anchors access to securamines

y g

ORGANIC SYNTHESIS

y

The northern extent of the boreal forest, seen here in Gates of the Arctic National Park and Preserve in Alaska, is controlled in part by proximity to open water in the Arctic Ocean.

g

T

he arctic is warming multiple times faster than the global average, and the resulting loss of albedo from melting ice and increasing tree cover is further increasing global climate change. Boreal forest is predicted to shift into the tundra zone as the climate warms, but thus far tree lines have moved northward slowly or not at all. Dial et al. linked tree line advance to declines in Arctic Ocean sea ice. Combining field and remotely sensed measurements from northern Alaska with published data from 60 other Arctic sites, the authors found that distance to sea ice predicted tree growth and recruitment, as well as the probability of forest advance. —BEL Science p. 877, 10.1126/science.adh2339

R ES EA RCH | I N S C I E N C E J O U R NA L S

Science p. 911, 10.1126/science.adk4432

Evidence for a neutron star in SN 1987A

SPECIES RANGE TUMOR IMMUNOLOGY

PolySTING takes out tumors Activation of the DNA-sensing cGAS/STING pathway in tumors has opposing hostprotective or tumor-promoting responses depending on the tissue context. STING is broadly expressed by host cells, making its role in tumor control complex. To investigate the role of immune cell subsets in STING-mediated antitumor immunity, Wang et al. generated a polymeric STING-activating nanoparticle (PolySTING) with a “shock-and-lock” mode of dual STING signaling. Activation of STING in host conventional type 1 dendritic cells was crucial for STING-mediated tumor rejection in mice. The presence of these cells in human tumors was correlated with better patient survival. —HMI

Fire fuels upslope moves

S

pecies are expected to move to cooler areas at higher elevations and latitudes as temperatures warm. Although shifts have been widely documented, many species have not been observed to move upslope. Wilson et al. investigated whether fire disturbance influences the ranges of plant species by killing adult trees and making space for seedlings to establish. By resurveying vegetation plots sampled more than 30 years previously, range shifts were quantified for 42 understory plant species in the Cascade Range of Washington state. Most species did not shift upslope between 1983 and 2015 except for some that experienced fire. Moderate fire may aid these species to shift their ranges, but predicting shifts remains a challenge. —BEL Ecology (2024) 10.1002/ecy.4242

Sci. Immunol. (2024) 10.1126/sciimmunol.adj3945

BIOCHEMISTRY

Recording regulatory activity in live cells

Science p. 890, 10.1126/science.adg0812

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

QUANTUM CHEMISTRY

On the role of error cancellation in DFT Kohn–Sham density functional theory (DFT) is the most widely used electronic structure theory, and through the development of advanced exchange-correlation functionals, has become a powerful practical tool for predicting chemical and material properties for a wide range of systems. It is imperative to understand in detail how various approximations work within this formalism because such an approach may be the key to improving them. Kanungo et al. performed a rigorous analysis of the prototype reaction H2 + F and found that unconventional error cancellation in Hartree–Fock (HF) density functionals leads to an improved reaction barrier height,

which could explain the success of the HF DFT method for chemical reactions. Further analysis of this unconventional error cancellation could be used to improve the existing functionals. —YS J. Phys. Chem. Lett. (2024) 10.1021/acs.jpclett.3c03088

,

Methods that record physiological regulatory activity in cells can provide a more complete view of cellular regulation than what can be gained from individual snapshots at a single time point. Huppertz et al. present a method in which modified proteins can stably record biological activity during a particular time period in which cells are exposed to a fluorescent substrate. The authors altered the HaloTag method so that a tag was split between two proteins of interest, thus monitoring their interaction. They demonstrated the approach by monitoring protein interactions that reflect G protein–coupled receptor activation or calciumdependent protein interactions in neurons. —LBR

y

842

Science p. 898, 10.1126/science.adj5796

y g

Spectroscopy measurements of a supernova remnant indicate the presence of a central neutron star.

Edited by Caroline Ash and Jesse Smith

y

The nearby supernova SN 1987A was visible to the naked eye, and its evolution has been observed over the ensuing decades. The explosion is thought to have produced a neutron star or black hole, but none has been directly detected. Fransson et al. observed a remnant of SN 1987A using near- and mid-infrared integral field spectroscopy. They identified emission lines of ionized argon that appear only near the center of the remnant. Photoionization models show that the line ratios and velocities can be explained by ionizing radiation from a neutron

IN OTHER JOURNALS

g

SUPERNOVAE

star illuminating gas from the inner parts of the exploded star. —KTS

p

and coenzyme A that serves as a handle for reactions involving thioesters, a key species in many biochemical transformations. Because it is strictly conserved and has such a central role in metabolism, pantetheine may have originated alongside other prebiotic components from simple organic and inorganic precursors. Fairchild et al. identified a set of reactions that selectively yield pantetheine in water under dilute conditions that may be reasonable for an early Earth setting. The route uses nitrile chemistry common to other prebiotic syntheses and thus fits into a broader picture for prebiotic cyanosulfidic reactions that are likely candidates for the chemistry that gave rise to life on early Earth. —MAF

CANCER

Not just the virus Epstein-Barr virus (EBV) commonly infects humans and is associated with several diseases. For example, acute infection with EBV causes a self-limited viral illness called mononucleosis. In the long run, the same virus predisposes patients to several types of cancers and other medical conditions. The specific cancer types associated with EBV vary in different parts of the world. science.org SCIENCE

susceptibility of older individuals to tuberculosis. —SHR

By clearing mature vegetation, fire in the Cascade Mountains promotes upslope plant colonization in response to climate change.

Mucosal Immunol. (2024) 10.1016/j.mucimm.2024.01.001

NANOMATERIALS

Giving nanocrystal superlattices the slip

g y

J. Am. Chem. Soc. (2024) 10.1021/jacs.3c10706

p

The ordering of metal nanocrystals (NCs) in dried self-assembled superlattices has been improved by using surface ligands that mimic thermotropic liquid crystals. Ning et al. designed dendritic ligands with chain ends functionalized with biphenyl groups. A superlattice of gold NCs bearing these ligands on a silicon substrate became more ordered after thermal annealing because the NCs could move and reorient. Molecular dynamics simulations suggested that these mesogenic ligands slide past each other and act as a lubricant, as opposed to locking through interdigitation and entanglement. —PDS

WEARABLE DEVICES

Motion capture in hand

Fixing failing cilia in CF The clearance of mucus from the airways involves the repetitive beating of millions of cilia SCIENCE science.org

Proc. Natl. Acad. Sci. U.S.A. (2024) 10.1073/pnas.2313089121

Not aged to perfection Alveolar lining fluid (ALF) is located at the interface between lung tissue and the external environment. Age-associated changes to the soluble components of ALF may impair immune defense mechanisms and favor infection and disease by lung pathogens. Olmo-Fontánez et al. subjected Mycobacterium tuberculosis to ALF taken from either elderly individuals or nonelderly adults. M. tuberculosis that had been exposed to ALF from older people replicated better within a lung epithelial cell line in vitro. Correlating with this, the infected epithelial cells produced fewer proinflammatory immune mediators. Alterations to ALF occurring with aging might be another factor contributing to the

Nat. Mach. Intell. (2024) 10.1038/s42256-023-00780-9

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

843

,

CYSTIC FIBROSIS

HOST DEFENSE

y

Cell Genom. (2024) 10.1016/j.xgen.2023.100474

on the surface of respiratory tract cells. In cystic fibrosis (CF), mucus clearance by cilia is compromised because mucus composition is altered and airway surfaces are dehydrated. Briottet et al. used live-cell microscopy of primary cultures of nasal epithelial cells from CF patients to assess the frequency, synchronization, orientation, and quantity of cilia beating. The authors also measured the height of the airway surface liquid layer and secreted mucins. The defects in ciliary beating in CF could be ameliorated by application of a lipid mediator, resolvin E1. A similar approach could also be useful for studying motile cilia in other diseases. —SMH

y g

Knowing that nasopharyngeal carcinoma is specifically linked to EBV in southern China, Xu et al. analyzed the genetic makeup of circulating EBV subtypes, as well as local human leukocyte antigens that determine how the virus is presented to the immune system. The authors found a strong interaction between EBV subtype 163364 and two host single-nucleotide polymorphisms in HLA that make the carrier vulnerable to infection with this EBV subtype. —YN

The need for accurate motion tracking spans a range of applications from remote surgery to gaming. High-end applications use markers and motion-capture cameras, but significant progress is also being made in wearable technology. Tashakori et al. developed a glove using embedded helical sensor yarns composed of electrospun polymer fibers wrapped with metal-coated nanofibers (NFs) and protected by a polydimethylsiloxane layer that also forms a matrix around them. Pressure or stretching changes the contact area of the NFs, generating measurable responses, whereas the helical structure ensures a linear response over a wide range of strains and stresses. Using machine learning algorithms, the gloves showed high accuracy in measuring changes in joint angles and in real-time tracking of hand motions. —MSL

R ES E ARCH

ALSO IN SCIENCE JOURNALS MEDICINE

The unmet challenges of Long Covid

Science p. 844, 10.1126/science.adf9621

NEPHROLOGY

Better recovery without SOX

OPTICS

HLAs and lung cancer risk Human leukocyte antigens (HLAs) are cell surface molecules that recognize antigenic peptides and present them to T cells to initiate an immune response. Krishna et al. performed an immunogenetic analysis of associations between HLA class II loci and lung cancer risk. They analyzed two large population-level datasets, one from the UK Biobank and the other from Finland’s FinnGen. Heterozygosity at HLA class II loci was found to be associated with reduced risk of developing lung cancer. Individuals who were current or former smokers benefited from this protective effect, but individuals who had never smoked did not. Smokers with lung cancer were found to have somatic loss of HLA class II heterozygosity, suggesting a functional role for HLA-related immune surveillance in cancer risk. —PNK

Science p. 848, 10.1126/science.adj2609

SOLAR CELLS

A better buried interface In all-perovskite tandem solar cells, a lead-tin perovskite is used instead of silicon as the narrow band gap cell, but nonuniform nucleation and fast crystallization has limited film quality and device efficiency. Gao et al. now show that aminoacetamide hydrochloride can strongly coordinate the precursor components in solution, which homogenizes the crystallization process and also passivates the buried perovskite interface (see the Perspective by Chen and Zhao). The authors achieved a certified power conversion efficiency of 24.5% for a 20-square-centimeter module made by blade-coating the layers. —PDS Science p. 855, 10.1126/science.adj6088; see also p. 827, 10.1126/science.adn7930

Science p. 847, 10.1126/science.adi3808

Insects’ sweet sensations

Increasing precision of array computing

,

In non-Abelian (non-commutative) systems, the final state is dependent on the exact sequence of steps performed getting there. Research on nonAbelian phenomena has recently undergone rapid developments in photonics and acoustics, introducing a new perspective on controlling light and sound at the fundamental level. Such manipulation also provides a versatile platform for studying non-Abelian physics more generally and exploiting the effect for

IN-MEMORY COMPUTING

y

Not what you do but how you do it

843-B

IMMUNOGENETICS

y g

Science p. 845, 10.1126/science.add6371

Science p. 846, 10.1126/science.adj7026

to either monosaccharides or disaccharides, as confirmed by accompanying electrophysiology measurements. The structures revealed a tetrameric ion channel with a sugar-binding site near the extracellular face in a deep, hydrophilic pocket. A mutant with constitutive activity enabled visualization of an open state of one of the channels with a wide pore consistent with nonselective cation conduction. —MAF

y

Acute kidney injury from a variety of causes can be temporary and followed by tissue recovery or it can result in long-term damage, fibrosis, and eventual progression to chronic kidney disease. It was not known what determines the outcome of this process, but some previous work implicated a transcription factor called SOX9 as a possible culprit. Using mouse models of acute kidney injury, Aggarwal et al. identified a SOX9 switch, showing that regenerating kidney cells transiently activate SOX9, but then some of these cells deactivate it and some do not. This subsequent deactivation, or lack thereof, with resulting fibroproliferative effects through the WNT signaling pathway, is what distinguishes cells making a healthy recovery from those undergoing fibrosis and longterm injury. —YN

of PRDM9. They found that crossover events occurred at both PRDM9-determined hotspots as well as promoter-like regions. This suggests that multiple processes, rather than the presence of PRDM9 alone, underlie the recombination landscape in a species. —CNS

g

Science p. 828, 10.1126/science.adl0867

possible applications. Yang et al. summarize the fundamentals, highlight recent advances, and discuss the future directions of non-Abelian physics in photonics and acoustics. — ISO

p

Chronic postinfection symptoms after COVID-19 illness, which is referred to as Long Covid, are widespread, affecting people of all ages and having unknown impacts on long-term health. Understanding and treating the condition is complicated by the multisystemic and diverse symptoms that can arise. In a Perspective, Al-Aly and Topol identified key issues that need to be prioritized. It is crucial to understand the mechanisms of Long Covid and the wide-reaching effects that it has on the risk of developing other chronic diseases. Although preventive measures such as vaccination and antivirals could be an effective strategy to reduce the incidence of Long Covid, therapeutics are also needed. There has been slow progression in undertaking randomized controlled trials of repurposed and newly developed drugs that could improve patient health. Long Covid also highlights the need to understand postinfectious conditions so that the needs of people with such chronic illnesses can finally be met. —GKA

Edited by Michael Funk

The ability to detect food at a molecular level is crucial for identifying high-quality sources of nutrition. Humans use canonical G protein–coupled receptors for taste, but insects are known to have distantly related gustatory receptors that function as ligand-gated ion channels. Ma et al. determined the cryo–electron microscopy structures of two gustatory receptors from the fruit fly Drosophila that bind and respond

In-memory computing can execute a large vector-matrix multiplication (VMM) within one computing cycle, but the relatively low precision of these analog methods has precluded its use with conventional computing. Song et al. show that for one-memristor-one-transistor arrays, most of the VMM can be performed to arbitrarily high precision before being output as a digital result (see the Perspective by Aimone and

RECOMBINATION

Reexamining PRDM9’s role in corn snakes Where meiotic recombination occurs in the genome varies across species. In humans and some other mammals, binding sites of the protein PRDM9 determine recombination hotspot locations, whereas in other species without PRDM9, these tend to occur in CpG islands and near promoters. Hoge et al. identified historic patterns of linkage disequilibrium and crossover events from pedigrees in the corn snake, which has a single intact ortholog

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

RECEPTORS

science.org SCIENCE

Agarwal). After a computation step by an array, subarrays dynamically compensate for residual errors of the previously programmed array. This method was used experimentally to solve partial differential equations with remarkable precision ( 3 (see Fig. 2B, right), the non-Abelian topological charges turn to be generalized quaternion groups (13). Non-Abelian topological charges have been used to study the geometry, topology, and physics of defects from a homotopy perspective. In material science, real-space non-Abelian topo-

23 February 2024

logical charges have been elegantly applied to describe the disclination line defects in biaxial nematic liquid crystals (13, 82, 84). They are assemblies of brick-like (see the inset of Fig. 2D) molecules that self-organize to form mesophases (82). The brick has three different principal axes related to height, width, and length, and the three axes (red, green, and blue) define a frame that indicates the orientations of the molecule 4 of 14

,

Fig. 2. Theoretical aspects of non-Abelian topological charges. (A) Abelian topological classification of 1D topological bands with one single bandgap (13). (B) Non-Abelian topological classification of 1D topological bands with two or more bandgaps (13). (C and D) 2D extended band structure and the corresponding eigenstate frame rotations for Abelian and non-Abelian topological charges. The position and type of band degeneracies in the extended 2D systems can predict the Abelian and non-Abelian topological edge states of the 1D

RES EARCH | R E V I E W

,

5 of 14

y

We now address non-Abelian synthetic gauge fields in light and sound, primarily focusing on synthetic non-Abelian magnetic fields that are associated with the vector potentials; we will briefly discuss synthetic non-Abelian electric fields toward the end of this section. The recent development of this topic has drawn inspiration from other platforms, in particular cold atomic systems, for which we would like to draw readers’ attention to the relevant

y g

Non-Abelian gauge fields

y

23 February 2024

tems (68), which were found to support zerorefractive-index propagation in photonics (90). Further evolution of the nodal-link structure leads to PT-symmetric triple degeneracy in three dimensions (74, 75), as shown in Fig. 2F. The underlying topology of the triple degeneracy can be well described by the Euler number (74, 75, 91–93), for example, defined by integrating Euler curvature on the sphere surrounding the triple point, which leads to a universal higher-order bulk-boundary correspondence (94). Actually, there are two kinds of PT-symmetric triple degeneracies (75, 95); the other triple points, so-called nexus triple point degeneracies, cannot be captured by the Euler number. Researchers have proposed the possibility of exploring new topological charges beyond the paradigmatic Chern number, that is, “punctured-Chern” invariant (96, 97). Recently, the second Euler number in 4D synthetic matters has also been discussed (98). Non-Abelian topological charges demonstrate the noncommuting properties that may enable different ways to manipulate wave packets and inspire applications in information transmission and processing. Owing to the versatile manipulation of different degrees of freedom in acoustics and photonics, these theoretical models have found extensive experimental realization, as shown in Fig. 3, A to D. The path-dependent node collision of Dirac points has been observed in acoustic systems using Kagome lattice formed by cylindrical resonators (Fig. 3A) and ideal metamaterials with three variable geometry parameters (Fig. 3B). The non-Abelian nodal links have been experimentally demonstrated in acoustic crystals (68) and photonic biaxial hyperbolic metamaterials (Fig. 3C), which illustrates the topological constraints imposed by non-Abelian charges. The photonic biaxial hyperbolic metamaterials offer a natural platform for implementing the three-band continuum models directly deriving from Maxwell’s equations. (13, 15). The three-band system serves as the minimal non-Abelian topological model, on which artificial sublattices were designed and mapped to the non-Abelian topological charges by rotating the eigenstate frame (Fig. 3D) (14). Interplaying with other crystalline symmetry, photonic, acoustic, and cold-atomic setups will further fuel the excitement in this research direction.

g

Yang et al., Science 383, eadf9621 (2024)

contrast to the Abelian bulk-edge correspondence that applies to an individual bandgap, the non-Abelian one predicts both the position and number of the edge states for multiple bandgaps. In 3D crystals, a nodal line is a 1D degeneracy curve in momentum space that arises because of a gap closing in the eigenvalue spectrum. When considering multiple bands, non-Abelian topological charges can be used to characterize the topological link structure of the multiple nodal-line systems, with an example shown in Fig. 2E, left. The crossings between adjacent bands form the linked nodal lines threading through each other. As shown by Wu et al. (13), non-Abelian charges can be used to explain various topological constraints on the nodallink configurations. For example, in Fig. 2E: A pair of nodal lines of different colors cannot move across each other, a closed nodal ring can only encircle an even number of nodal lines of the other color, and nodes formed by consecutive pairs of bands anticommute, whereas all nodes formed by more distant pairs of bands commute. Another feature is that the sign of the charge assigns an orientation to the nodal lines (see Fig. 2E, left), and the sign of a nodal line is reversed each time it goes under a nodal line of the adjacent bands. This is due to the nontrivial braiding rules arising from the noncommutativity of the quaternion charge (13). The local band structure near the nodal-line degeneracy is topologically equivalent to a 2D Dirac cone. Figure 2E, right, shows the 2D band structure on the cutting plane at a fixed k3, where three Dirac points can be clearly seen (two are in the first gap, and one is in the second gap). The non-Abelian topological charges of nodal links relate to how they are braided together and determine the outcome of the pathdependent node collision process (13, 16, 60–62). The two Dirac points formed between the first and second bands cannot be removed along the path L2 as shown in Fig. 2E, left, because the non-Abelian topological charge is C(L2) = (+k) × (+k) = −1. Conversely, the charge of L1 loop in Fig. 2E, left, is C(L1) = (+k) × (−k) = 1, indicating trivial topology; the loop can hence be shrunk to a point, and thus the two lower-band Dirac points can annihilate by bringing them together along a path enclosed by L1. For a system consisting of three bands, the annihilation of two Dirac points formed between the two lower bands depends on the positions of Dirac points between two higher bands. The above statement also indicates that two nodal ring linking is not allowed (13). More generally, on a plane protected by C2T symmetry, the arguments of 2D Dirac points can be applied to Weyl points as well. Therefore, one can also braid Weyl points on the C2T invariant plane (60). It is worth mentioning that the non-Abelian topological charges are also related to triple degeneracies in 2D sys-

p

(13, 14). Let us consider a loop enclosing a defect. The molecule frame rotates along the loop, and the frame must return to itself after going around the loop. The frame rotation can be p or 2p. The charge of +i corresponds to the frame rotation of p around the red axis (13, 14). Similarly, the other two non-Abelian topological charges +j and +k indicate the frame rotation of p around the green and blue axes, respectively. Therefore, the non-Abelian topological charges are also called noncommutative “frame-rotation charges,” whose underlying topology can be explained using Dirac’s belt trick (85). As mentioned earlier, topological defects appear not only in nematics but also in topological bands. Here, the molecules (or frames) in liquid crystals can be directly mapped onto the eigenstate frames of the topological bands in three-band PT-symmetric systems. The topological structure of the 1D bands (Fig. 2, A and B) are clearer after extending the 1D Hamiltonian H(k) onto a 2D plane (14). Figure 2, C and D, shows the extended 2D band structure. The original Hamiltonian H(k) is exactly located on the unit circle k12 + k22 = 1 (see white or gray circles in Fig. 2, C and D) of the extended Hamiltonian. Each nontrivial topological charge represents a band degeneracy in the 2D system, which exhibits Dirac-cone dispersion in the range k12 + k22 < 1 (Fig. 2, C and D). They represent an obstruction that cannot be removed unless topological phase transition happens with the degeneracy point moving out of the unit circle corresponding to bandgap closing and reopening. For Abelian topological descriptions, we consider only one bandgap. The 2D Hamiltonian carries one Dirac point, that is, the linear intersection of the two bands in Fig. 2C. But for multiple bandgap descriptions, different non-Abelian topological charges (Fig. 2D) correspond to the band degeneracies (Dirac points) that appear in different bandgaps accordingly. As such, frame rotations and topological charges of 1D Hamiltonians are closely related to band degeneracies in a high-dimensional extended Hamiltonian. For 1D systems, the formation of edge states can be visualized from the 2D extended plane, where the edge states of the 1D system are inherently related to the topological degeneracy points encircled by the 1D Hamiltonian, that is, the existence of enclosed Dirac points directly predicts the appearance of edge states. It is so-called bulk-edge correspondence (86–89), which itself is an elegant theory that connects the properties of an infinite periodic system and those of an exposed edge of a truncated bulk, for example, each Dirac point implies one nontrivial edge state. From this viewpoint, the bulk-edge correspondence of non-Abelian topological charges can be explained heuristically, for example, by counting the number of Dirac points on the 2D extended plane (14). In

RES EARCH | R E V I E W

A

B

D

C

t=2 Unitcell 1

L

(t)

Lg(t)

L f(t)

2 3

z

y

y

x

x

t=8 88

z kH

5.

1

z

B(B′)

kz 5.

59

z kH

A′

A

kx

ky

x

1

ky 1

kx

p

1 1

y

,

6 of 14

y

23 February 2024

fields. Another criterion is based on the concept of the real-space Wilson loopW ≡ TrðW Þ that is a gauge-invariant quantity; it is required that the modulus of Wilson loop should differ from the dimensionality of the Hilbert space for the gauge fields to be non-Abelian. This criterion works well for many circumstances but has its caveat—a system in Abelian gauge fields such as a single eiqsz term results in decoupled spins; however, its Wilson loop still satisfies the non-Abelian condition jW j ≠ 2. The non-Abelian Aharonov-Bohm interference is particularly useful for experimentally detecting non-Abelian gauge fields and the resulting geometric phases. The original Aharonov-Bohm effect highlights the measurability of the gauge potential experienced by a spinless particle even though it never enters the region filled with the magnetic field. The generalized non-Abelian effect describes a spinful particle moving along two paths, where the path integrals on the inhomogeneous gauge fields are reversed to each other. The two paths form a closed loop, enabling interference measurements of the spin population that reflect the noncommutativity of the underlying gauge fields. This effect has stimulated long-standing theoretical interests on various physical platforms (18, 25, 99, 102–108) and has been realized using nonreciprocal fiber optics and electric circuits (20, 26) (Fig. 4C). In the experiments, projection measurements were performed along a certain basis to extract their associated population contrast at the interference. Because

y g

Yang et al., Science 383, eadf9621 (2024)

matrix-valued link variables between positions ′ r and r′: L ≡ P expi∫rr A  dl, where P denotes path-ordered integral. A real-space loop operator can thus be defined as W ≡ PP□ L, that is, a path-ordered product of link variables around the loop. For a square-lattice plaquette, a counterclockwise loop operator beginning from its bottom left corner can be explicitly expressed as W ¼ L4 1 L3 1 L2 L1 (Fig. 4B). The real-space curvature, that is, magnetic field, can thus be defined by B ≡ i logW for this plaquette. In the continuum limit, B reduces to B = ∇ × A – iA × A, an exact real-space counterpart to the multiband Berry curvature in momentum space, which is introduced at the beginning of this review. This loop operator enables several criteria for identifying non-Abelian gauge fields (100). Rigorously, the definition of non-Abelian gauge fields relies on examining the commutativity among different loop operators. It requires the existence of two different loop operators W and W′ that satisfy the noncommutativity condition WW′ ≠ W′W. Such a noncommutativity criterion has been applied to analytically examine the genuine non-Abelian conditions in non-Abelian Hofstadter models (101). Although introduced in a lattice context, the criterion above works equally well for continuum systems where the loop operators can be replaced with curvatures. In addition, there are loosely defined criteria for non-Abelian gauge fields. Broadly, a matrixvalued A or its noncommutative components could be referred to as non-Abelian gauge

y

reviews, for example, (99, 100), to facilitate a deeper understanding of the discussions here. We illustrate the basic concept of synthetic gauge fields using the picture of a particle hopping among lattices, where the particle picks up a phase q º∫A  dr during the hopping, where A is the real-space gauge fields. U(1) gauge fields couple to spinless particles whose Hilbert space forms a unit circle (Fig. 4A, left). Consequently, all geometric phases accumulated during the hopping are Abelian because all the rotation around the unit circle is commutative. By contrast, non-Abelian gauge fields couple to particles of internal degrees of freedom situated in an enlarged Hilbert space, for example, a unit sphere for SU(2) gauge fields (Fig. 4A, right). Because rotations around different axes of the sphere are noncommutative, the gauge fields and their geometric phases become non-Abelian. Crucially, synthetic gauge fields couple to the momentum that is associated with a sign flip of geometric phases when particles propagate toward opposite directions (Fig. 4A, bottom)—it is a requirement for maintaining Hermiticity, which is also the key to the time-reversal symmetry breaking for the U(1) gauge fields. This sign flip also distinguishes synthetic gauge fields from other related types of Hilbert space operations, such as singlequbit gates that typically suffice along a single direction. Similar to their Abelian counterparts, closed loops are needed to define real-space curvature, that is, the magnetic field. To see this, we define

g

Fig. 3. Acoustic and photonic realizations of non-Abelian topological charges. (A) Cylindrical acoustic resonators forming three-band kagome lattices and mapped acoustic band structures (16). (B) Unit-cell structure of acoustic metamaterial and experimental characterization of the collision stability of nodes (61). (C) The metallic metamaterials and experimentally probed non-Abelian nodal links (15). (D) Transmission line network and experimentally mapped eigenstate frame spheres (14). [(A) is adapted with permission from (16); (B) is adapted from (61); (C) is reprinted with permission from (15) Copyright 2020 by the American Physical Society; (D) is adapted with permission from (14)].

RES EARCH | R E V I E W

A

Forming a plaquette

B

eiθσ z

e iθ e

C

90

-1

L3 L4-1

e-iθσz

-iθ

non-Abelian Aharonov-Bohm interference

Sawtooth phase modulation

L2

Faraday rotator

L1

z

-k

k

current input

- WW’ ≠ W’ W - Matrix W - |Tr(W)| ≠ Dim(W)

Latitude α (deg)

k

-k

-45

90

-1

Non-Abelian criteria

-θ θ

0

0 + contrast in dB (a.u.)

W = L3 L4 L2L1 -θ θ

45

-90

light input and interference

Loop operator -1

Latitude α (deg)

Hopping phases coupled to momenta

voltage interference

45 0 -45 -90 -180

-90

0

90

180

System: Homogeneous media and planar cavity

E

F

Single defect

G link-engineered lattice models

Hexagonal photonic lattices

(m,n+1)

x

x

y

x

D

V

ky

kx

(m+1) σ z

A

(m,n)

(m+1,n)

H

Non-Abelian Hofstadter

Quasimomentum

7 of 14

,

However, observing ZB directly in relativistic electrons is highly improbable owing to the minuscule amplitude of the trembling motion and the extremely high oscillation frequency (on the order of 1021 rad/s). Therefore, ZB has been generalized into a universal wave phenomenon over the past century: It describes a trembling motion resulting from wave interference of quasi-degenerate modes, and alternative systems have been proposed to simulate this effect

y

couplint (19, 22). (F) Polaritonic graphene (top left) features Drssselhaus-type non-Abelian gauge fields near the K and K′ points, causing pseudospin switching on opposite sides of the degeneracy point (see color changes of the bands at the bottom left) and the dipolar magnetic field texture (top right: theory; bottom right: experiment) (24). (G) Proposal for realizing a non-Abelian Hofstadter model in the Landau gauge (top) using averaged photon transmission spectroscopy (bottom) (47, 101). [(C) is adapted with permission from (20, 26), respectively; (D) is adapted from (18, 23); (D) and (E) are adapted with permission from (22) © Optical Society of America; (E) is reprinted with permission from (19) Copyright 2014 by the American Physical Society; (F) is adapted with permission from (24); (G) is adapted with permission from (101) and reprinted with permission from (47) Copyright 2023 by the American Physical Society].

where A can be spatially and temporally dependent. Below, we present a summary of phenomena under this umbrella, categorized based on the degree of symmetry present in the systems, which spans from homogeneous media to lattice models with intentionally engineered link variables. Zitterbewegung (ZB) was initially proposed as a quantum mechanical interference between the solutions to the Dirac equation in the 1930s.

23 February 2024

m σz

y g

Fig. 4. Non-Abelian gauge fields. (A) Lattice illustration of U(1) Abelian gauge fields (top left) and SU(2) non-Abelian gauge fields (top right; yellow arrows represent a spinful internal degree of freedom). They lead to momentumdependent rotation in the Hilbert space, that is, a unit circle (bottom left) and a unit sphere (bottom right), respectively. The former is always commutative, whereas the latter can be noncommutative around different axes. (B) NonAbelian gauge flux and non-Abelian criteria are defined from a matrix-valued loop operator. Dim, dimension; Tr, trace. (C) Non-Abelian Aharonov–Bohm interference to examine gauge-field commutativity in fiber optics and electric circuits (20, 26). (D) Zitterbewegung analogs can be realized by coherent superposition of eigenstates of a homogeneous anisotropic non-Abelian medium and planar quantum-well cavities (18, 22, 23, 122). (E) Wave focusing around an impenetrable defect (white circles) embedded in a non-Abelian medium of Rashba spin-orbit

Yang et al., Science 383, eadf9621 (2024)

E

x

the input spin can be accurately controlled by optical and electrical means, the contrast can be measured around the entire surface of the Bloch sphere, where the creation, evolution, and annihilation of zeros and poles indicate the appearance of non-Abelian gauge fields. Similar to the Abelian counterpart, non-Abelian gauge fields A get incorporated in a Hamiltonian system through the minimal coupling described by the Peierls substitution H(p) → H(p – A),

H

Dresselhaus SOC

Intensity (a.u.)

Intensity (a.u.)

y

(m+1)θ σ y

y

Polaritonic lensing with Rashba SOC

Phenomena: Zitterbewegung analogs

mθ σ y

ky

g

kx

y

(m+1,n+1)

Intensity (a.u.)

1

0

y

ky

Intensity (a.u.)

Intensity (a.u.)

Frequency detuning

D

p

Longitude (deg)

RES EARCH | R E V I E W

Non-Abelian pumping

y ,

8 of 14

y g

The effect of non-Abelian gauge fields can manifest in the dynamic adiabatic evolution, or pumping, of an expanded Hilbert space. Because such pumping simultaneously involves multiple states, they are connected by BerryWilczek-Zee (BWZ) phase matrix (1), which gives rise to dynamic transition processes among multiple eigenstates described by nonAbelian holonomies. A typical example is the realization of nonAbelian mode braiding through pumping, wherein several degenerate modes are sequentially exchanged following braiding rules. Mathematically, the braiding of n strands is described by an infinite discrete group called a braid group, denoted Bn. Bn has a set of 2(n – 1) generators, denoted ti with i ∈ {1, 2, …, n – 1} and their inverses ti−1. ti executes the exchange of the i-th and i + 1–th strands with i + 1 over crossing i, and the corresponding inverse element denotes under crossing. The generators follow titjti = tjtitj if ji  j j ¼ 1, and titj = tjti if ji  j j ≥ 2. It is straightforward to see that Bn is non-Abelian for all n > 2, because titj ≠ tjti. These properties are schematically shown in Fig. 5A using B2 and B3. The generators of Bn can map to n-by-n matrices. Take B2 as an example, there is t1 → G1(2) = −isy and t1−1 → G1−1(2) = isy, with sy being the second Pauli matrix. It follows that the initial and end states of an arbitrary braid can be connected using such matrices. Because braid groups have natural inclusion characteristics, the matrix representation for the generators of Bn is transparent, for example, for B3, t1 → G1(3) = [G1(2), 0; 0, 1] and t2 → G2(3) = [1, 0; 0, G1(2)]. The non-Abelian characteristics of B3 naturally emerge, because G1(3)G2 (3) ≠ G2(3)G1(3).

y

23 February 2024

number in two dimensions), which may be obtained by summing over the products over the first Chern numbers of different subdimensions (134). In a fiber design using nonreciprocal gyroelectric materials, a rotation angle is used as a synthetic dimension to complement the 3D momenta to form the 4D space, which permits the construction of a topologically nontrivial second Chern number from non-Abelian Berry curvature, which in turn confirms the property of one-way transport (10). Moreover, synthetic non-Abelian Yang monopoles were created in 5D synthetic space using hyperfine ground states of rubidium (9) and photonic bianisotropic semimetal metamaterials (11). Interestingly, when certain symmetries are broken, a Yang monopole can transform into a pair of linked Weyl surfaces, and the second Chern number is characterized by the linking number between them (11, 136–138). Recently, nontrivial second Chern numbers in hyperbolic lattices were also realized in artificial circuit networks, featuring non-Abelian translational operations (139).

g

Yang et al., Science 383, eadf9621 (2024)

4D, top left) can also be created. Creating the Dresselhaus-type non-Abelian gauge fields for polaritonic cavities requires extra symmetry breaking (Fig. 4F). It has been theoretically predicted that hexagonal photonic graphene (120, 121) hosts the Dresselhaus-type fields at its K and K′ points (122), which was confirmed experimentally (24) by the dipolar pseudospin winding at K and K′ [instead of the quadrupolar winding at G similar to that of unpatterned cavities (123)] and the associated optical spin Hall effect. Aside from quantum-well structures, other anisotropic materials like perovskites (22, 124–127), perylene (128, 129), and liquid crystals (125, 126, 130) are also emerging for the versatile engineering of spin-orbit coupling via artificial non-Abelian gauge fields. The controllability of optical and acoustic systems enables the realization of various lattice models with non-Abelian gauge fields. The Qi-Wu-Zhang or half–Bernevig-Hughes-Zhang model (131), a celebrated model for Chern insulators, can be written in real space where sites are connected by matrix-valued non-Abelian hopping links. This model has been realized with electric circuits, allowing the visualization of chiral edge states (26) (Fig. 4G, top). So far, all the phenomena addressed above deal with spatially homogeneous non-Abelian gauge fields, and interests in inhomogeneous ones are also spawning. A typical example is a class of nonAbelian Hofstadter models that features linearly varying gauge fields (101). Such link arrangements give rise to nonsymmorphic chiral symmetries of nontrivial symmetry algebra (50). Proposals have been made toward realizing the models based on photonic synthetic dimensions (47) (Fig. 4G, bottom). Synthetic non-Abelian electric fields can be equally created and manipulated. In the nonAbelian setting, the synthetic electric fields are given by (18, 100) E ¼ ∇ϕ  @A=@t  i½ϕ; A , where ϕ is the scalar potential. The first two terms of the expression are inherited from the Abelian counterpart (as in electromagnetism), meaning that synthetic electric fields can be created from the spatial and temporal gradients of the scalar and vector potentials, respectively. Meanwhile, a third term, that is, the commutator between ϕ and A, also appears, which indicates that non-Abelian electric fields are even possible under temporally static and spatially uniform non-Abelian gauge potentials, a characteristic shared by the synthetic magnetic fields (via the A × A term). So far, non-Abelian electric fields have not been synthesized in light and sound experimentally, according to our knowledge. Synthetic dimensions enable the realization of non-Abelian gauge fields (i.e., connection, and the associated curvature) in higher-dimensional space. A quintessential example is the 4D quantum Hall effect (132–135) featured by the second Chern number (generalization of the first Chern

p

(109–116), and more broadly, relativistic quantum mechanics. Among other established approaches, non-Abelian gauge fields have recently emerged as a new way to demonstrate the ZB effect. Homogeneous anisotropic non-Abelian optical media with electromagnetic duality are sufficient to this end (18) (Fig. 4D, top). In this approach, an anisotropic medium, with a single gauge-field contribution (i.e., Abelian) in the permittivity and permeability tensors, exhibits two branches of modes forming a 1D Dirac point in the isofrequency contour. The addition of another gauge-field component introduces the coupling between the two modes, lifts the Dirac point, and enables the definition of the ZB beat wave number (i.e., describing the interference of the two eigenmodes) proportional to the Dirac mass of the particle under the trembling motion. ZB was also predicted for exciton polaritons in planar semiconductor microcavities (117) based on the TE-TM splitting (see Fig. 1), which provides an effective magnetic field that causes polaritonic precession (21). Meanwhile, polaritonic graphene was theoretically shown to exhibit the ZB effect in the p bands near the K point (24). The ZB effect was recently probed in hybrid organic-inorganic perovskites and GaAs/AlGaAs quantum wells (22, 23). In the angle- and polarization-resolved photoluminescence measurement, ZB was observed under a circular resonant pump that excites both polarization branches, whereas it was diminished under a single polarization excitation (Fig. 4D, bottom). In the planar polaritonic cavities, a TE-TM crossing can appear at a nonzero critical momentum when the static in-plane and the TE-TM fields compensate each other (19). At this crossing momentum, the polaritonic Hamiltonian can be reformulated into a minimal-coupling gauge-field Hamiltonian with a Rashba-type spin-orbit interaction. Notably, the quantum metric, the real part of the quantum geometric tensor, diverges at the TE-TM crossing (118). The crossing can be gapped under external magnetic fields, permitting measurement of nontrivial Berry curvature and quantum metric simultaneously (119). Near the TE-TM crossing, a lensing effect can be interpreted as ZB in the presence of a defect. In particular, when one of the polaritonic modes hits a defect, the spin-orbit interaction induces opposite group velocities for the scattered polaritons toward opposite directions, leading to the focusing effect in the total field intensity. This lensing effect was observed in the perovskite polaritonic platform (22), where a linear-polarized laser excites a polariton flow that hits a potential, splits into circular-polarized flows, and refocuses guided by the non-Abelian magnetic fields. A similar phenomenon should also occur for pure photonic systems, for example, a homogeneous non-Abelian media where the required mode dispersion (blue curve in Fig.

RES EARCH | R E V I E W

A

n=2

Rotation of two modes induces their braiding

B

Braid operations

E

Non-Abelian Thouless pumping

n=3

...

...

τ1

τ1

τ1-1

τ2

C

F

Non-Abelian braiding of three acoustic modes

τ2

τ2

C A B C

B A

D

≠

A B C

B

C Br aid in

1

g

τ1τ2

τ2τ1

Injection

A di re

Braiding

ct io

z

n

0

p

n tio ec r i d

ing aid |P| Br

Braiding of Majorana-like modes bounded by gauge vortex

τ1

n=3

n=3

τ1

Input

Input

Interferometer

Non-Abelian braiding of multiple photonic modes g

G

Non-Abelian quantum holomony of photons A

1.0

0.5

Braiding direction

Input Braiding direction

Input

A B C D E

W

E

y

C

0

A B C D E

y

arrows indicate the injection position at the inputs (29). (E) Non-Abelian Thouless pumping in generalized Lieb lattices (31). Under a particular gauge sequence, an optical zero mode is pumped to hop across the lattice (left). By switching the gauge sequence, the optical mode stays at the same unit cell (right). (F) Braiding of two topological zero modes bounded by Kekulé vortices in photonic waveguides (27). (G) Realization of U(3) gauge structure in photonic waveguides and non-Abelian two-photon holonomy (45). [(B) and (C) are adapted with permission from (28); (D) is adapted with permission from (29); (E) is adapted with permission from (31); (F) is adapted with permission from (27); (G) is adapted from (45)].

y g

Fig. 5. Non-Abelian pumping. (A) Braiding operations of two and three strands. (B) Two-strand braiding manifests as the SO(2) local rotation of two orthogonal vectors induced by the SO(3) global rotation. (C) Non-Abelian braiding realized in coupled acoustic waveguides. Herein, two different sections of coupled waveguides are connected in different orders. The acoustic input in both cases is at waveguide A, but the output is detected at waveguide C (left) and B (right), respectively, which confirms the non-Abelian characteristics (28). (D) Five-mode non-Abelian braiding realized in coupled photonic waveguides. The top panel is a schematic of the waveguide array. The bottom panels are the measured results at the outputs, with bright spots indicating strong optical intensity. The red

,

Note that all Gi(n) are orthogonal matrices with unity determinants, which suggests that they are elements in SO(n). Therefore, it is possible to emulate braiding using saliently designed rotation in an n-dimensional space. One route is to consider the adiabatic evolution of degenerate states. For example, consider a Hamiltonian H = (0M×M, tT; t, 0N×N) with t ∈ ℝMN , a total of n ¼ jM  N j eigenstates are pinned at zero energy because of the sublattice symmetry C−1HC = −H, with C = (−1M×M, 0; 0, 1N×N). These degenerate states form an n-dimensional subspace. When the entries in t are driven by external parameters, the degenerate states undergo adiabatic pumping. Yang et al., Science 383, eadf9621 (2024)

Such a multistate evolution is captured by an n-dimensional BWZ phase matrix. Because all eigenvectors are real, the BWZ matrix belongs to SO(n), which can vary the composition of the states within the subspace. For a 2D subspace, the matrix is O(W) = (cosW, −sinW; sinW, cosW), where W is the solid angle enclosed by the loop of in the parameter space spanned by t. It becomes clear that G1(2) = O(p/2), which realizes the generating operation of B2 (Fig. 5B). The generalization to Bn can be realized by using the natural inclusion property. Waveguide systems are a good platform for realizing the adiabatic evolution of the abovementioned H. For example, acoustic rectan-

23 February 2024

gular waveguide arrays coupled by air bridges have been realized (28), where the coupling magnitudes among the waveguides are tunable by adjusting the positions of the air bridges. The positions slowly vary along the guiding direction to adiabatically pump the guided modes as they propagate down the waveguides. The two generators of B3 were successfully realized in two waveguide arrays with different pumping profiles. The braiding effects manifest as the swapping of dwelling waveguides between the input and output ports. Furthermore, by connecting the two waveguide arrays in different orders, it was observed that the same input mode was converted to different output modes, 9 of 14

RES EARCH | R E V I E W

,

10 of 14

y

23 February 2024

y g

So far, we have focused on Hermitian systems, but non-Hermitian systems are another realm in which non-Abelian effects can emerge. The non-Hermitian formalism describes open systems in which energy exchange with external environments is permitted. Unlike Hermitian systems wherein bands are naturally ordered according to their energy, non-Hermitian spectra, that is, the eigenvalues, are generically complex functions of system parameters, which presents an additional layer of topology. Notably, nonHermitian spectral topology can be nontrivial even for a 1D single-band system without symmetry constraints because the energy is a map from a 1D parameter space, for example, Bloch wavenumber, to a complex plane, on which nontrivial winding can readily emerge (151). Here, we address the non-Abelian phenomena in nonHermitian systems. For comprehensive accounts of non-Hermitian topology, the readers are referred to existing reviews such as (78, 152, 153). Using the fundamental homotopy group, the space of N-dimensional non-Hermitian Hamiltonians are topologically classified as braid groups BN (154, 155), which are non-Abelian groups for N > 2 (as discussed in “Non-Abelian pumping”). Different from the non-Abelian topological classification of multiple bands discussed in “Non-Abelian topological charges” that considers eigenvector rotations, this topology mainly rests upon the geometry of non-Hermitian spectral manifolds, which are self-intersecting complex Riemannian sheets

y

Non-Abelian characteristics of non-Hermitian systems

instead of isolated surfaces. Figure 6A shows an example of a three-band non-Hermitian system, where the first and second bands (green and blue), and the second and third bands (blue and orange) coalesce at different parameters, forming two order-2 exceptional points (EPs). It can be seen that when a closed parametric loop encloses these EPs, which are spectral branch-point singularities, the eigenvalues of different states smoothly connect to one another by crossing the intersection curves called branch cuts. Hence, encircling an EP essentially braids the eigenvalues. For example, the loops L1 and L2 encircle the two different EPs in Fig. 6A, swapping the eigenvalues in the process. It is then clear that encircling the two EPs in different order produces different eigenvalue braids, as shown in Fig. 6B. Such non-Abelian braids have been observed in acoustic systems (Fig. 6B) (156, 157). Leveraging on the non-Hermitian eigenvalue topology, synthetic platforms were developed with ring resonator coupling and modulation designs to realize the single- or two-band knots (158, 159). Subsequently, the multiband nonAbelian braids have also been realized in cavity optomechanics (Fig. 6C) (160). Eigenvalue knots are also experimentally realized in acoustic systems (161, 162). The nontrivial braids can also map to different knot structures when the eigenvalue braids are projected to a 3D Euclidean space. Eigenvalue braids or knots appearing from non-Hermitian Bloch bands also exhibit nonAbelian characters when three or more bands are involved (Fig. 6D) (163–165). Another key difference between non-Hermitian and Hermitian systems is that the eigenvectors of the former are not orthogonal. This feature, coupled with the fact that non-Hermitian eigenvectors are fiber bundles sticking to the spectral manifolds, means that eigenvectors can also smoothly evolve into one another in parallel transport. This effect was readily captured by the fractional geometric phase produced when an EP is encircled (166–171), that is, the state evolution on the eigenvalue manifold is not holonomic even when the parametric loop is closed, and multiple loops in the parameter space are required for the states to recover with a quantized geometric phase. It is also the origin of the non-Abelian exchange of non-Hermitian states (157, 172). An example in acoustics is reported in (157), wherein two order-2 EPs formed by two coalescing states are found on the spectral manifold of a threestate non-Hermitian system. Because these order-2 EPs are formed by different states, encircling them in different sequences produces different state-permutation outcomes. The non-Abelian characteristics are confirmed by stroboscopic measurement of the acoustic wave functions under a constant gauge. Mode dynamics presents another intriguing scenario. Because of the intrinsic instability

g

Yang et al., Science 383, eadf9621 (2024)

is the topological modes bounded by a gauge vortex. For example, a honeycomb lattice under Kekulé modulation, which takes the form of a real-space vortex gauge, sustains Majorana-like zero modes bounded to the vortex core (145). Such modulated graphene and the zero modes have been realized in 2D photonic (146) or phononic crystals (147). Multiple spatially wellseparated vortices can simultaneously bound multiple zero modes at different locations. A way to realize braiding is to slowly interchange the positions of the vortices in real space, which swaps the relevant zero modes. To realize such effects in optics, waveguide arrays based on the modulated lattices were constructed, and the evolution of the vortices is encoded in the propagation directions (148). A photonic experiment successfully demonstrated the viability of this approach (27) (Fig. 5F). More recently, nonAbelian quantum holonomy was realized with indistinguishable photons in coupled waveguide systems (45, 46), indicating the possibility for realizing more complex braiding structures (Fig. 5G). Proposals also suggest the realization of Majorana-like zero modes in the classical-mechanical analog of Kitaev model (149), which are also candidates for realizing braiding operations (150).

p

which confirms the non-Abelian characteristics of B3 (Fig. 5C). Based on a similar principle, photonic experiments were performed by fabricating an array of meandering optical waveguides etched in glass substrates using femtosecond laser writing (29), wherein the waveguides are evanescently coupled so the coupling strengths are tuned by changing their separations (Fig. 5D). A similar approach can also be used to braid topological edge modes in a Y-junction formed by a 1D topological lattice (140, 141). Thouless pumping is a topological phenomenon characterized by the quantized transport of charge under adiabatic evolution of the system Hamiltonian (142). The braiding effect can be incorporated with such pumping to realize non-Abelian Thouless pumping. Lattice systems with a dispersionless bulk band are ideal for this demonstration. In an openboundary lattice, such a flat band consists of a set of highly degenerate modes that has zero group velocity. For example, optical waveguides can be arranged into a 1D generalized Lieb lattice (30), in which a flat band exists at zero energy. A position-dependent gauge field drives the spatial evolution of the guiding modes and makes them sequentially hop from one unit cell to the next even though they have zero group velocity in the transverse directions. The transverse motion is thus a generalized form of Thouless pumping. Furthermore, it is experimentally observed that changing the spatial order of the gauge field produces different hopping sequences, meaning that the pumping is non-Abelian in character (Fig. 5E). Such non-Abelian Thouless pumping is successfully realized in optics (31) using on-chip photonic waveguides. A demonstration in acoustics is also reported (32). Presently, most studies on non-Abelian holonomies have focused on adiabatic evolution in systems with perfectly degenerate bands everywhere in the parameter space. However, the requirement for perfectly degenerate bands is usually only approximately satisfied in realistic systems. Recently, it was shown that, in contrast to conventional wisdom, non-Abelian holonomy could also exist in systems with isolated degeneracies between multiple energy bands. Two groups independently showed that the transition between different states could be arbitrarily controlled by introducing abrupt turns when the evolution path traverses isolated degeneracies (32, 143). These works suggest that U(N) holonomy may be used to describe the evolution of states in physical systems with N bands connected by a finite number of isolated degeneracies across the entire parameter space. This new approach greatly broadens the choice of parameter spaces to achieve non-Abelian holonomy. Braiding operations can also emerge in specially designed special-unitary and unitary operations. A primary candidate of this approach is the Majorana zero modes (144). One example

RES EARCH | R E V I E W

A

Non-Hermitian eigenvalue manifolds

B Non-Abelian braiding of three non-Hermitian bands: Acoustics

2

1

L1

C

Non-Abelian braiding of three non-Hermitian bands: Cavity optomechanics

1

3

control laser

2

3

probe 120 laser

Multi-tone driving

L2L1

L1L2

80 +

Re (E)

L2

0

3

T=4K

3 1 Modes

D

E

–40

150

–80 –120

/2π]

Im[λ

0

(Hz)

Non-Abelian Wilson lines via dissipative couplings

g

Bands

Knot

p

Braid

Knot and braid structures of non-Hermitian bands

0 50 100

) (Hz

2

2π]

2

/ Re[λ

Resonator 1

40

Yang et al., Science 383, eadf9621 (2024)

23 February 2024

11 of 14

,

Non-Abelian topological charges for a set of bandgaps are characterized by matrix-like entities, which complement and enrich the established integer classifications for a single bandgap. The noncommuting properties imply phenomena such as multiple-choice topological phase transition paths (14). We expect

nonreciprocal non-Abelian gauge fields can drive non-Hermitian topological phase transition despite the lack of gauge flux in one dimension and break the gauge invariance of non-Abelian Wilson loops in 2D non-Hermitian systems (180). We expect that the rich interplay among non-Hemiticity (78, 152, 153), nonreciprocity (181), and the contents reviewed here will lead to further fundamental discoveries and practical applications, for example, nonreciprocal optoelectronics (182, 183). Perhaps one of the most attractive applications of braiding is in quantum computation. Non-Abelian braiding is one of the essential components of universal quantum logic. The realization of non-Abelian braiding in light and sound, therefore, not only expands our capability for wave manipulations but may also bring new toolsets for implementing topological quantum-logic operations based on bosonic platforms (184–187). Toward this goal, nonAbelian gauge fields could play an important role in synergy with nonlinearity for generating many-body photonic effects (188).

y

Conclusion and outlook

further realizations of non-Abelian topological charges with photonics and acoustics in the near future, thereby enabling mutual crossfertilization between different research fields. We will see that non-Abelian defects may also play pivotal roles in morphogenesis and cosmology, where singularities called cosmic strings seem to correspond with the defect lines (81). Presently, the non-Abelian topological phase based on eigen-frame rotation mainly focuses on PT-symmetric Bloch systems; the expansions toward other directions deserve further exploration and study, for example, the Floquet multi-gap topology (177). Non-Abelian gauge fields have so far mostly been studied in Hermitian systems, and their interplay with non-Hermiticity deserves further exploration. Indeed, the presence of nonHermiticity in lattice systems has been treated as imaginary Abelian gauge fields in some studies of non-Hermitian skin effect [e.g., (178, 179)], whose further interplay with nonAbelian gauge fields is thus anticipated. A recent endeavor in this direction shows that

y g

problem for states with higher imaginary eigenvalues and the nonorthogonality of the non-Hermitian eigenmodes, non-Hermitian mode evolutions inevitably leave the adiabatic path and move to the state with higher loss (173). Such a nonadiabatic effect has been observed and leveraged for asymmetric switching of waveguide modes (174, 175). This nonadiabaticity also causes non-Abelian Wilson line evolution of untouched dissipative bands in time-multiplexed photonic networks (176) (Fig. 6E). If nonadiabatic transition can be precisely controlled, it may also function as a particular operation for tailoring non-Abelian mode dynamics in non-Hermitian systems.

Si3N4 membrane controlled by cavity optomechanics and the measured braiding eigenvalue spectrum (160). (D) The eigenvalue braids and the corresponding knot structures appear in non-Hermitian Bloch bands (163). (E) Non-Abelian Wilson line of isolated bands obtained through dissipative coupling in synthetic time dimension (176). [(B) is adapted with permission from (156, 157); (C) is adapted with permission from (160); (D) is reprinted with permission from (163) Copyright 2021 by the American Physical Society; (E) is adapted from (176)].

y

Fig. 6. Non-Abelian characteristics emerging in non-Hermitian systems. (A) The real parts of the eigenvalue Riemann surfaces of a three-band nonHermitian system. The red circles mark two order-2 EPs. The blue and red curves are the eigenvalue trajectories when the two EPs are encircled individually by the parametric loops labeled L1 and L2, respectively. (B) Distinct eigenvalue braids produced by concatenating the two loops in different orders, L1L2 (left) and L2L1 (right), and the coupled acoustic cavities (middle) that realized the nonAeblian eigenvalue braids and non-Abelian state exchanges (156, 157). (C) A

RES EARCH | R E V I E W

RE FE RENCES AND N OT ES

2.

4. 5.

6.

8.

11.

12.

13.

14.

15.

Yang et al., Science 383, eadf9621 (2024)

23 February 2024

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

55.

56.

57. 58.

59.

60.

61.

62.

63.

64.

65.

12 of 14

,

10.

44.

y

9.

43.

T. Dai et al., Topologically protected quantum entanglement emitters. Nat. Photonics 16, 248–257 (2022). doi: 10.1038/ s41566-021-00944-2 X. Lu, A. McClung, K. Srinivasan, High-Q slow light and its localization in a photonic crystal microring. Nat. Photonics 16, 66–71 (2022). doi: 10.1038/s41566-021-00912-w J. Wang et al., Experimental observation of Berry phases in optical Möbius-strip microcavities. Nat. Photonics 17, 120–125 (2023). doi: 10.1038/s41566-022-01107-7 V. Neef et al., Three-dimensional non-Abelian quantum holonomy. Nat. Phys. 19, 30–34 (2023). doi: 10.1038/ s41567-022-01807-5 M. Kremer, L. Teuber, A. Szameit, S. Scheel, Optimal design strategy for non-Abelian geometric phases using Abelian gauge fields based on quantum metric. Phys. Rev. Res. 1, 033117 (2019). doi: 10.1103/PhysRevResearch.1.033117 D. Cheng, K. Wang, S. Fan, Artificial non-Abelian lattice gauge fields for photons in the synthetic frequency dimension. Phys. Rev. Lett. 130, 083601 (2023). doi: 10.1103/ PhysRevLett.130.083601; pmid: 36898123 Y. Zhao, Y.-X. Huang, S. A. Yang, Z2-projective translational symmetry protected topological phases. Phys. Rev. B 102, 161117 (2020). doi: 10.1103/PhysRevB.102.161117 Y. X. Zhao, C. Chen, X.-L. Sheng, S. A. Yang, Switching spinless and spinful topological phases with projective PT symmetry. Phys. Rev. Lett. 126, 196402 (2021). doi: 10.1103/ PhysRevLett.126.196402; pmid: 34047612 Y. Yang et al., Non-Abelian nonsymmorphic chiral symmetries. Phys. Rev. B 106, L161108 (2022). doi: 10.1103/ PhysRevB.106.L161108 H. Xue et al., Projectively enriched symmetry and topology in acoustic crystals. Phys. Rev. Lett. 128, 116802 (2022). doi: 10.1103/PhysRevLett.128.116802; pmid: 35363008 T. Li et al., Acoustic Möbius insulators from projective symmetry. Phys. Rev. Lett. 128, 116803 (2022). doi: 10.1103/ PhysRevLett.128.116803; pmid: 35362999 Y. Meng et al., Spinful topological phases in acoustic crystals with projective PT symmetry. Phys. Rev. Lett. 130, 026101 (2023). doi: 10.1103/PhysRevLett.130.026101; pmid: 36706409 J. Hu, S. Zhuang, Y. Yang, Synthetic gauge fields enable highorder topology on Brillouin real projective plane. arXiv:2306.15477v2 [cond-mat.mes-hall] (2023). K. Klitzing, G. Dorda, M. Pepper, New method for highaccuracy determination of the fine-structure constant based on quantized Hall resistance. Phys. Rev. Lett. 45, 494–497 (1980). doi: 10.1103/PhysRevLett.45.494 D. J. Thouless, M. Kohmoto, M. P. Nightingale, M. den Nijs, Quantized Hall conductance in a two-dimensional periodic potential. Phys. Rev. Lett. 49, 405–408 (1982). doi: 10.1103/PhysRevLett.49.405 T. Ozawa et al., Topological photonics. Rev. Mod. Phys. 91, 015006 (2019). doi: 10.1103/RevModPhys.91.015006 G. Ma, M. Xiao, C. T. Chan, Topological phases in acoustic and mechanical systems. Nat. Rev. Phys. 1, 281–294 (2019). doi: 10.1038/s42254-019-0030-x A. P. Schnyder, S. Ryu, A. Furusaki, A. W. W. Ludwig, Classification of topological insulators and superconductors in three spatial dimensions. Phys. Rev. B 78, 195125 (2008). doi: 10.1103/PhysRevB.78.195125 A. Bouhon et al., Non-Abelian reciprocal braiding of Weyl points and its manifestation in ZrTe. Nat. Phys. 16, 1137–1143 (2020). doi: 10.1038/s41567-020-0967-9 H. Qiu et al., Minimal non-abelian nodal braiding in ideal metamaterials. Nat. Commun. 14, 1261 (2023). doi: 10.1038/ s41467-023-36952-9; pmid: 36878907 J. Ahn, S. Park, B.-J. Yang, Failure of Nielsen-Ninomiya theorem and fragile topology in two-dimensional systems with space-time inversion symmetry: Application to twisted bilayer graphene at magic angle. Phys. Rev. X 9, 021013 (2019). doi: 10.1103/PhysRevX.9.021013 B. Peng, A. Bouhon, R.-J. Slager, B. Monserrat, Multigap topology and non-Abelian braiding of phonons from first principles. Phys. Rev. B 105, 085115 (2022). doi: 10.1103/ PhysRevB.105.085115 B. Peng, A. Bouhon, B. Monserrat, R.-J. Slager, Phonons as a platform for non-Abelian braiding and its manifestation in layered silicates. Nat. Commun. 13, 423 (2022). doi: 10.1038/ s41467-022-28046-9; pmid: 35058473 A. Tiwari, T. Bzdušek, Non-Abelian topology of nodal-line rings in PT-symmetric systems. Phys. Rev. B 101, 195130 (2020). doi: 10.1103/PhysRevB.101.195130

y g

7.

42.

y

3.

F. Wilczek, A. Zee, Appearance of gauge structure in simple dynamical systems. Phys. Rev. Lett. 52, 2111–2114 (1984). doi: 10.1103/PhysRevLett.52.2111 C. A. Mead, The geometric phase in molecular systems. Rev. Mod. Phys. 64, 51–85 (1992). doi: 10.1103/ RevModPhys.64.51 A. Bohm, A. Mostafazadeh, H. Koizumi, Q. Niu, J. Zwanziger, The Geometric Phase in Quantum Systems: Foundations, Mathematical Concepts, and Applications in Molecular and Condensed Matter Physics (Springer, 2013). K. G. Wilson, Confinement of quarks. Phys. Rev. D Part. Fields 10, 2445–2459 (1974). doi: 10.1103/PhysRevD.10.2445 D. Vanderbilt, Berry Phases in Electronic Structure Theory: Electric Polarization, Orbital Magnetization and Topological Insulators (Cambridge Univ. Press, 2018). doi: 10.1017/ 9781316662205 A. Alexandradinata, J. Höller, C. Wang, H. Cheng, L. Lu, Crystallographic splitting theorem for band representations and fragile topological photonic crystals. Phys. Rev. B 102, 115117 (2020). doi: 10.1103/PhysRevB.102.115117 T. Christensen, H. C. Po, J. D. Joannopoulos, M. Soljačić, Location and topology of the fundamental gap in photonic crystals. Phys. Rev. X 12, 021066 (2022). doi: 10.1103/ PhysRevX.12.021066 V. Gupta, B. Bradlyn, Wannier-function methods for topological modes in one-dimensional photonic crystals. Phys. Rev. A 105, 053521 (2022). doi: 10.1103/ PhysRevA.105.053521 S. Sugawa, F. Salces-Carcoba, A. R. Perry, Y. Yue, I. B. Spielman, Second Chern number of a quantum-simulated non-Abelian Yang monopole. Science 360, 1429–1434 (2018). doi: 10.1126/ science.aam9031; pmid: 29954972 L. Lu, H. Gao, Z. Wang, Topological one-way fiber of second Chern number. Nat. Commun. 9, 5384 (2018). doi: 10.1038/ s41467-018-07817-3; pmid: 30568189 S. Ma et al., Linked Weyl surfaces and Weyl arcs in photonic metamaterials. Science 373, 572–576 (2021). doi: 10.1126/ science.abi7803; pmid: 34326241 C.-K. Chiu, J. C. Teo, A. P. Schnyder, S. Ryu, Classification of topological quantum matter with symmetries. Rev. Mod. Phys. 88, 035005 (2016). doi: 10.1103/RevModPhys.88.035005 Q. Wu, A. A. Soluyanov, T. Bzdušek, Non-Abelian band topology in noninteracting metals. Science 365, 1273 (2019). doi: 10.1126/science.aau8740; pmid: 31467188 Q. Guo et al., Experimental observation of non-Abelian topological charges and edge states. Nature 594, 195–200 (2021). doi: 10.1038/s41586-021-03521-3; pmid: 34108697 E. Yang et al., Observation of non-Abelian nodal links in photonics. Phys. Rev. Lett. 125, 033901 (2020). doi: 10.1103/ PhysRevLett.125.033901; pmid: 32745405

B. Jiang et al., Experimental observation of non-Abelian topological acoustic semimetals and their phase transitions. Nat. Phys. 17, 1239–1246 (2021). doi: 10.1038/ s41567-021-01340-x 17. M. Aidelsburger, S. Nascimbene, N. Goldman, Artificial gauge fields in materials and engineered systems. C. R. Phys. 19, 394–432 (2018). doi: 10.1016/j.crhy.2018.03.002 18. Y. Chen et al., Non-Abelian gauge field optics. Nat. Commun. 10, 3125 (2019). doi: 10.1038/s41467-019-10974-8; pmid: 31311929 19. H. Terças, H. Flayac, D. D. Solnyshkov, G. Malpuech, Non-Abelian gauge fields in photonic cavities and photonic superfluids. Phys. Rev. Lett. 112, 066402 (2014). doi: 10.1103/PhysRevLett.112.066402; pmid: 24580697 20. Y. Yang et al., Synthesis and observation of non-Abelian gauge fields in real space. Science 365, 1021–1025 (2019). doi: 10.1126/science.aay3183; pmid: 31488687 21. E. Sedov, Y. Rubo, A. Kavokin, Zitterbewegung of excitonpolaritons. Phys. Rev. B 97, 245312 (2018). doi: 10.1103/ PhysRevB.97.245312 22. L. Polimeno et al., Experimental investigation of a non-Abelian gauge field in 2D perovskite photonic platform. Optica 8, 1442 (2021). doi: 10.1364/OPTICA.427088 23. S. Lovett et al., Observation of Zitterbewegung in photonic microcavities. Light Sci. Appl. 12, 126 (2023). doi: 10.1038/ s41377-023-01162-x; pmid: 37221208 24. C. Whittaker et al., Optical analogue of Dresselhaus spin– orbit interaction in photonic graphene. Nat. Photonics 15, 193–196 (2021). doi: 10.1038/s41566-020-00729-z 25. K. Osterloh, M. Baig, L. Santos, P. Zoller, M. Lewenstein, Cold atoms in non-Abelian gauge potentials: From the Hofstadter “moth” to lattice gauge theory. Phys. Rev. Lett. 95, 010403 (2005). doi: 10.1103/PhysRevLett.95.010403; pmid: 16090589 26. J. Wu et al., Non-Abelian gauge fields in circuit systems. Nat. Electron. 5, 635–642 (2022). doi: 10.1038/s41928-022-00833-8 27. J. Noh et al., Braiding photonic topological zero modes. Nat. Phys. 16, 989–993 (2020). doi: 10.1038/s41567-020-1007-5 28. Z.-G. Chen, R.-Y. Zhang, C. T. Chan, G. Ma, Classical nonAbelian braiding of acoustic modes. Nat. Phys. 18, 179–184 (2022). doi: 10.1038/s41567-021-01431-9 29. X.-L. Zhang et al., Non-Abelian braiding on photonic chips. Nat. Photonics 16, 390–395 (2022). doi: 10.1038/ s41566-022-00976-2 30. V. Brosco, L. Pilozzi, R. Fazio, C. Conti, Non-Abelian Thouless pumping in a photonic lattice. Phys. Rev. A 103, 063518 (2021). doi: 10.1103/PhysRevA.103.063518 31. Y.-K. Sun et al., Non-Abelian Thouless pumping in photonic waveguides. Nat. Phys. 18, 1080–1085 (2022). doi: 10.1038/ s41567-022-01669-x 32. O. You et al., Observation of non-Abelian Thouless pump. Phys. Rev. Lett. 128, 244302 (2022). doi: 10.1103/ PhysRevLett.128.244302; pmid: 35776444 33. A. B. Khanikaev et al., Photonic topological insulators. Nat. Mater. 12, 233–239 (2013). doi: 10.1038/nmat3520; pmid: 23241532 34. C. He et al., Photonic topological insulator with broken time-reversal symmetry. Proc. Natl. Acad. Sci. U.S.A. 113, 4924–4928 (2016). doi: 10.1073/pnas.1525502113; pmid: 27092005 35. M. G. Silveirinha, P · T · D symmetry-protected scattering anomaly in optics. Phys. Rev. B 95, 035153 (2017). doi: 10.1103/PhysRevB.95.035153 36. F. Liu, J. Li, Gauge field optics with anisotropic media. Phys. Rev. Lett. 114, 103902 (2015). doi: 10.1103/ PhysRevLett.114.103902; pmid: 25815934 37. S. Dufferwiel et al., Spin textures of exciton-polaritons in a tunable microcavity with large TE-TM splitting. Phys. Rev. Lett. 115, 246401 (2015). doi: 10.1103/PhysRevLett.115.246401; pmid: 26705642 38. M. Hafezi, E. A. Demler, M. D. Lukin, J. M. Taylor, Robust optical delay lines with topological protection. Nat. Phys. 7, 907–912 (2011). doi: 10.1038/nphys2063 39. M. Hafezi, S. Mittal, J. Fan, A. Migdall, J. Taylor, Imaging topological edge states in silicon photonics. Nat. Photonics 7, 1001–1005 (2013). doi: 10.1038/nphoton.2013.274 40. S. Mittal, E. A. Goldschmidt, M. Hafezi, A topological source of quantum light. Nature 561, 502–506 (2018). doi: 10.1038/ s41586-018-0478-3; pmid: 30202090 41. S. Mittal, V. V. Orre, D. Leykam, Y. D. Chong, M. Hafezi, Photonic anomalous quantum Hall effect. Phys. Rev. Lett. 123, 043201 (2019). doi: 10.1103/PhysRevLett.123.043201; pmid: 31491276

g

1.

16.

p

More optical and acoustic degrees of freedom can be leveraged for non-Abelian phenomena. Orbital angular momenta have been proposed and widely realized as a synthetic spatial dimension, that is, to label lattice sites, in frequencydegenerate optical cavities and quantum walks (189–195). Nevertheless, to our knowledge, orbital angular momenta have not been used as a pseudospin degree of freedom for non-Abelian operations, which should also be possible. For example, one can couple angular momentum modes m = ±1 with a q = 1 q-plate to form the pseudospin in a cavity, where suitable dispersion should be created to minimize the leakage to other undesirable high-order angular momenta. Besides, many of the realizations reviewed here are in the low-frequency domain. Hence, there is a growing interest in realizing non-Abelian effects in the Terahertz and optical regimes in miniaturized platforms. This is particularly attractive in the context of integrated platforms, where previously uncharacterized topological building blocks are emerging (43, 44, 196).

RES EARCH | R E V I E W

92.

93.

94.

95.

96.

97.

98.

99.

100.

101.

102.

104.

105.

106.

108.

111.

112.

113.

114.

23 February 2024

13 of 14

,

110.

y

109.

y g

107.

115. X. Zhang, Observing Zitterbewegung for photons near the Dirac point of a two-dimensional photonic crystal. Phys. Rev. Lett. 100, 113903 (2008). doi: 10.1103/ PhysRevLett.100.113903; pmid: 18517788 116. F. Dreisow et al., Classical simulation of relativistic Zitterbewegung in photonic lattices. Phys. Rev. Lett. 105, 143902 (2010). doi: 10.1103/PhysRevLett.105.143902; pmid: 21230830 117. I. A. Shelykh, A. V. Kavokin, Y. G. Rubo, T. Liew, G. Malpuech, Polariton polarization-sensitive phenomena in planar semiconductor microcavities. Semicond. Sci. Technol. 25, 013001 (2009). doi: 10.1088/0268-1242/25/1/013001 118. O. Bleu, D. D. Solnyshkov, G. Malpuech, Measuring the quantum geometric tensor in two-dimensional photonic and exciton-polariton systems. Phys. Rev. B 97, 195422 (2018). doi: 10.1103/PhysRevB.97.195422 119. A. Gianfrate et al., Measurement of the quantum geometric tensor and of the anomalous Hall drift. Nature 578, 381–385 (2020). doi: 10.1038/s41586-020-1989-2; pmid: 32076220 120. T. Jacqmin et al., Direct observation of Dirac cones and a flatband in a honeycomb lattice for polaritons. Phys. Rev. Lett. 112, 116402 (2014). doi: 10.1103/ PhysRevLett.112.116402; pmid: 24702392 121. V. Sala et al., Spin-orbit coupling for photons and polaritons in microstructures. Phys. Rev. X 5, 011034 (2015). doi: 10.1103/PhysRevX.5.011034 122. A. V. Nalitov, G. Malpuech, H. Terças, D. D. Solnyshkov, Spinorbit coupling and the optical spin Hall effect in photonic graphene. Phys. Rev. Lett. 114, 026803 (2015). doi: 10.1103/ PhysRevLett.114.026803; pmid: 25635557 123. C. Leyder et al., Observation of the optical spin Hall effect. Nat. Phys. 3, 628–631 (2007). doi: 10.1038/nphys676 124. M. S. Spencer et al., Spin-orbit-coupled exciton-polariton condensates in lead halide perovskites. Sci. Adv. 7, eabj7667 (2021). doi: 10.1126/sciadv.abj7667; pmid: 34851673 125. K. Łempicka-Mirek et al., Electrically tunable Berry curvature and strong light-matter coupling in liquid crystal microcavities with 2D perovskite. Sci. Adv. 8, eabq7533 (2022). doi: 10.1126/sciadv.abq7533; pmid: 36197989 126. Y. Li et al., Manipulating polariton condensates by RashbaDresselhaus coupling at room temperature. Nat. Commun. 13, 3785 (2022). doi: 10.1038/s41467-022-31529-4; pmid: 35778391 127. L. Polimeno et al., Tuning of the Berry curvature in 2D perovskite polaritons. Nat. Nanotechnol. 16, 1349–1354 (2021). doi: 10.1038/s41565-021-00977-2; pmid: 34675412 128. J. Ren et al., Nontrivial band geometry in an optically active system. Nat. Commun. 12, 689 (2021). doi: 10.1038/s41467020-20845-2; pmid: 33514702 129. J. Ren et al., Realization of exciton‐mediated optical spin‐ orbit interaction in organic microcrystalline resonators. Laser Photonics Rev. 16, 2100252 (2022). doi: 10.1002/ lpor.202100252 130. K. Rechcińska et al., Engineering spin-orbit synthetic Hamiltonians in liquid-crystal optical cavities. Science 366, 727–730 (2019). doi: 10.1126/science.aay4182; pmid: 31699934 131. X.-L. Qi, Y.-S. Wu, S.-C. Zhang, Topological quantization of the spin Hall effect in two-dimensional paramagnetic semiconductors. Phys. Rev. B 74, 085308 (2006). doi: 10.1103/PhysRevB.74.085308 132. T. Ozawa, H. M. Price, N. Goldman, O. Zilberberg, I. Carusotto, Synthetic dimensions in integrated photonics: From optical isolation to four-dimensional quantum Hall physics. Phys. Rev. A 93, 043827 (2016). doi: 10.1103/PhysRevA.93.043827 133. M. Lohse, C. Schweizer, H. M. Price, O. Zilberberg, I. Bloch, Exploring 4D quantum Hall physics with a 2D topological charge pump. Nature 553, 55–58 (2018). doi: 10.1038/ nature25000; pmid: 29300006 134. O. Zilberberg et al., Photonic topological boundary pumping as a probe of 4D quantum Hall physics. Nature 553, 59–62 (2018). doi: 10.1038/nature25011; pmid: 29300011 135. Z.-G. Chen, W. Zhu, Y. Tan, L. Wang, G. Ma, Acoustic realization of a four-dimensional higher-order Chern insulator and boundary-modes engineering. Phys. Rev. X 11, 011016 (2021). doi: 10.1103/PhysRevX.11.011016 136. B. Lian, S.-C. Zhang, Five-dimensional generalization of the topological Weyl semimetal. Phys. Rev. B 94, 041105 (2016). doi: 10.1103/PhysRevB.94.041105 137. B. Lian, S.-C. Zhang, Weyl semimetal and topological phase transition in five dimensions. Phys. Rev. B 95, 235106 (2017). doi: 10.1103/PhysRevB.95.235106 138. J.-Y. Chen, B. Lian, S.-C. Zhang, Doubling theorem and boundary states of five-dimensional Weyl semimetal.

y

103.

G. F. Lange, A. Bouhon, B. Monserrat, R.-J. Slager, Topological continuum charges of acoustic phonons in two dimensions and the Nambu-Goldstone theorem. Phys. Rev. B 105, 064301 (2022). doi: 10.1103/PhysRevB.105.064301 F. N. Ünal, A. Bouhon, R.-J. Slager, Topological Euler class as a dynamical observable in optical lattices. Phys. Rev. Lett. 125, 053601 (2020). doi: 10.1103/PhysRevLett.125.053601 M. Ezawa, Topological Euler insulators and their electric circuit realization. Phys. Rev. B 103, 205303 (2021). doi: 10.1103/PhysRevB.103.205303 P. M. Lenggenhager, X. Liu, T. Neupert, T. Bzdušek, Universal higher-order bulk-boundary correspondence of triple nodal points. Phys. Rev. B 106, 085129 (2022). doi: 10.1103/ PhysRevB.106.085129 W. Liu, H. Wang, B. Yang, S. Zhang, Unveiling real triple degeneracies in crystals: Exploring link and compound structures. arXiv:2307.01228 [cond-mat.mes-hall] (2023). A. Das, S. Pujari, Topological character of three-dimensional nexus triple point degeneracies. Phys. Rev. B 102, 235148 (2020). doi: 10.1103/PhysRevB.102.235148 A. Das, E. Cornfeld, S. Pujari, Punctured-Chern topological invariants for semimetallic band structures. Phys. Rev. Lett. 130, 186202 (2023). doi: 10.1103/PhysRevLett.130.186202; pmid: 37204898 A. Bouhon, Y.-Q. Zhu, R.-J. Slager, G. Palumbo, Second Euler number in four dimensional synthetic matter. arXiv:2301.08827 [cond-mat.mes-hall] (2023). J. Dalibard, F. Gerbier, G. Juzeliūnas, P. Öhberg, Colloquium: Artificial gauge potentials for neutral atoms. Rev. Mod. Phys. 83, 1523 (2011). doi: 10.1103/RevModPhys.83.1523 N. Goldman, G. Juzeliūnas, P. Öhberg, I. B. Spielman, Lightinduced gauge fields for ultracold atoms. Rep. Prog. Phys. 77, 126401 (2014). doi: 10.1088/0034-4885/77/12/126401; pmid: 25422950 Y. Yang, B. Zhen, J. D. Joannopoulos, M. Soljačić, Non-Abelian generalizations of the Hofstadter model: Spin-orbit-coupled butterfly pairs. Light Sci. Appl. 9, 177 (2020). doi: 10.1038/ s41377-020-00384-7; pmid: 33088494 T. T. Wu, C. N. Yang, Concept of nonintegrable phase factors and global formulation of gauge fields. Phys. Rev. D Part. Fields 12, 3845–3857 (1975). doi: 10.1103/PhysRevD.12.3845 P. A. Horváthy, Non-Abelian Aharonov-Bohm effect. Phys. Rev. D Part. Fields 33, 407–414 (1986). doi: 10.1103/ PhysRevD.33.407; pmid: 9956634 M. G. Alford, J. March-Russell, F. Wilczek, Discrete quantum hair on black holes and the non-abelian Aharonov-Bohm effect. Nucl. Phys. B 337, 695–708 (1990). doi: 10.1016/ 0550-3213(90)90512-C M. Chaichian, P. Prešnajder, M. M. Sheikh-Jabbari, A. Tureanu, Aharonov–Bohm effect in noncommutative spaces. Phys. Lett. B 527, 149 (2002). doi: 10.1016/S03702693(02)01176-0 A. Jacob, P. Öhberg, G. Juzeliūnas, L. Santos, Cold atom dynamics in non-Abelian gauge fields. Appl. Phys. B 89, 439 (2007). doi: 10.1007/s00340-007-2865-6 X.-D. Zhang, Z. Wang, L.-B. Hu, Z.-M. Zhang, S.-L. Zhu, Detecting unambiguously non-Abelian geometric phases with trapped ions. New J. Phys. 10, 043031 (2008). doi: 10.1088/ 1367-2630/10/4/043031 M. Fruchart, Y. Zhou, V. Vitelli, Dualities and non-Abelian mechanics. Nature 577, 636–640 (2020). doi: 10.1038/ s41586-020-1932-6; pmid: 31959986 W. Zawadzki, T. M. Rusin, Zitterbewegung (trembling motion) of electrons in semiconductors: A review. J. Phys. Condens. Matter 23, 143201 (2011). doi: 10.1088/0953-8984/23/14/ 143201; pmid: 21422506 J. Y. Vaishnav, C. W. Clark, Observing Zitterbewegung with ultracold atoms. Phys. Rev. Lett. 100, 153002 (2008). doi: 10.1103/PhysRevLett.100.153002; pmid: 18518102 M. Hasan et al., Wave packet dynamics in synthetic nonAbelian gauge fields. Phys. Rev. Lett. 129, 130402 (2022). doi: 10.1103/PhysRevLett.129.130402; pmid: 36206409 X. Zhang, Z. Liu, Extremal transmission and beating effect of acoustic waves in two-dimensional sonic crystals. Phys. Rev. Lett. 101, 264303 (2008). doi: 10.1103/ PhysRevLett.101.264303; pmid: 19437645 L.-G. Wang, Z.-G. Wang, S.-Y. Zhu, Zitterbewegung of optical pulses near the Dirac point inside a negative-zero-positive index metamaterial. Europhys. Lett. 86, 47008 (2009). doi: 10.1209/0295-5075/86/47008 S. Longhi, Photonic analog of Zitterbewegung in binary waveguide arrays. Opt. Lett. 35, 235–237 (2010). doi: 10.1364/OL.35.000235; pmid: 20081979

g

Yang et al., Science 383, eadf9621 (2024)

91.

p

66. H. Park, S. Wong, X. Zhang, S. S. Oh, Non-Abelian charged nodal links in a dielectric photonic crystal. ACS Photonics 8, 2746–2754 (2021). doi: 10.1021/acsphotonics.1c00876 67. D. Wang et al., Observation of non-Abelian charged nodes linking nonadjacent gaps. Phys. Rev. Lett. 129, 263604 (2022). doi: 10.1103/PhysRevLett.129.263604; pmid: 36608196 68. M. Wang et al., Experimental observation of non-Abelian earring nodal links in phononic crystals. Phys. Rev. Lett. 128, 246601 (2022). doi: 10.1103/PhysRevLett.128.246601; pmid: 35776454 69. H. Park, S. Wong, A. Bouhon, R.-J. Slager, S. S. Oh, Topological phase transitions of non-Abelian charged nodal lines in spring-mass systems. Phys. Rev. B 105, 214108 (2022). doi: 10.1103/PhysRevB.105.214108 70. X.-Q. Sun, S.-C. Zhang, T. Bzdušek, Conversion rules for weyl points and nodal lines in topological media. Phys. Rev. Lett. 121, 106402 (2018). doi: 10.1103/PhysRevLett.121.106402; pmid: 30240246 71. S. Chen, A. Bouhon, R.-J. Slager, B. Monserrat, Non-Abelian braiding of Weyl nodes via symmetry-constrained phase transitions. Phys. Rev. B 105, L081117 (2022). doi: 10.1103/ PhysRevB.105.L081117 72. T. Bzdušek, M. Sigrist, Robust doubly charged nodal lines and nodal surfaces in centrosymmetric systems. Phys. Rev. B 96, 155105 (2017). doi: 10.1103/PhysRevB.96.155105 73. J. Ahn, D. Kim, Y. Kim, B.-J. Yang, Band topology and linking structure of nodal line semimetals with Z2 monopole charges. Phys. Rev. Lett. 121, 106403 (2018). doi: 10.1103/ PhysRevLett.121.106403; pmid: 30240267 74. P. M. Lenggenhager, X. Liu, S. S. Tsirkin, T. Neupert, T. Bzdušek, From triple-point materials to multiband nodal links. Phys. Rev. B 103, L121101 (2021). doi: 10.1103/ PhysRevB.103.L121101 75. S. Park, Y. Hwang, H. C. Choi, B.-J. Yang, Topological acoustic triple point. Nat. Commun. 12, 6781 (2021). doi: 10.1038/ s41467-021-27158-y; pmid: 34811349 76. A. Bouhon, R.-J. Slager, Multi-gap topological conversion of Euler class via band-node braiding: Minimal models, PTlinked nodal rings, and chiral heirs. arXiv:2203.16741 [cond-mat.mes-hall] (2022). 77. T. Jiang et al., Four-band non-Abelian topological insulator and its experimental realization. Nat. Commun. 12, 6471 (2021). doi: 10.1038/s41467-021-26763-1; pmid: 34753932 78. K. Ding, C. Fang, G. Ma, Non-Hermitian topology and exceptional-point geometries. Nat. Rev. Phys. 4, 745–760 (2022). doi: 10.1038/s42254-022-00516-5 79. C. Fang, Y. Chen, H.-Y. Kee, L. Fu, Topological nodal line semimetals with and without spin-orbital coupling. Phys. Rev. B 92, 081201 (2015). doi: 10.1103/PhysRevB.92.081201 80. C. Fang, H. Weng, X. Dai, Z. Fang, Topological nodal line semimetals. Chin. Phys. B 25, 117106 (2016). doi: 10.1088/ 1674-1056/25/11/117106 81. S. Fumeron, B. Berche, Introduction to topological defects: From liquid crystals to particle physics. Eur. Phys. J. Spec. Top. 232, 1813–1833 (2023). doi: 10.1140/epjs/ s11734-023-00803-x 82. N. D. Mermin, The topological theory of defects in ordered media. Rev. Mod. Phys. 51, 591–648 (1979). doi: 10.1103/ RevModPhys.51.591 83. W. F. Basener, Topology and Its Applications (Wiley, 2006). 84. M. Kléman, L. Michel, G. Toulouse, Classification of topologically stable defects in ordered media. J. Physique Lett. 38, 195–197 (1977). doi: 10.1051/ jphyslet:019770038010019500 85. M. Staley, Understanding quaternions and the Dirac belt trick. Eur. J. Phys. 31, 467–478 (2010). doi: 10.1088/0143-0807/ 31/3/004 86. R. Jackiw, C. Rebbi, Solitons with fermion number ½. Phys. Rev. D Part. Fields 13, 3398–3409 (1976). doi: 10.1103/ PhysRevD.13.3398 87. Y. Hatsugai, Chern number and edge states in the integer quantum Hall effect. Phys. Rev. Lett. 71, 3697–3700 (1993). doi: 10.1103/PhysRevLett.71.3697; pmid: 10055049 88. M. Z. Hasan, C. L. Kane, Colloquium: Topological insulators. Rev. Mod. Phys. 82, 3045–3067 (2010). doi: 10.1103/ RevModPhys.82.3045 89. G. M. Graf, M. Porta, Bulk-edge correspondence for twodimensional topological insulators. Commun. Math. Phys. 324, 851–895 (2013). doi: 10.1007/s00220-013-1819-6 90. X. Huang, Y. Lai, Z. H. Hang, H. Zheng, C. T. Chan, Dirac cones induced by accidental degeneracy in photonic crystals and zero-refractive-index materials. Nat. Mater. 10, 582–586 (2011). doi: 10.1038/nmat3030; pmid: 21623377

RES EARCH | R E V I E W

139.

140.

141.

142.

143.

144.

145.

147.

149.

151.

152.

153.

155.

158.

159.

160.

Yang et al., Science 383, eadf9621 (2024)

166.

167.

168.

169.

170.

171.

172.

173.

174.

175.

176.

177.

178.

179.

180.

23 February 2024

AC KNOWLED GME NTS

Funding: We acknowledge support from the National Key R&D Program of China grant 2022YFA1404400; the National Natural Science Foundation of China (NSFC) Excellent Young Scientists Fund (HKU 12222417); NSFC grant 12322412; Hong Kong Research Grants Council 12302420, 12301822, 16307522, 16307821, 16307621, RFS2223-2S01, AoE/P-502, and STG3/E-704/23-N; the New Cornerstone Science Foundation; the start-up fund of The University of Hong Kong; and an Asian Young Scientist Fellowship. Competing interests: The authors declare no competing interests. License information: Copyright © 2024 the authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original US government works. https://www.science.org/about/science-licenses-journalarticle-reuse Submitted 23 November 2022; accepted 17 January 2024 10.1126/science.adf9621

14 of 14

,

157.

165.

y

156.

164.

y g

154.

163.

y

150.

162.

181. D. L. Sounas, A. Alu’, Non-reciprocal photonics based on time modulation. Nat. Photonics 11, 774–783 (2017). doi: 10.1038/ s41566-017-0051-x 182. X. Guo, Y. Ding, Y. Duan, X. Ni, Nonreciprocal metasurface with space-time phase modulation. Light Sci. Appl. 8, 123 (2019). doi: 10.1038/s41377-019-0225-z; pmid: 31871675 183. W. Yang et al., A self-biased non-reciprocal magnetic metasurface for bidirectional phase modulation. Nat. Electron. 6, 225–234 (2023). doi: 10.1038/ s41928-023-00936-w 184. M. Burrello, A. Trombettoni, Non-Abelian anions from degenerate landau levels of ultracold atoms in artificial gauge potentials. Phys. Rev. Lett. 105, 125304 (2010). doi: 10.1103/ PhysRevLett.105.125304; pmid: 20867652 185. T. Keilmann, S. Lanzmich, I. McCulloch, M. Roncaglia, Statistically induced phase transitions and anyons in 1D optical lattices. Nat. Commun. 2, 361 (2011). doi: 10.1038/ ncomms1353; pmid: 21694709 186. E. Kapit, P. Ginsparg, E. Mueller, Non-Abelian braiding of lattice bosons. Phys. Rev. Lett. 108, 066802 (2012). doi: 10.1103/PhysRevLett.108.066802; pmid: 22401101 187. L. Yuan, M. Xiao, S. Xu, S. Fan, Creating anyons from photons using a nonlinear resonator lattice subject to dynamic modulation. Phys. Rev. A 96, 043864 (2017). doi: 10.1103/ PhysRevA.96.043864 188. I. Carusotto, C. Ciuti, Quantum fluids of light. Rev. Mod. Phys. 85, 299–366 (2013). doi: 10.1103/RevModPhys.85.299 189. X.-W. Luo et al., Quantum simulation of 2D topological physics in a 1D array of optical cavities. Nat. Commun. 6, 7704 (2015). doi: 10.1038/ncomms8704; pmid: 26145177 190. X.-F. Zhou et al., Dynamically manipulating topological physics and edge modes in a single degenerate optical cavity. Phys. Rev. Lett. 118, 083603 (2017). doi: 10.1103/ PhysRevLett.118.083603; pmid: 28282161 191. F. Cardano et al., Quantum walks and wavepacket dynamics on a lattice with twisted photons. Sci. Adv. 1, e1500087 (2015). doi: 10.1126/sciadv.1500087; pmid: 26601157 192. F. Cardano et al., Detection of Zak phases and topological invariants in a chiral quantum walk of twisted photons. Nat. Commun. 8, 15516 (2017). doi: 10.1038/ncomms15516; pmid: 28569741 193. B. Wang, T. Chen, X. Zhang, Experimental observation of topologically protected bound states with vanishing Chern numbers in a two-dimensional quantum walk. Phys. Rev. Lett. 121, 100501 (2018). doi: 10.1103/PhysRevLett.121.100501; pmid: 30240260 194. M. Yang et al., Topological band structure via twisted photons in a degenerate cavity. Nat. Commun. 13, 2040 (2022). doi: 10.1038/s41467-022-29779-3; pmid: 35440661 195. M. Yang et al., Realization of exceptional points along a synthetic orbital angular momentum dimension. Sci. Adv. 9, eabp8943 (2023). doi: 10.1126/sciadv.abp8943; pmid: 36696496 196. Z. Zhang et al., Spin-orbit microlaser emitting in a fourdimensional Hilbert space. Nature 612, 246–251 (2022). doi: 10.1038/s41586-022-05339-z; pmid: 36385532

g

148.

161.

271–275 (2022). doi: 10.1038/s41586-022-04796-w; pmid: 35831605 Q. Zhang et al., Observation of acoustic non-Hermitian Bloch braids and associated topological phase transitions. Phys. Rev. Lett. 130, 017201 (2023). doi: 10.1103/ PhysRevLett.130.017201; pmid: 36669209 Z. Li, K. Ding, G. Ma, Eigenvalue knots and their isotopic equivalence in three-state non-Hermitian systems. Phys. Rev. Res. 5, 023038 (2023). doi: 10.1103/ PhysRevResearch.5.023038 H. Hu, E. Zhao, Knots and non-Hermitian Bloch bands. Phys. Rev. Lett. 126, 010401 (2021). doi: 10.1103/ PhysRevLett.126.010401; pmid: 33480787 C.-X. Guo, S. Chen, K. Ding, H. Hu, Exceptional non-Abelian topology in multiband non-Hermitian systems. Phys. Rev. Lett. 130, 157201 (2023). doi: 10.1103/ PhysRevLett.130.157201; pmid: 37115861 J. L. K. König, K. Yang, J. C. Budich, E. J. Bergholtz, Braid-protected topological band structures with unpaired exceptional points. Phys. Rev. Res. 5, L042010 (2023). doi: 10.1103/PhysRevResearch.5.L042010 C. Dembowski et al., Experimental observation of the topological structure of exceptional points. Phys. Rev. Lett. 86, 787–790 (2001). doi: 10.1103/PhysRevLett.86.787; pmid: 11177940 C. Dembowski et al., Encircling an exceptional point. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 69, 056216 (2004). doi: 10.1103/PhysRevE.69.056216; pmid: 15244913 T. Gao et al., Observation of non-Hermitian degeneracies in a chaotic exciton-polariton billiard. Nature 526, 554–558 (2015). doi: 10.1038/nature15522; pmid: 26458102 K. Ding, G. Ma, M. Xiao, Z. Zhang, C. T. Chan, Emergence, coalescence, and topological properties of multiple exceptional points and their experimental realization. Phys. Rev. X 6, 021007 (2016). doi: 10.1103/ PhysRevX.6.021007 S. T. Schindler, C. M. Bender, Winding in non-Hermitian systems. J. Phys. A Math. Theor. 51, 055201 (2017). doi: 10.1088/1751-8121/aa9faf W. Tang, K. Ding, G. Ma, Direct measurement of topological properties of an exceptional parabola. Phys. Rev. Lett. 127, 034301 (2021). doi: 10.1103/PhysRevLett.127.034301; pmid: 34328755 Q. Zhong, M. Khajavikhan, D. N. Christodoulides, R. El-Ganainy, Winding around non-Hermitian singularities. Nat. Commun. 9, 4808 (2018). doi: 10.1038/s41467-01807105-0; pmid: 30442951 M. Berry, R. Uzdin, Slow non-Hermitian cycling: Exact solutions and the Stokes phenomenon. J. Phys. A Math. Theor. 44, 435303 (2011). doi: 10.1088/1751-8113/44/43/435303 J. Doppler et al., Dynamically encircling an exceptional point for asymmetric mode switching. Nature 537, 76–79 (2016). doi: 10.1038/nature18605; pmid: 27454554 H. Xu, D. Mason, L. Jiang, J. G. Harris, Topological energy transfer in an optomechanical system with exceptional points. Nature 537, 80–83 (2016). doi: 10.1038/nature18604; pmid: 27454555 M. Parto, C. Leefmans, J. Williams, F. Nori, A. Marandi, NonAbelian effects in dissipative photonic topological lattices. Nat. Commun. 14, 1440 (2023). doi: 10.1038/s41467-02337065-z; pmid: 36922499 R.-J. Slager, A. Bouhon, F. N. Ünal, Non-Abelian Floquet braiding and anomalous Dirac string phase in periodically driven systems. Nat. Commun. 15, 1144 (2024). doi: 10.1038/ s41467-024-45302-2; pmid: 38326295 L. Jin, Z. Song, Bulk-boundary correspondence in a nonHermitian system in one dimension with chiral inversion symmetry. Phys. Rev. B 99, 081103 (2019). doi: 10.1103/ PhysRevB.99.081103 C. H. Lee, R. Thomale, Anatomy of skin modes and topology in non-Hermitian systems. Phys. Rev. B 99, 201103 (2019). doi: 10.1103/PhysRevB.99.201103 Z. Pang, B.T.T. Wong, J. Hu, Y. Yang, Synthetic non-Abelian gauge fields for non-Hermitian systems. Phys. Rev. Lett. 132, 043804 (2024). doi: 10.1103/PhysRevLett.132.043804

p

146.

Phys. Rev. B 100, 075112 (2019). doi: 10.1103/ PhysRevB.100.075112 W. Zhang, F. Di, X. Zheng, H. Sun, X. Zhang, Hyperbolic band topology with non-trivial second Chern numbers. Nat. Commun. 14, 1083 (2023). doi: 10.1038/s41467-023-36767-8; pmid: 36841813 P. Boross, J. K. Asbo’th, G. Sze’chenyi, L. Oroszla’ny, A. Pa’lyi, Poor man’s topological quantum gate based on the SuSchrieffer-Heeger model. Phys. Rev. B 100, 045414 (2019). doi: 10.1103/PhysRevB.100.045414 Y. Wu, H. Liu, J. Liu, H. Jiang, X. C. Xie, Double-frequency Aharonov-Bohm effect and non-Abelian braiding properties of Jackiw-Rebbi zero-mode. Natl. Sci. Rev. 7, 572–578 (2020). doi: 10.1093/nsr/nwz189; pmid: 34692076 R. Citro, M. Aidelsburger, Thouless pumping and topology. Nat. Rev. Phys. 5, 87–101 (2023). doi: 10.1038/s42254-02200545-0 C. D. Brown et al., Direct geometric probe of singularities in band structure. Science 377, 1319–1322 (2022). doi: 10.1126/ science.abm6442; pmid: 36108029 C. Nayak, S. H. Simon, A. Stern, M. Freedman, S. D. Sarma, Non-Abelian anyons and topological quantum computation. Rev. Mod. Phys. 80, 1083–1159 (2008). doi: 10.1103/ RevModPhys.80.1083 C.-Y. Hou, C. Chamon, C. Mudry, Electron fractionalization in two-dimensional graphenelike structures. Phys. Rev. Lett. 98, 186809 (2007). doi: 10.1103/PhysRevLett.98.186809; pmid: 17501599 X. Gao et al., Dirac-vortex topological cavities. Nat. Nanotechnol. 15, 1012–1018 (2020). doi: 10.1038/s41565020-0773-7; pmid: 33077965 C.-W. Chen et al., Mechanical analogue of a Majorana bound State. Adv. Mater. 31, e1904386 (2019). doi: 10.1002/ adma.201904386; pmid: 31682285 T. Iadecola, T. Schuster, C. Chamon, Non-Abelian braiding of light. Phys. Rev. Lett. 117, 073901 (2016). doi: 10.1103/ PhysRevLett.117.073901; pmid: 27563965 E. Prodan, K. Dobiszewski, A. Kanwal, J. Palmieri, C. Prodan, Dynamical Majorana edge modes in a broad class of topological mechanical systems. Nat. Commun. 8, 14587 (2017). doi: 10.1038/ncomms14587; pmid: 28230164 Y. Barlas, E. Prodan, Topological braiding of non-Abelian midgap defects in classical metamaterials. Phys. Rev. Lett. 124, 146801 (2020). doi: 10.1103/PhysRevLett.124.146801; pmid: 32338983 K. Kawabata, K. Shiozaki, M. Ueda, M. Sato, Symmetry and topology in non-Hermitian physics. Phys. Rev. X 9, 041015 (2019). doi: 10.1103/PhysRevX.9.041015 Y. Ashida, Z. Gong, M. Ueda, Non-Hermitian physics. Adv. Phys. 69, 249–435 (2020). doi: 10.1080/ 00018732.2021.1876991 E. J. Bergholtz, J. C. Budich, F. K. Kunst, Exceptional topology of non-Hermitian systems. Rev. Mod. Phys. 93, 015005 (2021). doi: 10.1103/RevModPhys.93.015005 C. C. Wojcik, X.-Q. Sun, T. Bzdušek, S. Fan, Homotopy characterization of non-Hermitian Hamiltonians. Phys. Rev. B 101, 205417 (2020). doi: 10.1103/PhysRevB.101.205417 Z. Li, R. S. Mong, Homotopical characterization of nonHermitian band structures. Phys. Rev. B 103, 155129 (2021). doi: 10.1103/PhysRevB.103.155129 W. Tang et al., Exceptional nexus with a hybrid topological invariant. Science 370, 1077–1080 (2020). doi: 10.1126/ science.abd8872; pmid: 33243886 W. Tang, K. Ding, G. Ma, Experimental realization of nonAbelian permutations in a three-state non-Hermitian system. Natl. Sci. Rev. 9, nwac010 (2022). doi: 10.1093/nsr/nwac010; pmid: 36523566 K. Wang, A. Dutt, C. C. Wojcik, S. Fan, Topological complexenergy braiding of non-Hermitian bands. Nature 598, 59–64 (2021). doi: 10.1038/s41586-021-03848-x; pmid: 34616054 X.-Q. Sun, C. C. Wojcik, S. Fan, T. Bzdušek, Alice strings in non-Hermitian systems. Phys. Rev. Res. 2, 023226 (2020). doi: 10.1103/PhysRevResearch.2.023226 Y. S. S. Patil et al., Measuring the knot of non-Hermitian degeneracies and non-commuting braids. Nature 607,

RES EARCH

SOX9 switch links regeneration to fibrosis at the single-cell level in mammalian kidneys Shikhar Aggarwal†, Zhanxiang Wang†, David Rincon Fernandez Pacheco†, Anna Rinaldi, Alex Rajewski, Jasper Callemeyn, Elisabet Van Loon, Baptiste Lamarthée, Ambart Ester Covarrubias, Jean Hou, Michifumi Yamashita, Haruhiko Akiyama, S. Ananth Karumanchi, Clive N. Svendsen, Paul W. Noble, Stanley C. Jordan, Joshua J. Breunig, Maarten Naesens, Pietro E. Cippà, Sanjeev Kumar*

INTRODUCTION: In mammalian organs, including the kidney, some injured tissue displays scarless recovery, whereas adjacent areas exhibit impaired regeneration and progressive fibrosis, highlighting inefficient tissue regeneration capabilities. The precise mechanism through which the regeneration response culminates in fibrosis while at the same time driving scarless tissue restoration in the same microenvironment is unclear.

SOX9

SOX9on

Proximal tubular epithelial cells SOX9neg SOX9 /CDH6neg

SOX9

vated SOX9 early (SOX9on cells) in ischemiareperfusion injury– and rhabdomyolysis-induced AKI to CKD models revealed SOX9on-off and SOX9on-on lineages. We identified a continuum of dynamic SOX9 activity tightly linked to the restorative status of apicobasolateral polarity in these lineages. In vitro, cells lacking tight junction activated SOX9, which then switched off upon monolayer formation. Single-cell RNA

Fibroblast

NaKATPase

SOX9on-on

CDH6

Early injury/ repair response Injury reactivates SOX9 SOX9on/CDH6neg

SOX9-lineage upon regeneration /CDH6neg SOX9

SOX9-lineage with unrestored polarity/cell-cell adhesion leads to SOX9on-on/CDH6+ cell state CDH6+ WNTs+ cells +

SMA+

Unrestored apicobasolateral polarity

SOX9

SOX9on

SOX9

Reprogramming to a nephron progenitor-like state SOX9on-on

CDH6+

▪

Fibrosis +++ Dynamic SOX9 activity determines regeneration with or without fibrosis. Comparison between the two initially committed lineages for epithelium regeneration uncovered SOX9on → SOX9on-off status upon regeneration, and SOX9on → SOX9on-on activity in lineages with unrestored apicobasolateral polarity (demarcated by SOX9on-on induced CDH6+ cells). Single SOX9+CDH6+ cells reprogram to a progenitor-like state, producing ample WNTs to drive fibrosis during sustained effort to regenerate the epithelium through SOX9on-on activity. Aggarwal et al., Science 383, 845 (2024)

23 February 2024

The list of author affiliations is available in the full article online. *Corresponding author. Email: [email protected] †These authors contributed equally to this work. Cite this article as S. Aggarwal et al., Science 383, eadd6371 (2024). DOI: 10.1126/science.add6371

READ THE FULL ARTICLE AT https://doi.org/10.1126/science.add6371 1 of 1

,

the epithelium (SOX9on-on)

of epithelial restorative status, the activity of which connected regeneration to fibrosis at the singlecell level in mammalian kidneys. Our study revealed that in contrast to SOX9on/SOX9on-off cells, SOX9on-on activity highlighted by SOX9+CDH6+ cell state, during its sustained efforts to regenerate the lineage, generated substantial singlecell WNT activity to activate adjacent fibroblasts, thus driving AKI to CKD. Therefore, we have also identified how WNTs might be spatiotemporally regulated in the damaged tissue microenvironment. More work will be needed to elucidate how apicobasolateral polarity and tight junctions regulate the SOX9 switch.

y

Successful regeneration

+

CONCLUSION: We uncovered a dynamic sensor

y

(PTEC) injury and cell death leads to acute kidney injury (AKI). Post-AKI fibrosis can progress to chronic kidney disease (CKD) and end-stage kidney failure. After AKI, injured PTECs rapidly reactivate SOX9 to repair the epithelium. We hypothesized that the lineages of PTECs reactivating SOX9 would subsequent-

RESULTS: Lineage tracing of cells that acti-

g

RATIONALE: Proximal tubular epithelial cell

ly silence SOX9 upon regeneration, whereas those unable to fully restore the epithelium would maintain SOX9 activity. If confirmed, this would establish a model system to compare regeneration in two initially committed lineages, one efficiently restoring (SOX9on-off) and the other unable to fully restore the epithelium (SOX9on-on) from the onset of a single insult.

y g

NEPHROLOGY

sequencing (scRNA-seq) revealed SOX9on-on cells in a distinct regenerating state demarcated by Cadherin 6 (CDH6). The SOX9on-off lineages healed without fibrosis, whereas adjacent SOX9on-onCDH6+ lineages displayed robust, intimate association with myofibroblasts. Integration of SOX9 cell-specific bulk RNA-seq with scRNA-seq analysis unveiled single Sox9+Cdh6+Sox11+Wnt4pos- and Wnt7bposexpressing cells. Myofibroblasts were identified as injury-induced WNT-responsive cells (WRCs) located adjacent to the SOX9on-onCDH6+ cells. Lineage-tracing WRCs and Ctnnb1-deficient myofibroblasts highlighted WRCs’ central role in fibrosis generation and maintenance. Removing the WNT secretory apparatus from SOX9on-on cells reduced myofibroblasts. Further, removal of SOX9on-on activity reduced fibrosis, WNT4 ligand, and the identified early WRC molecular signature (Axin2+Nkd1+Nkd2+). Integration of scRNA-seq and SOX9 lineage– specific, single-nuclei assay for transposaseaccessible chromatin sequencing (snATAC-seq) analysis revealed the reprogramming of Sox9on-on nuclei to a progenitor-like state, whereas the Sox9on-off nuclei regained the epigenetic landscape of healthy, mature PTECs. Time-resolved, lineage-specific SOX9 genomic occupancy assay directly connected SOX9on and SOX9on-on activity to epithelial regeneration through a distinct time-specific, genomic occupancy profile, and uncovered SOX9on-on specific target genes associated with chromatin remodeling, a stem cell–like state, and WNT signaling. Wnt2 was identified as a direct SOX9on-on activity–specific target gene, with a marked reduction in Wnt2 after removal of SOX9on-on activity. In transplanted human kidneys, the cluster containing SOX9+CDH6+WNT2B+ single cells strongly correlated with fibrosis.

p

RESEARCH ARTICLE SUMMARY

◥

RES EARCH

RESEARCH ARTICLE

◥

NEPHROLOGY

SOX9 switch links regeneration to fibrosis at the single-cell level in mammalian kidneys Shikhar Aggarwal1,2†, Zhanxiang Wang1,2†, David Rincon Fernandez Pacheco1,2†, Anna Rinaldi3, Alex Rajewski4, Jasper Callemeyn5, Elisabet Van Loon5, Baptiste Lamarthée5‡, Ambart Ester Covarrubias6, Jean Hou7, Michifumi Yamashita7, Haruhiko Akiyama8, S. Ananth Karumanchi2,6, Clive N. Svendsen1,2, Paul W. Noble9, Stanley C. Jordan6, Joshua J. Breunig1,2, Maarten Naesens5, Pietro E. Cippà3,10, Sanjeev Kumar1,2,6*

23 February 2024

1 of 13

,

Aggarwal et al., Science 383, eadd6371 (2024)

First, injury-induced SOX9+ cells were fate-mapped, using previously validated Sox9IRES-CreERT2/+: R26RtdT/+ animals in a model of long-term survival–compatible bilateral ischemia reperfusion injury (IRI) that induced AKI to progress to CKD (5, 10). In human AKI, return of serum creatinine (a biochemical marker of kidney function) to normal baseline level reflects recovery from acute tubular necrosis through regeneration of tubular epithelial cells. In our

y

*Corresponding author. Email: [email protected] †These authors contributed equally to this work. ‡Present address: Université de Franche Comté, EFS BFC, Inserm Right, 25000 Besançon, France.

Identification of a dynamic SOX9/CDH6 switch within the repairing epithelia

y g

Board of Governors Regenerative Medicine Institute, CedarsSinai Medical Center, Los Angeles, CA 90048, USA. Department of Biomedical Sciences, Cedars-Sinai Medical Center, Los Angeles, CA 90048, USA. 3Division of Nephrology, Ente Ospedaliero Cantonale, CH-6900 Lugano, Switzerland. 4Applied Genomics, Computation, and Translational Core, Cedars-Sinai Medical Center, Los Angeles, CA 90048, USA. 5Department of Microbiology, Immunology and Transplantation, KU Leuven, BE-3000 Leuven, Belgium. 6 Division of Nephrology, Department of Medicine, CedarsSinai Medical Center, Los Angeles, CA 90048, USA. 7 Department of Pathology and Laboratory Medicine, CedarsSinai Medical Center, Los Angeles, CA 90048, USA. 8 Department of Orthopaedic Surgery, Gifu University Graduate School of Medicine, Gifu 501-1194, Japan. 9Division of Pulmonary and Critical Care Medicine, Department of Medicine, Women’s Guild Lung Institute, Cedars-Sinai Medical Center, Los Angeles, CA 90048, USA. 10Faculty of Biomedical Sciences, Università della Svizzera Italiana, CH6900 Lugano, Switzerland. 2

y

1

might only be a few cell distances away heal without fibrosis, have remained elusive. Our previous work, substantiated by other studies, identified Sox9 (a SRY-related highmobility-group box family of transcription factor) activation as a fundamental epithelial injury– induced repair response (Sox9+) (10, 11), and proposed that Sox9 activity observed 4 weeks after AKI might represent regions of unresolved injury and repair processes. However, the precise characterization of such processes, including their relationship to the initial Sox9+ cell, remained unexamined. To investigate this, we first tested our hypothesis that the descendants of Sox9+ cells would silence Sox9 upon regeneration, whereas persistence of Sox9 activity was predicted for the Sox9 lineage with features of unrestored epithelia. If true, then this would provide an opportunity to conduct head-to-head, spatiotemporal examination of the regenerated lineage compared with the progeny that could not fully restore the epithelia right from the onset of a single-inciting insult within with same microenvironment.

g

D

epending upon the nature and severity of the initial injury, the damaged tissue microenvironment of adult mammalian organs can demonstrate progressive scarring, whereas other foci exhibit scarless tissue recovery (1–3). This spatial heterogeneity in interstitial fibrosis seen across diverse organs is also frequently observed during progression of acute kidney injury (AKI) to chronic kidney disease (CKD) (4, 5). AKI, defined by an abrupt decline in kidney function, is often caused by acute proximal tubular epithelial cell injury (6, 7) with selective damage per se able to drive interstitial fibrosis (8). Failed tubule recovery has been broadly linked with post-AKI fibrosis (9). However, the precise mechanism that determines why one focus progressively scars, whereas other initially injured foci that

p

The steps governing healing with or without fibrosis within the same microenvironment are unclear. After acute kidney injury (AKI), injured proximal tubular epithelial cells activate SOX9 for self-restoration. Using a multimodal approach for a head-to-head comparison of injury-induced SOX9 lineages, we identified a dynamic SOX9 switch in repairing epithelia. Lineages that regenerated epithelia silenced SOX9 and healed without fibrosis (SOX9on-off). By contrast, lineages with unrestored apicobasal polarity maintained SOX9 activity in sustained efforts to regenerate, which were identified as a SOX9on-on Cadherin6pos cell state. These reprogrammed cells generated substantial single-cell WNT activity to provoke a fibroproliferative response in adjacent fibroblasts, driving AKI to chronic kidney disease. Transplanted human kidneys displayed similar SOX9/CDH6/WNT2B responses. Thus, we have uncovered a sensor of epithelial repair status, the activity of which determines regeneration with or without fibrosis.

model, serum creatinine returned to near normal levels by day 10. Therefore, the day 10 damaged tissue microenvironment, which correlated with the clinically meaningful kidney function recovery, was investigated first (Fig. 1A and fig. S1, A and B). A strong basolateral expression of sodium-potassium ATPase ion pumps (ATP1A1basolateral/high) and a thick, apical lotus tetragonoglobus lectin–enriched brush border (LTLthick) characterized uninjured proximal tubular epithelia (fig. S1C). Consistent with our previous report (10), injured epithelial cells with disrupted basolateral polarity activated SOX9, with tdT+ cells reporting such cells at 48 hours after IRI (Fig. 1A). On day 10 after injury, the initially labeled tdT+ cells had expanded (fig. S1D), and nearly the entire Sox9 lineage, which regenerated the epithelia marked by restoration of apicobasolateral polarity, had silenced SOX9 (tdT+Sox9neg; Fig. 1A and fig. S1, E and F). Conversely, the Sox9 lineage that exhibited unrestored epithelia, marked by cytoplasmic flattening and disrupted apicobasolateral polarity, maintained the SOX9+ state, thereby demonstrating sustained Sox9 activity (IRI day 10: SOX9+tdT+; Fig. 1A; fig. S1, F to H; and tables S1 and S2). The unrestored lineages were mainly observed within the proximal tubules of the outer medullary region, which undergo relatively extensive cell loss due to exaggerated susceptibility to injury (7) (fig. S1H). A subset of such lineages displayed morphological features typically associated with tubulogenesis, such as prominent cytoplasmic projections with elongated nuclei suggestive of filopodia formation, likely reflecting an attempt to repopulate the extensively denuded tubule, as well as luminal clusters with long cellular axis oriented in all directions (12) (Fig. 1A and fig. S1, G and H). Similar to IRIinduced AKI, we observed distinct Sox9 lineages with dynamic Sox9 activity tightly linked to their restorative status in rhabdomyolysisinduced AKI (rhabdo-AKI), another distinct and clinically relevant model of toxic AKI (fig. S2, A and B, and table S2). Therefore, hereafter, the terms Sox9on, Sox9on-off, and Sox9on-on will be used to describe the following cells: Sox9on for the initially injured proximal tubular epithelial cells (PTECs) that activated SOX9 at 48 hours after injury, Sox9on-off for the Sox9on lineage that regenerated the epithelia, and Sox9on-on for the counterpart that weren’t able to restore the epithelia by day 10 after AKI. To study the Sox9on and Sox9on-on cells, we performed single-cell RNA sequencing (scRNAseq) 48 hours and 10 days after IRI (Fig. 1B and figs. S3 and S4). After confirming cell-typespecific distinct clustering of single-cell datasets (Fig. 1B and figs. S3 and S4), we focused on the proximal tubule (fig. S5A). Single-cell analysis showed heterogeneous Sox9+ cells at 48 hours, whereas the day 10 population demonstrated relative homogeneity (Fig. 1C and

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

A

48h

IRI: day 0 48h

TAM

or

d10 trace

day 10

ATP1A1 tdT SOX9 DAPI

Sox9IRES-CreERT2/+:R26RtdT/+

SOX9 activity within its lineage at day10 post IRI IRI day 10: % tdT+ cells

ATP1A1pos ATP1A1neg SOX9 pos

SOX9 on C57BL6/J

IRI

D PTECs: day10 Sox9 pos vs. 48h Sox9 pos

48h d10 Single cell sequencing

Sox9

Antigen processing and presentation of exogenous peptide antigen via MHC class II

MHC protein complex Flotillin complex

day 10

48h

Positive regulation of macrophage cytokine production

Microvillus membrane Fascia adherens

Distal tubule development Collecting duct develpment

Apicolateral plasma membrane

Nephric duct morphogenesis

Catenin complex Positive regulation of axon regeneration

Extracellular exosome Microvillus Brush border membrane Lamellipodium

Average expression

E

Sox9

Lcn2

8 12 4 Enrichment Score

8 12 16 4 Enrichment Score

75 100 % Expressed Fascia adherens

F

Distal tubule development

Cdh6

g

Catenin complex

Atp6v1g3

Calb1

Scnn1b

Slc14a2

Mki67

Havcr1

Slc12a1

Lrp2

Hnf4a

Slc5a12

0

25 50

Mesenchymal to epithelial transition involved in metanephros morphogenesis Negative regulation of programmed necrotic cell death Cell differentiation involved in metanephros development 0

p

Podocytes Interstitial cells Immune Endothelial Injured distal tubules Intercalated cells Principal cells Connecting tubule Distal convoluted tubule Ascending thick timb Descending thin limb Cycling proximal tubule Injured proximal tubule PTEC Lrp2- S2/3 PTEC Lrp2+ S2/3 Proximal tubule - S1 (PT epithelial cells, PTEC)

SOX9 on-on

C

d0

IRI

Slc7a13

B

SOX9 on-off

48h

Sox9: day10

day 10

Sox9: 48h Cdh6 Ctnna1 Ctnnb1 Ctnnd1 Smad7

Ctnnb1 Dsp

Sptan1

Pax8

Mesenchymal to epithelial transition involved in metanephros morphogenesis

Apicolateral plasma membrane

y

0.5

Sox9: day10

0.0

Sox9: 48h Cldn3 Cldn7

- 0.5

Bmp4 Ctnnb1 Pax8

day 10

48h

I

ACTB

d10 trace

48h

day 10

ATP1A1 CDH6 tdT DAPI

1

2

Exp.

1

ATP1A1 CDH6

2

1

2

SOX9

,

SC C

y

IRI day10: % of tdTposCDH6pos cells acc to A ATP1A1 status

CDH6

SC C

SC = subconfluent C = confluent

ATP1A1neg

Fig. 1. scRNA-seq reveals a dynamic SOX9/CDH6 switch within the repairing nephron epithelium. (A) Identification of a dynamic SOX9 switch. Shown are the schema of lineage-tracing of injury-induced SOX9pos cells and coimmunoanalysis for SOX9 and the basolateral polarity marker ATP1A1, showing that cells with disrupted basolateral polarity activated SOX9. Note the co-localization of tdT with SOX9pos cells at 48 hours after IRI (IRI 48 hours: SOX9on cells, yellow arrows). By day 10 after IRI, the lineage of SOX9on cells that restored basolateral polarity had silenced SOX9 (SOX9on-off cell state, regenerated epithelia), whereas the lineage with unrestored ATP1A1 maintained SOX9 activity (SOX9on-on cells, yellow arrows). (B) Schema of isolating single cells to Aggarwal et al., Science 383, eadd6371 (2024)

2

Relative levels

or

1

Relative levels

% of CDH6pos cells expressing SOX9

48h

ATP1A1 CDH6 tdT DAPI

H

Exp.

Subconfluent Confluent

CDH6

ACTB

IRI: day 0 TAM

Confluent

SOX9

IRI: 48h day10

Sox9IRES-CreERT2/+:R26RtdT/+

SC C

Subconfluent

y g

CDH6 SOX9

Single cell dissociation & seeding

SOX9 off

% of SOX9+ DAPI+ cells

SOX9 on

p < 0.001

SOX9ZO1DAPI

C57BL6/J

Ctnnb1 Ctnnd1 Krt8

DAPI

G

Pou3f3 Umod Z-Score Expression

23 February 2024

compare IRI 48 hours Sox9pos (n = 2) versus day 10 Sox9pos cells, with dot plot showing average gene expression values and percentage of cells expressing cell-type–specific markers of differentiation, injury (Havcr1 and Lcn2), and repair (Sox9) response. (C) Time-resolved UMAP projection of Sox9pos cells within the proximal tubule cluster. Arrows show distinct clusters at day 10 after IRI containing Sox9pos cells compared with 48 hours after IRI. (D) Topmost enriched GO cellular components and biological processes in IRI day 10 Sox9pos cells versus 48 hour Sox9pos cells. (E) Heatmap of genes driving the GO terms. Shown is the enrichment of epithelial cell-cell adhesion machinery, including Cdh6, within day 10 Sox9pos cells. (F) Time-resolved UMAP projection of Cdh6pos 2 of 13

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

cells (arrows) within the PT cluster. (G and H) Coimmunostaining for CDH6 and SOX9 (G) and ATP1A1 (H) showing CDH6 expression restricted to day 10 SOX9pos cells (G) and to the Sox9 lineage with unrestored ATP1A1-based basolateral polarity versus the lineage that restored polarity (H), indicating that CDH6 demarcates SOX9on-on cells (SOX9posCDH6pos cell state). (I) Coimmunostaining for SOX9 and the epithelial tight junction marker ZO-1 and

,

3 of 13

y

23 February 2024

y g

In IRI- and rhabdo-induced AKI to CKD models, head-to-head comparison between the two lineages showed that most of the aSMA+ myofibroblasts were conspicuously located adjacent to Sox9on-on cells (Fig. 2A; fig. S10, A to D; and tables S1 and S2). A similar robust, intimate association was observed with the CDH6pos lineage (Fig. 2B). The outer medullary region exhibited most of the aSMA response paralleling Sox9on-on activity (fig. S10, A to D). Substantial juxtaposition at single-cell spatial distance was confirmed by Sox9IRES-CreERT2/+:R26RtdT/+: Acta2-GFP [Acta2 encoding aSMA, green fluorescent protein (GFP)] animals (Fig. 2B, fig. S10C, and tables S1 and S2). To ascertain whether a distinct Acta2 negative profibrotic celltype, for example, ACTA2neg COL1A1+ cells, might encase the CDH6neg cells, we searched our scRNAseq datasets. Consistent with known responses in fibrotic kidneys, Acta2-, Col1a1-, and Col3a1-expressing cells were observed within

y

SOX9posCDH6pos cells form a central hub of myofibroblast generation and maintenance

the Pdgfrb+ interstitial cluster, the latter signature predominantly contributed by the day 10 Pdgfrb+cells (fig. S11A). Unlike the lung, no distinct Acta2negCol1a1+ or Col3a1+ demarcated cell clusters were observed (fig. S11A). These findings were cross-validated by searching published snRNA-seq datasets of post-AKI kidneys (fig. S11B). Coimmunoanalysis confirmed that interstitial cells co-expressed ACTA2 and COL1A1, with a marked paucity of COL1A1+ cells around the regenerated lineage (fig. S11C). Up to 88% of Pdgfrb+ cells that activated aSMA were the resident Pdgfrb+ fibroblasts located within single-cell distance of Sox9on-on cells (fig. S11, D and E, and table S2). The above findings highlight that Sox9on-off lineages heal without fibrosis and implicate the CDH6pos lineage (SOX9posCDH6pos) as a possible cell state that generates myofibroblasts through a short-range secretory ligand. To identify the secretory ligand(s), we next performed RNA-seq–based profiling of IRI day 14 tdT+ versus 48 hour tdT+ cells, with Slc34a1+tdT+ cells serving as normal, uninjured control PTECs (Fig. 2C and fig. S12, A and B). Sox9 transcripts were enriched within both tdT+ populations versus normal control PTECs (~3-fold and ~13-fold), respectively (Fig. 2C and fig. S12C). Quantitative polymerase chain reaction (qPCR) validated Sox9 enrichment and confirmed Cdh6 activation specifically within IRI day 14 tdT+ cells versus 48 hour tdT+ cells (Fig. 2D), demonstrating the fidelity of the Sox9 reporter animals for Sox9 activity. Genes such as Ctnnb1, Ctnnd1, Pou3f3, and Smad7, which contributed to the topmost enriched GO terms in single day 10 Sox9+cells (Fig. 1E), were also enriched within IRI day 14 tdT+ versus 48 hour tdT+ cells (fig. S12, D and E). The GO analysis enriched term “kidney development” showed Sox11 enrichment within IRI day 14 tdT+ versus 48 hour tdT+ cells (fig. S12E). The renal role of Sox11 is relatively unknown, except for its role in human and murine nephrogenesis (19). scRNA-seq analysis revealed that Sox11+ cells constituted a subset of the Cdh6pos population (Fig. 2E; see also Fig. 1, C and F). Thus, the above validatory studies confirmed that our bulk Sox9 cell-type–specific RNA-seq study could be used reliably to identify possible secretory ligands that might engage the adjacent fibroblast. The Wnt signaling pathway, consisting of Wnt ligands, was among the top enriched GO terms within IRI day 14 tdT+ cells versus 48 hour tdT+ cells, although it was conspicuously

g

Aggarwal et al., Science 383, eadd6371 (2024)

regenerated lineage silenced SOX9/CDH6 activity. Therefore, this finding supports a continuous, dynamic SOX9/CDH6 axis despite the interstitium becoming progressively fibrotic and inflamed. A similar dynamic activity was also observed in rhabdo-AKI (fig. S8). Further, on similar lines, in vitro replicating subconfluent cells lacking cell-cell contact activated SOX9 and CDH6, which returned to baseline upon quiescent, confluent, monolayer formation, the latter characterized by acquisition of ZO-1, a tight junction protein (Sox9on-off cell state; Fig. 1I). We also leveraged published databases that used single-nuclear sequencing to cross-validate the Sox9/Cdh6 switch (fig. S9A) (17, 18). Consistent with our scRNAseq analysis, singlenuclear trajectory analysis of Sox9+ nuclei demonstrated distinct Sox9 lineages. One Sox9 lineage restored the transcriptome back to normal PTECs and switched off Sox9. By contrast, another lineage, which was marked by sustained Sox9 activity, acquired a distinct transcriptome characterized by Cdh6+ nuclei (fig. S9, B to E). Thus, snRNA-seq analysis further substantiated our findings. We have therefore identified a dynamic SOX9/CDH6 switch tightly linked with the restorative status of the proximal tubule epithelia, with Sox9on-on cell state (SOX9posCDH6pos) most probably highlighting ongoing attempts to regenerate the epithelia.

p

fig. S5B). Gene ontology (GO) term analysis of the differentially expressed genes showed enrichment of terms related to the formation of proximal tubular epithelia and function in Sox9+ cells at 10 days versus 48 hours (Fig. 1D). For example, in the category “cellular component,” the terms “lamellipodium,” “fascia adherens,” and “catenin complex” were enriched, in addition to “microvillus,” the latter a characteristic of PTECs. The cadherin-catenin-actin complex and lamellipodium are essential for the formation and maturation of epithelia (13, 14) (Fig. 1D and fig. S6A). In addition to Ctnn1a and Ctnnb1, Cdh6, which encodes cadherin 6 (k-cadherin), also contributed to the enriched term “catenin complex” (Fig. 1, D to F). Cdh6 is essential for the formation of a fully polarized nephron epithelium during nephrogenesis (15). Spatiotemporal mapping confirmed that CDH6 expression was restricted to IRI day 10 SOX9+ cells and to the Sox9 lineage with Sox9on-on activity, which exhibited disrupted or absent apicobasolateral polarity (Fig. 1, G and H, and table S1). Consistent with the observed LTLlow/absent apical brush border of Sox9on-on cells, our single-cell analysis also predicted low Lrp2 (encoding the apical brush border component megalin) status of Cdh6+ cells (fig. S6A). The presence of CDH6+LRP2low cells was confirmed at day 10 after injury (fig. S6B). Additionally, the genes associated with apicobasolateral polarity (Fat1, Myo9a), both linked with proximal tubule formation and function (16), were enriched in Sox9on-on versus Sox9on cells (fig. S6C). Complementing the GO category cellular component, the biological processes terms linked with nephrogenesis (Bmp4, Ctnnb1, and Pax8) were enriched within Sox9on-on versus Sox9on cells (Fig. 1E). Haploinsufficiency of SOX9, BMP4, CTNNB1, and PAX8, respectively, led to murine and human kidney hypoplasia and/or malformations. We validated Bmp4 enrichment within Sox9on-on cells (fig. S6, D and E). Next, we fate-mapped Sox9on-on cells by injecting tamoxifen at day 10 after IRI (fig. S7). Consistent with the results of a coimmunoanalysis study (Fig. 1G), ~90% of the initially labeled cells were CDH6+. Two weeks later, at least a subset of Sox9on-on cells regenerated the epithelia, consistent with the “ongoing attempt to regenerate” prediction of the scRNAseq analysis (fig. S7, A to D, and table S2). Although the proportion of regenerated tubules in the damaged outer medullary region was smaller compared with the outer and inner cortices, the

immunoblot for SOX9 and CDH6 showing that cells that lacked cell-cell contact (subconfluent state) activated SOX9 and CDH6, which waned upon restoration of tight junctions (confluent monolayer). Box inset highlights the region depicted at high magnification. All n = 3 animals per time point unless otherwise stated. Data are shown as mean ± SEM. Scale bars, 100 mm. For total cells counted, see table S1.

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

p g y y g y ,

Fig. 2. CDH6 status is tightly linked to repair with or without aSMApos myofibroblast response through single-cell Wnt activity. (A) Schema of lineage-tracing of IRI-induced Sox9pos cells and coimmunostaining showing aSMApos myofibroblasts adjacent to SOX9postdTpos cells (Sox9on-on cells, arrows) but no detectable aSMA activity around the Sox9on-off cells (arrow). Left and right magnified panels correspond to the foci highlighted by corresponding left and right box insets. (B) Coimmunoanalysis showing aSMApos myofibroblasts encasing the CDH6pos Sox9 lineages at single-cell spatial distance; by contrast, the paucity of such response around the CDH6neg Sox9 lineage can be observed (arrow). Circle and box highlight such foci, with the magnified panel corresponding to the foci highlighted by the box inset. (C) Schemas for purifying Slc34a1pos normal PTECs (control tdTpos cells) and Sox9pos cells 48 hours and day 14 after IRI and principal components analysis (PCA) plot showing distinct Aggarwal et al., Science 383, eadd6371 (2024)

23 February 2024

transcriptomic profiles of purified tdTpos cells. Sox9 transcripts were enriched in purified tdTpos cells. (D) qPCR of purified tdTpos cells confirming Sox9 and Cdh6 enrichment. (E) UMAP projections for Sox11 at 48 hours and day 10 after IRI. Arrows highlight the distinct day 10 cluster composed of these cells. Single Sox9pos and Cdh6pos cells also contributed to this cluster (Fig. 1, B and E). (F) Volcano plot demonstrating enrichment of Wnt ligands within day 14 versus 48 hour tdTpos cells. (G) UMAP projections for Wnt4 and Wnt7b at 48 hours and day 10 after IRI. Arrows highlight the distinct day 10 cluster, which contained such single Wnt4pos and Wnt7bpos cells. Note that the same distinct cluster was composed of day 10 Sox9pos, Cdh6pos, and Sox11pos cells [see also Fig. 1, B and E, and this figure (E)]. (H) Upset plot analysis of scRNAseq datasets showing subset of single Cdh6pos cells enriched with Wnt4, including Cdh6posSox11pos subsets. (I) qPCR of purified tdTpos cells confirming Wnt4 enrichment in day 14 4 of 13

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

post-IRI tdTpos cells versus their 48-hour counterparts. (J) Schema for labeling early Axin2pos cells after IRI and immunoanalysis for CDH6 showing colocalization of tdTpos and GFPpos cells with tdTposGFPpos cells located adjacent to CDH6pos subsets. Circles and box highlight single-cell, biologically active

y g y ,

5 of 13

y

23 February 2024

no effect on overall fibrosis (26), thus ruling out stromal Wnt4 as the main driver of postAKI fibrosis. The above findings reveal the SOX9posCDH6pos cell state as the central source of continuous, biologically active Wnt signal driving fibrosis after AKI. Next, we sought to determine whether Sox9on-on activity per se is the critical component of such profibrotic niches in vivo. Our previous work highlighted the essential role of Sox9 activation in PTEC regeneration. Proximal tubule– specific removal of Sox9 before injury led to impaired tubular repair and renal function recovery compared with animals with intact Sox9 activity (10). Based on the findings of our scRNAseq and Sox9+ cell-type–specific bulk RNA-seq studies, we hypothesized that SOX9posCDH6pos cells with Sox9on-on likely recruited essential downstream proregenerative programs by day 10 after AKI. Therefore, it might be feasible to remove Sox9on-on activity during the progression of AKI to CKD. The precise molecular signature of early WRCs after AKI remains unknown. Therefore, to determine whether removal of Sox9on-on activity leads to the abrogation of fibrosis and the obliteration of Wnt niches, we established a molecular signature of WRCs. To this end, we used Axin2CreERT2/+:R26RtdT/+: Acta2-GFP animals (Fig. 3D). Comparison of the transcriptome between the earliest WRCs (tdTpos GFP GFPpos) versus other myofibroblasts (tdTneg GFPpos) cells confirmed the Wnt signaling pathway activation as one of the top-most enriched GO terms within the tdTpos cells; Axin2, Lef1, Nkd1, and Lgr6 were enriched, whereas Fzd7, encoding Frizzled 7, a Wnt receptor, was down-regulated (Fig. 3, E to G, and table S3), suggesting that Fzd7 might be the Wnt-sensing receptor activating Pdgfrb+ cells. Other top enriched GO terms included “response to fibroblast growth factor” and “regulation of mitotic cell cycle,” suggesting that the early Axin2+ myofibroblast subset may have greater proliferative properties compared with other myofibroblast populations, consistent with lineage-tracing and Ki67-based coimmunoanalysis studies (Fig. 3F and fig. S15, A and B). qPCR validated Axin2 enrichment within tdTposGFPpos versus tdTnegGFPpos cells (Fig. 3H). Our scRNA-seq study analysis not only further endorsed induction of Axin2 activity within the same cluster containing Acta2+ Col1a1+Col3a1+Pdgfrb+ cells at day 10 after IRI, but consistent with the above findings, also showed induction of Nkd1+ cells specifically restricted to the above day 10 postIRI cluster (Fig. 3I). Nkd2, reported as a marker of terminal

g

Aggarwal et al., Science 383, eadd6371 (2024)

Nearly half of the replicating myofibroblasts were Axin2+ (fig. S15B). Further, comparative blinded analysis of the b-catenin–deficient lineage of the earliest aSMA+ myofibroblasts (Ctnnb1-cKO) versus the lineage with intact b-catenin activity (Ctnnb1-wt, controls) showed that Ctnnb1-cKO displayed significantly less expansion (P < 0.01) and lacked the intense aSMA response at day 28 after IRI (fig. S15C). Moreover, the characteristic strong aSMA response adjacent to Sox9on-on cells was conspicuously deficient in most of the Ctnnb1-cKO cells (8.3 ± 2.4% versus 79.2 ± 2.6%, Ctnnb1cKO versus Ctnnb1-wt, P < 0.01; fig. S15C and table S2). Thus, canonical b-catenin signaling is essential for the maintenance of aSMA+ activity within the myofibroblasts after AKI. To determine whether SOX9on-onCDH6pos cells secrete Wnt and drive fibrosis, we performed a head-to-head comparison between tdT-demarcated Sox9+cells lacking Wls versus and the intact Wls secretory apparatus (Fig. 3A). Wls encodes Wntless, a highly conserved, transmembrane protein essential for Wnt secretion and function (25). Successful recombination of the floxed Wls allele was confirmed by two different approaches (fig. S16, A to D). Blinded analysis revealed a substantial reduction in aSMA response around the tdT+SOX9: Wls-cKO versus controls (tdT+SOX9: Wls-Het; Fig. 3A and table S1). Both groups had a similar proportion of SOX9on-on cells (Fig. 3B). Further scrutiny of Sox9: Wls-cKO animals showed that a proportion of SOX9+ cells underwent successful recombination, as shown by tdT expression, thus generating a microenvironment mosaic for Wls activity: SOX9+tdTpos (Wls-cKO) and SOX9+tdTneg (Wls-WT) cells, respectively. Leveraging the observed mosaicism, a direct-blinded comparison for aSMA activity around SOX9+tdTpos and SOX9+tdTneg tubules revealed a significant (P < 0.01) reduction in myofibroblasts around the SOX9+tdTpos tubules compared with their SOX9+tdTneg counterparts (Fig. 3C, fig. S16E, and table S1). Macrophages could be a collateral source of Wnts. We did not detect Wnt7b, Wnt9b, Wnt11, or Wnt4 induction in fluorescence-activated cell sorting (FACS)–purified LyzM Cre-derived tdT+F4/80+cells compared with uninjured resident tdT+F4/80+cells despite an ~5-fold increase in F4/80+cells (fig. S17, A to B). Consistent with this finding, blinded analysis for aSMA+ myofibroblasts or overall fibrotic responses showed no effect of Wls removal within the LyzM Cre-derived immune cells (fig. S17, C to F). Although kidney stromal cells express Wnts, removal of Wnt4 within these cells had

p

devoid of the canonical Wnt-b catenin pathway activity reporter Axin2 (Fig. 2F and fig. S12E). Mammalian Wnt proteins are a family of lipidmodified glycoproteins that signal within a typical range of just one or two cells in a juxtracrine or autocrine manner (20, 21). Integration with scRNA-seq datasets identified single cells with de novo activation of Wnt4, and Wnt7b specifically within Cdh6pos cells, including a Cdh6+Sox11+ subset that also expressed Wnt4 (Fig. 2, G and H). qPCR and RNA-scope studies confirmed Wnt4+ induction within CDH6pos cells, thus validating RNA-seq findings (Fig. 2I and fig. S12F). Using a previously validated and published WNT4 antibody (22, 23), coimmunostainings substantiated the RNA scope studies. A strong WNT4 expression was strictly restricted to Sox9on-on cells, with such cells displaying a single-cell spatial and tight association with GFPpos myofibroblasts, indicating that Sox9on-on cells maintain biologically active Wnt niches during the progression of postAKI fibrosis (fig. S13, A and B, and table S2). Thus, the SOX9posCDH6pos cell state might be the secretory cells that generate and maintain Wnt-enriched niches after AKI. Next, we used Axin2CreERT2/+:R26RmT/mG animals, in which Axin2 activity reports cells with ongoing canonical Wnt signaling (20), to ascertain the identity of Wnt-responsive cells (WRCs). In a healthy adult kidney, a subset of interstitial aSMAneg PDGFRB+ cells residing within the inner medulla region expressed membrane GFP (mGFP), demarcating resident WRCs (fig. S14, A to C). Contrary to the previous report (24); no resident WRCs were detected within the nephron epithelia of the cortices and outer medulla (fig. S14, A to C). After AKI, the animals treated with only corn oil (the vehicle for tamoxifen) did not demonstrate mGFP+ cells, confirming tamoxifen dependence and ruling out injury-induced spontaneous Cre activation (fig. S14, D and E). The kidneys of the tamoxifen-treated animals showed that subsets of aSMA+ myofibroblasts mounted Axin2 activity, as shown by mGFP. By contrast, the resident Axin2+ cells remained aSMAneg (fig. S14, F to I). Thus, myofibroblasts represent de novo WRCs after injury. To determine the spatiotemporal relationship between the earliest WRCs, CDH6pos cells and myofibroblasts, we scrutinized Axin2CreERT2/+:R26RtdT/+:Acta2-GFP mice. Examination uncovered subsets of CDH6pos cells forming a biologically active Wnt niche at the single-cell level (Fig. 2J and table S1). Genetic lineage studies showed that the descendants of Axin2+cells formed the bulk of the scar tissue by 4 weeks after IRI (fig. S15A).

Wnt-enriched niches robustly linked with subsets of CDH6pos cells. Magnified panels correspond to the foci highlighted by box inset. All n = 3 animals per time point unless otherwise stated. Data are shown as mean ± SEM. Scale bars, 100 mm. For total cells counted, see table S1.

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

p g y y g y ,

Aggarwal et al., Science 383, eadd6371 (2024)

23 February 2024

6 of 13

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

Fig. 3. Sox9on-on cell state is the main driver of myofibroblast formation and maintenance. (A to C) Schema of Wls removal from Sox9on-on cells, outer medullary region representative image, and blinded coimmunoanalysis for SOX9 and aSMA (n = 5 animals/group). No difference can be seen in Sox9on-on cells between the two groups (B). The mosaic tissue–damaged microenvironment displayed significantly (P < 0.01) reduced aSMApos myofibroblasts adjacent to tdTpos demarcated Wls-KO tubules (arrows) versus tdTneg tubules with intact Wls [(C), arrowheads]. (D) Schema of isolating early Axin2pos cells after IRI with FACS plot showing that nearly all tdTpos cells co-expressed GFP. (E) PCA plot of purified tdTposGFPpos and tdTnegGFPpos cells. (F) GO analysis showing Wnt pathway among the top 10 terms, confirming that Axin2pos cells demarcated early WRCs. (G and H) Volcano plot showing the molecular signature of early WRCs, with qPCR confirming Axin2 enrichment within tdTposGFPpos cells, thus validating reporter animals. (I) UMAP projection of Axin2, Nkd1, and Nkd2 cells. Such cells were within the same PdgfrbposCola1posCol3a1pos cluster (see also fig. S9). (J) Experimental outline

,

7 of 13

y

23 February 2024

y g

Having identified the biological relevance of Sox9on-on activity, we next aimed to better define and understand the epigenetic features of the Sox9on-on cell state. To this end, we generated single-nuclei assay for transposaseaccessible chromatin sequencing (snATAC-seq) profiles of day 10 Sox9 descendants after injury. To ensure sufficient availability of single Sox9on-on nuclei for meaningful analysis, we dissected the inner cortices and outer medulla region, the site of predominant Sox9on-on cells, and the single-nuclei isolated from freshly enriched lineage-traced tdTpos cells were subjected to snATAC-seq (Fig. 4A). Integration of snATAC-seq with the scRNA-seq profile dem-

y

The SOX9on-onCDH6poscell state highlights sustained efforts to regenerate the epithelium by attaining a progenitor-like cell state

onstrated that the nuclei subclustered in a nephron cell-type–specific manner based on differentially open or closed chromatin accessibility state (Fig. 4B and fig. S20, A and B). The differential chromatin accessibility of Sox9 led to the clustering of lineage-traced tdTpos nuclei into “Sox9on-on” and “Sox9on-off” nuclei (Fig. 4C). The Sox9on-off nuclei exhibited a relatively open chromatin accessibility state of Hnf4a (Fig. 4C) and its downstream gene, Lrp2 (fig. S20B), the markers of mature functional PTECs, thus providing strong evidence that the Sox9 lineage regenerates functional PTECs. By contrast, the Sox9on-on nuclei demonstrated a relatively open chromatin accessibility state of Cdh6, Wnt4, and Wnt7b, in addition to Sox11 (Fig. 4, C and D, and fig. S20B), further showing the Sox9on-onCdh6pos cell state to be Wnt enriched (Fig. 4D). Consistent with scRNAseq analysis (Fig. 1B), snATACseq analysis also provided evidence for UMOD+ thick ascending limb of Loop of Henle and AQP2+ collecting duct epithelial cells to activate SOX9 after injury (fig. S20, C and D). To uncover the transcriptional regulators that characterize the distinct outcomes of Sox9-descendants, the on-on and on-off nuclei were subjected to trajectory analysis (Fig. 4, E and F). Transcriptional regulators with an essential role in nephrogenesis, such as Sox11, Nfat5, Maz, and Pax8, displayed dynamic transcriptional changes along the pseudotime in the Sox9on-on cell state (29–31) (Fig. 4G). The dynamic reparative process was associated with cell proliferation, with ~22% of Sox9on-on descendants in the S-phase of the cell cycle (Fig. 4H). Moreover, consistent with the prediction of snATACseq pseudotemporal analysis, the expression of both SMARCC1 and RUNX1 was restricted to the Sox9on-on lineage compared with its onoff counterpart (Fig. 4, I and J). Smarcc1, an ATP-dependent chromatin-remodeling complex, maintains proliferation, pluripotency, and self-renewal of embryonic stem cells (32), whereas Runx1 drives muscle regeneration

g

Aggarwal et al., Science 383, eadd6371 (2024)

activity and the main site of de novo injuryinduced Wnt4 response, displayed maximal blunting of the Wnt4 response upon removal of Sox9on-on activity (Fig. 3R). These findings provided further evidence for Sox9on-on activity in generating biologically active Wnt4 niches. The resident Wnt4 expression within the papilla and inner medulla region (a region with no Sox9on-on cells) remained intact and thus served as a robust, internal positive control (Fig. 3, R and S). Further, in addition to the panel of profibrotic genes, the herein identified early molecular signature of injuryinduced WRCs (Axin2, Nkd1, and Nkd2) showed reductions in Sox9-cKO versus Sox9-WT animals (Fig. 3T). Genetic fate mapping of the Sox9on-ondeficient cells revealed that a significantly (P < 0.01) larger proportion of Sox9on-on-deficient cells had restored polarity compared with their counterparts with intact activity (fig. S19). This finding suggests that the reduced fibrosis secondary to the removal of Sox9on-on activity might serve as a relatively favorable milieu for the epithelium to restore itself. Thus, Sox9on-on activity transforms the SOX9posCDH6pos cell state into Wnt-secreting cells, a central hub for myofibroblast formation and maintenance.

p

myofibroblasts in a previous study (27) and in our molecular profiling study of AKI to CKD (5), was also enriched in early Acta2+Col1a1+ Pdgfrb+cluster (Fig. 3I). Thus, these findings not only validated the Axin2 reporter but also established a molecular signature of early WRCs. To remove Sox9on-on activity during the transition from AKI to CKD, doxycycline injections to Sox9 knock-out (Sox9-cKO) and control wildtype (Sox9-WT, Ctrl) animals were administered 1 week after AKI. The cells expressing tdT underwent successful doxycycline-mediated recombination. First, doxycycline dependence of the system was confirmed (fig. S18A). Further, in the absence of doxycycline, Sox9-cKO animals displayed an injury-induced, early SOX9 activation response, with nearly half of SOX9+ cells expressing Ki67 akin to animals with intact Sox9 activity, thus ruling out spontaneous injuryinduced Cre activation and preservation of early Sox9 repair responses (fig. S18B). Doxycycline led to obliteration of Sox9on-on activity in the Sox9-cKO (tdT+Sox9KO) animals; by contrast, Sox9-WT animals displayed intact Sox9on-on activity (tdT+Sox9+) (fig. S18C). To ensure rigorous comparison between similar damaged tissue microenvironments, a blinded, head-to-head comparison between Havcr1+ (also called kidney injury molecule-1, Kim1) regions of Sox9-cKO and Sox9-WT animals was performed. The kidneys lacking Sox9on-on activity displayed a significantly reduced fibrotic signature, including aSMA+ myofibroblasts (Kim1+Sox9WT versus Kim1+Sox9KO, 69.0 ± 3.6% versus 28.2 ± 5.8%, respectively, P < 0.01; Fig. 3, J to O, and table S1). Before the removal of Sox9on-on activity, the proportions of SOX9 and aSMA cells were similar in Sox9-WT and Sox9-cKO animals (fig. S18D). Using previously validated shRNA to knock down Sox9 (28), we found a marked reduction in Wnt4 in primary nephron tubular epithelial cells in Sox9 knock down cells versus controls (Fig. 3, P and Q). Next, we confirmed Sox9Wnt4 link in vivo (Fig. 3, R to T). The outer medulla region, the predominant site of Sox9on-on

of removal of Sox9on-on activity during the AKI to CKD transition. (K and L) Trichrome staining (K) and quantitative scores of interstitial fibrosis (L) (blinded analysis, n = 5 Sox9-Ctrl and 6 Sox9-cKO animals). (M and N) Coimmunostaining for SOX9, KIM1, and aSMA (M) and blinded coimmunoanalysis showing head-head comparison between KIM1+ regions for aSMA activity (N) in Sox9-cKO versus Sox9-Ctrl animals. (O) qPCR analysis of genes associated with fibrosis in the kidneys from Sox9-cKO versus Sox9-Ctrl animals. (P and Q) Western blot confirming SOX9 knock-down in subconfluent primary TECs (P), with qPCR showing Sox9 and Wnt4 down-regulation (Q). (R and S) RNAscope image (R) and analysis (S) showing significant (P < 0.01) Wnt4 reduction. (T) qPCR analysis of Wnts under scrutiny and Axin2, Nkd1, and Nkd2 in the kidneys from Sox9-cKO versus Sox9-Ctrl. All images are representative images. Data are shown as mean ± SEM. *P < 0.05, **P < 0.01 unpaired two-sided Student’s t test; ***P < 0.01, paired Student’s t test. Scale bars in whole scanned images [(K) and (R)], 1000 mm; all others, 100 mm. For total cells counted, see table S1.

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

Single tdTpos nuclei isolation: snATACseq

IRI: day 0 TAM

H

PT: S2/3

Sox9IRES-CreERT2/+:R26RtdT/+ IRI: day 0 EDU

% Sox9on-on incorporating EDU

of enriched, lineage-traced Sox9 B Single-nuclei epithelial cells (tdTpos) in inner cortex and outer medulla

A Sox9IRES-CreERT2/+:R26RtdT/+

d10 trace

TAM

de novo PT cell state

d10 trace

EDU tdT SOX9 DAPI

Descending thin limb Ascending thick limb Injured distal tubule Distal convoluted tubule Connecting tubule Principal cells Intercalated cells

C

Sox9

Hnf4a

Sox9 on-on

EDU tdT SOX9 DAPI

Sox9

on-off

Log2 (NormCounts+1)

D

E

Log2 (NormCounts+1)

F Sox9 on-on

Havcr1

Sox9 on-off

p

NormCounts

G

Wnt7b

Log2 (NormCounts+1)

I

Smarcc1 Runx1 Smad2

SMARCC1 tdT CDH6 DAPI

CDH6

SMARCC1

g

Sox11

Sox9

y

Wnt4

Pax8 Nfat5 Hoxa10 Hoxb2

Log2 (NormCounts+1)

Sox9 on-on Sox9 on-on

Maz Hoxc10 Sp1

Cdh6

Sox9 on-off

Hnf4a Hnf1a Arnt Hif1a

Log2 (NormCounts+1)

J

RUNX1 tdT SOX9

RUNX1 SOX9 DAPI

Hnf1b

y g

GeneIntegrationMatrix

Log2 (NormCounts+1)

MotifMatrix

Pseudotime

y

SOX9

RUNX1

,

Fig. 4. Epigenetic reprogramming of single Sox9on-on nuclei to a nephron progenitor–like state contrasts with the Sox9on-off counterpart, which reverts to normal PTEC. (A) Schema of the workflow to obtain single nuclei of the FACS-enriched tdTpos Sox9-descendants at day 10 after injury. Rectangle (inset) highlights the dissected region showing the inner cortices and outer medulla, the site of relatively extensive PTEC loss. (B) Integration of scRNA-seq and snATAC-seq showing the cellular identity of single-tdTpos nuclei (n = 2 animals). (C) Identification of Sox9on-on and Sox9on-off single nuclei: Chromatin accessibility revealing clustering of tdTpos descendants based on relatively open and closed chromatin accessibility state of Sox9. Sox9on-off nuclei exhibited Aggarwal et al., Science 383, eadd6371 (2024)

23 February 2024

open chromatin accessibility for Hnf4a, a known marker of healthy, mature PTECs, suggesting that the Sox9 lineage that regenerated normal PTECs closed chromatin accessibility for Sox9. (D) UMAP representation of chromatin accessibility analysis showing the relatively open chromatin state of Havcr1, Wnt7b, Wnt4, and Cdh6 versus their accessibility state in Sox9on-off nuclei. (E and F) Sox9 trajectory analysis with scRNAseq-imputed gene expression (E) and pseudotime (F) scale showing the Sox9on-on → Sox9on-off transition. (G) Heatmaps of cross-platform linked genes involved in the transcriptional cascade during Sox9on-on → Sox9on-off transition. Note that the highlighted genes in Sox9on-on nuclei are linked with tissue development and/or nephrogenesis, 8 of 13

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

and the Sox9on-off nuclei displayed open chromatin accessibility for Hnf4b, another known marker of healthy, mature PTECs in addition to Hnf4a. (H) Schema of 5-ethynyl-2′-deoxyuridine (EDU) regime and coimmunostaining for SOX9 and EDU showing SOX9postdTposEDUpos cells. Two representative images are shown, with the lower panel demonstrating a cluster of such cells.

CDH6posWNT2Bpos cells marks fibrotic foci in human kidneys

y ,

9 of 13

y g

To determine the clinical relevance of our findings, we studied transplanted human kidneys. Immediately after kidney transplantation, ischemia reperfusion injury–induced AKI frequently leads to delayed graft function, which is an independent predictor of subsequent renal allograft loss and dysfunction (47). Renal allograft biopsies from such patients (n = 3, day 7 after kidney transplantation) and preimplantation biopsies obtained from the same allograft (controls) demonstrated a marked increase in SOX9+ cells within the AQP1+ PTECs (Fig. 5A). This was also confirmed at the transcriptional level in protocol biopsies obtained after reperfusion (48) (Fig. 5B). Further, a stratification of kidney allograft protocol biopsies performed 1 year after transplantation showing SOX9 levels demonstrated that patients with persistent SOX9 expression displayed increased interstitial fibrosis and reduced renal function compared with the cohort that displayed return of SOX9 activity to baseline levels (Fig. 5, C to E). SOX9 and CDH6 levels correlated with the transition from AKI to CKD in transplant biopsies according to a previously reported model (49) (Fig. 5F). In the same kidney transplant cohort, protocol biopsies obtained 3 and 12 months after transplantation revealed CDH6 as the topmost gene correlated with SOX9 activity (Fig. 5G). Such strong correlation was also noted with fibrosisassociated genes such as COL1A1 and ACTA2.

y

23 February 2024

generation and maintenance of Wnt-enriched niches. Integration of HOMER analysis of consensus sequence motifs associated with sites targeted by SOX9 CUT&RUN assay and transcription regulators uncovered by pseudotime trajectory analysis in our snATAC-seq datasets revealed top cofactors operating specifically in the Sox9on-on cell state (fig. S21M). These include binding motifs for Yy1, Nfat5, Sp1, Arnt::Hif1a, and Tcf4. Yy1, a structural regulator of enhancerpromotor loops and gene expression, is essential for embryonic stem cell viability (44), and Nfat5 is essential for nephrogenesis and protects against stress (29). Tcf4, which is essential for skin epithelia repair and homeostasis (45), was noted to be co-enriched with Wnt4 in nephron progenitors, and Hif1a-Sox9 axis is involved in chondrogenesis (46). Thus, our findings indicate that Sox9on-on activity might cooperate with these factors to regulate downstream gene expression.

g

Aggarwal et al., Science 383, eadd6371 (2024)

(Fig. 1D). Calmodulin binding was one of the top GO enriched “molecular function” terms (fig. S21, H and I). SOX factors are known to contain a calmodulin-binding domain, and this calcium ion–enabled interaction imports SOX9 to the nucleus, with subsequent target gene activation (36, 37). Most Sox9 target genes were specific either to 48 hours (91.2%) or day 10 (85.11%) after injury, indicative of distinct time-dependent genomic occupancy (fig. S21J). The top enriched GO terms at both time points showed links with epithelial development processes consistent with its role in epithelial restoration. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis revealed the top terms linked with the RhoGTPase, calcium, and epidermal growth factor receptor (EGFR) pathways, whereas by day 10, metabolic pathways were uncovered as the top pathway. Indeed, Sox9 is a critical mediator of metabolic processes in chondrogenic cells (38) and regulates components of the EGFR pathway in malignant cells (39). The maintenance of the pluripotent stem cell state, along with chromatin organization, appeared in day 10 SOX9+ cells as shown by both KEGG pathway and GO analysis, thus directly implicating sustained Sox9 activity in endogenous reprogramming to a progenitor-like state in at least a subset of Sox9on-on cells. Two well-known SOX9 target genes (40, 41) showed distinct time-specific binding. At 48 hours but not at day 10, Sox9 bound Col1a1; by contrast, Col2a1 was bound at day 10 but not at 48 hours (fig. S21K). Fgf/Fgfr/Sox9 forms a feed-forward loop to maintain cell identity and growth (42). Conversely, our data showed that Fgfr1 was targeted at both time points, and a higher engagement was observed by day 10. In day 10–specific target gene analysis, the Wnt signaling pathway, consisting of genes such as Wnt2, Wnt5a, Wnt5b, Tcf7, and Tcf7l2, and others, featured among the top pathways. Tcf7 and Tcf7l2 are known to cooperate with b-catenin in committed nephron progenitor cells (43). To confirm whether Wnt2 represents a direct transcriptional target of SOX9, we examined the effect of removal of Sox9on-on activity on Wnt2 mRNA expression. Wnt2 mRNA was markedly reduced in Sox9on-on cKO versus intact Sox9on-on activity (fig. S21L). These findings validate the CUT&RUN SOX9 genomic occupancy assay and uncover a direct Sox9/Wnt2 link, which might contribute to maintaining fibroproliferative response after AKI. Thus,we have provided further evidence for the direct role of Sox9on-on activity in the

p

and hematopoietic stem and progenitor cell specification (33). Thus, these findings not only validated snATAC-seq pseudotemporal analysis, but also identified them as potential candidates upstream of Sox9on-on activity. snATAC-seq analysis of the inner cortices and outer medulla region (Fig. 4A) revealed a sufficient population of Sox9on-on nuclei. Therefore, SOX9 CUT&RUN genomic occupancy assay in the day 10 lineage versus its parent Sox9on cells (48 hours after injury) provided an opportunity to examine the direct effects of sustained Sox9 activity within its lineage in vivo. We conducted a time-resolved, lineagespecific SOX9 CUT&RUN genomic occupancy assay (fig. S21A). An H3K4me3 genomic occupancy assay was used as a control. Because of the technical challenges involved in isolating high-quality, >90% viable fragile tdTposenriched cells under low-flow FACS conditions and the limited labeling possible by a single tamoxifen injection, a pellet of ~30,000 to 100,000 cells was subjected to a genomic occupancy/antibody/time-point assay. The cell pellet was expected to contain Sox9on-off cells, therefore, to reduce these cells, the lineage cells were enriched from the IC/OM region. Despite these limitations, (i) validation of appropriate insert size for a transcription factor (fig. S21B), (ii) distinct time- and antibodyspecific clustering of biological samples using Pearson correlation approaches and UMAP (uniform manifold approximation and projection) approaches (fig. S21, C and D), (iii) signal of SOX9 peak relative to the transcription start site (fig. S21E), and (iv) significant optimal Sox9 motif enrichment with Sox9 antibody versus input (P < 1E-300 at day 10, P < 0.02 at 48 hours; fig. S21F) or versus H3K4me3 antibody (P < 1E-300 at day 10) suggested that the data could be reliably mined for meaningful biological conclusions with respect to SOX9specific target binding. Consistent with the previous studies of genome-wide chromatin binding by SOX9, 10 to 15% of SOX9-bound regions occurred within transcriptional start or promoter sites (34, 35). Indeed, the enriched GO terms across both time points showed that the significantly bound region and the corresponding significantly called peaks and genes compared with their respective inputs [false discovery rate (FDR)–adjusted P value < 0.01; fig. S21G), were linked with the biological processes involved in epithelial regeneration in Sox9 48 hour and day 10 cells (fig. S21, H and I), consistent with the genetic-lineage tracing studies and scRNA-seq analysis of such cells

n = 3 animals. (I and J) Coimmunostaining for SMARCC1 and CDH6 showing SMARCC1 expression confined to day 10 CDH6pos Sox9 lineage cells (I), and RUNX1 and SOX9 coimmunostaining showing RUNX1 restricted to Sox9on-on versus Sox9on-off lineage (J) as predicted by (G), thus validating Sox9 lineage–specific snATAC-seq datasets. n = 2 animals. Scale bars, 100 mm.

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

p g y y g y ,

Fig. 5. Human renal allografts display dynamic SOX9/CDH6/WNT2B activity, with CDH6pos cells demarcating SOX9on-on activity and fibrotic foci. (A) Coimmunostaining for SOX9 and AQP1 in biopsies obtained before implantation (uninjured) and IRI-induced AKI after transplantation (same allograft) showing early SOX9 activation. (B) Box plot showing SOX9 levels at Aggarwal et al., Science 383, eadd6371 (2024)

23 February 2024

different time points within kidney transplant protocol biopsies (n = 163). (C and D) Dot plots (C) and box plots (D) showing categorization of patients and kidney function according to SOX9 levels at 1 year (comparison by Mann-Whitney U test; n = 35). (E) Histograms showing the number of patients with different degrees of kidney fibrosis, as estimated by interstitial fibrosis of cortical areas 10 of 13

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

(ci score) according to Banff classification (comparison by chi-square test; n = 39). (F) Box plot showing SOX9 and CDH6 levels categorized according to a previously reported model discriminating the transcriptome of kidney transplant biopsies in successful repair (1), transition to chronic injury (2), or CKD (3). LD, biopsies obtained from living donors at the time of transplantation (49). (G) Heatmap showing expression correlation of genes of interest at 3 and 12 months after transplantation (Spearman r correlation coefficient; n = 72). Arrows highlight the identified molecular signatures of SOX9/CDH6/WNTs, including AXIN2 and NKD1 in human kidneys. Note that CDH6 is the topmost correlated gene. (H) UMAP showing distinct PTEC clustering (29,180 genes × 24,070 cells). (I) Dot plot showing average gene expression values and percentage of cells expressing markers of differentiation and injury, SOX9 and CDH6, by each identified cluster of PTECs. Cluster 8 consisted of SOX9pos-, CDH6pos-, and WNT2Bposexpressing cells (see also figs. S22, C to E, and S23). (J) Feature plot displaying the normalized transcript expression for the respective genes. Magic mRNAassay of renal cells is depicted. (K) Analysis comparing the percentage of cluster 8

,

11 of 13

y

23 February 2024

y g

In this study, by establishing a model system that facilitated head-to-head comparison between the two initially committed lineages to regenerate the injured proximal tubular nephron epithelia but with divergent outcomes (one that expeditiously regenerated tissue versus the other that was unable to do so), we identified the unifying mechanism underpinning scarless versus fibrotic tissue repair at the single-cell level within the same microenvironment. We also identified how precisely Wntenriched niches are formed and maintained after injury. Until now, despite the prominent link between Wnt and tissue fibrosis after injury, this question has remained unanswered (51, 52). For example, in the setting of lung injury, the Wnt-inducing factor was highlighted as the “unknown factor” (52). Our study highlights the potential for abrogating fibrotic responses through a cell-state–specific intervention, even when introduced 1 week after injury, during the progression of AKI to CKD. This was illustrated by precise cell-state–specific genetic perturbation of Sox9on-on activity or removal of Wntless within the Sox9on-on cells.

y

Discussion

Our data suggest that the Soxon-on CDH6pos cell state signifies an ongoing regenerating phase during the transition from AKI to CKD. This was supported by (i) enrichment of biological and cellular processes linked with PTEC formation at the single-cell level; (ii) integrated analysis of scRNA-seq and Sox9 lineage–specific snATAC-seq datasets, unveiling a cascade of transcriptional regulators linked to nephrogenesis in the Sox9on-on nuclei; and (iii) distinct, time-specific SOX9 genomic occupancy within its lineage, indicating direct involvement in activating programs for nephron epithelia formation. Indeed, it would make biological sense for Sox9 in its sustained attempt to regenerate the unrestored epithelia to recruit Wnts, specifically Wnt4. Unexpectedly, Wnt2, which is essential for murine lung development but understudied in nephrogenesis, emerged as direct target gene of sustained SOX9 activity. Wnt2 was substantially decreased upon Sox9 activity removal, highlighting the direct role of Sox9on-on in provoking fibrosis during a sustained effort to regenerate the epithelia. WNT2B demonstrated stronger correlation with sustained SOX9 activity in transplanted human kidneys then did than WNT2. These findings hint at differential deployment of paralogous Wnt2 genes in the progression from AKI to CKD between humans and mice. Therefore, we have uncovered a link that explains how the tissue regeneration process culminates in fibrosis. The SOX9posCDH6pos cell state contrasts with other known maladaptive cell types implicated in postinjury organ fibrosis, including senescent, partial epithelial to mesenchymal, or cellcycle-arrested cells (2, 53, 54). However, it remains a possibility that with time these cell states might be subsequently attained due to the secondary effects of severely fibrotic adverse milieu. Sox9-expressing myofibroblasts have been linked with renal fibrosis (55). Our study, which used varied orthogonal approaches, did not reveal such cells in post-AKI fibrotic kidneys, although this does not completely rule out the possibility that this population exists.

g

Aggarwal et al., Science 383, eadd6371 (2024)

and human kidneys uncovered a robust link between dynamic Sox9on-on activity and Cdh6 response. Because cell-cell contact and/or adhesion disruption leads to Sox9 activation, with silencing of Sox9 upon cell-cell contact restoration, it would make biological sense that in the setting of prolonged cell-cell contact disruption, the persistent Sox9 transcriptional activity would induce a cadherin in its attempt to restore cell-cell adhesion. Therefore, we verified that Cdh6 not only demarcates Sox9on-on cells but might also be regulated by Sox9. We detected significant reduction of CDH6pos cells and Cdh6 mRNA (P < 0.01) in Sox9-cKO kidneys compared with the WT (Fig. 5, P and Q). Thus, SOX9posCDH6pos cells demarcate cells with SOX9on-on activity– driven WNT signaling niches and fibrosis after human AKI.

p

Although strong correlation was noted with WNT4, the correlation was more pronounced with other members of the WNT family in humans, particularly with WNT2B. Patients with a low immunological risk profile were uniformly managed in a single center using a protocol comprising tacrolimus-, mycophenolate mofetil–, and prednisone-based immunosuppression. Rejection episodes were rare. To obtain single-cell-level resolution of the above observed responses, we leveraged scRNAseq datasets involving 16 kidney transplant biopsies (50) (Fig. 5H and fig. S22, A and B). scRNA-seq showed that SOX9/CDH6/WNT2B– expressing cells predominantly resided in the same cluster (cluster 8), which also contained HAVCR1+ cells (Fig. 5, I and J, and fig. S22, C to E). This cluster demonstrated significant correlation with fibrosis (P < 0.05) (Fig. 5K), and time-resolved PTECs showed de novo emergence of cluster 8 after transplantation (Fig. 5L). Unlike Wnt2 (fig. S21L), Wnt2b did not exhibit a significant reduction upon removal of Sox9on-on activity (fig. S22F). Cluster 8 displayed dynamic activity and greater latent time compared with other PTEC clusters (fig. S23, A to E). Further, in another unbiased analysis to identify the driver gene that confers clusterspecific dynamic behavior, CDH6 was found to be the gene with the second-highest likelihood of underlying dynamic activity in HAVCR1+ cluster 8 (Fig. 5M). Akin to our findings in mice, CDH6pos LTLlow PTECs (Fig. 5N) and persistent SOX9 activity (fig. S24) displayed intimate association with aSMA+ myofibroblasts. Thus, these findings validated the identified dynamic axis in human kidneys at the single-cell level and revealed WNT2B as the likely WNT that may drive human fibrosis. Akin to mouse PTECs, subconfluent primary human PTECs mounted SOX9 activity, which subsided upon tight monolayer formation, further emphasizing the tight link of SOX9 activity with cell-cell contact status (Fig. 5O). Our transcriptomic profiling and spatiotemporal mapping studies involving both murine

and the interstitial fibrosis and tubular atrophy (IFTA) grade for each patient. Percentage of cluster 8 is calculated with respect to the PTEC number in each patient. P < 0.05 based on the Kruskal-Wallis test and Pearson correlation analysis. (L) UMAP showing time-resolved PTECs clustering after transplantation (shown as days after transplantation). Note the emergence of cluster 8 with time after transplantation. (M) Cluster-type–specific genes analysis revealed CDH6 as being among the top two driver genes that underlie the dynamic activity in cluster 8. (N) Coimmunostaining showing that CDH6pos cells displayed a tight intimate association with ACTA2pos myofibroblasts within human kidney allograft, with CDH6neg foci showing no ACTA2pos myofibroblasts. (O) Immunoblot showing that subconfluent human primary PTECs activated SOX9, which waned upon confluency. (P and Q) Coimmunoanalysis (P) and qPCR (Q) showing reduction in CDH6pos cells and Cdh6 mRNA, respectively, upon removal of Sox9on-on activity (blinded analysis, unpaired two-sided Student’s t test; data are shown as mean ± SEM). Scale bars in whole scanned image (P), 1000 mm; all others, 100 mm.

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

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

28.

29.

30.

31.

32.

12 of 13

,

27.

y

23 February 2024

14.

y g

Aggarwal et al., Science 383, eadd6371 (2024)

1. D. C. Rockey, P. D. Bell, J. A. Hill, Fibrosis—A common pathway to organ injury and failure. N. Engl. J. Med. 372, 1138–1149 (2015). doi: 10.1056/NEJMra1300575; pmid: 25785971 2. N. C. Henderson, F. Rieder, T. A. Wynn, Fibrosis: From mechanisms to medicines. Nature 587, 555–566 (2020). doi: 10.1038/s41586-020-2938-9; pmid: 33239795 3. P. Konieczny, S. Naik, Healing without scarring. Science 372, 346–347 (2021). doi: 10.1126/science.abi5770; pmid: 33888629 4. S. Kumar, Cellular and molecular pathways of renal repair after acute kidney injury. Kidney Int. 93, 27–40 (2018). doi: 10.1016/j.kint.2017.07.030; pmid: 29291820 5. J. Liu et al., Molecular characterization of the transition from acute to chronic kidney injury following ischemia/reperfusion. JCI Insight 2, e94716 (2017). doi: 10.1172/jci.insight.94716; pmid: 28931758 6. J. G. Abuelo, Normotensive ischemic acute renal failure. N. Engl. J. Med. 357, 797–805 (2007). doi: 10.1056/ NEJMra064398; pmid: 17715412 7. J. V. Bonventre, L. Yang, Cellular pathophysiology of ischemic acute kidney injury. J. Clin. Invest. 121, 4210–4221 (2011). doi: 10.1172/JCI45161; pmid: 22045571 8. I. Grgic et al., Targeted proximal tubule injury triggers interstitial fibrosis and glomerulosclerosis. Kidney Int. 82, 172–183 (2012). doi: 10.1038/ki.2012.20; pmid: 22437410 9. M. A. Venkatachalam et al., Acute kidney injury: A springboard for progression in chronic kidney disease. Am. J. Physiol. Renal Physiol. 298, F1078–F1094 (2010). doi: 10.1152/ ajprenal.00017.2010; pmid: 20200097 10. S. Kumar et al., Sox9 activation highlights a cellular pathway of renal repair in the acutely injured mammalian kidney. Cell Rep. 12, 1325–1338 (2015). doi: 10.1016/j.celrep.2015.07.034; pmid: 26279573 11. K. Zhang et al., In vivo two-photon microscopy reveals the contribution of Sox9+ cell to kidney regeneration in a mouse

13.

y

All animal procedures were approved by the Cedars-Sinai Medical Center Institutional Animal Care and Use Committee, and institutional review board–approved human kidney biopsies were used. All animals used in the study are described in table S4. To induce the IRIinduced transition from AKI to CKD, 9- to 12-week-old weight-matched (25 to 30 g) mice were subjected to long-term survival compatible bilateral renal IRI surgery. To induce the transition from rhabdo-AKI to CKD, adult anesthetized mice (9 to 12 weeks old) were administered an intramuscular injection of 50% hypertonic glycerol solution in each quadriceps muscle (total dose, 8 mg/kg). For induction of CreERT2 protein, mice were injected with tamoxifen dissolved in corn oil through an intraperitoneal injection. For induction of doxycycline-inducible Cre protein, mice were

RE FERENCES AND NOTES

12.

model with extracellular vesicle treatment. J. Biol. Chem. 295, 12203–12213 (2020). doi: 10.1074/jbc.RA120.012732; pmid: 32641493 B. L. Hogan, P. A. Kolodziej, Organogenesis: Molecular mechanisms of tubulogenesis. Nat. Rev. Genet. 3, 513–523 (2002). doi: 10.1038/nrg840; pmid: 12094229 W. J. Nelson, R. Nusse, Convergence of Wnt, beta-catenin, and cadherin pathways. Science 303, 1483–1487 (2004). doi: 10.1126/science.1094291; pmid: 15001769 V. Vasioukhin, C. Bauer, M. Yin, E. Fuchs, Directed actin polymerization is the driving force for epithelial cell-cell adhesion. Cell 100, 209–219 (2000). doi: 10.1016/S0092-8674 (00)81559-7; pmid: 10660044 S. P. Mah, H. Saueressig, M. Goulding, C. Kintner, G. R. Dressler , Kidney development in cadherin-6 mutants: Delayed mesenchyme-to-epithelial conversion and loss of nephrons. Dev. Biol. 223, 38–53 (2000). doi: 10.1006/dbio.2000.9738; pmid: 10864459 S. Thelen, M. Abouhamed, G. Ciarimboli, B. Edemir, M. Bähler, Rho GAP myosin IXa is a regulator of kidney tubule function. Am. J. Physiol. Renal Physiol. 309, F501–F513 (2015). doi: 10.1152/ajprenal.00220.2014; pmid: 26136556 Y. Kirita, H. Wu, K. Uchimura, P. C. Wilson, B. D. Humphreys, Cell profiling of mouse acute kidney injury reveals conserved cellular responses to injury. Proc. Natl. Acad. Sci. U.S.A. 117, 15874–15883 (2020). doi: 10.1073/pnas.2005477117; pmid: 32571916 D. Legouis et al., Altered proximal tubular cell glucose metabolism during acute kidney injury is associated with mortality. Nat. Metab. 2, 732–743 (2020). doi: 10.1038/ s42255-020-0238-1; pmid: 32694833 Y. Neirijnck et al., Sox11 gene disruption causes congenital anomalies of the kidney and urinary tract (CAKUT). Kidney Int. 93, 1142–1153 (2018). doi: 10.1016/j.kint.2017.11.026; pmid: 29459093 H. Clevers, K. M. Loh, R. Nusse, An integral program for tissue renewal and regeneration: Wnt signaling and stem cell control. Science 346, 1248012 (2014). doi: 10.1126/science.1248012; pmid: 25278615 H. F. Farin et al., Visualization of a short-range Wnt gradient in the intestinal stem-cell niche. Nature 530, 340–343 (2016). doi: 10.1038/nature16937; pmid: 26863187 B. Degirmenci, T. Valenta, S. Dimitrieva, G. Hausmann, K. Basler, GLI1-expressing mesenchymal cells form the essential Wnt-secreting niche for colon stem cells. Nature 558, 449–453 (2018). doi: 10.1038/s41586-018-0190-3; pmid: 29875413 S. Kim et al., The polycystin complex mediates Wnt/Ca(2+) signalling. Nat. Cell Biol. 18, 752–764 (2016). doi: 10.1038/ ncb3363; pmid: 27214281 Y. Rinkevich et al., In vivo clonal analysis reveals lineagerestricted progenitor characteristics in mammalian kidney development, maintenance, and regeneration. Cell Rep. 7, 1270–1283 (2014). doi: 10.1016/j.celrep.2014.04.018; pmid: 24835991 C. Bänziger et al., Wntless, a conserved membrane protein dedicated to the secretion of Wnt proteins from signaling cells. Cell 125, 509–522 (2006). doi: 10.1016/j.cell.2006.02.049; pmid: 16678095 D. P. DiRocco, A. Kobayashi, M. M. Taketo, A. P. McMahon, B. D. Humphreys, Wnt4/b-catenin signaling in medullary kidney myofibroblasts. J. Am. Soc. Nephrol. 24, 1399–1412 (2013). doi: 10.1681/ASN.2012050512; pmid: 23766539 C. Kuppe et al., Decoding myofibroblast origins in human kidney fibrosis. Nature 589, 281–286 (2021). doi: 10.1038/ s41586-020-2941-1; pmid: 33176333 W. Guo et al., Slug and Sox9 cooperatively determine the mammary stem cell state. Cell 148, 1015–1028 (2012). doi: 10.1016/j.cell.2012.02.008; pmid: 22385965 C. López-Rodríguez et al., Loss of NFAT5 results in renal atrophy and lack of tonicity-responsive gene expression. Proc. Natl. Acad. Sci. U.S.A. 101, 2392–2397 (2004). doi: 10.1073/pnas.0308703100; pmid: 14983020 M. Haller, J. Au, M. O’Neill, D. J. Lamb, 16p11.2 transcription factor MAZ is a dosage-sensitive regulator of genitourinary development. Proc. Natl. Acad. Sci. U.S.A. 115, E1849–E1858 (2018). doi: 10.1073/pnas.1716092115; pmid: 29432158 M. Bouchard, A. Souabni, M. Mandler, A. Neubüser, M. Busslinger, Nephric lineage specification by Pax2 and Pax8. Genes Dev. 16, 2958–2970 (2002). doi: 10.1101/gad.240102; pmid: 12435636 L. Ho et al., An embryonic stem cell chromatin remodeling complex, esBAF, is essential for embryonic stem cell

g

Methods summary

injected with doxycycline dissolved in 0.9% normal saline intraperitoneally. Enzymatic digestion, which was conducted on ice to isolate cells from the kidney for RNA or nuclei extraction studies, used B. Licheniformis Cold Active Protease, DNase1, and Liberase TL. Cell-type–specific bulk RNAseq libraries were constructed using the Universal plus mRNAseq with NuQuant kit from NuGEN. Sample libraries were sequenced on the NovaSeq platform (Illumina) using 150–base pair pairedend sequencing. Single-cell RNA libraries were obtained using the Chromium platform. Sox9 lineage–specific single nuclei for snATAC-Seq studies were isolated according to the 10xGenomics protocol using the low-input version and with a 1:5 diluted lysis buffer in nucleasefree water. Isolated cell samples were immediately processed with a Chromium Next GEM Single Cell ATAC Kit v2 (10xGenomics). Cell pellets of ~100,000 FACS-enriched Sox9 lineage cells were subjected to Sox9 genomic occupancy assay, and the libraries generated from the immune-enriched DNA samples using the Illumina kit were analyzed using Partek Flow software v10. RNAscope based in situ hydridization assay was performed on 12-mmcut, optimal cutting temperature (OCT)– embedded cryosections according to the manufacturer’s protocol. The qPCR primers and primary and secondary antibodies used in this study are detailed in tables S5 to S7. Most of the imaging was performed on the Zeiss 780 confocal system. Unpaired, two-sided Student’s t test was used to compare two independent groups.

p

Across phyla, damage-induced repair response represents a fundamental tissue survival mechanism. How does an injured tubular epithelial cell temporally sense its reparative state in vivo in damaged-tissue microenvironment that lacks resident stem or progenitor cell population? Is there a unifying, central “onoff” molecular switch? For example, in yeast, a single transcriptional control mechanism enables cells to respond to fluctuating nutrient concentrations (56). Herein, we identified SOX9 as a dynamic, fundamental, intrinsic transcriptional link between loss of epithelial integrity and regenerative response in vivo. Recently, Yamanaka and colleagues linked apicobasolateral polarity, maintained by the tight junction protein ZO-1, with receptivity to signaling proteins and multicellular patterning in an in vitro human gastrulation platform (57). Genetic removal of ZO-1 led to sustained BMP4 signaling pathway (BMP4/pSMAD1/5) activation and the ensuing distinct cell-state specification. We found that the Sox9on-on cell state is Bmp4/Smads enriched (Fig. 1E and figs. S6D and S20B) and is a subset attaining a nephron progenitor-like cell state with time. It is tempting to speculate that a similar enhanced receptivity to the as-yet elusive signaling proteins in such disrupted epithelial cells drives Sox9 activity and results in a distinct cell state. However, how precisely such disruption might regulate Sox9 remains to be established, which is a limitation of the current study. Further, a pharmacological approach to perturb the identified pathway remains unidentified, but our findings lay the ground for drug discovery and precise cell-state–specific genetic perturbation strategies to retard fibrosis. In the present study, we have shown that the duration of the regeneration response is a key determinant of healing with or without fibrosis, with the SOX9posCDH6pos cell state interconnecting the transition from AKI to CKD.

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

33.

34.

35.

36.

37.

38.

40.

42.

44.

45.

46.

We thank S. Chiamthamachinda for technical assistance with quantification, immunostaining, and genotyping; and I. Zafar for technical assistance with immunostaining and genotyping; C. Santiskulvong at the Applied Genomics, Computation, and Translational Core and flow core facility of Cedars-Sinai Medical Center; the Technology Center for Genomics & Bioinformatics (TCGB) at David Geffen School of Medicine, University of California, Los Angeles (UCLA) for cell-type–specific RNAsequencing; and the UAB-UCSD O’Brien Center, Birmingham, AL, for serum creatinine measurements. Funding: This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health (grant R01 DK118265 to S.K.); the American Heart Association (grant 18CDA34110416 to S.K.); the American Society of Nephrology (John Merrill Transplant Scholar Grant to S.K.); UCLA CTSI (S.K.); the One Legacy Foundation (S.K.); and the Department of Defense (grant CDMRP KC200178 to S.K and J.B.). Work in the P.E.C. laboratory is supported by the Swiss National Foundation (Sinergia grant CRSII5_202302) and by the Balli and Gianella foundations. The funders had no role in study design, data collection and

SUPPLEMENTARY MATERIALS

science.org/doi/10.1126/science.add6371 Materials and Methods Figs. S1 to S24 Tables S1 to S7 References (58–69) MDAR Reproducibility Checklist

y g

47.

ACKN OWLED GMEN TS

y

43.

interpretation, or the decision to submit the work for publication. Author contributions: S.A. conducted bone marrow–derived macrophage experiments, and isolated single cells from the scRNA-seq and RNAscope experiments. Z.W. modified Sox9 shRNA and conducted the shRNA experiments. S.A. and Z.W. conducted experiments, including, breeding, genotyping, AKI models, cellular studies, coimmunostaining, and confocal microscopy, and analyzed data. D.R.F.P. conducted and analyzed lineage-specific snATAC-seq, time-resolved lineage-specific CUT&RUN, and with S.K designed these experiments. A.Ri. analyzed published singlenuclear and human single-cell datasets. A.Ra. analyzed raw singlecell and cell-type–specific bulk RNA sequencing data. A.E.C. performed coimmunostaining and genotyping. M.Y. analyzed histopathology. C.N.S. contributed to single-cell experiments, contributed crucial reagents, and analyzed data. J.C., E.V.L., and B.L. contributed human kidney single-cell RNAseq datasets. J.H., H.A., S.A.K., P.W.N., and S.C.J. contributed crucial reagents and analyzed data. J.B. performed dynamic modeling of human single-cell RNA-seq datasets. M.N. and P.E.C. contributed and analyzed human kidney transcriptomic studies, single-cell RNAseq database of kidney transplants, and published single-nuclear datasets. S.K. conceptualized the study and designed all experiments, conducted confocal microscopy, analyzed results, and wrote the manuscript with input from all authors. Competing interests: S.K., S.C.J., and S.A. are inventors on provisional patent application (US63/605, 777) submitted by Cedars-Sinai Medical Center that covers “Cadherin6 expression status in determination of renal fibrosis and related uses thereof.” The remaining authors declare no competing interests. Data and materials availability: The datasets generated during this study are available in the Gene Expression Omnibus (GEO) database under accession no. GSE249781. These include single-cell sequencing (scRNA-seq) datasets (GSE196929); Sox9 cell-typespecific bulk RNAseq datasets (GSE249778); Axin2CreERT2/+: R26RtdT/+:Acta2-GFP cell-type–specific bulk RNAseq datasets (GSE249777); Sox9 lineage–specific, time-resolved genomic occupancy assay datasets (GSE249776); and Sox9 lineage–specific snATACseq data (GSE249780). Accession codes of the published data in GEO used in this study are as follows: RNAseq data for human kidney transplant biopsies: GSE126805 and single-nuclei RNAseq: GSE151167, GSE139107, and GSE163863). The accession codes for the published human scRNAseq data have been deposited in the European repository Biostudies under accession code E-MTAB-12051. License information: Copyright © 2024 the authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original US government works. https://www.science.org/about/sciencelicenses-journal-article-reuse

g

41.

48. P. E. Cippà et al., A late B lymphocyte action in dysfunctional tissue repair following kidney injury and transplantation. Nat. Commun. 10, 1157 (2019). doi: 10.1038/s41467-01909092-2; pmid: 30858375 49. P. E. Cippà et al., Transcriptional trajectories of human kidney injury progression. JCI Insight 3, e123151 (2018). doi: 10.1172/ jci.insight.123151; pmid: 30429361 50. B. Lamarthée et al., Transcriptional and spatial profiling of the kidney allograft unravels a central role for FcyRIII+ innate immune cells in rejection. Nat. Commun. 14, 4359 (2023). doi: 10.1038/s41467-023-39859-7; pmid: 37468466 51. B. D. Humphreys, Mechanisms of renal fibrosis. Annu. Rev. Physiol. 80, 309–326 (2018). doi: 10.1146/annurev-physiol-022516034227; pmid: 29068765 52. O. Burgy, M. Königshoff, The WNT signaling pathways in wound healing and fibrosis. Matrix Biol. 68-69, 67–80 (2018). doi: 10.1016/j.matbio.2018.03.017; pmid: 29572156 53. L. Yang, T. Y. Besschetnova, C. R. Brooks, J. V. Shah, J. V. Bonventre, Epithelial cell cycle arrest in G2/M mediates kidney fibrosis after injury. Nat. Med. 16, 535–543, 1p, 143 (2010). doi: 10.1038/nm.2144; pmid: 20436483 54. J. M. Wolstein et al., INK4a knockout mice exhibit increased fibrosis under normal conditions and in response to unilateral ureteral obstruction. Am. J. Physiol. Renal Physiol. 299, F1486–F1495 (2010). doi: 10.1152/ajprenal.00378.2010; pmid: 20861074 55. S. Raza et al., SOX9 is required for kidney fibrosis and activates NAV3 to drive renal myofibroblast function. Sci. Signal. 14, eabb4282 (2021). doi: 10.1126/scisignal. abb4282; pmid: 33653921 56. C. Ricci-Tam et al., Decoupling transcription factor expression and activity enables dimmer switch gene regulation. Science 372, 292–295 (2021). doi: 10.1126/science.aba7582; pmid: 33859035 57. I. Vasic et al., Loss of TJP1 disrupts gastrulation patterning and increases differentiation toward the germ cell lineage in human pluripotent stem cells. Dev. Cell 58, 1477–1488.e5 (2023). doi: 10.1016/j.devcel.2023.05.019; pmid: 37354899

p

39.

self-renewal and pluripotency. Proc. Natl. Acad. Sci. U.S.A. 106, 5181–5186 (2009). doi: 10.1073/pnas.0812889106; pmid: 19279220 P. Li et al., Epoxyeicosatrienoic acids enhance embryonic haematopoiesis and adult marrow engraftment. Nature 523, 468–471 (2015). doi: 10.1038/nature14569; pmid: 26201599 S. Ohba, X. He, H. Hojo, A. P. McMahon, Distinct Transcriptional Programs Underlie Sox9 Regulation of the Mammalian Chondrocyte. Cell Rep. 12, 229–243 (2015). doi: 10.1016/j.celrep.2015.06.013; pmid: 26146088 B. M. Fuglerud et al., SOX9 reprograms endothelial cells by altering the chromatin landscape. Nucleic Acids Res. 50, 8547–8565 (2022). doi: 10.1093/nar/gkac652; pmid: 35904801 V. R. Harley, R. Lovell-Badge, P. N. Goodfellow, P. J. Hextall, The HMG box of SRY is a calmodulin binding domain. FEBS Lett. 391, 24–28 (1996). doi: 10.1016/ 0014-5793(96)00694-1; pmid: 8706923 A. Argentaro et al., A SOX9 defect of calmodulin-dependent nuclear import in campomelic dysplasia/autosomal sex reversal. J. Biol. Chem. 278, 33839–33847 (2003). doi: 10.1074/jbc.M302078200; pmid: 12810722 N. van Gastel et al., Lipid availability determines fate of skeletal progenitor cells via SOX9. Nature 579, 111–117 (2020). doi: 10.1038/s41586-020-2050-1; pmid: 32103177 F. Liu et al., EGFR Mutation Promotes Glioblastoma through Epigenome and Transcription Factor Network Remodeling. Mol. Cell 60, 307–318 (2015). doi: 10.1016/ j.molcel.2015.09.002; pmid: 26455392 D. M. Bell et al., SOX9 directly regulates the type-II collagen gene. Nat. Genet. 16, 174–178 (1997). doi: 10.1038/ ng0697-174; pmid: 9171829 C. D. Oh et al., Identification of SOX9 interaction sites in the genome of chondrocytes. PLOS ONE 5, e10113 (2010). doi: 10.1371/journal.pone.0010113; pmid: 20404928 P. A. Seymour et al., A Sox9/Fgf feed-forward loop maintains pancreatic organ identity. Development 139, 3363–3372 (2012). doi: 10.1242/dev.078733; pmid: 22874919 Q. Guo et al., A b-catenin-driven switch in TCF/LEF transcription factor binding to DNA target sites promotes commitment of mammalian nephron progenitor cells. eLife 10, e64444 (2021). doi: 10.7554/eLife.64444 pmid: 33587034 A. S. Weintraub et al., YY1 is a structural regulator of enhancerpromoter loops. Cell 171, 1573–1588.e28 (2017). doi: 10.1016/ j.cell.2017.11.008; pmid: 29224777 H. Nguyen et al., Tcf3 and Tcf4 are essential for long-term homeostasis of skin epithelia. Nat. Genet. 41, 1068–1075 (2009). doi: 10.1038/ng.431; pmid: 19718027 J. C. Robins et al., Hypoxia induces chondrocyte-specific gene expression in mesenchymal cells in association with transcriptional activation of Sox9. Bone 37, 313–322 (2005). doi: 10.1016/j.bone.2005.04.040; pmid: 16023419 E. Huang et al., Three-year outcomes of a randomized, double-blind, placebo-controlled study assessing safety and efficacy of C1 esterase inhibitor for prevention of delayed graft function in deceased donor kidney transplant recipients. Clin. J. Am. Soc. Nephrol. 15, 109–116 (2020). doi: 10.2215/ CJN.04840419; pmid: 31843975

Submitted 24 June 2022; resubmitted 8 July 2023 Accepted 11 January 2024 10.1126/science.add6371

y ,

Aggarwal et al., Science 383, eadd6371 (2024)

23 February 2024

13 of 13

RES EARCH

RESEARCH ARTICLE SUMMARY

◥

IMMUNOGENETICS

An immunogenetic basis for lung cancer risk Chirag Krishna†, Anniina Tervi†, Miriam Saffern†, Eric A. Wilson†, Seong-Keun Yoo, Nina Mars, Vladimir Roudko, Byuri Angela Cho, Samuel Edward Jones, Natalie Vaninov, Myvizhi Esai Selvan, Zeynep H Gümüs ¸, FinnGen, Tobias L. Lenz, Miriam Merad, Paolo Boffetta, Francisco Martínez-Jiménez, Hanna M. Ollila*‡, Robert M. Samstein*‡, Diego Chowell*‡

INTRODUCTION: Whether the host immune sys-

y

▪

The list of author affiliations is available in the full article online. *Corresponding author. Email: [email protected] (D.C.); [email protected] (R.M.S.); hanna.m.ollila@ helsinki.fi (H.M.O) †These authors contributed equally to this work. ‡These authors contributed equally to this work. Cite this article as C. Krishna et al., Science 383, eadi3808 (2024). DOI: 10.1126/science.adi3808

y g

RESULTS: In both the UK Biobank and FinnGen, we found that heterozygosity at the HLA class II (HLA-II) loci was associated with reduced risk of lung cancer over more than a decade of followup. HLA-II heterozygosity was associated with reduced risk of lung cancer in both current and former but not never-smokers, suggesting that smoking-derived antigens may augment the immune response to early neoplastic disease. HLA-II homozygosity conferred substantial lifetime risk of disease (e.g., in the UK Biobank, 13.9% for current smokers homozygous at HLA-DRB1) and was independent of

CONCLUSION: The association of HLA-II heterozygosity with reduced risk of lung cancer implies that genetic variation in immunosurveillance is a feature of cancer susceptibility, together with environmental exposures, hereditary risk, and DNA replication errors. Our findings broaden understanding of the role of the host immune system in cancer risk and may motivate the incorporation of immunogenetics into lung cancer screening programs.

g

RATIONALE: Understanding the molecular determinants of cancer risk is critical for early cancer detection and strategies to limit cancer mortality. Tobacco smoking increases lung cancer risk and is associated with a heightened somatic mutation rate that drives neoplastic potential, but whether there are additional risk

factors that further modify lung cancer susceptibility, even among smokers, is unclear. The HLA heterozygote advantage theory posits that an HLA genotype encoding two different allomorphs enables presentation of a more diverse antigenic peptide repertoire to the immune system—and subsequent superior immune control of infected or cancerous cells—than does an HLA genotype encoding two equivalent allomorphs. Thus, heterozygous HLA allomorphs may present more neoantigens arising from smoking-derived somatic mutations. In this study, we evaluated the effect of HLA heterozygosity on lung cancer risk, leveraging genetic and longitudinal clinical data from the UK Biobank and FinnGen together with multimodal genomic analyses of nonmalignant and lung tumor samples.

p

tem naturally protects against cancer has long been the subject of intense debate. The cancer immunosurveillance theory ascribes a protective function to the adaptive immune system, whereby T cell–mediated recognition of neoantigens presented by the major histocompatibility complex suppresses early neoplasia. Studies in mice have provided support for the cancer immunosurveillance theory, yet evidence for a protective role of the immune system against cancer in humans has been relatively lacking. In lung cancer, genetic variation in the human leukocyte antigen (HLA) locus is linked to tumor evolution and treatment outcomes, but whether HLA polymorphisms reduce lung cancer risk—which would imply a role of the host immune system in preventing lung cancer—is currently unclear. Populationscale biobank analysis coupling host genetics with longitudinal clinical data enables a systematic investigation of how HLA polymorphism influences lung cancer risk together with smoking and other established risk factors.

known clinical and genetic risk factors, including a genome-wide polygenic risk score. Heterozygosity of amino acid sites within the HLA-II peptide binding groove was also associated with reduced risk of lung cancer, whereas analysis of single-cell RNA-sequencing data from nonmalignant and tumor lung samples showed that lung macrophages and epithelial cells express HLA-II and are affected by smoking. Analysis of tumor genomes from the The Cancer Genome Atlas (TCGA) cohort, the Pan-Cancer Analysis of Whole Genomes (PCAWG) cohort, and Hartwig Medical Foundation cohort revealed widespread loss of heterozygosity (LOH) of the HLA-II loci in lung cancer, with rates of LOH equaling those of HLA-I. An analysis of neoantigen repertoires between lung cancer tumors with and without HLA-II LOH showed that HLA-II LOH favors the loss of alleles with larger neopeptide repertoires, underscoring the importance of the HLA-II loci and the CD4+ T cell response in lung cancer.

READ THE FULL ARTICLE AT https://doi.org/10.1126/science.adi3808

y , The immunogenetics of lung cancer risk. Genetic epidemiological analyses in the UK Biobank and FinnGen coupled with multimodal genomics reveal that HLA-II heterozygosity is associated with reduced lung cancer risk in smokers. The data suggest that heterozygous HLA allomorphs promote immune control of early neoplasia through presentation of diverse smoking-related antigens. Thus, genetic variation in immunosurveillance is a key cancer risk factor. APC, antigen-presenting cell. Krishna et al., Science 383, 847 (2024)

23 February 2024

1 of 1

RES EARCH

RESEARCH ARTICLE

◥

IMMUNOGENETICS

An immunogenetic basis for lung cancer risk Chirag Krishna1†, Anniina Tervi2†, Miriam Saffern3,4†, Eric A. Wilson3,4,5†, Seong-Keun Yoo3,4,5, Nina Mars2, Vladimir Roudko3,4, Byuri Angela Cho3,4,5, Samuel Edward Jones2, Natalie Vaninov3,4, Myvizhi Esai Selvan6, Zeynep H Gümüs ¸6,7, FinnGen§, Tobias L. Lenz8, Miriam Merad3,9,4,10,11, Paolo Boffetta12,13, Francisco Martínez-Jiménez14,15, Hanna M. Ollila1,2,16,17*‡, Robert M. Samstein3,4,7,18*‡, Diego Chowell3,9,4,5*‡

23 February 2024

1 of 17

,

Krishna et al., Science 383, eadi3808 (2024)

y

*Corresponding author. Email: [email protected] (D.C.); [email protected] (R.M.S.); hanna.m.ollila@ helsinki.fi (H.M.O) †These authors contributed equally to this work. ‡These authors contributed equally to this work. §FinnGen consortium members are listed in the supplementary materials.

y g

Broad Institute of MIT and Harvard, Cambridge, MA 02142, USA. 2Institute for Molecular Medicine, Finland (FIMM), HiLIFE, University of Helsinki, Helsinki 00290, Finland. 3The Marc and Jennifer Lipschultz Precision Immunology Institute, Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA. 4Department of Immunology and Immunotherapy, Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA. 5Icahn Genomics Institute, Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA. 6Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA. 7 Center for Thoracic Oncology, Tisch Cancer Institute, Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA. 8 Research Unit for Evolutionary Immunogenomics, Department of Biology, Universität Hamburg, 20146 Hamburg, Germany. 9 Department of Oncological Sciences, Tisch Cancer Institute, Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA. 10Division of Hematology and Medical Oncology, Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA. 11 Human Immune Monitoring Center, Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA. 12Department of Medical and Surgical Sciences, Alma Mater Studiorum University of Bologna, 40138 Bologna, Italy. 13Stony Brook Cancer Center, Stony Brook University, New York, NY 11794, USA. 14Vall d’Hebron Institute of Oncology, Barcelona 08035, Spain. 15Hartwig Medical Foundation, Amsterdam 1098 XH, the Netherlands. 16Center for Genomic Medicine, Massachusetts General Hospital and Harvard Medical School, Boston, MA 02114, USA. 17Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Boston, MA 02114, USA. 18 Department of Radiation Oncology, Mount Sinai Hospital, New York, NY 10029, USA.

y

1

The role of smoking in lung cancer risk and mortality was initially defined in seminal work by Richard Doll over 70 years ago (5) and validated in countless studies since then, including recent meta-analyses highlighting a severe doseresponse relationship between the number of packs smoked and mortality of lung cancer and other diseases (6). Genetic studies have implicated germline genetic variation in lung cancer risk, including mutations in TP53, epidermal growth factor receptor (EGFR), and others (1, 7). Together, these studies have established lung cancer as a multifactorial disease with diverse genetic and environmental triggers (8). However, our understanding of the full spectrum of lung cancer risk factors and how they interact remains incomplete; for example, genome-wide association studies (GWAS) explain only a tiny proportion of the genetic variability in lung cancer risk (9). Indeed, there exists wide variability in lung cancer risk even among smokers (9, 10). The importance of the immune system in conferring protection against pathogens is well established (11). However, there is a longstanding debate regarding whether the immune system also protects against cancer. The cancer immunosurveillance hypothesis, initially developed by Ehrlich, Thomas, and Burnet (12–16), posits that lymphocytes constantly survey tissues for neoplastic cells presenting mutation-derived neoantigens, an activity that could trigger an effective immune response that eliminates developing cancers. The cancer immunoediting theory suggests that the immune system plays dual protective and promoting roles in neoplastic transformation (17). Moreover, large cohort studies have noted an increased risk of diverse infection-related and unrelated cancer among solid organ trans-

g

L

ung cancer is currently the leading cause of worldwide cancer mortality (1–3). Although diagnosis rates for advanced-stage disease continue to decline, rates for earlystage disease have increased (2), highlighting the need for research clarifying the factors underpinning lung cancer risk. Smoking causes lung cancer through DNA damage and other mechanisms and accounts for more than 80% of lung cancer deaths (4).

p

Cancer risk is influenced by inherited mutations, DNA replication errors, and environmental factors. However, the influence of genetic variation in immunosurveillance on cancer risk is not well understood. Leveraging population-level data from the UK Biobank and FinnGen, we show that heterozygosity at the human leukocyte antigen (HLA)-II loci is associated with reduced lung cancer risk in smokers. Finemapping implicated amino acid heterozygosity in the HLA-II peptide binding groove in reduced lung cancer risk, and single-cell analyses showed that smoking drives enrichment of proinflammatory lung macrophages and HLA-II+ epithelial cells. In lung cancer, widespread loss of HLA-II heterozygosity (LOH) favored loss of alleles with larger neopeptide repertoires. Thus, our findings nominate genetic variation in immunosurveillance as a critical risk factor for lung cancer.

plant recipients (18). A plausible interpretation of these seminal studies is that abrogation or differences in the strength of immune surveillance may lead to variations in cancer risk (19–21). Lung cancer is an exemplary disease for the study of immunosurveillance in cancer because the healthy lung is among the most heavily T cell–infiltrated tissues (22). Additionally, metastatic lung cancers demonstrate encouraging responses to immune checkpoint blockade (ICB) agents targeting T cells via the PD-L1/PD-1 and CTLA-4 axes (23–25), highlighting an important role for neoantigen-driven cytotoxic activity in the disease. Furthermore, key studies investigating the basis for ICB response in lung cancer and other tumor types (26) have shown that the elevated mutation rate caused by smoking (23) promotes increased visibility of neoantigens to cytotoxic T cells (27). Thus, these previous studies suggest that there may exist interactions between smoking and the immune system in the development of lung cancer. Yet, such interactions—and the role of the immune system in cancer risk in general—are not well understood. One clue as to how the immune system is involved in lung cancer risk has arisen from GWAS, which have implicated individual singlenucleotide polymorphisms (SNPs) and alleles of the human leukocyte antigen (HLA) class I (HLA-I) and II (HLA-II) genes in lung cancer susceptibility (28–30). The HLA genes are highly polymorphic (31) and encode the major histocompatibility complex (MHC) molecules, which serve as critical gatekeepers of the adaptive immune response through the presentation of self- and foreign antigens for recognition by T cells. Previous work has highlighted the somatic loss of HLA-I as a mechanism of immune evasion in lung cancer (32, 33). Furthermore, HLA-I and HLA-II genotypes influence the oncogenic-driver landscape (34, 35). However, whether and how HLA polymorphism interacts with smoking and other risk factors in driving lung cancer risk over time has not currently been addressed. The heterozygote advantage hypothesis is a foundational principle of the evolution of the HLA system and of HLA-mediated protection against disease. According to this hypothesis (36), individuals heterozygous at HLA are afforded greater protection against disease because they present more antigens for T cell recognition through their two different HLA allomorphs than do homozygous individuals and consequently clear infected or neoplastic cells more efficiently. Although evidence for heterozygote advantage theory has been demonstrated most clearly in the context of clinical outcomes among individuals who already have the disease [i.e., in delaying progression to AIDS among individuals with HIV (37, 38), clearance of hepatitis B (39), or response to ICB in metastatic cancer (40–46)], whether there

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

Table 1. Clinical and demographic characteristics of lung cancer cases and controls in UK Biobank and FinnGen.

Characteristic

UK Biobank Full Cohort

UK Biobank Lung Cancer

FinnGen Full Cohort

FinnGen Lung Cancer

Total individuals Healthy controls

391,182 384,928

2468

183,163 179,233

3480

y g

..................................................................................................................................................................................................................... ..................................................................................................................................................................................................................... .....................................................................................................................................................................................................................

.....................................................................................................................................................................................................................

Smoking status Current (%) 40,674 (10.5%) 1107 (44.9%) 50,179 (27.4%) 2464 (70.8%) Former (%) 135,410 (35.0%) 1035 (41.9%) 43,493 (23.7%) 666 (19.1%) ..................................................................................................................................................................................................................... Never (%) 211,312 (54.5%) 326 (13.2%) 89,041 (48.6%) 350 (10.1%) ..................................................................................................................................................................................................................... ..................................................................................................................................................................................................................... .....................................................................................................................................................................................................................

.....................................................................................................................................................................................................................

Sex Male 177,259 (54.2%) 1270 (48.5%) 89,052 (48.6%) 2724 (78.3%) Female 210,137 (45.8%) 1297 (51.4%) 93,661 (51.1%) 756 (21.7%) ..................................................................................................................................................................................................................... ..................................................................................................................................................................................................................... .....................................................................................................................................................................................................................

.....................................................................................................................................................................................................................

Age (I.Q.R.)

58 (50-63)

67 (63-71)

63 (49-74)

75 (70-80)

.....................................................................................................................................................................................................................

23 February 2024

2 of 17

,

Lung cancer subtype ..................................................................................................................................................................................................................... Adenocarcinoma (%) 700 (28.3%) 801 (23.0%) ..................................................................................................................................................................................................................... Squamous (%) 338 (13.7%) 679 (19.5%) ..................................................................................................................................................................................................................... Small cell (%) 192 (7.8%) 336 (9.7%) ..................................................................................................................................................................................................................... Other/unknown (%) 1238 (50.2%) 1664 (47.8%) .....................................................................................................................................................................................................................

y

Krishna et al., Science 383, eadi3808 (2024)

y

We sought to examine the effects of HLA heterozygosity on lung cancer risk at the population level, defined here as the odds ratio (OR) or hazard ratio (HR) corresponding to diagnosis or death due to lung cancer as a function of HLA zygosity and other clinical variables. Thus, we first assembled individual-level genetic, clinical, environmental, and longitudinal clinical data from the UK Biobank and FinnGen (48, 49) (table 1). The UK Biobank and FinnGen are distinctive in size and scope, with rich longitudinal phenotypic and healthrelated information available through linkage to medical records for each participant followed over time. In addition, the UK Biobank and FinnGen consist of roughly 500,000 and 350,000 genotyped individuals from the UK and Finland, respectively, including the imputation of genotypes at the classical HLA-I (HLA-A, HLA-B, HLA-C) and HLA-II (HLA-DRB1, HLA-DQB1, HLA-DQA1, HLA-DPB1, HLA-DPA1) genes. FinnGen in particular has employed a Finnish-specific reference panel for HLA imputation (50). Thus, the UK Biobank and FinnGen are well suited to address whether HLA heterozygosity affects cancer risk. After performing quality control of HLA genotypes as recommended by the UK Biobank and

was the proportion of smokers in each category among lung cancer cases (for instance, 70.8% current smokers in FinnGen as compared with 41.8% current smokers in the UK Biobank). The percentage of male lung cancer patients (78.3%) far exceeded the percentage of female lung cancer patients (21.7%) in FinnGen, whereas the distribution was more balanced in the UK Biobank. Although several prior studies have used the imputed HLA genotypes provided by the UK Biobank for bespoke analyses (54), we undertook several additional quality checks to validate the quality of imputed HLA genotyping in the UK Biobank. We first compared the allele frequency of 2-field (i.e., four-digit) alleles in the UK Biobank to population-level allele frequencies from the Allele Frequency Net Database (AFND) (55) (Fig. 1A); the frequencies were highly correlated (P < 0.0001; Spearman’s r = 0.91), suggesting that allele genotyping in the UK Biobank is representative of the allele genotypes in the wider UK population. We observed similar results comparing allele frequencies in FinnGen with Finnish population allele-frequency data (P < 0.0001; Spearman’s r = 0.85) (Fig. 1B). We then directly compared allele frequencies in the UK Biobank with those in FinnGen (Fig. 1C); allele frequencies were generally correlated (P < 0.0001; Spearman’s r = 0.66) except for a few HLA-I and HLA-II alleles that approached allele frequencies of up to 10% in the individual cohorts. The strong correlations between allele frequencies in the UK Biobank or FinnGen with the general

g

Immunogenetic and demographic characterization of individuals in the UK Biobank and FinnGen

filtering out any individuals with a cancer diagnosis before the start of the UK Biobank study (48) (fig. S2; Materials and methods), a cohort of 391,182 individuals was identified for further analysis. The primary clinical endpoint of interest in our study was a first diagnosis or death due to lung cancer [defined by ICD-10 codes (ICD, International Statistical Classification of Diseases and Related Health Problems)] over a roughly 14-year follow-up period, with participants recruited between March 2007 and October 2011. We documented 2468 individuals in the UK Biobank fitting these criteria of lung cancer case, with the remaining individuals designated as healthy controls (N = 384,928) after excluding individuals with missing data (fig. S2). Consistent with previous reports, the most common histological subtypes of lung cancer were adenocarcinoma (N = 700), squamous cell carcinoma (N = 338), and small cell carcinoma (N = 192); with the remaining patients representing other or missing histologies. Of the lung cancer cases, 86.8% recorded as current or former smokers and the remainder as never-smokers. Gender was roughly evenly split between males and females in both cases and controls. To replicate our findings from the UK Biobank, we assembled lung cancer case and control data from FinnGen (51–53) by using the same criteria applied to the UK Biobank. We identified 3480 lung cancer cases in FinnGen (N = 183,163 total individuals after filtering; fig S3). Although the distribution of lung cancer subtypes in FinnGen was similar to that in the UK Biobank, a key difference

p

exists a protective effect of HLA heterozygosity against the development of lung cancer (or other cancer types) is currently unknown. Such an effect, together with established risk factors such as smoking and age, may underscore HLA heterozygosity and the immune system in general as critical factors in lung cancer risk. In this study, we hypothesized that heterozygosity at the HLA genes is associated with reduced lung cancer risk over time, on the basis of the assumption that two different HLA allomorphs will present more neoplastic antigens than will a single HLA allomorph (47), thus increasing the likelihood of a cytotoxic reaction against mutated neoplastic cells. To test this hypothesis, we leveraged clinical, genetic, environmental, and longitudinal data from two largescale population cohorts: the UK Biobank (N = 391,182) and FinnGen (N = 183,163) (fig. S1). We then employed multiple approaches, including fine-mapping and structural analyses of the peptide binding groove and genomic profiling of the healthy lung through single-cell RNA sequencing (scRNA-seq), to clarify the mechanisms underlying HLA-mediated protection against lung cancer. Lastly, we investigated somatic loss of heterozygosity (LOH) of the HLA-I and HLA-II loci in lung cancer tumors from The Cancer Genome Atlas (TCGA) cohort, the Pan-Cancer Analysis of Whole Genomes (PCAWG) cohort, and the Hartwig Medical Foundation cohort.

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

A

D

B

C

E

F

G

p g

I

y

H

y g y ,

Fig. 1. HLA genotype and associations with lung cancer risk in UK Biobank and FinnGen. (A) Correlation of HLA allele frequencies in the UK Biobank with mean allele frequencies across England, Scotland, and Wales was obtained from the Allele Frequency Net Database (AFND). P value computed using Spearman correlation. (B) Correlation of HLA allele frequencies in FinnGen with allele frequencies from Finland obtained from AFND. P value calculated with Spearman correlation. (C) Correlation of HLA allele frequencies in UK Biobank with allele frequencies in FinnGen. P value calculated with Spearman correlation. (D) Rates of heterozygosity at four-digit allele resolution in UK Biobank. (E) Rates of heterozygosity at four-digit allele resolution in FinnGen. HLA-DPA1 genotypes were not imputed in FinnGen and are thus left gray. Krishna et al., Science 383, eadi3808 (2024)

23 February 2024

(F) Distribution of age at onset among lung cancer cases compared with age at first assessment in UK Biobank. (G) Distribution of age at onset among lung cancer cases compared with age at first assessment in FinnGen. (H) Multivariable logistic regression analyses testing heterozygosity at the indicated locus together with all clinical and demographic covariates for associations with lung cancer case and control status in UK Biobank. Forest plots depict odds ratio from logistic regression and 95% CI. (I) Multivariable logistic regression analyses testing heterozygosity at the indicated locus and all clinical and demographic covariates for associations with lung cancer case and control status in FinnGen. Forest plots depict OR from logistic regression and 95% CI. 3 of 17

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

y ,

4 of 17

y g

23 February 2024

y

Having validated the high quality of HLA genotyping in the UK Biobank, we next asked whether HLA heterozygosity is associated with reduced lung cancer risk in the UK Biobank by performing a multivariable logistic regression analysis. We controlled for clinical and demographic covariates that are known to influence lung cancer risk and outcomes in the UK Biobank (62). We reasoned that a multivariable model accounting for all covariates would be especially critical given the drastic difference in age between lung cancer cases in the UK Biobank (median 67) (table 1 and Fig. 1F) and those in FinnGen (median 75) (table 1 and Fig. 1G). Specifically, we fit an independent multivariable logistic regression for each HLA locus testing heterozygosity at the locus as a predictor together with clinical and demographic covariates, including smoking status (Materials and methods) (Fig. 1H). The outcome was a binary variable indicating diagnosis or death due to lung cancer (N = 2468 in the UK Biobank) or healthy control (N = 384,928 in the UK Biobank) (table 1). In addition to the eight multivariable models fit for each HLA-I (HLA-A, HLA-B, HLA-C) and HLA-II (HLA-DRB1, HLA-DQB1, HLA-DQA1, HLA-DPB1, HLA-DPA1) locus, we fit two additional models for maximal heterozygosity at HLA-I (6 alleles versus 11 million SNPs corresponding to a mean heterozygosity (p) of 0.0032 per site across 10-kb windows (fig. S2A).

y

1

histone H3 lysine 4 (H3K4me3) and H3 lysine 36 (H3K36me3) (9). In mice, the SET domain that performs these modifications is necessary for PRDM9-dependent DSB formation (9). PRDM9 also has two N-terminal domains that are essential for its role in recombination, but their specific functions remain unclear (10–12). This four-domain gene structure originated before the ancestor of vertebrates but has been lost, in whole or in part, at least 13 times independently (13). In species that are missing N-terminal domains [e.g., several lineages of ray-finned fish (12, 13)], lack PRDM9 entirely [e.g., birds (14, 15)], or carry a pseudogene for PRDM9 [e.g., canids (16)], crossover rates are elevated near CpG islands and transcription start sites (TSS) (12, 15, 17). These promoter-like features are typically marked by H3K4me3, but not H3K36me3 (18), through molecular processes that serve purposes other than initiating meiotic recombination (19). A possible explanation is that in the absence of the marks made by PRDM9, the machinery responsible for making DSBs defaults to residual sites of H3K4me3 (7) either because the mark causes its recruitment, say through a reader protein, or simply because such regions are more accessible. Consistent with the notion that species lacking PRDM9 rely on other H3K4me3 marks, when Prdm9 is knocked out in B6 mice or SHR rats, ~92% and ~99% of DSBs occur at residual sites of H3K4me3, respectively (6, 7). The knockout experiments in mice and rats, alongside the report of a human mother homozygous for a nonfunctional PRDM9 (20), further indicate that in species with a complete PRDM9, there exists an alternative mechanism of recombination that can serve as a backup, albeit a less efficient one (7, 8, 21).

g

M

eiotic recombination begins with the intentional creation of double-strand breaks (DSBs) in the genome, the repair of which results in crossovers and noncrossovers (1). In vertebrates as well as other taxa, such DSBs are concentrated in short genomic segments (typically one or two kilobases) known as “recombination hotspots.” The mechanisms that specify the genomic location of hotspots differ among species. In mice, humans, and other mammals, the gene PRDM9 (PR/SET Domain 9) plays a key role (2–6). In mouse and rat, outside of the pseudoautosomal region (6, 7), almost all hotspot locations are determined by PRDM9 binding. For example, congenic mouse strains with different Prdm9 alleles share only ~1% of their DSB hotspots (7), and introducing a human PRDM9 allele into B6 mice shifts the location of >97% of DSB hotspots (8). PRDM9 is a zinc finger protein that binds DNA with sequence specificity and trimethylates

Hoge et al., Science 383, eadj7026 (2024)

p

In some mammals, notably humans, recombination occurs almost exclusively where the protein PRDM9 binds, whereas in vertebrates lacking an intact PRDM9, such as birds and canids, recombination rates are elevated near promoter-like features. To determine whether PRDM9 directs recombination in nonmammalian vertebrates, we focused on an exemplar species with a single, intact PRDM9 ortholog, the corn snake (Pantherophis guttatus). Analyzing historical recombination rates along the genome and crossovers in pedigrees, we found evidence that PRDM9 specifies the location of recombination events, but we also detected a separable effect of promoter-like features. These findings reveal that the uses of PRDM9 and promoter-like features need not be mutually exclusive and instead reflect a tug-of-war that is more even in some species than others.

In mammals in which PRDM9 is known to direct recombination, the zinc finger domain of the protein shows evidence of positive selection at residues in contact with DNA, driving rapid evolution of the binding affinity (3, 12). Modeling work suggests that this rapid evolution is an expected consequence of the roles of PRDM9 in recombination (22, 23). Similar evidence for positive selection is also seen in all nonmammalian vertebrates with a complete PRDM9 ortholog that have been surveyed to date (and only in those species with a complete ortholog) (12, 13). These findings suggest that PRDM9 directs recombination outside of mammals (12). We tested this prediction by examining recombination patterns in an exemplar species with a single, complete PRDM9 ortholog, the corn snake (Pantherophis guttatus). We inferred historical recombination rates from patterns of linkage disequilibrium (LD) across the genome and identified crossover events in pedigrees.

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

C

A

B

p g

Hoge et al., Science 383, eadj7026 (2024)

23 February 2024

2 of 10

,

In our unrelated sample of corn snakes, expected pairwise LD decays on the order of 10 kb (fig. S5). Given the levels of heterozygosity,

LD-based inferences provide an estimate of the population recombination rate, which reflects an average over the many generations ancestral to the sample (32). In this respect, the immense diversity of PRDM9 zinc finger alleles presents a challenge, because each individual allele may be too rare or short-lived to have left a discernible footprint in the decay of LD. Nonetheless, the set of computationally predicted PRDM9 binding sites is associated with a significant increase in the mean population recombination rate, in genomic regions far (>10 kb) from a promoter-like feature (Fig. 2A and fig. S10). Despite their diversity, corn snake alleles of PRDM9 sometimes share a subset of fingers in common (Fig. 1C). These common submotifs may be helpful in identifying binding sites of PRDM9 alleles (33). To hone in on such shared features, we considered the predicted binding affinities of all sets of five or more consecutive fingers observed in at least five alleles [fig. S11; see section 3.3 in (25)]. The predicted binding sites of each of these “shared zinc fingers” are associated with increased recombination, and in many cases, the decay with distance is steeper than it is for the union of binding sites from all zinc finger alleles (Fig. 2B). Of particular

y

The population recombination rate increases both near PRDM9 binding sites and near promoter-like features

this scale of LD decay suggests that the ratio of mutation to recombination is comparable to, or perhaps a bit lower than, what is found in humans and other species in which recombination rates have been reliably estimated from patterns of LD (29). Using the genetic variation found in the 24 unrelated individuals, we therefore proceeded to infer LD-based recombination maps using the programs LDhelmet (29) and Pyrho (30). A comparison of the genetic maps obtained by the two approaches suggested that estimates based on LDhelmet may be more reliable [fig. S6 (25)]. In what follows, we therefore relied on results from LDHelmet after excluding any 10-kb region in which the recombination rates estimated by the two methods differed by >10-fold (2.2% of the autosomal genome); qualitative results were similar if based on the Pyrho map instead (fig. S7). At a 10-kb scale, the resulting map has a mean estimated population recombination rate of 0.0068/bp and a median of 0.0027/bp (fig. S8A). As expected, recombination rates per base pair increase with GC content (fig. S8B) and are higher on microchromosomes [defined as chromosomes 10 kb) from any predicted PRDM9 binding site (Fig. 2A). The magnitude of the increase dwarfs that at PRDM9 but is in fact not directly comparable: CpG islands and TSS are readily identified compared with true PRDM9 binding sites and are relatively stable over evolutionary time scales. By contrast, PRDM9

y

note is the clear elevation in the mean rate around binding sites of a sequence of 11 zinc fingers that is found in five different PRDM9 alleles (henceforth, the “Shared 11-ZF” allele), including PRDM9-A, the most common allele in our sample (Fig. 2B). Thus, despite the diversity of PRDM9 zinc finger alleles, patterns of LD support the use of PRDM9 binding in directing recombination in snakes.

1-kb windows (thinned to be 10 kb apart) and the predictors are the (binary) presence or absence of one or more predicted PRDM9 binding sites, TSS, or CpG islands. Covariates include the background recombination rate (1-Mb scale), GC content, and distance to telomeres [see (25)]. Results are reported for data from the autosomes (circles), only scaffolds assigned to macrochromosomes (squares), and only microchromosomes (diamonds). (D and E) Overlap of hotspots (D) and matched coldspots (E) far from a promoter-like feature with the predicted binding sites for the Shared 11-ZF allele. The observed values are shown with solid lines. The overlap expected by chance is shown with the shaded distribution and is based on 500 replicates, in which each hotspot (D) or coldspot (E) was placed at random within 5 Mb of the original location conditional on there not being a gap in the genome sequence [see (25)]. Although hotspots and coldspots were matched for base composition [see (25)], that need no longer be the case once we conditioned on them lying far from a promoter-like feature, driving the slight difference between the null distributions in (D) and (E).

y g

Fig. 2. An increase in the population recombination rate is seen around both PRDM9 binding sites and promoter-like features. (A) Mean population recombination rate in 100-bp windows as a function of distance to the nearest predicted PRDM9 binding site (blue) or promoter-like feature (orange). When considering one feature, we conditioned on windows >10 kb from the other feature; thus, when focusing on predicted PRDM9 binding sites, we only considered windows that are >10 kb from promoter-like features (i.e., TSS or CpG island). The recombination rate is relative to the mean rate 8 to 10 kb away. Shaded regions represent the central 95% CI obtained by bootstrapping [see (25)]. (B) Mean population recombination rate in 100-bp windows as a function of distance from predicted binding sites for sets of zinc fingers shared among PRDM9 alleles. This plot is conditional on the windows being far from promoter-like features. The “Shared 11-ZF” allele shared among five PRDM9 alleles is shown in purple, and the set of all PRDM9 alleles, equivalent to the curve in (A), is shown in blue. (C) Point estimate and 95% CI for the coefficients of a linear model, in which the response variable is the (log) recombination rate in

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

B

A

g

rat snake (bottom). Note that rattlesnakes, a sister species to European viper, were included only to infer the ancestral state in substitutions [see (25)]. GC* around hotspots that are close (3 kb away from the center.

p

Fig. 3. Footprints of recombination in divergence data. (A) Ratio of losses in the corn snake lineage relative to the black rat snake (magenta) and the ratio of gains in the corn snake lineage relative to the black rat snake (green) for motifs enriched in recombination hotspots far from promoter-like features (>10 kb) ordered by motif enrichment significance (top to bottom). The 95% CIs were obtained by bootstrapping >5 Mb regions with at least one gain or loss event in either of the two lineages. Motifs that are close matches to subsets of at least one PRDM9 allele (Pearson correlation > 0.8; fig. S19) are marked with an asterisk. (B) Increased flux to GC (GC*) as a function of distance from corn snake autosomal hotspots, in sliding windows of 500 bp with a 100-bp offset, for the lineages leading to European viper (top), corn snake (middle), and black

C

y

Because the chromosome on which a DSB occurs is repaired based on information from its homolog, recombination converts strong PRDM9 binding sites to weaker PRDM9 binding sites in heterozygotes (36, 37). In the absence of any countervailing force, this process leads to the more rapid loss of PRDM9 binding sites over evolutionary time than expected under neutrality (3, 38). To look for a signal of this phenomenon in the lineage ancestral to corn snakes, we used a multispecies whole-genome alignment that we had previously generated, which includes corn snake and the closely related black rat snake, Pantherophis obsoletus (39). First, we focused on predicted binding sites of the Shared 11-ZF allele, because it shows a clear association with increased population recombination rates, suggesting that it may be old enough to also have left a footprint in divergence data. There are more losses than gains of the binding sites of this allele in the corn snake lineage (P = 0.0018 using a permutation test; fig. S18), and similarly, an excess of losses over gains is seen for 16 of the 22 PRDM9 alleles (if we consider them to be independent, P = 0.019, using a one-tailed Wilcoxon signedrank test). Next, we identified eight kmers significantly overrepresented in recombination hotspots that are far from promoter-like features relative to their matched coldspots (25); for seven of the eight, there are more losses 4 of 10

,

23 February 2024

Divergence at PRDM9 binding sites

y

Hoge et al., Science 383, eadj7026 (2024)

We also used the LD-based genetic map in corn snakes to infer the presence of 13,580 autosomal recombination hotspots, defined as genomic regions of >2 kb with recombination rates at least 5-fold the background rate of the flanking regions (25). Given these criteria, the median estimated heat is 8.6-fold and the median estimated length ~3 kb (fig. S16). Hotspots far (>10 kb) from promoter-like features overlap with predicted PRDM9 binding sites more often than expected by chance (P = 0.002, as assessed by reshuffling hotspot locations 500 times within 5 Mb of the starting location to control for background recombination rate; fig. S17A). This overlap cannot be explained simply by local base composition effects, because GC-matched coldspots far from promoterlike features overlap with PRDM9 binding sites no more than would be expected by chance (P = 0.61, fig. S17A). The same finding was obtained for sites predicted to be bound by the Shared 11-ZF allele (hotspots P = 0.002, coldspots P = 0.3; Fig. 2, D and E). Analogously, hotspots overlap promoter-like features that are far from any PRDM9 binding site significantly more than would be expected by chance (P = 0.002), but the same is not true for GCmatched coldspots (P = 0.34, fig. S17B). Thus, as expected from the analysis of population recombination rates, hotspot locations inferred from LD are associated with both predicted PRDM9 binding sites and promoter-like features.

y g

rate in 1-kb bins [Fig. 2C; P < 10−22 for PRDM9 binding sites and CpG islands and P = 0.02 for TSS; see (25)], and their interaction term is not significantly different from 0 (P = 0.18). The model further reveals an apparent difference between the 10 microchromosomes and eight macrochromosomes (24): Despite a high density of genes on microchromosomes (31), there appears to be a weaker effect of CpG islands and TSS (controlling for background recombination rates), but a comparable effect of predicted PRDM9 binding sites (Fig. 2C). These findings stand in contrast to the conclusion reached in an analysis of LD patterns in the prairie rattlesnake (Crotalus viridis), in which, based on three of the zinc fingers of the PRDM9 reference genome sequence, the authors concluded that PRDM9 directs recombination to promoter-like features (35). We reanalyzed their genomic data after collecting six complete prairie rattlesnake PRDM9 alleles from three individuals (fig. S12). Although we confirmed that rattlesnake PRDM9 alleles are predicted to bind GC-rich genomic regions (fig. S13) such as CpG islands, we found separable effects of PRDM9 binding sites and promoter-like features on population recombination rates, as in corn snakes (fig. S14). During our analyses, however, we noted that the prairie rattlesnake genome assembly may be unreliable around CpG islands (fig. S15), so a firm conclusion for this second snake species awaits further validation.

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

LD-based genetic maps represent a mixture of the effects of many different PRDM9 zinc finger alleles, past and present, and LD-based estimates can be biased in the presence of population structure or changes in population size (29, 43–45), as well as by subtle errors in the reference genome (fig. S15). We therefore sought to confirm our findings of a dual usage of PRDM9 and promoter-like features using a more direct and robust approach. We collected genome sequences from two corn snake pedigrees with five F1s each, as well as two F2s from one of the pedigrees, thereby obtaining information about 24 meioses (Fig. 1B). Calling crossovers from phase switches of informative markers [following (46); see (25)], we identified 324 crossover events, of which 172 could be delimited to within an interval of 10 kb). In the second column (orange) is the overlap of crossovers with promoter-like features far (>10kb) from the PRDM9 binding sites of alleles carried by the parents. The solid line is the observed overlap for the subset of crossovers, n, that satisfy the criteria. The frequency distribution represents the overlap for 3000 sets of simulated crossovers obtained by placing the observed interval lengths down at random within 5 Mb of the original crossover interval, conditional on it containing at least two informative markers and there not being a gap in the genome sequence at that location [see (25)]. The rank of the observation relative to realizations under the null is given as a percentile (77). The three rows present results for all scaffolds, only those assigned to macrochromosomes, and only those assigned to microchromosomes [see (25)].

features, as previously reported in birds (15) and yeast (40). By contrast, recombination hotspots that are far from promoter-like features, a subset of which are likely specified by PRDM9 binding in corn snakes, show a much weaker signal of elevated GC* in the corn snake branch and no discernible GC* peaks in the other two snake lineages (Fig. 3B). This observation is consistent with corn snakes having two classes of hotspots: those near promoter-like features (and possibly other sites with H3K4me3 but not H3K36me3), which are conserved among distantly related species, and those at PRDM9 binding sites, which are rapidly repositioned because of changes in the binding affinity of PRDM9. Notably, the signal of GC* at hotspots overlapping promoter-like features is less pronounced on microchromosomes than on macrochromosomes (Fig. 3C), and this difference holds even after accounting for the increased background recombination rate on microchromosomes (fig. S21), consistent with a reduced use of promoter-like features on microchromosomes (Fig. 2C). A caveat to our interpretation is the possibility that the observed increase in recombination at promoter-like features is driven by PRDM9 binding at motifs that we failed to identify computationally. If so, then the marked increase in GC* at hotspots overlapping promoter-like features but not at PRDM9 binding sites points to a high level of stability in the PRDM9 binding to promoters. In the face of rapid turnover of the PRDM9 binding affinity, this explanation would therefore require that PRDM9 alleles binding promoters be strongly favored in corn snakes and in other snake lineages (35) (Fig. 3C). At odds with this possibility, the corn snake predicted PRDM9 binding sites do not overlap promoter-like features more often than would be expected by chance (fig. S13).

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

A

C

B

E p g

D

y

23 February 2024

which they occurred more frequently than would be expected by chance (P = 0.0006; Fig. 4A). Inversely, crossovers far from the predicted PRDM9 binding sites of the focal parent overlap promoter-like features more often than would be expected by chance (P = 0.007; Fig. 4B). Qualitative conclusions are the same if a larger set of 239 crossovers with a resolution 4 kcal/mol less stable than securine A (7) (corresponding to an equilibrium ratio of 7:1 of >103). Collectively, our data suggest that securamine A (1) is kinetically stable and that the ring-opening to securine A (7) is irreversible.

g

Completion of the synthesis by photochemical nitrene insertion

face of the lactam, which would require a facially selective reduction of the unsaturated g-lactam 15. A single diastereomer ( 1 H NMR analysis) of the hydrogenation product was obtained when 15 was treated with Crabtree’s catalyst [(tricyclohexylphosphine)(1,5-cyclooctadiene)(pyridine)iridium(I) hexafluorophosphate] and dihydrogen (46). Acylation of the hydrogenation product with 4-nitrobenzoyl chloride provided the acetanilide 39 (79% overall). Although NMR analysis suggested that the reduction proceeded with the desired stereoselectivity (3JH12–20 ~ 0 Hz), we carried out many attempts to obtain single crystals of 39 to confirm this assignment with x-ray analysis. However, the compound presented as thin, microcrystalline platelets unsuitable for conventional crystallography. Accordingly, we turned to microcrystal electron diffraction (MicroED) (47, 48), which confirmed that the acetanilide 39 possessed the relative configuration required for the proposed C–H amination (fig. S3). Our MicroED studies also revealed that the BOM ether populated four conformers in the solid state, which explains our inability to grow uniform single crystals of 39 (figs. S4 and S5). The azide 38 was prepared by means of sequential hydrogenation of 15 and exposure to tert-butyl nitrite (tBuONO) and trimethylsilyl azide (TMSN3; 64% from 15, 800 mg scale). Irradiation of solutions of 38 in dichloromethane with ultraviolet (UV) light (mediumpressure Hg lamp, 100 W, borosilicate filter) provided N5-BOM securamine A (14) as a 5:1 mixture of (Z)- and (E)-enamide diastereomers (43%). Removal of the BOM ether (boron trichloride) proceeded with concomitant opening of the pyrroloindoline to the indole, affording securine A (7, 70%; 17 steps, 12 purifications, 3.0% overall). NMR spectroscopic data for natural and synthetic securine A (7) in dimethyl sulfoxide-d6 (DMSO-d6) were in agreement (table S1) (2). However, whereas Christophersen reported that dissolution of natural securine A (7) in chloroform-d induced isomerization to securamine A (1) (2), we observed line broadening and no evidence of the pyrroloindoline framework (fig. S6). We reasoned that the solvent used might have contained acidic impurities (2), but we did not observe isomerization when phosphoric, trifluoroacetic, or hydrochloric acid were added to the NMR sample (fig. S6). Fortunately, we were able to obtain securamine A (1) by reordering the steps in the sequence. Treatment of the azide 38 with boron trichloride provided the NH imidazole 40 (86%). Irradiation of 40 with UV light then generated securamine A (1; 16%; 17 steps, 12 purifications, 1.4% overall). The diminished yield of this transformation relative to 38→14 (43%) points toward the free imidazole in 40, which may divert the nitrene from productive C–H insertion

p

the N-acyl chloroaminal 30, which was observed as the corresponding hemiaminal in LC-MS analysis. Enone hydrochlorination (30→31) was followed by elimination to the enamide (31→32) and ring contraction to provide the b,g-unsaturated-g-lactam 33. Tautomerization to the hydroxypyrrole 34 and protonation at C12 provided the observed product 29. On the basis of the available data, an alternate sequence proceeding through 35 instead of 32 cannot be excluded. This cascade reaction forms the cis-enamide, neopentylic alkyl chloride, g-lactam, and a nine-membered ring— four challenging substructures present in the securamines—in a single step. Selective reduction of the nitroarene in the presence of other reducible functionalities was achieved by treating the lactam 29 with tetrahydroxydiboron and 4,4′-bipyridine (39) to provide the complex lactam 15. Because intermediates 28 and 29 were difficult to obtain in analytically pure form, we advanced them without purification in preparative-scale experiments. By this approach, the lactam 15 was obtained in three steps and 49% yield from 27 (>20:1 diastereomeric ratio, 1H NMR analysis; 2.6 g scale). We envisioned that isomerization of the lactam 15 to an N-acyliminium ion, followed by ring closure (N13–C12 bond formation), would provide N5-BOM securamine A (14). However, we were unable to actualize this strategy, potentially because of the basicity of the imidazole and/or aniline. To circumvent this, we developed a two-step procedure to obtain the alternative N-acyliminium precursor 37 (Fig. 2C). Treatment of the aniline 15 with tertbutyldimethylsilyl trifluoromethanesulfonate (TBSOTf) and 2,6-lutidine provided the siloxypyrrole 36. Exposure of 36 to tin tetrachloride (40) at –78°C generated a 3:2:1 mixture of the b,g-unsaturated-g-lactam 37, 15, and 12-epi-15. Unfortunately, all attempts to induce cyclization of 37 by using Lewis or Brønsted acids [such as mercury(II) acetate or trifluoroacetic acid] resulted in reversion to 15. Additionally, we noted that 12-epi-15 spontaneously isomerized to 15 upon standing in dichloromethane-d2 at 23°C, suggesting that 15 is the thermodynamically favored stereoisomer. We presume that this epimerization proceeds through an aromatic hydroxypyrrole analogous to that of 34 (Fig. 2B).

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

SUPPLEMENTARY MATERIALS

science.org/doi/10.1126/science.adl6163 Materials and Methods Supplementary Text Figs. S1 to S12 Tables S1 to S14 References (51–70) Data S1 Submitted 27 October 2023; accepted 22 January 2024 10.1126/science.adl6163

6 of 6

,

We thank K. Hansen (Marbio, UiT The Arctic University of Norway) for kindly providing a sample of natural securamine I (6) and for helpful discussions. We thank S. Ohlinger (Wavefunction) for assistance with calculations. Funding: Financial support from the National Science Foundation (CHE-1954319 and Graduate Research Fellowship to N.M.B.) is gratefully acknowledged. Author contributions: B.W.A., N.M.B., and V.G. carried out the experimental work. B.Q.M and M.D.C. carried out MicroED analysis of 39. B.W.A., N.M.B., V.G., and S.B.H. wrote the manuscript. S.B.H. directed the study. Competing interests: None. Data and materials availability: Crystallographic data for 39 has been deposited in the Cambridge Crystallographic Data Centre (CCDC# 2294680). License information: Copyright © 2024 the authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original US government works. https://www.science.org/about/sciencelicenses-journal-article-reuse

y

23 February 2024

AC KNOWLED GME NTS

y g

Alexander et al., Science 383, 849–854 (2024)

1. M. L. Ciavatta et al., Mar. Drugs 18, 200 (2020). 2. L. Rahbaek, U. Anthoni, C. Christophersen, P. H. Nielsen, B. O. Petersen, J. Org. Chem. 61, 887–889 (1996). 3. L. Rahbæk, C.Christophersen, J. Nat. Prod. 60, 175–177 (1997). 4. K. Ø. Hansen et al., J. Nat. Prod. 80, 3276–3283 (2017). 5. K. Ø. Hansen et al., Nat. Prod. Commun. 16, 1934578X2199618 (2021). 6. S. C. Chaffee, “Synthetic studies toward the securamine alkaloids,” thesis, Yale University (2001). 7. P. Korakas, “Progress toward the total synthesis of securamine A: Construction of the securamine macrocyclic core,” thesis, Yale University (2004). 8. P. Korakas, S. Chaffee, J. B. Shotwell, P. Duque, J. L. Wood, Proc. Natl. Acad. Sci. U.S.A. 101, 12054–12057 (2004). 9. P. Aggarwal, “Studies directed toward total synthesis of securine A and securamine A,” thesis, Penn State Univeristy (2007). 10. S. Kikuchi, “Studies toward the biomimetic synthesis of marine alkaloids,” thesis, Harvard University (2011). 11. P. E. Murray, “Approaches to the synthesis of the marine alkaloid chartellamide A: A novel 2-(4-imidazoly)indole,” thesis, The University of Manchester (1998). 12. X. Lin, “Part one: Generation and cyclization of nitrogen radicals. Application to a formal total synthesis of (+/−)-peduncularine. Part two: Studies towards the total synthesis of chartelline A,” thesis, Pennsylvania State University (2002). 13. P. S. Baran, R. A. Shenvi, C. A. Mitsos, Angew. Chem. Int. Ed. 44, 3714–3717 (2005). 14. J. M. Rohde, “Synthetic approaches to aplydactone and chartellamide A inspired by biosynthetic hypotheses,” thesis, Scripps Research Institute (2005). 15. C. Sun, “Chapter one: Stereoselective total synthesis of the racemic and natural enantiomer of the marine tunicate alkaloid lepadiformine. Chapter two: Studies directed towards a total synthesis of chartelline A,” thesis, Pennsylvania State University (2005). 16. P. S. Baran, R. A. Shenvi, J. Am. Chem. Soc. 128, 14028–14029 (2006). 17. C. Sun, X. Lin, S. M. Weinreb, J. Org. Chem. 71, 3159–3166 (2006). 18. P. J. Black, E. A. Hecker, P. Magnus, Tetrahedron Lett. 48, 6364–6367 (2007). 19. J. E. Camp, “Chapter one: Studies directed towards a total synthesis of the chartellamides. Chapter two: Studies directed towards a total synthesis of the chartellines,” thesis, Pennsylvania State University (2008).

y

We devised synthetic routes to eight isolates representing the three skeletal classes in secu-

RE FERENCES AND NOTES

20. E. A. Hecker, “Studies toward the total synthesis of (±)-chartelline C and (−)-platensimycin,” thesis, University of Texas at Austin (2008). 21. S. Kajii, T. Nishikawa, M. Isobe, Chem. Commun. (27): 3121–3123 (2008). 22. R. A. Shenvi, “Pure and applied science in the chemical syntheses of marine alkaloids chartelline C and cortistatin A,” thesis, Scripps Research Institute (2008). 23. K. Iwasaki et al., Angew. Chem. Int. Ed. 51, 9160–9163 (2012). 24. Y. Nakane, A. Nakazaki, T. Nishikawa, J. Org. Chem. 85, 7534–7542 (2020). 25. H.-H. Liao, S. Miñoza, S.-C. Lee, M. Rueping, Chem. Eur. J. 28, e202201112 (2022). 26. I. Saito, T. Matsuura, M. Nakagawa, T. Hino, Acc. Chem. Res. 10, 346–352 (1977). 27. T. Hino, M. Nakagawa, Heterocycles 8, 743 (1977). 28. M. R. Iesce, F. Cermola, F. Temussi, Curr. Org. Chem. 9, 109–139 (2005). 29. L. Grassi, C. Cabrele, Amino Acids 51, 1409–1431 (2019). 30. H. H. Wasserman, K. Stiller, M. B. Floyd, Tetrahedron Lett. 9, 3277–3280 (1968). 31. H. H. Wasserman, M. S. Wolff, K. Stiller, I. Saito, J. E. Pickett, Tetrahedron 37, 191–200 (1981). 32. P. Kang, C. S. Foote, J. Am. Chem. Soc. 124, 9629–9638 (2002). 33. M. J. Davies, Biochem. Biophys. Res. Commun. 305, 761–770 (2003). 34. F. Souquet et al., Synthesis 52, 2970 (2020). 35. A. Umehara, H. Ueda, H. Tokuyama, Tetrahedron 79, 131809 (2021). 36. Y. He, Y. Chen, H. Du, L. A. Schmid, C. J. Lovely, Tetrahedron Lett. 45, 5529–5532 (2004). 37. M. M. Więcław, S. Stecko, Eur. J. Org. Chem. 2018, 6601–6623 (2018). 38. M. A. Blanchette et al., Tetrahedron Lett. 25, 2183–2186 (1984). 39. M. Jang, T. Lim, B. Y. Park, M. S. Han, J. Org. Chem. 87, 910–919 (2022). 40. N. Langlois, P. K. Choudhury, Tetrahedron Lett. 40, 2525–2528 (1999). 41. S. C. Patel, N. Z. Burns, J. Am. Chem. Soc. 144, 17797–17802 (2022). 42. D. A. Davydov, M. A. Giricheva, Y. B. Malysheva, G. K. Fukin, A. V. Budruev, J. Org. Chem. 88, 14998–15006 (2023). 43. T. J. Pearson et al., Science 381, 1474–1479 (2023). 44. B. C. G. Soderberg, Curr. Org. Chem. 4, 727–764 (2000). 45. T. G. Driver, Org. Biomol. Chem. 8, 3831–3846 (2010). 46. R. H. Crabtree, M. W. Davis, J. Org. Chem. 51, 2655–2661 (1986). 47. T. Gruene et al., Angew. Chem. Int. Ed. 57, 16313–16317 (2018). 48. C. G. Jones et al., ACS Cent. Sci. 4, 1587–1592 (2018). 49. S. Einarsson, B. Josefsson, P. Möller, D. Sanchez, Anal. Chem. 59, 1191–1195 (1987). 50. P. Ruiz-Sanchis, S. A. Savina, F. Albericio, M. Álvarez, Chemistry 17, 1388–1408 (2011).

g

Conclusions

ramines and securines. The hydrochlorination cascade (28→29) (Fig. 2A) we discovered installs many of the challenging functional groups in the targets in a single step while also achieving a productive remodeling of the skeletal framework of our intermediates. Sequential photochemical C–H amination and cycloaddition of singlet dioxygen enabled rapid access to the pyrroloindoline residue and hexacyclic scaffold of the targets. The success of the selective oxidation of the bromoimidazole residue lends support to the hypothesis that divergent biosynthetic oxidations of the indole-bearing securines and masked-indole securamines lead to skeletally distinct metabolites. Although we cannot recapitulate the previously reported securamine-securine equilibrium (2), our data make clear that synthetic samples of securines do not readily interconvert with their pyrroloindoline counterparts. Fortunately, the photochemical C–H amination we developed allowed us to circumvent this unexpected challenge. Pyrroloindolines are widespread in secondary metabolites (50), and we anticipate that this amination approach will be of general utility in accessing this substructure. This study also underscores the utility of MicroED as an indispensable tool for complex molecule structure determination.

p

resolve the aniline (±)-48 by acylation with (S)-(–)-fluorenylethylchloroformate (FLEC-Cl) on preparative scales (fig. S8) (49). This resolution provides a path to establish the absolute configuration of a broad range of isolates. Analogous to the securamine A (1) series, the C16bromoaniline 48 was converted to the azide 49 by reduction and diazotransfer (64% overall) (Fig. 3C). Securamine B (2) was obtained by removing the benzyloxymethyl ether protecting group (boron trichloride) and C–H amination (medium-pressure Hg lamp, 100 W, borosilicate filter; 29% from 49; 17 steps, 12 purifications, 1.8% overall). Alternatively, the sequence of C–H amination followed by protecting group removal provided securine B (8; 55% from 49; 17 steps, 12 purifications, 3.5% overall). When these C–H amination reactions were carried out in dichloromethane, we observed minor amounts of aryl chloride 50, potentially arising from homolysis of the C16–Br bond and abstraction of chloride from solvent. Fortunately, this was suppressed by using dibromomethane as solvent. Spectroscopic data for natural and synthetic securamine B (2) and securine B (8) were in agreement (tables S4 and S5). Direct photooxidation of securamine B (2) (rose bengal and methanol) provided securamine C (4, 22% from 48; 18 steps, 12 purifications, 1.4% overall). The reported propensity of the oxidized securamine scaffolds to act as a Michael acceptor (4) under ambient conditions motivated us to consider the conjugate reduction of securamine C (4) to provide securamine G (5) (3). Addition of tetrabutylammonium borohydride (Bu4NBH4) to solutions of securamine C (4) in dichloromethane generated securamine G (5, 50%; 19 steps, 13 purifications, 0.7% overall). Alternatively, bromination of securamine C (4) [N-bromosuccinimide (NBS)] provided securamine I (6, 34%; 19 steps, 13 purifications, 0.5% overall) (4). Natural securamine C (4) (2), G (5) (3), and I (6) (4) were characterized in chloroform-d, but we found that synthetic samples of these compounds were insoluble in chloroform-d (fig. S9). It seems plausible that the presence of variable amounts of methanol, water, and other organic impurities in samples of natural securamines C (4) and I (6) may have promoted their dissolution in chloroform-d (4). Graphical reproductions of NMR spectra for securamines A (1), B (2), and G (5) and securines A and B (7 and 8), are not available (2, 3). We therefore characterized them in DMSO-d6 (tables S6 to S8). We obtained natural securamine I (6) and found that it was identical to synthetic material by means of NMR analysis in DMSO-d6 and with LC-MS co-injection (figs. S10 to S12 and table S8).

RES EARCH

SOLAR CELLS

Homogeneous crystallization and buried interface passivation for perovskite tandem solar modules Han Gao1†, Ke Xiao1†, Renxing Lin1†, Siyang Zhao1, Wenliang Wang2, Sergey Dayneko3, Chenyang Duan1, Chenglong Ji2, Hongfei Sun1, Anh Dinh Bui4, Chenshuaiyu Liu1, Jin Wen1, Wenchi Kong1, Haowen Luo1, Xuntian Zheng1, Zhou Liu1, Hieu Nguyen4, Jin Xie2, Ludong Li1, Makhsud I. Saidaminov3, Hairen Tan1* Scalable fabrication of all-perovskite tandem solar cells is challenging because the narrowbandgap subcells made of mixed lead-tin (Pb-Sn) perovskite films suffer from nonuniform crystallization and inferior buried perovskite interfaces. We used a dopant from Good’s list of biochemical buffers, aminoacetamide hydrochloride, to homogenize perovskite crystallization and used it to extend the processing window for blade-coating Pb-Sn perovskite films and to selectively passivate defects at the buried perovskite interface. The resulting all-perovskite tandem solar module exhibited a certified power conversion efficiency of 24.5% with an aperture area of 20.25 square centimeters.

1 of 5

,

23 February 2024

y

Gao et al., Science 383, 855–859 (2024)

y g

*Corresponding author. Email: [email protected] †These authors contributed equally to this work.

y

National Laboratory of Solid State Microstructures, College of Engineering and Applied Sciences, Frontiers Science Center for Critical Earth Material Cycling, Nanjing University, Nanjing 210023, China. 2State Key Laboratory of Coordination Chemistry, Jiangsu Key Laboratory of Advanced Organic Materials, Chemistry and Biomedicine Innovation Center (ChemBIC), School of Chemistry and Chemical Engineering, Nanjing University, Nanjing, 210023, China. 3 Department of Chemistry, University of Victoria, Victoria, BC V8P 5C2, Canada. 4Research School of Electrical, Energy and Materials Engineering, College of Engineering and Computer Science, Australian National University, Canberra, NSW 2600, Australia.

The ratio of dimethylformamide (DMF) and dimethyl sulfoxide (DMSO) affected the crystallization and morphology of the gas-assisted blade-coated Pb-Sn perovskite films. We found that DMF:DMSO = 9:1 (v/v) was optimal for Pb-Sn perovskite film morphology and devices made by means of gas-assisted blade-coating (figs. S1 and S2 and table S1), whereas the inapposite solvent ratio gives rise to form the PbSn perovskite films with voids. We then studied the quality of perovskite films as a function of the blade-coating processing window (Dt), which we define as the sum of times required for blade-coating the perovskite ink (Dt1) and for transferring the wet film to a heating stage (Dt2) (fig. S3). While coating on the 2.5- by 2.5-cm substrates, both Dt1 and Dt2 = 5 s. The control ink with DMF:DMSO = 9:1 (v/v) showed a short processing window, which led to pinholes on blade-coated Pb-Sn perovskite films afterward (fig. S4). Incorporation of AAH (fig. S5) into precursor ink extended Dt to 100 s and resulted in pinhole-free Pb-Sn perovskite films (Fig. 1A and fig. S6) with large and vertically oriented grains (figs. S7 and S8). The processing window affected the performance of Pb-Sn PSCs. The control PSCs exhibited high sensitivity to Dt, and performance decreased steeply with increasing Dt, whereas the AAH Pb-Sn PSCs maintained similar performance with prolonged Dt (fig. S9 and table S2). Next, we studied uniformity of blade-coated Pb-Sn perovskite films on 6- by 6-cm substrates. The photoluminescence (PL) map of the AAH Pb-Sn perovskite film was more homogeneous and stronger than that of the control film (Fig. 1B and fig. S10). We patterned eight independent solar cells with an area of 0.48 cm2 on these substrates (fig. S11A, inset); the AAH Pb-Sn PSCs demonstrated a higher average PCE than that of the control devices (20.8 versus 17.2%), with a much smaller deviation (0.1 versus 0.5%) (Fig. 1C, fig. S11B, and table S3). Additionally, the external quantum efficiency (EQE) spectra also indicated improved uniformity of AAH films (fig. S12). We then sought to understand why AAH improved the uniformity of Pb-Sn perovskite films. We found that AAH has intermolecular interactions with nearly all perovskite precursor constituents, including PbI2, SnI2, formamidinium iodide (FAI), and solvents (Fig. 2A). X-ray photoelectron spectra (XPS) confirmed the presence of AAH in the crystalline film (fig. S13). Absorption spectra suggest negligible

g

1

(23). The introduction of solvents with a high boiling point and low vapor pressure, such as N-methyl-2-pyrrolidinone and g-butyrolactone, does slow the volatilization rate of solvents, but their own removal from thick Pb-Sn perovskite films presents challenges (24–26). Furthermore, nonradiative recombination at the buried bottom perovskite–hole transport layer (HTL) interface limits the performance of blade-coated Pb-Sn perovskite solar cells (27, 28). The topto-bottom crystallization direction of perovskite films makes it challenging to control the buried interface (26, 29), and predeposited passivation layers on HTL can be removed by solvent during a subsequent perovskite coating process (30). Therefore, it is crucial to develop strategies for effective passivation of buried interfaces. We posited that the challenges mentioned above can be addressed by using an additive that coordinates strongly with precursor ingredients to impede the crystallization, maintains near-neutral pKa values to prevent deprotonation of organic cations (where pKa is the negative logarithm of the acid dissociation constant Ka), exhibits high solubility in polar solvents to accumulate at the buried interface during crystallization, and consists of a short chain to enable efficient charge transport. The first three criteria are also applied in the selection of 20 Good’s buffers, a list of zwitterionic salts used in biochemistry research (31). To fulfill the requirement of a short chain, we identified aminoacetamide hydrochloride (AAH) as a suitable candidate from the Good’s buffer list. We report that the use of AAH from the Good’s buffer list enables a wide processing window and hence homogeneous crystallization of Pb-Sn perovskite films. We show that AAH sinks to the bottom of the perovskite interfaces and passivates the buried junction. We achieved a PCE of 21.4% for blade-coated

Uniform, large-area, narrow-bandgap Pb-Sn perovskite films

p

T

he power conversion efficiency (PCE) of all-perovskite tandem solar cells has increased (1–6) and reached 28% for laboratory-scale small area devices (7). However, the champion reported certified PCE of all-perovskite tandem modules (aperture area of 20.25 cm2) is only 21.7% (8–10). This substantial efficiency gap between individual cells and modules is caused by the nonuniformity of large-area lead-tin (Pb-Sn) narrow-bandgap (NBG) perovskites induced by their rapid uncontrollable crystallization during the gas-quenched blade-coating process (11–14). This phenomenon is attributed to the rapid aggregation of perovskite building blocks and solvent volatilization during crystallization (15, 16). Lewis base-acid adducts—which are molecules with a lone electron pair on nitrogen, oxygen, sulfur, or phosphorus—can coordinate Pb2+and Sn2+ and slow perovskite crystallization (17–20). Control over residual coordinating solvent during crystallization is necessary to achieve uniform Sn-containing perovskite films (21, 22). Because film deposition for upscaling is a time-demanding process, it is essential to extend the coating processing window to deposit uniform large-area perovskite films

Pb-Sn PSCs and 26.8% for 1-cm2 all-perovskite tandem solar cells made with fully scalable fabrication techniques. Last, we fabricated an all-perovskite tandem solar module with a certified PCE of 24.5% (aperture area of 20.25 cm2) and a high geometric fill factor (GFF) of 96.1%.

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

variation in the energy bandgap of Pb-Sn perovskite with AAH (fig. S14A). The x-ray diffraction patterns showed similar crystallinity and crystallographic orientation with a single perovskite phase (fig. S14B). The decreased binding energies of Pb 4f7/2 and Sn 3d5/2 (Fig. 2B) indicated the formation of Lewis acid-base AAH-Pb/Sn adducts, which in turn increased the solubility of inorganic salts (fig. S15) (14). Two-dimensional (2D) 1H-1H spin diffusion nuclear magnetic resonance (NMR) spectra revealed the interaction between AAH with perovskite’s organic cation FAI (fig. S16) (32). Fourier-transform infrared (FTIR) spectroscopy analysis demonstrated the presence of

t =40 s

t =100 s

t=70 s

B

Control

AAH

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0

0

31.5

21

0.86

78

20

30.0

29.0

0.76 Control AAH

74

72

29.5

0.78

PCE (%)

0.80

30.5

FF (%)

Jsc (mA cm-2)

Voc (V)

0.82

76

18 17 16

70

Control AAH

19

Control AAH

Control AAH

Fig. 1. Uniformity of blade-coated Pb-Sn perovskite films and solar cells. (A) Top-view scanning electron microscopy (SEM) images of blade-coated Pb-Sn perovskite films at different processing windows. (B) PL mapping images of control and AAH samples of 4.5- by 4.5-cm region at the center of 6- by 6-cm substrates. The scale bar shows the normalized PL intensity. (C) Performance uniformity of control and AAH PSCs deposited on 6- by 6-cm substrates, expressed by dispersion degrees of performance parameters calculated from current density-voltage (J-V) curves of eight individual PSCs (aperture area of 0.48 cm2, shown in fig. S11). Gao et al., Science 383, 855–859 (2024)

23 February 2024

2 of 5

,

31.0

0.84

y

C

1

y g

1 cm

1

Given the high solubility of AAH in polar solvents (31), and the top-to-bottom crystallization direction of perovskite films (29), our crystallization model proposes that AAH would mainly localize at the bottom interface (fig. S20 and supplementary text 2). Time-of-flight secondary-ion mass spectrometry (TOF-SIMS) showed a gradual increase in the content of AAH+ ions from the top to the bottom of perovskite film (Fig. 3A and fig. S21), indicating its accumulation at the buried interface of perovskite–PEDOT:PSS, with improved adhesion of the perovskite–PEDOT:PSS interface (fig. S22). We then studied the effect of buried interface passivation on perovskite’s photophysical properties. Ultrafast transient absorption spectroscopy exhibited similar dynamics of both sides of the AAH Pb-Sn perovskite film (fig. S23). The steady-state PL emission of the AAH Pb-Sn film measured from both sides were identically strong and higher than those of the control film (Fig. 3B). The PL decay lifetimes (t) probed from both sides of the AAH Pb-Sn perovskite film were similar (t = 595.0 ns from top and t = 551.4 ns from bottom) and much longer than those of the control film (t = 134.3 ns from top and t = 76.2 ns from bottom) (Fig. 3C). Steady-state PL spectra and time-resolved PL lifetimes of perovskite films measured off their bottom surface through the PEDOT:PSS–ITO side and when peeled off exhibit stronger intensity and longer decays for AAH-containing films (fig. S24). We thus conclude that AAH reduces charge carrier trapping by passivating the bottom surfaces of blade-coated Pb-Sn perovskites films. Electroluminescence quantum yield (ELQY) was further used to analyze nonradiative recombination and open-circuit voltage (Voc) loss of devices (34). At current densities equivalent to the short-circuit current density (Jsc) under a simulated 1-sun, the ELQYs of control and AAH devices were 0.15 and 1.68%, corresponding to Voc losses of 168 and 106 mV for the control and AAH devices, respectively (fig. S25). This finding is further supported by closer-to-unity ideality factor (fig. S26) and higher built-in potential of AAH PSCs compared with control devices (fig. S27). To further quantify interfacial nonradiative recombination losses, the PL quantum yield (PLQY) of neat perovskite, HTL-perovskite, and HTL-perovskite–electron transport layer (ETL) half stacks were measured to estimate the quasi-Fermic level splitting (QFLS) and voltage losses (35). The AAH sample exhibited a substantially reduced voltage loss of 23 mV, whereas the control sample showed a higher value of 42 mV upon introducing HTL, exceeding the corresponding voltage losses at perovskite–ETL interfaces (fig. S28). This indicates that AAH mainly

y

1 µm

Passivation of buried perovskite interfaces

g

AAH

Control

t =10 s

the Pb-Sn perovskite precursor and retarded the crystallization process (figs. S18 and S19).

p

A

intermolecular hydrogen bonding between AAH and DMF in the downshift of the –C=O stretching vibrational mode of DMF upon dissolution of AAH (Fig. 2C) (33). Density functional theory (DFT) calculations suggested that hydrogen bonds formed between the –C=O of DMF and –NH3+ of AAH. The simulated shift in FTIR was also consistent with the experimental results (fig. S17 and supplementary text 1). Gas chromatography (GC) analysis showed that AAH helped in maintaining DMF in gas-assisted wet perovskite and in extending the processing window (Fig. 2D and table S4). Thus, we conclude that AAH+, rather than Cl–, facilitated intermolecular interactions with the components in

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

A

Precursor solution

Control

AAH

AAH+ [Pb/SnI6]4FA+/MA+ DMF DMSO (Pb/Sn)-O bond

B

C

Control AAH Sn 3d5/2

Transmittance

1 -C=O

D

3000 2500

Control, t=30 s AAH, t=30 s Control, t=90 s AAH, t=90 s

2000 1500

p

Intensity (a.u.)

Pb 4f7/2

DMF DMF-AAH

Concentration (µmol/L)

Hydrogen bond

1000 500

1655 cm-1

1661 cm-1

0 140

138

488

136

Binding energy (eV)

486

1800

484

1700

1600

1500

1400

1300

DMF

Wavenumber (cm-1)

Binding energy (eV)

DMSO

g Fig. 2. Mechanism of extending the processing window. (A) Schematic diagram of intermolecular bonding in the ink. (B) Pb 4f and Sn 3d XPS in perovskite films. (C) FTIR spectra of pure DMF and AAH dissolved in DMF (DMF-AAH). (D) GC of the control and AAH gas-assisted blade-coated wet perovskite films with different delay times.

y

Perovskite

Control-top Control-bottom AAH-top AAH-bottom

PEDOT:PSS

104 10

B

Pb2+ Sn2+ In+ AAH+

3

102

C

100

0.1

50

100

150

200

250

300

950

350

Sputter time (s)

1000

1050

1100

1150

30

PCE (%)

FF (%)

Jsc (mA cm-2)

80

78

29

0.78

76

20 19 18 17

28

0.76

16 74

Control

AAH

Control

AAH

Control

AAH

3000

4000

E

35 30

,

21

31

0.80

2000

y

22

82

0.82

1000

Time (ns)

32 0.86 0.84

0

Wavelength (nm)

Current density (mA cm-2)

0

Voc (V)

Control-top = 134.3 ns Control-bottom = 76.2 ns AAH-top = 595.0 ns AAH-bottom = 551.4 ns

0.01

101

D

1

y g

Intensity (counts)

105

ITO substrate

Intensity (a.u.)

106

PL intensity

A

Control

AAH

25

Reverse Forward

20

VOC JSC FF (V) (mA cm-2) (%)

15 10 5

PCE (%)

Rev. Fwd.

0 0.0

0.2

0.4

0.6

Voltage (V)

0.8

Fig. 3. Optoelectronic performance of blade-coated Pb-Sn perovskite films and solar cells. (A) TOF-SIMS of AAH perovskite films deposited on ITO–PEDOT: PSS substrates. (B) Steady-state and (C) time-resolved PL spectra of control and AAH Pb-Sn perovskite films deposited on bare glass (excitation from both the perovskite and glass sides). (D) Comparison of photovoltaic performance between control and AAH narrow-bandgap PSCs with aperture area of 0.049 cm2. (E) J-V curves of the champion Pb-Sn PSC. Gao et al., Science 383, 855–859 (2024)

23 February 2024

3 of 5

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

state PCE of 21.1%), with a Voc of 0.856 V, a Jsc of 31.2 mA cm−2, and a fill factor (FF) of 80.0% (Fig. 3E and fig. S31). We attribute the performance improvement of Pb-Sn PSCs primarily to AAH+ rather than Cl– (fig. S32). Photovoltaic performance of Pb-Sn PSCs also improved while posttreating perovskite films with AAH (fig. S33).

minimizes the buried interfacial nonradiative recombination loss. Next, we evaluated the effect of AAH on photovoltaic performance of blade-coated PbSn PSCs. Comparative analysis with control devices showed improvement of all photovoltaic figures of merit for AAH-based devices (Fig. 3D and fig. S29), particularly for an optimized concentration of AAH (fig. S30). The AAH devices (aperture area of 0.049 cm2) delivered an average PCE of 20.3 ± 0.5% for more than 80 devices and a champion PCE of 21.4% (steady-

perovskite tandem solar cells using fully scalable processing techniques (supplementary materials, materials and methods). To enhance the performance of wide-bandgap (WBG) subcells, a self-assembled monolayer of [4-(3,6-dimethyl9H-carbazol-9-yl)butyl] phosphonic acid was introduced to reduce surface nonradiative recombination on NiO hole transporter layer (36). The blade-coated WBG PSC (aperture area of 1.05 cm2) delivered a champion PCE of 17.9%, with a Voc of 1.31 V, a Jsc of 17.2 mA cm−2, and a FF of 79.6% (fig. S34 and table S5).

All-perovskite tandem solar modules

We then used the Pb-Sn NBG (~1.25 eV) perovskite with AAH to fabricate monolithic all-

NBG perovskite WBG perovskite 1 µm

Aperture area: 1.05 cm2

C

15

100 80

Control AAH

10

VOC JSC (V) (mA cm-2)

5

FF (%)

PCE (%)

p

Current density (mA cm-2)

B

EQE (%)

A

60

20

Control AAH

g

0

0 0.0

0.5

1.0

1.5

400

2.0

75

80

70

70 Mechanical scribing Laser scribing 60

65 15.0 11.25 9.0

7.5

6.4

5.6

F

2.0 P1 P2 P3

1.5

1.0 220 µm

Aperture area: 20.25 cm2

0.5

Rev, PCE=24.9% Fwd, PCE=24.8% 0

3

6

0.9

0.6

0.3

Aperture area: 64 cm2 Rev, PCE=23.8% Fwd, PCE=23.3%

12

15

18

0

5

10

15

20

25

30

y g

Voltage (V)

Voltage (V)

1.0

FACs: T80=656 h

0.8 FAMA: T80=330 h

y

0.6

,

Normalized PCE

9

1000

1.2

0.0

0.0

5.0

Width of subcells (mm)

G

Current density (mA cm-2)

90

Current density (mA cm-2)

80

E

GFF (%)

FF (%)

100

800

y

85

600

Wavelength (nm)

Voltage (V)

D

Back 15.5 mA cm-2

Front 15.5 mA cm-2

40

0.4

FAMA FACs

0.2

MPP tracking under 100 mW cm -2 illumination Encapsulated module operating under ambient condition

0.0 0

100

200

300

400

500

600

700

Time (h)

Fig. 4. Photovoltaic performance of all-perovskite tandem solar cells and modules. (A) Cross-section SEM image of all-perovskite tandem solar cells. (B) Photovoltaic performance of 1.05-cm2 all-perovskite tandem solar cell with and without AAH. (C) EQE curves of the champion AAH tandem device (aperture area of 1.05 cm2). (D) The FF and GFF of tandem modules as a function of subcells’ width made by means of mechanical or laser scribing. (E) J-V curves of a champion laser-scribed tandem module (aperture area of 20.25 cm2). (Inset) Photo of the module (left) and an optical microscopy image Gao et al., Science 383, 855–859 (2024)

23 February 2024

of the 220-mm-wide dead area (right). (F) J-V curves of the champion tandem module, with aperture area of 64 cm2. (G) Continuous maximum power point (MPP) tracking of encapsulated FAMA or FACs tandem solar cell under simulated air mass coefficient (AM) 1.5 G illumination (100 mW cm−2, multicolor lightemitting diode simulator) in ambient air with a RH of 30 to 50%. The device temperature was kept at ~50°C during operation under illumination. FAMA and FACs represent compositions of FA0.7MA0.3Pb0.5Sn0.5I3 and FA0.8Cs0.2Pb0.5Sn0.5I3 for the NBG perovskite layers, respectively. 4 of 5

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

Conclusions

RE FERENCES AND NOTES

science.org/doi/10.1126/science.adj6088 Materials and Methods Supplementary Text Figs. S1 to S52 Tables S1 to S8 References (42–48) Submitted 8 July 2023; resubmitted 8 August 2023 Accepted 26 December 2023 10.1126/science.adj6088

5 of 5

,

SUPPLEMENTARY MATERIALS

y

23 February 2024

Funding: This work was financially supported by the National Key R&D Program of China (2022YFB4200304), National Science Fund for Distinguished Young Scholars (T2325016), National Natural Science Foundation of China (U21A2076, 61974063, and 62305150), Natural Science Foundation of Jiangsu Province (BE2022021, BE2022026, BK20202008, and BK20190315), Fundamental Research Funds for the Central Universities (0213/14380206, 0205/14380252, and 0213/14380236), Frontiers Science Center for Critical Earth Material Cycling Fund (DLTD2109), and Program for Innovative Talents and Entrepreneur in Jiangsu. All theoretical calculations were performed at the High Performance Computing Center of Nanjing University. M.I.S. is grateful to the Canadian Foundation for the Canada Research Chairs Program (CRC-2019-00297) for financial support. Author contributions: H.T. conceived the idea and directed the overall project. H.G., K.X., and R.L. fabricated all the devices and conducted the characterization. W.W. performed the GC measurement and DFT calculation. C.J. performed the NMR measurement. A.D.B. and H.N. performed the PL-mapping measurement. S.Z., C.D., H.S., C.L, J.W., W.K., H.L., X.Z., Z.L., J.X, and L.L. carried out device fabrication and materials characterization. M.I.S. developed the crystallization model. H.T. supervised the project and assisted in data analysis. H.G., K.X., R.L., S.D., M.I.S., and H.T. wrote the manuscript. All authors discussed the results and commented on the paper. Competing interests: H.T., H.G., and R.L. are inventors on a patent application related to this work filed by Nanjing University and Renshine Solar Co. H.T. is the founder, chief scientific officer, and chairman of Renshine Solar Co., a company that is commercializing perovskite PVs. The other authors declare that they have no conflict of interest. Data and materials availability: All data are available in the main text or the supplementary materials. License information: Copyright © 2024 the authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original US government works. https://www.science.org/about/science-licensesjournal-article-reuse

y g

1. J. Kromdijk et al., Science 354, 857–861 (2016). 2. T. Leijtens et al., Sustain. Energy Fuels 2, 2450–2459 (2018). 3. R. Lin et al., Nat. Energy 4, 864–873 (2019). 4. Q. Jiang et al., Science 378, 1295–1300 (2022). 5. R. Lin et al., Nature 603, 73–78 (2022). 6. R. He et al., Nature 618, 80–86 (2023). 7. R. Lin et al., Nature 620, 994–1000 (2023). 8. B. Abdollahi Nejand et al., Nat. Energy 7, 620–630 (2022). 9. X. Dai et al., Nat. Energy 7, 923–931 (2022). 10. K. Xiao et al., Science 376, 762–767 (2022). 11. Y. Zong, Z. Zhou, M. Chen, N. P. Padture, Y. Zhou, Adv. Energy Mater. 8, 1800997 (2018). 12. W. Ke et al., ACS Energy Lett. 3, 1470–1476 (2018). 13. W. Ke, C. C. Stoumpos, M. G. Kanatzidis, Adv. Mater. 31, e1803230 (2020). 14. K. Xiao et al., Nat. Energy 5, 870–880 (2020). 15. M. Fievez et al., Sol. Energy Mater. Sol. Cells 230, 111189 (2021). 16. P. Li et al., Nano Energy 46, 203–211 (2018). 17. J. W. Lee et al., Chem 3, 290–302 (2017). 18. F. Zhang et al., Energy Environ. Sci. 11, 3480–3490 (2018). 19. W. Chen et al., J. Mater. Chem. C Mater. Opt. Electron. Devices 11, 151–160 (2022). 20. J. Zhu et al., ACS Energy Lett. 6, 3425–3434 (2021).

AC KNOWLED GME NTS

y

We demonstrated 20.25 cm2 all-perovskite tandem solar modules to reach a certified efficiency of 24.5%. To achieve this performance, we used a short-chain buffer from Good’s list to homogenize perovskite crystallization and passivate buried interfaces. All-perovskite tandem solar cells can practically achieve 30% efficiency (38). Low Jsc remains the main bottleneck because of optical losses owing to light reflection, parasitic absorption, inefficient absorption in the Pb-Sn subcell, and a large dead area in modules (10, 39). These challenges can be addressed with light management, thickening Pb-Sn absorber layer with long carrier diffusion lengths, and decreasing the dead area in modules. The Voc loss attributed to the perovskite-C60 interface can be suppressed through exploiting the scalable posttreatment passivation method, such as chemical passivation deposited by bladecoating or just integrating with 3D/3D heterojunction as in our previous work, but its uniform deposition remains to be developed (supplementary text 4) (7, 40). In terms of stability, substituting back metal electrodes with inert ones, such as conductive transparent oxides, and tunnel recombination junction with a thermally stable one would further strengthen the durability of all perovskite tandem modules (41).

21. K. O. Brinkmann et al., ACS Appl. Mater. Interfaces 11, 40172–40179 (2019). 22. C. Wang et al., Nat. Energy 7, 744–753 (2022). 23. J. B. Whitaker et al., Sustain. Energy Fuels 2, 2442–2449 (2018). 24. M. Yang et al., Nat. Energy 2, 17038 (2017). 25. Y. Deng et al., Sci. Adv. 5, eaax7537 (2019). 26. S. Chen et al., Science 373, 902–907 (2021). 27. Z. Ni et al., Nat. Energy 7, 65–73 (2022). 28. Z. Zheng et al., Adv. Mater. 34, e2109879 (2022). 29. S. Chen et al., Sci. Adv. 7, eabb2412 (2021). 30. B. Chen et al., Adv. Mater. 33, e2103394 (2021). 31. N. E. Good et al., Biochemistry 5, 467–477 (1966). 32. A. Q. Alanazi et al., J. Am. Chem. Soc. 141, 17659–17669 (2019). 33. P. Zhao et al., Adv. Mater. 30, e1802763 (2018). 34. J. Liu et al., Nat. Energy 1, 16089 (2016). 35. S. Rühle, Sol. Energy 130, 139–147 (2016). 36. L. Li et al., Nat. Energy 7, 708–717 (2022). 37. M. A. Green et al., Prog. Photovolt. Res. Appl. 31, 3–16 (2023). 38. M. Jošt, L. Kegelmann, L. Korte, S. Albrecht, Adv. Energy Mater. 10, 1904102 (2020). 39. Z. Yang et al., Nat. Commun. 10, 4498 (2019). 40. J. Thiesbrummel et al., Adv. Energy Mater. 13, 2202674 (2023). 41. H. Gao et al., Sol. RRL 5, 2100814 (2021).

g

Gao et al., Science 383, 855–859 (2024)

under 85% relative humidity (RH) (fig. S47B), 80% of its initial PCE within 200 hours under damp-heat tests (85% RH and 85°C) (fig. S47C), and 86% of its initial PCE under 200 thermal cycling tests (–40°C to 85°C) (fig. S47D).

p

The blade-coated all-perovskite tandem solar cells, consisting of ~380-nm thick WBG perovskite and ~1000-nm-thick NBG perovskite layers (Fig. 4A), demonstrated a champion efficiency of 26.8% (aperture area of 1.05 cm2) with a Voc of 2.15 V, a Jsc of 15.5 mA cm−2, and a FF of 80.5% (Fig. 4, B and C, and fig. S35). We then proceeded to fabricate all-perovskite tandem solar modules on 6- by 6-cm substrates. The optimization of module fabrication is detailed in supplementary text 3 and figs. S36 to S38. We adjusted the width of subcells from 11.3 to 5.0 mm and observed increasing FF of tandem modules (fig. S39) resulting from improved conductivity of front transparent cells. For the best-PCE eight-subcell module with a cell width of 5.625 mm (table S6), we achieved a high GFF of 96.1% and a narrow dead width of 220 mm (Fig. 4, D and E, inset). We also found that AAH could reduce Sn4+ back to Sn2+ in NBG perovskite ink (figs. S40 and S41) as well as reduce the presence of Sn4+ in the perovskite film (fig. S42), ultimately enabling laser scribing in module fabrication under ambient conditions (fig. S43). We fabricated 40 all-perovskite tandem solar modules that showed an average PCE of 23.3 ± 0.7% (fig. S44). The champion tandem module exhibited a PCE of 24.9% under reverse scan, with a Voc of 17.2 V, a Jsc of 1.85 mA cm−2, and a FF of 78.2% (Fig. 4E). The device exhibited a minor hysteresis between the reverse and forward scans (24.9 versus 24.8%) and a steady-state PCE of 24.7% at bias voltage of 14.5 V (fig. S45 and table S7). Considering a GFF of 96.1%, the active-area efficiency of the tandem module reached 25.9%. The module received a certified PCE of 24.5% (stabilized) by an accredited independent photovoltaic calibration and measurement laboratory (Japan Electrical Safety and Environment Technology Laboratories) (fig. S46). This PCE, now included in the Solar Cell Efficiency Tables (version 61) (37), represents the highest efficiency among perovskite solar minimodules. The perovskite film uniformity with AAH also maintained a high efficiency of 23.8% even when the aperture area was increased to 64 cm2 (Fig. 4F). To assess the operational stability of the tandem solar cells, we kept encapsulated devices at a temperature of ~50°C in ambient air (figs. S47A and S48) and measured their performance under simulated 1-sun illumination (multicolor light-emitting diode solar simulator with a spectrum profile equivalent to 100 mW cm−2). The AAH tandem module maintained 80% (T80) of initial PCE (table S8) after 330 hours under these conditions. When we used MA-free Pb-Sn perovskite (FACs), the operational stability of the device further increased to T80 = 656 hours (Fig. 4G). The encapsulated AAH device retained 90% of its initial PCE within 2000 hours

RES EARCH

CLIMATE CHANGE

Chemistry-albedo feedbacks offset up to a third of forestation’s CO2 removal benefits James Weber1*, James A. King1, Nathan Luke Abraham2,3, Daniel P. Grosvenor4,5, Christopher J. Smith6,7, Youngsub Matthew Shin2, Peter Lawrence8, Stephanie Roe9, David J. Beerling1, Maria Val Martin1* Forestation is widely proposed for carbon dioxide (CO2) removal, but its impact on climate through changes to atmospheric composition and surface albedo remains relatively unexplored. We assessed these responses using two Earth system models by comparing a scenario with extensive global forest expansion in suitable regions to other plausible futures. We found that forestation increased aerosol scattering and the greenhouse gases methane and ozone following increased biogenic organic emissions. Additionally, forestation decreased surface albedo, which yielded a positive radiative forcing (i.e., warming). This offset up to a third of the negative forcing from the additional CO2 removal under a 4°C warming scenario. However, when forestation was pursued alongside other strategies that achieve the 2°C Paris Agreement target, the offsetting positive forcing was smaller, highlighting the urgency for simultaneous emission reductions.

y g

Simulations

*Corresponding author. Email: [email protected] (J.W.); [email protected] (M.V.M.)

Weber et al., Science 383, 860–864 (2024)

Land surface cover (forest cover change at 2095 relative to 2015)

4C_SSP3

Simulation conditionsb

SSP3 SSP3-7.0 (deforestation, (High warming up to 4°C, small –290 Mha) ................................................................................... air pollution decrease) 4C_MF Maxforest (extensive forestation, +750 Mha) ................................................................................... 2C_SSP1 SSP1 SSP1-2.6 (forestation, +300 Mha) (Low warming up ................................................................................... to 2°C, large 2C_MF Maxforest (extensive forestation, air pollution decrease) +750 Mha)

DBVOC DGlobal emissions tree cover (MF – SSP) (MF – SSP) at 2050 (2095) at 2050 (2095)c 15% (26%)

17 to 19% (32 to 38%)

6% (10%)

8% (11 to 13%)

................................................................................... a

b Simulations performed at years 2050 and 2095. Well-mixed greenhouse gases, anthropogenic and biomass-burning c emissions, and sea-surface temperatures. Range shows model variation.

23 February 2024

1 of 5

,

a

y

Leverhulme Centre for Climate Change Mitigation, School of Biosciences, University of Sheffield, Sheffield S10 2TN, UK. Centre for Atmospheric Science, Yusuf Hamied Department of Chemistry, University of Cambridge, Cambridge CB2 1EW, UK. 3National Centre for Atmospheric Science, Yusuf Hamied Department of Chemistry, University of Cambridge, Cambridge CB2 1EW, UK. 4Centre for Environmental Modelling and Computation (CEMAC), School of Earth and Environment, University of Leeds, Leeds LS2 9JT, UK. 5Met Office Hadley Centre, Exeter EX1 3PB, UK. 6Institute for Climate and Atmospheric Science, School of Earth and Environment, University of Leeds, Leeds LS2 9JT, UK. 7 International Institute for Applied Systems Analysis, A-2361 Laxenburg, Austria. 8NCAR Earth System Laboratory, Climate and Global Dynamics Division, National Center for Atmospheric Research, Boulder, CO 80307, USA. 9World Wildlife Fund, Washington, DC 20037, USA.

y

Table 1. Modeling experiments in UKESM1 and CESM2.

1

2

g

CH4; drive O3 production or loss depending on the chemical environment; and produce oxidation products, which can add to or form aerosols that interact with solar radiation. Changes to atmospheric composition have been shown to be important in the net climatic impact of instantaneous global deforestation (10), and that which occurred between years 1850 and 2000 from cropland expansion (11). However, the atmospheric composition’s response to proposed reforestation and afforestation programs under different 21st-century future climate pathways, as well as their effects on climate, have received less consideration. We present an assessment of climate feedbacks from a large-scale afforestation, reforestation, and forest enhancement (hereafter, all three are referred to as “forestation”) scenario. To

p

R

eforestation and afforestation are widely proposed nature-based strategies for atmospheric carbon dioxide (CO2) removal (CDR) and climate change mitigation (1). These strategies have the potential to provide additional benefits for biodiversity; multiple ecosystem services, including reduced soil erosion and climate resilience; and forestry products and local cooling through transpiration (2–4). The Bonn Challenge, the New York Declaration on Forests, and the UN Decade on Ecosystem Restoration set a target to restore 350 Mha of degraded and deforested lands by 2030 (5). However, wide-scale forest expansion drives biophysical feedbacks within the Earth system that may lead to warming. For example, darker forests decrease surface albedo, which can substantially offset the cooling effects of carbon sequestration in some regions of the world (6, 7). Forests also release substantial quantities (760 TgC yr−1) of biogenic volatile organic compounds (BVOCs) that affect the greenhouse gases ozone (O3) and methane (CH4), as well as organic aerosols, with complex impacts on climate (8, 9). Chemical reactions of BVOCs deplete the hydroxyl radical (OH), increasing

mitigate possible single-model bias (8), we performed the same experiments in two state-ofthe-art climate models, UKESM1 (12) and CESM2 (13), which feature interactive atmospheric chemistry, aerosols, and BVOC emission schemes. We used a land surface cover scenario, which we named Maxforest, that expands forests from 2015 land cover in biomes where trees are expected to thrive: through reforestation (of rangeland, secondary forest, and secondary nonforest in forest biomes), forest enhancement (of forests where tree cover density is less than its potential), and afforestation (of rangeland, secondary forest, and secondary nonforest in nonforest biomes where tree cover is >10%) (14). The Maxforest scenario represents a nearbiophysical maximum for forestation given constraints on the rate of forestation and excluding expansion on croplands, pasturelands, urban lands, and protected areas designated by the International Union for Conservation of Nature (supplementary materials, Maxforest Scenario). This scenario resulted in additional tree cover of 500 Mha by 2050 rising to 750 Mha in 2095 (relative to 2015) (Fig. 1A), with approximately 55% from afforestation, 25% from reforestation, and 20% from forest enhancement by 2095. Although large-scale forestation presents certain risks and trade-offs (1), we used this theoretical biophysical maximum forestation scenario for our assessment to best detect biophysical changes. We compared Maxforest to two well-established future scenarios: SSP3-7.0 (regional rivalry), which features resource-intensive consumption, diminished technology development, and very low climate change mitigation efforts leading to global warming up to 4°C above preindustrial temperatures; and SSP1-2.6 (sustainability), which is characterized by inclusive development, environmental management, and lower

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

g y y g y ,

Weber et al., Science 383, 860–864 (2024)

p

resource- and energy-intensive consumption with much stronger efforts to mitigate climate change, limiting warming to 500 mM) and completely failed to yield 14 at low concentration ( -nitrile selectivity 11

major H 2N

CN 16G

CN 10

OH

OH H N

HO

H N

HO

CN 28

O

N NH2

28

11 4.6 H2N

X OH

4.4

4.3

4.2

4.1

4.0

3.9

Chemical Shift (ppm)

13

OH CN

H N

HO

S

O

O

F Quantitative cysteamine ligation

N

1

17

20

H N

H N

17

pantetheine 1

4.4

4.3

4.2

4.1

4.0

3.9

3.8

3.7

Chemical Shift (ppm)

Fig. 1. Overview of prebiotic pantetheine synthesis by nitrile-activation. (A) Coenzyme A. (B) Selective, highyielding nitrile-mediated pathway to pantetheine 1. Me, methyl. iPr, isopropyl. 1H nuclear magnetic resonance (NMR) spectra showing selectivity of (C) the aqueous aldol reaction of aldehydes 2 (22 mM), 3 (17 mM), 5 (22 mM), and 6 (22 mM) in PBS (pH 7, 500 mM) after (i) 20 min at 20°C, followed by sodium cyanide (NaCN) (300 mM) at 20°C; and (ii) 1 day at 60°C, followed by NaCN (300 mM) at 20°C, yielding pantoic acid nitrile 23 (57%). (D) Pantoic acid and proteinogenic amino acid differentiation upon reaction of aldehydes 2 (20 mM), 3 (17 mM), 4 (20 mM), 5 (31 mM), and 6 (20 mM), with NH3 (500 mM) and NaCN (150 mM) after 1 day in PBS (pH 9.5, 500 mM) at 20°C, yielding proteinogenic aminonitriles 16G, 16A, 16V, and 16S, and pantoic acid amidine 26. *, methanol. (E) Pantoylation during the stoichiometric competition of b-alanine-nitrile 10 (6.2 mM) and glycine nitrile 16G (6.2 mM) with aldehyde 4 (3.1 mM) and HCN (4.7 mM) in PBS (pH 9, 31 mM) at 20°C after 3 days, yielding pantothenic acid nitrile 11 (44%). The noncanonical a-homolog 20 was not observed. (F) Activating agent–free pantetheine synthesis upon reaction of pantothenic acid nitrile 11 (500 mM) with cysteamine 13 (2 equiv.) in PBS (pH 7, 500 mM) at 20°C after 60 days, yielding pantetheine 1 (93%). Ppm parts per million. Fairchild et al., Science 383, 911–918 (2024)

23 February 2024

2 of 8

,

SH

O

O

y

OH HO

y g

H N

HO

4.5

SH

y

H 2N

3.3

g

E

7

3.4

p

26

23

reaction vessel (fig. S11). The reaction failed completely when dried lactone 8, b-alanine 12, and cysteamine 13 were not sealed in an airtight reaction vessel before heating (fig. S10) owing to the sublimation of lactone 8 (Fig. 2D). To compound the problems of dry-state heating, we discovered pantothenic acid 14 decomposes to b-alanine 12 and pantoic acid 9 under hot-dry conditions (fig. S7). The drystate synthesis of pantetheine 1 is thus not prebiotically plausible because it demands an artificially sealed reaction vessel (20), and therefore, alternative substrates are needed for the effective prebiotic synthesis of pantetheine 1. Aminonitriles are prebiotic precursors of amino acids (7, 35), but their hydrolysis to amino acids dissipates the energy stored within the nitrile moiety. However, taking advantage of the latent nitrile activation of a-aminonitriles 16, we recently reported a chemoselective synthesis of proteinogenic a-peptides in water (31, 37). These mechanisms bypassed a–amino acids to generate a-peptides without the electrophilic carboxylate-activation that would be necessary with amino acids. Electrophilic activation is not only incompatible with various proteinogenic amino acid side chains (31, 37), but it is also incompatible with pantetheine 1 synthesis. To demonstrate this, our attempts to synthesize pantetheine 1 from cysteamine 13 and carboxylic acids (i.e. pantoic acid 9, b-alanine 12, and pantothenic acid 14) were thwarted by a myriad of detrimental reactions (Fig. 2C) (supplementary materials, supplementary text 3). Chief amongst these problems were the incompatibility of cysteamine 13 with electrophilic carboxylate activation (fig. S24) (45) and the fragmentation of pantothenic acid 14 (fig. S20). Reflection upon our recent a-peptide coupling strategies (31, 37) led us to suspect that latent nitrile activation could be exploited to achieve pantetheine 1 synthesis and overcome these problems. Different chemistries are required for selective synthesis of pantetheine 1, which contains an a–hydroxy acid and a b–amino acid, rather than proteinogenic a–amino acids (31, 37). Specifically, we hypothesized that b-alanine-nitrile 10 (pKaH = 7.8, where KaH is the conjugate acid dissociation constant) would possess a key nucleophilic advantage over b-alanine 12 (pKaH = 10.5) to allow selective coupling of b-alanine-nitrile 10 with lactone 8 to generate pantothenic acid nitrile 11 in water. 11 would retain latent activation, within its nitrile moiety, which would allow an activating agent–free reaction with cysteamine 13 to furnish pantetheine 1. To test the first element of our hypotheses, we incubated lactone 8 with b-aminonitrile 10 in water. We observed pantothenic acid nitrile 11 in up to 94% yield, but near-quantitative coupling required high (>100 mM) lactone 8 concentration (table S5). Nevertheless, incubating lactone 8 with equimolar b-alanine-nitrile

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

A OH

9

OH

O

HO CO2H

pH < 4

OH H2N

O

9

8

OH

pH > 9

CO2H

8

CO2H 9

12

50 mM 4.3

4.4

4.2

4.1

4.0

3.9

3.8

3.7

Chemical Shift (ppm)

B

OH H 2N

CO2H

O

HO

minor

O

12

9

H N

HO

O

CO2H

15

15

14 14

8

H 2N

OH

major

CO2H

500 mM

H N

HO O

4.4

15

4.3

4.2

4.1

4.0

3.9

3.8

3.7

p

Gly

CO2H

Chemical Shift (ppm)

C H N

HO

O

SH

CO2H

O

O

HO

EDC

H2N

H2 N

13

H N

Me2N

CO2H

SH N

12

14

H N

g

OH

EDC 13

8

y

D

12 8 O

HO

O

H 2N

H2N

CO2H

CO2H

i.

12

12 8

y g

ii. 4.5

4.0

3.5

3.0

2.5

Chemical Shift (ppm)

Latent electrophilic nitrile activation

We next investigated the latent activation of pantothenic acid nitrile 11. We had previFairchild et al., Science 383, 911–918 (2024)

ously observed that b-alanyl-nitriles resisted reaction with thiol nucleophiles, which blocked their unwanted incorporation into peptides by thiol-catalyzed peptide ligation (31). However, the ambident nucleophilicity of cysteamine 13 and irreversible thiazoline formation were found to switch on b-alanine-nitrile reactivity. Incubating pantothenic acid nitrile 11 with

23 February 2024

cysteamine 13 thus led to the formation of thiazoline 17 in good-to-excellent yield across a broad pH range (table S7). Incubating pantothenic acid nitrile 11 with cysteamine 13 at neutral pH directly yielded pantetheine 1 (93%) (fig. S33). Therefore, in three high-yielding, activating agent–free steps, pantetheine 1 was produced through the nucleophilicity of b-alanine-nitrile 3 of 8

,

10 and b-alanine 12 returned 84% pantothenic acid nitrile 11 with only 3% acid 14 (Fig. 3A) in a clear demonstration of the superior reactivity of b-alanine-nitrile 10.

activation of 14 with EDC resulted in fragmentation of pantoyl-amide 14, whereas the rapid reaction of 13 with EDC blocked onward reaction and synthesis of pantetheine 1. See the supplementary materials (supplementary text 3) for further details. (D) Failed synthesis of pantothenic acid 14 by wet-dry cycling. 1 H NMR spectra show (i) a solution of pantolactone 8 (500 mM) and b-alanine 12 (500 mM, pH 7.0) and (ii) the products of this solution after a slow stream of air was passed over it for 2.5 days at 20°C, followed by heating of the residue at 100°C for 24 hours and then dissolution in water (1 mL), which revealed the sublimation of pantolactone 8 (>99%). See the supplementary materials (supplementary text 2) for further details.

y

Fig. 2. Failed carboxylic acid–mediated pathways to pantetheine. (A and B) 1H NMR spectra show the following results: (A) Incubating lactone 8 (50 mM) with b-alanine 12 (2 equiv.) in PBS (pH 9, 500 mM) at 20°C gave hydrolysis product pantoic acid 9 (94%) after 8 days. (B) Incubating lactone 8 (500 mM) with Gly (2 equiv.) and b-alanine 12 (2 equiv.) in PBS (pH 9, 500 mM) at 20°C gave hydrolysis product pantoic acid 9 (59%) and the noncanonical pantoyl-a-glycine 15 (29%) as the major products after 7 days, alongside only 11% of canonical pantothenic acid 14 as the minor product. (C) Failed coupling of cysteamine 13 and pantothenic acid 14 with model electrophilic activating agent 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC). Electrophilic

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

A

11

OH H2N

minor

CO2H

O

HO

H N

HO O

12

CO2H 14

O

9 H2N

8

major

CN

14

OH

H N

HO

10

O

CN 4.4

11

4.3

4.2

4.1

4.0

3.9

3.8

3.7

Chemical Shift (ppm)

B

11 OH H 2N

CN

minor

O

HO

H N O

18

9

19

O

19 H 2N

8

CN

HO

OH

major

CN 10

O

p

H N

HO

CN 4.3

4.4

11

4.2

4.1

4.0

3.9

3.8

3.7

Chemical Shift (ppm)

C

20 H 2N

HO

O

HO

O

16G

9

H N

minor

CN

CN

8

8

i.

20

y

O 8

20 11

g

OH

H 2N

OH

major

CN

H N

HO 10

O

CN 11

ii. 4.4

Pantetheine 1 possesses a b-alanyl motif, so we next questioned whether lactone 8 would discriminate b-alanine-nitrile 10 from its shorter and longer homologs, a-aminonitrile 16G and g-aminonitrile 18 (Fig. 3). g-Aminonitrile 18 (pKaH = 10.2) is substantially more basic than b-alanine-nitrile 10, and so 18 did not effectively couple with lactone 8. Indeed, the reaction of lactone 8 with equimolar 10 and 18 selectively produced pantothenic acid nitrile 11 (71%) alongside only 10% of g homolog 19 (Fig. 3B). Under the same conditions, we observed the reaction of lactone 8 with equimolar b-alanine-nitrile 10 and glycine-nitrile 16G, which produced nearly equal amounts of pantothenic acid nitrile 11 Fairchild et al., Science 383, 911–918 (2024)

3.7

PBS (pH 9, 500 mM) at 20°C yielded pantothenic acid nitrile 11 (71%) as the major product after 2 days, alongside 9 (19%) and 19 (10%). (C) Incubating lactone 8 (500 mM) with b-alanine-nitrile 10 (2 equiv.) and glycine nitrile 16G (2 equiv.) in PBS (pH 9, 500 mM) at 20°C yielded (i) 11:20 (~1:1) after 6 hours, and then (ii) canonical pantothenic acid nitrile 11 as the major product (11:20, >5:1) after 6 days.

and pantoyl-a-glycyl-nitrile 20 after 6 hours. However, upon further incubation, an unanticipated equilibration yielded nitrile 11 as the major product (11:20, >5:1) after 6 days (Fig. 3C). This dynamic reactivity was confirmed by incubating isolated nitrile 20 with b-alanine-nitrile 10, which yielded pantothenic acid nitrile 11 in up to 93% yield (table S10). These results demonstrate that the reactivity of b-alanine-nitrile 10 markedly favors the synthesis of the canonical structure of pantetheine 1 over both shorter and longer homologs in water. One-pot multicomponent synthesis of pantetheine from pantolactone in water

We next investigated the one-pot multicomponent synthesis of pantetheine 1 (Fig. 4A). Incubating lactone 8 (500 mM) with b-alanine-nitrile 10 (2 equiv.) and cysteamine 13 (2 equiv.) yielded

23 February 2024

3.8

thiazoline 17 (57%) after 3 days at pH 9 (Fig. 4B). Moreover, incubating lactone 8 with b-alaninenitrile 10, cysteamine 13, and glycine-nitrile 16G under the same conditions furnished thiazoline 17 (33%) as the major pantoyl-amide product (figs. S52 and S53). During this four-component reaction, the rapid reaction of glycine-nitrile 16G with cysteamine 13 suppressed the reaction of the a-aminonitrile with lactone 8, favoring the addition of b-alanine-nitrile 10 to lactone 8. We also observed that cysteamine 13 was released back into solution to drive thiazoline 17 synthesis. The thiazoline 17 synthesized in these multicomponent reactions was observed to hydrolyze in near-quantitative yield at neutral or acidic pH (pH 7 to 4) to yield pantetheine 1 (up to 57% yield from lactone 8). Our results demonstrate a nitrile-directed multicomponent synthesis of pantetheine 1 from lactone 8 in water, 4 of 8

,

Nitrile-controlled b-alanyl–selective pantoylation

4.1 4.0 3.9 Chemical Shift (ppm)

y

10 and the latent electrophilicity of pantothenic acid nitrile 11 in water.

4.2

y g

Fig. 3. Chemoselective pantoylation of aminonitriles. (A to C) 1H NMR spectra show the following results: (A) Incubating lactone 8 (500 mM) with b-alanine-nitrile 10 (2 equiv.) and b-alanine 12 (2 equiv.) in PBS (pH 9, 500 mM) at 20°C yielded pantothenic acid nitrile 11 (84%) as the major product after 2 days, alongside 9 (11%) and 14 (3%). (B) Incubating lactone 8 (500 mM) with b-alanine-nitrile 10 (2 equiv.) and g-aminobutyric-acid-nitrile 18 (2 equiv.) in

4.3

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

A

B H 2N

SH

O

HO

OH

H N

13 HO

O

N

H2N

via 11

CN

9

S

O

8

17

17

10

i. 1

C

9 OH

H N

H N

HO

SH

O

O

ii.

1 11

4.3

4.2

4.1

4.0

3.9

3.8

3.7

Chemical Shift (ppm)

p

D

E

O H

O

H

OH

2

O

OH

23

OH

HCN CN

g

10

PBS

3

i.

23

4

8

H N

HO

y

OH

9

11

O

HO

O

CN

ii.

O 11

8 4.7

4.5

4.4

4.3 4.2 4.1 4.0 Chemical Shift (ppm)

3.9

3.8

3.7

hydroxypivaldehyde 4 reacts in situ with HCN and b-alanine-nitrile 10 to yield pantoic acid nitrile 23, which undergoes intramolecular g-hydroxyl–catalyzed interrupted nitrile hydrolysis to yield pantolactone 8. This locks the final carbon atom of pantoic acid’s carbon framework into lactone 8, while retaining chemical activation toward amide-bond formation. (E) 1H NMR spectra of 4 (3.1 mM), 10 (6.3 mM), and NaCN (3.4 mM) in PBS (pH 7, 31 mM) after (i) 10 min and (ii) 11 days at 20°C, yielding lactone 8 (54% from 4) and pantothenic acid nitrile 11 (13% from 4). For experimental details, see the supplementary materials (pages S113–S115).

y g

Fig. 4. Multicomponent syntheses of pantetheine through lactone 8. (A) One-pot multicomponent reaction of lactone 8, b-alanine-nitrile 10, and cysteamine 13 yields pantetheine 1 in water. (B) 1H NMR spectra of lactone 8 (500 mM), 10 (2 equiv.), and 13 (2 equiv.) in PBS (pH 9, 500 mM) at 20°C. (i) After 4 days, this yielded 17 (57% from 8), and then (ii) after in situ (rapid) hydrolysis of 17 at pH 4, yielded pantetheine 1 (57% from 8). For experimental details, see the supplementary materials (pages S64–S66). (C) Molecular structure of pantothenic acid nitrile 11, as determined by single crystal x-ray diffraction. (D) Aldol product

4.6

y

Chemoselective aldol synthesis of hydroxypivaldehyde and pantolactone at neutral pH

We found that the conversion of formaldehyde 2 and isobutyraldehyde 3 to hydroxypivaldehyde 4 was highly effective at neutral pH and catalyzed by phosphate (46, 47). For example, incubating 2 (22 mM) and 3 (17 mM) in Fairchild et al., Science 383, 911–918 (2024)

phosphate buffer solution (PBS) at pH 7 gave hydroxypivaldehyde 4 (94%) after 2 days at 60°C (fig. S61). To test the selectivity of this aldol reaction, we incubated 2, 3, and another enolizable aldehyde, acetaldehyde 5, at pH 7. We again observed the formation of 4 (94%) after 2 days, but now alongside quantitative recovery of acetaldehyde 5 (fig. S64). Incubation of 2, 3, acetaldehyde 5, and glycolaldehyde 6 also yielded hydroxypivaldehyde 4 as the major aldol product (figs. S67 to S70 and table S25) with excellent recovery of acetaldehyde 5. Partial conversion of glycolaldehyde 6 to dihydroxyacetone, which is a C3 sugar precursor of nucleic acids, amino acids, and lipids (7, 12, 35), was also observed (table S25). Subsequent addition of HCN led to in situ quantitative

23 February 2024

conversion of aldehydes 2 to 6 to their respective cyanohydrins 21 to 25 (figs. S67 to S70); however, continued incubation under the same conditions provided a mild one-pot conversion of cyanohydrin 23 to lactone 8 (54% from hydroxypivaldehyde 4) (Fig. 4E and fig. S80). Encouraged by the facile synthesis of lactone 8, we next investigated the chemoselectivity required to integrate pantoate and proteinogenic a-aminonitrile syntheses (7, 35). Differentiation of pantoate from proteinogenic a-aminonitriles and in situ formation of pantoyl-amides

The selective concurrent synthesis of pantoate and proteinogenic a-aminonitriles 16 represents an intrinsic challenge because the pantoate 5 of 8

,

albeit at high concentration. However, we suspected that further investigation of the role of nitriles in pantoic acid synthesis would resolve the apparent need for high reagent concentrations. Therefore, we turned our attention to the origins of pantoic acid precursors from the aldehydes generated by prebiotic reduction of HCN (7, 12).

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

A O

OH

OH H 2N

H2N

R

OH

OH HO

CN

10 or 12

16G or Gly

4

non-canonical -homolog blocked by 5-exo-dig cyclization

23

HO

CO2H

minor

O

CN

H N

HO N

NH

HO

R=CO2H

H N

HO

NH

14

O

OH

OH minor

NH2

28

27

R=CN

CN

20

X

OH H N

H N

HCN

R

O

OH H N

HO

major

CO2H

OH

OH

major

7

H N

HO

O

NH3

H 2O CN

–NH3

NH

15

11

NH

OH

NH

NH3

26

B

+13

O

p

HO

10

HO

NH2

CN

O

29

intramolecular -OH catalyzed transamidation favors canonical pantethenic acid nitrile

H N

HO

pantetheine 1

7

C

D

23

23

26

g y

7

7

7 23

i.

i.

i. 9

9

11

y g

15

15 14

28

ii.

ii.

ii. 4.6

4.5

4.4

4.3

4.2

4.1 4.0

9

11

3.9

3.8

3.7

3.6

4.6

4.5

4.4

4.3

4.2

4.0

3.9

3.8

3.7

precursor, hydroxypivaldehyde 4, must not form an a-aminonitrile. The aminonitrile of hydroxypivaldehyde 4 would have an a-amine, not the canonical pantoate a-hydroxyl moiety. Conversely, at the same time and under the same conditions, amino acid precursors alde-

4.5

4.4

4.3

4.2

4.1

4.0

3.9

3.8

3.7

Chemical Shift (ppm)

Gly (6.3 mM), and NaCN (4.7 mM) in PBS (pH 9, 31 mM) at 20°C after (i) 10 min and (ii) 8 days yielded 9 (65%), 15 (16%), and 14 (9%). (C) 4 (3.1 mM), 10 (6.2 mM), 16G (6.2 mM), and HCN (4.7 mM) in PBS (pH 9, 31 mM) at 20°C after (i) 10 min and (ii) 3 days, yielding 11 (44%). See fig. S85 for spectra that demonstrate that b-alanine-nitrile 10 outcompetes b-alanine 12 (2:1) in a direct stoichiometric competition. (D) 4 (20 mM), NaCN (30 mM), and NH3 (500 mM) in PBS (pH 9.5, 500 mM) at 20°C; after (i) 4.5 hours, this yielded 26 (82%), (ii) then following the addition of 10 (40 mM), yielded 11 (46% from 4) after 6 days. See fig. S90 for one-pot synthesis of pantetheine 1 through in situ formation of 26.

hydes 2, 3, 5, and 6 must form a-aminonitriles 16 that possess an a-amine moiety necessary for proteinogenic a-peptide synthesis (31, 37). We found that incubating aldehydes 2 to 6 under Strecker conditions (7, 35) with cyanide and ammonia led to the complete differentiation

23 February 2024

4.6

of peptide and pantoate precursors (fig. S73). Chemoselective formation of proteinogenic a-aminonitrile 16G, 16V, 16A, and 16S was observed, but hydroxypivaldehyde 4 was crucially excluded from a-aminonitrile synthesis by rapid formation of a-hydroxy-amidine 26. 6 of 8

,

Fig. 5. Chemoselective multicomponent reaction cascades. (A) Nitriles (R=CN): Incubating hydroxypivaldehyde 4, b-alanine-nitrile 10, glycine-nitrile 16G, and HCN selectively yields canonical pantothenic acid nitrile 11. a-Homolog 20 synthesis is blocked by cyclization of amidine 27. Incubating aldehyde 4, HCN, and NH3 yields a,g-dihydroxyamidine 26, which undergoes selective transamidation with b-alanine-nitrile 10 to yield canonical pantothenic acid nitrile 11. Amino acids (R=CO2H): Incubation of aldehyde 4, b-alanine 12, glycine Gly, and HCN selectively yields nonnatural 15, not canonical pantothenic acid 14. (B to D) 1H NMR spectra show the following results: (B) 4 (3.1 mM), 12 (6.3 mM),

Fairchild et al., Science 383, 911–918 (2024)

4.1

Chemical Shift (ppm)

y

Chemical Shift (ppm)

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

7 of 8

,

We thank K. Karu (UCL, mass spectrometry) and A. E. Aliev (UCL, NMR spectroscopy). Funding: This work was supported by the

y

23 February 2024

AC KNOWLED GME NTS

y g

1. S. L. Miller, Science 117, 528–529 (1953). 2. C. de Duve, Proc. Natl. Acad. Sci. U.S.A. 84, 8253–8256 (1987). 3. C. de Duve, American Scientist 83, 428–437 (1995). https://www.jstor.org/stable/29775520.

y

The selective syntheses of pantetheine 1 in water challenges the persistent dogma that, despite it being the “solvent of life” (48), water is problematic (or even a “poison”) for prebiotic chemistry (49). We observed highly effective nitrile-activated amide bond formations in water, even below physiological CoA concentrations within modern cells (50). This chemistry not only favors the canonical structure of 1 but also closely aligns with previously reported prebiotic pathways to a-peptides, RNA, and lipids (7, 12, 31, 35, 37). Therefore, our results suggest that 1 would have been a product of cyanosulfidic reaction pathways prior to the emergence of life on Earth. Once available, it is simple to envisage how pantetheine 1 could have been deployed at the origins of life, for example, as a (nucleotidecoded) catalyst or a cofactor to enhance the functional limitations of early ribozymes (14, 26, 28) or peptide catalysts (31). This would mirror its essential role in augmenting the functional repertoire of enzymes in extant biochemistry (14, 17, 22) and provide a mechanism to couple 1 to the evolutionary development of life. RE FERENCES AND NOTES

The synthesis of pantothenic acid nitrile 11 through the in situ formation of iminolactone 7 suggested an inherent mechanism to block Fairchild et al., Science 383, 911–918 (2024)

Conclusions

4. C. Huber, G. Wächtershäuser, Science 276, 245–247 (1997). 5. W. Martin, M. J. Russell, Philos. Trans. R. Soc. Lond. B Biol. Sci. 362, 1887–1925 (2007). 6. K. Ruiz-Mirazo, C. Briones, A. de la Escosura, Chem. Rev. 114, 285–366 (2014). 7. B. H. Patel, C. Percivalle, D. J. Ritson, C. D. Duffy, J. D. Sutherland, Nat. Chem. 7, 301–307 (2015). 8. S. Islam, M. W. Powner, Chem 2, 470–501 (2017). 9. A. Whicher, E. Camprubi, S. Pinna, B. Herschy, N. Lane, Orig. Life Evol. Biosph. 48, 159–179 (2018). 10. K. B. Muchowska, S. J. Varma, J. Moran, Chem. Rev. 120, 7708–7744 (2020). 11. L. F. Wu, J. D. Sutherland, Emerg. Top. Life Sci. 3, 459–468 (2019). 12. D. J. Ritson, J. D. Sutherland, Nat. Chem. 15, 1470–1477 (2023). 13. F. Lipmann, Science 173, 875–884 (1971). 14. H. B. White III, J. Mol. Evol. 7, 101–104 (1976). 15. A. Eschenmoser, E. Loewenthal, Chem. Soc. Rev. 21, 1–16 (1992). 16. J. E. Goldford, H. Hartman, T. F. Smith, D. Segrè, Cell 168, 1126–1134 (2017). 17. A. Kirschning, Angew. Chem. Int. Ed. 60, 6242–6269 (2021). 18. A. D. Goldman, B. Kacar, J. Mol. Evol. 89, 127–133 (2021). 19. F. Pietrocola, L. Galluzzi, J. M. Bravo-San Pedro, F. Madeo, G. Kroemer, Cell Metab. 21, 805–821 (2015). 20. A. D. Keefe, G. L. Newton, S. L. Miller, Nature 373, 683–685 (1995). 21. S. N. Semenov et al., Nature 537, 656–660 (2016). 22. R. Leonardi, Y.-M. Zhang, C. O. Rock, S. Jackowski, Prog. Lipid Res. 44, 125–153 (2005). 23. G. Fuchs, Annu. Rev. Microbiol. 65, 631–658 (2011). 24. C. T. Walsh, B. P. Tu, Y. Tang, Chem. Rev. 118, 1460–1494 (2018). 25. M. A. Fischbach, C. T. Walsh, Chem. Rev. 106, 3468–3496 (2006). 26. R. R. Breaker, G. F. Joyce, J. Mol. Evol. 40, 551–558 (1995). 27. F. Huang, C. W. Bugg, M. Yarus, Biochemistry 39, 15548–15555 (2000). 28. V. R. Jadhav, M. Yarus, Biochimie 84, 877–888 (2002). 29. W. E. Kowtoniuk, Y. Shen, J. M. Heemstra, I. Agarwal, D. R. Liu, Proc. Natl. Acad. Sci. U.S.A. 106, 7768–7773 (2009). 30. J. G. Bird et al., Nature 535, 444–447 (2016). 31. C. S. Foden et al., Science 370, 865–869 (2020). 32. T. P. Begley, C. Kinsland, E. Strauss, Vitam. Horm. 61, 157–171 (2001). 33. K. B. Muchowska, J. Moran, Science 370, 767–768 (2020). 34. J. Franke, C. Hertweck, Cell Chem. Biol. 23, 1179–1192 (2016). 35. S. Islam, D.-K. Bučar, M. W. Powner, Nat. Chem. 9, 584–589 (2017). 36. S. L. Miller, G. Schlesinger, J. Mol. Evol. 36, 308–314 (1993). 37. P. Canavelli, S. Islam, M. W. Powner, Nature 571, 546–549 (2019). 38. A. P. Johnson et al., Science 322, 404 (2008). 39. E. T. Parker et al., Proc. Natl. Acad. Sci. U.S.A. 108, 5526–5531 (2011). 40. A. S. U. Choughuley, R. M. Lemmon, Nature 210, 628–629 (1966). 41. F. Raulin, G. Toupance, J. Mol. Evol. 9, 329–338 (1977). 42. S. L. Miller, G. Schlesinger, J. Mol. Evol. 36, 302–307 (1993). 43. E. T. Parker et al., Orig. Life Evol. Biosph. 41, 201–212 (2011). 44. J. Singh et al., J. Am. Chem. Soc. 144, 10151–10155 (2022). 45. Acetyl phosphate has been proposed to be a potential carboxylic acid activating agent (9). However, incubation of pantoic acid 9, b-alanine 12, and cysteamine 13 with acetyl phosphate in water led only to N-acetylation of 13 and hydrolysis of acetyl phosphate. See the supplementary materials for further details. 46. A. J. Coggins, M. W. Powner, Nat. Chem. 9, 310–317 (2017). 47. Á. F. Magalhães, M. W. Powner, Chem. Commun. 58, 13519–13522 (2022). 48. P. Ball, Chem. Rev. 108, 74–108 (2008). 49. M. Marshall, Nature 588, 210–213 (2020). 50. B. D. Bennett et al., Nat. Chem. Biol. 5, 593–599 (2009).

g

Blocking nonbiological pantoyl-a–amino acid analogs

the synthesis of (noncanonical) a-homologs by intramolecular (5-exo-dig) cyclization (i.e., 27 to 28) (Fig. 5). We carried out competition reactions with b-alanine-nitrile 10 and a-aminonitrile 16G in anticipation that pantothenic acid nitrile 11 would emerge as the only pantoyl-amide capable of onward reaction with cysteamine 13 and would thus selectively yield pantetheine 1. Reaction of aldehyde 4, b-alaninenitrile 10, glycine-nitrile 16G, and cyanide resulted in a highly chemoselective formation of pantothenic acid nitrile 11 (44% from aldehyde 4). Only a trace yield of aminoimidazole 28 (60,000 animals) with a comprehensive time-calibrated phylogeny that we anchored with genomic data (5400 loci) from 1018 species. Due to shifts in the dynamics of speciation and phenotypic evolution, snakes have transformed the trophic structure of animal communities through the recurrent origin and diversification of specialized predatory strategies. Squamate biodiversity reflects a legacy of singular events that occurred during the early history of snakes and reveals the impact of historical contingency on vertebrate biodiversity.

23 February 2024

1 of 6

,

Title et al., Science 383, 918–923 (2024)

y

*Corresponding author. Email: [email protected] †These authors contributed equally to this work. ‡Deceased

y g

Department of Ecology and Evolution, Stony Brook University, Stony Brook, NY 11794, USA. 2Environmental Resilience Institute, Indiana University, Bloomington, IN 47408, USA. 3Museum of Zoology and Department of Ecology and Evolutionary Biology, University of Michigan, Ann Arbor, MI 48109, USA. 4Department of Biology, California State University, Dominguez Hills, Carson, CA 90747, USA. 5Department of Biology and Environmental Sciences, Auburn University at Montgomery, Montgomery, AL 36117, USA. 6Department of Biological Sciences, The George Washington University, Washington, DC 20052, USA. 7Department of Vertebrate Zoology, National Museum of Natural History, Smithsonian Institution, Washington, DC, 20560, USA. 8Biology Department, University of Puerto Rico at Mayagüez, Mayagüez 00680, Puerto Rico. 9Department of Environmental Science, Policy, and Management, University of California, Berkeley, Berkeley, CA 94720, USA. 10Museum of Vertebrate Zoology, University of California, Berkeley, Berkeley, CA 94720, USA. 11Science Group: Fossil Reptiles, Amphibians and Birds Section, Natural History Museum, London SW7 5BD, UK. 12Research Department of Cell and Developmental Biology, University College London, London WC1E 6BT, UK. 13Biological Sciences, University of Adelaide, Adelaide, SA 5005, Australia. 14Departamento de Sistemática e Ecologia, Universidade Federal da Paraíba, João Pessoa, Paraíba 58051-900, Brazil. 15Departamento de Zoologia, Universidade de Brasília, Brasília, Distrito Federal 70910-900, Brazil. 16Institute of Biotechnology, Helsinki Institute of Life Science, University of Helsinki, 00014 Helsinki, Finland. 17South Australian Museum, North Terrace, Adelaide, SA 5000, Australia. 18Research School of Biology, The Australian National University, Canberra, ACT 2600, Australia. 19Department of Integrative Biology, The University of Texas at Austin, Austin, TX 78712, USA. 20Department of Ecology and Evolutionary Biology, University of Michigan, Ann Arbor, MI 48109, USA. 21Sam Noble Museum and Department of Biology, University of Oklahoma, Norman, OK, USA.

y

1

g

spect to ecology, morphology, and biogeography (8–10). These observations suggest that evolutionary dynamics in snakes are qualitatively different from those in lizards. If so, the origin of snakes is potentially consistent with a Simpsonian view of macroevolution (11, 12), whereby major biodiversity expansions occur through qualitative phase shifts into new adaptive zones. These phase shifts can be conceptualized as macroevolutionary “singularities”: patterns of rapid change across multiple organismic and ecological axes that, when viewed retrospectively through the prism of geological time, are sufficiently clustered together so as to seem virtually instantaneous. The term “singular” also refers to the fact that these transformations typically appear unpredictable from prior character states and phylogenetic position alone; such transitions have been documented within birds (13, 14), mammals (15), and other taxa (16). Here, we characterize the tempo and mode of ecological and morphological innovation across squamate reptiles to address the role of macroevolutionary singularities in generating large-scale patterns of lizard and snake biodiversity. In particular, we test the extent to which phenotypic shifts have predictable consequences for evolutionary diversification across squamates.

p

F

rom a phylogenetic perspective, snakes are indisputably nested within lizards (1, 2) and are simply one example of a particularly species-rich and cosmopolitan group of “scaled reptiles” (Squamata). Yet, unlike lizards, snakes engage with human emotions in a visceral manner unmatched by almost any other group of organisms and for this reason have played important cultural roles in human societies (3). Most readily known for their lack of limbs and unique prey-capture strategies, snakes exhibit an incredible degree of ecomorphological diversity and specialization. The ~4000 extant snake species include shovelsnouted burrowers that hunt desert scorpions, slender arboreal predators that prey on tree snails, and paddle-tailed marine forms that probe reef crevices for fish eggs and eels. However, more than 25 clades of lizards have independently evolved limblessness (4), and other lineages also evolved dietary specialization, venom, highly mobile skulls, and/or advanced chemoreception (5–7)—all attributes typically associated with snakes. This convergence raises fundamental questions about how and why particular traits have influenced squamate diversification more generally. Viewed across multiple traits, snakes nonetheless appear distinct from lizards with re-

We constructed a genomic backbone phylogeny for 1018 species of squamates, sequencing an average of 4.4 Mb across 4945 loci per taxon (figs. S1 and S2 and data S1). This phylogeny was then used as a scaffold upon which we added additional species from GenBank, yielding a species-level, dated phylogeny containing 6885 of the 10,759 squamate taxa (fig. S3 and data S2). We then time-calibrated a phylogenomic tree subsampled to 134 tips that spanned 31 fossil calibrations (figs. S4 and S5 and data S1) and used the resulting node dates as secondary calibrations for our full tree (fig. S6). With this tree, we quantified macroevolutionary dynamics across a range of ecological, morphological, and environmental traits. Traits were collated from a variety of sources and augmented with primary natural history data from our fieldand museum-based research programs (17), including a dietary dataset (n = 68,547 records) from preserved stomach contents and field observations that spanned 1314 species of snakes and lizards. For each species, we then computed a simple statistical index of net innovation, Y, defined as the absolute difference in phenotype between the focal species and the inferred ancestral state for all extant squamates (17). We calculated indices for net trophic innovation (Ydiet), morphological innovation in skull shape (Yskull), presacral vertebral count (Yvert), body elongation (Yelong), and chemosensory processing (Ychem). For the same traits, we also computed an index of absolute, branch-specific change along individual branches (fig. S7). To estimate evolutionary rates, we developed a tip-rate metric for univariate and multivariate phenotypic data that is similar to the widely used “diversification rate” (DR) statistic for speciation rate (18); we refer to this metric as the “TR” statistic (17). We then used a variancepartitioning analysis to identify nodes that best account for species-level phenotypic variation [canonical phylogenetic ordination (CPO) (19)]. We also characterized heterogeneity in phenotypic macroevolution with multiprocess evolutionary models that allowed for shifts in both (trait) state and evolutionary rates, in which “state shifts” are defined as large jumps in phenotype on individual branches (17).

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

Limb-reduced lizards

Snakes

Value B

C

D E

F

G

K

H J

L

diet

elong

skull

p

Net innovation ( )

Higher A Median Lower

g

TR elong y

Rates (TR)

TR diet

0 50

E

K

100

B

J

D G

150 200

L

F

A

H

C

rates of evolution for innovation in skull shape (Yskull), elongation (Yelong), and diet composition (Ydiet); these results are even more striking for colubriform snakes (node L). Tip values are residuals from tree-wide median value; additional traits are shown in fig. S8.

y

Fig. 1. Phenotypic innovation and evolutionary rates across squamates. Net innovation (Y) and evolutionary rate (TR) are shown across 246 squamate clades (bottom), subsampled from our full time-calibrated phylogeny of 6885 species (fig. S3). Snakes show greater phenotypic innovation and faster

A Dibamidae G Anguiformes H Iguania B Gekkota C Scincoidea I Snakes J Scolecophidia D Teioidea E Amphisbaenia K Alethinophidia F Lacertoidea L Colubriformes

y g

Million years ago

TR skull

,

Results Phenotypic innovation across snakes and lizards

Snakes have undergone large transformations along multiple phenotypic axes (Figs. 1 and 2 and figs. S7 to S10) relative to other squamates. These shifts distinguish them both quantitatively and qualitatively from the ancestral squamate phenotype and from their closest living (nonsnake) relatives. Estimates of net innovation range from 1.6- to 3-fold higher in snakes relative to lizards (Fig. 1 and fig. S8). These results suggest that the origin or early diversification of snakes was associated with massive shifts in traits associated with feeding, locomotion, and sensory processing. Title et al., Science 383, 918–923 (2024)

Rates for morphological and ecological traits generally increased early in the evolutionary history of snakes (Fig. 1). Snakes occupy distinct morphological space and exhibit elevated rates of morphological evolution relative to lizards, as evidenced by their greater body elongation (mean snake elongation index and rate >6× and >24× that of lizards; fig. S11), higher vertebral counts (mean count and rate for snakes >5× and >80× that of lizards; figs. S11 and S12), highly distinctive skull shape [mean snake rate >3× that of lizards; fig. S11 (9)], and greater rate of body mass evolution (snake rate >12× that of lizards; fig. S13). Patterns of morphological evolution are mirrored by ecological trait evo-

23 February 2024

lution: The rate of trophic niche evolution was 3.2-fold faster in snakes than in lizards, and snake diets have diverged much more from the ancestral squamate diet relative to lizards (Fig. 1). Across a range of traits and rates, both CPO and multiprocess models reveal that much of the phenotypic variation across snakes and lizards can be explained by a single event that occurred early in the evolutionary history of snakes (Fig. 2 and figs. S14 and S15). Multiprocess phenotypic models generally recovered a phase shift in both rate and state associated with the ancestor of extant snakes or an associated early divergence within snakes (Fig. 2 and figs. S16 and S17), implying both decoupled 2 of 6

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

Elongation

Variance explained (cumulative)

Diet

1.0 0.8

Model fitting Canonical phylogenetic ordination

Mass

Skull

Climate

Clade Snakes Alethinophidia

1.0

1.0

1.0

1.0

0.8

0.8

0.8

0.8

0.6

0.6

0.6

0.6

0.6

0.4

0.4

0.4

0.4

0.4

0.2

0.2

0.2

0.2

0.2

0.0

0.0

0.0

0.0

1

3 2 4 Rank order

5

1

3 2 4 Rank order

5

1

3 2 4 Rank order

Fig. 2. Phase shifts associated with the early evolution of snakes explain trait variation across extant squamates. (Top row) Multiprocess models generally find strong support for major phase shifts in both trait state and rates associated with all snakes (red) or alethinophidian snakes (orange; 88% of snake diversity). (Bottom row) Cumulative trait variance across extant squamates

3 2 4 Rank order

5

explained by discrete phase shifts on single branches, as inferred from two distinct unsupervised learning methods. The total variance explained by five largest shift nodes (top row) is shown as triangles; results from stepwise canonical phylogenetic ordination are shown as squares. Individual node effects account for much less variance in climate (far right) relative to other traits.

Previous analyses of trophic macroevolution across lizards (7) have formulated the “deephistory” hypothesis of squamate diets, in which resource use is deeply conserved, in contrast to dynamic patterns of trophic niche evolution observed in some vertebrate radiations (21–23). Our diet analyses recovered elements of this hypothesis (7, 24) (fig. S20), but this legacy effect is dwarfed by the trophic shifts and expansions that have occurred in snakes throughout the Cenozoic (25). The origin of alethinophidian snakes (all snakes but blindsnakes; 88% of snake species; Fig. 2) accounts for nearly 40% of the variance in species-level diet state across squamates (fig. S14G). This shift, accompanied by an overall acceleration in the rate of trophic niche evolution in colubriform snakes (Fig. 1, TRdiet), led to a dramatic expansion of the squamate trophic universe. Overall, our results highlight the power of primary natural history data obtained from museum specimens and field observations to reveal major shifts in the ecological structure of the biota. Climate and life-history evolution

We quantified speciation rates across squamates, using phylogenetic imputation to infer placements for the 3872 squamate species for which

3 of 6

,

Speciation rates in lizards and snakes

genetic data were not available [fig. S23, (17)]. Snakes show substantially elevated rates of speciation relative to lizards [mean lizard lCLaDS = 0.08, lBAMM = 0.09; snake lCLaDS = 0.18, lBAMM = 0.21 (CLaDS, cladogenetic diversification rate shift model; BAMM, Bayesian analysis of macroevolutionary mixtures)] (Fig. 4 and figs. S24 and S25D), including other well-studied lizard clades thought to be the result of rapid radiation (e.g., Anolis and sphenomorphine skinks) (Fig. 4, C and D). Approximately 45% of the variance in tip-level speciation rates across squamates is explained by the node spanning colubriform snakes (lCLaDS = 0.20; lBAMM = 0.23; figs. S14 and S26). Outside of snakes, the most important rate shift involved the ancestor of the species-rich Liolaemus lizard clade from montane and arid South America (~260 species), which accounts for ~11% of the variance in tip-level rates. For higher taxa, speciation rate is correlated with species richness (n = 119 subfamilies; Spearman’s r = 0.63) (Fig. 4C), but this positive correlation is driven by the presence of very small clades with low speciation rates. For clades with at least 50 species—collectively accounting for 96% of global squamate diversity—there is effectively no relationship between richness and rate (n = 42, r = 0.03). Similar decoupling between speciation rates and diversity patterns, both for taxonomic groups and geographic regions, has now been observed in a wide range of taxa (18, 26, 27). We then tested whether previously proposed key innovations predict speciation rates across squamates more generally. We found that individual traits had minimal ability to explain speciation-rate variation (Fig. 4F and fig. S25E), even for those traits that have evolved multiple times across squamates [maximum coefficient of determination (R2) = 0.01]. To the extent that these traits contributed to snake diversification, their influence is statistically confounded with a macroevolutionary singularity associated with

y

Climatic niche and life history traits do not appear to be evolving along distinct evolutionary trajectories in snakes relative to lizards, thus departing from patterns observed for morphology and diet (Fig. 2). We find high lability of a key life-history trait (parity mode), and in climatic niche parameters (figs. S8 and S9). Despite some phylogenetic signal in climatic niche parameters (figs. S14 and S15), there is little evidence that such conservatism persists across large phylogenetic scales (figs. S21 and S22). Variation in climatic niche across squamates thus cannot be explained by deep divergences in their evolutionary past (Fig. 2 and figs. S14L and S15R).

23 February 2024

1

y g

Title et al., Science 383, 918–923 (2024)

5

y

Snakes occupy much greater dietary space relative to lizards (Fig. 3A), and their diets are generally nonoverlapping owing to the disproportionate consumption of vertebrate prey by snakes (Fig. 3B and figs. S18 and S19). Snakes also use an aquatic and semiaquatic prey base that is largely untouched by lizards (separation along diet PC 2, Fig. 3, A and B, and fig. S18) (PC, principal component). Although snakes as a whole use a greater diversity of food resources than lizards, the diets of individual snake species are consistently narrower than those of lizards (Fig. 3D). Snakes have evolved a range of phenotypes to specialize on viscous prey (worms, slugs), slippery prey (eggs, eels, and caecilians), and noxious and dangerous prey (e.g., venomous snakes, mammals, centipedes, and scorpions), as well as snails and other “armored” resources (Fig. 3, A and B). A small number of lizards (e.g., some monitor lizards, Heloderma, and Lialis) feed heavily on vertebrates, mainly exploiting prey already used by snakes. By contrast, finer partitioning of dietary categories among arthropod-feeding squamates (Fig. 3C) reveals that snakes and lizards consume largely nonoverlapping resources, with blindsnakes feeding almost exclusively on insect brood—larval and pupal ants and termites in particular—and with several other snake lineages specializing on spiders, centipedes, and scorpions. Conversely, arthropodfeeding lizards consume far more adult insects spanning multiple taxonomic categories, and lizard niche breadths are broader than those of arthropod-feeding snakes (Fig. 3C).

3 2 4 Rank order

g

Trophic structure of squamate biodiversity

0.0 1

p

rate dynamics in snakes relative to other squamates and also a major shift (jump) in phenotypic state. These findings are consistent with those of several recent studies that demonstrated phase shifts in the tempo and mode of cranial evolution associated with the origin and early radiation of snakes (8, 20).

5

Other Relative support Higher Lower

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

A More terrestrial prey

Anolis punctatus 5

B Varanus gouldii

Thecadactylus rapicauda

Prey items Plants

Lizards

Crotalus cerastes

Insects

Snakes

Arachnids & centipedes Decapods Worms & mollusks Fishes

PC 2 (14.2%)

0

Storeria occipitomaculata

Amphibians Squamates

Thamnophis radix

Mammals & birds

Lizards Snakes 5

3 6 9 12

Enyalioides palpebralis

Snakes

Nerodia fasciata

p

More aquatic prey

Species' dietary breadth

Lizards 10

More invertebrate prey

5

0

5

More vertebrate prey

PC 1 (53.5%)

g

C Snake taxa within typical lizard dietary space (n = 72)

D All squamate taxa in the diet dataset (n = 1314) Adult insects

Arachnids Centipedes Insect larvae, pupae, eggs 40

20

Lizards 0

y

Amerotyphlops reticulatus

Density

20

Excess prey utilization relative to lizards (%) Sonora semiannulata ta

Data points

Snakes Lizards

Tantilla nigriceps

Snakes 0

5

10

15

Species' dietary breadth

Global squamate biodiversity has been heavily shaped by a singularity in macroevolutionary Title et al., Science 383, 918–923 (2024)

10

12

number of species assigned to the group. (C) Apparent dietary overlap between some snakes and lizards (A) masks differences in the ways that these groups use invertebrate prey. Arthropod-feeding snakes tend to eat spiders, centipedes, and scorpions (e.g., Sonora semiannulata and Tantilla nigriceps), or soft-bodied larval insects (e.g., Amerotyphlops reticulatus), and they also consume far fewer adult insects relative to lizards. (D) Dietary-niche breadths are much greater for lizards than for snakes, even after accounting for intraspecific sampling variability.

dynamics that occurred in association with the origin of snakes and with their most successful constituent clade (Colubriformes, 3100 species) (Fig. 1), a radiation that shows no sign of slowing toward the present (25). Numerous hypotheses have been proposed to account for the dramatic evolutionary radiation of snakes, including shifts in feeding mechanisms, extreme cranial kinesis, body elongation, limb reduction, and venom (8, 25, 29, 30). Furthermore, snakes show several genomic attributes, such as transposable element proliferation (31) and metabolic protein redesign (32), that might contribute to evolutionary versatility of the general snake body plan

23 February 2024

8

(20). However, our results highlight the challenge of linking specific innovations to the overall “success” of the snake phenotype. Replicated innovations in our dataset, including chemosensory innovation, limb loss, and body elongation, show no predictable effects on speciation rate (Fig. 4). Moreover, counterfactual experiments have occurred throughout squamate evolutionary history that support a decoupling between innovation per se and macroevolutionary regime shifts. For example, lizard (nonsnake) lineages that have evolved potent venom (Heloderma) have scarcely diversified, and perhaps the most extreme instance 4 of 6

,

Conclusions

6

y

the origin of snakes (Fig. 4E and fig. S27) (28). We also considered several biogeographic predictors of speciation rate, finding no evidence for associations with latitude (fig. S28) or biogeographic region (fig. S29). Similar to other vertebrate clades (18, 26), the latitudinal gradient in squamate diversity appears largely decoupled from speciation rates, and geographic variability in speciation rate is largely a function of the relative fraction of snakes versus lizards in the biota.

4

y g

Fig. 3. Fundamental ecological differences between snakes and lizards. (A) Multidimensional trophic niche as inferred from 68,547 individual squamates (stomach contents, field observations). Individual points correspond to species (n = 1314); point size denotes dietary-niche breadth. (B) Similarity path through diet space as inferred by principal graph analysis, illustrating taxonomic structure of resource differences between snakes and lizards. Nodes (circles) are groupings of species that consume similar resources; colors represent fractional resource use of each inferred node, and node sizes are proportional to the

2

Species' dietary breadth

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

A

Lizard clades including limb-reduced taxa

Snakes

Colubriform snakes

B Dipsadines Colubrines

Cobras, sea snakes & allies New World pit vipers Liolaemines

0.2

Dipsadines Colubrines

0.1

Anoles

"True" geckos

Sphenomorphines

0

250

500

750 1000 1250

Species richness

Liolaemus melanops 0.6

0.2

0.0

rds kes rds es les es rms kes is) liza na liza phin anolaeminbrifo sna droph all all s ed mor lio −coluriform (Hy duc no s −re sphe noncolub nake b s limalian r sea t s Au

Typical Limb-reduced Snakes lizards lizards

Lower

0.30 0.10

0.03

300

1000

Elongation index (log)

0.04 0.03 0.02 0.01

3000

0.00

sis ode ory ateode ape unt phy ape iet unt ine m ns lim m /sh co ra sh dl co a ll k rity ose c ing ize imb eogkull u br sk pahem oragdy s git/l g s rte f o di c ve b

rates across all snakes and lizards and for several key groups. (E) Speciation rates for individual species as a function of body elongation. Snakes are elongate and have fast rates of speciation (D), but elongate and/or limb-reduced lizards show no trends toward faster rates. Phylogenetic distribution of limb reduction is indicated by blue branches in the phylogeny shown in (A). (F) Phylogenetic regression analysis reveals minimal effect of any phenotypic traits on speciation rates (R2 < 0.01).

of the biota through time may be partly attributable to phenotypic changes that confer increased survivorship at the species or clade level (40, 41). Determining why snakes and other groups—from bats to beetles, and from angiosperms to acanthomorph fishes—have experienced such dramatic shifts in evolutionary tempo and mode is, as Darwin put it, an “abominable mystery.” Yet the major features of biological diversity may owe more to these revolutionary events than to the causal structures we infer by extrapolating contemporary microevolution through geological time.

y g y

REFERENCES AND NOTES

,

(e.g., sexual selection and trophic flexibility) might facilitate comparison of putative causal processes or agents across diverse branches of the Tree of Life. Nonetheless, the action of these and other innovations may be highly contingent on geographic and other circumstances peculiar to individual clades (35, 36). The view of phenotypic evolution emerging from squamates and other taxa (15, 20, 37) is that many of the most spectacular changes in the history of life have involved macroevolutionary singularities: rare, transformative events that are unpredictable from evolutionary time, phylogenetic position, and other factors. In this study, the recurrent evolution of snake-like traits in other groups of nonsnake lizards has failed to generate comparable evolutionary outcomes, suggesting that replicated patterns of adaptive radiation observed in many groups of organisms (35, 38) may be superseded at deeper phylogenetic scales by systemic shifts that ultimately give rise to higher taxa (11). The extent to which such shifts are qualitatively distinct from those associated with young radiations is likely to remain a challenging problem. Perhaps most importantly, the pruning effect of extinction, through the selective elimination of ecologically and phenotypically intermediate clades, may exacerbate the distinctiveness of clades that comprise the “survivors” that we observe in the present day (39). Thus, transformation

23 February 2024

Median 0.05

y

Title et al., Science 383, 918–923 (2024)

F 1.00

Speciation rate Higher

g

Fig. 4. Speciation rates in snakes are consistently elevated relative to lizards. (A) Subsampled phylogeny of 246 clades (as shown in Fig. 1) illustrating (B) speciation rates (CLaDS) relative to tree-wide median rate. Rates for each tip represent the mean rate for all species assigned to each terminal clade. (C) Species richness and tip-averaged speciation rate for major clades within squamates (subfamily level). (D) Median and interquartile range for tip speciation

Cobras, sea snakes & allies

NW Noncolubriforms pit vipers

E Hydrophis Hydro major m

0.4

of cranial kinesis in a nonsnake squamate occurs in a species-poor group of legless geckos (33). Conversely, one lineage of snakes—the splitjaw snakes (Bolyeridae)—has further increased the flexibility of the alethinophidian trophic apparatus through dramatic skeletal innovations not seen in any other vertebrates (34). Yet this innovation was unaccompanied by any substantive ecological or species diversification, and the lineage is currently represented by a tiny clade of endangered or recently extinct island endemics (Casarea dussumieri and Bolyeria multocarinata). Snakes thus appear to represent an ongoing adaptive radiation characterized by sustained ecological and species diversification that was triggered ultimately by one or more unknown and perhaps unknowable singular events. The principal challenge in resolving the causal basis of such singularities involves a lack of phylogenetic replication. By focusing on traits specific to squamates (e.g., cranial kinesis and venom), evolutionary biologists are necessarily limited to a small number of statistically independent data points or contrasts (28). However, some hypotheses, particularly those associated with genomic change or molecular evolutionary rates (20), could be tested across vastly greater phylogenetic scales (e.g., all vertebrates or metazoans). Likewise, the development of taxonindependent ontologies to describe key processes

Liolaemines

Phylogenetic R 2

0.3

D

Speciation rate (log CLaDS)

Lizards Non-colubriforms Colubriform snakes

Anoles

p

Speciation rate (CLaDS)

C

Australian sphenomorphines

Speciation rate (CLaDS)

"True" geckos

1. C. L. Camp, Bull. Am. Mus. Nat. Hist. 48, 289–480 (1923). 2. T. Townsend, A. Larson, E. Louis, J. R. Macey, Syst. Biol. 53, 735–757 (2004). 3. H. W. Greene, Snakes: The Evolution of Mystery in Nature (Univ. of California Press, 1997). 4. J. J. Wiens, M. C. Brandley, T. W. Reeder, Evolution 60, 123–141 (2006). 5. K. Schwenk, Brain Behav. Evol. 41, 124–137 (1993). 6. K. Metzger, in Topics in Functional and Ecological Vertebrate Morphology, P. Aerts, K. D’Aout, A. Herrel, R. Van Damme, Eds. (Shaker Publishing, Düren and Maastricht, 2002), pp. 15–46. 7. L. J. Vitt, E. R. Pianka, Proc. Natl. Acad. Sci. U.S.A. 102, 7877–7881 (2005). 8. A. Watanabe et al., Proc. Natl. Acad. Sci. U.S.A. 116, 14688–14697 (2019). 9. F. O. Da Silva et al., Nat. Commun. 9, 376 (2018). 10. U. Roll et al., Nat. Ecol. Evol. 1, 1677–1682 (2017). 11. G. G. Simpson, Tempo and Mode in Evolution (Columbia Univ. Press, 1944).

5 of 6

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

37. C. Venditti, A. Meade, M. Pagel, Nature 479, 393–396 (2011). 38. D. L. Mahler, T. Ingram, L. J. Revell, J. B. Losos, Science 341, 292–295 (2013). 39. D. N. Reznick, R. E. Ricklefs, Nature 457, 837–842 (2009). 40. D. Jablonski, Science 231, 129–133 (1986). 41. G. C. Williams, Natural Selection: Domains, Levels, and Challenges (Oxford Univ. Press, 1992). 42. P. O. Title et al., Data from: The macroevolutionary singularity of snakes, Dryad (2023); https://doi.org/10.5061/ dryad.p5hqbzkvb. 43. P. O. Title et al., Data from: The macroevolutionary singularity of snakes, Version v1, Zenodo (2023); https://doi.org/ 10.5281/zenodo.8250112. ACKN OWLED GMEN TS

SUPPLEMENTARY MATERIALS

science.org/doi/10.1126/science.adh2449 Materials and Methods Figs. S1 to S44 Tables S1 to S5 References (44–249) MDAR Reproducibility Checklist Data S1 to S3

g

We thank G. Schneider and F. Kraus for sample access, P. Skipwith and S. Potter for logistical support, and A. Devault, J. Enk, M. R. Marchan-Rivadeneira and the team at UMich ARC for technical support. We also thank A. Davis Rabosky, members of the Rabosky Lab, and two anonymous reviewers for comments and suggestions that substantially improved this manuscript. Approximately 80% of the dietary records in our dataset came from preserved voucher specimens, and we thank the curatorial staff of the many natural history collections worldwide that made this research possible. For photographs, we thank T. Schramer, R. Recoder, M. T. Rodrigues, J. Grummer, J. Farquhar, J. Crowe-Riddell, and A. Davis Rabosky. Pictures by R. Brown and B. Trapp were modified to compose manuscript figures. Funding: This work was supported by NSF Division of Environmental Biology grant 1754398 (D.L.R. and S.S.); the Packard Fellowship for Science and Engineering (D.L.R.); Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (G.R.C.); the National Council for Scientific and Technological Development (G.R.C.); the FAPDF (G.R.C.); California State University, Dominguez Hills RSCA (S.S.); NSF Division of Environmental Biology grant 1754425 (G.C.C.); NSF Division of Biological Infrastructure grant 0905765 (R.A.P.) and 1930030 (S.A.S.); NSF Division of Environmental Biology grant 1441719 (R.A.P.); the J. William Fulbright Fellowship (T.J.C.); NSF Division of Environmental Biology grant 1501711 (T.J.C.); Australian Research

Council (C.M.); Australian Research Council grant DE130101567 (M.E.H.J.); National Council for Scientific and Technological Development grant 304715/2021-2 (D.O.M.); Fundação de Apoio à Pesquisa do Estado da Paraíba-PB (D.O.M.); Coordenação de Aperfeiçoamento de Pessoal de Nível Superior-PRINT UFPB grant AUXPE 88881.312082/2018-01 (D.O.M.); NSF Division of Environmental Biology grants 0415430, 9200779, and 9505518 (L.J.V.), and a postdoctoral fellowship with the Environmental Resilience Institute at Indiana University (P.O.T.). Author contributions: Project conceptualization and planning: D.L.R., P.O.T., S.S., M.C.G., and G.C.C. Data collection: S.S., T.JC., M.R.G., G.C.C., L.B.Q.C., G.R.C., N.S., M.E.H.J., P.O.T., D.L.R., M.C.G., and L.J.V. Funding acquisition: D.L.R., R.A.P., G.C.C., G.R.C., and S.S. Resources: D.L.R., R.A.P., G.C.C., L.B.Q.C., G.R.C., N.D.P., S.C.D., C.M., D.O.M., E.R.P., S.A.S., and L.J.V. Method development: M.C.G., D.L.R., and P.O.T. Data analysis: P.O.T., S.S., M.C.G., D.L.R., and G.C.C. Data visualization: P.O.T., S.S., M.C.G., I.P., and D.L.R. Writing of primary draft: D.L.R., P.O.T., and S.S. Editing, commenting on, and approving final draft: All authors but E.R.P. (deceased). Competing interests: The authors declare that they have no competing interests. Data and materials availability: Accession numbers (NCBI) for genomic data are available in supplemental data S1. Data and scripts used in data processing, analysis, and visualization are available on Dryad (42), Zenodo (43), and GitHub (https://github.com/macroevolution/squamata). License information: Copyright © 2024 the authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original US government works. https://www. science.org/about/science-licenses-journal-article-reuse

p

12. M. J. Landis, J. G. Schraiber, Proc. Natl. Acad. Sci. U.S.A. 114, 13224–13229 (2017). 13. C. R. Cooney et al., Nature 542, 344–347 (2017). 14. G. Navalón, A. Bjarnason, E. Griffiths, R. B. J. Benson, Nature 611, 306–311 (2022). 15. A. Goswami et al., Science 378, 377–383 (2022). 16. J. W. Valentine, D. Jablonski, D. H. Erwin, Development 126, 851–859 (1999). 17. Materials and methods are available as supplementary materials. 18. W. Jetz, G. H. Thomas, J. B. Joy, K. Hartmann, A. O. Mooers, Nature 491, 444–448 (2012). 19. N. P. Giannini, Syst. Biol. 52, 684–695 (2003). 20. T. R. Simões, O. Vernygora, M. W. Caldwell, S. E. Pierce, Nat. Commun. 11, 3322 (2020). 21. O. Seehausen, C. E. Wagner, Annu. Rev. Ecol. Evol. Syst. 45, 621–651 (2014). 22. D. Schluter, The Ecology of Adaptive Radiation (Oxford Univ. Press, 2000). 23. T. Price, Speciation in Birds (W. H. Freeman, 2008). 24. T. J. Colston, G. C. Costa, L. J. Vitt, Biol. J. Linn. Soc. Lond. 101, 476–486 (2010). 25. M. C. Grundler, D. L. Rabosky, PLOS Biol. 19, e3001414 (2021). 26. D. L. Rabosky et al., Nature 559, 392–395 (2018). 27. L. F. Henao Diaz, L. J. Harmon, M. T. C. Sugawara, E. T. Miller, M. W. Pennell, Proc. Natl. Acad. Sci. U.S.A. 116, 7403–7408 (2019). 28. J. C. Uyeda, R. Zenil-Ferguson, M. W. Pennell, Syst. Biol. 67, 1091–1109 (2018). 29. B. G. Fry et al., Mol. Cell. Proteomics 7, 215–246 (2008). 30. H. W. Greene, Integr. Comp. Biol. 23, 431–441 (2015). 31. G. I. M. Pasquesi et al., Nat. Commun. 9, 2774 (2018). 32. T. A. Castoe et al., Proc. Natl. Acad. Sci. U.S.A. 110, 20645–20650 (2013). 33. F. C. Patchell, R. Shine, J. Herpetol. 20, 59–64 (1986). 34. T. H. Frazzetta, Am. Nat. 104, 55–72 (1970). 35. C. E. Wagner, L. J. Harmon, O. Seehausen, Nature 487, 366–369 (2012). 36. A. de Queiroz, Syst. Biol. 51, 917–929 (2002).

Submitted 22 February 2023; accepted 2 January 2024 10.1126/science.adh2449

y y g y ,

Title et al., Science 383, 918–923 (2024)

23 February 2024

6 of 6

Post date 23 February 2024

Erratum Erratum for the Report “Humans can discriminate more than 1 trillion olfactory stimuli” by C. Bushdid et al. When a correction to the Report “Humans can discriminate more than 1 trillion olfactory stimuli” was posted on 18 August 2016, a formal Erratum was not issued. However, at the request of the authors, Science is now indexing this as a formal Erratum, which replaces the previous correction. As noted in the correction on 18 August 2016, the mathematical model used to extrapolate the number of olfactory stimuli that humans can discriminate was based on a number of assumptions. The authors inadvertently failed to state four of them: First, the model assumed that odor stimulus space is isotropic, so that changing any component in any mixture in any direction is the same as any other direction.

p

Second, the authors assumed that olfactory perceptual space is high-dimensional. The authors agree that their extrapolation fails in regimes of low dimensionality (1), and subsequent analysis indicates that the results hold if the dimensionality of olfactory representations is D ≥ 25 (2).

g

Third, the model used a simplified approximation to obtain the total number of spheres that can be packed in a space by dividing the overall volume of the space by the volume of a single sphere. The actual calculation of the number of packable spheres using the “spherical code” problem establishes a rigorous lower bound that is a factor of 10 smaller than our estimate. Fourth, the authors considered any pair of statistically significantly discriminable stimuli as discriminable regardless of effect size. Their estimates hold if a conventional effect size of 20% is imposed.

y

It has been pointed out that a model not constrained by these assumptions causes the number of discriminable stimuli to become arbitrarily small or large (1, 3). Any exponential function will be sensitive in this way, and the goal of the model was not to identify the exact number of discriminable olfactory stimuli but an estimate of the order of magnitude of human discriminatory power across a population of human subjects. The authors thank their colleagues for bringing these unstated assumptions to their attention.

y g y ,

RE FE RENCES AND N OT ES

1. M. Meister, eLife 4, e07865 (2015). 2. M. O. Magnasco, A. Keller, L. B. Vosshall, bioRxiv 10.1101/022103 (2015). 2. R. C. Gerkin, J. B. Castro, eLife 4, e08127 (2015). 10.1126/science.ado6457

SCIENCE science.org

ERRATUM POST DATE • 23 FEBRUARY 2024

WORKING LIFE By Charlotte Goeyers

Embracing my difference

M

science.org SCIENCE

,

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

y

926

Charlotte Goeyers is a Ph.D. student at Ghent University.

y g

“Being neurodivergent is not easy, but I am discovering that my traits are valuable.”

y

Will my colleagues take me seriously when I need extra time to verbalize my thoughts or process their words? Will they accept that I need daily medication to concentrate? Will shying away from work parties or group lunches limit my opportunities? Colleagues I shared my diagnoses with were very understanding. My boyfriend, friends, and family had no doubt in my abilities and urged me to continue—with a few modifications to help me thrive. I bought noise-canceling earplugs, created sensory-friendly playlists, and invested in an at-home work setup. I embraced artificial intelligence tools to break down big tasks into smaller steps and to help me verbalize my thoughts. I learned to diligently outline discussion topics for meetings. I began to say “no” to social events when I didn’t have the energy—and tried not to feel guilty about this necessary trade-off to maintain my health. I accepted that my energy can spike and plunge, from peaks of hyperfocus to valleys of fatigue, and that I can time my tasks to my energy levels. Being neurodivergent is not easy, but I am discovering that my traits are valuable. My brain excels in pattern recognition, which is useful when I am identifying species through a microscope or writing code. My ability to hyperfocus allows me to quickly become a self-taught expert. I still struggle with spontaneous social events, but if I have time to prepare myself, I am able to attend and even enjoy them. In embracing my neurodivergent brain, I discovered that what once seemed like insurmountable hurdles were, in fact, stepping stones to becoming more authentic and more myself. j

g

Looking back, maybe it shouldn’t have been a surprise. As a kid, I had trouble connecting with peers and preferred to play alone, reading books, making art, or teaching myself to play piano and guitar. At university, I struggled to maintain friendships and skipped almost all my lectures, as I was never able to follow them. As a graduate student, I couldn’t shake the feeling that I didn’t belong, that I had tricked everyone into believing I’m more intelligent than I am and that if I opened my mouth everyone would discover I’m a fraud who scammed my way into a Ph.D. The further I got in my Ph.D., the more my anxiety took over. The chronic stress led to irritable bowel syndrome, and I could barely muster the courage to leave the house. After one doctor told me it was “all in my head” and I just needed to manage my diet and stress, I sought mental health professionals, leading to my diagnoses. Amid the shock, I was excited to finally have clarity. As I immersed myself in the relevant literature and familiarized myself with my neurodivergent brain, my life started to make sense. Socializing can be difficult because my autistic brain often misses or doesn’t understand unwritten rules. Busy conferences and crowded events are draining because my sensitive brain is easily overwhelmed by bright lights, loud noises, and strong smells. My ADHD makes it challenging to process auditory information, which explains my difficulties following lectures and conversations. I had subconsciously learned to hide these traits, internalizing the difficulties I faced and blaming myself. I had become exhausted from masking. Yet alongside the relief I felt at knowing what was “wrong” with me, I worried about my professional future in academia.

p

y heart was pounding. My vision blurred. While my fellow conference attendees discussed the dinner menu, I was trying to avert a panic attack. “Why am I like this? For once in my life, I must get myself together.” This wasn’t the first time something like this had happened. When I had to socialize at meetings, I got overwhelmed and my brain stopped functioning. During group discussions, panic took over. For years I didn’t know why. But when my body finally rebelled during my Ph.D., I discovered the reason: I am neurodivergent, and have both attention-deficit/hyperactivity disorder (ADHD) and autism. Recognizing this facet of my identity brings new worries, and the risk of stigma. But knowing who I am has allowed me to thrive.

AAAS NEWS & NOTES

p

Ana Maria Porras is recognized for her social media engagement through crafting.

AAAS Honors 2024 Award Recipients g

Awards recognize contributions to science diplomacy, mentoring, public engagement, and more

Climate scientist and science communicator Katharine Hayhoe is the recipient of the 2024 AAAS Mani L. Bhaumik Award for Public Engagement with Science. The award, which recognizes scientists and engineers who demonstrate excellence in their contribution SCIENCE science.org

AAAS Early Career Award for Public Engagement with Science Ana Maria Porras is the recipient of the 2024 AAAS Early Career Award for Public Engagement with Science, which recognizes earlycareer scientists and engineers—those who are within 7 years of completing their terminal degree—who demonstrate excellence in their activities that engage the public with science. Porras, an assistant professor of biomedical engineering at the University of Florida, has combined arts and crafts, advocacy for Spanish speakers and multilingualism in science, and online and in-person 23 FEBRUARY 2024 • VOL 383 ISSUE 6685

839

,

AAAS Mani L. Bhaumik Award for Public Engagement with Science

to public engagement with science, was established in 1987. Since 2019, it has been endowed by quantum physicist Mani L. Bhaumik, whose support of the award has placed renewed emphasis on recognizing meaningful dialogue and exchange between the recipient and various publics. Hayhoe is the chief scientist at The Nature Conservancy and the Paul Whitfield Horn Distinguished Professor and Endowed Chair in Public Policy and Public Law in Public Administration at Texas Tech University. She has demonstrated an extraordinary commitment to communicating about climate change, particularly with religious communities, and to encouraging other scientists to engage in dialogue with their own communities, with a focus on shared values and beliefs. Her TED Talk, “The Most Important Thing You Can Do About Climate Change: Talk About It,” has garnered more than 4 million views. She is the author of Saving Us: A Climate Scientist’s Case for Hope and Healing in a Divided World and coauthor of A Climate for Change: Global Warming Facts for Faith-Based Decision alongside her husband, a pastor. She also reaches audiences through her digital series, “Global Weirding: Climate, Politics, and Religion with Katharine Hayhoe.”

y

Joel R. Primack is the recipient of the 2024 AAAS Philip Hauge Abelson Prize, which honors an individual for significant contributions to the advancement of science in the United States through their research, policy work, or public service. Primack, distinguished professor of physics emeritus at the University of California, Santa Cruz, receives the award for creating opportunities for scientists to engage in policy, supporting policy-makers in their evidence-informed decision-making, advancing scientific knowledge in the field of physics and cosmology, and improving public understanding of science. Primack originated the first congressional fellowship program for scientists, which originated from workshops he launched as a graduate student at Stanford University and later grew into the AAAS Science & Technology Policy Fellowship program. He is a founder of the Union of Concerned Scientists, helped to create the American Physical Society Forum on Physics and Society, proposed the APS program of studies on public policy issues, contributed to the founding of the NSF Science for Citizens program, and created and provided initial leadership for the AAAS Program on Science and Human Rights. A leader in the emergence of the field of particle astrophysics, Primack continues to teach undergraduate and graduate students and advise doctoral candidates at UCSC. He has publicly communicated the importance of science, working with leading planetariums and writing about cosmology for popular audiences.

y g

PHOTO: COURTESY OF IF/THEN COLLECTION

AAAS Philip Hauge Abelson Prize

y

Each year, AAAS bestows awards that recognize scientists, engineers, innovators, public servants, and authors for their contributions to science and society. In 2024, these awards recognize a broad span of accomplishments, from mentoring the next, more diverse generation of chemical scientists, to using cutting-edge science to advance human rights, to connecting with faith communities about climate change. Learn more about the accomplished individuals and teams and the transformative works recognized by AAAS in 2024.

AAAS NEWS & NOTES engagement to highlight diverse voices in science and build bridges between scientists, future scientists, and the broader public. Porras combines her interest in art and science by crocheting stuffed microbes and sharing her work on Instagram, launching #MicrobeMondays to discuss microorganisms and later #MicroMartes to share content in Spanish. Her social media engagement has led to other outreach, including writing an article about the importance of science communication in multiple languages, serving on the organizing committee for the Inclusive SciComm symposium, and working with the podcast Story Collider to produce its first “Stories of Science” show in Spanish. Porras has also volunteered with Science Clubs Colombia, an organization in her home country that creates project-oriented, STEM-focused workshops for children.

AAAS David and Betty Hamburg Award for Science Diplomacy

science.org SCIENCE

,

Subaru of America, Inc., and AAAS have partnered to award the 2024 AAAS/Subaru Prize for Excellence in Science Books, which recognizes exceptional books about science that engage readers of all ages. The prizewinning books for 2024 are: Whale Fall: Exploring an Ocean-Floor Ecosystem by Melissa Stewart and illustrated by Rob Dunlavey. The Planets Are Very, Very, Very Far Away: A Journey Through the Amazing Scale of the Solar System by Mike Vago. Human Body Learning Lab: Take an Inside Tour of How Your Anatomy Works by Betty Choi. The Sounds of Life: How Digital Technology Is Bringing Us Closer to the Worlds of Animals and Plants by Karen Bakker.

y

23 FEBRUARY 2024 • VOL 383 ISSUE 6685

AAAS/Subaru Prize for Excellence in Science Books

y g

840

An interdisciplinary team of more than 80 researchers led by Indigenous scholars has received the 2024 AAAS Newcomb Cleveland Prize for an article that showcased findings about the rapid spread of horses through Indigenous societies in North America. The AAAS Newcomb Cleveland Prize, supported by The Fodor Family Trust, was established in 1923 and is awarded annually to the author or authors of an outstanding paper published in the Research Articles section of Science. Recipients of the award, the oldest awarded by AAAS, collectively receive $25,000. “Early dispersal of domestic horses into the Great Plains and Northern Rockies” was published in the 31 March 2023, issue of Science. Horses originated in North America more than 4 million years ago before evolving and dispersing into Eurasia. Study results show that horses were quickly integrated into Indigenous lifeways upon their reintroduction. The winning team—composed of Indigenous and Western scientists—conducted an interdisciplinary study of historic archaeological horse remains that integrated genomic, isotopic, radiocarbon, and paleopathological evidence. The researchers found that horses were integrated into Indigenous life many decades earlier than was previously understood by the Western scientific community, findings that align with the oral histories of multiple Indigenous groups. “This award creates a global platform of recognition demonstrating the strength inherent in combining Indigenous and Western scientific systems, methods, and methodologies to create a powerful new path forward in the sciences,” said coauthor Yvette Running Horse Collin, an Oglala Lakota scientist and geneticist.

y

For his long-standing commitment to advancing diversity in the chemical sciences through recruitment and mentoring of students from groups traditionally underrepresented in science, Luis A. Colón is the recipient of the 2024 AAAS Lifetime Mentor Award. The Lifetime Mentor Award honors an individual with more than 25 years of experience who has mentored significant numbers of underrepresented students: women of all backgrounds, men from underrepresented minority groups (African American, Native American, and Hispanic), and people with disabilities. Since 1995, Colón has worked to recruit students from his undergraduate alma mater—the University of Puerto Rico at Cayey—to conduct research at the University at Buffalo, creating a diverse research community in a department that previously had no Hispanic representation. Colón has directly mentored 51 graduate students, but his impacts are even more widely felt. One peer who nominated Colón for the Lifetime Mentor Award estimated that more than one-third of faculty members from underrepresented groups hired in chemistry departments in recent years have a connection to Colón’s program or those it inspired.

AAAS Newcomb Cleveland Prize

g

AAAS Lifetime Mentor Award

Eric Stover is the recipient of the 2024 AAAS Award for Scientific Freedom and Responsibility, which honors scientists, engineers, or their organizations whose exemplary actions have demonstrated scientific freedom and/or responsibility in challenging circumstances. Stover, the faculty director of the Human Rights Center at the University of California, Berkeley, School of Law and adjunct professor of law and public health, is honored for his work that combines pioneering scientific methods and technology with a commitment to human rights to hold accountable perpetrators of mass human rights violations and advance justice for their victims. Stover has led forensic investigations of mass graves in Argentina, Guatemala, Honduras, Chile, Brazil, Iraq, El Salvador, Bosnia, Croatia, and Rwanda. Stover is also a founding member of the International Campaign to Ban Land Mines, the organization that received the 1997 Nobel Peace Prize. His work in the Balkans helped lead to the conviction of the perpetrators of the Srebenica massacre. His work in Uganda has led to the establishment of a school for girls and women who were victims of the Lord’s Resistance Army.

p

Tareq Abu Hamed, executive director of the Arava Institute for Environmental Studies in Israel, receives the 2024 AAAS David and Betty Hamburg Award for Science Diplomacy for his leadership in using science to build relationships across the Middle East, particularly between Israelis and Palestinians, to work together to address mutual environmental concerns. The award recognizes an individual or a small group working together in the scientific, engineering, or foreign affairs communities making an outstanding contribution to furthering science diplomacy. First established in 1992 as the International Scientific Cooperation Award, AAAS renamed its Award for Science Diplomacy in 2021 for David and Betty Hamburg, who championed science diplomacy throughout their careers in science and medicine. Abu Hamed is a graduate of universities in Turkey and Israel, holds a PhD in chemical engineering, and is the first Palestinian to head an academic institute in Israel. As the executive director of the Arava Institute for Environmental Studies (AIES), Abu Hamed builds trust and fosters cooperation to address the effects of the climate crisis on vulnerable communities. As AIES leader, a role he has held since 2021, Abu Hamed launched the Center for Climate Change Policy and Research. Outside of his work at AIES, Abu Hamed was the highestranking Palestinian in the Israeli government when he served first as deputy chief scientist, then acting chief scientist of Israel’s Ministry of Science, Technology, and Space in 2015 and 2016.

AAAS Award for Scientific Freedom and Responsibility