16 FEBRUARY 2024, VOL 383 ISSUE 6684 
Science

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

EDI TO R I A L

Science communication at scale

O

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

g y g y ,

PHOTO: CAMERON DAVIDSON

y

“…the secret sauce is always to make it personal.”

10.1126/science.ado5736

SCIENCE science.org

p

times it’s wrong,” but rather, it’s “getting less and less n the television program MythBusters, which wrong as we go.” aired on the Discovery Channel from 2003 to Savage agreed that science is experiencing enormous 2016, the hosts Adam Savage and Jamie Hynedifficulty in changing strong views about topics like vacman tested popular myths and ideas arising cines and climate change. I asked him whether he ever everywhere from folklore to popular culture, changed the mind of an adult who had already decided designating them as either “busted” or “conwhether a myth should be busted or confirmed. Alas, he firmed.” The show used engaging and entersaid he had seen the opposite many times: People with taining variations on the scientific method to answer entrenched views held to them even after convincing questions such as whether you get wetter when walktests. He gave the example of a plane taxiing along a ing or running in the rain or what the fastest way is runway that is moving, like a conveyor belt, in the oppoto board an airplane. I hear frequently from young site direction. If the speed of the plane and the opposing scientists that the show inspired their generation to speed of the runway are equal, many people believe they join the scientific enterprise. I sat down with Savage will cancel out so that the plane will not take off. Affor an interview about what we can learn about public ter an elaborate demonstration involving an ultralight engagement in science. plane that took off on a tarp towed in the opposite diAs viewers will know, Savage exudes excitement and rection by a truck, Savage said a common response was, passion about his subjects. “For me,” he said, “the secret “We don’t agree now, even though sauce is always to make it personal. they’ve proved it. They must have It’s always to find that intersection gotten something wrong.” Daunting with my emotionality about the things words for science communicators. that I’m talking about…” I wanted to know if Savage could Given the current focus on reimagine one more episode of Mythsearch integrity, I wanted to know Busters that would demonstrate the how Savage reacted when he got reality of human-caused climate credible questions about the methchange. He said he had considered odology and conclusions reached this topic multiple times but couldn’t on MythBusters. “We took them sefigure out how to use the drama of riously,” he said. “Revisiting a myth Adam Savage, the MythBusters model to conceptuwas one of my favorite things.…Behost of MythBusters alize the sum of many small actions. fore the end of the first season, we Savage felt that understanding the were getting word that we had gotscale of the whole planet was someten some experimental methodology thing that didn’t fit the MythBusters rubric: “I’m alwrong. And there were several stories…that we tackled ways about building the biggest possible model you multiple times based on new data, new information, can to demonstrate that you have satisfied the condiand new methods of execution.” He had to push the Discovery Channel to let his team revisit episodes. tions to the greatest possible degree.” As successful as “Discovery always had to be dragged to do the revisit was, the MythBusters relied on the idea that seeing its,” he said. “But to me, it was the most scientific thing is believing. we did.” Fortunately, these episodes did well in the ratSavage is very aware that he may have helped creings, increasing the network’s support for them over ate a generation of scientists and inventors. “It has time, which is a great lesson for institutions and aubeen a remarkable progression over the past 20-plus thors who drag their feet when examining criticisms years,” he reflected. “People were saying, ‘Hey, you got of published papers. When we talked about trust in me through high school chemistry,’ and now they’re science, Savage felt that when a different result arises, saying, ‘I’ve sold three startups and MythBusters got people tend to think that science “got it wrong” to beme interested in science at the very beginning.’” There gin with. The message of science as a self-correcting is much that scientists and communicators can learn process is obviously not getting through. As he noted, from that legacy. it’s not as if science “get[s] things right and then some–H. Holden Thorp

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

683

NEWS

Air pollution, which dims a nighttime view of the Zhenhai Tower in Huai’an, China, also raises the risk of suicide.

IN BRIEF Edited by Jeffrey Brainard

p

PUBLIC HEALTH

16 FEBRUARY 2024 VOL 383 ISSUE 6684

| The U.S. Census Bureau announced last week it will not change questions about disability on its annual American Community Survey, after critics said the proposed revision would reduce the estimated prevalence of disabled people in the nation from 14% to as low as 8%. Census officials had said the new questions were recommended by a scientific advisory committee and reflected international standards for measuring disability. DIVERSITY

Early hunters used stone wall? | An ancient line of stones running for nearly 1 kilometer across the Baltic Sea’s floor was likely a wall built to help people hunt reindeer at the end of the last ice age, a study has concluded. The stones, found off Germany’s coast, were originally placed along the edge of an

A R C H A E O L O GY

science.org SCIENCE

,

684

Census backtracks on disability

But the proposal, posted in October 2023, drew more than 12,000 comments, many of which decried what they saw as a lack of engagement with the disability community. The bureau plans to publish another notice this year on the questions to be used in the 2025 survey and will also meet with researchers, activists, and other stakeholders to discuss ongoing efforts to improve collection of disability data.

y

| NASA’s premier center for space science, the Jet Propulsion Laboratory, announced last week it would lay off 530 employees, about 8% of its workforce, along with 40 contractors, because of uncertainty over this year’s budget and lawmakers’ qualms about skyrocketing costs for the Mars Sample Return (MSR) mission. After arrival at Mars as early as 2030, the $10 billion effort would pick up rock samples collected by the Perseverance rover, load them in a rocket, and ferry them back to Earth. A Senate appropriations bill for this year would provide the mission only $300 million, a 63% cut from 2023. Congress has delayed passing spending bills because of disagreements about overall federal spending levels. Planning to spend more

P L A N E TA RY S C I E N C E

than the Senate’s proposed level would be unsound, NASA Administrator Bill Nelson said in a statement. NASA is reviewing MSR’s plans as scientists worry its escalating costs are depleting the agency’s funding for other research.

y g

NASA cuts jobs at storied lab

a longer term impact. The research supports previous findings that airborne particles smaller than 2.5 microns (called PM2.5) not only harm physical health, but also quickly alter brain chemistry and are associated with depression and inability to cope with crises. As in most other countries, China’s suicide rate declined during the past decade, in part because of rising prosperity. The study, published this week in Nature Sustainability, attributed nearly 10% of the overall decline in suicide rates between 2013 and 2017 to a campaign China started in 2013 to reduce air pollution.

y

A

study in China that builds on evidence linking heavy air pollution to increased suicide risk credits air-quality improvements with preventing more than 45,000 suicides. To distinguish pollution’s effect from that of other risk factors, a research team relied on an unusual indicator, incidences of thermal inversions—when warmer air traps cold air and airborne pollutants close to the ground, increasing people’s exposure. The research team found that inversions, typically lasting 2 to 3 hours, were associated with an increase in suicide rates within 1 week but did not have

g

China reduces suicides by cutting air pollution

ancient lake or bog until rising sea levels submerged them, preserving the wall for millennia. Led by an instinct to follow straight lines, herds of hoofed animals may have followed it into the reach of waiting hunters, the researchers write this week in the Proceedings of the National Academy of Sciences. Estimated to have been built about 10,000 years ago, the structure is the earliest and largest of its kind identified in Europe, and challenges views that Stone Age hunters were highly mobile.

New Zealand scraps reforms

bed in southern France is offering scientists a rare glimpse at 470-million-year-old life forms, including well-preserved soft tissue such as sponge bodies. The site, discovered by amateur paleontologists, dates to the Ordovician period, when marine life flourished and primitive plants began to emerge on land. In a paper published this week in Nature Ecology & Evolution, researchers describe more than 400 fossils—collectively called the Cabrières biota, after the region in France where they were found—including mollusks, conical-shelled creatures called hyoliths, and arthropods such as trilobites. Fossilized soft tissue from this period is rare, but the site contains many examples of the ancient marine creatures’ internal organs, such as hyolith guts. The squishy tissues fossilized between layers of mudstone and siltstone, turning into iron oxide crystals that retained the tissue’s shape over many millions of years.

ALASKAPOX DEATH Health officials in Alaska reported this week the first death from Alaskapox, a viral disease similar to smallpox that causes skin lesions. The immunocompromised man may have contracted it through a scratch from a stray cat he was caring for that hunted small mammals, a known reservoir for the virus, the officials said. Only six other cases in Alaska have been reported since the first, in 2015; those people had mild symptoms and recovered.

SCIENCE science.org

16 FEBRUARY 2024 VOL 383 ISSUE 6684

685

,

Farmers protested pesticide regulations last week outside the European Parliament in Strasbourg, France.

FALSE VACCINE STORIES More than 300 different false tales about vaccines are circulating on social media and in search results, according to an analysis by Newsguard, a company that tracks false stories to rate the trustworthiness of online news sources and combat misinformation. For example, one claims that blood clots reported by embalmers are proof of widespread deaths caused by COVID-19 vaccines. But clots commonly form after death, and embalmers often don’t know the deceased’s vaccination status, Newsguard says.

y

DARWIN’S TROVE A website has released a 300-page catalog detailing 7400 books, pamphlets, and journals contained in the personal library of naturalist Charles Darwin. The reference lists hundreds of works not previously known to have been included and new details on others, such as their authors and dates. John van Wyhe at the National University of Singapore and colleagues spent 18 years compiling the catalog for the online Complete Library of Charles Darwin, which also provides links to digitized copies of the works. The printed originals are kept at the University of Cambridge; Darwin’s home, Down House; and other institutions.

y g

PHOTO: FREDERICK FLORIN/AFP VIA GETTY IMAGES

| A newly discovered fossil

y

| Researchers in Europe have raised the alarm about what they see as European leaders backtracking on green legislation such as reducing pesticide use, against expert advice. Last week, European

R E G U L AT I O N

PA L E O N T O L O GY

FAILED ANTIBODIES More than 50% of commercial antibodies marketed to detect and characterize proteins linked to Alzheimer’s disease and other neurological illnesses failed to work, researchers found. In one of the first systematic evaluations of its kind, reported in eLife, they tested 614 antibodies to 65 proteins involved in the diseases. A hopeful sign: Two-thirds of the proteins were covered by at least one highperforming antibody. After the research team informed manufacturers about the failing ones, they altered the recommended usage for 153 and removed 73 more from the market.

g

EU retreats on pesticide cuts

Site preserves ancient innards

IN OTHER NEWS

p

| Scientists in New Zealand this week criticized the government’s decision to cancel a multiyear reform plan to overhaul research funding and improve career opportunities. The move to scrap the Te Ara Paerangi Future Pathways reform program came from the new, center-right ruling coalition, which was sworn in in November 2023 and has vowed to cut spending to finance tax cuts. The previous government had designed the reform program to better align research priorities with the country’s needs, boost funding, and improve support for early-career researchers. It also included NZ$450 million ($275 million) to build and renovate laboratories around the capital city of Wellington. The New Zealand Association of Scientists says reform was overdue and ending the Pathways plan could lead to funding reductions. POLICY

Commission President Ursula von der Leyen announced she was shelving the Sustainable Use Regulation—a controversial, stalled proposal to halve the use of chemical pesticides by 2030, which she said had become a “symbol of polarization.” The announcement came as farmers across the continent continued protesting environmental regulations and other issues and 4 months before EU elections, which are expected to pull the union further to the right. The move follows efforts at the European and national levels to tone down other pieces of green legislation. In France, scientists criticized the government after it announced on 1 February that it would pause Écophyto, a national plan to slash pesticide use.

NE WS

IN DEP TH

p g y

Michael Mann (center) arriving last week at a court in Washington, D.C. His defamation lawsuit succeeded after 12 years of litigation.

SCIENCE AND THE MEDIA

Jury rules noted climate scientist was defamed Verdict punishes bloggers who compared Michael Mann’s work to molestation

16 FEBRUARY 2024 VOL 383 ISSUE 6684

science.org SCIENCE

,

686

move, in 2022, to the University of Pennsylvania, Mann is a frequent source in news stories and a combative personality on social media, where he regularly butts heads with those he believes are misrepresenting climate science or seeking to delay action on climate change. At the heart of Mann’s lawsuit are two 25-year-old scientific papers that he led. The studies combined historical records with tree rings and other temperature proxies going back 1000 years to show that temperatures stayed largely flat until the past century, when they rose sharply. A key chart from the papers, dubbed the “hockey stick” because of its shape, was used in a 2001 U.N. climate report and has remained an icon of climate science, underscoring the outsize impact of human activity on temperature. For nearly as long as it has existed, however, Mann and his hockey stick have been besieged by criticism, scientific and otherwise. Follow-up work, some by climate

y

L

ast week, a jury awarded Michael Mann, a prominent climate scientist, more than $1 million in damages after finding he was defamed by the writers of two blog posts 12 years ago that compared his work on global warming to child molestation. The ruling is a victory for climate researchers who have been frequently attacked online over their work and especially for Mann, who has faced the brunt of it, says Lauren Kurtz, executive director of the Climate Science Legal Defense Fund. In a statement, Mann said, “I hope this verdict sends a message that falsely attacking climate scientists is not protected speech.” At the same time, the ruling could end up chilling public criticism of science, says Gene Policinski, a senior fellow at the Freedom Forum, a nonpartisan foundation focused on free speech. People might be

more vague or circumspect in their commentary, to the detriment of the public, he says. “It’s important in today’s world for people to be aware of research that’s going on and having people both praise and criticize it openly.” Chilling debate is a risk, and countering speech with more speech remains the ideal, says David Schulz, an expert in media law at Yale University. But with the decline of the mainstream media, rising political polarization, and ever-present social media, that ideal can seem quaint. In society today, counterspeech is often ineffective, he says, because it will never be heard by people trapped in their bubbles. “We live in a society where people have difficulty knowing what’s true and what’s not,” he says. “These types of lawsuits are one of the few tools we have as a society to hold accountable people who abuse their free speech rights.” A longtime paleoclimatologist at Pennsylvania State University (Penn State) until his

y g

By Paul Voosen

N E WS

By Daniel Clery

T

16 FEBRUARY 2024 VOL 383 ISSUE 6684

687

,

he first x-ray survey of the universe in decades has cataloged and measured the biggest lumps in the cosmos: clusters of hundreds or thousands of galaxies. Their masses show how fast matter clumped together over cosmic history, and they tell a story that is at once satisfying and frustrating: The universe is about as lumpy as the standard cosmological theory predicts. “This will make many theorists very sad,” says Vittorio Ghirardini of the Max Planck Institute for Extraterrestrial Physics (MPE), a member of the mission, called eROSITA (extended Roentgen Survey with an Imaging Telescope Array). “They want to find something new, but the theory of everything is working fine.” eROSITA is a German instrument launched in 2019 onboard a Russian satellite called Spektr-RG, which also carries a smaller Russian telescope (Science, 14 June 2019, p. 1020). Its aim is to complete the first census of x-ray sources across the entire sky since ROSAT, another German-led mission, did so in the 1990s. On 31 January, the eROSITA team released its first batch of data, cataloging 900,000 sources across half the sky—already more than were known from 60 years of x-ray astronomy. The list includes 710,000 supermassive black holes at the hearts of distant galaxies, 180,000 stars hot enough to emit x-rays in our Milky Way, as well as supernova remnants, pulsars, and other objects. “The quality of images was stunning, even to us,” says principal investigator Andrea Merloni of MPE. The catalog also contains 12,000 galaxy clusters, among the largest structures in the universe. Each one teems with galaxies and is pervaded by thin gas, compressed by the

y

SCIENCE science.org

eROSITA telescope shows galaxies’ “clumpiness” matches predicted effect of dark energy, dark matter

y g

methodological “tricks,” such as Mann’s technique for combining historical and proxy temperature records, could be misused by opponents of climate action—as they ultimately were. Numerous institutions, including Penn State and the National Science Foundation, investigated the leaked emails and cleared Mann of any research misconduct. But he continued to face online attacks. One came in 2012 from Rand Simberg, then a blogger at the Competitive Enterprise Institute, following the arrest of Jerry Sandusky, a serial child molester and Penn State football coach. Simberg likened the case to the university’s investigation of Mann, saying Mann “molested and tortured data” to create his hockey stick. Mark Steyn, a conservative author and broadcaster, then quoted Simberg’s post in a blog hosted by the National Review, calling Mann’s work “fraudulent.” Mann fought back before and after the email leaks, often with vitriol. Correspondence revealed in the trial showed that he called one of his critics “human filth” and suggested that a retired atmospheric scientist with contrarian views advanced her early career by dating someone on the faculty. In 2011, he sued Timothy Ball, a prominent Canadian climate change doubter, for defamation after Ball accused him of scientific fraud, a case that was dismissed in 2019 because of its extensive delays. In 2012 he

X-ray survey bolsters theory of universe’s expansion

y

David Schulz, Yale University

ASTRONOMY

g

“ These types of lawsuits are one of the few tools we have as a society to hold accountable people who abuse their free speech rights.”

sued Simberg and Steyn. Under U.S. law, Mann had to clear a high bar to prove defamation. He needed to show that the accused acted with malice and knowingly made false claims or showed a “reckless disregard” for the truth. In closing the case, Mann’s lawyer, John Williams, said Simberg and Steyn had shown that recklessness, likening them to those who still deny that Donald Trump lost the 2020 U.S. presidential election. “We don’t know for certain why they continue to deny that in the face of overwhelming evidence to the contrary,” he said. The defense spent days walking the jury through critiques of Mann’s hockey stick, stressing that Simberg fully believed the arguments he was making. “Rand is just a guy. Just a blogger. Voicing his truly held opinions on a topic he believes is important,” Victoria Weatherhead, Simberg’s lawyer, said in closing statements. Mann’s legal team tried to show how the blogs had caused him emotional and financial damage, another criterion for defamation. Mann said on the stand “he felt like a pariah.” Mann’s lawyers also showed his research grants dwindling after 2012, although they provided no evidence that the blogs spurred this decline. Around the same time, Simberg’s lawyers noted, he published a successful book, increasing his fame and leading to lucrative speaking engagements. In ruling for Mann, the six-person jury in the Superior Court of the District of Columbia awarded him just $2 in actual damages due to the defamation, but then added $1 million in punitive damages against Steyn; Simberg faces $1000 in such damages. In a statement, Simberg painted the verdict as a partial win, noting the small size of his fine. Kurtz says it’s notable that the jury awarded far more damages for Steyn’s claims of fraud—an accusation of research misconduct that cuts to the heart of a scientist’s professional standing—than for Simberg’s hyperbolic comparisons of Mann to Sandusky. Although Mann was the victim of defamation, Kurtz says cases that lob the same accusation at scientists are on the rise. Just last year, one of her organization’s clients was sued for debunking and deplatforming a blogger’s climate-conspiracy videos. “I do have concerns about using defamation to hassle climate scientists—but that’s already happening.” Steyn has indicated he will appeal the decision. Mann’s plans have also emerged: He will team up to write a book with Peter Hotez, a vaccine scientist and a veteran of online clashes over vaccination during the pandemic. It will be titled Science Under Siege: The Anti-Science Forces That Threaten Humanity. j

p

contrarians, found statistical flaws in his methods. Critics also cast doubt on the whole record by noting that tree rings in the far north tracked global temperatures less faithfully after 1960. But the underlying insight held up. A consortium of other scientists, using nearly 700 proxy records, found in 2017 that recent warming is unmatched in the past 2000 years. All the reconstructions show the exact same thing, says Jessica Tierney, a paleoclimatologist at the University of Arizona: “The temperatures we’re experiencing today are really unprecedented.” Indeed, she says, temperatures may now be the warmest in the past 100,000 years. The barrage of attacks against Mann reached a peak after “climategate”: a 2009 leak of emails from the University of East Anglia. The trove included requests from fellow researchers to Mann asking that he delete emails, for fear that their discussion of

NE WS | I N D E P T H

eROSITA, an orbiting German x-ray telescope (left), charted nearly 1 million point sources amid a foreground of diffuse, hot gas in the Milky Way (right).

p y ,

science.org SCIENCE

y g

16 FEBRUARY 2024 VOL 383 ISSUE 6684

eROSITA has also weighed in on the search for the mass of neutrinos. The big bang gave birth to countless numbers of these tiny particles, but they interact with other particles so rarely that measuring their mass is extremely challenging. Because of their abundance throughout space, however, their gravity does affect the distribution of matter. From its survey, the eROSITA team estimated a combined mass of at most 0.22 electronvolts for the three different types of neutrino. That value is less than one-millionth of the mass of an electron and—again—is in line with other measurements. “The more independent measurements we have on the clustering of matter, the better constraints we can put on neutrino masses,” Dunkley says. The eROSITA researchers hope to improve the accuracy of their measurements. In 2 years they will likely release the rest of the data they gathered during the first 2.5 years of observing. But eROSITA may never get to complete its planned 4-year survey. In February 2022, days after Russia’s invasion of Ukraine, the team put the instrument into hibernation after Germany forbade any research collaboration with Russia. There has been no observing since. “It’s in limbo,” says Merloni, who now has no contact with his colleagues in Russia. eROSITA’s results are a foretaste of what researchers will be grappling with a few years from now when they start getting data from several upcoming large-scale surveys, including Europe’s Euclid space telescope and the Vera C. Rubin Observatory, a U.S.-backed telescope in Chile. Stefano Borgani, director of the Astronomical Observatory of Trieste, says eROSITA’s results “really demonstrate how far we can go in the coming years.” j

y

688

universe to be less clumpy than lambdaCDM predicts—a hint that dark energy is stronger than it should be. eROSITA researchers use a different approach. By looking at the x-ray brightness of each cluster and combining that with data from surveys at other wavelengths, they can estimate its mass. With mass estimates for all of their clusters, they can calculate S8: how structured, or clumpy, matter is across the cosmos. They find it is close to the theoretical prediction. “We don’t find a tension,” says team member Esra Bulbul of MPE. “There is no need for new physics.” Ofer Lahav of University College London, a senior scientist in the Dark Energy Survey, one of the weak lensing studies, says it’s “intriguing” that the new finding contradicts the three weak lensing studies. But Jo Dunkley, a cosmologist at Princeton University, is not yet ready to say the S8 tension has disappeared. “Something is going on here,” she says. “It’s too soon to say that weak lensing is the one with the problem.” The eROSITA team also derived values for other cosmological numbers, including w, a parameter that describes dark energy’s strength. According to lambda-CDM it should be exactly –1. Most measurements of it are close to that figure and eROSITA concurs, putting it at –1.12 with error bars of +/–0.12. For Ghirardini, this is eROSITA’s most significant result. “This is 70% of the content of the universe,” he says. “We want to know what it is.” Not all theorists want measurements of S8 and w that agree with expectations, however. “Many of us are hoping to see something that departs from what we have,” Dunkley says, because anomalies might point the way to a new understanding of the cosmos.

g

throng of galaxies until it is hot enough to emit x-rays. Using a subset of those clusters, 5259 of them, eROSITA researchers measured the evolution of the universe. Current theory holds that the universe has three main ingredients: roughly 5% normal matter, the stuff of stars, planets, and the gas in between; 25% mysterious dark matter, invisible stuff whose gravity holds rotating galaxies together and distorts light from background galaxies; and 70% dark energy, an innate pressure in empty space that is accelerating the expansion of the universe. The theory is referred to as lambda-CDM, with lambda referring to a constant that gives space its inherent springiness and CDM standing for cold dark matter. Lambda-CDM describes the universe very well, but in a few respects the nearby universe doesn’t look quite like it should if it had evolved according to lambda-CDM. Cosmologists debate whether these tensions are due merely to measurement bias or are an unsettling hint that lambda-CDM is wrong. One mismatch concerns the Hubble constant, the expansion rate of the universe: In our cosmic neighborhood, the galaxies are flying apart faster than lambdaCDM says they should (Science, 3 November 2023, p. 499). A second tension, less well known, concerns a parameter called S8, which describes how clumpy the universe is. Optical surveys can measure this by studying the shapes of distant galaxies to see how the images are distorted by the gravitational influence of all the clumped-up matter, both dark and light, between the galaxy and Earth, a phenomenon called weak gravitational lensing. Clumping reflects the tug of war between gravity and dark energy. Three optical surveys have each found the nearby

SCIENCE POLICY

What’s at stake for science in ‘abortion pill’ case Supreme Court decision this summer could gut FDA’s authority over drugs By Meredith Wadman

N

PHOTO: ANDREW CABALLERO-REYNOLDS/AFP VIA GETTY IMAGES

,

689

y

16 FEBRUARY 2024 VOL 383 ISSUE 6684

y g

SCIENCE science.org

y

Who’s suing FDA, and why? In 2022, a group of antiabortion physicians and medical organizations filed a lawsuit challenging the approval of the drug in 2000 as rushed and improper. The suit also argued that FDA decisions in 2016 and 2021 that made mifepristone easier to access put patients at risk. The plaintiffs said they had standing to sue because they would be injured by the workload when mifepristonerelated emergencies “overwhelm” medical systems. In an April 2023 opinion, Texas district court Judge Matthew Kacsmaryk agreed, citing a now-retracted 2021 paper that concluded medical abortions produced more emergency room (ER) visits than surgical abortions. The judge, who suspended mifepristone’s approval, also referred to a 2022 paper, just retracted as well, which found that ER visits following mifepristone use were more likely than surgical abortions to result in hospital admission to ensure an abortion is complete.

brief filed last month on behalf of FDA by the American College of Obstetricians and Gynecologists, the American Medical Association, and 14 other medical societies. And an analysis in Nature Medicine this week concludes that U.S. telehealth medication abortions are as safe as in-person medication abortions. The key retracted paper paints a different picture. It used a Medicaid database to analyze differences in ER visits after medication and surgical abortions. The authors found that from 1999 to 2015, an ER visit for any reason was 22% more likely to occur after a mifepristone-aided abortion than after a surgical aborThe Supreme Court will hear a mifepristone case on 26 March. tion; an abortion-related ER visit was 53% more likely. But Ushma In August 2023, a panel of the Fifth Upadhyay, a public health scientist at the Circuit Court of Appeals restored FDA’s University of California, San Francisco, says original approval of the drug. But it agreed ER visits after mifepristone use are a poor with the district court that FDA’s regulatory proxy for adverse events, because about 50% actions in 2016 and 2021 easing access to of such visits do not lead to treatment. She mifepristone should be reversed. FDA and notes that people having a medication aborthe drug’s maker appealed to the Supreme tion experience bleeding and cramping, and Court, which has stayed any change in acmany seek assurance in ERs. cess for now. What’s at stake for FDA? Nearly everything, according to groups Is mifepristone safe? that filed briefs in support of the agency. Hundreds of studies and clinical trials say They argue that FDA needs free rein to let it is, including a 2018 consensus report scientific expertise guide its decisions. The by the National Academies of Sciences, Fifth Circuit’s opinion, finding that FDA was Engineering, and Medicine. Serious ad“arbitrary and capricious” in easing access verse events—blood transfusions, major to mifepristone, calls the foundation of its surgery, or hospital admission—occurred authority into question, they say. For exin 0.31% of more than 11,000 patients in ample, the appeals court asserted that FDA a large California study. “It is exceedingly should have considered the “cumulative rare for patients to experience even minor effect” on patient safety of loosening several complications from medication abormifepristone restrictions. Drugmakers and tion,” according to a friend-of-the-court

g

What is mifepristone? Mifepristone, used in more than half of U.S. abortions in 2020, is part of a two-drug combination taken to end early pregnancies. It blocks the action of progesterone, a hormone essential to maintaining pregnancy. Originally, mifepristone had to be prescribed and dispensed in person by a physician within 7 weeks of the last menstrual period. Today, in states that haven’t made abortion illegal, a prescription can be obtained in a telemedicine appointment, including with physician assistants or nurse practitioners, and filled by mail. It can now be taken through 10 weeks after the last menses.

Its scientific underpinnings were in the spotlight last week, when the journal Health Services Research and Managerial Epidemiology retracted two articles that raised questions about the safety of mifepristone. They were cited by the conservative district court judge in Texas who first ruled to suspend mifepristone’s FDA approval. Sage Publishing said it retracted both studies for “a lack of scientific rigor”; because one reviewer was affiliated with the antiabortion Charlotte Lozier Institute (CLI), a funder of the described research; and because the authors disclosed no conflicts of interest even though all but one were affiliated with antiabortion groups. Here are the issues.

p

ext month, the U.S. Supreme Court will hear oral arguments in a case that could hamper abortion access and have profound implications for the authority of the Food and Drug Administration (FDA). The case, which involves disputes over the safety of the abortion drug mifepristone, is an appeal arising from two lower court decisions that, taken together, challenge FDA’s approval of the drug and relatively recent easing of access to it. Daniel Aaron, a physician and law professor at the University of Utah, says conservatives see the case as “an opportunity to hinder the FDA while encumbering access to an important medication for women’s health.”

NE WS | I N D E P T H

their advocates say that is an unrealistic demand, with no basis in law. The Fifth Circuit also faulted the FDA Adverse Events Reporting System (FAERS). In essence, the court found that because FDA had in 2016 removed a requirement that mifepristone prescribers report nonfatal adverse events, the agency could not rely on FAERS data in regulating access to the drug. Industry groups and other critics of the ruling say that could force FDA to completely ignore warning signs for other drugs.

I

,

science.org SCIENCE

y

16 FEBRUARY 2024 VOL 383 ISSUE 6684

Hacking convection wasn’t enough, however: moist air was also needed. The model showed that moist, high-altitude winds from n the United Arab Emirates, water is the Persian Gulf would suffice. When condimore valuable than oil. To support its cittions were ripe, the model found, a 20-squareizens, the nation relies on expensive dekilometer solar field would increase a salination plants and campaigns of cloud storm’s total rainfall by nearly 600,000 cubic seeding from aircraft, which spray partimeters—equivalent to 1 centimeter of rain cles into passing clouds to trigger rainfall. falling across an area the size of Manhattan. But according to a new modeling study, If such rainstorms occurred 10 times in one there may be another way to stir up a rainsummer, they would provide enough water to maker: with city-size solar farms that create support more than 30,000 people for a year. their own weather. The heat from dark solar The new work makes sense, and it’s “very panels can cause updrafts that sometimes stimulating,” Lu says. “They are targeting a lead to rainstorms, providing water for tens real solution.” One concern, however, is that of thousands of people. “Some solar farms are the simulated solar panels were darker than getting up to the right size right now,” says most manufacturers make them. Some solar Oliver Branch, a climate scientist at the Unipanels are even reflective, designed to cool versity of Hohenheim who led the work, pubtheir surroundings, Lu says. lished last month in the journal Earth System Still, Branch hopes the Dynamics. “Maybe it’s not idea could soon be tested. science fiction that we can Solar farms in China and produce this effect.” elsewhere are nearly big Few studies have exenough, he says. If they amined how renewwere built in the right able energy might shift spots, it wouldn’t take weather patterns. In 2020, much to darken the panresearchers found that els and to plant dark, implausibly large solar drought-tolerant crops befarms, taking up more tween panel rows. than 1 million square The UAE funded the kilometers in the Sahara study. Whether it will test desert, could boost local the scheme remains to be rainfall and cause vegSolar farms may soon be big enough seen. The country “is cometation to flourish. But the to create their own weather. bounty would come with a mitted to study the potencost, the researchers found: By altering wind tial implementation of all robust strategies, patterns, the solar farms would push tropical such as optimizing convection,” says Alya rain bands north. “If you push those northAl Mazrouei, director of the UAE’s Research ward, that’s not good news for the Amazon,” Program for Rain Enhancement Science. For says Zhengyao Lu, a climate scientist at Lund now, she adds, the country is deeply commitUniversity and lead author of the 2020 study. ted to its cloud seeding program, carrying out Branch and his co-authors wanted to look some 300 missions each year. at solar farms of more realistic sizes. They Branch and his colleagues say the scheme turned to a weather model produced by the could also work well in places such as NaU.S. National Center for Atmospheric Remibia and Mexico’s Baja Peninsula. They’re search that can account for land surface also trying to improve the realism of their sochanges. They modeled the solar farms as lar panel simulations by cross-checking them nearly black fields that absorbed 95% of the with field measurements at existing solar sunlight, surrounded by relatively reflecfarms. Ultimately, Branch hopes the rainmaktive sand. When the solar farms exceeded ing potential of solar farms will lead to more 15 square kilometers, they found, the inbeing built. “If you can provide evidence that creased heat they absorbed appreciably ina huge solar farm produces rainfall,” he says, creased the updrafts, or convection, that “that might give impetus to increase the size drive cloud formation. of them.” j

y g

690

By Paul Voosen

y

With reporting by David Malakoff.

Updrafts from dark solar panels could fuel storms

g

Will the retractions affect the case? Several medical-legal experts said it’s unlikely the retractions will influence the high court, which is officially constrained to consider only arguments and briefs filed by 30 January, 5 days before the retractions were announced. But Arthur Caplan, a bioethicist at New York University’s Grossman School of Medicine, argues the retractions are vital to the case. “The whole basis of claims of danger from mifepristone to women sits on these papers. There’s nothing else in the literature,” he says. “If these papers fall, then the argument that upper courts are reviewing falls apart.” j

Giant solar farms could provoke rainclouds in the desert

p

What were the methodological issues with the retracted papers? Sage flagged “misleading” data presentations in the 2021 paper, which plotted all ER admissions after medication abortion using a scale of 1000s, and on the same graph plotted abortion-related ER visits on a scale of 100s. In a written rebuttal, the authors write that the dual axis “provided sophisticated readers a faster way to visualize the degree of change in all … outcome variables over time.” Sage also publicly said there were “material” errors in the authors’ analysis of the data in the two papers cited in the district court opinion. In a private letter to the authors, which they provided to Science, Sage specifies that they “artificially inflat[ed] the number of adverse events” by counting multiple visits by the same patient; that “conflating” ER visits with adverse events without examining diagnoses or treatments “may not be a valid or rigorous approach”; and that one paper’s conclusion that the miscoding of incomplete abortions as miscarriages caused serious adverse events was “inaccurate and unsupported by the data.” James Studnicki, a health services researcher with CLI and the first author of all three papers, rejects those claims. “No single specific finding in any of the three papers has been explicitly challenged, let alone invalidated. … These retractions are meritless,” he tells Science.

ATMOSPHERIC SCIENCE

Wildebeests, zebras, and Thomson’s gazelles (not shown) all migrate through the Serengeti for food.

Each species’ dining habits set the table for the next wave By Elizabeth Pennisi

PHOTO: JASON DONALDSON

,

691

y

16 FEBRUARY 2024 VOL 383 ISSUE 6684

y g

SCIENCE science.org

y

E

ach April, some 200,000 zebras begin their journey 800 kilometers northward through the Serengeti in Tanzania and into Kenya. About 1.3 million wildebeests join the throng, and after nearly 1 month come 400,000 Thomson’s gazelles. The procession is “breathtaking,” says T. Michael Anderson, a savanna ecologist at Wake Forest University. On p. 782, he and colleagues clarify the timing and other details of this iconic migration, uncovering complex forces that order each species’ movements and showing how fire and unusual weather can disrupt them. By watching the herds with camera traps for 8 years, tracking them with GPS, and collecting feces to understand their diets, the group has “unraveled the mysteries of one of most impressive natural spectacles in the world,” says Duncan Kimuyu, an ecologist at Karatina University. Diet, the authors find, is a major factor shaping the migration. “Each species affects the availability of forage for others,” says Emily Bennitt, a wildlife ecologist at the University of Botswana. Before humans dominated much of the landscape, large grazing mammals from buffalo to antelopes made seasonal migrations across many parts of the world, following the growth of new vegetation. But today, “The populations of many of the migratory species are declining rapidly,” Kimuyu says.

One of the few remaining unimpeded large mammal migrations unfolds in the Serengeti, with three species believed to travel in reverse order of size. It was thought that zebras, weighing in at 230 kilograms, moved out first, then the 150-kilogram wildebeests, and last the 20-kilogram Thomson’s gazelles. Why had provoked debate. “We assumed it was a game of follow the leader,” Anderson says. He and some other scientists theorized that by cropping the tall vegetation, the bigger animals ahead enable smaller species to reach lower growing plants—a process called facilitation. Others thought fear of predators such as lions played a role. A few ecologists argued that the staggered pattern was a way to reduce competition for food along the way. To better understand the migration, Anderson and colleagues deployed about 200 camera traps across 1000 square kilometers, starting in 2012; an army of trained volunteer scientists, and later an artificial intelligence program, scanned the images to identify species. The group also analyzed movements of 54 wildebeests and 28 zebras wearing GPS collars and collected 232 dung samples, which had DNA of plants the animals ate. The researchers were surprised to find little time separation between the zebras and wildebeests—Anderson now views their movement as a comigration, although zebras may reach patches of ungrazed land first. And camera traps revealed no evidence that predators influenced the migration’s dynamics.

g

Diet explains Serengeti grazers’ predictable migration

p

ECOLOGY

Plant DNA in the dung samples yielded another surprise: Each species has its own dietary preferences, making choices among the grass and other plants that naturalists had missed. Coarse, tall grasses that are poor in nutrients make up more than 90% of zebras’ diet, their dung showed. Zebras have wide muzzles that scoop up large quantities of grasses indiscriminately and digestive tracts that rapidly process this roughage. The researchers conclude the animals stay just ahead to get first dibs on dense patches of tall grass and avoid competition with numerous wildebeests. “If zebras do not have the capacity to only eat the best stuff, they at least want to get some of the good stuff,” says Elizabeth Le Roux, a large-mammal ecologist at Aarhus University. As wildebeests then crowd out zebras, their narrower muzzles can more effectively pick out the “good stuff,” such as legumes, the DNA of which is found in their dung. Then the gazelles arrive. With smaller stomachs and slower digestion, “gazelles must focus on high-nutrient foods, which are insufficient to fill a zebra’s or wildebeest’s stomach, to meet their nutritional needs,” says Saeideh Esmaeili, an ecologist at Colorado State University. These foods, forbs and other young herbaceous plants, become much more accessible after the zebras and wildebeests eat down all the tall grasses, Anderson and his colleagues report. Tender new grass shoots also appear by the time the gazelles arrive. “This profound paper shows that a mixture of processes act together to allow species to depend on each other, coexist, and remain in the [migration],” says Anthony Sinclair, an emeritus ecosystem ecologist at the University of British Columbia who has worked in the Serengeti since 1965. Fires or intense rains can alter the migration timing, strengthening the link with diet. After a fire burns the savanna, the new growth is particularly nutrient rich, so in those years, gazelles didn’t follow the wildebeests so closely, Anderson explains. In contrast, rain early in the year stimulated excess growth of tall grasses, making the gazelles more dependent on the larger grazers. These natural experiments also suggest climate change may upend the animals’ incredible journey across the Serengeti. “A future with more or more widespread fires might threaten the migration dynamics,” says Corinna Riginos, an ecologist with the Nature Conservancy. “Migration is an important process for keeping herds healthy and abundant—lose the migration, and you may lose a lot of animals.” j

SCIENCE science.org

1

Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, MD, USA. 2Howard Hughes Medical Institute, Johns Hopkins University School of Medicine, Baltimore, MD, USA. Email: [email protected]

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

703

,

10.1126/science.adn7791

C

ombination antiretroviral therapy (ART) for HIV-1 infection reduces plasma virus levels such that clinical assays for viremia are negative. Because of unexpected cooperativity in dose-response relationships and drug synergies, ART completely inhibits new infection events, halting viral evolution and disease progression (1). However, ART is not curative owing to a persistent latent reservoir of HIV-1, the half-life of which is long enough to guarantee lifetime persistence, even with optimal ART (2). Within weeks, interruption of ART leads to rebound viremia, with subsequent disease progression toward fatal immunodeficiency. Therefore, lifelong ART is necessary. People who acquire HIV-1 during their 20s face 50 to 60 years of continuous treatment with multiple antiretroviral drugs. Children infected perinatally face a lifetime of treatment. Thus, there is a pressing need for a cure, which, in a practical sense, amounts to preventing viral rebound after ART interruption. Present efforts focus on eliminating the reservoir. The latent reservoir consists primarily of resting memory CD4+ T cells that carry transcriptionally silent proviruses, which are unaffected by ART or immune responses (2). Although other reservoirs may exist in the central nervous system and elsewhere, the reservoir in resting CD4+ T cells is readily demonstrable in blood samples from all people with HIV and provides simple, coherent explanations for the hallmark features of HIV-1 persistence: stability for decades, wide distribution in tissues, latency, and clonality (2). The establishment of the latent reservoir must be understood in the context of a fundamental transition that lymphocytes make between resting and activated states. Encounter with antigen drives the transformation of resting CD4+ T cells into lymphoblasts, which proliferate and differentiate into effector cells. This blast transformation involves substantial increases in size and metabolic activity, extensive changes in gene expression, and entry into the cell cycle. At the conclusion of

an immune response, most activated cells die, but a subset survive and revert back to a resting state as long-lived memory cells. The HIV-1 promoter contains conserved binding sites for the activation-dependent host transcription factors nuclear factor of activated T cells (NFAT) and nuclear factor kB (NF-kB), demonstrating that HIV-1 and related viruses evolved to replicate in activated CD4+ T cells, which are biosynthetically capable of producing virions. Most infected lymphoblasts die quickly but not before producing enough virus to sustain the infection. One model suggests that a state of latent infection may be established when activated cells are infected during the transition back to a resting memory state that is nonpermissive for viral gene expression, which is due in part to sequestration of NFAT and NF-kB (2). The cells then persist through normal mechanisms that maintain immunologic memory. In this model, HIV-1 latency is a consequence of normal immune physiology. Subsequent cellular activation can induce latent proviruses, leading to rebound viremia if ART is interrupted. Cure has been achieved in a few people who received hematopoietic stem cell transplantation (HSCT) from donors homozygous for a deletion in the HIV-1 co-receptor C-C chemokine receptor type 5 (CCR5) (3). HSCT has substantial mortality and can be carried out only when required to treat an otherwise fatal malignancy. In reported HIV-1 cure cases, host origin immune cells, including latently infected cells, were eliminated in a virus-independent fashion by the conditioning regimen and graft-versus-host disease (GVHD), which together can cause near complete replacement of host white blood cells with HIV-1–resistant donor cells. Virus released by any surviving infected host cells cannot replicate in the HIV-1–resistant, donor-derived cells that make up the reconstituted immune system. Three interesting but unsuccessful cases of HSCT in people with HIV-1 infection used donors who were wild type for CCR5 (4, 5). ART was continued for several years after transplantation to prevent HIV-1 infection of donor cells. After reconstitution, host immune cells were present at levels of ≤0.001%. This is a 100,000-fold reduction in infected cells, assuming that latently infected

y

N.A. is supported by the National Institutes of Health (R35GM143108 and R01CA266100), the V Foundation for Cancer Research, and the Boettcher Foundation. T.R.C. is an investigator of the Howard Hughes Medical Institute.

By Janet D. Siliciano1 and Robert F. Siliciano1,2

y g

AC KN OWL EDG M ENTS

Cure strategies are confounded by basic reservoir biology

y

1. C. G. Kinzig et al., Science 383, 763 (2024). 2. N. Arnoult, J. Karlseder, Nat. Struct. Mol. Biol. 22, 859 (2015). 3. G.-L. Yu, E. H. Blackburn, Cell 67, 823 (1991). 4. K. M. Kramer, J. E. Haber, Genes Dev. 7, 2345 (1993). 5. S. J. Diede, D. E. Gottschling, Curr. Biol. 11, 1336 (2001). 6. C. N. Sprung, G. E. Reynolds, M. Jasin, J. P. Murnane, Proc. Natl. Acad. Sci. U.S.A. 96, 6781 (1999). 7. E. L. Denchi, T. de Lange, Nature 448, 1068 (2007). 8. P. Margalef et al., Cell 172, 439 (2018). 9. S. Kim et al., Semin. Cell Dev. Biol. 54, 177 (2016). 10. F. Müller et al., Cell 67, 815 (1991). 11. R. Stanyon et al., Chromosome Res. 16, 17 (2008). 12. G. B. Morin, Nature 353, 454 (1991). 13. J. Nandakumar et al., Nature 492, 285 (2012). 14. R. Sandhu, M. Sharma, D. Wei, L. Xu, Proc. Natl. Acad. Sci. U.S.A. 118, e2024889118 (2021).

HIV cure: The daunting scale of the problem

g

RE F ER E NC ES AND NOTES

MEDICINE

p

larly important in cancer, where telomerase is reactivated in 90% of cases. Chromosome rearrangements can drive cancer by creating oncogenes or increasing their copy number or by eliminating tumor-suppressor genes. However, high instability is not sustainable for cell growth. Kinzig et al. propose that the healing of DSBs by telomerase may allow cancer cells to cope with genome instability. Some species have adopted telomere capping of programmed DSBs as a developmental mechanism. In ciliated protozoa, a few very large germinal chromosomes are broken, giving rise to a high number of small somatic chromosomes, which are stabilized by neotelomere formation (3). Germline genes in the nematode Ascaris are eliminated during early development through programmed DSBs and telomere healing (10). From an evolutionary standpoint, it is possible that neotelomere formation has contributed to speciation. Indeed, the great apes have many instances of distal deletions or inversions (11), which may have necessitated the action of telomerase on nontelomeric substrates. It is noteworthy how effective human cells are at suppressing this treacherous activity of telomerase, even in contexts where ATR is suppressed and telomerase activity is high. A potential simple explanation could be the lack of telomeric sequences and TPP1 at DSBs. Telomerase RNA typically pairs with five nucleotides at the telomeric end, and although it can still add repeats with as few as two nucleotides of complementarity (1, 12), these are likely inefficient primers. Similarly, TPP1 aids in telomerase recruitment to telomeres (13). Telomerase is capable of extension without TPP1, but this activity is minimal (14). Future research may reveal whether these factors, combined with the protective role of ATR, are sufficient to prevent telomerase action at DSBs or whether additional safeguarding mechanisms exist. j

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

host cells are eliminated to the same extent as uninfected host cells. Assays for infected cells were negative, and analytical treatment interruptions (ATIs) were carried out. Initially, viremia remained undetectable, but sudden exponential rebound was observed in these three recipients at 2.8, 7.4, and 7.9 months after ATI. During these time intervals, there was little to prevent viral replication, given the absence of ART and the elimination of host lymphocytes by GVHD. The simplest explanation for the delayed rebound is persistence of a small number of latently infected host cells, one of which eventually became activated. The rebound viruses were monophyletic, which is consistent with the reactivation of a single latently infected cell. These cases illustrate the fundamental problem that precludes cure: HIV-1 can persist

in reservoir size caused by an intervention. If ART is stopped, rapid viral rebound can occur when a single cell from the reservoir becomes activated. During ART, multiple recently activated cells from the reservoir are producing virus simultaneously, giving rise to a trace level of residual viremia that can be detected with ultrasensitive assays (typically ~1 copy of HIV-1 RNA per ml of plasma) (2). The time to rebound after an ATI is the time required for replication-competent viruses released by recently activated CD4+ T cells to replicate exponentially, which increases viremia from this trace level to the typical pre-ART set point levels of 10,000 to 100,000 copies/ml. This normally takes only a few weeks after ART interruption. Long delays in rebound occur only if the reservoir has been reduced to such an extent that activation leading to

CRISPR

Excision or silencing of latent provirus

Total reservoir cells

Transcriptionally Transcriptionally silent provirus active provirus

106 104 102

10

0

No reduction in latent reservoir

102

105-fold reduction in latent reservoir

101

ART

100 5 10 15 20 0 Time after ART interruption (years)

virus production is an isolated event separated by a long time interval from the next relevant activation event. Long delays involve a switch from a growth-limited regime to an activation-limited regime (7). This requires a multilog reduction in the reservoir. Modeling studies suggest that, on average, a 2000-fold reduction is required to delay rebound by 1 year (7). However, the HSCT cases suggest that even larger reductions may be required to achieve substantial delays. To date, no intervention other than HSCT has convincingly reduced the reservoir, and the basic biology of the reservoir confounds attempts to produce multilog reductions. HIV-1 latency is a consequence of normal immune system physiology and is thus more difficult to target than abnormal processes such as the uncontrolled proliferation of tumor cells. The absence of viral gene expression in latently infected cells adds an enormous additional complication. Although large reductions in tumor burden may facilitate immune

science.org SCIENCE

,

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

103

y

704

104

y g

for long periods in a latent state, and then a single cell can give rise to exponential rebound. Reservoir reductions delay rebound, but only multilog reductions in HIV-1 infected cells produce clinically meaningful delays (see the figure). The same conclusions follow from cases in which ART was initiated immediately after infection, resulting in very small reservoirs (6), and from mathematical modeling of rebound (7). These considerations prompt a reassessment of the likelihood that cure can be achieved by reservoir reduction. For this purpose, reservoir size must be considered on a logarithmic scale, the way oncologists view tumor burden. Cancer chemotherapy can reduce tumor burden by many logs, affording remission, the length of which is generally proportional to the log reduction caused by chemotherapy. For HIV-1, however, there is a more complex and less favorable relationship between the time to viral rebound after treatment interruption and the log reduction

105

y

LRA

108

Rebound after ART interruption Plasma HIV RNA (copies/ml)

The latent reservoir

Targeting the latent reservoir Resting memory Cytotoxic T cell CD4+ T cell

g

Strategies to reduce the latent reservoir include latency reversing agents (LRAs) to allow killing of infected cells and CRISPR editing of the provirus. However, substantial latent reservoir reductions by these approaches have not been demonstrated, and even 105-fold reductions may have minimal impact on viral rebound. For example, only short delays in rebound of 2 to 8 months after antiretroviral therapy (ART) interruption have been observed in hematopoietic stem cell transplantation cases in which host immune cells were reduced by 105-fold (4, 5).

p

Reservoir reduction and viral rebound

clearance of residual tumor cells, the same is not true for the latent reservoir of HIV-1 infected cells. The immune system cannot detect the infected cells unless latency is reversed. Furthermore, although overcoming exhaustion with immune checkpoint–blocking antibodies has considerably improved cancer therapy, it will not allow recognition of latently infected cells. In addition, the normal biology of resting CD4+ T cells gives them a much wider tissue distribution than metastatic cancer cells. To promote encounters between rare antigen-specific T cells and dendritic cells presenting the relevant antigen, resting T cells continually recirculate through the secondary lymphoid organs and are thus widely distributed throughout the body (8). Finally, cells harboring latent HIV-1 can undergo extensive proliferation as part of the normal T cell response to antigen (9), and recent studies demonstrate that owing to infected cell proliferation, the frequency of latently infected cells may be increasing slowly in people who have been on ART for more than two decades (10). Thus, the reservoir consists of widely distributed mobile cells that are invisible to the immune system and are capable of proliferation. The most commonly discussed cure strategy envisions overcoming the problem of latency by inducing viral gene expression with latency reversing agents (LRAs) (11) to allow the killing of infected cells by immune effector cells. The most difficult aspect of this “shock-and-kill” strategy is the identification of effective LRAs. Again, this problem is related to T cell biology. HIV-1 evolved to replicate in activated CD4+ T cells that have a transcriptional environment that is profoundly different from that of resting cells harboring latent proviruses. Latency can be reversed by global T cell activation, but this causes unacceptable toxicity. Not surprisingly, finding a way to turn on a single activation-dependent promoter (the HIV-1 long terminal repeat) without cellular activation has proven difficult. The LRAs with the greatest activity in in vitro or ex vivo studies activate pathways downstream of the T cell receptor, but concerns about toxicity remain. Importantly, although regimens with substantial toxicity can be justified for cancer treatment, the same is not true for HIV-1 cure, given the available alternative of lifelong ART. Another major problem is that poorly understood stochastic features limit proviral induction even when the T cell is fully activated. Only a minority of replication-competent proviruses are induced by a single round of global T cell activation; the remainder proliferate and retain the ability to produce virus after subsequent activation events (12). For LRAs that do not induce T cell activation, the fraction of proviruses induced may be even lower, mak-

By Jillian R. Jaycox1,2, Yile Dai1, Aaron M. Ring2

I

Department of Immunobiology, Yale School of Medicine, New Haven, CT, USA. 2Division of Translational Science and Therapeutics, Fred Hutchinson Cancer Center, Seattle, WA, USA. Email: [email protected]

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

705

,

1

y

nvestigating the causes of individual variation in health outcomes has led to transformative insights into human biology and advances in nearly every branch of medicine. Historically, emphasis has been placed on how genetic factors contribute to phenotypic variation within populations. However, an emerging concept is that self-reactive antibodies (autoantibodies) represent a critical yet largely underexplored factor that influences human health and disease. Investigating autoantibodies and their protective as well as pathological roles in disease may unlock new treatment paradigms, much like the prior study of genetics. Generated by the humoral immune system, antibodies are capable of specifically binding to virtually any biomolecule target (broadly termed “antigens”) (1). Although the primary function of antibodies is to provide adaptive immunity against pathogens, invariably some antibodies arise that bind to self-antigens. These autoantibodies elicit a wide range of biological effects, including altering the activity of their targets and immunomodulation (see the figure). Every person carries a distinct array of autoantibodies—an “autoantibody reactome”— offering a potential avenue for trait diversity that mirrors the way genetic differences influence phenotypes. Autoantibodies are usually known for their etiologic role in mediating autoimmune diseases. Canonically, autoantibodies can drive pathological inflammation within nearly any tissue, notably affecting the skin, joints, muscles, and central nervous system as well as organs such as the thyroid and pancreas (2). Similarly, autoantibodies can trigger distinctive syndromes marked by highly specific biological effects, akin to the distinct impact observed with Mendelian single-gene mutations, because they interfere with essential pathways in the body. Notable examples include myasthenia gra-

vis, a neuromuscular disease caused by autoantibodies that inhibit the acetylcholine receptor, and the hyperthyroidism in Grave’s disease that is driven by autoantibodies that activate the thyrotropin receptor. Less appreciated are the more subtle phenotypic effects of autoantibodies that are disease-modifying or even clinically silent until their activity is unmasked in states of stress. A key example of this phenomenon was revealed during the COVID-19 pandemic in which type I interferon (IFN-I)– neutralizing autoantibodies were found to confer up to 200-fold increased risk of death from COVID-19 (3). Although they are apparently clinically silent in most circumstances, the prevalence of IFN-I autoantibodies sharply increases with age, peaking at ~4% of individuals over 70 years old (4). Consistent with their substantial clinical influence and overall frequency, it is estimated that 20% of all COVID-19 deaths are associated with the presence of IFN-I autoantibodies (4). These findings underscore the ability of autoantibodies to reveal both key biological insights [for example, the critical importance of IFN-I in host immunity to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)] and the profound impact that autoantibodies may exert at a population level.   However, autoantibodies are not uniformly deleterious, and in some instances, they may provide protective effects that ameliorate or prevent disease. IFN-I autoantibodies are again instructive. Systemic lupus erythematosus (SLE) is an autoimmune disease that is characterized by elevated IFN-I signaling in >50% of patients. Intriguingly, ~5% of SLE patients have autoantibodies that neutralize IFN-I signaling (5). In contrast to COVID-19, these autoantibodies are associated with substantially lower disease activity, presumably by attenuating pathological IFN-I pathway function (5).  This counterintuitive observation emphasizes the dualistic nature of autoantibodies, demonstrating their ability to confer protective benefits in the very diseases they are typically implicated in causing. The protective effects of autoantibodies are apparent across numerous diseases. For

y g

SCIENCE science.org

Autoantibodies influence a wide range of conditions beyond autoimmune diseases

y

1. B. L. Jilek et al., Nat. Med. 18, 446 (2012). 2. J. D. Siliciano, R. F. Siliciano, Annu. Rev. Pathol. 17, 271 (2022). 3. G. Hütter et al., N. Engl. J. Med. 360, 692 (2009). 4. T. J. Henrich et al., Ann. Intern. Med. 161, 319 (2014). 5. N. W. Cummins et al., PLOS Med. 14, e1002461 (2017). 6. K. Luzuriaga et al., N. Engl. J. Med. 372, 786 (2015). 7. A. L. Hill et al., Proc. Natl. Acad. Sci. U.S.A. 111, 13475 (2014). 8. M. Lee et al., Blood 120, 1432 (2012). 9. F. R. Simonetti et al., J. Clin. Invest. 131, e145254 (2021). 10. N. F. McMyn et al., J. Clin. Invest. 133, e171554 (2023). 11. N. M. Archin et al., Nature 487, 482 (2012). 12. K. J. Kwon et al., Sci. Transl. Med. 12, eaax6795 (2020). 13. D. R. Collins et al., Nat. Rev. Immunol. 20, 471 (2020). 14. J. D. Gunst et al., Nat. Med. 28, 2424 (2022). 15. L. N. Bertagnolli et al., Proc. Natl. Acad. Sci. U.S.A. 117, 32066 (2020). 10.1126/science.adk1831

Decoding the autoantibody reactome

g

RE F ER E NC ES AND NOTES

IMMUNOLOGY

p

ing it harder to achieve the requisite multilog reductions. Even if a way can be found to safely induce all replication-competent proviruses, it will be necessary to ensure that all of the infected cells are killed, a problem confounded by immune exhaustion and viral escape mutations. Other strategies that focus on the excision of latent proviruses or permanently silencing them face similar problems related to the scale of the effect needed to produce a cure. An alternative approach to reservoir reduction, which is known as “functional cure,” involves the induction of immune responses that will keep viral replication in check so that viremia remains undetectable with clinical assays. In this situation, disease progression and virus transmission are unlikely even though the reservoir persists. Precedent for immune control comes from the rare individuals (1/300) who spontaneously control HIV-1 replication without ART (13). Extensive studies indicate that control is most likely mediated by HIV-1–specific cytolytic T lymphocytes. Unfortunately, it has not yet been possible to induce this degree of control in most people with HIV-1 using therapeutic vaccines. Broadly neutralizing antibodies (bNAbs) to the HIV-1 envelope protein can delay rebound and contribute to lower levels of postrebound viremia in recipients with sensitive virus, although some of the observed effects may be related to direct neutralization by trace residual levels of the bNAbs (14). It is also important to note that the autologous neutralizing antibody response can prevent outgrowth of a substantial but variable fraction of HIV-1 variants in the latent reservoir (15). Together, these results suggest that preventing viral rebound and thus achieving a functional cure may depend on enhancing virus-specific humoral immunity to cover all reservoir variants. Although shock-and-kill strategies are unlikely to produce the degree of reservoir reduction required for cure, reservoir reduction will facilitate immune control, and thus these strategies may be a useful adjunct to immune-based cure efforts. j

I NS I GHTS

PERSPECTIVES MOLECULAR BIOLOGY

Telomerase misbehaves after a breakup Suppressing telomerase action at broken DNA preserves genome integrity By Nausica Arnoult1 and Thomas R. Cech2

y g y ,

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

y

At natural chromosome ends, single-stranded telomeric DNA repeats ( ) provide a primer for extension by the RNA-protein enzyme telomerase (top right). Most double-strand breaks are repaired, but neotelomere formation can occur with a potentially disastrous loss of distal sequences. How telomerase initiates at nontelomeric DNA ( ) is unknown, but once initiated, the reaction proceeds normally (bottom right). Normal extension by telomerase

Telomere

Extension by telomerase

TPP1 G Telomerase Single-stranded POT1 telomeric DNA G AG G G T TA G C A A U CCCA A UC Telomerase RNA

DSB Repair

702

g

Telomerase action at natural telomeres and double-strand breaks

1

Department of Molecular, Cellular and Developmental Biology, University of Colorado Boulder, Boulder, CO, USA. 2 Department of Biochemistry and BioFrontiers Institute, University of Colorado Boulder, Boulder, CO, USA. Email: [email protected]; [email protected]

p

C

ells must distinguish between broken chromosome ends and natural ends, or telomeres. Broken chromosomes require repair, but telomeres need to be protected from repair because joining two telomeres would result in disastrous chromosome fusions. Telomerase is the enzyme that maintains chromosome ends by adding repeated DNA sequences. If directed to a DNA break, telomerase would erroneously create a new telomere, precipitating the loss of essential genetic material distal from the break site. Although extensive research has elucidated how telomeres evade DNA repair pathways, the question of telomerase’s misguided activity at DNA breaks in human cells has remained enigmatic. On page 763 of this issue, Kinzig et al. (1) report that telomerase can cause harmful neotelomere formation in human cells. They also show that such events are kept to a minimum by the naturally low prevalence of telomerase and the intervention of the ataxia telangiectasia and Rad3-related (ATR) cellular signaling pathway. Telomeres consist of repetitive DNA sequences protected by a protein complex called shelterin. In the absence of telomerase, telomeres shorten with each cell division, eventually reaching a point where they no longer suppress DNA damage signaling, which leads to cellular senescence (2). In the germ line, stem cells, and most cancers, telomerase counters this shortening by adding telomeric repeats to chromosome ends. Telomerase is a ribonucleoprotein complex with reverse transcriptase activity that uses its intrinsic RNA template to elongate telomeres. It is recruited to telomeres by the shelterin component TPP1 and binds to the

telomere’s single-stranded end through base pairing with its RNA template (see the figure). Unlike telomeres, a DNA double-strand break (DSB) lacks both TPP1 and telomeric DNA sequences that could prime telomerase. Thus, it has not been clear whether telomerase can add neotelomeres at DNA breaks in humans. Indications of this possibility do exist. Patients with deleted distal chromosome ends have telomeric sequences at the DNA breakpoint. Furthermore, neotelomere formation occurs in other organisms, such as ciliated protozoa (3), and can be promoted by induced DSBs in yeast and mice (4–6). To address the frequency and mechanism of neotelomere formation in human cells, Kinzig et al. engineered a cell line in which they could produce a DSB by CRISPR-Cas9 or the I-SceI endonuclease. They found that telomerase can add telomeric repeats to DSBs, which are then converted to functional telomeres. Neither the ataxia telangiectasia mutated (ATM) pathway—typically activated in response to DSBs—nor common DNA repair factors prevented this inappro-

priate action. Rather, Kinzig et al. identified ATR, a DNA damage signaling protein that coordinates the response to replication stress and whose action at natural telomeres is suppressed by the shelterin protein protection of telomeres 1 (POT1) (7). Although this finding is unexpected, one should consider that in the absence of an insult, such as ionizing radiation, most DSBs are the result of replication stress. Stalled replication forks can reverse, resulting in a structure called a chicken foot. Telomerase acts on such structures when they arise at telomeres (8). Future research is needed to elucidate the prevalence of neotelomere formation at replication-induced DSBs and at chicken foot structures as well as the function of ATR in that context. There are contexts in which DSBs are intentionally created. During meiosis, the enzyme Spo11 generates DSBs that drive recombination between homologous chromosomes, ensuring correct alignment and contributing to genetic diversity (9). Given that meiotic cells exhibit high telomerase activity and can generate more than 200 DSBs, it is plausible that some may be converted to neotelomeres. The resulting chromosome truncations would often be incompatible with embryonic development. Thus, neotelomere formation at meiotic DSBs could cause early miscarriage. Alternatively, neotelomeres formed close to chromosome ends, resulting only in the loss of nonvital genetic material, would lead to germinal structural variants with terminal deletions. Indeed, Kinzig et al. culled published DNA sequences and found 36 patients harboring terminal chromosome deletions associated with neotelomere formation. Neotelomere formation may be particu-

Chromosome

Neotelomere formation

DSB extension by telomerase DSB point Nontelomeric Extension by Telomerase G DNA telomerase G TTAG C A A U CCCA A UC Loss of distal sequences DSB, double-strand break; POT1, protection of telomeres 1.

science.org SCIENCE

SCIENCE science.org

1

Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, MD, USA. 2Howard Hughes Medical Institute, Johns Hopkins University School of Medicine, Baltimore, MD, USA. Email: [email protected]

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

703

,

10.1126/science.adn7791

C

ombination antiretroviral therapy (ART) for HIV-1 infection reduces plasma virus levels such that clinical assays for viremia are negative. Because of unexpected cooperativity in dose-response relationships and drug synergies, ART completely inhibits new infection events, halting viral evolution and disease progression (1). However, ART is not curative owing to a persistent latent reservoir of HIV-1, the half-life of which is long enough to guarantee lifetime persistence, even with optimal ART (2). Within weeks, interruption of ART leads to rebound viremia, with subsequent disease progression toward fatal immunodeficiency. Therefore, lifelong ART is necessary. People who acquire HIV-1 during their 20s face 50 to 60 years of continuous treatment with multiple antiretroviral drugs. Children infected perinatally face a lifetime of treatment. Thus, there is a pressing need for a cure, which, in a practical sense, amounts to preventing viral rebound after ART interruption. Present efforts focus on eliminating the reservoir. The latent reservoir consists primarily of resting memory CD4+ T cells that carry transcriptionally silent proviruses, which are unaffected by ART or immune responses (2). Although other reservoirs may exist in the central nervous system and elsewhere, the reservoir in resting CD4+ T cells is readily demonstrable in blood samples from all people with HIV and provides simple, coherent explanations for the hallmark features of HIV-1 persistence: stability for decades, wide distribution in tissues, latency, and clonality (2). The establishment of the latent reservoir must be understood in the context of a fundamental transition that lymphocytes make between resting and activated states. Encounter with antigen drives the transformation of resting CD4+ T cells into lymphoblasts, which proliferate and differentiate into effector cells. This blast transformation involves substantial increases in size and metabolic activity, extensive changes in gene expression, and entry into the cell cycle. At the conclusion of

an immune response, most activated cells die, but a subset survive and revert back to a resting state as long-lived memory cells. The HIV-1 promoter contains conserved binding sites for the activation-dependent host transcription factors nuclear factor of activated T cells (NFAT) and nuclear factor kB (NF-kB), demonstrating that HIV-1 and related viruses evolved to replicate in activated CD4+ T cells, which are biosynthetically capable of producing virions. Most infected lymphoblasts die quickly but not before producing enough virus to sustain the infection. One model suggests that a state of latent infection may be established when activated cells are infected during the transition back to a resting memory state that is nonpermissive for viral gene expression, which is due in part to sequestration of NFAT and NF-kB (2). The cells then persist through normal mechanisms that maintain immunologic memory. In this model, HIV-1 latency is a consequence of normal immune physiology. Subsequent cellular activation can induce latent proviruses, leading to rebound viremia if ART is interrupted. Cure has been achieved in a few people who received hematopoietic stem cell transplantation (HSCT) from donors homozygous for a deletion in the HIV-1 co-receptor C-C chemokine receptor type 5 (CCR5) (3). HSCT has substantial mortality and can be carried out only when required to treat an otherwise fatal malignancy. In reported HIV-1 cure cases, host origin immune cells, including latently infected cells, were eliminated in a virus-independent fashion by the conditioning regimen and graft-versus-host disease (GVHD), which together can cause near complete replacement of host white blood cells with HIV-1–resistant donor cells. Virus released by any surviving infected host cells cannot replicate in the HIV-1–resistant, donor-derived cells that make up the reconstituted immune system. Three interesting but unsuccessful cases of HSCT in people with HIV-1 infection used donors who were wild type for CCR5 (4, 5). ART was continued for several years after transplantation to prevent HIV-1 infection of donor cells. After reconstitution, host immune cells were present at levels of ≤0.001%. This is a 100,000-fold reduction in infected cells, assuming that latently infected

y

N.A. is supported by the National Institutes of Health (R35GM143108 and R01CA266100), the V Foundation for Cancer Research, and the Boettcher Foundation. T.R.C. is an investigator of the Howard Hughes Medical Institute.

By Janet D. Siliciano1 and Robert F. Siliciano1,2

y g

AC KN OWL EDG M ENTS

Cure strategies are confounded by basic reservoir biology

y

1. C. G. Kinzig et al., Science 383, 763 (2024). 2. N. Arnoult, J. Karlseder, Nat. Struct. Mol. Biol. 22, 859 (2015). 3. G.-L. Yu, E. H. Blackburn, Cell 67, 823 (1991). 4. K. M. Kramer, J. E. Haber, Genes Dev. 7, 2345 (1993). 5. S. J. Diede, D. E. Gottschling, Curr. Biol. 11, 1336 (2001). 6. C. N. Sprung, G. E. Reynolds, M. Jasin, J. P. Murnane, Proc. Natl. Acad. Sci. U.S.A. 96, 6781 (1999). 7. E. L. Denchi, T. de Lange, Nature 448, 1068 (2007). 8. P. Margalef et al., Cell 172, 439 (2018). 9. S. Kim et al., Semin. Cell Dev. Biol. 54, 177 (2016). 10. F. Müller et al., Cell 67, 815 (1991). 11. R. Stanyon et al., Chromosome Res. 16, 17 (2008). 12. G. B. Morin, Nature 353, 454 (1991). 13. J. Nandakumar et al., Nature 492, 285 (2012). 14. R. Sandhu, M. Sharma, D. Wei, L. Xu, Proc. Natl. Acad. Sci. U.S.A. 118, e2024889118 (2021).

HIV cure: The daunting scale of the problem

g

RE F ER E NC ES AND NOTES

MEDICINE

p

larly important in cancer, where telomerase is reactivated in 90% of cases. Chromosome rearrangements can drive cancer by creating oncogenes or increasing their copy number or by eliminating tumor-suppressor genes. However, high instability is not sustainable for cell growth. Kinzig et al. propose that the healing of DSBs by telomerase may allow cancer cells to cope with genome instability. Some species have adopted telomere capping of programmed DSBs as a developmental mechanism. In ciliated protozoa, a few very large germinal chromosomes are broken, giving rise to a high number of small somatic chromosomes, which are stabilized by neotelomere formation (3). Germline genes in the nematode Ascaris are eliminated during early development through programmed DSBs and telomere healing (10). From an evolutionary standpoint, it is possible that neotelomere formation has contributed to speciation. Indeed, the great apes have many instances of distal deletions or inversions (11), which may have necessitated the action of telomerase on nontelomeric substrates. It is noteworthy how effective human cells are at suppressing this treacherous activity of telomerase, even in contexts where ATR is suppressed and telomerase activity is high. A potential simple explanation could be the lack of telomeric sequences and TPP1 at DSBs. Telomerase RNA typically pairs with five nucleotides at the telomeric end, and although it can still add repeats with as few as two nucleotides of complementarity (1, 12), these are likely inefficient primers. Similarly, TPP1 aids in telomerase recruitment to telomeres (13). Telomerase is capable of extension without TPP1, but this activity is minimal (14). Future research may reveal whether these factors, combined with the protective role of ATR, are sufficient to prevent telomerase action at DSBs or whether additional safeguarding mechanisms exist. j

NE WS

FEATURES

A RIVER IN FLUX

Extreme flooding and droughts may be the new norm for the Amazon, challenging its people and ecosystems By Daniel Grossman, in Manaus, Brazil; Photography by Dado Galdieri

p g y

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

,

692

y

Reporting for this story was supported by the Pulitzer Center, the Pendleton Mazer Family Fund, and Abby Rockefeller and Lee Halprin.

y g

J

ochen Schöngart darts back and forth along an escarpment just above the Amazon River, a short water taxi ride from downtown Manaus, Brazil. It’s still early this October morning in 2023, but it’s already hot and his face is beaded with sweat. “Look, there’s a piece of ceramic!” he says, nodding to a worn shard lodged between boulders, likely a relic of an earlier civilization. It’s not the only one. Schöngart, a forest scientist at the National Institute of Amazon Research (INPA), stoops and stares at the bedrock at his feet. Well below the river’s normal level for this time of year, the rock bears a gallery of life-size faces, perhaps carved during a megadrought 1000 years ago. Now, they have been exposed again by a new drought, the worst in the region’s modern history. In the previous 4 months, only a few millimeters of rain have fallen in this city of 2 million at the confluence of the Negro and Amazon rivers. Normally it gets close to a half a meter during the same period. The Amazon sank steadily beginning in June, as it does most years during the dry season. But by mid-October, the port’s river gauge reached the lowest level observed since the record began in 1902. Freighters coming up from the Atlantic Ocean—the city’s primary supply line—were blocked by shoals. Factories furloughed workers. Making matters worse, the drought coincided with a series of weekslong heat waves. In September and October, withering conditions persisted across the Amazon, and temperatures here peaked at 39∞C, 6∞C above normal. Des-

A victim of heat and drought, a freshwater dolphin is ferried to a makeshift necropsy suite near the city of Tefé, Brazil. scienc sci en nce.o ee.o orrg g SCIENCE SC SCI C ENC CI EN C E science.org

p g

iccated jungle set ablaze by farmers enveloped the city in choking smoke. Then, in the season’s most freakish episode, a sandstorm blotted out the Sun. Drought and heat are only half of the story of the changes unfolding in the heart of the world’s largest rainforest. Schöngart and collaborators’ research on the river here has shown that for decades, while dry-season low water has been plummeting, rainy-season high water has been rising. The city has experienced frequent major flooding in recent years because of heavy rains across much of the Amazon Basin, forcing the officials to erect temporary wooden walkways above streets of the historic waterfront. Schöngart and other researchers expect such changes to intensify as global climate warms. The current drought provided a grim preview, killing river dolphins and fish, and threatening livelihoods for communities along the river. If the combination of higher highs and lower lows becomes the new norm, the ramifications could extend throughout the Amazon Basin and even beyond, threatening the very existence of the forest—which harbors much of the planet’s biodiversity, has a far-reaching influence over regional and global climate, and sustains millions of people. “We are undergoing massive changes in the hydrological cycle” of the Amazon Basin, Schöngart says. The question now, he says, is whether its ecosystems and people can adapt. ABOVE THE GROWL of an outboard engine, Ayan Fleis-

PHOTO: CREDIT GOES HERE AS SHOWN; CREDIT GOES HERE AS SHOWN P PHOTO PHO HOTO HO OTO O O: CRE CR RED DIT DI TG O SH OES ERE A SS SHO OWN; WN; C REDIT REDI R ED ED EDI DIT GOE G GO ES S HER H HE RE R EA SS HO HOWN H OWN O N

,

693

y

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

y g

SCIENCE science.org S C ENC SCI E N C E sc ssci cien ci eenc nce nc cee. e.o ..o org g

y

chmann hollers orders to the boatman of a skiff plying the waters of Lake Tefé in October 2023. Fleischmann, a hydrologist at the Mamirauá Institute for Sustainable Development in Tefé, a lakeside city of 70,000, is monitoring the extreme conditions in the central Amazon, 600 kilometers upstream from Manaus, in one of the regions hit hardest by the drought. We coast to what looks like a remnant of a fence—a bleached wooden post jutting obliquely above the surface. Few would suspect it marks a temperature monitoring station. That’s intentional, Fleischmann says. “We put it on this kind of pole so that no one thinks it’s important,” he says, grinning. He snags the unprepossessing post and fishes up a wristwatch-size data logger tied to a length of twine. The average water temperature for this time of year is 30°C, but the sensor recently recorded a high of 39.1°C. Fleischmann says shallower parts of the lake may have reached 41°C on that same day in late September. The air that month was hot, too, about 1.5∞C above average. Meanwhile, drought has shrunk the lake; by late October, Lake Tefé’s surface had sunk 6.5 meters below the average annual low. Together the low water and high air temperature took a deadly toll on wildlife. Within sight of Fleischmann’s temperature logger, four crouching figures cruise along Lake Tefé’s shore in a flatbottomed jon boat, their faces wrapped in surgical masks. They beach the boat on a mudflat and lug ashore a bundle wrapped in a tarp. As they unfurl it in an open-sided tent that serves as a makeshift surgical suite, a dead dolphin thuds onto a metal dissection table. The stench of rotted flesh wafts up from the peeling and discolored corpse. Half a dozen people lounging in the shade leap to their feet and pull on Tyvek suits, rubber gloves, and masks. One of them, Mariana Lobato, a researcher at Mamirauá, says

NE WS | F E AT U R E S

A future of extremes Wetter wet seasons and drier dry seasons may be in store for the Amazon Basin in the final 30 years of this century. Computer modeling suggests annual maximum flows will increase up to 50% in the Amazon River’s headwaters in the Andes to the west; meanwhile, annual minimum flows will decline by 20% to 50% throughout most of the region. Predicted change 50%

,

science.org SCIENCE

y

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

y g

694

y

of the Amazon Basin’s main rivers fell dramatically and, in places, dried up completely in September and October, says Jose Marengo, a climate scientist at Brazil’s National Center for Monitoring and Early Warning of Natural Disasters. Heat blanketed 60% of the rainforest, with temperatures 2°C to 5°C above normal highs.

g

MANAUS AND TEFÉ are not anomalies. Most

p

Dry season flow

50%

Amazon Basin

the torpedo-shaped mammal is a female tucuxi (Sotalia fluviatilis), the smaller of two freshwater dolphin species that live here. It has probably been decaying for a day or two, she says. Lobato grabs a scalpel and cuts into its abdomen. “We’re measuring the fat,” she says, pointing to a creamcolored layer as thick as her thumb. The ample fat indicates that this tucuxi did not die of starvation. She and her colleagues slice deeper, cutting out bits of lung, brain, and other organs for later study. The bodies began appearing in late September, recalls oceanographer Miriam Marmontel, who leads Mamirauá’s marine mammal group. Responding to reports from boaters, Mamirauá staff hauled 19 carcasses from the lake. In the days that followed, Marmontel says, her workers discovered dolphins “in agony, circling around themselves and not being able to dive.” They soon died. On 28 September, the day Fleischmann’s sensor logged the record high heat, 70 more corpses were discovered. All told they collected more than 200 dead dolphins, including both resident species—about 15% of the lake’s population—in the span of a few weeks. “It was something we could never expect. It really hit hard,” Marmontel says. The record water temperatures were an obvious suspect for the calamity, but Marmontel’s team conducted more than 100 necropsies to eliminate other explanations. They found no evidence of infectious disease or a toxin. “We think that climate change is the major culprit,” she says. That worries her, she says, because it’s the only possibility with no local solution. On the same hot September day, shimmering patches of dead fish coated large swaths of Lake Tefé’s surface—likely another casualty of the heat. A research team at INPA has shown that Amazonian fish can’t withstand temperatures higher than 35∞C to 37∞C. “Anything above that is a no-go,” says Alexandre Pucci Hercos, leader of the fish biology group at Mamirauá. “An extra degree may not seem like much, but it’s a big difference.” Today, Lake Tefé has hundreds of species of fish. But if, as anticipated, the harmful conditions become more frequent, Hercos says, “some species will migrate while others will become extinct.”

y ,

PHOTO: DADO GALDIERI/HILAEA MEDIA

y g

695

y

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

g

SCIENCE science.org

p

international collaboration of climate sciencurrent linked to climate change that carThe immediate causes, climate scientists tists, found that climate change has multiries heat from the Indian Ocean around say, lie thousands of kilometers away, in plied the likelihood of precipitation as low the Cape of Good Hope into the South Atpools of anomalously warm water in the Paas that seen in the Amazon Basin in 2023 lantic. The resulting temperature differencific and Atlantic oceans. by a factor of 10. Its impact was even larger tial alters continental-scale wind patterns, One well-established influence on temon the likelihood of an agricultural drought, perhaps funneling extra moisture from the perature and precipitation in the Amazon in which low rainfall and high temperaocean basins into the Amazon. is the El Niño-Southern Oscillation, a cyclitures combine to parch the soil, stressing Whatever their precise cause, the floods cal fluctuation in surface temperatures in crops and wild vegetation alike. The modare taking a toll. In June 2021, the river the central and eastern equatorial Pacific eling concluded that climate change has reached a historic high in Manaus, flooding Ocean. In its warm phase, El Niño, these increased the likelihood of an agricultural tens of thousands of homes and prompting regions are hotter than normal, creating drought as deep as that of 2023 by a factor authorities to build 9 kilometers of boardlow pressure zones in the overlying atmoof 30. Put more plainly, the drought would walks above the streets. Villagers near Lake sphere that steer the moist air responsible have been unlikely had climate change not Tefé say the stilts their houses sit on aren’t for rain off its normal pathways—and, ofwarmed the planet. Marengo isn’t ready always high enough anymore to keep them ten, away from the Amazon. The El Niño to say the drought was caused by climate dry when it floods. During unusual wet that began in 2023 appeared unusually seasons, some families early in the year and is install temporary floors, predicted to be one of 1 meter or more above the strongest on record the original, and enter by the time it ends. through windows rather Another player in than the front door. last year’s drought was Forests, too, may be unusual warmth in the at risk. Some trees in Atlantic just north of the low-lying floodplains the equator, which, like of the Amazon and its El Niño, drives moist tributaries are adapted air off course and has to up to 10 months of been linked to previous inundation, far more Amazon droughts. The than normal trees can Amazon Assessment withstand. But even Report 2021, an encythese trees have their clopedic appraisal of limits. In a 2020 study, the people, ecology, and Schöngart and colclimate of the Amazon, leagues documented a notes that of 15 megapattern of tree mortality droughts recorded becoinciding with years of tween 1906 and 2021, high water in Jaú Nasix coincided with El tional Park in the central Niño and three with Amazon, concluding that warm tropical North Submerged trees in the Uatumã River flood plain, below a hydroelectric plant, provide a laboratory for more intense flooding in Atlantic waters. Addresearchers studying how similar forests will respond to extra flooding caused by climate change. recent decades is already ing the 2023 drought, damaging these trees. all four of this century’s change, but he sees it as a “sample” of what’s In the future, excessively high water droughts occurred when both ocean regions had warmed. to come. “It seems like we are looking now could also threaten some wildlife, says Climate change is an obvious suspect in the at some of those things that could happen Rafael Rabelo, research coordinator at observed changes in the drought-fostering in the next decades.” Mamirauá. He’s particularly concerned ocean conditions, although the mechanisms Climate change might also be a culprit about the black squirrel monkey (Saimiri underlying its role are not clear. Some rein the extreme rainfall that has been occurvanzolinii), which only lives near Tefé in search suggests global warming might be ring during wet seasons. Of the 18 “flood 800 square kilometers of the seasonally increasing the strength and frequency of emergencies” declared in Manaus since inundated jungle called igapó. Mamirauá El Niños (Science, 2 February, p. 472). And 1902, half, including the four largest, have is conducting a modeling study to explore the broader increase in ocean temperataken place since 2000. “Recent floods not whether the higher flood waters expected tures caused by global warming could also only occur more often but also have become in the future will damage the igapó, which be contributing, providing a backdrop for more severe,” Schöngart and colleagues could put the primates themselves at risk. anomalous areas of hotter water in the Atwrote in a 2018 paper in Science Advances. lantic and Pacific. Global sea surface temIn the same paper, Schöngart and his “IF THE DROUGHT continues, we are worried peratures have been rising steadily for more colleagues propose a mechanism for the that we will run out of food,” says Márcio than a century, but in 2023 the global averflooding. Beginning in the 1990s, they da Silva Santos, the tuxaua, or chief, of the age broke previous records every month benote, warming Atlantic surface water proIndigenous village of Betel. A concrete and ginning in April, a trend that has continued duced a pronounced temperature disparity packed-dirt staircase of 130 steps descends into this year. between the Atlantic and the Pacific basins steeply from the village, on a bluff overA modeling experiment released last that has continued to grow. One driver, looking the Amazon River near Tefé, to the month by World Weather Attribution, an they suggest, is an anomalous warm ocean shrunken river. The steps are the first stage

p g science.org SCIENCE

,

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

Villagers tried unsuccessfully to extinguish the blaze with buckets of water and makeshift brooms. Luckily, a rain shower came that night and extinguished it. The farmers were spooked and too fearful of another fire to burn more fields. In community meetings, the village and three neighboring communities decided that next year they’ll try to prepare some fields without fire. The plan will require a tractor they hope to borrow from a regional development agency. A few minutes down the Amazon by boat from Betel, another Indigenous community, Porto Praia, has seen a different mix of troubles. An island off the town’s waterfront creates a kilometer-wide channel that separates the village from the Amazon’s main stem. In September, the channel completely dried up, revealing an expanse of sand dunes as tall as a person. Fleischmann calls such dune fields “Sahara on the Amazon.” Drier dry seasons and wetter wet seasons are conspiring to produce the dunes, he says. Fueled by extreme wet-season rainfall in headwaters, floods are scouring riverbeds more powerfully, eroding upstream shorelines and carrying heavier loads of sediment that get deposited where the current slows, such as off Porto Praia. Then they are ex-

y

696

crop for many of the region’s Indigenous people, has also been disrupted. In Betel, the heat this season is so intense that farmers can’t work their fields after 9 a.m. Santos says small streams that provide water used for processing cassava root and turning it into flour have dried up, forcing villagers to wait for the end of the drought to process the food. What concerns Santos most, though, is an unprecedented wildfire that occurred one evening in September. As is customary, villagers had prepared a cassava field by cutting down stalks remaining after the previous harvest and leaving them to dry before setting them on fire. This traditional slash-and-burn method disposes of discarded vegetation, kills weeds, and produces nutrient-rich ash. Normally, such fires peter out where the prepared field ends and the surrounding primary forest begins. But this time the flames kept going. First, they spread into a neighboring field, destroying a patch of açaí palm saplings that would have eventually produced berries for local consumption and sale to outsiders. Then, for the first time since the town was founded 53 years ago, the fire escaped into wild forest. “We had never seen fire like this,” Santos says. “We cannot explain it.”

y g

of a grueling trip made necessary by the drought. Santos grunts under the weight of a handmade wooden canoe he and three other Kambeba tribe members are taking turns carrying, two at a time, down the slope. When they reach the water, they position the canoe on the gunnels of a motorized skiff, ferry it across the Amazon, then hoist it back onto their shoulders and haul it through the jungle. Finally, they arrive at a long skinny body of water, Lake Catuano, a former meander left behind when the Amazon changed course long ago. They cast nets and drop lines, hoping to capture basket loads of pacu, piranha, curimatã, and other choice fish. Santos says he would rather fish a short boat ride from his village, in the Amazon proper. But since the drought took hold, there’s too much competition from commercial fishers who’ve abandoned Lake TefÈ’s literally overheated fishery—and from river dolphins that stake out the pools where fish gather in the depleted river. The dolphins relentlessly rip nets and steal the catch unless they’re constantly shoed away—or beaten back with a jabunque, a cudgel fishers carry for this purpose. It’s not only fishing that has become harder. Planting and harvesting of cassava, a staple starch and the most important cash

y

Vanuza de Barros Gomes holds a fishing net torn by dolphins vying with people for fish in the shrunken Amazon River near the city of Tefé, Brazil.

y ,

697

y g

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

y

SCIENCE science.org

g

met by more than 20% over most of their than 33%. Those dire changes unfolded in posed by drought. The town’s dunes first lengths. The low-water level of most of the the models even with a substantial global appeared during the 2022 dry season; with Xingu, a river in the east that today carries reduction in fossil fuel use. the 2023 drought they made the waterfront half as much water as the Mississippi, will Forsberg says similar extremes are likely landlocked for the first time. decline by more than 50%. That could upto afflict southern Brazil and other parts of Until the rains return and the river rises end the financial rationale for the recently South America, because conditions in the again, Porto Praia villagers—who have no completed Belo Monte power complex, one Amazon Basin influence rainfall patterns road access to Tefé, the nearest commercial of the world’s largest hydroelectric stafor thousands of kilometers around. A 2019 area—have to walk 1 hour to get to a boat, tions, Forsberg says. The team published paper in the Journal of Climate estimated doubling their travel time. Porto Praia’s one these modeling results in Climatic Change that “aerial rivers” of moisture from the Amaschool has closed because most of its teachin 2016, and Forsberg says events like last zon contribute 16% of the rainfall in the La ers commute from Tefé. The villagers have Plata River Basin, one of the also taken to fishing at night world’s largest catchments, because carrying gear and extending from the southern baskets of fish across the An area hard hit edge of the Amazon Basin for dunes to their preferred lake The central Amazon experienced some of the worst effects of the 2023 drought. Water more than 2000 kilometers during the heat of the day in Lake Tefé reached 39.1°C, 9°C above average. In the Indigenous village of Betel, Brazil, to Buenos Aires, Argentina. is unbearable. “It’s hard to fishers hauled their gear to tiny Lake Catuano to avoid competition from other fishers Those aerial flows are now fish during daytime nowaand dolphins at their usual spots in the Amazon. In nearby Porto Praia, sand dunes left in peril, Forsberg says, which days because of the heat,” behind by receding water cut off access to the river. could shrink rivers, parch says Anilton Bras da Silva, crops, and hamper power chief of Porto Praia. “In past plants on which millions of droughts, we could manage.” people depend. “It’s going to have disastrous effects … BRUCE FORSBERG, an ecoLakee T Lak Tefé efé Temperature that are going to leak down logist who has studied the gauge to all of South America.” Amazon for more than 4 decades, says Porto Praia Lake Catuano Tefé BEFORE I LEFT BRAZIL, Santos, and Betel will probably Sand dunes the tuxaua of Betel, invited see worse. Forsberg directs me for a meal. He built a fire the Large-Scale BiosphereBetel 0 10 and skewered a jaraqui, a Atmosphere Experiment, a Porto Praia fish the size of a dinner plate long-running international km and thin as my palm, with a collaboration focused on slender green branch. While links between the Amazoit roasted over the coals my nian rainforest and global interpreter, Diana Mayra climate. He and a team of Kˆhler, said, “Comeu jaraBrazilian and U.S. scientists qui, nao sai mais daqui”—“If have modeled how warming you’ve tasted jaraqui, you’ll from a continued increase never leave here”—an Amain fossil fuel emissions zonian maxim that rhymes would affect river flows. charmingly in Portuguese. On a large monitor at his Once, Santos could count office in Manaus, he pulls on the Amazon to yield up maps of the region, its rivers color coded to indiplenty of jaraqui and other cate projected changes in delicacies. Now, he fears that water flow in the final 3 dethe changes sweeping the cades of this century. A wetregion will spark disastrous season map shows most of competition for the bounty. the Amazon’s tributaries as He’s willing to avoid fishing sinuous dark blue and teal in the river to help the dollines, indicating the model’s phins survive. “We don’t disQuezia Gomes, an Indigenous girl, plays in the dry Amazon riverbed near Betel. prediction that almost all of turb them so that they don’t the Amazon’s headwaters in have to suffer any more.” But the Andes will discharge 20% to 50% more year’s drought suggest the basic pattern he’s worried that more harvesting in skinny water by the end of the century. Wet-season they predicted is now starting to occur. Lake Catuano could put Betel at odds with discharge in the main stem of the Amazon A paper published last year in the Jourother villages that also sometimes rely on it. itself will increase up to 20% over most of nal of Hydrology paints a similar picture. “This is a delicate situation for us.” its length, suggesting serious flooding in Looking ahead 40 years, modeling based The same could be said for the great Manaus and much of the rest of the region. on four distinct climate models forecast river sliding past the village, and the forest In contrast, reddish hues color the rivmany headwater regions will experience all around. j ers on a map showing dry season flows. Acfloodwater increases of more than 50%, cording to the model, low-water discharge while low water will hit downstream reDaniel Grossman is a science journalist specializing of every major river in the basin will plumgions hardest, with flow decreases of more in climate change.

p

CREDITS: (GRAPHIC) D. AN-PHAM/SCIENCE; (PHOTO) DADO GALDIERI/HILAEA MEDIA

N E WS | F E AT U R E S

By Jillian R. Jaycox1,2, Yile Dai1, Aaron M. Ring2

I

Department of Immunobiology, Yale School of Medicine, New Haven, CT, USA. 2Division of Translational Science and Therapeutics, Fred Hutchinson Cancer Center, Seattle, WA, USA. Email: [email protected]

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

705

,

1

y

nvestigating the causes of individual variation in health outcomes has led to transformative insights into human biology and advances in nearly every branch of medicine. Historically, emphasis has been placed on how genetic factors contribute to phenotypic variation within populations. However, an emerging concept is that self-reactive antibodies (autoantibodies) represent a critical yet largely underexplored factor that influences human health and disease. Investigating autoantibodies and their protective as well as pathological roles in disease may unlock new treatment paradigms, much like the prior study of genetics. Generated by the humoral immune system, antibodies are capable of specifically binding to virtually any biomolecule target (broadly termed “antigens”) (1). Although the primary function of antibodies is to provide adaptive immunity against pathogens, invariably some antibodies arise that bind to self-antigens. These autoantibodies elicit a wide range of biological effects, including altering the activity of their targets and immunomodulation (see the figure). Every person carries a distinct array of autoantibodies—an “autoantibody reactome”— offering a potential avenue for trait diversity that mirrors the way genetic differences influence phenotypes. Autoantibodies are usually known for their etiologic role in mediating autoimmune diseases. Canonically, autoantibodies can drive pathological inflammation within nearly any tissue, notably affecting the skin, joints, muscles, and central nervous system as well as organs such as the thyroid and pancreas (2). Similarly, autoantibodies can trigger distinctive syndromes marked by highly specific biological effects, akin to the distinct impact observed with Mendelian single-gene mutations, because they interfere with essential pathways in the body. Notable examples include myasthenia gra-

vis, a neuromuscular disease caused by autoantibodies that inhibit the acetylcholine receptor, and the hyperthyroidism in Grave’s disease that is driven by autoantibodies that activate the thyrotropin receptor. Less appreciated are the more subtle phenotypic effects of autoantibodies that are disease-modifying or even clinically silent until their activity is unmasked in states of stress. A key example of this phenomenon was revealed during the COVID-19 pandemic in which type I interferon (IFN-I)– neutralizing autoantibodies were found to confer up to 200-fold increased risk of death from COVID-19 (3). Although they are apparently clinically silent in most circumstances, the prevalence of IFN-I autoantibodies sharply increases with age, peaking at ~4% of individuals over 70 years old (4). Consistent with their substantial clinical influence and overall frequency, it is estimated that 20% of all COVID-19 deaths are associated with the presence of IFN-I autoantibodies (4). These findings underscore the ability of autoantibodies to reveal both key biological insights [for example, the critical importance of IFN-I in host immunity to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)] and the profound impact that autoantibodies may exert at a population level.   However, autoantibodies are not uniformly deleterious, and in some instances, they may provide protective effects that ameliorate or prevent disease. IFN-I autoantibodies are again instructive. Systemic lupus erythematosus (SLE) is an autoimmune disease that is characterized by elevated IFN-I signaling in >50% of patients. Intriguingly, ~5% of SLE patients have autoantibodies that neutralize IFN-I signaling (5). In contrast to COVID-19, these autoantibodies are associated with substantially lower disease activity, presumably by attenuating pathological IFN-I pathway function (5).  This counterintuitive observation emphasizes the dualistic nature of autoantibodies, demonstrating their ability to confer protective benefits in the very diseases they are typically implicated in causing. The protective effects of autoantibodies are apparent across numerous diseases. For

y g

SCIENCE science.org

Autoantibodies influence a wide range of conditions beyond autoimmune diseases

y

1. B. L. Jilek et al., Nat. Med. 18, 446 (2012). 2. J. D. Siliciano, R. F. Siliciano, Annu. Rev. Pathol. 17, 271 (2022). 3. G. Hütter et al., N. Engl. J. Med. 360, 692 (2009). 4. T. J. Henrich et al., Ann. Intern. Med. 161, 319 (2014). 5. N. W. Cummins et al., PLOS Med. 14, e1002461 (2017). 6. K. Luzuriaga et al., N. Engl. J. Med. 372, 786 (2015). 7. A. L. Hill et al., Proc. Natl. Acad. Sci. U.S.A. 111, 13475 (2014). 8. M. Lee et al., Blood 120, 1432 (2012). 9. F. R. Simonetti et al., J. Clin. Invest. 131, e145254 (2021). 10. N. F. McMyn et al., J. Clin. Invest. 133, e171554 (2023). 11. N. M. Archin et al., Nature 487, 482 (2012). 12. K. J. Kwon et al., Sci. Transl. Med. 12, eaax6795 (2020). 13. D. R. Collins et al., Nat. Rev. Immunol. 20, 471 (2020). 14. J. D. Gunst et al., Nat. Med. 28, 2424 (2022). 15. L. N. Bertagnolli et al., Proc. Natl. Acad. Sci. U.S.A. 117, 32066 (2020). 10.1126/science.adk1831

Decoding the autoantibody reactome

g

RE F ER E NC ES AND NOTES

IMMUNOLOGY

p

ing it harder to achieve the requisite multilog reductions. Even if a way can be found to safely induce all replication-competent proviruses, it will be necessary to ensure that all of the infected cells are killed, a problem confounded by immune exhaustion and viral escape mutations. Other strategies that focus on the excision of latent proviruses or permanently silencing them face similar problems related to the scale of the effect needed to produce a cure. An alternative approach to reservoir reduction, which is known as “functional cure,” involves the induction of immune responses that will keep viral replication in check so that viremia remains undetectable with clinical assays. In this situation, disease progression and virus transmission are unlikely even though the reservoir persists. Precedent for immune control comes from the rare individuals (1/300) who spontaneously control HIV-1 replication without ART (13). Extensive studies indicate that control is most likely mediated by HIV-1–specific cytolytic T lymphocytes. Unfortunately, it has not yet been possible to induce this degree of control in most people with HIV-1 using therapeutic vaccines. Broadly neutralizing antibodies (bNAbs) to the HIV-1 envelope protein can delay rebound and contribute to lower levels of postrebound viremia in recipients with sensitive virus, although some of the observed effects may be related to direct neutralization by trace residual levels of the bNAbs (14). It is also important to note that the autologous neutralizing antibody response can prevent outgrowth of a substantial but variable fraction of HIV-1 variants in the latent reservoir (15). Together, these results suggest that preventing viral rebound and thus achieving a functional cure may depend on enhancing virus-specific humoral immunity to cover all reservoir variants. Although shock-and-kill strategies are unlikely to produce the degree of reservoir reduction required for cure, reservoir reduction will facilitate immune control, and thus these strategies may be a useful adjunct to immune-based cure efforts. j

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

The influence of autoantibodies on health and disease Autoantibodies elicit a diverse array of biological effects that result in disparate health outcomes, with both pathogenic and protective clinical effects. Emerging autoantibody discovery technologies such as antigen microarrays, phage immunoprecipitation sequencing (PhIP-seq), and rapid extracellular antigen profiling (REAP) now enable “autoantibody-wide association studies” to identify putatively causal autoantibodies present within the autoantibody reactome, analogous to genome-wide association studies. Disease-modifying autoantibodies identified in this way can elucidate new drug targets and provide a template for therapeutic development.

Biological impacts of autoantibodies Amyloid plaque Autoantibody Antigen

Neutralization and inhibition

Receptor activation

Cell-specific depletion

Tumor cell opsonization

Inflammation

Molecular clearance

Decoding the autoantibody reactome Disease-specific autoantibody atlas -Log(p-value) Patient ID

amyloid-b that are prone to form plaques that are implicated in neurodegeneration— are found in healthy individuals and decrease with aging, particularly in patients with Alzheimer’s disease (AD) (9). This observation suggests that such antibodies may provide protection from AD and could provide therapeutic benefit. Building on these findings, monoclonal antibodies that target amyloid-b peptides have been developed and found to promote plaque clearance within the brain and to slow cognitive decline in preclinical AD models and clinical trials of AD patients (10). Indeed, an amyloid-b antibody recently approved by the US Food and Drug Administration (FDA), aducanumab, was developed from an autoantibody obtained from a cohort of older individuals who exhibited no indications of cognitive deterioration or displayed unusually gradual cognitive decline (10). This demonstrates that autoantibodies not only highlight potential therapeutic paradigms but, in some cases, can also be advanced as therapeutic drugs themselves. As a general theme, autoantibody associations with health outcomes can provide crucial information about what to drug (what gene product) and when to drug it (which disease indication). For example, protective IFN-I–blocking autoantibodies in SLE appear

science.org SCIENCE

,

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

Pathogenic autoantibodies

y

706

Protective autoantibodies

y g

example, the presence of autoantibodies against tumor-associated antigens (TAAs) has long been appreciated to represent a positive prognostic factor for patients with a variety of cancers. Multiple studies have linked TAA-reactive autoantibodies—such as anti–mucin 1 (MUC1) in various epithelial cancers and anti–human epidermal growth factor receptor 2 (HER2) in breast cancer—with better survival and recurrence outcomes in patients (6, 7). These autoantibodies have been hypothesized to elicit immune-directed tumor cell killing, contributing to improved tumor control. In other cases, determining the specific molecular targets of tumor-reactive autoantibodies led to the identification of new TAAs, including NY-ESO, MAGE, BAGE, GAGE, and HOM-MEL-40 (8). Cancer autoantibodies have thus provided a key line of evidence to support a role for the immune system in tumor surveillance and revealed previously unidentified targets for cancer therapy. Autoantibodies may have other effects on tumors, such as modulating therapeutic responses, but this requires further research. Autoantibodies in neurodegenerative diseases provide another example of beneficial autoreactivity. Autoantibodies that recognize amyloidogenic peptides—short fragments of

Odds ratio

y

Yeast display (REAP)

g

Phage display (PhIP-seq)

Autoantibody

Antigen array

p

Autoantibody-wide association study

Detection methods

to mimic the therapeutic benefits of anifrolumab (a therapeutic monoclonal antibody used in SLE that targets the IFN-I receptor IFNAR1), and HER2 autoantibodies in breast cancer mirror the pharmacology of trastuzumab (a HER2 monoclonal antibody used in the treatment of HER2+ malignancies) (5, 6). However, potential therapeutic targets can be identified not only from protective autoantibodies, which should be therapeutically mimicked, but also from deleterious autoantibodies. Deleterious autoantibodies highlight pathways whose unperturbed functions are necessary for optimal health outcomes, implying that an ideal therapeutic agent would exert the opposite effect. For example, although autoantibodies that neutralize IFN-I exacerbate COVID-19 severity, administration of recombinant type-III interferon, which has similar biological properties to those of IFN-I, has shown promise as a COVID-19 therapeutic (11). Notably, the function of autoantibodies can inform how a particular target should be drugged (what pharmacologic mechanism of action). Autoantibodies can be conceptualized as natural biologic drugs with an extraordinary array of potential impacts on human physiology. Examples include directly activating or inhibiting signaling receptors, stabilizing and extending the circulating halflife of ligands, or promoting their clearance (2). Additionally, autoantibodies can exert immunomodulatory functions by interacting with Fc receptors expressed on immune effector cells to drive antibody-directed cellmediated cytotoxicity (ADCC), phagocytosis (ADCP), or complement-directed killing (CDC), depleting cells that express the target antigen on their surface (1). Integrating knowledge of autoantibody-clinical associations with an understanding of their functions can thus enable the development of comprehensive therapeutic hypotheses that link drug targets to specific therapeutic indications and potential mechanisms of action. Although autoantibodies can exert large clinical effects and offer insights into therapeutic development, their impact on physiology has probably been underestimated. This is likely because there have been no comprehensive surveys to determine how common autoantibodies are throughout the human population at a proteome-wide scale. Central to this challenge has been a lack of experimental tools for unbiased, high-throughput autoantibody detection. Consequently, disease-modifying autoantibodies have largely been discovered through hypothesis-driven approaches informed by known biology or through challenging experimental techniques with limited throughput and scalability. Nevertheless, the emergence of nextgeneration autoantibody detection methodol-

10.1126/science.abn1034

A

ll pathogens and their hosts engage in an arms race to gain the upper hand. Understanding how plants defend themselves could play a key role in the world’s food security. One of the first lines of defense in plants is the cell wall. Pathogens, therefore, use enzymes called polygalacturonases (PGs) to degrade it, and at the same time, to suppress the plant’s immune response. Plants, in turn, counteract by inhibiting PGs with inhibitory proteins (PGIPs). On page 732 of this issue, Xiao et al. (1) report the molecular details of how plant PGIPs do not just inhibit PGs, but also convert this virulence factor into an enzyme that triggers defense responses instead. PGs are one of many virulence factors (or in the case of plant pathogens, effectors) that pathogens use to support infection. They secrete these molecules and deliver some of them into host cells. Hosts have evolved strategies not only to counteract these virulence factors, but also to use them as triggers of broader immune responses to eliminate pathogens. In plants, there are two interconnected tiers of immune responses (see the figure). Pattern-triggered immunity involves pattern-recognition receptors located in the plant cell plasma membrane that recognize conserved pathogen molecules and provide resistance to a broad range of pathogens. Intracellular immune receptors from the nucleotide binding leucine-rich repeat family recognize pathogen effectors delivered into the plant cell and initiate effector-triggered immunity. This often leads 1

Botanical Institute, Christian-Albrechts University, Kiel, Germany. 2Max Planck Institute for Molecular Biology, Plön, Germany. 3School of Chemistry and Molecular Biosciences, The University of Queensland, Brisbane, QLD, Australia. Email: [email protected] 16 FEBRUARY 2024 • VOL 383 ISSUE 6684

707

,

The authors thank E. Wang, P. J. Utz, J. Craft, A. Wang, A. Iwasaki, and C. Hunter for valuable discussions. Y.D. and A.M.R. are inventors of a patent application describing the REAP method and receive royalties from Yale University. A.M.R. is the founder and a director of Seranova Bio, the commercial licensee of REAP.

By Elisha Thynne1,2 and Bostjan Kobe3

y

ACK NOWL EDG M E N TS

Plants commandeer a pathogen’s virulence factor to bolster immunity

y g

1. K. Murphy, C. Weaver, Janeway’s Immunobiology (Garland Science/Taylor & Francis Group, ed. 9, 2016). 2. R. J. Ludwig et al., Front. Immunol. 8, 603 (2017). 3. P. Bastard et al., Science 370, eabd4585 (2020). 4. P. Bastard et al., Sci. Immunol. 6, eabl4340 (2021). 5. H. F. Bradford et al., Cell Rep. Med. 4, 100894 (2023). 6. M. L. Disis et al., J. Clin. Oncol. 15, 3363 (1997). 7. Y. Kotera, J. D. Fontenot, G. Pecher, R. S. Metzgar, O. J. Finn, Cancer Res. 54, 2856 (1994). 8. Y. T. Chen et al., Proc. Natl. Acad. Sci. U.S.A. 94, 1914 (1997). 9. M. Britschgi et al., Proc. Natl. Acad. Sci. U.S.A. 106, 12145 (2009). 10. J. Sevigny et al., Nature 537, 50 (2016). 11. G. Reis et al., N. Engl. J. Med. 388, 518 (2023). 12. W. H. Robinson et al., Nat. Med. 8, 295 (2002). 13. H. B. Larman et al., Nat. Biotechnol. 29, 535 (2011). 14. E. Y. Wang et al., Cell Rep. Methods 2, 100172 (2022). 15. E. Y. Wang et al., Nature 595, 283 (2021).

Mixedorganism enzyme in plant defense

y

R EFER ENC ES AN D N OT ES

PLANT SCIENCE

g

SCIENCE science.org

thousands of distinct autoantibody classes have been observed, individual autoreactivities seen in PhIP-seq and REAP datasets are usually rare, often present in forest area available 25.0 million ha Intact forest not needing restoring 19.2 million ha

42.7 million ha

Portion of current restoration displaced to nonforest 70.1 million ha

Degraded forest for reforestation 63.5 million ha

p Portion of current restoration commitment in degraded forest 63.5 million ha

g

AFR100 commitments as percentage of forest area No commitment 1–25% 25–50% 50–100% >100% AFR100 commitment–no forest

Total area of degraded forest across AFR100 countries (176.3 million ha)

y

1 School of Environmental Sciences, University of Liverpool, Liverpool, UK. 2Department of Zoology and Entomology, University of Pretoria, Pretoria, South Africa. 3Centre for African Ecology, School of Animal, Plant and Environmental Sciences, University of the Witwatersrand, Johannesburg, South Africa. 4Copernicus Institute of Sustainable Development, Utrecht University, Utrecht, Netherlands. 5Centre for African Conservation Ecology, Nelson Mandela University, Gqeberha, South Africa. 6South African Environmental Observation Network, G-F-W Node, Pietermaritzburg, South Africa. 7Environmental Change Institute, School of Geography and the Environment, University of Oxford, Oxford, UK. Email: [email protected]; [email protected]

700

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

science.org SCIENCE

,

nas with native grasses (not tree planting), we could only find evidence of one project actively restoring the grass layer (Moilo Grass Seedbank and Maasai Wilderness Conservation Trust, Kenya); all other projects are focused on increasing the number of trees. These data from on-the-ground restoration projects illustrate that the large areas committed to restoration in nonforest systems are the principal target of restoration by tree-planting programs involving seedling planting (76% of projects) or agroforestry (49% of projects). Furthermore, almost 60% of agroforestry projects use non-native species; introduced species can be particularly problematic when they are invasive (e.g., Grevillea robusta). Of the UN Restoration Principles (12), it is questionable whether two are being met in nonforest ecosystems (“Benefits to Nature and People” and “Addresses Causes of Degradation”) as biodiversity outcomes can be poor, and increasing woody cover in open ecosystems is itself a cause of degradation (4–7 ). Most projects have as their aim improving livelihoods and restoring degraded

y

dicates that total degraded forest habitat covers 176.3 million ha (see the figure). In countries with degraded forest, once the area of degraded forest restoration is accounted for from their total commitment, 112.8 million ha of degraded forest across Africa still requires restoration; this means that a third of low- and medium-integrity forest (63.5 million ha) is earmarked for restoration currently under AFR100. Furthermore, if the total committed area for restoration were focused on truly degraded forest, instead of nonforest systems, three-quarters (75.8%) of degraded forests in AFR100 countries could be restored (see the figure) and extensive afforestation avoided. Examining on-the-ground restoration projects across AFR100 countries in the reforestation.app database (n = 99) with known locations (n = 67), 52% are in savanna or grasslands [with most variously inaccurately classified in the database as Tropical dry forest, Tropical rainforests, or Tropical moist forest; see (10)]. Despite the AFR100 initiative purporting to support restoration of grasslands and savan-

y g

what extent the forest area available for restoration actually requires restoration: That is, for each country, how much forest is degraded versus intact (i.e., not in need of restoration)? We estimated this using the Forest Landscape Integrity Index (FLII) (11) assuming that forests of low and medium integrity, but not high integrity, need restoration. Taking into consideration the extent of forest requiring restoration (i.e., sum of area classified as low and medium forest integrity; see SM), an additional 19.2 million ha of forest has been pledged that exceeds the area of degraded forested habitat (e.g., Ethiopia, Kenya, Somalia) (see the figure). Overall, a total of 70.1 million ha, or 52.5% of the total area committed, is in nonforest ecosystems, principally savannas and grasslands (see the figure); this is greater than the area of France, indicating that tree-based restoration in Africa could cover vast areas of nonforest habitat. Conversely, the area for true forest restoration in AFR100 countries covers less than half of the area pledged (47.5%). Across the 35 countries signed up to AFR100, our analysis based on FLII in-

“…vegetation definitions that only consider tree cover are problematic for tropical grassy biomes…”

RE FE REN C ES AN D N OT ES

SU PP LE ME N TARY M AT E RIALS

science.org/doi/10.1126/science.adj0899

10.1126/science.adj0899 16 FEBRUARY 2024 • VOL 383 ISSUE 6684

701

,

N.S. was supported by the Trapnell Fund. C.L.P. and N.S. contributed equally to this work. M.t.B. was supported by a Vidi grant from the Dutch Research Council (NWO) and acknowledges E. Leltz, J.P.G.M. Cromsigt, and G.I.H. Kerley for valuable discussions that contributed to the ideas presented here.

y

AC KN OW LE DG M E N TS

y g

1. C. Hanson, K. Buckingham, S. DeWitt, L. Laestadius, The Restoration Diagnostic (World Resources Institute, 2015). 2. W. J. Bond, N. Stevens, G. F. Midgley, C. E. R. Lehmann, Trends Ecol. Evol. 34, 963 (2019). 3. S. Maginnis, J. Rietbergen-McCracken, W. Jackson, “Restoring forest landscapes: An introduction to the art and science of forest landscape restoration,” ITTO Technical Series no. 23 (2005); http://www.itto. int/direct/topics/topics_pdf_download/ topics_id=10640000&no=1&file_ext=.pdf. 4. C. L. Parr, C. E. R. Lehmann, W. J. Bond, W. A. Hoffmann, A. N. Andersen, Trends Ecol. Evol. 29, 205 (2014). 5. P. F. Scogings, For. Ecol. Manage. 546, 121381 (2023). 6. N. Stevens, W. J. Bond, A. Feurdean, C. E. R. Lehmann, Annu. Rev. Environ. Resour. 47, 261 (2022). 7. J. D. Wieczorkowski, C. E. R. Lehmann, Glob. Change Biol. 28, 5532 (2022). 8. E. Buisson, S. Archibald, A. Fidelis, K. N. Suding, Science 377, 594 (2022). 9. E. Buisson et al., Biol. Rev. Camb. Philos. Soc. 94, 590 (2019). 10. E. Dinerstein et al., Bioscience 67, 534 (2017). 11. H. S. Grantham et al., Nat. Commun. 11, 5978 (2020). 12. FAO, IUCN CEM and SER, “Principles for ecosystem restoration to guide the United Nations Decade 20212030” (2021); https://www.fao.org/documents/card/ en/c/CB6591EN. 13. F. A. O. Silveira et al., J. Appl. Ecol. 59, 1967 (2022). 14. A. Wurz et al., Nat. Commun. 13, 4127 (2022). 15. N. Sasaki, F. E. Putz, Conserv. Lett. 2, 226 (2009).

y

forests (13) within global restoration efforts. At the root of the problem is that the grassy biomes remain fundamentally misunderstood and consequently are misclassified as forest (4, 6). Guidelines for FLR based on the FAO definition of forest consider any area with a minimum of >0.1 ha in size, a minimum of 10% tree cover, and a minimum tree height of 2 m as forest. The definition has been criticized (4, 5, 15) because it may not adequately differentiate natural and plantation forest (which differ in carbon and biodiversity values) and is based solely on vegetation structure, meaning that open systems with trees, such as savanna, can be misclassified. Yet the definition persists, in part because the focus is on trees and canopy cover is easily measured with remote sensing. FLR targets areas with low tree cover in climates that can support forest, often erroneously assuming that these areas are deforested and degraded and therefore represent opportunities for restoration (1, 2). But vegetation definitions that only consider tree cover are problematic for tropical grassy biomes because they fail to recognize the grassy layer under the canopy that is a defining feature of these systems and means they are structurally, functionally, and compositionally distinct from forests (4). Although FLR states that tree cover should not exceed that considered “ecologically appropriate for a particular location” (1), it is unclear who decides what is appropriate. Grassy systems vary hugely in woody cover in space and time [e.g., 5 to 80% cover (4)], and such a vague approach with no consideration of the inherent ecological differences between forest and open ecosystems is fraught with problems. Increasing tree cover in open ecosystems globally represents a major threat, not only

for the ecosystems themselves but ultimately for society as a whole (4, 6), and treeplanting actions will exacerbate already problematic woody encroachment. We urge a paradigm shift away from the structural focus on trees to include the distinctive and important characteristics and ecology of grassy, nonforest systems. Although not perfect, an improvement would be the use of biome maps, such as RESOLVE ecoregions, which would enable large areas of nonforest to be masked and true forests to be identified more accurately (5). It is essential that the differing characteristics of forest and nonforest degradation are recognized because this will determine restoration actions and enable genuinely degraded systems to be restored with greater sensitivity. Ultimately, the right trees and the right number need to be planted in the right place. But, until the definition of forest is revised, there will always be the double jeopardy of afforestation of ancient grasslands and deforestation of virgin forest. We must act to avoid a situation where we cannot see the savanna for the trees, and these precious grassy systems are lost irrevocably. j

g

SCIENCE science.org

layer is not helpful for most of these degradation examples. Although here we demonstrate the scale of inappropriate restoration across Africa, nonforest systems are threatened globally (e.g., the rich grasslands of Chapada dos Veadeiros National Park, Brazil, with up to 400 species of plants/ha, are being planted with trees). Our analyses indicate that tree planting is widespread across nonforest systems, and highlight the hegemony of trees and

p

and/or deforested areas, but critical information on what aspects of degradation are being restored (9), as well as monitoring and performance indicators, are seldom openly available, and finding information on these aspects of projects is challenging. Given that these projects received considerable funding (>$1 billion in development finance and $148 million from private sector commitments) from Global North governments (e.g., German Federal Ministry for the Environment), IUCN, UN, the FAO, Global Environmental Facility, and local and international nongovernmental organizations (e.g., Nature Conservancy), it is essential that the projects are transparent and accountable, or they will be at risk of greenwashing. There are likely a combination of drivers underlying our findings, including lack of ecological awareness among the public and policy-makers (13), large financial incentives, forest definitions, and poor program management. We argue that although FLR allows for multiple benefits of tree-based restoration, the widespread use of agroforestry for restoration should receive greater scrutiny, especially in nonforested ecosystems. Although agroforestry can enhance biodiversity (particularly when starting from a low baseline) and increase ecosystem health (e.g., increasing soil fertility, reducing soil erosion), and therefore assist with recovery of degraded land, this is not always the same as restoration because agroforests support fewer species than forests (14). Agroforestry, particularly with non-native species, in grassy systems is problematic because ecosystem processes such as fire and grazing, which are critical to the functioning of tropical grassy systems, often conflict with agroforestry. Therefore, although agroforestry can offer considerable benefit to human well-being, ecological integrity and functionality are not automatically enhanced. Instead, in grassy systems, other land uses may be more appropriate and compatible with restoration across large areas—these include, for example, sustainable livestock and wildlife farming. Of course, nonforested systems also require restoration when degraded (through soil erosion, loss of herbaceous layer from over-grazing and woody plant encroachment, suppression of fire, and overharvesting of trees) (6, 9). However, it is essential to identify nonforest ecosystems correctly so that they receive appropriate restoration interventions (e.g., resting land from grazing, seeding with grasses, clearance of woody encroachment). The dominant focus on trees to “regain ecological functionality” (1) rather than restoration of the ground

A statue of a great auk overlooks a waterway on Iceland’s Valahnukur coast.

B O OKS et al .

The end of the auk By Mark V. Barrow Jr.

709

,

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

y

SCIENCE science.org

10.1126/science.adn2788

y g

PHOTO: MJ PHOTOGRAPHY/ALAMY STOCK PHOTO

I

y

ting it into appropriate historical context. The great auk (Pinguinus impennis) was a large flightless seabird that inhabited the n The Last of Its Kind, Icelandic anthroNorth Atlantic and once numbered in the pologist Gísli Pálsson explores a largely millions. With a glossy black back, white forgotten expedition to Iceland that the belly, small head and wings, and webbed feet, British naturalists John Wolley and Alit grew to about 70 cm tall and weighed apfred Newton made in 1858. Like many proximately 5 kg as an adult. Fast, agile swimof their Victorian contemporaries, both mers, the species survived by foraging fish in men were avid bird and egg collectors, and coastal waters and bred on a limited number they ventured to this remote Nordic island of small, remote rocky islands. on an obsessive quest for an eluBecause it was conspicuous, sive bird—the great auk, also lacked an innate fear of humans, known as the garefowl. Although and moved awkwardly on land, Wolley and Newton failed to find the great auk proved highly vulthe much-coveted species during their 2-month journey, they nerable to the European fisherreturned home with several hunmen who began plying the North dred pages of interviews with the Atlantic in growing numbers at men who had hunted the rare bird the end of the 15th century. In the The Last of Its Kind: and the women who had skinned late 18th century, the exploitation The Search for the it to sell to eager foreign collectors. of the species for food, oil, and Great Auk and the Compiled in a set of five note- Discovery of Extinction bait became overshadowed by books handwritten in English, the devastating industrial-scale Gísli Pálsson Princeton University Icelandic, Danish, and German harvest of adult birds for their Press, 2024. 328 pp. that now reside in the Cambridge feathers, which were used in matUniversity Library, the Gare-Fowl tresses and quilts. As the species Books offer an invaluable window onto a became increasingly rare, egg and specimen lost species and the people who pushed it collectors also keenly sought to possess exbeyond the brink of extinction. Pálsson has amples of it, boosting economic incentives to done a wonderful job not only of mining take the final survivors. this singular primary source but also of putThe last known successful hunt, which occurred on Eldey Island—a 3-ha skerry about 10 km off the coast of the Reykjanes The reviewer is at the Department of History, Virginia Peninsula in southwest Iceland—was underPolytechnic Institute and State University, Blacksburg, VA taken at the behest of a Reykjavik merchant 24061, USA. Email: [email protected]

g

An anthropologist recounts how a pair of 19th-century naturalists traced an elusive bird’s extinction in real time

p

HISTORY OF SCIENCE

hoping to cash in on the growing demand for the vanishing bird. In 1844, a crew of 14 hardy Icelanders spent 12 hours rowing an open boat through treacherous waters to reach the island, where they managed to find and kill just two great auks. Fourteen years later, Wolley and Newton believed that the great auk might still be clinging to life on Eldey Island. However, their hopes for encountering it there were dashed by a long spell of bad weather that made an already risky voyage to the site downright foolhardy. Undaunted, the two men decided to learn as much as they could about the species by interviewing the peasants and fishermen who had hunted and eaten it and, in the process, become familiar with its behavior. In short, according to Pálsson, “from ornithologists,” Wolley and Newton “became anthropologists.” Although plagued with translation issues and inconsistencies that are part and parcel of much anthropological research, the Gare-Fowl Books “provide a vivid, almost real-time account of the extinction of the great auk.” Wolley’s premature death from an apparent brain ailment a little more than a year after the two men returned to England derailed any possibility for quick publication of the two naturalists’ research. Wolley bequeathed his massive egg collection to Newton, who produced a four-volume catalog, Ootheca Wolleyana (1864–1907). The release of Symington Grieve’s The Great Auk, or Garefowl in 1885 thwarted any remaining hopes Newton had of writing his own authoritative monograph on the species using his and Wolley’s Icelandic data, although he did manage to produce an article-length account of their work in 1861. Moved by his experience with the great auk and other declining avian species, Newton became a prominent bird conservation advocate. In 1869, he helped write the Sea Birds Preservation Act, the first British legislation protecting wild birds, and he was a strong supporter of the Society for the Protection of Birds from its inception two decades later. The Last of Its Kind is a fascinating, important, and timely book about a major icon of extinction. Although the rapid demise of the passenger pigeon and the nearly simultaneous narrow escape of the American bison would soon overshadow its disappearance, Pálsson reminds readers that the passing of the great auk represents a historic milestone in the discovery and problematization of human-caused wildlife loss. j

I NS I GHTS | B O O K S

SCIENCE AND SOCIETY

Managing expectations

Fluke: Chance, Chaos, and Why Everything We Do Matters Brian Klaas Scribner, 2024. 336 pp.

A political scientist urges readers to embrace the chaos and complexity of life By Michael Travisano

y g y ,

science.org SCIENCE

y

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

g

7 10

p

is difficult to understand which problems are hard and why. The outsize consequences of certain indispiring researchers are frequently adviduals’ behavior are highlighted in several vised to manage their expectations in historical vignettes, including one from the order to maximize success and miniUS Civil War and one from Darwin’s travels mize disappointment and stress. But on the HMS Beagle. The takeaways from these what if the commonplace basis for setstories are familiar to any geneticist as exting expectations is deeply flawed, as amples of pleiotropy and epistasis, although Brian Klaas argues in Fluke: Chance, Chaos, Klaas does not make this comparison. In soand Why Everything We Do Matters?  Here, ciety as in genetics, individual interactions Klaas uses examples from history to show are complex and multifaceted, involving spethat chance and contingency play surpriscific idiosyncrasies, partially due to context ingly large roles in determining success, and timing. Individual behaviors—whether including in life-and-death scenarios. He by people or genes—occur in the context of emphasizes that seemingly small, inconsethe rest of the system. quential events can have This interconnectedness outsize consequences is a major determinant for both the individuof the functional whole, als directly involved even though individual and others quite well contributions are not all removed. Pulling from equally consequential. diverse disciplines, inOf course, the basis cluding evolutionary bifor human interactions ology, probability, social is different from genetic science, and complexity interactions, and Klaas science, Klaas explains makes clear the potenwhy flukes are to be extial for outsize consepected, that narratives quences from human are tools for understandinteractions, both for ing the world, and that ill and for benefit. He embracing chaos is the emphasizes that maxibest approach for setting mizing efficiency can expectations. lead to fragility—as was Evolutionary biology evident in material supplays an important role in Fluke because the baply chains during the sis of Klaas’s argument is COVID-19 pandemic— Chance events can have outsize effects. Recognizing this helps us make realistic predictions. that the future is highly and that seemingly contingent on the present, and the present on tect patterns, even if they are false, suggests trivial unique knowledge can provide inthe past. With replicate populations and tens possible actions one might take, offering the sights that yield transformational change beof thousands of generations, Escherichia coli illusion of control. cause such information can help in pattern experimental evolution is an ideal example Klaas argues that the efforts required to recognition. of this phenomenon, and Klaas leverages the discern truth in an exceptionally complex Fluke  closes with a discussion on deterresearch of Richard Lenski, Zachary Blount, world can lead to inaction and potentially to minism and the merits of a complex, chaand other evolutionary biologists to systemdeath, as in cases of predation, when quick reotic world. At this point in the text, Klaas atically demonstrate this in the beginning of sponses are essential. However, false pattern has made the case that reality is tightly conthe book.  detection can be problematic because it can nected, with interweaving strands that touch In some E. coli populations, subsequent lead to conspiracy theories and an inability upon one another in nonlinear ways. Pull generations look drastically different from to admit to uncertainty. To Klaas, the phrase on any strand and the structure changes in those in other populations, he reveals. Chance “I don’t know” is vastly underused, leading to numerous ways, many of which have undifferences in the appearance and fixation of a false understanding of the causal basis of predictable consequences. Recognizing this reality, an underappreciation of the complexcomplexity makes setting reasonable expecThe reviewer is at the Department of Ecology, Evolution, and ity of the world, and an impaired ability to tations possible without the necessity for ilBehavior, the BioTechnology Institute, and the Minnesota identify opportunities for success. If we are lusionary control, he concludes. j Center for Philosophy of Science, University of Minnesota, unable to admit to uncertainty, he argues, it 10.1126/science.adn5394 St. Paul, MN 55108, USA. Email: [email protected]

A

mutations lead to alternative outcomes and alternative experimental histories. With this grounding, Klaas then covers a lot of material in the chapters that follow, pulling from many disciplines to motivate viewpoints without necessarily attempting to work out details and specifics. If chance and contingency are so important, why do humans often attribute outcomes to regularity and order? It is not that the importance of chance and contingency has not been repeatedly previously identified. Still, no one wants their successes or failures to depend on chance events outside their control. More importantly, the ability to de-

This artist’s rendering shows the Thirty Meter Telescope, one of two proposed extremely large US telescopes.

LET TERSS

p

Edited by Jennifer Sills

IMAGE: TMT INTERNATIONAL OBSERVATORY

1Director, National

Optical-Infrared Astronomy Research Laboratory (NOIRLab), National Science Foundation, Tucson, AZ 85719, USA. 2Executive Director, Thirty Meter Telescope International Observatory, Pasadena, CA 91124, USA.

10.1126/science.adn7832

Florida law undercuts US leadership in science In his In Depth story “Florida law blocks hiring of Chinese students” (15 December 2023, p. 1220), J. Mervis describes Florida’s SB 846 (1), which hinders and potentially prohibits hiring applicants from China or six other “countries of concern” unless the university obtains 16 FEBRUARY 2024 • VOL 383 ISSUE 6684

7 11

,

N.

Shelton3

1. Giant Magellan Telescope, “Founders” (2024); https:// giantmagellan.org/founders/. 2. TMT International Observatory, “Partners” (2022); https://www.tmt.org/page/partners. 3. Gordon and Betty Moore Foundation, “Thirty Meter Telescope” (2023); https://www.moore.org/initiativestrategy-detail?initiativeId=thirty-meter-telescope. 4. I. Schultz, “Planned giant telescope gets a huge influx of cash,” Gizmodo (2022). 5. National Academies, “Decadal Survey on Astronomy and Astrophysics 2020 (Astro2020)” (2021); https:// www.nationalacademies.org/our-work/decadalsurvey-on-astronomy-and-astrophysics-2020astro2020. 6. AURA, “Open letter to NSF director” (2023); https:// www.aura-astronomy.org/blog/2023/07/05/ open-letter-to-nsf-director/. 7. US Senate Committee on Commerce, Science, and Transportation, “CHIPS and Science implementation and oversight” (2023), video timestamp 50:10; https://www.commerce.senate.gov/2023/10/ chips-and-science-implementation-and-oversight. 8. US Senate, “Division B—Commerce, Justice, Science, and Related Agencies Appropriations Act, 2023” (2023), p. 117; https://www.appropriations.senate.gov/ imo/media/doc/Division%20B%20-%20CJS%20 Statement%20FY23.pdf.

y

P. Kirshner2*, Robert

RE FE REN C ES AN D N OT ES

y g

SCIENCE science.org

Patrick McCarthy1, Robert

Magellan Telescope, Pasadena, CA 91101, USA. *Corresponding author. Email: [email protected]

y

In his Editorial, “Extremely large telescopes at risk” (24 November 2023, p. 857), M. S. Turner calls for the National Science Foundation (NSF) to fund the construction of one US extremely large telescope (ELT), either the Giant Magellan Telescope in Chile or the Thirty Meter Telescope in Hawaii. We agree with Turner that NSF should act, but we advocate US government funding for both telescopes. The vision articulated by the National Academies’ 2020 Decadal Survey on Astronomy (Astro2020) ranked the US Extremely Large Telescope Program as the top priority for ground-based astronomy. The program offers a bold vision of broad ELT access to any US-based scientist. Funding both telescopes would make this vision a reality. Each telescope has committed international partners that include the most active countries in observational astronomy outside Europe: Australia, Brazil, Canada, Japan, India, Israel, and Korea (1, 2). Together with their US partners, they are on a path to contribute a substantial portion of the 3 billion US dollars required for each of the telescopes (3, 4). If NSF agrees to fund the balance of each project, the telescopes could be functional in a decade and productive for the next 50 years. Two telescopes would provide a wider suite of instruments to address the full

3President, Giant

g

Fund two extremely large US telescopes

scope of opportunities outlined in Astro2020 and enable validation of discoveries with complementary techniques. Because they would be located in different hemispheres and at different longitudes, there would be more opportunities to observe transient astronomical sources discovered by other facilities. A whole-sky system would allow scientists to analyze the atmospheres of rare objects, such as the nearest stars with planets, for signs of biological activity. The two telescopes together could also study colliding neutron stars detected by the Laser Interferometer Gravitational-Wave Observatory, which can occur anywhere in the sky. This work would provide insight into the origin of the elements and the expansion rate of the Universe. The US federal government should match the European Southern Observatory’s 39-m telescope in Chile by funding both smaller US projects, which, combined, would have similar area and twice as many observing nights. NSF does not currently have access to the required funds, but US scientists and citizens have shown substantial interest in bringing the projects to fruition (5, 6), and Congress has indicated interest in authorizing the money (7, 8). We urge a bold vision that will, over a decade of construction, empower US astronomers to continue their pioneering work.

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

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

1Department

of Anatomical Sciences, Faculty of Medicine, Tehran Medical Sciences, Azad University, Tehran, Iran. 2Department of Statistics, University of Qom, Qom, Iran. *Corresponding author. Email: [email protected] R E F E R E N C ES A N D N OT ES

1. “Animals used in research: Overview,” MSPCA-Angell (2023); https://www.mspca.org/animal_protection/ lab-animal-welfare-overview/. 2. “Laboratory animals, the silent victims of the progress of science: The silent suffering and death of laboratory animals in the vacuum of law and education,” Iranian Students’ News Agency (2009); https://www.isna.ir/ news/8809-05131 [in Farsi]. 3. Ministry of Health, Government of Iran, “Guide to the care and use of laboratory animals in scientific affairs in the Islamic Republic of Iran” (2020); https://ethics.acecr.ac.ir/UploadedFiles/gFiles/ AnimalCareGuideline-1399.pdf [in Farsi]. 4. S. Ahmadi-Noorbakhsh et al., “A comprehensive guide for the care and use of laboratory animals,” ScientificResearch Journal of Shahed University (2015); https:// www.sid.ir/paper/31139/en [in Farsi]. 5. National Centre for the Replacement, Refinement, and Reduction of Animals in Research, “The 3Rs” (2022); https://nc3rs.org.uk/who-we-are/3rs. 6. R. K. Boettger, C. Lam, IEEE Trans. Prof. Commun. 56, 272 (2013). 7. D. C. Montgomery, Design and Analysis of Experiments (Wiley, ed. 9, 2017), Section 6.5. 8. “Collection of laws, guidelines and ethical guidelines in biomedical research in Iran,” Pasteur Institute of Iran (2023) [in Farsi]. 9. “The production of transgenic animals at Razi Vaccine and Serum Research Institute: Institute/ethics committee supervises the issue,” Iranian Students’ News Agency (2017); https://www.isna.ir/news/97081908964/ [in Farsi]. 10. H. Feredowsian, “Stop torturing animals in the name of science,” Scientific American (2021). 10.1126/science.adn7909

science.org SCIENCE

,

7 12

Atarodsadat Mostafavinia1 and Seyed K. Ghoreishi2*

y

1. Florida Senate, “CS/CS/SB 846: Agreements of educational entities with foreign entities” (2023); https://www.flsenate.gov/Session/Bill/2023/846. 2. United Faculty of Florida–University of Florida, “Update: International student recruitment and ‘countries of concern’” (2023); https://uff-uf.org/update-internationalstudent-recruitment-and-countries-of-concern/. 3. S. M. Rovito, D. Kaushik, S. D. Aggarwal, MIT Science Policy Review, J. Stahlhut, Y. Petri, Eds. (2021); https:// sciencepolicyreview.org/2021/08/impactinternational-scientists-engineers-students-usresearch-output/. 4. K. Ferguson, B. Amlani, T. P. Kimbis, “Pressing forward in supporting postdocs: 2023 National Postdoctoral Association institutional policy report” (2023); https:// cdn.ymaws.com/www.nationalpostdoc.org/resource/ resmgr/docs/2023_NPA_Instit_Policy_Repor.pdf. 5. NIH Advisory Committee to the Director, “Working group on re-envisioning NIH-supported postdoctoral training” (2023), p. 20; https://www.acd.od.nih.gov/documents/ presentations/12152023_Postdoc_Working_Group_ Report.pdf. 6. University of Florida faculty, “Subject: Urgent

Globally, about 192.1 million animals are used annually for experimental purposes and testing (1). Iran has conducted animalbased research dating back to the time of Ibn Sina (2). Although, like many countries, Iran regulates the use and treatment of laboratory animals, it lacks sufficient enforcement. To ensure ethical treatment of research animals, Iran needs better data about their use, and institutions and individuals need to take responsibility for their care. Iran’s Ministry of Health, which oversees animal testing, requires that all animals are treated according to international ethical standards (3). However, some institutions that produce or use animals provide substandard treatment (2, 4), disregarding the ministry’s guidelines. To facilitate better enforcement of established regulations, Iran needs clear, accurate, and reliable information about the quantity of animals used in research. All Iranian research institutions that use animals should be required to collect and report annual data on the number of animals they use, the conditions in which the animals are kept, and the animals’ research purpose. Such data would allow Iran to track each institution’s adherence to the “3Rs” of animal research: Replacing animal-based experiments with alternative strategies, reducing the number of animals used, and refining techniques to minimize the pain and discomfort research animals experience (5). In addition to government enforcement, individual institutions should prioritize the replacement, reduction, and refinement of animal use. Iran’s medical schools and research institutions should establish an accessible database with comprehensive experimental details to prevent redundant animal studies. Ethics committees should not approve experiments that use more animals than necessary or methods that require animals when alternative methods are available. In some cases, technical sampling (using a small sample size) can replace experimental sampling (using a relatively large sample size) (6). For analysis of variance factorial designs with several factors, a single replication strategy that

y g

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

Ethical treatment of Iran’s research animals

y

Executive Director and Chief Executive Officer, National Postdoctoral Association, Rockville, MD 20855, USA. Email: [email protected]

10.1126/science.adn8952

tests the effects of lower-order interactions and ignores insignificant higher-order interactions should be given preference over the use of multiple strategies that each require animals (7). Ethics committees should also strictly supervise the treatment of animals during birth and rearing, the trial period, and sacrifice. Iran’s Pasteur Institute (8) and Razi Vaccine and Serum Research Institute (9) can serve as models of institutions that have implemented ethical animal use policies. Individuals should also take an active role in ethical animal treatment. All researchers must be trained to adhere to ethical principles and understand that unnecessary pain or discomfort in animals could affect the interpretation of data (10). Nongovernmental organizations should conduct animal rights campaigns to educate both scientists and the public. Maintaining international ethical standards for animal treatment is vital for Iran to assert its global standing in the field of medical sciences. Iran’s government, institutions, nongovernmental agencies, individual scientists, and public can work together to ensure ethical animal research.

g

Thomas P. Kimbis

clarification request regarding hiring international students with assistantship” (2023); https://docs.google. com/forms/d/e/1FAIpQLSezPVkPl6o1aVzy388egl1hho NKEpnkP0623ByfPptLc1KtYg/viewform.

p

a waiver from the state (2). Any state’s decision to shut out international scholars will push talented individuals elsewhere, cause delays in existing research efforts, threaten local job growth potential, and endanger US global leadership in research and development. Scientists born outside of the United States have contributed to global understanding and progress as well as to the US economy, society, health, and national security (3). About 58% of the 72,000 postdoctoral scholars (postdocs) working in the United States are internationals (4). Having obtained the highest level of education in their fields, many are in their late 20s and 30s and are poised to assume leadership positions across society. Welcoming postdocs from other nations drives innovation and builds stronger research institutions capable of greater accomplishments and economic development (3, 5). There is no credible evidence of a surplus of domestic-born postdoc scientists and researchers whose career pathways are put at risk by international scholars. Moreover, by turning away the brightest minds from the seven targeted countries, Florida increases the chances that these individuals will instead find opportunities in other countries, including US competitors, a lost opportunity for innovation that could threaten the very national security that SB 846 purports to protect. More than 300 faculty members have now signed the petition, cited by Mervis, that questions the policy and advocates for the freedom to hire top talent regardless of national origin (6). The National Postdoctoral Association, which supports US postdocs of all nationalities, joins their call.

PRIZE ES SAY GRAND PRIZE WINNER

Bingxu Liu

p

Bingxu Liu received undergraduate degrees from Zhejiang University and a PhD from Massachusetts Institute of Technology. He is currently a postdoctoral scholar at the Institute for Protein Design at the University of Washington, where he works on exploring and overcoming the limitations of natural biological systems. www.science.org/doi/10.1126/ science.adn9407

g Stimulator of interferon genes (STING), upon sensing danger signals, acts as a channel and induces the leakage of protons from Golgi bodies into the cytosol to trigger downstream innate immune defenses.

y

IMMUNOLOGY

Protons are the new first responders

Institute for Protein Design, University of Washington, Seattle, WA, USA. Email: [email protected] SCIENCE science.org

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

7 13

,

D

etection and clearance of invading pathogens are crucial for maintaining homeostasis in living systems, from bacteria to humans. First proposed by Charles Janeway in 1989, multiple sensors for sensing pathogen-associated molecular patterns (PAMPs) have now been discovered experimentally. Owing to the diversity of pathogens, human immune systems exhibit remarkable creativity in their detection of various PAMPs. The immune system can sense a wide spectrum of danger signals, which encompass macromolecular double-stranded DNA (dsDNA), lipopolysaccharides, flagel-

lin proteins, and small molecules such as cyclic dinucleotides from bacteria. The varied sensors of these particular signals typically converge in their downstream effects, for example, activation of nuclear factor kB (NF-kB) or interferon pathways (1, 2). Despite similarities in their downstream responses, some sensors exhibit additional behaviors upon activation. Stimulator of interferon genes (STING), a critical protein in the dsDNA response pathway, not only induces interferon [through TANK-binding kinase 1–interferon regulatory factor 3 (TBK1-IRF3)] or inflammation (through NF-kB) but also triggers noncanonical autophagy, NLR family pyrin domain– containing 3 (NLRP3) inflammasome activation, and cell death (3–5). STING therefore plays an important role in inducing multiple distinct defenses against

y

CREDITS: (PHOTO) IAN HAYDON/UNIVERSITY OF WASHINGTON; N. BURGESS/SCIENCE

By Bingxu Liu

y g

Human STING’s newfound function as a channel expands our understanding of immunity

Gabriele Casirati received his MD degree from Università degli Studi di Milano and trained as a hematologist at Università Vita-Salute San Raffaele. From 2020 to 2023, he worked as a research fellow at the DanaFarber Cancer Institute and received a PhD in molecular and translational medicine from Università MilanoBicocca. He is currently a postdoctoral research fellow at Boston Children’s Hospital. His research focuses on new gene-editing approaches to enhance targeted immunotherapies for hematological malignancies. www.science. org/doi/10.1126/science.adn9409 FINALIST

Carla Nowosad

RE FE REN CES A ND N OT ES

5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

K. A. Fitzgerald, J. C. Kagan, Cell 180, 1044 (2020). S. Liu et al., Science 347, aaa2630 (2015). B. Zhong et al., Immunity 29, 538 (2008). T. D. Fischer, C. Wang, B. S. Padman, M. Lazarou, R. J. Youle, J. Cell Biol. 219, e202009128 (2020). M. M. Gaidt et al., Cell 171, 1110 (2017). D. Gao et al., Science 341, 903 (2013). S.-R. Woo et al., Immunity 41, 830 (2014). R. Ulferts et al., Cell Rep. 37, 109899 (2021). D. Lu et al., Nature 604, 557 (2022). B. Liu et al., Science 381, 508 (2023). J. Wu et al., J. Exp. Med. 216, 867 (2019). J. Wu, N. Dobbs, K. Yang, N. Yan, Immunity 53, 115 (2020). M. Gentili et al., Nat. Commun. 14, 611 (2023). U. Tak, P. Walth, A. T. Whiteley, bioRxiv 10.1101/2023.07.24.550367 (2023). B. R. Morehouse et al., Nature 586, 429 (2020).

10.1126/science.adn9407

science.org SCIENCE

,

1. 2. 3. 4.

y

lease, or the loss of calcium homeostasis may be a secondary effect that results from proton leakage–induced voltage changes. STING activation holds great promise for activating the immune system to control cancer, but effects like STING-induced T cell death limit its therapeutic efficacy (12). Moreover, we previously showed that STING-induced noncanonical autophagy promotes STING degradation to terminate the STING response. Thus, biased activation of STING without activating its channel function may also provide much stronger STING-induced TBK1IRF3 activation (13). Our discovery raises the possibility that a biased activation of downstream pathways could potentially offer better therapeutic effects by preventing extensive T cell death while amplifying certain STING downstream signaling.

y g

Carla Nowosad received undergraduate degrees from the University of Warwick and a PhD from the Francis Crick Institute, National Institute for Medical Research Mill Hill. After completing her postdoctoral fellowship at the Rockefeller University, she started her laboratory in the Department of Pathology and the Translational Immunology Center at New York University Grossman School of Medicine in late 2021. Her research focuses on how B cells make decisions in the complex microenvironment of the intestine. www.science. org/doi/10.1126/science.adn9414

Although, to our knowledge, STING is the first known immune-sensing channel in humans, ion flow could be more widely used as a sensing and defense mechanism in other organisms. In plants, activation of specific innate immune sensors has been shown to induce ion leakage. In bacteria, 29,39-cyclic guanosine monophosphate–adenosine monophosphate (also an agonist for STING) can activate the channel protein Cap14 to induce chloride flow for antiphage defense (14). Interestingly, in some bacterial species, STING has been shown to contain a transmembrane domain, instead of the toll–interleukin-1 receptor domain found in most bacterial species, which is indicative of a divergence in STING function that occurred early on during evolution (15). The acquisition of channel function by STING during evolution suggests that certain innate immune protein domains (such as the STING ligand-binding domain) are generally compatible functional modules for different outputs. Our research demonstrates that innate immunity in different species may have gone through an active phase of evolution to gain new functions and new effectors that could be added to existing functions without substantially changing them (e.g., STING’s ligand-binding domain and its acquired channel function). For adaptive immunity, the chimeric antigen receptor (CAR) system has properties that allow us to recombine different recognition modules and signaling modules to achieve optimal responses in immunotherapy for cancer or autoimmune diseases. Our results with STING suggest the possibility of inventing better innate immune recognition molecules by combining recognition and effector domains, as has been done for CARs and adaptive immunity. Can humans do better than nature for innate immune systems? We do not know, but now we can try. j

y

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

Gabriele Casirati

g

7 14

FINALIST

p

viruses, bacteria, and tumors (6, 7). However, the contribution of each downstream pathway to a given disease or therapeutic scenario is hard to define, owing to the lack of mechanistic understanding of the activation process. My colleagues and I therefore began to inquire how a single protein could mediate so many downstream responses. Is this achieved through a common downstream mechanism, or do different parts of STING activate individual functions? The molecular mechanisms for interferon activation downstream of STING are well understood, yet detailed mechanisms for noncanonical autophagy and the NLRP3 inflammasome remained unclear. Previous work found that proton leakage from acidic organelles can induce two distinct biological processes: noncanonical autophagy and NLRP3 inflammasome activation (8). Inspired by this finding, we constructed genetically encoded pH reporters and discovered that STING can induce proton leakage from Golgi bodies. When we searched for genes responsible for STING-induced proton leakage, we failed to identify a specific proton channel in a genome-wide screen, which prompted us to shift our focus to the STING protein itself. A careful analysis of the full-length STING structure led us to hypothesize that STING itself could serve as the channel that mediated proton leakage. The discovery of C53, a previously uncharacterized small molecule that binds to our proposed pore region, provided us with a distinctive tool to block this proposed pore area (9). As we predicted, treatment with C53 inhibited STING-induced pH increase and inhibited proton leakage, which blocked STING-induced noncanonical autophagy and NLRP3 inflammasome activation. Notably, C53 treatment did not affect STING phosphorylation, which is required for downstream interferon production; this indicated a decoupling of activation of this pathway from inflammasome and noncanonical autophagy signaling. To our knowledge, STING is the first immune sensor in humans that induces ion flow upon sensing danger signals (10). The concept of STING as a channel offers insights into our understanding of STING’s effects at both the cellular and physiological levels. Besides noncanonical autophagy and inflammasome activation, other downstream effects of STING activation at the cellular level may also be explained by its channel function. For instance, STING activation can induce cell death in various cell types, with unclear mechanisms. In T cells, STING disrupts endoplasmic reticulum calcium homeostasis (11). It is possible that STING’s channel function also mediates calcium re-

PRIZE ES SAY FINALIST

Gabriele Casirati

Stem cells in disguise Epitope editing can empower targeted cancer immunotherapies By Gabriele Casirati

T

,

science.org SCIENCE

y

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

y g

7 14-A

y

Dana-Farber Cancer Institute, Boston, MA, USA. Email: [email protected]

g

he discovery that our immune system not only provides constant surveillance against the development of neoplastic disorders but may also be artificially reprogrammed to recognize and eradicate established malignancies has transformed our perspective on oncology treatment and prognosis. The exploitation of monoclonal antibodies (mAbs) has paved the road to antigendirected targeting of cancerous cells and led to the development of chimeric antigen receptor (CAR) T cells. This revolutionary cellular therapy has demonstrated unprecedented efficacy in relapsed or refractory acute B-lymphoblastic leukemia, with up to 90% complete responses (1). These results were made possible by the high-level, uniform expression of the CD19 target antigen and the tolerability of B cell aplasia, whose sustained depletion is compatible with patients’ long-term survival. Despite widespread enthusiasm, attempts to expand CAR-T cell application to other cancers have encountered unavoidable obstacles, e.g., the lack of safely actionable tumor-associated antigens unshared by healthy tissues. One such case is acute myeloid leukemia (AML), for which target antigens [e.g., CD33, CD123, fms-like tyrosine kinase 3 (FLT3), c-type lectinlike molecule-1] are expressed either by hematopoietic stem and progenitor cells (HSPCs) or differentiated myeloid cells (2, 3). As opposed to B-phenotype leukemias and lymphomas, prolonged CAR-mediated depletion of myeloid lineages and HSPCs inevitably results in severe defects of innate immunity, increased risk of infection, and potentially long-term hematopoietic aplasia. Administration of immunotherapies as a bridge therapy before allogeneic hematopoietic stem-cell transplantation (HSCT) or the tunable depletion of CAR-T cells are both viable treatment options. Nonetheless, limiting the immunotherapeutic effect to a relatively short time window may severely affect the ability to eradicate residual tumoral cells. Elimination of the target antigen by means of CRISPR-Cas knock-out

(KO) in donor HSPCs has been proposed to avoid on-target killing (4–6). However, this approach is only feasible for genes dispensable for hematopoietic development, thus increasing the risk of tumor immune escape through antigen loss or down-regulation. Furthermore, nuclease-mediated KO is not safely amenable to multiplexing, which limits the opportunity to simultaneously target more than one antigen (7). My colleagues and I hypothesized that precise engineering of the target epitope within healthy HSPCs used for bone marrow transplantation may disrupt CAR-T cell binding without affecting the functionality of the modified protein, thereby fully preserving stem cell repopulation and differentiation capacity. Critically, this strategy enables us to target genes fundamental for leukemia survival, regardless of their expression or role in healthy hematopoiesis, which results in potent anti-leukemia efficacy with minimal on-target or off-tumor toxicity. Our group demonstrated the possibility of endowing HSPCs with selective resistance to CAR-T cells aimed at three cytokine receptors with widespread expression on AML cells by means of epitope base editing (8). FLT3 and KIT are tyrosinekinase receptors expressed in 93 and 85% of AML cases, respectively (9, 10). CD123 is the a subunit of the interleukin-3 receptor, a type I cytokine receptor found in >75% of AML cases. These genes are expressed at various stages of hematopoietic development, and their overexpression in AML is associated with a lower survival rate and increased risk of relapse (11, 12). Through epitope mapping library experiments, we found that single–amino acid substitutions were sufficient to disrupt therapeutic mAb recognition, despite preservation of surface expression. Accordingly, in vitro killing assays with CD28-costimulated CAR-T cells demonstrated that cells expressing epitope-edited variants were resistant to CAR-mediated killing and did not elicit T cell activation or degranulation, whereas wildtype FLT3, KIT, or CD123 were eliminated. Because epitope engineering can be achieved by point mutations, we reasoned that base editing (BE), which does not require doublestrand DNA breaks or template donor, could be a suitable and safer option compared to homology-directed repair (13). By electropor-

p

Gabriele Casirati received his MD degree from Università degli Studi di Milano and trained as a hematologist at Università Vita-Salute San Raffaele. From 2020 to 2023, he worked as a research fellow at the Dana-Farber Cancer Institute and received a PhD in molecular and translational medicine from Università Milano-Bicocca. He is currently a postdoctoral research fellow at Boston Children’s Hospital. His research focuses on new gene-editing approaches to enhance targeted immunotherapies for hematological malignancies. www.science.org/doi/10.1126/ science.adn9409

IMMUNOLOGY

R EFER ENC ES A ND N OT ES

1. S. A. Grupp et al., N. Engl. J. Med. 368, 1509 (2013). 2. A. Isidori, N. Daver, A. Curti, Front. Oncol. 11, 671252 (2021). 3. S. Haubner et al., Leukemia 33, 64 (2019). 4. O. Humbert et al., Leukemia 33, 762 (2019). 5. F. Borot et al., Proc. Natl. Acad. Sci. U.S.A. 116, 11978 (2019). 6. M. Y. Kim et al., Cell 173, 1439 (2018). 7. M. L. Leibowitz et al., Nat. Genet. 53, 895 (2021). 8. G. Casirati et al., Nature 621, 404 (2023). 9. F. Kuchenbauer et al., Haematologica 90, 1617 (2005). 10. L. R. Valverde et al., Ann. Hematol. 72, 11 (1996). 11. X. Gao et al., PLOS ONE 10, e0124241 (2015). 12. A. E. Bras et al., Cytometry B Clin. Cytom. 96, 134 (2019). 13. H. A. Rees, D. R. Liu, Nat. Rev. Genet. 19, 770 (2018). 14. N. M. Gaudelli et al., Nat. Biotechnol. 38, 892 (2020).

g y

10.1126/science.adn9409

y g y ,

SCIENCE science.org

tools developed in our work can increase the therapeutic index of cancer immunotherapies and enable long-term antitumor maintenance. Additionally, through restriction of on-target activity to leukemia cells, epitope editing can reduce the antigen burden to which CAR-T cells are exposed and decrease undesired CAR-T cell stimulation and exhaustion. Finally, epitope editing can easily be multiplexed to enable combination therapies while avoiding overlapping toxicities. Though several questions with regard to the clinical feasibility of complex genome-engineered cellular therapies remain unanswered, we believe that epitope engineering is a key achievement for the development of safer and more effective immunotherapies when on-target or off-tumor toxicities are the key limiting factor to successful translation. j p

ating single-guide RNA with adenine base editing (sgRNA+ABE) variants, we achieved efficient epitope editing (~85%) in human CD34+ HSPCs, with minimal toxicity and a favorable off-target profile. Contrary to previous observations with nuclease-mediated editing, BE efficiencies were similar in bulk and more primitive, HSC-enriched subsets. Critically, epitope modification for FLT3, CD123, and KIT was possible with the same ABE, in principle allowing multiplexing (14). We next assessed the preservation of the edited receptors’ functionality (e.g., ligand affinity, activation of signal transduction, proliferation, transcriptional and phospho-proteomic responses) and found profiles comparable to those of wild-type receptors, both at baseline and upon ligand stimulation. Edited HSPCs were resistant to CAR- or mAb-mediated killing in vitro and confirmed that these stem cells “in disguise” could survive even when directly exposed to immunotherapies. Severe combined immunodeficient (SCID)–repopulating experiments in NBSGW mice showed preserved repopulation and multilineage differentiation capacity, both in primary and secondary recipients. To assess whether CAR-T cells could selectively eliminate AML cells while sparing epitope-edited hematopoiesis—one of the “holy grails” of AML immunotherapy— we sequentially engrafted NBSGW mice with HSPCs and human patient-derived AML xenografts (PDXs). Mice treated with FLT3-CAR achieved complete AML eradication. At the same time, we observed substantial depletion of CD19+ hematopoietic subsets (pre-B, pro-B), granulocytes, granulocyte-monocyte progenitors, and lymphoid-primed multipotent progenitors only in the adeno-associated virus integration site 1 (AAVS1) control group, whereas mice engrafted with FLT3-BE HPSC were protected. CAR-T cells exposed to FLT3-BE hematopoiesis displayed lower programmed cell death protein 1 expression compared to AAVS1-BE, suggesting decreased activation and exhaustion. With proof-of-concept data for our approach, we tested whether epitope editing allows multiplexing. Combined dual FLT3+CD123 BE was able to protect hematopoietic lineages in vivo when exposed to FLT3+CD123 CAR-T cells, which in turn could eradicate PDXs resistant to FLT3-targeting alone, underscoring the potential of multitarget immunotherapies. Our studies provided evidence that tumor-associated antigens shared by normal tissue can be safely targeted by precisely modifying the target epitope in healthy cells, which endows them with selective resistance and generates an artificial leukemia-restricted antigen. The innovative

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

7 14-B

PRIZE ES SAY FINALIST

Carla Nowosad

Who goes there? How B cells assess risk in the intestine By Carla Nowosad

M

,

science.org SCIENCE

y

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

y g

7 14-C

y

Department of Pathology, NYU Langone, Grossman School of Medicine, New York University, New York, NY, USA. Email: [email protected]

g

any students have sat through immunology classes and listened to animated lecturers who touted the immune system as the body’s natural army, poised and ready to fight off harmful invaders. Although this well-worn analogy holds some truth, I have come to learn that it is far from the whole picture. The reality is that immune cells encounter harmless foreign or selfentities far more often than dangerous pathogens. Therefore, the immune system is better compared with border control, in which immune cells encounter stimuli, assess them, and communicate with each other to decide whether the newcomer may be integrated into the local microenvironment. Although we know that nonpathogenic encounters are far from ignored by the immune system, the mechanisms used to turn a molecular blind eye to harmless stimuli are not well understood. A perfect example of this analogy in action is the intestine: a barrier organ under constant bombardment by foreign antigens from food, so-called friendly bacteria from the microbiota, and the occasional harmful pathogen (1). Our internal border agent, the B cell, contributes to intestinal peacekeeping through the secretion of massive amounts of immunoglobulin A (IgA) antibodies—the most of any anatomical site. Antibodies play well known roles in defense but also play vital, less-understood roles in homeostasis (2). When we are healthy, IgA accumulates in the gut lumen, where it controls the community structure of the microbiota (3). Accurate immune regulation of the microbiota is an area of intense research interest because a dysregulated microbiome plays a key role in many human disorders, including obesity, depression, autoimmune diseases, and cancer. During an infection, B cells undergo an iterative process to edit and improve their surface B cell receptor (BCR), a biological blueprint for secreted antibody. BCR evolution takes place in an organized B cell “training camp” called the germinal center (GC), where improved cells are selected

and expanded to form a clone of many identical high-affinity cells specialized to neutralize the incoming pathogen (4). GCs typically form exclusively during infection and contract upon pathogen clearance— except in the intestine, where they persist even in the absence of any observable inflammation (3, 5). For decades, researchers observed these unusual steady-state GCs but concluded that these structures were reserved for pathogen-driven responses and did not supply the healthy gut with microbiotaregulating IgA. Instead, researchers stated that homeostatic IgA formed outside of GCs and was low affinity and poly specific. Most of the evidence supporting this conclusion was garnered from mouse models that lacked essential components of the GC reaction (for example, T cells or positive selection signals) that maintained intact IgA responses (6). Even so, it was difficult to conceptualize how a pool of low-affinity, poly-specific IgA could exert such fine control over the complex microbiome. By contrast, human evidence strongly supported a GC origin for microbiota-regulating IgA (7, 8), which led me to decide to take a fresh look at whether steady-state GCs were functional—this time, without disrupting natural responses. To observe the evolution of homeostatic GCs in vivo with minimal experimental manipulation, I used the Confetti allele, a fate-mapping tool, to turn GCs technicolor and watch clonal selection proceed in the living mouse (9). In this system, each GC B cell is randomly assigned a different fluorescent color. If a GC B cell is successfully selected and expanded, its daughter cells retain their color stamp. Over time, as an expanded clone “wins” to overtake its GC, a once-technicolor GC instead becomes single-colored. When my colleagues and I looked for selection in gut-draining lymphoid organs, we could see that the intestine did contain single-color “winner” GCs that underwent selection at similar rates to those induced during infection (5). Winning in the GC occurs when strong selection pressure is applied by a driving antigen, which suggests the existence of an intestinal feedback loop, in which specific antigens from the environment drive the specificity of the IgA pool that is secreted into the gut lumen to modify microbiota composition.

p

Carla Nowosad received undergraduate degrees from the University of Warwick and a PhD from the Francis Crick Institute, National Institute for Medical Research Mill Hill. After completing her postdoctoral fellowship at the Rockefeller University, she started her laboratory in the Department of Pathology and the Translational Immunology Center at New York University Grossman School of Medicine in late 2021. Her research focuses on how B cells make decisions in the complex microenvironment of the intestine. www.science.org/ doi/10.1126/science.adn9414

IMMUNOLOGY

1. A. J. Macpherson, B. Yilmaz, J. P. Limenitakis, S. C. GanalVonarburg, Annu. Rev. Immunol. 36, 359 (2018). 2. A. Fleming, T. Castro-Dopico, M. R. Clatworthy, Scand. J. Immunol. 95, e13139 (2022). 3. O. Pabst, C. R. Nowosad, Semin. Immunol. 69, 101806 (2023). 4. G. D. Victora, M. C. Nussenzweig, Annu. Rev. Immunol. 40, 413 (2022). 5. C. R. Nowosad et al., Nature 588, 321 (2020). 6. A. Reboldi, J. G. Cyster, Immunol. Rev. 271, 230 (2016). 7. F. Barone et al., Gastroenterology 140, 947 (2011). 8. J. Benckert et al., J. Clin. Invest. 121, 1946 (2011). 9. J. M. Tas et al., Science 351, 1048 (2016). 10. S. Brugiroux et al., Nat. Microbiol. 2, 16215 (2016). 11. T. Rollenske et al., Nature 598, 657 (2021). 12. K. Moor et al., Nature 544, 498 (2017). 13. G. P. Donaldson et al., Science 360, 795 (2018). 10.1126/science.adn9414

g y y g y ,

SCIENCE science.org

R EFER ENC ES AN D N OT ES

p

To understand whether microbiotaspecific clones were predominantly winning in steady-state GCs, we painstakingly recovered hundreds of individual winner GCs, sequenced their BCRs, and produced recombinant versions of their antibodies. As we hoped, heavily selected winner clones did bind specifically to the microbiota and did not display considerable polyreactivity, which was previously implicated as the principal mode of binding for homeostatic IgA. We were curious to know whether microbiota-specific clones showed any bias for binding certain bacteria among the melee, which proved challenging to study given the sheer scale of the natural microbiome’s diversity. To narrow our focus, we instead isolated winner clones from mice with simplified microbiomes consisting of only 12 bacterial strains (10). With this simplified setup, we were able to match clones to their individual bacterial binding partners. Further, we observed that the BCR mutations acquired in the GC conferred improved binding toward that specific bug—bona fide experimental evidence for functional GC responses in the absence of inflammation. Contrary to the existing paradigm, this finding suggested that there are not independent pathways to categorize harmful versus tolerated bacterial encounters. Rather, intestinal GCs function as a common niche to evolve homeostatic and neutralizing antibody responses, with the in-built ability to mode-switch which antibody they select for, according to the level of perceived danger. Our discovery that the immune system crafts highly evolved antibody responses to maintain homeostasis has led to a number of exciting research questions. At present, it remains unclear precisely which bacteria are capable of driving GC-evolved IgA production, or whether all commensal bacteria will elicit GC-IgA responses in particular contexts (such as during bacterial outgrowth). Additionally, there is mounting experimental evidence that details how IgA acts on the bacteria it binds to, including the classic clearance response as well as several less familiar, regulatory mechanisms (such as modulating bacterial growth, altering bacterial metabolism, and regulating bacterial adherence) (11–13). Understanding how GCs select for unconventional antibody binding properties offers an opportunity to consider the “bugs-as-drugs” movement through an alternative immune lens. Dynamic alteration of bacterial strains in the microbiota through local production of IgA with desirable, modulatory functions could divert disease progression across a range of human diseases.

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

7 14-D

RESEARCH IN S CIENCE JOU R NA L S

Edited by Michael Funk

p g y Zebras in Serengeti National Park have both competitive and facultative interactions with other grazing species

ECOLOGY

Grazing in turns

BATTERIES

SCIENCE science.org

PHOTO: JASON DONALDSON

Science p. 739, 10.1126/science.adh5115

PHOTOCHEMISTRY

A floating melt Modern chemistry relies on vast quantities of toxic solvents. Conducting reactions in water could substantially mitigate the waste stream, but the low aqueous solubility of many reagents remains a stubborn barrier. This problem is especially pronounced

for photochemistry, which is not efficient in cloudy suspensions. Tian et al. report that certain pairs of solid reactants can form a lower-melting eutectic mixture that floats as a liquid on the water surface and is conveniently exposed for photochemistry. A segmented flow system effectively adapted the method to gram scale. —JSY Science p. 750, 10.1126/science.adl3092

PHYSICAL CHEMISTRY

Highly exciting roaming Roaming is an important reaction mechanism that does not follow a minimum energy pathway and thus bears specific

signatures in product state distributions. Over the past two decades, much experimental evidence has been found for roaming in both ground and first excited states of polyatomic molecules. Given the complex electronic structure and multiple product channels, roaming should be more common in highly excited electronic states, but experimental evidence has been lacking. Li et al. demonstrated that roaming is important for sulfur dioxide photodissociation from highly excited states. The experimental data were supported by full-dimensional dynamics simulations. The present work

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

7 15

,

In most solid electrolytes, the conduction pathways have a single coordination geometry. Han et al. designed electrolytes based on a Li7Si2S7I chemistry with ion arrangements similar to those in intermetallic systems. This leads to anion packing that alternates between hexagonal close packed structures and sheared face-centered cubic-like motifs to accommodate sulfur and iodine complexes, analogous to the structure of nickel-zirconium. The resulting materials have an interconnected set of

Science p. 782, 10.1126/science.adg0744

y

Packing for lithium ion conduction

15 crystallographically separate lithium sites with a variety of geometries and anion coordination, providing a diverse path of conduction pathways for lithium ions and thus high conductivity. —MSL

main herbivores within the migration: zebra, wildebeest, and gazelle. They found that competition pushes zebra ahead of the other larger species, wildebeest. Herbivory by wildebeest then facilitates the development of newer growth, which gazelle take advantage of as they trail the other two. —SNV

y g

T

he Serengeti migrations consist of millions of herbivores sharing food resources. For decades, there has been interest in understanding how these resources are divided up and what drives the division. Anderson et al. used ecological and animal movement data, fecal DNA, and “natural experiments” driven by rain and fire to characterize the divisions among the three

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

confirms that roaming should be expected to be a fairly common phenomenon in dynamics of highly excited states. —YS Science p. 746, 10.1126/science.adn3357

for marine plankton assemblages, scaling in consumption, growth, and predation gives rise to the higher levels of biological organization that underpin ecosystem function. —CA

IN OTHER JOURNALS Edited by Caroline Ash and Jesse Smith

Science p. 777, 10.1126/science.adk6901

CORONAVIRUS

SARS-CoV-2 human challenge study

Improving vascular access in the brain

Sci. Immunol. (2024)

10.1126/scirobotics.adh0298

OCEAN OXYGEN ECOLOGY

Size says it all

Science p. 727, 10.1126/science.adh4893

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

Genetic mutations are an important cause of deafness. Mutations in the Otoferlin (OTOF) gene cause autosomal-recessive deafness 9 (DFNB9). Adenoassociated virus (AAV)–mediated gene therapy has shown positive results in rodents, but safety and efficacy in humans had not been tested. Qi et al. report the first exploratory clinical trial in two children carrying OTOF mutations. Cochlear administration of dual AAV-OTOF vectors restored hearing in the injected ear in one patient and significantly improved hearing in the other individual a month after treatment. The exploratory trial was preceded by safety and efficacy work in mice and nonhuman primates. These results pave the way for the application of this approach to other forms

Adv. Sci. (2024) 10.1002/advs.202306788

NEUROSCIENCE

The eyes have it Complex object features are processed in high-level visual areas such as the inferior temporal cortex. Azadi et al. hypothesized that neural activity in this brain region plays a causal role in controlling eye movements. Using muscimol, a potent neural silencer, to reversibly inactivate clusters of face-selective neurons in macaques, the authors investigated the resulting effects on eye movements while the monkeys were freely viewing faces and other objects. Inactivation of face-selective neurons altered the overall structure of eye movements. The animals still found faces in the scene but neglected the eye contralateral to the science.org SCIENCE

,

7 16

The amount of oxygen that can be dissolved in the ocean is an inverse function of its temperature, a fact with worrisome implications for global warming. How oxygen is distributed in the ocean also matters for the health of the organisms that need it. Moretti et al. have shown that oxygen availability in the shallow subsurface waters of the tropical North Pacific actually rose during the Paleocene-Eocene Thermal Maximum (PETM), a warm interval that has been considered a geologic analog to anthropogenic warming. This effect may have helped to prevent a mass extinction during the PETM. —HJS

Improving hearing in children

of deafness caused by genetic mutations. —MMa

y

The size of an individual is a product of physiological constraints such as area-to-volume ratio relationships or the fractal scaling of circulatory systems. Scaling patterns throughout ecosystems are well documented and tend to obey a 3/4 exponent, yet the mechanisms by which they emerge are poorly understood. Starting with the premise that micro-level scaling constrains macro-ecology and even evolution, Wickman et al. used a mechanistic eco-evolutionary model to explore relationships such as prey and predator biomass and community productivity. The authors conclude that

Distributed effects

GENE THERAPY

y g

10.1126/sciimmunol.adj9285

y

Sci. Robot. (2024)

g

Acute ischemic stroke can lead to long-term debilitating conditions and death, but if patients are treated promptly, their chances of survival and recovery are greatly increased. Due to the complex network and architecture of blood vessels in the brain, there is an unmet need for approaches that enable effective navigation of catheters to remove blood clots after stroke. Dreyfus et al. describe a highly dexterous articulated continuum robot that is magnetically steered to improve the navigation of catheters in tortuous blood vessels. The device enabled successful endovascular navigation from the aorta to millimeter-sized cranial arteries in vivo, demonstrating its potential for nontraumatic access to occluded vessels in the brain. —AM

p

The COVID-19 pandemic has provided great insight into how humans fight respiratory virus infections, but the earliest stages of severe acute human respiratory syndrome coronavirus 2 (SARS-CoV-2) infection remain poorly characterized. Wagstaffe et al. conducted a human SARSCoV-2 infection challenge study enabling the analysis of the innate and adaptive immune responses during the early postexposure period. Of 34 seronegative young adults inoculated, 18 developed sustained infections, which were accompanied by a systemic interferon-dominated inflammatory response preceding that in nasal lining fluid. Modeling of the immune response identified CD8+ T cell and early mucosal immunoglobulin A responses as being strongly associated with viral control, suggesting that vaccines that optimally induce these responses may help to reduce transmission. —CO

MEDICAL ROBOTS

R ES E ARCH

ALSO IN SCIENCE JOURNALS IMMUNOLOGY

The many functions of autoantibodies

A balancing act in the liver

Science p. 719, 10.1126/science.adk3468

ANTIBIOTICS

Advantages of preorganization Many small-molecule antibiotics target the bacterial ribosome, and there are many corresponding ribosome modifications that confer resistance by reducing the binding affinity of these molecules. One way to boost their affinity would be to lock these molecules in an ideal conformation for binding. Wu et al. used insights from structural analysis of ribosomes bound to previously developed antibiotics to design a conformationally restricted molecule, dubbed cresomycin, that adopts the exact conformation necessary for binding. Computational, structural, and biochemical experiments confirmed the expected binding mode. Cresomycin potently inhibits Gram-negative and -positive bacteria, including multi-drugresistant strains, both in vitro and in a mouse infection model. —MAF Science p. 721, 10.1126/science.adk8013

PLANT SCIENCE

Immune receptors oligomerize

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

MOLECULAR BIOLOGY

Keeping telomerase in check The enzyme telomerase synthesizes telomeric repeats at the ends of linear chromosomes, allowing the shelterin protein complex to bind and protect the ends from the DNA damage response. If telomerase were to add telomeres to broken DNA ends, then genes distal to the break could be lost. Kinzig et al. found that human telomerase can act at broken DNA ends, thereby threatening genome integrity (see the Perspective by Arnoult and Cech). They show that this genome-destabilizing aspect of telomerase is averted by the ATR kinase component of the DNA damage response. —DJ Science p. 763, 10.1126/science.adg3224; see also p. 702, 10.1126/science.adn7791

IMMUNOLOGY

B cell marker turned adjuvant Upon infection, multicellular structures known as germinal centers (GCs) facilitate the maturation of B cells and the formation of a pool of antibodies that recognize multiple epitopes from a pathogen. The lipid globotriaosylceramide (Gb3) is highly expressed by B cells found within GCs, but its function was previously unknown. Using mouse models, Sharma et al. found that elevated levels of Gb3 in B cells promoted the formation of high-affinity antibodies; augmented responses to weaker, subdominant epitopes; and promoted antibody responses that were protective between different strains of influenza virus. Moreover, exogenous Gb3

Interspecies protein structure Many plant pathogens release cell wall–degrading enzymes to facilitate access into the cell. When cell walls degrade, long-chain wall fragments suppress immune responses, and short-chain fragments can enhance plant immunity. Using biochemical assays and protein structural analysis, Xiao et al. established how a plant polygalacturonase-inhibiting protein (PGIP) interacts with a fungal polygalacturonase enzyme (see the Perspective by Thynne and Kobe). They found that this interaction alters how the fungus binds to cell wall polysaccharides and influences substrate preferences and the types of degradation products generated to increase immunity. Proteins from two different species combine to make an enzymatic complex with different behavior from either alone. Additionally, the authors demonstrate the potential for engineering by modifying a related but inactive PGIP to become active. —MRS Science p. 732, 10.1126/science.adj9529; see also p. 707, 10.1126/science.adn8306

,

Plant immune responses are frequently detected by intracellular receptors known as nucleotide-binding leucine-rich repeat receptors (NLRs). Some NLRs work in pairs to transduce signals and initiate cell death response to prevent pathogen spread. Yang et al. found that some NLR pairs heterodimerize and are kept in this inactive state by receptor-like kinases such as BRI1-ASSOCIATED RECEPTOR KINASE 1 (BAK1). On activation, the heterodimers can oligomerize, which is required to initiate downstream

PLANT IMMUNITY

y

7 17-B

Science p. 720, 10.1126/science.adg0564

y g

The storage of nutrients after food consumption requires an appropriate balance, with the liver helping to direct incoming glucose toward the production of glycogen and/or fat. An excess of fat production can contribute to the development of fatty liver disease and inflammation, so this process requires careful regulation. Chen et al. found that the metabolite uridine diphosphate glucose (UDPG), which is produced during glycogenesis, is transported to the Golgi apparatus, where it blocks a protease called S1P and thereby inhibits lipogenesis. In both human organoids and a mouse model of nonalcoholic fatty liver disease, treatment with UDPG promoted glycogenesis and reduced the accumulation of

acted as a vaccine adjuvant in mice. —SHR

y

METABOLISM

Science p. 718, 10.1126/science.adi3332

responses. The authors provide mechanistic insight into how paired NLRs activate plant immunity. —MRS

g

Science p. 705, 10.1126/science.abn1034

lipids, suggesting a potential for therapeutic applications. —YN

p

Autoantibodies are usually known for their role in mediating autoimmune diseases, but evidence increasingly indicates that they can have protective as well as pathological roles in human health. In a Perspective, Jaycox et al. discuss how autoantibodies may confer an underappreciated level of heterogeneity in human biology, which could have diverse impacts in, for example, cancer, neurodegeneration, and infectious diseases. They also highlight the need for “autoantibody-wide association studies” reminiscent of such studies in genetics. By better understanding the diverse functions of autoantibodies, new biomedical paradigms might be revealed that could help guide drug development. —GKA

Edited by Michael Funk

MEDICINE

Strategies to cure HIV The nature of HIV-1 infection means that there is a latent reservoir of infected cells. Although antiretroviral therapy can halt viral evolution and disease progression, it is not a cure because the latent reservoir is not targeted. Because just one latent HIV-1–infected cell can give rise to disease, eliminating the reservoir is a major focus for developing an HIV cure. In a Perspective, Siliciano and Siliciano discuss current efforts science.org SCIENCE

to target the latent reservoir, drawing from examples of people who have been cured of HIV, for example, through hematopoietic stem cell transplantation. The authors discuss the importance of and challenges raised by having to extensively reduce the latent reservoir. —GKA Science p. 703, 10.1126/science.adk1831

CATALYSIS

Helping acid help you

ARCHAEOLOGY

Making rope in the European Paleolithic

Sci. Adv. (2024) 10.1126/sciadv.adh5217

CELL BIOLOGY

CANCER

Radio-sensitizing glioblastoma

Restoring epigenetic slips in cancer DNA methylation silences transcription and regulates cellular differentiation and fate determination. For myelodysplastic syndromes and acute myeloid leukemia, DNA-hypomethylating agents are a standard of care. However, focal hypomethylation due to deficiency in the de novo methyltransferase activities of DNMT3A and/or DNMT3B also contributes to disease. Li et al. defined and corrected focal hypomethylation patterns in the bone marrow of mice with various DNMT3A/B–deficient genotypes. Their study provides a potential strategy for reactivating de novo DNMT activity to correct focal hypomethylation in patients with clonal hematopoiesis or myeloid malignancy. —ASH

,

SCIENCE science.org

Sci. Signal. (2024) 10.1126/scisignal.abl5880

CANCER THERAPEUTICS

y

The standard of care for glioblastoma includes treatment with radiation therapy (RT) and temozolomide; however, long-term survival remains rare. To improve radiation therapy, Chen et al. investigated the use of AZD1390, a brain-penetrant inhibitor of the kinase ATM, and a radiosensitizer in combination with RT in mouse models. They found that this combination was beneficial for a cohort of patient-derived xenografts that were TP53 mutant, suppressing homologous recombination efficiency and extending survival. This represents a potential therapy that could benefit

The neuropeptide relaxin-3 binds to its receptor, RXFP3, to mediate functions such as anxiety, feeding, and cognition. Relaxin-3 is composed of an A-chain linked to a B-chain, with the A-chain scaffolding the B-chain for receptor presentation. Jayakody et al. characterized the binding properties and functional responses of RXFP3 to relaxin-3 and analogs that were modified with hydrocarbon “staples.” Their results showed that two such analogs stimulated biased signaling by RXFP3 through G proteins, in contrast to the unbiased relaxin-3. These findings may help in the design of drugs targeting RXFP3. —JFF

Science p. 771, 10.1126/science.adj3549

y g

A staple bias for relaxin-3

Graphene and the family of two-dimensional materials known as MXenes have important mechanical and electrical properties that make them potentially useful for making flexible energy storage devices, but it is challenging to assemble flakes of these materials into ordered, free-standing sheets. Yang et al. assembled films of graphene using the MXene Ti3C2Tx to provide bridging between the sheets. A key aspect of the process is the use of nanoconfined water, which mitigates wrinkling and aids with better particle alignment. In addition to high strength, the films have high electronic conductivity, making them potentially suitable for electromagnetic interference shields, current collectors, or electrodes. —MSL

y

Archaeologists have long been puzzled about the function of so-called perforated batons found at sites from the European Upper Paleolithic. Usually made of reindeer antler, they are commonly viewed as ritual objects or symbols of status. Conard and Rots examined a 35,000-year-old perforated baton made of mammoth ivory from Hohle Fels Cave in southwestern Germany using residue and use-wear analysis. They conducted fiber-making experiments with a replica and found that that the perforations served as effective guides for raw fibers, making it easier to make a strong, elastic rope than simply twisting fibers by hand. —MSA

Dense films of twodimensional materials

g

Science p. 757, 10.1126/science.adk4902

Sci. Transl. Med. (2024) 10.1126/scitranslmed.adj5962

LAYERED MATERIALS

p

Conventional Brønsted acids are among the oldest catalysts and are still the most effective. Because they operate by simple proton transfer, however, their accelerating effects are not easily amenable to improvement through structural modification. Westendorff et al. report that the application of an electrochemical potential can tilt protonation pre-equilibria and thereby accelerate an acid-catalyzed alcohol dehydration by up to 100,000-fold. The strategy was also effective in accelerating a Friedel-Crafts acylation, suggesting that it may have broad applicability to acid-catalyzed chemistry. —JSY

patients with TP53-mutant glioblastoma and warrants further study. —DLH

Sci. Adv. (2021) 10.1126/sciadv.adk8598 16 FEBRUARY 2024 • VOL 383 ISSUE 6684

7 17-C

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

confirms that roaming should be expected to be a fairly common phenomenon in dynamics of highly excited states. —YS Science p. 746, 10.1126/science.adn3357

for marine plankton assemblages, scaling in consumption, growth, and predation gives rise to the higher levels of biological organization that underpin ecosystem function. —CA

IN OTHER JOURNALS Edited by Caroline Ash and Jesse Smith

Science p. 777, 10.1126/science.adk6901

CORONAVIRUS

SARS-CoV-2 human challenge study

Improving vascular access in the brain

Sci. Immunol. (2024)

10.1126/scirobotics.adh0298

OCEAN OXYGEN ECOLOGY

Size says it all

Science p. 727, 10.1126/science.adh4893

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

Genetic mutations are an important cause of deafness. Mutations in the Otoferlin (OTOF) gene cause autosomal-recessive deafness 9 (DFNB9). Adenoassociated virus (AAV)–mediated gene therapy has shown positive results in rodents, but safety and efficacy in humans had not been tested. Qi et al. report the first exploratory clinical trial in two children carrying OTOF mutations. Cochlear administration of dual AAV-OTOF vectors restored hearing in the injected ear in one patient and significantly improved hearing in the other individual a month after treatment. The exploratory trial was preceded by safety and efficacy work in mice and nonhuman primates. These results pave the way for the application of this approach to other forms

Adv. Sci. (2024) 10.1002/advs.202306788

NEUROSCIENCE

The eyes have it Complex object features are processed in high-level visual areas such as the inferior temporal cortex. Azadi et al. hypothesized that neural activity in this brain region plays a causal role in controlling eye movements. Using muscimol, a potent neural silencer, to reversibly inactivate clusters of face-selective neurons in macaques, the authors investigated the resulting effects on eye movements while the monkeys were freely viewing faces and other objects. Inactivation of face-selective neurons altered the overall structure of eye movements. The animals still found faces in the scene but neglected the eye contralateral to the science.org SCIENCE

,

7 16

The amount of oxygen that can be dissolved in the ocean is an inverse function of its temperature, a fact with worrisome implications for global warming. How oxygen is distributed in the ocean also matters for the health of the organisms that need it. Moretti et al. have shown that oxygen availability in the shallow subsurface waters of the tropical North Pacific actually rose during the Paleocene-Eocene Thermal Maximum (PETM), a warm interval that has been considered a geologic analog to anthropogenic warming. This effect may have helped to prevent a mass extinction during the PETM. —HJS

Improving hearing in children

of deafness caused by genetic mutations. —MMa

y

The size of an individual is a product of physiological constraints such as area-to-volume ratio relationships or the fractal scaling of circulatory systems. Scaling patterns throughout ecosystems are well documented and tend to obey a 3/4 exponent, yet the mechanisms by which they emerge are poorly understood. Starting with the premise that micro-level scaling constrains macro-ecology and even evolution, Wickman et al. used a mechanistic eco-evolutionary model to explore relationships such as prey and predator biomass and community productivity. The authors conclude that

Distributed effects

GENE THERAPY

y g

10.1126/sciimmunol.adj9285

y

Sci. Robot. (2024)

g

Acute ischemic stroke can lead to long-term debilitating conditions and death, but if patients are treated promptly, their chances of survival and recovery are greatly increased. Due to the complex network and architecture of blood vessels in the brain, there is an unmet need for approaches that enable effective navigation of catheters to remove blood clots after stroke. Dreyfus et al. describe a highly dexterous articulated continuum robot that is magnetically steered to improve the navigation of catheters in tortuous blood vessels. The device enabled successful endovascular navigation from the aorta to millimeter-sized cranial arteries in vivo, demonstrating its potential for nontraumatic access to occluded vessels in the brain. —AM

p

The COVID-19 pandemic has provided great insight into how humans fight respiratory virus infections, but the earliest stages of severe acute human respiratory syndrome coronavirus 2 (SARS-CoV-2) infection remain poorly characterized. Wagstaffe et al. conducted a human SARSCoV-2 infection challenge study enabling the analysis of the innate and adaptive immune responses during the early postexposure period. Of 34 seronegative young adults inoculated, 18 developed sustained infections, which were accompanied by a systemic interferon-dominated inflammatory response preceding that in nasal lining fluid. Modeling of the immune response identified CD8+ T cell and early mucosal immunoglobulin A responses as being strongly associated with viral control, suggesting that vaccines that optimally induce these responses may help to reduce transmission. —CO

MEDICAL ROBOTS

Infrared shows stars being ripped apart

A

tidal disruption event (TDE) occurs when a star is ripped apart by the gravitational field of a supermassive black hole. Most TDEs are identified in soft x-ray or optical observations, which are not sensitive to events obscured by dust. Masterton et al. have searched for TDEs in a preexisting survey taken at infrared wavelengths, which are less affected by dust. They identified 12 TDEs, of which 11 do not appear in optical data, demonstrating that there is a population of dust-obscured TDEs that have been missed by previous surveys. The authors calculate that this population roughly doubles the TDE occurrence rate. —KTS

Nat. Mater. (2024) 10.1038/s41563-023-01783-y

WOMEN IN SCIENCE

Surprisingly, aquatic sponges, which are among the simplest of multicellular sessile organisms, can move their entire bodies. This movement occurs without a nervous system or muscles and pumps water through canals where filter feeding occurs. Ruperti et al. explored how the body of the non-model pond sponge Spongilla lacustris deflates and inflates to pump water. Starting with the known physiological role of nitric oxide signaling and using SCIENCE science.org

Curr. Biol. (2024) 10.1016/j.cub.2023.12.021

BATTERIES

Adv. Mat. (2024) 10.1002/adma.202310667

Caffeine enhances zinc batteries

2D MATERIALS

Zinc is a promising material for rechargeable batteries because it is more readily available than lithium and can also be processed without requiring an inert environment. However, it suffers from

Stacking graphene layers on top of each other at an angle has been shown to lead to a multitude of unconventional states. The properties of these materials

Imaging reconstruction

Genesis (2023) 10.1002/dvg.23578

16 FEBRUARY 2024 • VOL 383 ISSUE 6684

7 17

,

Just relax, stay cool

y

PHYSIOLOGY

dendrite formation and corrosion during cycling, and these issues are exacerbated as the size of the battery is reduced. Qu et al. added caffeine to the surface of a cross-linked polyacrylamide hydrogel electrolyte. They found that the caffeine passivated the surface of the zinc anode, thus preventing corrosion. Further, it helped form zinc complexes that aided in reversibility during cycling. The authors were able to fabricate on-chip Zn||MnO2 microbatteries that showed 80% depth of discharge over 700 cycles. —MSL

y g

Proc. Natl. Acad. Sci. U.S.A. (2024) 10.1073/pnas.2309906121

three-dimensional imaging and thermal proteome profiling, the authors identified an inflammatory state involving peptides called granulins that prompted deflation. Granulins, in addition to macrophage migration inhibitory factor, among other biological active peptides, initiated actomyosin contraction reminiscent of the vertebrate vascular system, indicating long evolutionary conservation of a relaxant-inflammatory pathway. —CA

Tunicates are a diverse group of marine invertebrate animals that are the closest relatives to vertebrates. Because of their simple embryology, anatomy, and genomes, they offer powerful model systems with which to explore vertebrate evolution and development. The history of tunicate research traces back almost 200 years. Nydam et al. have undertaken meticulous research into the contribution of the female scientists who pioneered the research on tunicates. From their work, we learned that at the beginning of the last century, Edith Sumner Byxbee studied pelagic tunicates collected in the tropical Pacific Ocean by the US Fish Commission ship “Albatross.” Françoise Monniot collaborated and succeeded her husband Claude Monniot in enriching our understanding of the taxonomy and distribution of ascidians, the largest subgroup of tunicates, around the world, especially in the tropics and the deep sea. These are just two of the many women contributors to the field who are little known to those working on these model organisms today. —DJ

y

Tidal disruption events revealed by infrared imaging show that they are more common than previously believed.

g

Heroines of tunicate research

Astrophys. J. (2024) 10.3847/1538-4357/ad18bb

inactivation hemisphere. These results indicate that high-level visual processing can determine the input to the visual system by steering the eyes toward the relevant parts of the visual field. —PRS

p

depend sensitively on the size of the “twist” angles between the layers. In addition, the atoms in such multilayers may slightly rearrange to assume a more energetically favorable configuration. To image the stacking configuration on a local level, Craig et al. used interferometric four-dimensional scanning transmission electron microscopy (4D-STEM). The researchers focused on graphene trilayers, studying a wide range of twist angles, as well as effects of strain. The detailed understanding of lattice reconstruction gained from this study highlights the ability of 4D-STEM to characterize complex multilayers. —JSt

BLACK HOLE PHYSICS

RES EARCH

RESEARCH ARTICLE SUMMARY



IMMUNOLOGY

The lipid globotriaosylceramide promotes germinal center B cell responses and antiviral immunity Pankaj Sharma†, Xiaolong Zhang†, Kevin Ly, Yuxiang Zhang, Yu Hu, Adam Yongxin Ye, Jianqiao Hu, Ji Hyung Kim, Mumeng Lou, Chong Wang, Quinton Celuzza, Yuji Kondo, Keiko Furukawa, David R. Bundle, Koichi Furukawa, Frederick W. Alt, Florian Winau*

INTRODUCTION: Viral infections, such as influ-

CD19 translocation CD19

BCR Gb3

Akt

FDC

PI3K TFH cell

GC B cell selection

p

MHC-II/TCR

y

FOXO1 degradation

p

y g

CD81

Antigen presentation

y

Light zone

g

Dark zone

RESULTS: We found that Gb3 specifically binds to the plasma membrane glycoprotein CD19, disrupting the interaction with its chaperone CD81 and allowing CD19 to recruit and activate kinases associated with BCR signaling. This signaling pathway, which was dependent on the presence of Gb3, triggered the degradation of the transcription factor FOXO1. Reduction in FOXO1 caused gene expression changes that allowed GC B cells to transition from the dark zone, where somatic hypermutation happens, to the light zone, where B cell selection occurs. Thus, the Gb3-dependent CD19 translocation mechanism facilitated BCR signaling and GC B cell dynamics that overall promoted affinity maturation of B cells. Conversely, in the absence of Gb3, CD19 remained in complex with CD81 and failed to translocate to the BCR, consequently abrogating downstream signaling, FOXO1 degradation, and GC B cell cycling. As a result, without Gb3, B cells were not able to undergo affinity maturation and failed to produce high-affinity antibodies. Furthermore, the lipid Gb3 supported the surface expression of MHC-II levels on GC B cells, which facilitated the activation of TFH cells. Through this mechanism, Gb3 could promote selection of B cell clones reactive with subdominant epitopes, which are antigens that remain unrecognized in a conventional immune response, indicating that Gb3 amplifies antibody diversity. We observed that, when used as an adjuvant in vaccination against influenza virus in mice, Gb3 triggered antibody responses reactive with the stalk of viral hemagglutinin, which, in contrast to its globular head, harbors conserved antigenic determinants. This broader Gb3-mediated B cell response proved to be cross-protective against other strains of influenza.

p

RATIONALE: Globotriaosylceramide (Gb3) is a glycosphingolipid abundantly expressed on GC B cells. Although used as a marker for these cells, it was not known whether Gb3 could regulate the function of B cells. We used genetically modified mouse models to deplete or increase the abundance of Gb3 in GC B cells to investigate its function in regulating humoral immune responses. In the dark zone of the GC, B cells undergo affinity maturation on the basis of a process called somatic hypermutation, by introducing mutations into antibody sequences to augment their affinity. B cell receptor (BCR) signaling triggers a downstream cascade that is required for migration of GC B cells out of the dark zone. Subsequently, B cells arrive in the light zone of the GC, where they perform antigen presentation through major histocompatibility

enza, impose a substantial burden on public health. In the absence of potent antiviral drugs, vaccination is the best strategy to control infectious disease. The goal of efficient vaccines is the induction of an immune response that is characterized by production of antibodies of high affinity and broad reactivity. The diversity of antibodies is especially important to facilitate cross-protective immunity in response to mutating viral strains. Antibody-producing B lymphocytes are central players in humoral immunity. After initial activation, B cells are recruited to anatomical structures called germinal centers (GCs), in which B cells mutate their antibody sequences to increase their affinity for antigen and processes of B cell selection occur that promote the diversity of the antibody repertoire.

complex class II (MHC-II) to T follicular helper (TFH) cells driving B cell selection and diversity. We therefore investigated somatic hypermutation, BCR signaling, GC B cell cycling, and MHCII presentation to understand the impact of the lipid Gb3 on antibody affinity and diversity.

,

rHA

Antibody affinity

Gb3 Affinity and diversity Gb3 as adjuvant

CONCLUSION: Our work introduces the lipid

Antibody diversity 100 80 60 40 20 0

Survival

CXCR4

Gb3 as a protagonist in the GC response, determining affinity as well as diversity of antibodies. From these findings, we propose that Gb3 has the potential to act as an adjuvant for vaccination against viral infections.

Gb3 immunized No Gb3 Antiviral immunity



Time

Gb3 drives affinity and diversity of antibody responses and protects against viral infection. In GC B cells, Gb3 binds to CD19 and disengages it from its chaperone CD81 for subsequent translocation to the BCR to mediate downstream signaling and affinity maturation. Moreover, Gb3 regulates MHC-II levels to facilitate selection of TFH cells and GC B cells of increased diversity. In vaccination with recombinant hemagglutinin (rHA), the use of Gb3 as an adjuvant triggers broadly neutralizing antibodies and broad protection against influenza. FDC, follicular dendritic cell; TCR, T cell receptor. [Created with BioRender.com] Sharma et al., Science 383, 720 (2024)

16 February 2024

The list of author affiliations is available in the full article online. *Corresponding author. Email: florian.winau@childrens. harvard.edu †These authors contributed equally to this work. Cite this article as P. Sharma et al., Science 383, eadg0564 (2024). DOI: 10.1126/science.adg0564

READ THE FULL ARTICLE AT https://doi.org/10.1126/science.adg0564 1 of 1

RES EARCH

RESEARCH ARTICLE



IMMUNOLOGY

The lipid globotriaosylceramide promotes germinal center B cell responses and antiviral immunity Pankaj Sharma1†, Xiaolong Zhang1†, Kevin Ly1, Yuxiang Zhang2, Yu Hu1, Adam Yongxin Ye2, Jianqiao Hu2, Ji Hyung Kim1‡, Mumeng Lou1, Chong Wang2, Quinton Celuzza2§, Yuji Kondo3, Keiko Furukawa4, David R. Bundle5, Koichi Furukawa4, Frederick W. Alt2, Florian Winau1*

16 February 2024

1 of 14

,

Sharma et al., Science 383, eadg0564 (2024)

y

*Corresponding author. Email: [email protected] †These authors contributed equally to this work. ‡Present address: Department of Biotechnology, College of Life Sciences and Biotechnology, Korea University, Seoul 02841, Republic of Korea. §Present address: Broad Institute, Cambridge, MA, USA.

To modulate Gb3 levels, we used mice that lack the key enzymes involved in synthesis [a1,4galactosyltransferase (A4Galt)] or degradation [a-galactosidase A (Gla)] of Gb3 (fig. S1A). To calibrate this model for Gb3 abundance, we immunized wild-type (WT), A4galt-KO (knockout), and Gla-KO mice with 4-hydroxyl-3nitrophenylacetyl coupled to ovalbumin (NP-OVA) (Fig. 1A and fig. S1B) and determined the number of Gb3-positive leukocytes. Compared to diverse B cell subsets (fig. S1, C and D) and other leukocytes (fig. S1E), high Gb3 expression was restricted to GC B cells. Whereas A4galt-KO mice did not express Gb3, WT showed ~20% and Gla-KO showed 50% Gb3pos GC B cells, the latter with increased surface expression (fig. S1, F and G). Thus, A4galt-KO, WT, and Gla-KO mice represent Gb3-deficient, homeostatic, and Gb3-abundant conditions, respectively, and hence were a suitable model to test the functional impact of various Gb3 levels on GC responses.

y g

Program in Cellular and Molecular Medicine, Boston Children’s Hospital, Department of Pediatrics, Harvard Medical School, Boston, MA, USA. 2Program in Cellular and Molecular Medicine, Boston Children’s Hospital, Department of Genetics, Harvard Medical School, The Howard Hughes Medical Institute, Boston Children’s Hospital, Boston, MA, USA. 3Department of Biochemistry II, Nagoya University Graduate School of Medicine, Nagoya, Japan. 4Department of Biomedical Sciences, Chubu University College of Life and Health Sciences, Kasugai, Japan. 5Department of Chemistry, University of Alberta, Edmonton, AB T6G 2G2, Canada.

Results Gb3 abundance on GC B cells is vital for antibody affinity maturation

y

1

signaling and TFH cell functions, overall, the specific roles of lipids in immunoregulation are still underexplored (10–12). In this study, we investigated the function of globotriaosylceramide (Gb3), which has been described for its role in apoptosis and as GC B cell marker for many years (13–17). Gb3 is a glycosphingolipid that is located in membranes of endothelial cells and tubular epithelial cells in the kidney (18). However, in the immune system, Gb3 is abundantly expressed on GC B cells. Despite its phenotypic role as marker CD77, the functional impact of Gb3 on B cell responses had not been addressed.

g

D

uring an infection or immunization, B lymphocytes enter the germinal center (GC) reaction and differentiate into antibodysecreting plasma or memory B cells (1). Within GCs, B cells recognize and acquire antigens on the basis of B cell receptor (BCR) affinities and compete for limiting amounts of T cell help, which is necessary for survival, proliferation, and selection of affinity-matured plasma and memory B cells (2). However, some pathogens, such as influenza virus, escape the host immune response because their surface antigens display a high frequency of point mutations, through a process known as antigenic drift (3, 4). The discovery of human broadly neutralizing antibodies, which are able to neutralize distantly related influenza strains, has spurred vaccination strategies aimed at broader humoral responses (5–7). The quality of selected antibodies is determined by GC B cell intrinsic signaling driven by the BCR, co-receptors such as CD19, and through interaction with T follicular helper (TFH) cells (8–10). Although there is evidence that the lipid microenvironment affects BCR

p

Influenza viruses escape immunity owing to rapid antigenic evolution, which requires vaccination strategies that allow for broadly protective antibody responses. We found that the lipid globotriaosylceramide (Gb3) expressed on germinal center (GC) B cells is essential for the production of high-affinity antibodies. Mechanistically, Gb3 bound and disengaged CD19 from its chaperone CD81, permitting CD19 to translocate to the B cell receptor complex to trigger signaling. Moreover, Gb3 regulated major histocompatibility complex class II expression to increase diversity of T follicular helper and GC B cells reactive with subdominant epitopes. In influenza infection, elevating Gb3, either endogenously or exogenously, promoted broadly reactive antibody responses and cross-protection. These data demonstrate that Gb3 determines the affinity and breadth of B cell immunity and has potential as a vaccine adjuvant.

After NP-OVA immunization, elevated Gb3 abundance in Gla-KO mice amplified frequencies of GC B, plasma, and TFH cells in the spleen and sustained long-lived plasmablasts in the bone marrow (Fig. 1B and fig. S1, H and I). In histology, mice rich in Gb3 showed bigger GCs compared with WT and A4galt KO mice (Fig. 1C). By contrast, Gb3-deficient A4galtKO animals exhibited poor differentiation of plasma and TFH cells (Fig. 1B and fig. S1I). Notably, Gla-KO mice showed higher titers of total immunoglobulin G (IgG), whereas Gb3 deficiency abrogated IgG2c class switch (Fig. 1D). Moreover, we controlled that longevity of the GC reaction as well as homeostasis of various leukocyte and B cell subsets were not altered in the Gb3-modulating mouse models (fig. S1, J and K). To verify the direct impact of the lipid Gb3 on the observed phenotype, we immunized A4galtKO mice with emulsified Gb3 in combination with NP-OVA (fig. S2A). Ten days after immunization, we detected ~20% Gb3pos GC B cells in the draining lymph node (fig. S2B). In addition, Gb3-treated mice showed increased total IgG as well as IgG2c class switch (fig. S2C). Because we used a Gb3-deficient host environment, these results suggest that this lipid may inherently drive GC B cell responses. Since we found up-regulation of Gb3 early upon B cell activation (fig. S2D), we also explored the impact of Gb3 on extrafollicular B cell responses, using immunization with NP adsorbed to the polymer Ficoll as well as Salmonella typhimurium infection (fig. S2, E and H). Both models revealed similar plasma cell differentiation and IgG titers among all strains (fig. S2, E to J), indicating that the Gb3 effects require the GC reaction. Interestingly, Gb3dependent IgG2c class switch exhibited an extrafollicular component (fig. S2, G and J). To validate the B cell intrinsic function of Gb3, we used mixed bone marrow chimera and found that Gb3-abundant GC B cells outcompeted WT cells and showed an increased proportion of IgG2c-expressing cells (fig. S3, A to D). Kinetic analysis of mixed chimera showed that the recruitment of B cells to the GC reaction and the ratio of TFH cells were not affected by Gb3 levels, which demonstrated that Gb3-dependent competitiveness was a specific feature of the GC (fig. S3, A to C). Additionally, we transferred naïve B cells from our mouse models to muMT-KO animals, which lack mature B cells, and observed that elevated GC B cell numbers and IgG2c class switch were intrinsically Gb3 dependent (fig. S3, E to G). We next measured antibody maturation by quantifying avidity and affinity of antigenantibody complexes. Antibodies from A4galtKO mice had low avidity, as indicated by the quick denaturation of antigen-antibody complexes in the presence of sodium thiocyanate

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

GC area (% of total area)

C 6 4 2

% GC B cells

B 8.0

% Plasma cells

A

8.0

6.0 4.0 2.0 0.0

6.0 4.0 2.0 0.0

106 105 104 WT Gla KO A4galt KO

102

2c

1

60 40 20 0

2M NaSCN

G

G

80

1.0

0.5

0.0

Ig

Ig

G

101

100

G

p

103

Avidity index (% bound antibodies)

107

IgG (NP4/NP30 ratio)

F

E

108

Ig

D

Anti-NP-OVA endpoint titer (Log10)

0

g

H

1.0 CAGGTCCAACTGCAGCAGCCTGGGGCTGAGCTTGTGAAGCCTGGGGCTTCAGTGAAGCTGTCCTGCAAGGCTTCTGGCTACACCTTCACCAGCTACTGGATGCACTGGGTGAAGCAGAGGCCTGGACGAGGCCTTGAGTGGATTGGAAGGATTGATCCTAATAGTGGTGGTACTAAGTACAATGAGAAGTTCAAGAGCAAGGCCACACTGACTGTAGACAAACCCTCCAGCACAGCCTACATGCAGCTCAGCAGCCTGACATCTGAGGACTCTGCGGTCTATTATTGTGCAAGA

0.5

CAGGTCCAACTGCAGCAGCCTGGGGCTGAGCTTGTGAAGCCTGGGGCTTCAGTGAAGCTGTCCTGCAAGGCTTCTGGCTACACCTTCACCAGCTACTGGATGCACTGGGTGAAGCAGAGGCCTGGACGAGGCCTTGAGTGGATTGGAAGGATTGATCCTAATAGTGGTGGTACTAAGTACAATGAGAAGTTCAAGAGCAAGGCCACACTGACTGTAGACAAACCCTCCAGCACAGCCTACATGCAGCTCAGCAGCCTGACATCTGAGGACTCTGCGGTCTATTATTGTGCAAGA

0.0

Mutation freq. (%)

CAGGTCCAACTGCAGCAGCCTGGGGCTGAGCTTGTGAAGCCTGGGGCTTCAGTGAAGCTGTCCTGCAAGGCTTCTGGCTACACCTTCACCAGCTACTGGATGCACTGGGTGAAGCAGAGGCCTGGACGAGGCCTTGAGTGGATTGGAAGGATTGATCCTAATAGTGGTGGTACTAAGTACAATGAGAAGTTCAAGAGCAAGGCCACACTGACTGTAGACAAACCCTCCAGCACAGCCTACATGCAGCTCAGCAGCCTGACATCTGAGGACTCTGCGGTCTATTATTGTGCAAGA

Mutation freq. (%)

y

CAGGTCCAACTGCAGCAGCCTGGGGCTGAGCTTGTGAAGCCTGGGGCTTCAGTGAAGCTGTCCTGCAAGGCTTCTGGCTACACCTTCACCAGCTACTGGATGCACTGGGTGAAGCAGAGGCCTGGACGAGGCCTTGAGTGGATTGGAAGGATTGATCCTAATAGTGGTGGTACTAAGTACAATGAGAAGTTCAAGAGCAAGGCCACACTGACTGTAGACAAACCCTCCAGCACAGCCTACATGCAGCTCAGCAGCCTGACATCTGAGGACTCTGCGGTCTATTATTGTGCAAGA

1.0

0.5

0.0

VH1-72 33W→L

VH1-72 33W→L CAGGTCCAACTGCAGCAGCCTGGGGCTGAGCTTGTGAAGCCTGGGGCTTCAGTGAAGCTGTCCTGCAAGGCTTCTGGCTACACCTTCACCAGCTACTGGATGCACTGGGTGAAGCAGAGGCCTGGACGAGGCCTTGAGTGGATTGGAAGGATTGATCCTAATAGTGGTGGTACTAAGTACAATGAGAAGTTCAAGAGCAAGGCCACACTGACTGTAGACAAACCCTCCAGCACAGCCTACATGCAGCTCAGCAGCCTGACATCTGAGGACTCTGCGGTCTATTATTGTGCAAGA

CAGGTCCAACTGCAGCAGCCTGGGGCTGAGCTTGTGAAGCCTGGGGCTTCAGTGAAGCTGTCCTGCAAGGCTTCTGGCTACACCTTCACCAGCTACTGGATGCACTGGGTGAAGCAGAGGCCTGGACGAGGCCTTGAGTGGATTGGAAGGATTGATCCTAATAGTGGTGGTACTAAGTACAATGAGAAGTTCAAGAGCAAGGCCACACTGACTGTAGACAAACCCTCCAGCACAGCCTACATGCAGCTCAGCAGCCTGACATCTGAGGACTCTGCGGTCTATTATTGTGCAAGA

1.0 CAGGTCCAACTGCAGCAGCCTGGGGCTGAGCTTGTGAAGCCTGGGGCTTCAGTGAAGCTGTCCTGCAAGGCTTCTGGCTACACCTTCACCAGCTACTGGATGCACTGGGTGAAGCAGAGGCCTGGACGAGGCCTTGAGTGGATTGGAAGGATTGATCCTAATAGTGGTGGTACTAAGTACAATGAGAAGTTCAAGAGCAAGGCCACACTGACTGTAGACAAACCCTCCAGCACAGCCTACATGCAGCTCAGCAGCCTGACATCTGAGGACTCTGCGGTCTATTATTGTGCAAGA

y

0.5

,

CAGGTCCAACTGCAGCAGCCTGGGGCTGAGCTTGTGAAGCCTGGGGCTTCAGTGAAGCTGTCCTGCAAGGCTTCTGGCTACACCTTCACCAGCTACTGGATGCACTGGGTGAAGCAGAGGCCTGGACGAGGCCTTGAGTGGATTGGAAGGATTGATCCTAATAGTGGTGGTACTAAGTACAATGAGAAGTTCAAGAGCAAGGCCACACTGACTGTAGACAAACCCTCCAGCACAGCCTACATGCAGCTCAGCAGCCTGACATCTGAGGACTCTGCGGTCTATTATTGTGCAAGA

Mutation freq. (%)

y g

CAGGTCCAACTGCAGCAGCCTGGGGCTGAGCTTGTGAAGCCTGGGGCTTCAGTGAAGCTGTCCTGCAAGGCTTCTGGCTACACCTTCACCAGCTACTGGATGCACTGGGTGAAGCAGAGGCCTGGACGAGGCCTTGAGTGGATTGGAAGGATTGATCCTAATAGTGGTGGTACTAAGTACAATGAGAAGTTCAAGAGCAAGGCCACACTGACTGTAGACAAACCCTCCAGCACAGCCTACATGCAGCTCAGCAGCCTGACATCTGAGGACTCTGCGGTCTATTATTGTGCAAGA

CAGGTCCAACTGCAGCAGCCTGGGGCTGAGCTTGTGAAGCCTGGGGCTTCAGTGAAGCTGTCCTGCAAGGCTTCTGGCTACACCTTCACCAGCTACTGGATGCACTGGGTGAAGCAGAGGCCTGGACGAGGCCTTGAGTGGATTGGAAGGATTGATCCTAATAGTGGTGGTACTAAGTACAATGAGAAGTTCAAGAGCAAGGCCACACTGACTGTAGACAAACCCTCCAGCACAGCCTACATGCAGCTCAGCAGCCTGACATCTGAGGACTCTGCGGTCTATTATTGTGCAAGA

0.0

VH1-72 33W→L CAGGTCCAACTGCAGCAGCCTGGGGCTGAGCTTGTGAAGCCTGGGGCTTCAGTGAAGCTGTCCTGCAAGGCTTCTGGCTACACCTTCACCAGCTACTGGATGCACTGGGTGAAGCAGAGGCCTGGACGAGGCCTTGAGTGGATTGGAAGGATTGATCCTAATAGTGGTGGTACTAAGTACAATGAGAAGTTCAAGAGCAAGGCCACACTGACTGTAGACAAACCCTCCAGCACAGCCTACATGCAGCTCAGCAGCCTGACATCTGAGGACTCTGCGGTCTATTATTGTGCAAGA

Fig. 1. Gb3 promotes the formation of germinal centers and antibody maturation. (A) Diagrammatic representation of the experimental setup. WT, Gla-KO, or A4galt-KO mice were immunized with NP-OVA adsorbed on alum. (B) FACS plots and percentage of GC B cells and plasma cells in the spleen on day 10 after NP-OVA immunization. The data shown are representative of three independent experiments with seven or eight mice per group. (C) Confocal microscopy of frozen sections from the spleen on day 10 after immunization. IgD (blue: naïve B cells), CD3 (red: T cells), and GL7 (yellow: GC B cells). GL7 staining was used to identify GCs, and the GC area was calculated in relation to the whole section. Bar graph depicts the percentage of GC area to the total Sharma et al., Science 383, eadg0564 (2024)

16 February 2024

CAGGTCCAACTGCAGCAGCCTGGGGCTGAGCTTGTGAAGCCTGGGGCTTCAGTGAAGCTGTCCTGCAAGGCTTCTGGCTACACCTTCACCAGCTACTGGATGCACTGGGTGAAGCAGAGGCCTGGACGAGGCCTTGAGTGGATTGGAAGGATTGATCCTAATAGTGGTGGTACTAAGTACAATGAGAAGTTCAAGAGCAAGGCCACACTGACTGTAGACAAACCCTCCAGCACAGCCTACATGCAGCTCAGCAGCCTGACATCTGAGGACTCTGCGGTCTATTATTGTGCAAGA

section area from five mice per group. Scale bars, 200 mm. (D) Serological analysis of immunoglobulin concentrations of indicated isotypes based on endpoint titers using ELISA on day 13 after immunization. Experiment was repeated three times with eight mice per group, and graph shows data from one representative experiment. (E) Avidity index of NP-OVA–specific IgG was measured by ELISA and expressed as the percentage of the endpoint titer values obtained with sodium thiocyanate treatment (2 M) compared with PBS on day 13 after immunization. Experiment was repeated twice with eight mice per group, and graph shows data from one representative experiment. (F) Antibody affinity assay. Sera were analyzed by ELISA, and high-affinity antibodies were 2 of 14

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

measured as a ratio of antibodies binding to BSA conjugated with NP4 (low valency) to NP30 (high valency). The graph shows data from one representative experiment repeated twice with eight mice per group. (G) Pie charts representing the distribution of productive VH1-72 clonotypes in the GCs of each mouse. Colors indicate the dominant clonotypes found among three different mouse genotypes. (H) Somatic hypermutation profiles of VH1-72 segments of the dominant clonotypes compared with background intrinsic mutation pattern of nonproductive VH1-72 alleles. Data are plotted as the percentage of reads containing mutations at each nucleotide. The VH sequences cover the CDR1 region containing a point mutation at position 98 that corresponds to the NP-specific high-affinity mutation W33L (indicated by arrow). Bar graphs depict frequencies of the W33L mutation in individual mice. Representative junctional structure of each IgH clonotype shown at the top was generated by WebLogo

,

3 of 14

y

16 February 2024

y g

Sharma et al., Science 383, eadg0564 (2024)

y

Within GCs, centroblasts in the dark zone undergo extensive proliferation as well as somatic hypermutation. Subsequently, centroblasts transit into the light zone to become centrocytes, which undergo affinity-based selection (8, 22). Given that the transition between the dark and light zone is critical for effective humoral responses, we next examined the impact of Gb3 on the composition of centroblasts and centrocytes. Analyzing GC B cells for the centroblast

and Akt. The low level of p-Akt in Gb3-deficient GC B cells resulted in impaired phosphorylation and concomitant accumulation of total FOXO1 (Fig. 2C). As not all GC B cells express Gb3, we sorted Gb3+ and Gb3− GC B cells from WT mice and explored their functional difference. Gb3+ GC B cells proved to be superior to their Gb3− counterparts in BCR downstream signaling, as indicated by amplified phospho-Akt activation (fig. S5C). Moreover, administration of Gb3 to Gb3-deficient animals showed that the signalingpromoting properties were a direct effect of this lipid on GC B cells (fig. S5D). Taken together, these data indicate that Gb3 activates the CD19PI3K-Akt axis, leading to FOXO1 degradation, which corresponds to CXCR4 down-regulation and efficient centroblast-to-centrocyte transition. In resting B cells, CD19 forms a complex with CD81, a tetraspanin required for CD19 trafficking to the B cell surface (29–31). Activated GC B cells exhibit pronounced phosphatase activity and transient activation of kinases after BCR cross-linking (32), which makes the phosphorylation of signaling molecules a delicate process. This challenge is overcome by the lateral translocation of CD19 from CD81 to the BCR signaling domain in activated B cells (12, 30, 33). Since Gb3 abundance drives strong CD19 downstream signaling, we hypothesized that the lipid Gb3 might facilitate CD19 disengagement from CD81 for translocation to the BCR nanocluster to form a functional signaling complex. To this end, we explored the interaction of BCR and CD81 with CD19 using proximity ligation assays (PLAs). Gb3-deficient GC B cells displayed exceptionally high PLA signal for CD81-CD19 proximity, while, in a complementary fashion, BCR-CD19 proximity was markedly reduced compared with WT and Gla-KO mice (Fig. 2D). By contrast, BCRCD19 PLA signal was augmented in Gla-KO GC B cells, indicating that Gb3 abundance dissociates CD19 from CD81 and amplifies the spatial relationship between CD19 and the BCR (Fig. 2D). In the BCR signaling cascade, PI3K is physically recruited to the cytoplasmic tail of CD19. When we performed PLA, GC B cells abundant in Gb3 showed increased interaction

g

Gb3 binds to CD19 for translocation to the BCR complex to trigger signaling

marker CXCR4 and the centrocyte molecule CD83 by flow cytometry, we found that centroblastto-centrocyte transition was hampered in the absence of Gb3 (Fig. 2, A and B). Quantitative imaging of GC B cell distribution mirrored our flow cytometry results as Gb3-deficient GC B cells were predominantly located in the dark zone, whereas GC B cells in Gla-KO mice were more numerous in the light zone when compared with WT and A4galt-KO (fig. S5A). These observations suggested that the lipid Gb3 is crucial for proper positioning of GC B cells in the light zone to allow for productive selection processes to occur. GC B cell proliferation in the dark zone in response to BCR activation is mainly regulated by the transcription factor FOXO1 (23–25). CD19 positively regulates BCR signaling, and cross-linking of BCR with CD19 reduces its activation threshold (10, 26). Furthermore, mice containing loss-of-function mutations at two signaling tyrosine residues of CD19 (Y482F and Y513F) lack selection of antigen-specific antibodies (27). Phosphorylation of CD19 activates the phosphatidylinositol 3-kinase–Akt (PI3K-Akt) pathway that leads to FOXO1 phosphorylation and triggers its proteasomal degradation, thus terminating the FOXO1-dependent gene signature that includes CXCR4 expression (25, 27, 28). Because we observed accumulation of CXCR4-expressing centroblasts in Gb3 deficiency, we next explored the impact of Gb3 on BCR signaling and downstream activation of the PI3K/Akt/FOXO1 pathway. We isolated GC B cells from our different mouse models and subjected them to BCR ligation with anti–IgM-F(ab)2 antibodies before analyzing the phosphorylation of signaling molecules using immunoblot. Modulation of Gb3 abundance had no differential effect on phosphorylation in the proximal signal cascade, involving CD79b and Syk (Fig. 2C and fig. S5B). However, Gb3-abundant GC B cells showed strong CD19 downstream signaling, markedly higher phosphorylation of Akt, and, consequently, low expression of total FOXO1 (Fig. 2C and fig. S5B). In contrast, CD19 downstream signaling was compromised in Gb3-deficient GC B cells, as they exhibited diminished phosphorylation of CD19, PI3K,

p

(NaSCN) (Fig. 1E). Furthermore, antibodies from A4galt-KO serum showed weak binding to bovine serum albumin (BSA) substituted with a low number of NP residues, suggesting impaired affinity maturation (Fig. 1F). To perform an in-depth analysis of the BCR repertoire in GC B cells, we used the recently described RepSHM-Seq method to detect specific BCR usage and assess somatic hypermutation (SHM) patterns in enriched clonotypes, where ≥90% nucleic acid sequence similarity in the complementaritydetermining region 3 (CDR3) was used as cutoff to define a clonotype (19). To examine heavychain variable gene segment (VH) enrichment, we compared VH usage of GC B cells with naïve B lymphocytes. In line with previous studies (20, 21), we observed enrichment of NP-specific VH1-72 in GC B cells in all immunized mice of the different genotypes (fig. S4). VH1-72 rearrangement in GC B cells generated three shared clonotypes among the top five clonotypes in all animal groups (Fig. 1G). The mutation analysis of clonotypes revealed that dominant VH1-72 B cell clones from all mice had G-to-T point mutations at residue 98 of CDR1, which corresponds to the mutation W33L (Trp33→Leu) that affords high affinity for NP antigen. Although there was no difference in the frequencies of W33L mutations in WT and Gb3-abundant mice, GC B cells lacking Gb3 showed drastically lower frequencies of the high-affinity W33L mutation in all dominant VH1-72 clonotypes (Fig. 1H). Altogether, these data demonstrate that Gb3 abundance expands GC B, plasma, and TFH cells, and mice deficient in Gb3 show a lack of antibody affinity maturation.

and aligned with germline VH (red), DH (blue), and JH (orange) sequences, with the deleted sequences shown in gray. CDR3 nucleotide sequences are shown in black, purple denotes amino acid sequences, and the clonotype consensus is shown in sequence-logo pictures comprising CDR3 nucleotide sequences with matching V and J segments, identical CDR3 length, and >90% sequence similarity. AID target hotspot motifs AGCT, TCGA, and DGYW (where D stands for G, T, or A; G stands for mutations at G-C pairs; Y stands for a pyrimidine base; and W stands for A or T) are highlighted, and the green color represents the error bar. In (G) and (H), the data are from three mice in WT and Gla-KO groups and four mice in the A4galt-KO group. All graphs represent mean ± SD, and data points in bar graphs represent individual mice. Significance was calculated by using Kruskal-Wallis H test with Dunn’s multiple comparison test. *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001.

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

A

B

C

D

F

p

E g y

G

H

I

y g y ,

Fig. 2. Gb3 binds to CD19 for translocation to the BCR complex and efficient signaling. (A) Diagrammatic representation of the experimental setup. WT, Gla-KO, or A4galt-KO mice were immunized with NP-OVA adsorbed on alum. (B) FACS plots and percentages of centroblasts (CB) and centrocytes (CC) in the spleen gated on GL7+, CD95+, and IgD− GC B cells on day 10 after immunization. The experiment was repeated three times (four or five mice per group). (C) Immunoblot showing BCR and CD19 downstream signaling molecules and transcription factors. Anti-IgM antibodies [IgM-F(ab)2] were used to stimulate FACS-sorted GC B cells for 2 or 5 min. U, unstimulated. (D) Proximity ligation assay performed on GC B cells to probe for vicinity between CD81 and Sharma et al., Science 383, eadg0564 (2024)

16 February 2024

CD19 (top panel; blue, DAPI; red, CD19:CD81 PLA signal), and BCR and CD19 (bottom panel; blue, DAPI; red, CD19:BCR PLA signal). Scale bars, 10 mm. (E) PLA performed on FACS-sorted GC B cells to probe for proximity between CD19 and Gb3 (blue, DAPI; red, CD19:Gb3 PLA signal). PLA signal was captured by confocal microscopy, and images were processed and analyzed by ImageJ software. Scale bars, 10 mm. In (D) and (E), experiments were repeated at least three times, and signals on more than 30 cells in different fields were calculated for statistical analysis. (F) Structures of different lipids and Gb3 analog used in the study. (G) FACS-sorted GC B cells from A4galt-KO mice were seeded with a complex of lipid and BSA for 2 hours at 37°C (see methods), and Akt 4 of 14

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

phosphorylation was quantified after stimulation of GC B cells with anti-IgM F (ab)2 to examine the effect of lipid reconstitution on Akt phosphorylation in GC B cells from Gb3-deficient mice. Histogram overlay (left) and mean fluorescence intensity (right panel) of pAkt staining measured by phospho-flow in GC B cells cultured with different lipids. The experiment was repeated twice with three samples per group. (H) Isothermal titration calorimetry to measure the binding between Gb3 and CD19. CD19 and the Gb3 analog were dissolved in PBS, and thermodynamic analysis of their binding was carried out at 25°C on a MicroCal

Gb3-mediated IgG2c isotype switch is dependent on type I interferon signaling

Gb3 drives B cell repertoire diversity

,

5 of 14

y

The generation of broad antibody responses, which are not limited to the most prominent antigenic determinants, but also recognize subdominant epitopes, is pivotal to effective immunity. Interclonal competition and repetitive mutation and selection processes increase affinity and impair clonal diversity (37). As Gb3abundant mice exhibited higher numbers of centrocytes, a cell population that undergoes selection in the light zone of GCs, we investigated whether high levels of Gb3 could help with the selection of B cell clones reactive against diverse epitopes. For this purpose, we performed rarefaction analysis of our Rep-SHM-Seq libraries to estimate for species richness, where each clonotype is defined as a unique species (Fig. 3, A and B). We found that, while there was no difference between WT and A4galt-KO, B cell clones from Gla-deficient animals showed higher diversity (Fig. 3B). We investigated whether higher clonal diversity in Gb3-abundant mice represents the selection of GC B cells reactive with subdominant epitopes. In the NP-OVA system for immunization, NP serves as the dominant epitope and OVA as subdominant antigen. After detection of GC B cell reactivities using fluorochromecoupled NP or OVA, analysis by flow cytometry

y g

16 February 2024

y

Given that type I interferon is critical for IgG2c class switching (34, 35), and as Gla-KO mice had elevated titers of IgG2c, we investigated whether Gb3-dependent IgG2c induction is regulated by signaling through interferon-a receptor (IFNAR1). Because type I interferon signaling is important for GCs to form (36), we aimed to test the impact of IFNAR1 specifically on IgG2c class switching after the initial phase of GC development. To this end, we injected anti-IFNAR1 blocking antibodies to mice with ongoing GC reactions at days 6 and 8 after immunization (fig. S6A). The amplified GC B cell response in Gb3-abundant animals was not affected by anti-IFNAR1 treatment (fig. S6B). By contrast, IFNAR1 blocking reduced the total IgG titers in Gla-deficient mice to WT levels, and IgG2c production was virtually abrogated (fig. S6C). To validate the impact of IFNAR signaling on Gb3-mediated IgG2c class switching further, we used Gla-deficient and Gla−/−/Ifnar−/− (double knockout, DKO) mice, and quantified IgG1- and IgG2c-positive GC B cells after NP-OVA immunization. Gladeficient and DKO mice showed comparable frequencies of IgG1-expressing cells, whereas IgG2c-positive GC B cells were not detectable in the absence of functional IFNAR signaling in DKO mice (fig. S6, D and E). We analyzed the titers of IgG1 and IgG2c antibodies in serum and found that IgG2c induction was specifically dependent on IFNAR signaling as we observed a reduced ratio of serum IgG2c to IgG1 in DKO animals (fig. S6F).

Furthermore, we explored the effect of Gb3 on type I interferon signaling and examined a possible molecular interaction between Gb3 and IFNAR1 on GC B cells, using PLAs. Analysis based on confocal microscopy revealed elevated PLA signal for Gb3 and IFNAR1 on GC B cells isolated from Gla-KO mice compared with WT, whereas GC B cells from A4galt-deficient mice, lacking Gb3, failed to produce any PLA signal (fig. S6G). In order to investigate Gb3-mediated IFNAR1 signaling functionally, we stimulated pure populations of GC B cells with IFNa and assessed the activation of downstream kinases. Immunoblot analysis showed augmented phosphorylation of Tyk2, JAK1, STAT1, and STAT2 in GC B cells from Gb3-abundant mice when compared with WT (fig. S6H). By contrast, GC B cells from Gb3-deficient animals displayed diminished activation of the respective kinases (fig. S6H). Taken together, these data demonstrated that interaction between Gb3 and IFNAR1 modulated the type I interferon signaling cascade and facilitated the selection of the IgG2c class of antibodies.

g

Sharma et al., Science 383, eadg0564 (2024)

binding with Ga2, iGb3, or Gb4. Next, to quantify binding affinity of Gb3 and CD19, we used isothermal titration calorimetry (ITC), in which the lipid is injected into a reaction chamber containing recombinant CD19 in solution. Because the titration of lipid into aqueous phase is challenging in ITC and the CD19-Gb3 interaction depends on the carbohydrate headgroup, we used the Gb3 analog with a short alkyl chain for the ITC assay. Binding data revealed a specific interaction between Gb3 and CD19 with a stoichiometry of 1:1 and a dissociation constant of 4.8 mM (Fig. 2H). Taken together, these results indicated that Gb3 binds to CD19 to enable its translocation to the BCR complex for efficient downstream signaling (Fig. 2I).

p

between CD19 and PI3K after BCR crosslinking (fig. S5E). Most notably, the PLA signal in A4galt-deficient GC B cells was virtually abolished (fig. S5E), suggesting that the lipid Gb3 is required for PI3K recruitment to CD19. Consistent with an earlier report suggesting a potential Gb3 binding site (16), our PLA between Gb3 and CD19 revealed close proximity between the lipid and the co-receptor (Fig. 2E), in contrast to the negative control CD45R (fig. S5F). Moreover, we excluded alterations of cholesterolrich membrane domains using cholera toxin subunit B (CTxB)–mediated detection of lipid rafts on GC B cells from the various mouse strains (fig. S5G). Overall, this indicates that direct interaction with Gb3 promotes the clustering of CD19 with the BCR complex and consequently amplifies CD19 downstream signaling after BCR stimulation. Next, we tested whether the lipid Gb3 is able to exert a direct effect on GC B cell stimulation and whether the carbohydrate headgroup and/or the lipid tail are required for its activationpromoting properties. We cultured Gb3-deficient GC B cells in the presence of either Gb3 or molecules that have overlapping lipid or sugar groups. Galabiaosylceramide (Ga2), isoglobotriaosylceramide (iGb3), and globotetraosylceramide (Gb4) share a similar ceramide backbone as Gb3 and only differ in their carbohydrate headgroups. Globotriaosylsphingosine (Lyso-Gb3) and a Gb3 analog contain the same sugar headgroup as Gb3 (Fig. 2F). Following incubation of the GC B cells with these different molecules, we measured phosphorylation of Akt after BCR cross-linking by phospho-flow (Fig. 2G). Exogenous addition of Gb3, Lyso-Gb3, or the Gb3-sugar analog restored Akt activation in A4galt-deficient GC B cells, albeit to different degrees (Fig. 2G). By contrast, Ga2, iGb3, and Gb4 had no effect on B cell activation. These results demonstrated the ability of exogenous Gb3 to reconstitute Gb3-deficient GC B cells. Furthermore, Gb3mediated activation of BCR downstream signaling depended on the lipid’s trisaccharide structure. We analyzed direct interaction between CD19 and different lipids using a protein lipid overlay (PLO) assay. Polyvinylidene difluoride (PVDF) membranes were loaded with different concentrations of lipids, and in accordance with our phospho-flow data, Gb3 strongly bound to CD19, followed by Lyso-Gb3 and Gb3 analog (fig. S5H). Of note, we did not see any CD19

ITC 200 instrument. (Top) The x axis depicts time, and the y axis represents rate of heat release (microcalories per second). (Bottom) The x axis represents molar ratio between CD19 and Gb3 analog, and the y axis depicts change in enthalpy. (I) A mechanistic scheme of the effect of Gb3 on the CD19 translocation and BCR downstream signaling pathway. All graphs represent mean ± SD. Data points represent individual mice (B), single cells [(D) and (E)], and individual samples (G). Significance was calculated by Kruskal-Wallis H test with Dunn’s multiple comparison test. *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001.

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

A

B

C

D

E

p

F

H g

60

40

20

y

0

G

0

y g

I

y ,

Fig. 3. Gb3 facilitates the selection of subdominant epitopes. (A) Schematic representation of the experimental setup. WT, Gla-KO, or A4galt-KO mice were immunized with NP-OVA adsorbed on alum. (B) Rarefaction curves showing the clonotype diversity for the three genotypes (purple, WT; red, Gla-KO; blue, A4galtKO). The solid dark curves and light shades show the mean and SEM (standard error of the mean) of the calculated diversity for each genotype. One-way ANOVA with Tukey’s HSD method was used to test the difference in diversity across different genotypes at 1000, 3000, and 5000 reads (n = 3 mice for WT and Gla-KO; n = 4 mice for A4galt-KO). (C) FACS plots and percentages of OVA-FITC+ and NP-PE+ among GL7+ and CD95+ GC B cells in the spleen on day 16 after immunization. The data are representative of three independent experiments with at least four mice per group. (D) Histogram overlays depicting the MHC-II expression on naïve B cells, centroblasts, and centrocytes as quantified by flow cytometry. Experiment was repeated at least three times with five mice per group. (E) (Top) Experimental Sharma et al., Science 383, eadg0564 (2024)

16 February 2024

setup of NP-OVA immunization and OT-II T cell transfer. (Bottom) FACS plots and percentages of OT-II TFH cells in the spleen on day 10 after immunization as quantified by flow cytometry. The experiment was repeated at least two times (five or six mice per group). (F) Experimental setup of NP-OVA immunization using different mouse models. (G) MHC-II expression on GC B cells on day 16 after immunization. Data are plotted as mean fluorescence intensity as quantified by flow cytometry (n = 5 mice per group). (H) FACS plots and percentages of OVA-FITC+ and NP-PE+ GC B cells in the spleen on day 16 after immunization. (I) Serological analysis of indicated antigen-specific immunoglobulin concentrations as detected by ELISA on day 16 after immunization. For (G) and (H), experiments were repeated three times with five mice per group. Graphs in (C) to (I) represent mean ± SD, and data points in graphs represent individual mice. Significance was calculated by Kruskal-Wallis H test with Dunn’s multiple comparison test. *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001; ns, not significant. 6 of 14

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

High Gb3 level cross-protects against influenza infection

,

7 of 14

y

On the basis of our results describing the immunological functions of Gb3 in GC B cells, we aimed at developing exogenous Gb3 as potential adjuvant for B cell responses upon vaccination. To this end, we investigated the kinetics and cell types of Gb3 uptake by administration of Gb3 and recombinant hemagglutinin (rHA) to A4galt-KO mice (fig. S9A). Compared to other lymph node cells, activated B cells preferentially incorporated exogenous Gb3 into their membranes, as indicated by surface staining (fig. S9B). High membrane incorporation of exogenous Gb3 correlated with reduced a-galactosidase A activity, suggesting a decelerated Gb3 metabolism in activated B cells (fig. S9C). Moreover, excision of the Gb3 injection site in mice did not alter the kinetics of Gb3 uptake in the lymph node, excluding a lipid-associated depot effect (fig. S9D). To test the adjuvant effect of Gb3 in antiviral vaccination, we immunized WT mice with rHA alone or in combination with either Gb3 or the well-established adjuvant alum (Fig. 4A). In a prime-boost model, both Gb3- and alumformulated rHA induced robust antibody responses (Fig. 4, B to D). However, mice immunized with Gb3-formulated rHA showed the strongest production of the IgG2c class of antibodies (Fig. 4D), higher avidity of antigenantibody complexes (Fig. 4E), and drastically elevated titers of antibodies reactive with the HA stalk compared with HA head recognition

y g

16 February 2024

Exogenous Gb3 is a potent adjuvant for influenza vaccination

y

In viral infection and vaccination, the generation of broadly neutralizing antibodies reactive with subdominant antigens is vital to ensure continued protection in the face of emerging mutant viruses (6, 7). Thus, we explored whether Gb3 abundance in GC B cells allows for the selection of diverse antibodies during respiratory flu infection. For this purpose, we infected our mouse models with influenza A virus (H1N1, strain PR8) and monitored the cellular and humoral immune responses. Unlike in A4galtdeficient mice, intranasal infection with 102 plaque forming units (PFU) of H1N1 triggered a strong GC response in mediastinal lymph nodes in Gla-KO animals (fig. S8, A and B), followed by increased titers of total IgG and IgG2c antibodies when Gb3 was abundant (fig. S8C). The viral antigen hemagglutinin (HA) consists of a globular head, which largely contains the dominant antigenic sites, and a head-supporting stalk domain that harbors subdominant epitopes. Interestingly, Gb3-abundant mice exhibited considerably increased titers of serum antibodies that reacted with the HA stalk region (fig. S8D), suggesting an amplified response to subdominant antigenic determinants mediated by Gb3. Moreover, Gla-deficient animals showed increased antibody titers recognizing recombinant H3, the most distantly related HA molecule compared with H1 (fig. S8E). We also purified IgG1 and IgG2c isotypes from serum and found that Gb3-deficient mice failed to produce high-avidity antibodies reactive with HA (fig. S8F). Next, we used live virus to examine the neutralization capacity of serum antibodies derived from our different mouse models. Compared to WT and A4galt-KO

animals, neutralization of H1N1 in vitro was augmented using serum from Gb3-abundant mice (fig. S8G). Notably, antibodies from Gladeficient animals also allowed for neutralization of a distantly related heterologous influenza virus (H3N2, strain HK68 X-31), demonstrating Gb3-mediated induction of cross-reactive antibodies (fig. S8H). Lastly, we performed in vivo challenge experiments, using H1N1 for priming followed by secondary infection with either H1N1 or H3N2 (fig. S8I). Compared to WT, Gla-KO mice showed protection against both influenza strains, as reflected by reduced weight loss and low viral burden in the lung (fig. S8, J to M). By contrast, A4galt-deficient mice suffered from weight loss and failed to control pulmonary viral load (fig. S8, J to M). Interestingly, only mice with high Gb3 levels were able to eradicate H3N2 virus in the lungs (fig. S8M), demonstrating Gb3-mediated cross-protection. Using lectin-based detection of sialic acid residues, we controlled for similar quantities of influenza virus receptors in the lungs of the various mouse strains (fig. S8N). Altogether, these data indicate that Gb3 in GC B cells drives the recognition of the HA stalk region, development of cross-reactive antibodies, and heterologous protection against influenza infection.

g

Sharma et al., Science 383, eadg0564 (2024)

of GC B and TFH cells in Gla-deficient animals were reduced upon lowered MHC-II expression in heterozygous MHC-II/Gla-KO mice (fig. S7, G to I). B cell transfer from the Gb3-modulating mouse models and MHC-II–haplosufficient strains to muMT-KO recipients showed that expanded TFH cell numbers were due to the increased MHC-II antigen-presenting functions of B cells rich in Gb3 specifically (fig. S7J). Most notably, the elevated response to subdominant OVA antigen in Gb3-abundant mice was virtually abolished when MHC-II expression was cut in half (Fig. 3H). These findings were reflected in serum antibody titers, with anti-OVA IgG in Gla-KO/MHC-II-Het animals brought to WT levels, whereas MHC-II haplosufficiency did not affect the antibody response to the dominant hapten NP (Fig. 3I). Thus, the GC response to subdominant epitopes was strictly dependent on Gb3-modulated MHC-II expression. Taken together, these results demonstrated that Gb3 regulates MHC-II–restricted antigen presentation to select TFH and GC B cells of broad diversity.

p

showed that Gla-KO mice displayed a sizable population of OVA-reactive B cells, whereas WT and Gb3-deficient animals only selected for NP-binding BCRs (Fig. 3C). Moreover, in contrast to WT and A4galt-KO and compared with NP-specific responses, Gb3-abundant mice showed increased IgG titers specific for OVA (fig. S7, A and B). Further, we were able to detect OVA-specific plasma cells in the bone marrow of Gla-deficient mice at day 60 after immunization, highlighting the lasting humoral response to subdominant epitopes that is mediated by Gb3 abundance (fig. S7C). To test whether the effect of Gb3 on the selection of subdominant epitopes is B cell intrinsic, we used our B cell transfer setup to muMT-KO mice that lack mature B lymphocytes (fig. S7D). We observed a higher frequency of OVApos GC B cells as well as elevated anti-OVA IgG titers in mice that received Gla-KO B cells (fig. S7, E and F), corroborating the B cell–intrinsic function of Gb3-mediated selection of B cell reactivity toward subdominant epitopes. The expansion of TFH cells and selection of diverse B cell clones depend on antigen presentation by GC B lymphocytes (38). Thus, taking into account that GC B cells show a wide range of MHC-II expression (39), we investigated whether Gb3 affects MHC-II levels as well as antigen presentation to TFH cells. In line with previous studies (39), we found lower MHC-II expression on WT centroblasts than on centrocytes and naïve B cells. However, mice rich in Gb3 maintained high amounts of MHC-II on centroblasts as well as centrocytes (Fig. 3D). By contrast, Gb3-deficient GC B cell subsets displayed a much lower level of surface MHC-II (Fig. 3D). We injected CD45.1-congenic OVAspecific CD4 T cells (OT-II) to NP-OVA immunized mice and quantified the generation of TFH cells (Fig. 3E). The increase of OVA-specific TFH cells in Gb3-abundant mice was virtually abrogated in the absence of Gb3 in A4galt-deficient mice compared to WT (Fig. 3E). The density of MHC-II/peptide complexes on antigen-presenting GC B cells during their interaction with TFH cells has been described to regulate B cell selection or entry into the GC reaction (40–42). Because we observed an elevated number of OVA-reactive GC B cells after hapten-based immunization in Gb3-abundant mice, we addressed whether the selection of GC B cells reactive with subdominant epitopes is dependent on Gb3-regulated MHC-II expression. We compared the GC response in mouse models expressing MHC-II from a homozygous or haplosufficient gene locus (Fig. 3F). Consistent with prior work (42), MHC-II haplosufficiency halved MHC-II surface expression on GC B cells in WT and Gb3-abundant mice, whereas the effect was negligible in mice with Gb3 deficiency, which is characterized by overall low amounts of MHC-II (Fig. 3G). After NPOVA immunization, the amplified frequencies

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

A

B

E

C

F

D

G

H

p

K

J

g

I

y

L

M

N

O y g y ,

Fig. 4. Exogenous Gb3 broadens the antibody response in influenza infection. (A) Experimental setup depicting animal groups, timeline of immunization, and sample collection schedule for rHA immunization. (B to D) Serological analysis of immunoglobulin concentrations of indicated isotypes specific for rHA was performed by ELISA. Graphs represent mean ± SD with five mice per group. The experiment was repeated three times. (E) Avidity of rHA-specific IgG was measured by ELISA and expressed as the percentage of the endpoint titer values obtained with sodium thiocyanate treatment (2 M) compared with PBS. Graphs represent mean ± SD with five mice per group. (F) Serum concentrations of immunoglobulins against rHA head and stalk were detected by ELISA. Graphs represent mean ± SD with five mice per group. (G) Enzyme-linked immunosorbent spot (ELISPOT) showing IgG1and IgG2c-producing plasma cells recognizing rHA on day 60 after immunization. The experiment was repeated three times (n = 4 mice for each group). (H) Serum concentrations of immunoglobulins against rHA of heterotypic influenza strains were Sharma et al., Science 383, eadg0564 (2024)

16 February 2024

detected by ELISA. The experiment was repeated three times (n = 6 mice for each group). (I and J) H1N1 or H3N2 neutralizing antibody (NA) titers in serum on day 28 after rHA immunization. Serum concentrations of immunoglobulins against rHA of heterotypic influenza strains were detected by ELISA. The experiment was repeated at least three times (n = 5 mice for each group). (K) Experimental setup depicting immunization and infection regimen. Immunized mice were challenged intranasally with 103 PFU of either H1N1 or H3N2 on day 28. i.n., intranasal. (L and M) Weight loss curve after H1N1 (PR8) or H3N2 (HK68, X-31) infection. Data shown in graphs represent three independent experiments with five mice per group. (N and O) H1N1 or H3N2 PFUs in lungs on day 5 after infection. The experiment was repeated twice with five mice per group. All bar graphs represent mean ± SD. Data points in graphs represent individual mice. Statistics were calculated by Kruskal-Wallis H test with Dunn’s multiple comparison test. *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001. 8 of 14

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

Discussion

,

9 of 14

y

RPMI medium 1640, ethylenediaminetetraacetic acid (EDTA), b-mercaptoethanol, penicillin and streptomycin, sodium pyruvate, L-glutamine, and TMB Substrate Solution were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Fetal bovine serum (FBS), acetylated trypsin, BSA, NaSCN, phosphate-buffered saline (PBS), PLA probe maker kit, and Duolink in situ detection reagents were procured from Millipore Sigma (Burlington, MA, USA). RPMI medium 1640 was supplemented with 10% FBS, penicillin (100 U/ml) and streptomycin (100 mg/ml), sodium pyruvate (1 mM), L-glutamine (2 mM), b-mercaptoethanol (50 mM), and HEPES (100 mM). Gb3 was purchased from Matreya LLC (State College, PA, USA), and other lipids used in the study were purchased from Avanti Polar Lipids (Alabaster, AL, USA). NP-OVA, NPFicoll, and NP-PE were procured from SouthernBiotech (Birmingham, AL, USA). B cell enrichment kit was purchased from Miltenyi Biotec (Bergisch Gladbach, Germany). Recombinant mouse CD19, hemagglutinin proteins from H1N1 (PR8), H3N2 (HK68, X-31), H5N1 (A/Vietnam/1194/2004), and H7N7 (A/chicken/ Netherlands/1/03) were purchased from Sino Biologicals Inc (Beijing, China). Hemagglutinin stalk protein (HA2 subunit) from influenza A (A/turkey/Ireland/1378/1983; #40164-V08B2L) was custom-purified by Sino Biologicals Inc. (Beijing, China). Influenza viruses (H1N1; PR8 and H3N2; HK8), Salmonella enterica subsp. enterica serovar Typhimurium (strain CDC 6516-60), and Madin-Darby Canine Kidney

y g

16 February 2024

Materials and methods Reagents y

We introduce herein previously unappreciated functions of the lipid Gb3—beyond its use as a mere marker molecule—that regulate the GC response at multiple levels (fig. S10G). We found that Gb3 enabled CD19 translocation to the BCR complex, amplified downstream signaling, and drove centroblast-to-centrocyte transition. CD19 is a central signaling hub in B cell activation, and its physical recruitment of PI3K is critical for signaling along the Akt/FOXO1 pathway (44). FOXO1 in turn is a major transcription factor that dominates the gene expression maintaining the dark zone of the GC, and, consequently, PI3K activation and FOXO1 degradation are a prerequisite for centroblast transition to the light zone (23, 25). Our data showed that Gb3 promoted BCR downstream signaling and amplified activation of the PI3K/ Akt/FOXO1 pathway. Increased FOXO1 degradation mediated by Gb3 allowed for efficient centroblast-to-centrocyte transition, thus driving the overall dynamic of the GC reaction and the opportunity for subsequent selection processes in the light zone. Activation of CD19 requires its dissociation from the resting CD81 complex and subsequent translocation to an active BCR nanocluster (12, 30). However, mechanisms controlling lateral translocation of CD19 in activated B cells are not clearly understood. Our work revealed that the glycolipid Gb3 could bind to CD19, and this idiosyncratic lipid– protein interaction promoted lateral translocation of CD19 to the BCR complex to facilitate the signaling cascade. Additionally, we observed that Gb3 facilitated MHC-II presentation, TFH cell differentiation, and subdominant epitope selection. B cells equipped with a high-affinity BCR are most successful in taking up antigen presented by follicular dendritic cells in the light zone, which makes them potent antigen-presenting cells for activation of TFH cells recognizing MHC-II (40). TFH cells in turn provide critical signals for GC B cell selection (8, 9). Thus, the variety of the TFH cell repertoire is mirrored in the selection of diverse GC B cell clones, including the ones reactive with subdominant epitopes (38). We found that Gb3 regulated MHC-II on GC B cells and that high Gb3 abundance promoted TFH cell differentiation. Using the NP-OVA system, proliferation of OVA-reactive TFH cells corresponded to the expansion of GC B cells recognizing OVA as subdominant epitope when Gb3 was abundant. Furthermore, MHC-II haplosufficient mice lacked any subdominant epitope reactivity, which highlights a role for Gb3 in MHC-II–restricted antigen presentation and subdominant epitope selection.

By infecting either Gla-deficient mice or vaccinated animals with influenza virus, we discovered three Gb3-mediated hallmarks in flu infection: IgG2c production, recognition of the HA stalk, and generation of cross-reactive antibodies that protect against diverse viral strains. IgG2c is important because it has proved to be the most effective isotype for influenza protection (45). This feature is based on the binding capacity of IgG2c to Fc receptors on myeloid cells, which amplifies antibody-dependent cellmediated cytotoxicity and clearance of infected host cells (46, 47). In contrast to the influenza virus HA head, which represents the main target of antibody responses, the stalk region harbors subdominant epitopes, which are less frequently mutated and therefore able to sustain effective immunity (7, 48). Thus, vaccine strategies aim at generating broadly neutralizing antibodies that are independent of seasonal antigenic drift and protect against a large range of viral strains (49). Using exogenous Gb3 for vaccination, we achieved cross-reactive, heterologous protection against influenza viruses, suggesting that Gb3 could serve as a future adjuvant toward a universal vaccine against influenza.

g

Sharma et al., Science 383, eadg0564 (2024)

the lipid Gb3 can activate CD19 downstream signaling in human B cells and has potential as an adjuvant in the human system, involving IgG1-based adjuvanticity.

p

(Fig. 4F). While both alum- and Gb3-adjuvanted animals had robust memory B cell responses, plasma cells secreting IgG2c were exclusively present in the Gb3-adjuvanted group at day 60 after immunization (Fig. 4G). Moreover, the antibodies isolated from the Gb3-adjuvant group bound more effectively to rHA molecules from group 1 (H1 and H5) and group 2 (H3 and H7) of influenza viruses (Fig. 4H). Whereas Gb3and alum-adjuvanted groups had similar H1N1 (PR8) neutralization titers, only Gb3 treatment facilitated efficient neutralization of the group 2 virus H3N2 (X-31) (Fig. 4, I and J). These data demonstrate that exogenous Gb3 acts as an adjuvant and promotes the production of cross-reactive antibodies. Next, we investigated whether Gb3-immunized mice were broadly protected against influenza infections with distantly related viral strains. To this end, we infected vaccinated animals with either H1N1 or H3N2 and monitored weight loss and virus multiplication in the lung (Fig. 4K). Only Gb3-adjuvanted rHA provided protection from both viral strains, as characterized by reduced weight loss (Fig. 4, L and M), and improved clearance of viruses in the lung (Fig. 4, N and O). Moreover, we purified IgG1 and IgG2c from serum of HA-vaccinated mice and observed that Gb3 increased the avidity of both isotypes reactive with HA (fig. S9, E and F) as well as their recognition of subdominant stalk epitopes (fig. S9G). Transfer of purified IgG1 and IgG2c from Gb3-adjuvanted mice to naïve WT animals revealed that both isotypes were protective against heterologous infection with H3N2 (fig. S9, H to J), while IgG2c showed stronger cross-protective potency, underscoring its important antiviral functions. Ultimately, we aimed at translating our findings on Gb3 as an adjuvant to the human system. We analyzed whether Gb3 is able to interact with CD19 to drive BCR downstream signaling in human GC B cells. For this assay, we sorted CD77pos and CD77neg GC B cells from human tonsils (fig. S10A). Performing PLA, we found a strong signal in confocal microscopy, indicative of proximity between Gb3 and CD19 in human GC B cells (fig. S10B). Functionally, after BCR cross-linking, CD77pos B cells displayed higher phosphorylation of Akt and degradation of FOXO1 than CD77neg cells (fig. S10C). We next compared the adjuvant effect of Gb3 with the clinically used adjuvant alum, using human tonsil organoids as previously described (43). Dissociated tonsil cells were cultured in vitro to form organoids and then stimulated with antigens and adjuvants (fig. S10D). Compared with stimulation with rHA alone, Gb3 improved GC responses as well as plasma cell differentiation (fig. S10E). In addition, Gb3 triggered a higher anti-rHA IgG1 titer than rHA alone and alum, reminiscent of our findings in the murine model regarding IgG2c production (fig. S10F). Taken together, these results demonstrate that

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

(MDCK) cell line were procured from ATCC. Alkaline Phosphatase Substrate Kit III for ELISPOT was procured from Vector Laboratories (Newark, CA, United States). Nitrocellulose membranes, PVDF membranes, 12% precast polyacrylamide gel, x-ray films, and Laemmli buffer were purchased from Bio-Rad Laboratories (Hercules, CA, USA). Chemiluminescence substrate (ECL plus) for immunoblots and bicinchoninic acid (BCA) protein assay kit were procured from Thermo Fisher Scientific (Waltham, MA, USA), and receptor destroying enzyme (RDE II) for hemagglutination inhibition (HI) test was purchased from Denka Seiken (Tokyo, Japan). Antibodies

16 February 2024

Total IgG, IgG1, and IgG2c antibodies specific for NP-OVA, OVA, or rHA were quantified using ELISA. High-binding flat-bottom 96-well plates (Corning Life Sciences, NY, USA) were coated with 1 mg/ml NP-OVA, OVA, or rHA in PBS; incubated overnight at 4°C; and blocked with PBS + 2% BSA for 1 hour at room temperature (RT). Subsequently, sera from vaccinated or H1N1-primed mice were added with an initial dilution of 1:100 and 1:4 serial dilutions in PBS + 1% BSA, before incubation for 2 hours at RT. Plates were then washed and incubated for 1 hour at RT with horseradish peroxidase (HRP)–conjugated anti-mouse IgG, IgG1, and IgG2c antibodies. Following incubation, plates were washed again and developed with tetramethylbenzidine for 5 min, then stopped with 1 M H2SO4. The optical density was read at 450 nm using a Versamax microplate reader with SoftMax Pro Version 5 (both from Molecular Devices, San Jose, CA, USA), and endpoint titers were calculated with a cutoff three times the optical density over background. To test the IgG1 released by human tonsil cells, rHAcoated plates were incubated with culture supernatant with an initial dilution of 1:25 and 1:4 serial dilutions. For measuring antibody affinity, plates were coated with either NP30-BSA or NP4-BSA (10 mg/ml). The serum titers of IgG reactive with each conjugate were determined as above. Results were presented as ratios of IgG titers against NP4-BSA versus NP30-BSA. For assessing antibody avidity, plates were incubated for 15 min with 2 M sodium isothiocyanate 10 of 14

,

Mixed bone marrow (BM) chimeric mice were generated following an adaptation of previously described methods (42, 51). Briefly, donor mice were euthanized, and their bones were surgically cleaned from surrounding tissue. Extremities of tibiae and femora were trimmed with sterile scissors, and bone mar-

ELISA measurement of antibody affinity and avidity

y

Generation of bone marrow chimeric mice

Six- to eight-week-old WT, Gla-KO, and A4galtKO mice were euthanized, and total splenic B cells were purified using a negative B cell isolation kit (Miltenyi Biotec). Purity of enriched B cells was determined by flow cytometry, and recipient muMT-KO mice were injected with 1 × 107 B cells intravenously. B cell–reconstituted muMT-KO mice were immunized 7 days after transfer with NP-OVA adsorbed on alum, as described above, and GC B cell and antibody responses were quantified on day 21 after B cell transfer.

y g

Sharma et al., Science 383, eadg0564 (2024)

Adoptive transfer of B cells to muMT-KO mice

Mice were immunized with 25 mg NP-OVA, 250 mg NP-Ficoll, or 10 mg of rHA (from H1N1; PR8) either alone or in combination with either 50 mg alum or 25 mg Gb3 in 200 ml PBS. For S. typhimurium infection, bacteria were cultured in fresh Luria broth for 24 hours. Thereafter, an additional 2-hour cultivation at 37°C was performed until the bacteria reached the mid-log phase of growth. Subsequently, bacteria were washed three times with sterile PBS, their concentration was assessed by measuring absorbance at 600 nm, and 1 × 105 colony-forming units were administered intraperitoneally. To test the adjuvant effect of Gb3, 10 ml of a 5 mg/ml Gb3 stock (in dimethyl sulfoxide) was suspended in saline solution containing antigens and emulsified with 10% Tween 80 for 30 min. Thereafter, mice were immunized with lipid-formulated antigens in 200 ml PBS. rHA-immunized mice were challenged with 1 × 103 PFU of either H1N1 (PR8) or H3N2 (HK68, X-31) viruses. To assess whether emulsified Gb3 persists as a depot at the injection site and undergoes gradual release, mice were immunized subcutaneously with Gb3 in the ears, and subsequently, ears were excised on day 1, 2, or 4 after immunization. In separate infection experiments, WT, Gla-KO, and A4Galt-KO mice were intranasally primed with 1 × 102 PFU of H1N1 and, 24 days after priming, were challenged with 1 × 103 PFU of either H1N1 or H3N2 viruses. In some experiments, 100 mg of polyclonal IgG1 or IgG2c antibodies isolated from rHA- and Gb3-immunized mice were administered intravenously into WT mice. One day later, recipient mice were challenged with 1 × 103 PFU of H3N2 virus. Subsequently, mice were monitored for weight loss, and 25% weight loss was considered as endpoint.

y

C57BL/6, B6.SJL (B6.SJL-Ptprca Pepcb/Boy), muMT-KO (B6.129S2-Ighmtm1Cgn/J), MHC-IIKO (B6.129S2-H2dlAb1-Ea/J), OT-II transgenic mice (B6.Cg-Tg(TcraTcrb)425Cbn/J), Ifnar1-KO (B6.129S2-Ifnar1tm1Agt/Mmjax), and Gla-KO (B6;129-Glatm1Kul/J) mice were acquired from the Jackson Laboratory. A4galt-deficient mice were generated by K. Furukawa (50). Mice were bred and maintained (4 or 5 mice per cage) in

Immunizations and influenza infection

g

Mice

row was flushed through a 70-mm nylon mesh strainer (Corning Life Sciences, NY, USA). Cell number and viability were determined by trypan blue exclusion. To generate BM chimera, WT.SJL (F1 generation of CD45.2 C57BL/6 and CD45.1 SJL) were lethally irradiated with 1100 rad and intravenously received 5 × 106 WT CD45.1 and CD45.2 BM cells from either C57BL/6, Gla-KO, or A4galt-KO mice (ratio of CD45.1:CD45.2 = 1:1). Mice were given drinking water containing 1% neomycin for one week and were used 6 weeks after BM transplantation.

p

Anti-mouse antibodies against CD3 (145-2C11), CD4 (GK1.5), CD45.1 (A20), CD45.2 (104), CXCR5 (L138D7), PD1 (29F.1A12), GL-7 (GL7), CD45R (RA3-6B2), IgG (Poly4053), IgG2c (MRG2c-67), CD93 (AA4.1), CD21/35 (76G), CD103 (2E7), CD11b (M1/70), F4/80 (QA17A29), and Gr-1 (RB6-8C5) as well as anti-human antibodies against CD3 (UCHT1), IgD (IA6-2), CD27 (M-T271), CD38 (HB-7), CD77 (5B5), and purified anti-mouse IgM (RMM-1) antibodies were purchased from BioLegend (San Diego, CA, USA). Purified antimouse CD19 (MB19-1), PI3K (H.843.0), CD79a (SP18) and pCD79a (polyclonal), anti-human CD19 (HIB19) and IgM (SA-DA4) antibodies were purchased from Thermo Fisher Scientific (Waltham, MA, USA). The following antibodies for Western blot analysis were purchased from Cell Signaling Technology (Danvers, MA, USA): anti-mouse CD19 (D4V4B), Akt (40D4), Syk (D3Z1E), PI3K (C73F8), FOXO1 (C29H4), b-actin (13E5), p-Akt (D9E), p-Syk (65E4), p-PI3K (E3U1H), p-FOXO1 (E1F7T), STAT-2 (D9J7L), Jak1 (D1T6W), Tyk2 (E9H4T), p-STAT-1 (polyclonal), p-Jak1 (D7N4Z), p-Tyk2 (D7T8A), and anti-human b-actin (8H10D10). Anti-mouse p-STAT-2 (polyclonal), STAT-1 (STAT1-79), and p-CD19 (polyclonal) were purchased from Thermo Fisher Scientific (Waltham, MA, USA), and anti-mouse CD77 (BGR23) was purchased from Creative Biolabs. Anti-mouse IgG-HRP (polyclonal), IgG1-HRP (polyclonal), IgG2c-HRP (polyclonal), and antihuman IgG1-HRP were purchased from SouthernBiotech (Birmingham, AL, USA). Cytofix/ Cytoperm buffer kit and Fc-blocking reagent were purchased from BD Biosciences (Franklin Lakes, NJ, USA).

the Harvard Medical School (HMS) animal barrier facility and fed on standard pellet diet and water ad libitum. Mice used in the experiments were males or females between 6 and 8 weeks old, and experimental groups were age- and sex-matched. All experimental mice were euthanized by exposure to a high dose of carbon dioxide (CO2) with cervical dislocation as a secondary method. All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at HMS (approval number IS00001618).

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

before the addition of HRP-conjugated antimouse IgG antibodies. Avidity was expressed as the log half-maximal effective concentration ratio of corresponding plates treated with or without sodium isothiocyanate. Purification of IgG1 and IgG2c from immune serum

ELISPOT

ITC experiments were performed using a MicroCal ITC microcalorimeter at 25°C and a 150 mM NaCl buffer at pH 7.4. Recombinant CD19 was dialyzed into PBS containing 20 mM HEPES, and the lyophilized Gb3 analog was suspended in dialysate to achieve buffer match. All experiments were performed with the CD19 ectodomain in the microcalorimeter cell at 300 mM, and the Gb3 analog in the syringe at 100 mM. The titrations consisted of a total of 35 injections (5 ml each), spaced 80 s apart. Protein concentrations were determined on the basis of the ultraviolet absorbance at 280 nm, and the molar extinction coefficient of the CD19 ectodomain was 6805 M−1 cm−1. Data were analyzed with MicroCal ITC Origin 7 software (Malvern Panalytical Ltd, Worcestershire, UK) using a one-site binding model. 11 of 14

,

16 February 2024

Isothermal titration calorimetry assay y

Sharma et al., Science 383, eadg0564 (2024)

y g

Leukocytes were harvested from spleen as well as inguinal and mediastinal lymph nodes by passage through a 40-mm mesh and incubated for 10 min with 1 mg/ml of anti-CD16/32 Fc-block in fluorescence-activated cell sorting (FACS) buffer (PBS supplemented with 1 mM EDTA, 2% heat-inactivated FBS, and 0.02% sodium azide). Cells were further labeled using fluorochromecoupled antibodies directed against the respec-

Confocal microscopy was performed to visualize GCs as well as the results of proximity ligation assays. Regarding the former, spleens from immunized mice were fixed in 4% paraformaldehyde in PBS for 1 hour at 4°C, followed by a five-time wash in PBS (10 min each). Thereafter, spleens were moved to 30% sucrose solution in PBS (overnight) and were frozen in TAK tissue-mounting media the next day. For immunostaining, 30-mM sections were cut and air dried for 1 hour before rehydration in PBS with 1% BSA for 10 min. Slides were then washed three times in PBST and blocked for 1 hour at RT with 2% BSA. Staining with primary antibodies directed against CD3, GL7, CD35, and IgD was performed in PBS with 2% mouse serum, 0.1% BSA, and 0.1% NaN3 at 4°C overnight. Subsequently, slides were mounted with anti-fade mounting medium, and images were taken using either an Olympus VS120 slide scanner or an Olympus Fluoview (FV) 1000 confocal microscope. Images were processed and analyzed using ImageJ/Fiji software (National Institutes of Health). Briefly, CD35 staining for follicular dendritic cells was used to demarcate the light zones within GCs. Thereafter, the proportion of GL7+ GC B cells in light and dark zones was assessed by measuring GL7+ mean fluorescence intensity in the respective areas, and the statistical analysis was carried out on the basis of 25 separate GCs in each group. In a separate experiment, quantifying the abun-

FACS-sorted GC B cells were plated on PTFEslides for 30 min at 37°C and were fixed for 15 min with 2% paraformaldehyde. For intracellular PLA, cells were permeabilized after fixation with 0.5% saponin in PBS for 30 min. PLA was carried out with Duolink in situ detection reagents following the manufacturer’s instructions (Millipore Sigma, MA, USA). Briefly, fixed cells were blocked using a solution that contained 25 mg/ml sonicated salmon sperm DNA and 250 mg/ml BSA in PBS. After blocking, cells were incubated with appropriate PLA probes directed against either CD19 (polyclonal), BCR (RMM-1), or CD81 (9F7). To test for PI3K recruitment to CD19, cells were stained with antibodies against the intracellular domain of CD19 (D4V4B) as well as PI3K (H.843.0). Thereafter, cells were washed with PBS thrice and were further incubated with secondary PLA probes binding to the corresponding primary antibodies. To study the interaction between Gb3 and CD19, anti-mouse CD19 and anti-Gb3 antibodies were directly tagged to plus and minus oligo probes, using the PLA probe maker kit following the manufacturer’s guidelines (Millipore Sigma, MA, USA), and fixed cells were directly labeled with tagged antibodies overnight. Subsequently, all probe-labeled samples were subjected to PLA signal amplification following the manufacturer’s instructions. Finally, cells were mounted on slides with DAPIcontaining mounting reagent. Images were taken using an Olympus Fluoview (FV) 1000 confocal microscope, and data were analyzed using ImageJ software (NIH).

y

Flow cytometry and cell sorting

Immunofluorescence microscopy

Proximity ligation assay

g

To quantify the antigen-specific antibody-secreting plasma cells by ELISpot assay, 96-well ELISpot plates (Millipore, Burlington, MA, USA) were coated for 1 hour at RT with 50 ml of 8 mg/ml goat-anti-mouse IgG (KPL, Gaithersburg, MD, USA) for assessing total IgG, or 50 ml of 20 mg/ml purified rHA proteins from H1N1 (PR8) or H3N2 (HK68, X-31) influenza virus strains. Wells were washed with PBS and then blocked for 1 hour at 37°C with 200 ml per well of RPMI containing 10% heat-inactivated FBS. Single marrow cells were extracted from both femurs from each mouse as described above, and cells were then washed twice with PBS containing 0.5% BSA and added to the precoated plates at 500,000 cells per well, followed by 1:2 serial dilutions down to 6250 cells per well. The plates were incubated for 4 hours in a 5% CO2humidified incubator at 37°C. Thereafter, the wells were washed six times with 0.1% Tween20 PBS (PBST), and phosphatase-labeled goatanti-mouse IgG detection antibody was added at 100 ml per well. Plates were then incubated overnight at 4°C, washed six times with PBST, before IgG-specific spots were developed using the Alkaline Phosphatase Substrate Kit III. Spot numbers were enumerated using an ImmunoSpot plate reader (Cellular Technology Limited, Cleveland, OH, USA).

dance of lipid rafts, purified GC B cells were fixed and stained with FITC-labeled CTxB overnight at 4°C, while nuclei were stained with 4′,6diamidino-2-phenylindole (DAPI). Images were acquired on an Olympus Fluoview (FV) 1000 confocal microscope and analyzed using ImageJ/ Fiji software.

p

Polyclonal IgG1 and IgG2c antibodies were affinity-purified from mouse serum immunized with rHA and Gb3. Anti-mouse IgG1 and IgG2c antibodies (SouthernBiotech, Birmingham, AL) were coupled to agarose beads using AminoLink coupling resin (Thermo Scientific), and sera were incubated with isotype-specific affinity columns overnight at 4°C. Thereafter, columns were washed three times with PBS, and bound antibodies were eluted using 0.1 M glycine buffer (pH 2.7) and neutralized immediately with 0.04 volumes 1 M Tris (pH 9). The presence and purity of rHA-specific isotypes in samples were assessed by ELISA.

tive antigens: CD45R, GL-7, CD93, CD95, CD138, IgG, IgG1, IgG2c, CXCR4, CD83, CD4, PD-1, CXCR5, CD77, Gr-1, F4/80, CD103, and CD11b. Direct detection of antigen-binding B cells was performed using NP-PE and OVA-FITC labeling of singlecell suspensions. For phospho-Akt staining, B cells were stimulated with anti-BCR antibodies for precisely the same time to avoid variations induced by intracellular phosphatases. Cells were surface-stained at 4°C for 30 min, followed by fixation and permeabilization using the Fixation/Permeabilization buffer kit according to manufacturer’s instructions (BD Biosciences, NJ, USA). This was followed by staining with anti-phospho-Akt antibody diluted in Perm/Wash buffer for 1 hour. For tetramer staining, 2 × 106 cells suspended in 50 ml PBS were incubated with MHC-II OVA (ISQAVHAAHAEINEAGR; 1:50 dilution) tetramers for 30 min at 37°C in the dark. All flow cytometry data were generated using a FACS Canto Flow Cytometer (BD Biosciences, NJ, USA). For sorting GC B cells, B cells were enriched using a B Cell Isolation Kit (Miltenyi Biotec, Bergisch Gladbach, Germany) from spleens of immunized mice, and subsequently, GC B cells (CD45R+, GL7+, CD95+) and naïve B cells (CD45R+, GL7−, CD38−) were sorted either using a FACS Aria II (Becton Dickinson) or SH-800 cell sorter (Sony).

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

Western blotting

a-Galactosidase A (a-Gal A) activity assay

16 February 2024

To identify the enrichment of antigen-specific GC clonotypes across different samples, we pooled the GC productive reads from all the mice of different genotypes together to carry out clonal clustering as described (19). In our study, a clonotype is considered as enriched in one sample if it makes up more than 0.3% of all productive junctions. The consensus CDR3 DNA sequence plot for each clonotype was generated by WebLogo (55). Clonotype diversity analysis was done by rarefaction using the R package iNEXT. The number of clonotypes was calculated for each sample at the number of reads from 100 to 5000. One-way analysis of variance (ANOVA) with Tukey’s Honestly Significant Difference (HSD) post-hoc method was used to test the difference in diversity across different genotypes at 1000, 3000, and 5000 reads. SHM analysis

Mutation frequency in the heavy-chain variable segment of CDRs was calculated by enumerating the number of nucleotide mismatches compared with the background intrinsic nonproductive allele for the selected mutation as described (19). For statistical analysis, we used Kruskal-Wallis H test with Dunn’s multiple comparison test to compare SHM rate of productive sequences in a clonotype across different genotypes. 12 of 14

,

Sharma et al., Science 383, eadg0564 (2024)

Splenic naïve and GC B cells were enriched using a B cell isolation kit followed by FACSsorting as described above. Rep-SHM-Seq was performed as described previously using our established bioinformatics pipeline (19, 53). We used bait primers close to the coding ends of the joining regions of the immunoglobulin heavy chain (JHs) to capture full-length V(D)J sequences for SHM analysis and used Illumina MiSeq 2 × 300–base pair (bp) paired-end sequencing (19, 54). As described before, mixed JH primers were designed on the basis of a highly degenerate region to minimize the amplification bias caused by different primers. Four hundred nanograms genomic DNA from FACS-sorted naïve or GC B cells was used as starting material to keep the level of polymerase chain reaction (PCR) amplification comparable among different samples. In addition, to minimize the potential cross-contamination due to FACS sorting, we further filtered reads by assigning mutated sequences to GC B cell samples and nonmutated sequences to naïve B cell samples. Libraries for sequencing were prepared as described (19). Briefly, genomic DNA was sonicated and subjected to linear amplification–mediated PCR using biotinylated

CDR3 clonotype analysis

y

GC B cells were isolated from Gb3-deficient A4galt-KO mice and purified as described above. For lipid incorporation, a complex of Gb3 or its analogs with defatted BSA was prepared in serum-free RPMI supplemented with 15 mM HEPES (pH 7.2) at a molar ratio of 1:1. Cells were seeded in V-shaped 96-well plates and incubated with the lipid/BSA solution for 2 hours at 37°C. Subsequently, cells were washed once with complete RPMI/FCS and once with PBS, before collection by centrifugation and resuspension in complete RPMI/FCS for further stimulation assays. Gb3 incorporation into B cells was confirmed by flow cytometry using anti-Gb3 antibody staining.

BCR sequencing

For each sample, the ratio of each functional V segment was calculated as percent usage among total productive and nonproductive V(D)J junctions. For GC versus naïve repertoire comparison, V segment usages from multiple samples were plotted in the same chart above and below the x axis along indicated chromosomal coordinates. A complete list of functional VH segments in the order plotted along the x axis is shown in table S1. Significantly enriched V segments in GCs were identified by paired Student’s t test with multiple test correction by false discovery rate (FDR), and GC-enriched VH segments with FDR value ≤ 15% were considered significant.

y g

Lipid loading onto GC B cells

Mouse lungs were lysed in RIPA buffer, and total protein content was quantified by a BCA kit (Thermo Scientific). Thereafter, proteins were diluted at 3 mg/ml in ELISA coating buffer (0.1 M sodium carbonate, pH 9), and highbinding ELISA plates were coated overnight with 100 ml of protein solution at 4°C, while different concentrations of biotinylated lectins were used as standard. Plates were washed three times with PBST (PBS + 0.05% Tween 20) and blocked with Carbo-Free blocking solution (Vector Laboratories) for 30 min. Subsequently, plates were incubated sequentially for one hour with 10 mg/ml biotinylated lectins (SNA and MAL II; Vector Laboratories) and VECTASTAIN Elite ABC-HRP (Vector Laboratories), and peroxidase activity was quantified using TMB substrate.

V segment usage analysis

y

Protein lipid overlay assays were performed as previously described (52). Various glycolipids were dissolved in a mixture of methanol and chloroform (1:1), spotted onto a PVDF membrane and air-dried. The membrane was further blocked with 4% fatty acid–free BSA in PBST buffer (PBS containing 0.1% Tween 20, pH 8.0) for 1 hour at RT. Thereafter, the membrane was incubated with histidine-tagged recombinant CD19 (5 mg/ml) at RT for 2 hours. Subsequently, the membrane was washed five times (5 min each) with PBST buffer and incubated for 1 hour at RT with a 1:4000 dilution of rabbit anti-histidine antibody. Next, the membrane was washed five times as described above and then incubated with a 1:10,000 dilution of HRP-labeled donkey anti-rabbit antibody. Finally, the membrane was washed in PBST buffer, and deposited lipids were detected by enhanced chemiluminescence on x-ray films.

Lectin ELISA

bait primers. Linear amplification–mediated PCR products were purified using Dynabeads MyOne streptavidin C1 magnetic beads (Thermo Fisher Scientific, MA, USA) and ligated to bridge adaptors. Adaptor-ligated products were amplified by nested PCR with indexed JH1/JH3 primers as well as primers annealed to the adaptor. PCR products were further tagged with Illumina sequencing adaptor sequences and size-selected by means of gel extraction. Libraries were sequenced by paired-end 300-bp sequencing on MiSeq, an Illumina sequencing platform (San Diego, CA, USA).

g

Protein lipid overlay assay

a-Gal A activity was quantified per the manufacturer’s protocol (Abcam). FACS-purified leukocyte subsets were counted, and 1 × 106 cells were lysed in a-Gal assay buffer. Thereafter, 40 ml of protein lysate from each sample was incubated with 20 ml of diluted a-Gal substrate for 2 hours, while 4-methylumbelliferone was used as standard. Mean fluorescence intensities (Ex/Em = 360/445 nm) were quantified using a fluorometer at 37°C, while total protein contents were quantified using a BCA kit (Thermo Scientific). Specific a-Gal A activity was calculated as total a-Gal A activity per milligram of protein.

p

FACS-sorted GC B cells were lysed using radioimmunoprecipitation assay (RIPA) buffer containing 1× protease and phosphatase inhibitors, and the protein concentrations of whole-cell lysates were quantified using the BCA protein assay kit. The samples were boiled in SDS loading Laemmli buffer (Bio-Rad) supplemented with b-mercaptoethanol and subsequently resolved using a 12% polyacrylamide gel, before transfer onto a nitrocellulose membrane. Next, the membrane was blocked with 5% BSA for 1 hour and then probed for the indicated proteins using specific primary antibodies overnight. All primary antibodies were used at 1:500, which were further probed with HRP-labeled donkey anti-rabbit antibody (Jackson ImmunoResearch, West Grove, PA, USA). HRP-labeled antibody/protein complexes were detected by incubating the membrane with enhanced chemiluminescent substrate and exposing the blot to x-ray films in the dark. X-ray films were scanned in gray scale, and densitometry analysis of processed images was carried out using ImageJ software.

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

Microneutralization and hemagglutination inhibition assays

16 February 2024

13 of 14

,

Sharma et al., Science 383, eadg0564 (2024)

y

Dissociated cells from human tonsils were procured from BioIVT (NY, USA) and stimulated

1. G. D. Victora, M. C. Nussenzweig, Germinal centers. Annu. Rev. Immunol. 40, 413–442 (2022). doi: 10.1146/annurev-immunol120419-022408; pmid: 35113731 2. S. Crotty, T follicular helper cell biology: A decade of discovery and diseases. Immunity 50, 1132–1148 (2019). doi: 10.1016/ j.immuni.2019.04.011; pmid: 31117010 3. D. Angeletti et al., Defining B cell immunodominance to viruses. Nat. Immunol. 18, 456–463 (2017). doi: 10.1038/ ni.3680; pmid: 28192417 4. C. J. Wei et al., Next-generation influenza vaccines: Opportunities and challenges. Nat. Rev. Drug Discov. 19, 239–252 (2020). doi: 10.1038/s41573-019-0056-x; pmid: 32060419 5. D. C. Ekiert et al., Antibody recognition of a highly conserved influenza virus epitope. Science 324, 246–251 (2009). doi: 10.1126/science.1171491; pmid: 19251591 6. J. R. Whittle et al., Broadly neutralizing human antibody that recognizes the receptor-binding pocket of influenza virus hemagglutinin. Proc. Natl. Acad. Sci. U.S.A. 108, 14216–14221 (2011). doi: 10.1073/pnas.1111497108; pmid: 21825125 7. M. Sangesland et al., Germline-encoded affinity for cognate antigen enables vaccine amplification of a human broadly neutralizing response against influenza virus. Immunity 51, 735–749.e8 (2019). doi: 10.1016/j.immuni.2019.09.001; pmid: 31563464

y g

Human tonsil organoids

RE FERENCES AND NOTES

y

At various time points after infection, lungs were removed and stored at −80°C until virus quantification was performed. Lungs were homogenized individually in 1 ml RPMI. Virus was titrated in duplicate from lung homogenates by standard plaque assay on MDCK cells, and titers were reported as PFU per milliliter RPMI. Briefly, 1 day before infection, MDCK cells were seeded in six-well plates. The next day, the cells were washed once with 2 ml Dulbecco’s modified Eagle’s medium (DMEM) and incubated with virus diluted in DMEM containing 2% BSA, 200 U/ml penicillin and 200 mg/ml streptomycin for 1 hour at 35°C. After incubation, the virus inoculum was removed, and the cells were overlaid with DMEM containing 1.6% agarose solution and 2 mg/ml L-(tosylamido-2-phenyl) ethyl chloromethyl ketone (TPCK)–treated trypsin (Sigma). The plates were incubated at 35°C for 48 hours. Subsequently, the plaques were visualized by staining with crystal violet.

For statistical analysis, samples were randomly selected. Data distribution was checked using Shapiro-Wilk test for normality, and thereafter, analysis for multiple comparison was carried out using Kruskal-Wallis H test with Dunn’s multiple comparison test. To compare two groups, we used either Mann-Whitney U test or Student’s t test. One-way ANOVA with Tukey’s HSD method was used to analyze the difference in B cell diversity across different mice, while paired Student’s t test with multiple test correction by FDR was used to determine the enrichment of VH segments in GC B cells compared with naïve B cells. Statistical analyses were performed using Prism (version 9, GraphPad Software) and R version 3.6.3.

g

Quantification of viral load in the lung

Statistical analysis

8. Z. Shulman et al., Dynamic signaling by T follicular helper cells during germinal center B cell selection. Science 345, 1058–1062 (2014). doi: 10.1126/science.1257861; pmid: 25170154 9. C. D. Allen, T. Okada, H. L. Tang, J. G. Cyster, Imaging of germinal center selection events during affinity maturation. Science 315, 528–531 (2007). doi: 10.1126/science.1136736; pmid: 17185562 10. A. Cherukuri, P. C. Cheng, H. W. Sohn, S. K. Pierce, The CD19/ CD21 complex functions to prolong B cell antigen receptor signaling from lipid rafts. Immunity 14, 169–179 (2001). doi: 10.1016/S1074-7613(01)00098-X; pmid: 11239449 11. G. Fu et al., Metabolic control of TFH cells and humoral immunity by phosphatidylethanolamine. Nature 595, 724–729 (2021). doi: 10.1038/s41586-021-03692-z; pmid: 34234346 12. S. Minguet et al., Caveolin-1-dependent nanoscale organization of the BCR regulates B cell tolerance. Nat. Immunol. 18, 1150–1159 (2017). doi: 10.1038/ni.3813; pmid: 28805811 13. C. M. Högerkorp, C. A. Borrebaeck, The human CD77- B cell population represents a heterogeneous subset of cells comprising centroblasts, centrocytes, and plasmablasts, prompting phenotypical revision. J. Immunol. 177, 4341–4349 (2006). doi: 10.4049/jimmunol.177.7.4341; pmid: 16982868 14. M. Mangeney, Y. Richard, D. Coulaud, T. Tursz, J. Wiels, CD77: An antigen of germinal center B cells entering apoptosis. Eur. J. Immunol. 21, 1131–1140 (1991). doi: 10.1002/ eji.1830210507; pmid: 1709864 15. V. Pascual et al., Analysis of somatic mutation in five B cell subsets of human tonsil. J. Exp. Med. 180, 329–339 (1994). doi: 10.1084/jem.180.1.329; pmid: 8006591 16. M. D. Maloney, C. A. Lingwood, CD19 has a potential CD77 (globotriaosyl ceramide)-binding site with sequence similarity to verotoxin B-subunits: Implications of molecular mimicry for B cell adhesion and enterohemorrhagic Escherichia coli pathogenesis. J. Exp. Med. 180, 191–201 (1994). doi: 10.1084/ jem.180.1.191; pmid: 7516406 17. T. Mori et al., Globotriaosyl ceramide (CD77/Gb3) in the glycolipid-enriched membrane domain participates in B-cell receptor–mediated apoptosis by regulating Lyn kinase activity in human B cells. Exp. Hematol. 28, 1260–1268 (2000). doi: 10.1016/S0301-472X(00)00538-5; pmid: 11063874 18. I. Morace et al., Renal globotriaosylceramide facilitates tubular albumin absorption and its inhibition protects against acute kidney injury. Kidney Int. 96, 327–341 (2019). doi: 10.1016/ j.kint.2019.02.010; pmid: 31101366 19. H. Chen et al., BCR selection and affinity maturation in Peyer’s patch germinal centres. Nature 582, 421–425 (2020). doi: 10.1038/s41586-020-2262-4; pmid: 32499646 20. A. L. Bothwell et al., Heavy chain variable region contribution to the NPb family of antibodies: Somatic mutation evident in a g2a variable region. Cell 24, 625–637 (1981). doi: 10.1016/ 0092-8674(81)90089-1; pmid: 6788376 21. J. Jacob, J. Przylepa, C. Miller, G. Kelsoe, In situ studies of the primary immune response to (4-hydroxy-3-nitrophenyl) acetyl. III. The kinetics of V region mutation and selection in germinal center B cells. J. Exp. Med. 178, 1293–1307 (1993). doi: 10.1084/jem.178.4.1293; pmid: 8376935 22. G. D. Victora et al., Germinal center dynamics revealed by multiphoton microscopy with a photoactivatable fluorescent reporter. Cell 143, 592–605 (2010). doi: 10.1016/ j.cell.2010.10.032; pmid: 21074050 23. D. Dominguez-Sola et al., The FOXO1 transcription factor instructs the germinal center dark zone program. Immunity 43, 1064–1074 (2015). doi: 10.1016/j.immuni.2015.10.015; pmid: 26620759 24. T. Inoue et al., The transcription factor Foxo1 controls germinal center B cell proliferation in response to T cell help. J. Exp. Med. 214, 1181–1198 (2017). doi: 10.1084/jem.20161263; pmid: 28351982 25. S. Sander et al., PI3 kinase and FOXO1 transcription factor activity differentially control B cells in the germinal center light and dark zones. Immunity 43, 1075–1086 (2015). doi: 10.1016/ j.immuni.2015.10.021; pmid: 26620760 26. Y. Xu, K. Fairfax, A. Light, N. D. Huntington, D. M. Tarlinton, CD19 differentially regulates BCR signalling through the recruitment of PI3K. Autoimmunity 47, 430–437 (2014). doi: 10.3109/08916934.2014.921810; pmid: 24953501 27. Y. Wang et al., The physiologic role of CD19 cytoplasmic tyrosines. Immunity 17, 501–514 (2002). doi: 10.1016/ S1074-7613(02)00426-0; pmid: 12387743 28. D. A. Tuveson, R. H. Carter, S. P. Soltoff, D. T. Fearon, CD19 of B cells as a surrogate kinase insert region to bind phosphatidylinositol 3-kinase. Science 260, 986–989 (1993). doi: 10.1126/science.7684160; pmid: 7684160

p

Serum samples were pretreated with receptordestroying enzyme overnight at 37°C before being heated at 56°C for 30 min and diluted with minimum essential medium to a final serum dilution of 1:10. Serum antibody titers were measured by hemagglutination inhibition (HI) and microneutralization (NA; neutralizing antibody) assays following standard procedures. For HI assays, we used U-bottom 96-well microtiter plates. Twofold serial dilutions of sera (25 ml) starting at 1:10 were diluted with PBS and mixed with an equal volume of 4 hemagglutinating units of either H1N1 or H3N2 virus. The mixture of diluted serum and virus was incubated for 30 min at RT. Fifty microliters of 0.5% turkey red blood cells were added to the antigen/serum mixture and incubated for 45 min at RT. HI antibody titers were expressed as the reciprocal of the highest serum dilution that could prevent hemagglutination. For the NA assay, 50 ml of serial twofold dilutions of the sera were prepared and subsequently incubated with equal volumes of 2 × 103 TCID 50/ml of either H1N1 (PR8) or H3N2 (HK68, X-31) for 1 hour at 35°C. Thereafter, 1.5 × 104 MDCK cells in 96-well plates were incubated with serum/virus mixtures at 35°C for 72 hours. Fifty microliters of culture supernatant from each well was incubated with 50 ml of 1% turkey red blood cell in 96well V bottom plate for 1 hour at RT. NA titers were defined as the highest dilution factor which can inhibit the infection of MDCK cells with virus.

with antigens as described previously (42). For culture of cryopreserved cells, frozen cells were washed twice in complete medium (RPMI with GlutaMAX, 10% FBS, 1 × nonessential amino acids, 1 × sodium pyruvate, 1 × penicillin/ streptomycin, 1 × Normocin from InvivoGen), and 2 × 107 cells per ml were suspended in complete RPMI. One hundred microliters of cell suspension was then plated per well onto permeable (0.4-mm pore size) membranes (24well size polycarbonate membranes in standard 12-well plate), with the lower chamber consisting of complete medium (1 ml for 12-well plates, 200 ml for 96-well plates) supplemented with 1 mg/ml of recombinant human B cell– activating factor (BAFF, BioLegend). One microgram per milliliter rHA from H1N1 (PR8) was then added directly to the cultured tonsil dissociated cells. For adjuvant testing, either alum (0.01%) or Gb3 (10 mg/ml) was added directly into the culture immediately after antigen addition. Cultures were incubated at 37°C and 5% CO2 humidity for up to 21 days.

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

41.

42.

43.

44.

45.

46.

47.

48.

50.

52.

AC KNOWLED GME NTS

We thank K. Arnett and the Center for Macromolecular Interactions at Harvard Medical School for assistance with ITC experiments. Funding: This work was supported by National Institutes of Health grants R01 AI077595 to F.W.A. and R01 AI136939 to F.W. F.W.A. is an investigator of the Howard Hughes Medical Institute. Illustrations used in the figures were created using BioRender.com. Author contributions: P.S. and F.W. designed the study, performed data analysis, and wrote the manuscript. P.S., X.Z., K.L., Y.Z., Y.H., J.H.K., M.L., and C.W. performed the experiments. Y.Z., A.Y.Y., Q.C., J.H., and F.W.A. analyzed BCR sequencing data. D.R.B. synthesized the Gb3 analog. Y.K., Ke.F., and Ko.F. generated A4galt-KO mice. All authors reviewed the manuscript. Competing interests: The authors declare that they have no competing financial interests. F.W. and P.S. are listed as inventors on a patent application filed by Boston Children’s Hospital (provisional application number: 63/452,962), which covers the use of the lipid Gb3 in vaccination. Data and materials availability: All data needed to evaluate the conclusions are available in the manuscript or the supplementary materials. Raw BCR sequencing data reported in this paper have been deposited in the Gene Expression Omnibus (GEO) database under the accession number GSE247797. Rep-SHM-Seq was performed as described previously using our established bioinformatics pipeline (19, 53). A4galt-KO mice are available from Ko.F. under a material agreement with Nagoya University, Japan. 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

51.

53. A. Y. Ye, Yyx2626/HTGTSrep: v0.1.0-alpha pre-release (v0.1.0-alpha), Zenodo (2023); https://doi.org/10.5281/ zenodo.10379657. 54. S. G. Lin et al., Highly sensitive and unbiased approach for elucidating antibody repertoires. Proc. Natl. Acad. Sci. U.S.A. 113, 7846–7851 (2016). doi: 10.1073/pnas.1608649113; pmid: 27354528 55. G. E. Crooks, G. Hon, J. M. Chandonia, S. E. Brenner, WebLogo: A sequence logo generator. Genome Res. 14, 1188–1190 (2004). doi: 10.1101/gr.849004; pmid: 15173120

g

49.

hypermutation. Nature 509, 637–640 (2014). doi: 10.1038/ nature13300; pmid: 24805232 Z. Jing, M. J. McCarron, M. L. Dustin, D. R. Fooksman, Germinal center expansion but not plasmablast differentiation is proportional to peptide-MHCII density via CD40-CD40L signaling strength. Cell Rep. 39, 110763 (2022). doi: 10.1016/ j.celrep.2022.110763; pmid: 35508132 C. H. Yeh, T. Nojima, M. Kuraoka, G. Kelsoe, Germinal center entry not selection of B cells is controlled by peptide-MHCII complex density. Nat. Commun. 9, 928 (2018). doi: 10.1038/ s41467-018-03382-x; pmid: 29500348 L. E. Wagar et al., Modeling human adaptive immune responses with tonsil organoids. Nat. Med. 27, 125–135 (2021). doi: 10.1038/s41591-020-01145-0; pmid: 33432170 S. J. Keppler et al., Wiskott-Aldrich syndrome interacting protein deficiency uncovers the role of the co-receptor CD19 as a generic hub for PI3 kinase signaling in B cells. Immunity 43, 660–673 (2015). doi: 10.1016/j.immuni.2015.09.004; pmid: 26453379 A. Watanabe et al., Antibodies to a conserved influenza head interface epitope protect by an IgG subtype-dependent mechanism. Cell 177, 1124–1135.e16 (2019). doi: 10.1016/ j.cell.2019.03.048; pmid: 31100267 D. J. DiLillo, P. Palese, P. C. Wilson, J. V. Ravetch, Broadly neutralizing anti-influenza antibodies require Fc receptor engagement for in vivo protection. J. Clin. Invest. 126, 605–610 (2016). doi: 10.1172/JCI84428; pmid: 26731473 D. J. DiLillo, G. S. Tan, P. Palese, J. V. Ravetch, Broadly neutralizing hemagglutinin stalk-specific antibodies require FcgR interactions for protection against influenza virus in vivo. Nat. Med. 20, 143–151 (2014). doi: 10.1038/nm.3443; pmid: 24412922 A. Impagliazzo et al., A stable trimeric influenza hemagglutinin stem as a broadly protective immunogen. Science 349, 1301–1306 (2015). doi: 10.1126/science.aac7263; pmid: 26303961 F. Krammer, P. Palese, Advances in the development of influenza virus vaccines. Nat. Rev. Drug Discov. 14, 167–182 (2015). doi: 10.1038/nrd4529; pmid: 25722244 T. Okuda et al., Targeted disruption of Gb3/CD77 synthase gene resulted in the complete deletion of globo-series glycosphingolipids and loss of sensitivity to verotoxins. J. Biol. Chem. 281, 10230–10235 (2006). doi: 10.1074/ jbc.M600057200; pmid: 16476743 M. Espéli et al., FcgRIIb differentially regulates pre-immune and germinal center B cell tolerance in mouse and human. Nat. Commun. 10, 1970 (2019). doi: 10.1038/ s41467-019-09434-0; pmid: 31036800 J. M. Kavran et al., Specificity and promiscuity in phosphoinositide binding by pleckstrin homology domains. J. Biol. Chem. 273, 30497–30508 (1998). doi: 10.1074/jbc.273.46.30497; pmid: 9804818

p

29. K. J. Susa, S. Rawson, A. C. Kruse, S. C. Blacklow, Cryo-EM structure of the B cell co-receptor CD19 bound to the tetraspanin CD81. Science 371, 300–305 (2021). doi: 10.1126/ science.abd9836; pmid: 33446559 30. K. J. Susa, T. C. Seegar, S. C. Blacklow, A. C. Kruse, A dynamic interaction between CD19 and the tetraspanin CD81 controls B cell co-receptor trafficking. eLife 9, e52337 (2020). doi: 10.7554/eLife.52337; pmid: 32338599 31. B. Zimmerman et al., Crystal structure of a full-length human tetraspanin reveals a cholesterol-binding pocket. Cell 167, 1041–1051.e11 (2016). doi: 10.1016/j.cell.2016.09.056; pmid: 27881302 32. W. Luo, F. Weisel, M. J. Shlomchik, B cell receptor and CD40 signaling are rewired for synergistic induction of the c-Myc transcription factor in germinal center B cells. Immunity 48, 313–326.e5 (2018). doi: 10.1016/j.immuni.2018.01.008; pmid: 29396161 33. P. K. Mattila et al., The actin and tetraspanin networks organize receptor nanoclusters to regulate B cell receptor-mediated signaling. Immunity 38, 461–474 (2013). doi: 10.1016/ j.immuni.2012.11.019; pmid: 23499492 34. P. P. Domeier et al., B-cell-intrinsic type 1 interferon signaling is crucial for loss of tolerance and the development of autoreactive B cells. Cell Rep. 24, 406–418 (2018). doi: 10.1016/j.celrep.2018.06.046; pmid: 29996101 35. C. L. Swanson et al., Type I IFN enhances follicular B cell contribution to the T cell–independent antibody response. J. Exp. Med. 207, 1485–1500 (2010). doi: 10.1084/ jem.20092695; pmid: 20566717 36. M. W. Dahlgren et al., Type I interferons promote germinal centers through B cell intrinsic signaling and dendritic cell dependent Th1 and Tfh cell lineages. Front. Immunol. 13, 932388 (2022). doi: 10.3389/fimmu.2022.932388; pmid: 35911733 37. R. K. Abbott et al., Precursor frequency and affinity determine B cell competitive fitness in germinal centers, tested with germline-targeting HIV vaccine immunogens. Immunity 48, 133–146.e6 (2018). doi: 10.1016/j.immuni.2017.11.023; pmid: 29287996 38. J. Merkenschlager et al., Dynamic regulation of TFH selection during the germinal centre reaction. Nature 591, 458–463 (2021). doi: 10.1038/s41586-021-03187-x; pmid: 33536617 39. O. Bannard et al., Ubiquitin-mediated fluctuations in MHC class II facilitate efficient germinal center B cell responses. J. Exp. Med. 213, 993–1009 (2016). doi: 10.1084/ jem.20151682; pmid: 27162138 40. A. D. Gitlin, Z. Shulman, M. C. Nussenzweig, Clonal selection in the germinal centre by regulated proliferation and

SUPPLEMENTARY MATERIALS

science.org/doi/10.1126/science.adg0564 Figs. S1 to S12 Table S1 MDAR Reproducibility Checklist Submitted 29 November 2022; resubmitted 22 August 2023 Accepted 20 December 2023 10.1126/science.adg0564

y g y ,

Sharma et al., Science 383, eadg0564 (2024)

16 February 2024

14 of 14

RES EARCH

RESEARCH ARTICLE SUMMARY



METABOLISM

Hepatic glycogenesis antagonizes lipogenesis by blocking S1P via UDPG Jie Chen†, Yabo Zhou†, Zhuohang Liu, Yan Lu, Yishen Jiang, Kexin Cao, Nannan Zhou, Dianheng Wang, Chaoqi Zhang, Ning Zhou, Keqing Shi, Lu Zhang, Li Zhou, Zhenfeng Wang, Huafeng Zhang, Ke Tang, Jingwei Ma, Jiadi Lv, Bo Huang*

Glucose

RESULTS: Using both mouse and human primary

hepatocytes, we found that glucose carbon swiftly flowed to glycogenesis before lipogenesis. Knockdown of glycogenic enzymes switched the priority to lipogenesis, emphasizing the antagonistic relationship. Moreover,

UDPG Glycogen

Glucose

Fatty acid

Golgi

S1P

S2P SREBP

SREBP

SREBP

SREBP

Cytoplasm

C

N

N

N

N Nucleus Lipogenic genes

Hepatic glycogenesis antagonizes lipogenesis by blocking S1P via UDPG. Upon feeding, hepatocytes take up and allow the conversion of food-derived glucose into glycogen (top). The intermediate metabolite UDPG can be transported into the Golgi apparatus, where it binds to S1P and abrogates S1P-mediated cleavage of SREBP, thereby inhibiting lipogenesis in hepatocytes (bottom). S2P, site-2 protease; SRE, sterol regulatory element. [Figure created with BioRender.com]. Chen et al., Science 383, 718 (2024)

16 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 J. Chen et al., Science 383, eadi3332 (2024). DOI: 10.1126/science.adi3332

READ THE FULL ARTICLE AT https://doi.org/10.1126/science.adi3332 1 of 1

,

SRE SREBP

SLC35F5

y

UDPG

y g

Glucose

CONCLUSION: After food uptake, the basic energy molecule glucose is transported to the liver via portal veins, leaving hepatocytes with the physiological task of dealing with excess glucose by using either glycogenesis or lipogenesis to store energy. Although insulin instructs hepatocytes to activate both glycogenesis and lipogenesis to store glucose molecules, our data show that glycogenesis in the liver—by virtue of the intermediate metabolite UDPG, which is shunted to the Golgi apparatus to inactivate S1P enzymatic activity—leads to the inhibition of SREBP cleavage, thus inhibiting the de novo synthesis of fatty acids. The identification of such a metabolic regulation, which probably also occurs in other glycogen-producing cells, provides insights into how hepatocytes precisely deal with sugar and lipid storage. Therapeutically, this UDPG-mediated regulation of lipogenesis may open avenues for the management of abnormal lipid metabolism in humans.

y

cogenesis or lipogenesis to store glucose energy, which, however, cause different consequences. Glycogen degradation promotes reactive oxygen species (ROS) clearance, but fatty acid oxidation leads to ROS generation. Given that increased ROS levels contribute to steatosis and further induce nonalcoholic fatty liver disease (NAFLD), hepatocytes may preferentially exploit glycogenesis to store glucose and supply energy when required.

g

RATIONALE: Hepatocytes may choose either gly-

eties now take up ample food and leave their hepatocytes with the physiological task of storing excess glucose in the form of glycogen or lipid. Despite the importance of these two forms of energy storage, the manner by which hepatocytes select glycogenesis or lipogenesis for glucose carbon storage remains unclear. In this study, we provide evidence that the storage of glucose as glycogen is prioritized over storage as lipid. We found that uridine diphosphate glucose (UDPG), a metabolic intermediate of glycogenesis, can be transported to the Golgi apparatus of hepatocytes, where it binds to site-1 protease (S1P) and inhibits S1Pmediated cleavage of sterol regulatory elementbinding proteins (SREBPs), thereby inhibiting lipogenesis and facilitating glycogenesis.

p

INTRODUCTION: Individuals in developed soci-

we found that UDPG, a key glycogenesis metabolite, could down-regulate lipogenic genes and thereby inhibit lipogenesis. SREBPs in the Golgi apparatus is cleaved by S1P to generate its active form, thus entering the nucleus to promote lipogenic gene expression. By using coimmunoprecipitation, click chemistry, liquid chromatography–tandem mass spectrometry (LC-MS/MS), and drug affinity responsive target stability (DARTS) assay, we demonstrate that UDPG directly binds to S1P in vivo and in vitro. This binding did not alter S1P mRNA expression but did down-regulate its protein expression in hepatocytes. By conducting molecular docking, point mutation assay, and circular dichroism spectra, we further demonstrate that the Asn438 site of S1P was required for UDPG binding; upon binding, the conformation of S1P was changed, resulting in S1P degradation through the ubiquitin-proteasome pathway, thereby impairing SREBP cleavage and subsequent inhibition of lipogenesis. Solute carrier family 35 member F5 (SLC35F5) was identified as the transporter that mediated UDPG entry into the Golgi apparatus. Administration of UDPG to mice resulted in the down-regulation of lipogenic genes, which reduced fatty acid synthesis and alleviated hepatic steatosis. Consistent results were also observed in primary human hepatocytes and tissues from patients with NAFLD, emphasizing the translational relevance of these findings.

RES EARCH

RESEARCH ARTICLE



METABOLISM

Hepatic glycogenesis antagonizes lipogenesis by blocking S1P via UDPG Jie Chen1†, Yabo Zhou1†, Zhuohang Liu1, Yan Lu1, Yishen Jiang1, Kexin Cao1, Nannan Zhou1, Dianheng Wang1, Chaoqi Zhang1,2, Ning Zhou3, Keqing Shi4, Lu Zhang5, Li Zhou1, Zhenfeng Wang1, Huafeng Zhang6, Ke Tang7, Jingwei Ma8, Jiadi Lv1, Bo Huang1,7*

16 February 2024

1 of 12

,

Chen et al., Science 383, eadi3332 (2024)

To explore the priority of glucose carbon flow to glycogenesis or lipogenesis, we replenished glucose-free medium–treated primary hepatocytes with glucose (Fig. 1A). We found that the hepatocytes swiftly used glucose to synthesize glycogen, and glycogen levels began to increase 15 min after replenishment, as evidenced by a

Next, we investigated the manner by which glycogenesis inhibited lipogenesis in hepatocytes. We speculated that intermediate metabolite(s) from glycogenesis exerted an inhibitory effect against lipogenesis. UDPG is a key metabolite during glycogenesis, which is regulated by both PGM1 and UGP2 (Fig. 2A). We found that knockdown of Pgm1 or Ugp2 resulted in decreases in UDPG in hepatocytes (Fig. 2B). However, the addition of UDPG inhibited glucose carbon flow to lipogenesis, as evidenced by decreased triglycerides (Fig. 2C) and reduced 13C-traced palmitic acid and oleic acid (Fig. 2D). In line with this, lipogenic genes, including Acaca, Fasn, and Scd1 were downregulated in UDPG-treated hepatocytes (Fig. 2E), suggesting that UDPG inhibits lipogenesis. Given that PGM1 catalyzes G6P to G1P and UGP2 catalyzes G1P to UDPG during glycogenesis, this finding raised the possibility that lipogenesis was regulated by G1P. However, the import of exogenous G1P into Ugp2-knockdown hepatocytes did not affect lipogenesis (fig. S2A). In addition, blockade of glycogen degradation by a glycogen phosphorylase inhibitor (GPI) increased glycogen and UDPG levels (Fig. 2, F and G), which also resulted in the reduction of lipogenesis in hepatocytes (Fig. 2, H and I, and fig. S2, B and C). By contrast, overexpression of the glycogen

y

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

Glycogenesis inhibits lipogenesis in hepatocytes

UDPG inhibits lipogenesis

y g

Department of Immunology and National Key Laboratory of Medical Molecular Biology, Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing 100005, China. 2Department of Thoracic Surgery, National Cancer Center/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing 100021, China. 3Department of Pathology, Sichuan Mianyang 404 Hospital, Sichuan 621000, China. 4Translational Medicine Laboratory, The First Affiliated Hospital of Wenzhou Medical University, Wenzhou 325035, China. 5Tianjin Centers for Disease Control and Prevention, Tianjin 300011, China. 6Department of Pathology, School of Basic Medicine, Tongji Medical College, Huazhong University of Science and Technology, Wuhan 430030, China. 7Department of Biochemistry and Molecular Biology, Tongji Medical College, Huazhong University of Science and Technology, Wuhan 430030, China. 8Department of Immunology, Tongji Medical College, Huazhong University of Science and Technology, Wuhan 430030, China.

y

1

ylase (ACACA) and fatty acid synthase (FASN) (9). The manner by which glucose selects glycogenesis or lipogenesis for carbon storage still remains unclear, despite the importance of these two forms of energy storage. A major difference between glycogen and fat storage probably lies in the regulation of reactive oxygen species (ROS) during their degradation to supply energy (10–15). The degradation of fatty acids via oxidation leads to ROS generation via reverse electron transport (16–18). However, glycogenolysis enhances ROS clearance by facilitating NADPH [reduced form of nicotinamide adenine dinucleotide phosphate (NADP+)] production via the pentose phosphate pathway (10–12, 15, 19). As a result, storing glucose via lipogenesis may pose a potential risk in that increased ROS levels contribute to steatosis and further induce nonalcoholic fatty liver disease (NAFLD) (20–22). To avoid this, hepatocytes may preferentially use glycogenesis to store glucose, thereby supplying energy when required. In this study, we hypothesized that glycogenesis is prioritized by antagonizing lipogenesis. We provide evidence that UDPG, a metabolic intermediate of glycogenesis, can be transported to the Golgi apparatus of hepatocytes, where it binds to site-1 protease (S1P) and inhibits S1P-mediated cleavage of sterol regulatory element–binding proteins (SREBPs), thereby inhibiting lipogenesis.

g

G

lycogenesis and lipogenesis are two fundamental pathways that cells use to store energy (1–3). After food consumption, the basic energy molecule glucose is transported to the liver via portal veins, leaving hepatocytes with the physiological task of how to deal with excess glucose. Insulin produced by pancreatic b cells guides hepatocytes to store glucose molecules as glycogen via glycogenesis or as fat via lipogenesis (4–6). Glycogenesis is initiated from glucose-6phosphate (G6P), which is converted to glucose1-phosphate (G1P) and results in the transfer of glucose from uridine diphosphate glucose (UDPG) to glycogen by glycogen synthase (5, 7, 8). During lipogenesis, glucose is oxidized via glycolysis to generate pyruvate, which is then converted into citrate in the mitochondria. The citrate is then transported to the cytosol, where it is split into acetyl–coenzyme A (acetyl-CoA) and oxaloacetate. Cytosolic acetyl-CoA is used to synthesize fatty acids by acetyl-CoA carbox-

p

The identification of mechanisms to store glucose carbon in the form of glycogen rather than fat in hepatocytes has important implications for the prevention of nonalcoholic fatty liver disease (NAFLD) and other chronic metabolic diseases. In this work, we show that glycogenesis uses its intermediate metabolite uridine diphosphate glucose (UDPG) to antagonize lipogenesis, thus steering both mouse and human hepatocytes toward storing glucose carbon as glycogen. The underlying mechanism involves transport of UDPG to the Golgi apparatus, where it binds to site-1 protease (S1P) and inhibits S1P-mediated cleavage of sterol regulatory element–binding proteins (SREBPs), thereby inhibiting lipogenesis in hepatocytes. Consistent with this mechanism, UDPG administration is effective at treating NAFLD in a mouse model and human organoids. These findings indicate a potential opportunity to ameliorate disordered fat metabolism in the liver.

colorimetric assay (Fig. 1B). By contrast, lipogenesis was initiated 1 hour later (Fig. 1B). This priority of glycogenesis was also confirmed by 13C tracing (Fig. 1, C to E). To exclude the possible influence of serum cholesterol on lipogenesis (23, 24), we additionally pretreated the cells with serum-free and glucose-free medium for 2 hours, which caused the lowest glycogen level (fig. S1A), and similar results were obtained (fig. S1, B to D), suggesting that hepatocytes store glucose in the form of glycogen as a priority. Liver phosphoglucomutase 1 (PGM1) is the first enzyme to catalyze glycogenesis. Knockdown of Pgm1 resulted in a swift switch of glucose carbon flow from glycogenesis to lipogenesis (Fig. 1, F to I, and fig. S1E). Consistent with this, knockdown of the gene encoding UDPG pyrophosphorylase 2 (Ugp2), the second glycogenic enzyme that catalyzes G1P to UDPG, also resulted in the switch to lipogenesis (fig. S1, F to J). Transfecting RNA interference (RNAi)–resistant Pgm1 or Ugp2 vectors into the knockdown cells rescued glycogenesis and inhibited the carbon flow to lipogenesis (Fig. 1, J to L, and fig. S1, K to M). In addition, using 13C-glucose tracing in Pgm1+/− mice, we found that more carbon flowed to fatty acids in these mice relative to their wild-type counterparts (fig. S1N). Together, these results suggest that glycogenesis likely antagonizes lipogenesis in hepatocytes during glucose carbon assimilation.

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

p g y

16 February 2024

mobilizes the intermediate metabolite UDPG to antagonize hepatic lipogenesis.

,

Chen et al., Science 383, eadi3332 (2024)

P2Y14 is recognized by UDPG and transduces signals in various cell types (26, 27). Blocking P2Y14 with either PPTN (4,7-disubstituted 2naphthoic acid derivative) or siRNA (small interfering RNA) did not alter the effect of UDPG on hepatic lipogenesis (fig. S2, J to O). Given the important role of insulin in lipogenesis (28), we additionally treated the 2-hourfasted hepatocytes with 13C-glucose in the presence of UDPG, insulin, or UDPG and insulin. Although lipogenesis lagged behind glycogenesis in the presence of insulin, we found that UDPG still maintained its inhibitory effect on lipogenesis even in the presence of insulin (fig. S2, P to T). This result was not due to UDPG treatment inactivating insulin signaling, given the unaltered phosphorylation of AKT and S6K (fig. S2U). Together, these results suggest that glycogenesis

y

phosphorylase liver form (PYGL) down-regulated glycogen and UDPG levels (Fig. 2, J and K), leading to enhanced lipogenesis (Fig. 2, L and M, and fig. S2, D and E). In addition, we excluded the regulation of lipogenesis by glycogen itself, based on the Gys2 knockdown, which resulted in a decrease of glycogen and triglyceride (Fig. 2N) but an increase of UDPG in hepatocytes (fig. S2F). In line with this, 13 C-glucose tracing showed that M+6 UDPG changed inversely with lipogenesis in primary hepatocytes, which had been subjected to glucose and serum-free treatment in advance (Fig. 2O and fig. S2G). UDPG is normally converted to uridine diphosphate glucuronic acid (UDPGA) by UDP-glucose 6-dehydrogenase (UGDH) (25). We found that the addition of UDPGA had no effect on lipogenesis in hepatocytes (Fig. 2P and fig. S2, H and I). The purinergic receptor

serum- and glucose-free medium for 2 hours, followed by 1 hour of 13C-glucose treatment. M+6 labeled glucose in glycogen (H) and the incorporation of 13 C into palmitic acid and oleic acid (I) were detected by LC-MS/MS. (J to L) Hepatocytes were transfected with Pgm1 siRNA for 48 hours in the presence or absence of exogenous RNAi-resistant Pgm1 cDNA. Glycogen (J) and TG (K) levels were measured by a colorimetric assay. The hepatocytes were treated with serum- and glucose-free medium for 2 hours, followed by 1 hour of 13C-glucose treatment. The incorporation of 13C into palmitic acid and oleic acid were analyzed by LC-MS/MS (L). OE, overexpression. Data are presented as means ± SEM. p values were calculated using one-way analysis of variance (ANOVA).

y g

Fig. 1. Glycogenesis antagonizes lipogenesis in hepatocytes. (A) Workflow for the discovery of glucose carbon flow to glycogen or lipid. Primary hepatocytes were washed and cultured in glucose-free medium for 2 hours, followed by glucose or 13C-glucose treatment. (B) Glycogen and triglyceride (TG) levels were measured by a colorimetric assay. (C) Schematic of the carbon flow in the storage of glucose. HK, hexokinase; OAA, oxaloacetic acid; TCA, tricarboxylic acid cycle. (D to E) M+6-labeled glucose in glycogen (D) and M+4-labeled palmitic acid (E) were analyzed by LC-MS/MS. (F and G) Hepatocytes were transfected with Pgm1 siRNA for 48 hours, then glycogen (F) and TG (G) levels were measured by a colorimetric assay. siNC, siRNA negative control. (H and I) Hepatocytes, transfected with Pgm1 siRNA for 45 hours, were treated with

UDPG disrupts SREBP cleavage in the Golgi apparatus

Next, we investigated the molecular basis by which UDPG inhibits lipogenesis. SREBP1a, SREBP1c, and SREBP2 are the key transcription factors involved in lipid synthesis. SREBP-1 expression is regulated by insulin, and SREBP-2 expression relies more on sterols (29). SREBP1c, as a dominant form of SREBP1 in the liver, is crucial for hepatic lipogenesis; it is initially present in the endoplasmic reticulum (ER) membrane in the full-length form, which interacts with SREBP cleavage-activating protein (SCAP) (30, 31). In the absence of cholesterol, SCAP escorts SREBP into COPII vesicles and then into the Golgi, where S1P and site-2 protease 2 of 12

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

p g y g

and oleic acid was analyzed by LC-MS/MS. (J to L) Hepatocytes were transfected with Pygl-overexpressing plasmid (Pygl-OE) for 24 hours, glycogen levels were measured by a colorimetric assay (J), and the amount of UDPG was analyzed by LC-MS/MS (K). TG levels were measured by a colorimetric assay (L). (M) Hepatocytes with or without Pygl overexpression were treated with serum- and glucose-free medium for 2 hours, followed by 1 hour of 13C-glucose treatment. The incorporation of 13C into palmitic acid and oleic acid were analyzed by LC-MS/MS. (N) Hepatocytes were transfected with Gys2 siRNA for 48 hours, then glycogen (left) and TG (right) levels were measured by a colorimetric assay. (O) Hepatocytes were treated with serum- and glucose-free medium for 2 hours, followed by the addition of 13C-labeled glucose (2 mg/ml). The amount of intracellular 13C-UDPG M+6 was analyzed by LC-MS/MS. (P) Hepatocytes were treated with UDPGA for 48 hours, then TG levels were measured. Data are presented as means ± SEM. p values were calculated using one-way ANOVA [(B) to (D), (N), and (P)] or unpaired two-tailed Student’s t test [(F) to (M)]. ns, not significant.

y

Fig. 2. UDPG is identified as an intermediate metabolite to inhibit lipogenesis. (A) Schematic representation of glycogen metabolism. (B) Hepatocytes were transfected with Pgm1 or Ugp2 siRNA for 48 hours, and the amounts of UDPG were detected by LC-MS/MS. (C to E) Hepatocytes were treated with UDPG for 48 hours, then TG levels were measured by a colorimetric assay (C). The hepatocytes were treated with serum- and glucose-free medium for 2 hours, followed by 1 hour of 13C-glucose treatment. The incorporation of 13 C into palmitic acid and oleic acid was analyzed by LC-MS/MS (D). The expression of Acaca, Fasn, Scd1, Acly, and Acss2 was determined by real-time polymerase chain reaction (PCR) (E). PBS, phosphate-buffered saline. (F to H) Hepatocytes were treated with 20 mM GPI for 24 hours. Glycogen levels were measured by a colorimetric assay (F). The amount of UDPG was analyzed by LC-MS/MS (G). TG levels were measured by a colorimetric assay (H). DMSO, dimethyl sulfoxide. (I) Hepatocytes, stimulated with 20 mM GPI for 24 hours, were treated with serum- and glucose-free medium for 2 hours, followed by 1 hour of 13C-glucose treatment. The incorporation of 13C into palmitic acid

y

Chen et al., Science 383, eadi3332 (2024)

line with this notion, knocking down Pgm1 or Ugp2 to reduce UDPG levels increased nSREBP quantity (Figs. 2B and 3E), concomitant with the up-regulation of lipogenic genes (fig. S3D), which, however, was reversed by the addition of UDPG (Fig. 3, F and G, and fig. S3, E and F), in parallel with reduced lipogenesis (Fig. 3H and fig. S3G). Using Flag-SREBP1c– transfected hepatocytes (Flag epitope tag located at the N terminus), we found that the Flag tag could be transferred from the ER to the nucleus of hepatocytes upon cholesterol deprivation (Fig. 3I); however, the addition of UDPG did not affect the entry of the Flag tag into the Golgi but did disrupt entry into the nucleus (Fig. 3, I and J). Apart from SREBP1c, carbohydrate responsive element-binding pro-

16 February 2024

tein (ChREBP) is another transcriptional activator of hepatic lipogenesis (34). However, UDPG treatment did not affect the mRNA and protein (total and nuclear) levels of ChREBP in hepatocytes (fig. S3, H to J). In addition, UDPG still maintained its inhibition on lipogenesis in ChREBPa-knockdown hepatocytes (fig. S3, K to N). Together, these results suggest that UDPG negatively regulates SREBP cleavage in the Golgi apparatus, thereby inhibiting hepatic lipogenesis. UDPG blocks SREBP cleavage by binding S1P

Next, we sought to elucidate how UDPG regulates SREBP cleavage. Using fluorescencelabeled UDPG to treat cholesterol-deprived hepatocytes, we found that UDPG was not 3 of 12

,

(S2P) cleave SREBP into n-SREBP (N terminus of SREBP), which transactivates lipogenic genes (24, 31–33). In UDPG-treated hepatocytes, SREBP1a, SREBP1c, and SREBP2 mRNAs were not affected (Fig. 3A), whereas the cleaved form of SREBP1 was inhibited (Fig. 3B). Consistent with this, less n-SREBP1 entered the nucleus in UDPG-treated hepatocytes (Fig. 3C), concomitant with the down-regulation of Scd1, Fasn, and Acaca (Fig. 2E). Similar results were obtained for nuclear SREBP2 and its target genes (Hmgcr and Hmgcs1) (Fig. 3D and fig. S3, A and B), suggesting that UDPG interferes with SREBP cleavage. In addition, consistent with the previous results (fig. S2T), we found that UDPG treatment also decreased n-SREBP1 levels in the presence of insulin (fig. S3C). In

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

p g y

16 February 2024

riched in fatty acid metabolic pathways (fig. S4B). Meanwhile, using human embryonic kidney (HEK) 293T cells that were forced to overexpress S1P and SREBP1c, we found that S1P-mediated SREBP1c cleavage was inhibited by the addition of UDPG (Fig. 4I). Such UDPG-mediated effects, however, did not happen in S1P-knockdown hepatocytes (fig. S4, C to F). Using a biolayer interferometry (BLI) assay, we found that UDPG directly bound to S1P (Fig. 4J and fig. S4G). To explore the structural basis underlying UDPG regulation of S1P activity, we performed a molecular docking simulation using Molecular Operating Environment (MOE) software based 4 of 12

,

Chen et al., Science 383, eadi3332 (2024)

n-SREBP. We speculated that UDPG bound S1P and/or S2P, thus interfering with SREBP cleavage. Using hemagglutinin (HA)–tagged SREBP1c, S1P, or S2P to incubate with UDPG, we found that UDPG bound S1P rather than SREBP1c or S2P, as evidenced by LC-MS/MS (Fig. 4, E and F). Moreover, we used a click chemistry approach to pull down proteins linked to UDP-2-N3-Glc (Azido-UDPG) for MS (35). The result showed that UDPG only bound S1P rather than S2P or SREBP (Fig. 4, G and H). Analyses by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) showed that UDPG-binding proteins were en-

y

localized in the ER but was present in the Golgi (Fig. 4A), as evidenced by liquid chromatography– tandem mass spectrometry (LC-MS/MS) (Fig. 4B). Using fasting hepatocytes, we found that the supply of glucose allowed an initial increase of the Golgi-to-cytosolic UDPG ratio, which reached a peak 1 hour later and subsequently decreased, correlating with increased hepatocellular lipogenesis (fig. S4A). However, knockdown of Pgm1 or Ugp2, or overexpression of Pygl, reduced the content of UDPG in the Golgi (Fig. 4, C and D). S1P and S2P cleave SREBP at the luminal loop in the Golgi and membranespanning domain, respectively, to generate

hepatocytes were stimulated with 200 mM UDPG for 48 hours, and then TG levels were measured by a colorimetric assay. (I and J) After transfection with Flag-SREBP1c plasmid for 8 hours, hepatocytes were treated with UDPG for 32 hours and then cells were cultured in cholesterol-depleted medium (5% lipoprotein-deficient serum plus 10 mM sodium mevalonate and 1 mM lovastatin) with or without 25-hydroxycholesterol (25-HC) for 16 hours. Golgi fractions, ER fractions, and nuclear extracts were prepared, and SREBP1 was analyzed by Western blot (I). The localization of SREBP1 was observed under confocal microscope (scale bars are 20 mm) (J). FBS, fetal bovine serum; PDI, protein disulfide isomerase. Data are presented as means ± SEM. p values were calculated using one-way ANOVA [(A), (D), (F), and (H)] or unpaired two-tailed Student’s t test (C).

y g

Fig. 3. UDPG regulates SREBP cleavage in the Golgi apparatus. (A to C) Hepatocytes were treated with UDPG for 48 hours, then SREBP1a, SREBP1c, and SREBP2 expression was determined by real-time PCR (A). SREBP1was detected by Western blot (B) and confocal microscopy (scale bars are 20 mm) (C). DAPI, 4′,6-diamidino-2-phenylindole; n, nuclear; p, precursor. (D) Hepatocytes were treated with UDPG for 48 hours, then Hmgcr and Hmgcs1 expression was determined by real-time PCR. (E) Hepatocytes were transfected with Pgm1 or Ugp2 siRNA for 48 hours, then SREBP1 and SREBP2 were detected by Western blot. (F and G) Pgm1 siRNA transfected hepatocytes were stimulated with UDPG for 48 hours, then the localization of SREBP1 was determined by confocal microscopy (scale bars are 20 mm) (F), and SREBP1 and SREBP2 were detected by Western blot (G). (H) Pgm1 or Ugp2 siRNA–transfected

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

p g y y g y ,

Fig. 4. UDPG blocks SREBP cleavage by binding S1P. (A) Hepatocytes, transfected with pcDNA3.1-Golgi-mCherry or pcDNA3.1-ER-mCherry for 24 hours, were treated with UDPG-fluorescein conjugate in cholesterol-free medium for 12 hours, and imaged by confocal microscopy (scale bars are 20 mm). (B) Hepatocytes were treated with UDPG in cholesterol-free medium for 12 hours. UDPG levels in the ER and Golgi were analyzed by LC-MS/MS. (C and D) Hepatocytes were transfected with Pgm1, Ugp2 siRNA (C), or Pygl-OE plasmid (D), and Golgi was isolated for UDPG analysis. (E and F) HEK293T cells were transfected with pCMV-HA, pCMV-m-SREBP1c-HA, pCMV-m-Mbtps1HA, or pCMV-m-Mbtps2-HA plasmids for 48 hours. Cell lysates were Chen et al., Science 383, eadi3332 (2024)

16 February 2024

immunoprecipitated with anti-HA agarose beads, followed by the incubation of beads with 200 mM UDPG for 3 hours at 4°C. The amounts of UDPG bound to the beads were detected by LC-MS/MS. The workflow for the discovery of UDPG binding proteins by immunoprecipitate is shown in (E). Relative UDPG levels were calibrated to HA-tagged total protein concentration, as shown in (F). (G and H) Workflow for the discovery of UDPG binding proteins by a click chemistry approach (G). The top 20 enriched proteins were identified (H). (I) HEK293T cells overexpressing S1P and SREBP1c were treated with 200 mM UDPG for 24 hours, and SREBP1 in cell lysates was determined by Western blot. (J) Binding [dissociation constant (KD)] between murine S1P and UDPG was 5 of 12

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

measured by BLI, and 400 mM UDPGA was used as a control. (K) Electrostatic surface representation of S1P onto which UDPG docks. The most energetically favored pose of UDPG in the potential binding pocket was predicted by MOE software (left). In the zoomed-in structure, eight labeled residues are potential sites of mutations designed to inhibit the binding of S1P and UDPG (right). (L) HEK293T cells were transfected with HA-mbtps1–overexpressing plasmid. Forty-eight hours after transfection, cell lysates were immunoprecipitated with anti-HA agarose beads, and then the beads were incubated with 200 mM UDPG for 3 hours at 4°C. The amounts of UDPG bound to the beads were detected by LC-MS/MS. (M) HEK293T cells transfected with the indicated S1P and SREBP1c plasmids were treated with 200 mM UDPG for 24 hours, and then SREBP1 was detected by

SLC35F5 transports UDPG into the Golgi apparatus

6 of 12

,

16 February 2024

y

UDPG inhibition of hepatic lipogenesis offers an opportunity to manipulate and alleviate hepatic

y g

UDPG treatment alleviates hepatic steatosis in a mouse model

y

Given that glycogenesis occurs in the cytosol where UDPG is generated, we further investigated how UDPG was transported from the cytosol into the Golgi. Solute carrier family 35 (SLC35), a type III transmembrane protein that contains seven subfamilies from SLC35A to SLC35G, is thought to mediate the transport of cytosolic nucleotide sugars to the lumen of Golgi vesicles (37–39). By examining mRNA expression, we identified a large number of SLC35 members in the mouse hepatocytes (Fig. 5A). The top 10 candidates (a1, a3, a4, b1, d1, d2, e1, e2, f5, and f6) were knocked down using siRNAs. Flow cytometric analysis of the lipid-binding dye BODIPY showed that the fluorescence intensity was highest upon knockdown of Slc35f5 (Fig. 5B and fig. S5A). As a result, such Slc35f5 knockdown inhibited the entry of UDPG into the Golgi, as evidenced by both LC-MS/MS and fluorescent microscopy (Fig. 5, C and D), concomitant with the up-regulation of n-SREBP1, lipogenic gene expression, and fatty acid synthesis in hepatocytes (Fig. 5, E to G, and fig. S5, B and C). To validate the in vitro results in vivo, we knocked down Slc35f5 in hepatocytes in mice using hydrodynamics-based gene delivery (40). We found that the flow of glucose to glycogen was reduced but the flow to triglycerides was increased in the liver tissue of Slc35f5-knockdown mice (fig. S5, D to F). By contrast, forced overexpression of Slc35f5 promoted UDPG entry into the Golgi and subsequently down-regulated n-SREBP1, lipogenic gene expression, and fatty acid synthesis in cells (Fig. 5, H to L, and fig. S5, G and H). Moreover, UDPG treatment generated a stronger inhibition of lipogenesis in combination with Slc35f5 overexpression (Fig. 5, J to L, and fig. S5, G and H). Together, these results suggest that hepatocytes mobilize SLC35F5 to transport UDPG into the Golgi apparatus.

steatosis. After 12 hours of fasting, mice were intraperitoneally injected with glucose [2 g per kg body weight (g/kg)]. By dissecting the liver with a colorimetric assay, we found that glycogen levels started to increase 15 min after replenishment, and lipogenesis was detected in hepatocytes 6 min later (Fig. 6A). Notably, the lipogenic time point could be further delayed by intraperitoneal injection of UDPG (Fig. 6B). Such exogenous UDPG entry into hepatocytes was verified by intraperitoneally injecting the UDPG analog UDP-2-N3-Glc into the peritoneal cavity (fig. S6A). In line with the rise and fall of glycogenesis and lipogenesis, 13C-labeled glucose in vivo tracing showed that hepatic M+6 UDPG levels increased at the beginning but decreased later (fig. S6B). Then, we used UDPG (8 mg/kg once per day) to treat the mice with a normal diet for 7 days (Fig. 6C). The results showed that although the total triglyceride content decreased by only 13% (fig. S6C), the amount of newly synthesized fatty acids was markedly reduced compared with that in untreated mice (Fig. 6D), prompting us to speculate that UDPG may be a potential agent against NAFLD. To this end, we fed mice a high-fat diet (HFD) for 16 weeks to develop NAFLD, as evidenced by hepatic ballooning degeneration, lipid droplet accumulation, and the accumulation of triacylglycerol and cholesterol in the liver and serum (Fig. 6, E to H). Notably, periodic acid–Schiff (PAS) staining and LC-MS/MS analysis showed that glycogen content and UDPG levels were reduced in the liver (Fig. 6, I and J), concomitant with the down-regulation of Pgm1 and Ugp2 expression (Fig. 6K), suggesting that hepatic glycogenesis is limited in NAFLD mice. The HFD mice were continually fed a HFD but received UDPG treatment from the eighth week onward for 8 weeks. We found that UDPG treatment had no detectable influence on food intake but could reduce weight (Fig. 6L and fig. S6D). However, hematoxylin and eosin (H&E) staining showed that UDPG treatment prevented ballooning degeneration and lipid droplet accumulation, and UDPG treatment reduced triacylglycerol and cholesterol in the liver and serum (Fig. 6, F to H), concomitant with the downregulation of Acaca, Fasn, and Scd1 and the

g

Chen et al., Science 383, eadi3332 (2024)

promotes its degradation to inhibit SREBP cleavage in the Golgi apparatus.

p

on the Uniprot website (https://www.uniprot. org/uniprot/). As shown in Fig. 4K and fig. S4H, the binding affinities of UDPG with S1P from the 2D and 3D binding modes were estimated to be −7.77949 kcal/mol, and residues N328, N329, E435, N438, M522, E592, and T593 displayed potential interactions with UDPG (N, Asn; E, Glu; M, Met; T, Thr). By constructing the corresponding mutant vector [N328A (Asn328→Ala), N329A, E435A, N438A, M522A, E592A, or T593A] with an HA tag, we found that N438A and T593A mutations had no effect on S1P expression and folding but disrupted the binding of UDPG to S1P (Fig. 4L and fig. S4, I to L). Consistent results were obtained from the analysis of n-SREBP1c expression under mutated conditions (Fig. 4M and fig. S4M). Knock-in of mutated S1P (N438A) into Hep3B cells also resulted in the inhibition of UDPG binding to S1P and the cleavage of SREBP1 (fig. S4, N to P). In addition to SREBPs, S1P cleavage of other substrates, such as ATF6 and GlcNAc1-phosphoacyltransferase, was also inhibited by UDPG (fig. S4, Q and R). We also found that UDPG did not alter S1P mRNA but downregulated its protein expression in hepatocytes (Fig. 4N and fig. S4S), suggesting that UDPG is likely to affect the protein stability of S1P. In line with this, circular dichroism spectra revealed the conformational change of S1P upon UDPG binding (fig. S4T). Conformational change commonly induces protein degradation through the ubiquitin-proteosome pathway (36). Use of the proteasome inhibitor MG132, but not autophagy inhibitor bafilomycin A1 (BafA1), blocked UDPG-induced S1P degradation (Fig. 4O). Further use of the protein synthesis inhibitor cycloheximide (CHX) demonstrated that UDPG treatment promoted the degradation of S1P but not mutated S1P (N438A) (Fig. 4P). Consistent with this, UDPG treatment augmented ubiquitination of S1P but not N438A-mutated S1P (Fig. 4Q). Two lysine sites (K433 and K443) were located close to N438. Mutating them into K433R and K443R (R, Arg), respectively, did not affect the binding of UDPG to S1P (fig. S4U); however, K443R mutation inhibited UDPGmediated S1P ubiquitination and degradation (fig. S4, V and W). Together, these results established that UDPG interacts with S1P and

Western blot. (N) Hepatocytes were treated with UDPG for 24 hours, and then S1P was quantified by Western blot. (O) S1P degradation in hepatocytes with UDPG treatment was determined by Western blot. BAF, bafilomycin (a lysosome inhibitor); CHX, cycloheximide; MG132 (a proteasome inhibitor). (P) HEK293T cells were transfected with plasmids expressing S1PWT or S1PN438A (WT, wild type). Twelve hours later, cells were treated with CHX (20 mg/ml) and UDPG (200 mM) for 12 hours. S1P degradation was determined by Western blot. (Q) Coimmunoprecipitation and immunoassay for extracts of HEK293T cells transfected with HA-S1PWT or HA-S1PN438A along with UDPG (200 mM) and MG132 (20 mM) by anti-HA beads. IP, immunoprecipitate. Data are presented as means ± SEM. p values were calculated using one-way ANOVA [(C), (F), and (L)] or unpaired two-tailed Student’s t test [(B) and (D)].

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

p g y acid and oleic acid was analyzed by LC-MS/MS (G). (H and I) Hepatocytes were transfected with Slc35f5-overexpressing plasmid (Slc35f5-OE) for 24 hours. The localization of UDPG in hepatocytes was detected by immunofluorescence staining (scale bars are 20 mm) (H). UDPG levels in the Golgi were analyzed by LC-MS/MS (I). (J to L) Hepatocytes with or without Slc35f5 overexpression were treated with 200 mM UDPG for 24 hours, and SREBP1 was determined by Western blot (J). TG levels were measured by a colorimetric assay (K). Hepatocytes were treated with serum- and glucose-free medium for 2 hours, followed by 1 hour of 13C-glucose treatment. The incorporation of 13C into palmitic acid and oleic acid was analyzed by LC-MS/MS (L). Data are presented as means ± SEM. p values were calculated using one-way ANOVA [(C), (F), (G), (K), and (L)] or unpaired two-tailed Student’s t test (I).

y g

Fig. 5. SLC35F5 mediates the transport of UDPG into the Golgi apparatus. (A) The mRNA expression of Slc35 members in mouse hepatocytes was determined by real-time PCR (left). The top 10 most expressed genes were identified (right). (B) BODIPY fluorescence levels of siRNA-transfected hepatocytes were determined by flow cytometry. (C) Hepatocytes were transfected with Slc35f5 or Slc35d2 siRNA for 48 hours, and then UDPG levels in the Golgi were analyzed by LC-MS/MS. (D to G) Hepatocytes were transfected with Slc35f5 siRNA for 48 hours, and the localization of UDPG in hepatocytes was detected by immunofluorescence staining (scale bars are 20 mm) (D). SREBP1 in whole-cell lysates was determined by Western blot (E). TG levels were measured by a colorimetric assay (F). The hepatocytes were then treated with serum- and glucose-free medium for 2 hours, followed by 1 hour of 13C-glucose treatment. The incorporation of 13C into palmitic

y

Chen et al., Science 383, eadi3332 (2024)

Glycogenesis antagonizes lipogenesis via UDPG in patients with NAFLD

Finally, we sought to validate our findings in human hepatocytes and NAFLD patient samples. First, using human S1P, S2P, and SREBP1c, we confirmed the binding of UDPG to human S1P rather than human S2P or SREBP1c, as evidenced by LC-MS/MS analysis and BLI assay (Fig. 7, A and B). Correspondingly, mutation of the key residues (N438A or T593A) disrupted the binding of human S1P to UDPG (fig. S7A). By cotransfecting human S1P and SREBP1c vectors into HEK293T cells, we also verified that UDPG disrupted the cleavage of human SREBP1c by S1P (fig. S7B). Then, we isolated primary hepatocytes from the liver tissues of patients with hepatic hemangioma.

16 February 2024

We demonstrated that the glycogen content increased (Fig. 7C), but triglyceride and human n-SREBP1 levels decreased, upon UDPG treatment ex vivo (Fig. 7, D and E, and fig. S7C). In addition, using serum- and glucose-free medium, we also demonstrated that the primary human hepatocytes started to increase glycogen within 30 min, but triglycerides increased 2 hours later in response to glucose supplementation (Fig. 7F). This priority of glycogenesis was also confirmed by 13C tracing (Fig. 7, G and H). By collecting normal and NAFLD liver tissue (fig. S7D), we found that glycogen and UDPG levels were reduced in NAFLD tissues (n = 5) (Fig. 7, I and J), concomitant with the down-regulation of PGM1 and UGP2, compared with those in normal tissues (n = 5) 7 of 12

,

reduced n-SREBP1 (Fig. 6M and fig. S6E). In line with these results, UDPG treatment restored glycogenesis, as evidenced by the increased glycogen content and Pgm1 and Ugp2 expression (Fig. 6, I and K). In parallel, this UDPG treatment neither affected blood glucose levels, glucose tolerance, and ketone body levels (b-hydroxybutyrate) nor altered liver gluconeogenic enzymes and insulin signaling (fig. S6, F to J). Moreover, the administered UDPG could be transferred to the Golgi, where it bound S1P (Fig. 6, N and O). In addition, we found that SLC35F5 was up-regulated in the Golgi after UDPG treatment (Fig. 6P). Together, these results suggest that UDPG may be a potential agent for the treatment of hepatic steatosis.

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

p g y y g

Chen et al., Science 383, eadi3332 (2024)

acid–to–palmitic acid ratio) to induce steatosis in isolated primary hepatocytes (41, 42), followed by UDPG treatment. We found that UDPG treatment resulted in the increase of glycogen contents (fig. S7G), decrease of triglyceride

16 February 2024

and n-SREBP1 (fig. S7, H and I), and downregulation of the expression of ACACA, FASN, and SCD1 (fig. S7J). Additionally, UDPG treatment increased SLC35F5 mRNA expression in the cells (fig. S7K). Finally, we used oleic acid 8 of 12

,

(Fig. 7, K and L). ACACA, FASN, and SCD1 expression was up-regulated in NAFLD hepatocytes, concomitant with increased n-SREBP1 expression (fig. S7, E and F). To further verify these results, we used free fatty acids (2:1 oleic

remaining on the HFD. H&E and Oil Red O staining of liver tissues were performed (scale bars are 20 mm) (F). TG and cholesterol (TC) in liver (G) and serum (H) were measured by a colorimetric assay. Glycogen content of liver tissues was determined by PAS staining (scale bars are 20 mm) (I). UDPG levels were analyzed by LC-MS/MS (J). Pgm1 and Ugp2 expression was determined by real-time PCR (K). Body weights of mice were measured (L). SREBP1 in liver tissues was determined by Western blot (M). C57BL/6J mice were intraperitoneally injected with glucose-UDP-fluorescein conjugate three times, then UDPG location in isolated hepatocytes was determined by immunofluorescence staining (scale bars are 20 mm) (N). UDPG levels in the Golgi were analyzed by LC-MS/MS (O), and SLC35F5 expression was determined by Western blot (P). Data are presented as means ± SEM. p values were calculated using one-way ANOVA [(G), (H), (J) to (L), and (O)] or unpaired two-tailed Student’s t test [(B) and (D)].

y

Fig. 6. UDPG administration alleviates hepatic steatosis in mouse HFD model. (A) After 12 hours of fasting, C57BL/6J mice were treated with glucose [2 g/kg intraperitoneally (ip)] for the indicated time, and glycogen and TG levels in livers were measured by a colorimetric assay (n = 6 mice per group at each time point). (B) C57BL/6J mice were injected with UDPG (8 mg/kg ip) or PBS, and then the food was removed overnight. The mice were then treated with glucose (2 g/kg ip) for 1 hour, and TG levels in livers were analyzed by a colorimetric assay (n = 6 mice per group at each time point). (C and D) Schematic of experimental design (C). C57BL/6J mice were treated with UDPG (8 mg/kg ip) for 1 week. After 12 hours of fasting, the mice were treated with 13C-glucose (2 g/kg ip) for 1 hour, and M+4 labeled palmitic acid and oleic acid were analyzed by LC-MS/MS (D). ND, normal diet. (E to P) Schematic of experimental design (n = 6 mice per group) (E). After 8 weeks on a HFD, C57BL/6J mice were treated with UDPG (8 mg/kg) or vehicle for 8 weeks while

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

p g y y g

Chen et al., Science 383, eadi3332 (2024)

noids (Fig. 7, M and N, and fig. S7, L and M). Together, these results suggest that UDPG inhibits lipogenesis in human hepatocytes and patients with NAFLD.

16 February 2024

Discussion

In this work, we report the regulation of lipogenesis by glycogenesis in hepatocytes. After a meal is consumed, the flow of excess glucose 9 of 12

,

to induce steatosis in human hepatocyte organoids (43, 44). We found that, indeed, UDPG treatment increased glycogen content and decreased triglyceride content in the orga-

colorimetric assay (n = 3 biological replicates per group). (G and H) Primary human hepatocytes were treated with serum- and glucose-free medium for 2 hours, followed by 2 mg/ml 13C-glucose treatment. M+6 labeled glucose in glycogen (G) and M+4 labeled palmitic acid (H) were analyzed by LC-MS/MS. (I to L) Surgical liver tissues with or without NAFLD were collected, and glycogen levels were measured by a colorimetric assay (n = 5 biological replicates per group) (I), UDPG levels were measured by LC-MS/MS (J), and PGM1 and UGP2 expression was measured by immune histochemical staining (scale bars are 50 mm) [(K) and (L)]. (M and N) Human hepatocyte organoids were pretreated with 300 mM oleic acid (OA) for 24 hours followed by stimulation with UDPG for another 24 hours. Glycogen levels were evaluated by PAS staining (dark pink) (scale bars are 20 mm) (M). TG levels were evaluated by labeling lipid droplets with BODIPY 493/503 (scale bars are 20 mm) (N). BSA, bovine serum albumin. Data are presented as means ± SEM. p values were calculated using one-way ANOVA (A) or unpaired two-tailed Student’s t test [(C), (D), and (I) to (L)].

y

Fig. 7. Glycogenesis antagonizes lipogenesis by UDPG in patients with NAFLD. (A) HEK293T cells were transfected with pCMV-HA, pCMV-h-SREBP1cHA, pCMV-h-S1P-HA, or pCMV-h-S2P-HA plasmids for 48 hours. Cell lysates were immunoprecipitated with anti-HA agarose beads, followed by the incubation of beads with 200 mM UDPG for 3 hours at 4°C. The amounts of UDPG bound to the beads were detected by LC-MS/MS. Relative UDPG levels were calibrated to HA-tagged total protein concentration. (B) Binding (KD) between human S1P and UDPG was measured by BLI, and 400 mM UDPGA was used as a negative control. (C to E) Primary human hepatocytes isolated from normal liver regions of donors were treated with 200 mM UDPG for 48 hours (n = 2 or 3 biological replicates per group), then glycogen (C) and TG (D) levels were measured by a colorimetric assay, and SREBP1 was determined by digital Western blot (52) (E). (F) Primary human hepatocytes were cultured in serum- and glucose-free medium for 2 hours, followed by 2 mg/ml glucose treatment. Glycogen and TG levels of hepatocytes were measured by a

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

Materials and methods summary Animals and cell lines

The UDP-2-N3-Glc pulldown–MS experiment was performed by following a previous protocol (35). Briefly, hepatocytes were treated

16 February 2024

10 of 12

,

UDP-2-N3-Glc pulldown–MS and MS data analysis

y

Primary hepatocytes were isolated from liver tissues of patients with hepatic hemangioma by two steps of collagenase digestion, and cells were filtered through a 70-mm filter and collected by 5 min of centrifugation at 70g. Adult hepatocytes were mixed with human HepMedium and Matrigel (with a ratio of 1:3) and seeded in 24-well plates. After solidification, 500 ml of medium was added. The components of human Hep-medium are shown in table S2. During culturing, medium was refreshed every 2 to 3 days.

To assess hepatic de novo lipogenesis activity, mice were fasted for 12 hours and then were intraperitoneally administered 2 g/kg 13Cglucose (389374, Sigma-Aldrich) and euthanized. Their liver tissues were collected and frozen in liquid nitrogen. To assess the 13C-labeled fatty acid, 20 mg of frozen tissue sample was homogenized in 100 ml of water using a TissueLyser (JX-24) for 3 min at 40 Hz. Fatty acids were extracted from the homogenate using a mixture composed of 200 ml of methanol and 400 ml of chloroform and incubated for 30 min. After centrifugation at 14,000g for 15 min at 4°C, 300 ml of the chloroform layer was isolated and evaporated to dryness. The residues were mixed with 50 ml of extraction solvent (20% methanol/water). After centrifugation at 20,000g for 10 min at 4°C, supernatant extracts were analyzed by LC-MS/MS. To assess cells’ de novo lipogenesis activity, we treated hepatocytes with serum- and glucose-free medium for 2 hours, followed by the replenishment of 13Cglucose. The cells were collected and homogenized in water. Fatty acids were extracted from the homogenate using a mixture composed of methanol and chloroform (1:2) and incubated for 30 min. For quantitative analyses, a high-performance LC (HPLC)–MS/MS system (AB Sciex LLC, USA) was applied. The separation was performed on a BEH C18 column (1.8 mm, 2.1 mm by 100 mm, Waters Corp., USA). The mobile phase consisted of solvent A (water/acetonitrile, 10/90, containing 10 mM ammonium formate and 0.1% formic acid) and solvent B (water/acetonitrile, 2/98, containing 0.1% formic acid). Gradient elution was carried out for 8.0 min at a flow rate of 0.3 ml/min. Gradient conditions were as follows: 0 to 3.0 min, 0.1% B; 3.0 to 3.1 min, 0.1% B to 99% B; 3.1 to 5.0 min, 99% B; 5.0 to 5.1 min, 99% B to 0.1% B; 5.1 to 8.0 min, 0.1% B. A 5-ml aliquot of each sample was injected into the column. The column temperature was kept at 40°C. All samples were kept at 4°C throughout the analysis. MS was performed on a Sciex 6500+ triple quadrupole mass spectrometer equipped with a Turbo V ion source. All compounds were detected in negative electrospray

y g

Organoid culture of human hepatocytes

Metabolite analysis

y

Five- to 8-week-old male C57BL/6J mice were purchased from the Center of Medical Experimental Animals of the Chinese Academy of Medical Science (Beijing, China). These animals were maintained in the Animal Facilities of the Chinese Academy of Medical Science under pathogen-free conditions. All studies involving mice were approved by the Animal Care and Use Committee of the Chinese Academy of Medical Science (ACUC-A02-2021-042). Mice were fed either a HFD (#D12492, Research Diets) or a no-fat diet (#D12451, Research Diets) for a total of 16 weeks, starting at 8 weeks of age. Nutrient composition of the diets is shown in table S1. HEK293T and Hep3B cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. The cell lines were assessed for mycoplasma contamination, and the test results were negative.

with 200 mM UDP-2-N3-Glc for 12 hours (for control, cells were treated with 200 mM UDPG). Cells were washed with phosphate-buffered saline (PBS) twice, cross-linked by ultraviolet light– induced cross-linking (UVC) (254 nm, Scientz03II) at 0.3 J/cm2, and then lysed in urea lysis buffer (200 mM Tris pH8, 4% CHAPS, 1 M NaCl, 8 M Urea). The Click-&-Go Click Chemistry Capture Kit (Click Chemistry Tools, #1065) was used to isolate UDP-2-N3-Glc–linked proteins according to the manufacturer’s instructions. Samples were digested on-bead per the kit protocol and sent for LC-MS/MS analysis on an Orbitrap Fusion mass spectrometer.

g

Chen et al., Science 383, eadi3332 (2024)

Given that increases in de novo lipogenesis are major contributors to NAFLD, UDPG might be a promising candidate for NAFLD treatment, offering several advantages: (i) Given its extracellular presence in peripheral blood, UDPG might be suitable for intravenous administration (49); (ii) the administration of UDPG is potentially patient friendly, considering its physiological presence in the blood (10, 50); (iii) the uptake of UDPG by hepatocytes, as evidenced in this study, demonstrates its feasibility in regulating hepatic lipogenesis; and (iv) UDPG may facilitate glycogen metabolism, which favors ROS clearance by the regulation of the pentose phosphate pathway (10–12). Our findings are consistent with the notion of increased lipogenesis as a distinct characteristic of NAFLD (51). The identified UDPG-mediated regulation of lipogenesis may open therapeutic avenues for the management of abnormal lipid metabolism in humans.

p

can be directed to glycogenesis or lipogenesis in the liver. However, the intermediate metabolite UDPG from glycogenesis may be shunted via the transporter SLC35F5 to the Golgi apparatus, where UDPG binds to S1P and promotes its ubiquitination for degradation, thereby disrupting S1P-mediated SREBP cleavage and subsequent lipogenesis. Glycogens have long been thought to only store and supply energy in the liver and muscles (1, 3). Our previous studies demonstrated that glycogen synthesis and degradation regulate cellular redox homeostasis by activating the pentose phosphate pathway (10–12, 15). However, abnormal glycogen metabolism may lead to tumorigenesis (45, 46). In addition, we demonstrated that proinflammatory M1 macrophages use glycogen synthesis to activate P2Y14 signaling for STAT1 expression and phosphorylation (10). Therefore, the physiological function of glycogen is not limited to energy storage, and its pathophysiological role is worthy of comprehensive investigation. In the present study, we further highlight this concept by providing evidence that the glycogenic metabolite UDPG regulates lipogenesis. We found that the UDPG levels increase at the beginning but decrease later in conjunction with the rise and fall of glycogenesis and lipogenesis. This UDPG-mediated regulation appears to be very specific, because UDP-glucuronate does not inhibit lipogenesis. In addition, although insulin signaling stimulates glycogenesis and lipogenesis, we found that the effect of UDPG goes beyond insulin signaling. However, to achieve the regulation, UDPG needs to gain access to the Golgi apparatus, where it is transported into the Golgi lumen. The solute carrier family of nucleotide sugar transporters is designated SLC35, and it facilitates the transport of a cytosolic nucleotide sugar that crosses the membrane into the ER or Golgi lumen (47, 48). In this work, we demonstrate that SLC35F5 mediates the entry of UDPG into the Golgi lumen, where UDPG binds to S1P. S1P is a membrane-bound serine protease. By sequence analysis, we found that S1P is highly conserved among human and other species (fig. S8). We also observed that the purported UDPG binding sites on the S1P sequence are highly conserved (fig. S8), suggesting that the inhibitory effect of UDPG on lipogenesis may be applicable across species. As a key player, S1P controls lipogenesis in hepatocytes and other cells by cleaving SREBP and initiating the release of the N-terminal fragment with transcriptional activity for genes that encode enzymes involved in cholesterol and fatty acid biosynthesis, as well as the lowdensity lipoprotein receptor. Using LC-MS/ MS and BLI, we verified that UDPG binds to S1P, thus inhibiting S1P-mediated cleavage of SREBP.

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

ion mode. Curtain gas, collision gas, nebulizer gas (GS1), and turbo-gas (GS2) were set at 35, 9, 60, and 60 psi, respectively. The electrospray voltage was −4500 V, and the turbo-ion spray source temperature was 500°C. Nitrogen was used as the collision gas. The system was operated in multiple reaction monitoring (MRM) mode. The quantitative transitions, and their dwell time, co-recollision energy, declustering potential, and rentiom time are shown in table S3. Data acquisitions were performed using Analyst 1.7.1 software (Applied Biosystems, USA). Multiquant 3.0.3 software (Applied Biosystems, USA) was used to quantify all analytes. Statistical analysis

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

29.

32.

33.

34.

16 February 2024

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

AC KNOWLED GME NTS

We thank W. M. Tong (Chinese Academy of Medical Sciences, China) for help with evaluation of the H&E analysis of cells and tissues from NAFLD patients. Funding: This work was supported by the National Key Research and Development Program of China (2022YFA1206000), the National Natural Science Foundation of China (82388201), the Chinese Academy of Medical Sciences (CAMS) Innovation Fund for Medical Sciences (CIFMS) (2021-I2M1-021), and the Haihe Laboratory of Cell Ecosystem Innovation Fund (22HHXBSS00009). Author contributions: B.H. conceived the project. J.C., Y.Z., Z.L., Y.L., Y.J., K.C., Na.Z., D.W., C.Z., K.S., Ni.Z.,

11 of 12

,

31.

39.

y

30.

38.

y g

28.

37.

y

1. P. J. Roach, A. A. Depaoli-Roach, T. D. Hurley, V. S. Tagliabracci, Glycogen and its metabolism: Some new developments and old themes. Biochem. J. 441, 763–787 (2012). doi: 10.1042/ BJ20111416; pmid: 22248338 2. W. Shao, P. J. Espenshade, Sugar makes fat by talking to SCAP. Cancer Cell 28, 548–549 (2015). doi: 10.1016/ j.ccell.2015.10.011; pmid: 26555169 3. H. G. Hers, The control of glycogen metabolism in the liver. Annu. Rev. Biochem. 45, 167–190 (1976). doi: 10.1146/ annurev.bi.45.070176.001123; pmid: 183599 4. A. R. Saltiel, C. R. Kahn, Insulin signalling and the regulation of glucose and lipid metabolism. Nature 414, 799–806 (2001). doi: 10.1038/414799a; pmid: 11742412 5. P. Sonksen, J. Sonksen, Insulin: Understanding its action in health and disease. Br. J. Anaesth. 85, 69–79 (2000). doi: 10.1093/bja/85.1.69; pmid: 10927996 6. P. V. Röder, B. Wu, Y. Liu, W. Han, Pancreatic regulation of glucose homeostasis. Exp. Mol. Med. 48, e219 (2016). doi: 10.1038/emm.2016.6; pmid: 26964835 7. A. Blanco, G. Blanco, Medical Biochemistry (Academic Press, 2017), pp. 283–323. 8. H. Zhang, J. Ma, K. Tang, B. Huang, Beyond energy storage: Roles of glycogen metabolism in health and disease. FEBS J. 288, 3772–3783 (2021). doi: 10.1111/febs.15648; pmid: 33249748 9. J. A. Menendez, R. Lupu, Fatty acid synthase and the lipogenic phenotype in cancer pathogenesis. Nat. Rev. Cancer 7, 763–777 (2007). doi: 10.1038/nrc2222; pmid: 17882277 10. J. Ma et al., Glycogen metabolism regulates macrophagemediated acute inflammatory responses. Nat. Commun. 11, 1769 (2020). doi: 10.1038/s41467-020-15636-8; pmid: 32286295 11. R. Ma et al., A Pck1-directed glycogen metabolic program regulates formation and maintenance of memory CD8+ T cells. Nat. Cell Biol. 20, 21–27 (2018). doi: 10.1038/s41556-0170002-2; pmid: 29230018 12. K. Tang et al., Hypoxia promotes breast cancer cell growth by activating a glycogen metabolic program. Cancer Res. 81, 4949–4963 (2021). doi: 10.1158/0008-5472.CAN-21-0753; pmid: 34348966 13. M. Gao et al., Role of mitochondria in ferroptosis. Mol. Cell 73, 354–363.e3 (2019). doi: 10.1016/j.molcel.2018.10.042; pmid: 30581146 14. T. M. McIntyre, S. L. Hazen, Lipid oxidation and cardiovascular disease: Introduction to a review series. Circ. Res. 107, 1167–1169 (2010). doi: 10.1161/CIRCRESAHA.110.224618; pmid: 21071714 15. H. Zhang et al., Ketogenesis-generated b-hydroxybutyrate is an epigenetic regulator of CD8+ T-cell memory development.

36.

g

RE FE RENCES AND N OT ES

35.

in mice and humans. J. Clin. Invest. 122, 2176–2194 (2012). doi: 10.1172/JCI41636; pmid: 22546860 W. Miao et al., Glucose dissociates DDX21 dimers to regulate mRNA splicing and tissue differentiation. Cell 186, 80–97.e26 (2023). doi: 10.1016/j.cell.2022.12.004; pmid: 36608661 H. Dou et al., Structural basis for autoinhibition and phosphorylation-dependent activation of c-Cbl. Nat. Struct. Mol. Biol. 19, 184–192 (2012). doi: 10.1038/nsmb.2231; pmid: 22266821 Z. Song, Roles of the nucleotide sugar transporters (SLC35 family) in health and disease. Mol. Aspects Med. 34, 590–600 (2013). doi: 10.1016/j.mam.2012.12.004; pmid: 23506892 C. B. Hirschberg, P. W. Robbins, C. Abeijon, Transporters of nucleotide sugars, ATP, and nucleotide sulfate in the endoplasmic reticulum and Golgi apparatus. Annu. Rev. Biochem. 67, 49–69 (1998). doi: 10.1146/annurev. biochem.67.1.49; pmid: 9759482 J. L. Parker, R. A. Corey, P. J. Stansfeld, S. Newstead, Structural basis for substrate specificity and regulation of nucleotide sugar transporters in the lipid bilayer. Nat. Commun. 10, 4657 (2019). doi: 10.1038/s41467-019-12673-w; pmid: 31604945 L. Faletti et al., TNFa sensitizes hepatocytes to FasL-induced apoptosis by NFkB-mediated Fas upregulation. Cell Death Dis. 9, 909 (2018). doi: 10.1038/s41419-018-0935-9; pmid: 30185788 Z. Peng et al., METTL3-m6A-Rubicon axis inhibits autophagy in nonalcoholic fatty liver disease. Mol. Ther. 30, 932–946 (2022). doi: 10.1016/j.ymthe.2021.09.016; pmid: 34547464 J. Ye et al., Fucoxanthin attenuates free fatty acid-induced nonalcoholic fatty liver disease by regulating lipid metabolism/ oxidative stress/inflammation via the AMPK/Nrf2/TLR4 signaling pathway. Mar. Drugs 20, 225 (2022). doi: 10.3390/ md20040225; pmid: 35447899 M. Kimura et al., En masse organoid phenotyping informs metabolic-associated genetic susceptibility to NASH. Cell 185, 4216–4232.e16 (2022). doi: 10.1016/j.cell.2022.09.031; pmid: 36240780 H. Hu et al., Long-term expansion of functional mouse and human hepatocytes as 3D organoids. Cell 175, 1591–1606.e19 (2018). doi: 10.1016/j.cell.2018.11.013; pmid: 30500538 C. E. Zois, E. Favaro, A. L. Harris, Glycogen metabolism in cancer. Biochem. Pharmacol. 92, 3–11 (2014). doi: 10.1016/ j.bcp.2014.09.001; pmid: 25219323 Q. Liu et al., Glycogen accumulation and phase separation drives liver tumor initiation. Cell 184, 5559–5576.e19 (2021). doi: 10.1016/j.cell.2021.10.001; pmid: 34678143 B. Hadley et al., Nucleotide sugar transporter SLC35 family structure and function. Comput. Struct. Biotechnol. J. 17, 1123–1134 (2019). doi: 10.1016/j.csbj.2019.08.002; pmid: 31462968 N. Ishida, M. Kawakita, Molecular physiology and pathology of the nucleotide sugar transporter family (SLC35). Pflugers Arch. 447, 768–775 (2004). doi: 10.1007/s00424-003-1093-0; pmid: 12759756 S. Kook, J. Cho, S. B. Lee, B.-C. Lee, The nucleotide sugar UDP-glucose mobilizes long-term repopulating primitive hematopoietic cells. J. Clin. Invest. 123, 3420–3435 (2013). doi: 10.1172/JCI64060; pmid: 23863713 E. R. Lazarowski, T. K. Harden, UDP-sugars as extracellular signaling molecules: Cellular and physiologic consequences of P2Y14 receptor activation. Mol. Pharmacol. 88, 151–160 (2015). doi: 10.1124/mol.115.098756; pmid: 25829059 J. E. Lambert, M. A. Ramos-Roman, J. D. Browning, E. J. Parks, Increased de novo lipogenesis is a distinct characteristic of individuals with nonalcoholic fatty liver disease. Gastroenterology 146, 726–735 (2014). doi: 10.1053/ j.gastro.2013.11.049; pmid: 24316260 L. Zhou et al., Tumor cell-released kynurenine biases MEP differentiation into megakaryocytes in individuals with cancer by activating AhR-RUNX1. Nat. Immunol. 24, 2042–2052 (2023). doi: 10.1038/s41590-023-01662-3; pmid: 37919525

p

All experiments were performed with at least three biological replicates. Results are expressed as mean values ± SEM as indicated, and data were analyzed by two-tailed Student’s t test or one-way analysis of variance (ANOVA). The p < 0.05 was considered statistically significant. The analysis was conducted using Graphpad 8.0 software.

Chen et al., Science 383, eadi3332 (2024)

16.

Nat. Cell Biol. 22, 18–25 (2020). doi: 10.1038/ s41556-019-0440-0; pmid: 31871320 F. J. Roca, L. J. Whitworth, H. A. Prag, M. P. Murphy, L. Ramakrishnan, Tumor necrosis factor induces pathogenic mitochondrial ROS in tuberculosis through reverse electron transport. Science 376, eabh2841 (2022). doi: 10.1126/science. abh2841; pmid: 35737799 S. Castelli, P. De Falco, F. Ciccarone, E. Desideri, M. R. Ciriolo, Lipid catabolism and ROS in cancer: A bidirectional liaison. Cancers 13, 5484 (2021). doi: 10.3390/cancers13215484; pmid: 34771647 M. G. Rosca et al., Oxidation of fatty acids is the source of increased mitochondrial reactive oxygen species production in kidney cortical tubules in early diabetes. Diabetes 61, 2074–2083 (2012). doi: 10.2337/db11-1437; pmid: 22586586 E. Favaro et al., Glucose utilization via glycogen phosphorylase sustains proliferation and prevents premature senescence in cancer cells. Cell Metab. 16, 751–764 (2012). doi: 10.1016/ j.cmet.2012.10.017; pmid: 23177934 Q. Zhao et al., Targeting mitochondria-located circRNA SCAR alleviates NASH via reducing mROS output. Cell 183, 76–93.e22 (2020). doi: 10.1016/j.cell.2020.08.009; pmid: 32931733 A. Zhuge et al., Microbiota-induced lipid peroxidation impairs obeticholic acid-mediated antifibrotic effect towards nonalcoholic steatohepatitis in mice. Redox Biol. 59, 102582 (2023). doi: 10.1016/j.redox.2022.102582; pmid: 36584600 G. Serviddio et al., Uncoupling protein-2 (UCP2) induces mitochondrial proton leak and increases susceptibility of non-alcoholic steatohepatitis (NASH) liver to ischaemiareperfusion injury. Gut 57, 957–965 (2008). doi: 10.1136/ gut.2007.147496; pmid: 18308829 M. S. Brown, J. L. Goldstein, A proteolytic pathway that controls the cholesterol content of membranes, cells, and blood. Proc. Natl. Acad. Sci. U.S.A. 96, 11041–11048 (1999). doi: 10.1073/pnas.96.20.11041; pmid: 10500120 T. Yang et al., Crucial step in cholesterol homeostasis: Sterols promote binding of SCAP to INSIG-1, a membrane protein that facilitates retention of SREBPs in ER. Cell 110, 489–500 (2002). doi: 10.1016/S0092-8674(02)00872-3; pmid: 12202038 X. Wang et al., UDP-glucose accelerates SNAI1 mRNA decay and impairs lung cancer metastasis. Nature 571, 127–131 (2019). doi: 10.1038/s41586-019-1340-y; pmid: 31243371 J. Meister et al., The G protein-coupled receptor P2Y14 influences insulin release and smooth muscle function in mice. J. Biol. Chem. 289, 23353–23366 (2014). doi: 10.1074/ jbc.M114.580803; pmid: 24993824 T. A. Jokela et al., Extracellular UDP-glucose activates P2Y14 Receptor and Induces Signal Transducer and Activator of Transcription 3 (STAT3) Tyr705 phosphorylation and binding to hyaluronan synthase 2 (HAS2) promoter, stimulating hyaluronan synthesis of keratinocytes. J. Biol. Chem. 289, 18569–18581 (2014). doi: 10.1074/jbc.M114.551804; pmid: 24847057 G. I. Smith et al., Insulin resistance drives hepatic de novo lipogenesis in nonalcoholic fatty liver disease. J. Clin. Invest. 130, 1453–1460 (2020). doi: 10.1172/JCI134165; pmid: 31805015 H. Shimano, R. Sato, SREBP-regulated lipid metabolism: Convergent physiology – divergent pathophysiology. Nat. Rev. Endocrinol. 13, 710–730 (2017). doi: 10.1038/nrendo.2017.91; pmid: 28849786 I. Shimomura, H. Shimano, J. D. Horton, J. L. Goldstein, M. S. Brown, Differential expression of exons 1a and 1c in mRNAs for sterol regulatory element binding protein-1 in human and mouse organs and cultured cells. J. Clin. Invest. 99, 838–845 (1997). doi: 10.1172/JCI119247; pmid: 9062340 L.-P. Sun, J. Seemann, J. L. Goldstein, M. S. Brown, Sterol-regulated transport of SREBPs from endoplasmic reticulum to Golgi: Insig renders sorting signal in Scap inaccessible to COPII proteins. Proc. Natl. Acad. Sci. U.S.A. 104, 6519–6526 (2007). doi: 10.1073/pnas.0700907104; pmid: 17428919 J. Luo, H. Yang, B.-L. Song, Mechanisms and regulation of cholesterol homeostasis. Nat. Rev. Mol. Cell Biol. 21, 225–245 (2020). doi: 10.1038/s41580-019-0190-7; pmid: 31848472 M. S. Brown, A. Radhakrishnan, J. L. Goldstein, Retrospective on cholesterol homeostasis: The central role of Scap. Annu. Rev. Biochem. 87, 783–807 (2018). doi: 10.1146/ annurev-biochem-062917-011852; pmid: 28841344 F. Benhamed et al., The lipogenic transcription factor ChREBP dissociates hepatic steatosis from insulin resistance

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

Lu.Z., Li.Z., and J.L. performed the experiments. J.C., Y.Z., Z.W., and K.T. developed methodology. J.C., H.Z., J.M., and B.H. performed data analysis. B.H., Y.Z., and J.C. wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Materials and reagents used in this study are listed in table S4. Materials and reagents described in this study are either commercially available or available on request from the corresponding author. 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 SUPPLEMENTARY MATERIALS

science.org/doi/10.1126/science.adi3332 Materials and Methods

Figs. S1 to S8 Tables S1 to S7 References (53, 54) MDAR Reproducibility Checklist Submitted 19 April 2023; resubmitted 27 October 2023 Accepted 20 December 2023 10.1126/science.adi3332

p g y y g y ,

Chen et al., Science 383, eadi3332 (2024)

16 February 2024

12 of 12

RES EARCH

RESEARCH ARTICLE SUMMARY



PLANT SCIENCE

Paired plant immune CHS3-CSA1 receptor alleles form distinct hetero-oligomeric complexes Yu Yang, Oliver J. Furzer, Eleanor P. Fensterle, Shu Lin, Zhiyu Zheng, Nak Hyun Kim, Li Wan, Jeffery L. Dangl*

NLR oligomers, called resistosomes, mediates NLR function. However, oligomerization status and resistosome formation of paired NLRs are not currently understood.

Clade 2 Per-0

BAK1/BIRs LRR

CHS3

TIR NB-ARC

LRR

Cell death

Heterodimer ?

AE

?

BB-loop, two CSA1 NADase active sites opening



Clade 3 Ws-2

The activation and distinct hetero-oligomer formation of paired NLRs. The TNL CHS3 and CSA1 alleles form a resting state heterodimer, and this inactive state is maintained by the negative regulators BAK1 and BIRs. Upon modulation of BAK1 and/or BIRs activity, the heterodimers of the CHS3-CSA1 pair dimerize and likely form a distinct dimer of heterodimers oligomer. The interaction of TIR domains induces opening of two CSA1 NADase active sites, which are essential for CHS3-CSA1 pair–mediated cell death. Yang et al., Science 383, 719 (2024)

16 February 2024

The list of author affiliations is available in the full article online. *Corresponding author. Email: [email protected] Cite this article as Y. Yang et al., Science 383, eadk3468 (2024). DOI: 10.1126/science.adk3468

READ THE FULL ARTICLE AT https://doi.org/10.1126/science.adk3468 1 of 1

,

TIR NB-ARC

y

CSA1

y g

BB-loop, two CSA1 NADase active sites opening

AE

CONCLUSION: A stepwise mechanism of paired TNL CHS3-CSA1 alleles immune activation is proposed on the basis of our findings and on the basis of previous biochemical studies. In the resting state, the majority of CHS3 and CSA1 from clades 2 or 3 form heterodimers and cannot further oligomerization because of the negative regulation by BAK1 and BIR proteins. Upon modulation of the BAK1 and/or BIRs by pathogens, host damage–associated patterns, or other mechanisms, suppression is relieved and the heterodimers of the CHS3CSA1 pair further dimerize and likely form a distinct “dimer of heterodimers” oligomer. The interaction of CHS3 and CSA1 TIR domains leads to opening of the two CSA1 nicotinamide adenine dinucleotide hydrolase (NADase) active sites to initiate immune response and cell death. Our work uncovers the diverse oligomerization requirements of different CHS3-CSA1 alleles and suggests that structural heterogeneity can be expected even among alleles of closely related paired NLRs.

y

RESULTS: We found that the sensor CHS3 and executor CSA1 from clade 2 or clade 3 form the same size oligomers only in combination

g

RATIONALE: NLRs are the most variable gene families in plants. Our recent studies demonstrated that allelic Arabidopsis TNL CHS3-CSA1 pairs can be divided into three phylogenetic clades. CHS3 is the sensor, and CSA1 is the executor. The CHS3-CSA1 TNL alleles evolved two different types of pairs and two separable regulatory modes for their activation across these three clades. We deployed two functionally distinct CHS3-CSA1 allelic pairs in a model to study the diverse oligomerization requirements and specific mechanism of activation of this paired TNL family.

p

INTRODUCTION: Plants rely on an innate immune system to defend themselves against pathogen invasion. Nucleotide-binding (NB) leucine-rich repeat (LRR) receptors (NLRs) are important components of intracellular innate immunity across all kingdoms. Plant NLRs are classified into two major groups, CNLs and TNLs, based on the presence of different N-terminal domains: Most NLRs express either a coiled-coil (CC) domain (CNLs) or a Toll-like interleukin-1 receptor (TIR) domain (TNLs). Some NLRs function as singletons, with one NLR protein mediating both pathogen recognition and subsequent downstream immune activation. However, some NLRs are encoded in a head-to-head orientation on the chromosome and function together—these are called paired NLRs. One member of each paired NLR is a so-called sensor that recognizes pathogen proteins or their activities and then activates an associated executor NLR to initiate immune signaling. Ligand-induced formation of various types of

with their genetically encoded partners. Selfassociation of either CHS3 or CSA1 is weak in the absence of its partner, but heterocomplexes of CHS3 and CSA1 are readily detected when they are coexpressed. CHS3 and CSA1 from clade 2 or clade 3 oligomerize and form hetero-oligomers but not separate homooligomers. Mutations in the AE interface and the BB-loop of both CHS3 and CSA1 TIR domains suggested that the AE interface in the TIR domain of both CHS3 and CSA1 mediates essential head-to-head interaction for heterodimer formation and that the BB-loop in the TIR domains mediates essential head-to-tail interaction and dimerization of the heterodimers, eventually forming a heterotetramer. Mutational analysis of CHS3-CSA1 pairs from both clades also defined clade-specific TIR domain requirements for function and oligomerization, suggesting that TNL CHS3-CSA1 pairs from different clades form a distinct hetero-oligomeric complex. Coexpression of the previously defined negative regulators BAK1 or BIRs with the CHS3-CSA1 pair inhibited oligomerization but not heterodimer formation. These results suggest that paired NLRs exist at rest as heterodimers but form distinct hetero-oligomeric complexes to function and that the inactive state of paired NLRs can be maintained through the suppression of oligomerization by the negative regulators.

RES EARCH

RESEARCH ARTICLE



PLANT SCIENCE

Paired plant immune CHS3-CSA1 receptor alleles form distinct hetero-oligomeric complexes Yu Yang1,2, Oliver J. Furzer1,2, Eleanor P. Fensterle1, Shu Lin3, Zhiyu Zheng1, Nak Hyun Kim1,2, Li Wan3, Jeffery L. Dangl1,2*

16 February 2024

1 of 11

,

Yang et al., Science 383, eadk3468 (2024)

We used blue native–polyacrylamide gel electrophoresis (BN-PAGE) to test oligomerization of CHS3-CSA1 TNL pairs from each of the three clades. We detected oligomerization of clade 2 and clade 3 CHS3-CSA1 pairs, as detailed below; therefore, we focused on them. Negative results for the clade 1 pair are not shown. Coexpression of paired CHS3 and CSA1 from clade 2 or clade 3 was sufficient to trigger strong hypersensitive-like cell death in Nicotiana tabacum, but neither CHS3 nor CSA1 expression alone was sufficient (43). This cell death phenotype was dependent on the conserved glutamic acid (E) in the CSA1 TIR domain and intact P-loops of both CHS3 and CSA1 (43). The conserved glutamic acid (E) in the CSA1 TIR domain is indispensable for NADase catalytic activity (43), and the P-loop typically conserved in NLR proteins is necessary for adenosine 5′-triphosphate (ATP) binding and NLR oligomerization (22–26). We recapitulated these findings using CHS3-CSA1 pairs from the clade 2 accession Per-0 and the clade 3 accession Ws-2 (fig. S1, A to C). We then used wild-type CHS3-CSA1 pairs from clade 2 Per-0 or clade 3 Ws-2 and mutants derived from them to assay for oligomerization using BN-PAGE. We observed that oligomerization into a high–molecular weight band at apparent molecular mass of ~720 kDa was only detected when CHS3 and CSA1 were

y

*Corresponding author. Email: [email protected]

CHS3-CSA1 TNL pairs oligomerize

y g

Department of Biology, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. 2Howard Hughes Medical Institute, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. 3National Key Laboratory of Plant Molecular Genetics, CAS Center for Excellence in Molecular Plant Sciences, Institute of Plant Physiology and Ecology, Chinese Academy of Sciences, Shanghai, China.

y

1

cation channel to mediate Ca2+ influx (18–20). By contrast, many TNLs also function as sensor NLRs but require unlinked, ancient, and conserved helper RNLs to activate immune responses. The sensor TNLs RPP1 (RECOGNITION OF PERONOSPORA PARASITICA 1) and ROQ1 (RECOGNITION OF XOPQ 1) oligomerize to form tetrameric resistosomes upon activation (21, 22). TNL resistosomes are nicotinamide adenine dinucleotide hydrolases (NADases) that generate nucleotide-derived small signaling molecules (21–23). These induce the heterodimerization of either of two EDS1 heterodimers that in turn are required for the oligomerization and activation of downstream RNLs (24–29). RNLs, like CNLs, function as cation channels to mediate Ca2+ influx, defense, and cell death (24–29). Thus, both autonomous singleton CNLs and TNLs acting through RNLs end up signaling through resistosome-mediated calcium influx. Similarly, activation of a large class of sensor NLRs from a variety of solanaceous species enables the oligomerization of a limited array of differentially redundant helper NLRs called NRCs (17, 30, 31). NRC (NLR-REQUIRED for CELL DEATH) oligomers do not obviously contain sensor NLRs, which leads to an activationrelease model for NLRs in the NRC immune receptor network (30–32). In contrast to singletons and sensor-helper networks, paired NLR genes are encoded in head-to-head orientation, adjacent to one another. Such pairs make up ~10% of the NLR repertoire in Arabidopsis genomes (33). Each pair consists of a specialized “sensor” for effector perception and an “executor” for immune activation (34–44). To date, postactivation oligomerization and requirements for the oligomerization of paired NLRs are still elusive.

g

P

lant intracellular nucleotide-binding leucine-rich repeat receptors (NLRs) play essential roles in innate immune systems. Some NLRs directly bind virulence factor effectors that are secreted and delivered into the plant cell by pathogens. Other NLRs indirectly detect host proteins modified by effectors (known as guardees or decoys of guardees). Either mode of NLR activation results in effector-trigged immunity (ETI) (1–6), usually accompanied by rapid calcium influx, a respiratory oxidative burst, transcriptional reprogramming, and cell death at the infection site that is referred to as the hypersensitive response (HR) (1–6). NLRs are broadly classified into three subgroups on the basis of their Nterminal domains. Their N termini typically contain Toll-like interleukin-1 receptor (TIR) domains, coiled-coil (CC) domains, or RPW8 (RESISTANCE TO POWDERY MILDEW 8)–like CC domains. NLRs carrying these domains are named TNLs (TIR-NLRs), CNLs (CC-NLRs), or RNLs (CCR-NLRs) (7–13), respectively. NLRs can function as singletons, in genetically linked sensor-executor pairs (also known as paired NLRs or NLR pairs), or in genetically unlinked sensor-helper networks (14–17). Singleton NLRs include the Arabidopsis CNL ZAR1 (HOPZ-ACTIVATED RESISTANCE 1), which forms an oligomeric pentamer resistosome upon effector recognition. This resistosome relocates to the plasma membrane, where it acts as a

p

Plant intracellular nucleotide-binding leucine-rich repeat receptors (NLRs) analyzed to date oligomerize and form resistosomes upon activation to initiate immune responses. Some NLRs are encoded in tightly linked co-regulated head-to-head genes whose products function together as pairs. We uncover the oligomerization requirements for different Arabidopsis paired CHS3-CSA1 alleles. These pairs form resting-state heterodimers that oligomerize into complexes distinct from NLRs analyzed previously. Oligomerization requires both conserved and allele-specific features of the respective CHS3 and CSA1 Toll-like interleukin-1 receptor (TIR) domains. The receptor kinases BAK1 and BIRs inhibit CHS3-CSA1 pair oligomerization to maintain the CHS3-CSA1 heterodimer in an inactive state. Our study reveals that paired NLRs hetero-oligomerize and likely form a distinctive “dimer of heterodimers” and that structural heterogeneity is expected even among alleles of closely related paired NLRs.

Arabidopsis allelic TNL CHS3-CSA1 (CHILLING SENSITIVE 3–CONSTITUTIVE SHADEAVOIDANCE 1) pairs have been divided into three phylogenetic clades (33, 43) (fig. S1A). CHS3 is the proposed sensor NLR, and CSA1 is the proposed executor NLR in this pair (33, 43). The clade 1 sensor CHS3 proteins contain a putative effector-binding integrated decoy domain (ID) (44), which is lacking in clades 2 and 3 (33, 43) (fig. S1A). CHS3-CSA1 pairs in these three clades evolved two separable regulatory modes. One is mediated by the ID domain on clade 1 CHS3 sensors, whereas the other relies on CHS3-CSA1 pairs from all clades detecting the presence or perturbation of receptor-like kinase (RLK) proteins BAK1 (BRI1-ASSOCIATED RECEPTOR KINASE) and BIRs (BAK1-INTERACTING RECEPTOR-LIKE KINASEs) (43). In this work, we analyze the activation and oligomerization of different CHS3-CSA1 pairs from clades 2 and 3. We find that these TIR domain (TNL) pairs form heterodimers that oligomerize into distinct hetero-oligomeric complexes with common and clade-specific structural requirements. Moreover, BAK1 and BIRs act as negative regulators that inhibit the oligomerization—but not heterodimer formation—of the CHS3-CSA1 pair to maintain their inactive state. Kinase activity is not required for the negative regulatory function of these RLKs.

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

Per-0 + -

+ - + - + + - -

- + - + + - +

+

-

-

-

kDa

- + - + + - +

+

+

-

-

-

-

C

- 1236 - 1048

- 720 - 480 - 242

- 720 - 480 - 242

PS

PS

Ws-2 + -

+ - + - + +

- + - + + -

-

-

-

+

+

+

-

-

-

-

+ -

+ - + - + + - -

- + - + + - +

+

+

-

-

-

- 720 - 480 - 242

- 720 - 480 - 242

- 130

-HA

-V5

PS

Fig. 1. CSA1 and CHS3 oligomerize in a partner- and P-loop–dependent manner. (A and B) Both CSA1 and CHS3 from clade 2 Per-0 (blue) (A) or clade 3 Ws-2 (gray) (B) are capable of oligomerization dependent on their respective paired partner. Note that the clade 2 and clade 3 color schemes are maintained throughout. 16 February 2024

- 130 PS

- 130

-HA

PS

- 130

-V5

Ws-2 CSA1-HA CHS3-HA CSA1-V5 CHS3-V5 EV

+ +

+ +

+ + -

+ + -

kDa

- 1236 - 1048 -HA

- 720 - 480 - 242

-HA

- 130

-V5

- 130

,

- 1236 - 1048

-V5

kDa

- 1236 - 1048 - 720 - 480 - 242

-HA

kDa

- 1236 - 1048

-HA

Yang et al., Science 383, eadk3468 (2024)

-

+ + -

y

BN-PAGE

kDa

+ -

+ + -

y g

+ CSA1_E88A-HA CSA1_p-loop-HA CHS3-V5 CHS3_p-loop-V5 -

+ +

PS

D

CSA1-HA

SDS-PAGE

SDS-PAGE

- 130

-V5

BN-PAGE

- 130

+ +

kDa

-V5

-HA

EV

Per-0 CSA1-HA CHS3-HA CSA1-V5 CHS3-V5 EV

y

SDS-PAGE

+ - + - + + - -

- 1236 - 1048

-HA

B

+ -

The above results suggest, but do not prove, that CHS3 and CSA1 form a hetero-oligomer. If true, then the proposed oligomeric state of the CHS3-CSA1 TNL pair would be different from the homo-oligomerization of activated NRC

g

BN-PAGE

-

+ -

CHS3-CSA1 TNL pairs form a hetero-oligomeric complex

p

+ CSA1_E85A-HA CSA1_p-loop-HA CHS3-V5 CHS3_p-loop-V5 EV + CSA1-HA

formed the same size oligomers (Fig. 1, C and D). Overall, the sensor CHS3 and executor CSA1 from clade 2 or clade 3 oligomerize in a manner dependent on their respective pair partner and form the same size oligomers.

SDS-PAGE

A

molecular weight = ~136 kDa) in both clade 2 Per-0 and clade 3 Ws-2, we swapped tags between CSA1 and CHS3 to compare the oligomer size. Coexpression of the new constructs resulted in similar cell death phenotypes (fig. S1, D and E). In addition, we used hemagglutinin (HA) antibody to blot HA-tagged CHS3 coexpressed with V5-tagged CSA1 or HA-tagged CSA1 coexpressed with V5-tagged CHS3 to demonstrate the formation of oligomers of similar sizes (Fig. 1, C and D). These results suggest that sensor CHS3 and executor CSA1 from both clade 2 Per-0 and clade 3 Ws-2

BN-PAGE

coexpressed and that both sensor CHS3 and executor CSA1 from clade 2 Per-0 or clade 3 Ws-2 were able to oligomerize (Fig. 1, A and B). We then explored whether the TIR catalytic residue of CSA1 and intact P-loops are required for CHS3 and CSA1 oligomerization. We found that the TIR catalytic activity of CSA1 was dispensable for oligomerization of CSA1 and CHS3 but that intact P-loops of CSA1 and CHS3 were required (Fig. 1, A and B). Because the size of the CHS3 monomer (predicted molecular weight = ~125 kDa) is slightly smaller than that of CSA1 (predicted

PS

PS

(Left) HA-tagged CSA1. (Right) V5-tagged CHS3. Amino acid abbreviations: E, glutamic acid; A, alanine. HA and V5 are the epitope tags. p-loop indicates P-loop dead mutant. PS (ponceau stain) indicates protein loading. (C and D) CSA1 and CHS3 from either clade 2 (C) or clade 3 (D) form similar size oligomers. 2 of 11

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

class helper NLRs after sensor NLR activation (30–32). We performed coimmunoprecipitation (co-IP) assays to test for self-association of CHS3 and CSA1 in the presence or absence of the respective pair partner. We made HF- and Myc-tagged CHS3 or CSA1, and these new constructs functioned as previously reported (43) (fig. S2). We coexpressed two differently tagged CSA1s or CHS3s with either empty vector (EV) or with its partner and performed

A

co-IP assays. We found that self-association of either CSA1 or CHS3 was weak or not detectable in the absence of its partner (Fig. 2, A and B, and fig. S3, A and B). Coexpression with the partner greatly increased the co-IP of differentially tagged CSA1 or CHS3 from either the clade 2 or clade 3 pair (Fig. 2, A and B, and fig. S3, A and B). Meanwhile, a heterocomplex containing both CHS3 and CSA1 was readily detected if they were coexpressed (Fig. 2, A

B

Per-0

C

Ws-2

+ +

+

+ +

+ -

-

CSA1-HF CSA1-Myc

+ -

+ +

+

+ +

+ -

-

CHS3-V5 CHS3-Myc

+ -

+ -

+ -

-

+ -

+ -

+

+ +

+ +

CHS3-V5 CHS3-Myc

+ -

-

+ -

+ -

+

+ +

EV

+

+

+

-

+

+

-

kDa

+ +

EV

+

+

+

-

+

+

-

-Flag

- 130

-Flag

- 130

-V5

- 130

-V5

- 130

-Myc

- 130

-Myc

- 130

- 130

- 130

-V5

- 130

-Myc

- 130

+ +

+

-

-

-Flag SDS-PAGE

+ + -

+

-

kDa

+ -

+ +

+ +

+

-

-

kDa

+ -

+ + -

+

-

kDa

- 1236 - 1048

- 1236 - 1048

- 1236 - 1048

- 1236 - 1048

- 720 - 480 - 242

- 720 - 480 - 242

- 720 - 480 - 242

- 720 - 480 - 242

- 130

-Flag

Fraction number

kDa 13 15 16 17 18 19 20 22 24 26 28 30 32 36 130-

* CSA1 * CHS3

-Flag

Ws-2

CSA1-HA CSA1-HF CSA1_p-loop-HA CHS3-HA CHS3-HF EV

+ -

+ +

+ +

+ -

+ + -

+

-

-

+

-

BN-PAGE

-HA

-HA

Fraction number

D

kDa

+ -

+ +

+ +

+ -

+ + -

+

-

-

+ -

kDa

- 1236 - 1048

- 1236 - 1048

- 720 - 480 - 242

- 720 - 480 - 242

-HA

-Flag

- 130

- 130

- 130

- 1236 - 1048

- 1236 - 1048

- 1236 - 1048

- 1236 - 1048

- 1236 - 1048

- 720 - 480 - 242

- 720 - 480 - 242

- 1236 - 1048

- 720 - 480 - 242

- 720 - 480 - 242

- 720 - 480 - 242

- 720 - 480 - 242

- 130

- 130

SDS-PAGE

y g

BN-PAGE

kDa

+ -

Input

+ +

3 5 7 9 11 13 15 1719 21 23 25 27 29 3133 35 37 39

y

F + -

Salt peak

Void volume

kDa

-Flag

Per-0

CSA1-HA CSA1-HF CSA1_p-loop-HA CHS3-HA CHS3-HF EV

Input

-Myc-IP

-Myc-IP

-Myc

High oligomer peak ~700 kDa

g

-V5

- 130

Per-0 CHS3-Flag+CSA1-Flag

p

E

- 130

Input

+ -

Input

CSA1-HF CSA1-Myc

-Flag

and B). We then coexpressed the CHS3-CSA1 pair from clade 2 Per-0 in insect cells and purified the proteins for size-exclusion chromatography (SEC) (fig. S3, C and D). We found that purified clade 2 Per-0 CSA1 and CHS3 proteins formed a high–molecular weight complex that coeluted in fractions from a size-exclusion column consistent with heterooligomer formation (Fig. 2, C and D). We performed co-IP followed by BN-PAGE to further

-Flag SDS-PAGE

-Flag

- 130

-HA

- 130

-HA

- 130

- 130

Fig. 2. The CHS3-CSA1 TNL pair forms a hetero-oligomeric complex. (A and B) Self-association of CSA1 or CHS3 is weak in the absence of its partner in both clade 2 (A) and clade 3 (B). HF, 6×His-3×Flag. (C) SEC shows the high oligomer formation of the clade 2 Per-0 CHS3-CSA1 pair. SEC analysis of the purified CHS3-CSA1 pair proteins from clade 2 Per-0 on a Superose 6 Increase Yang et al., Science 383, eadk3468 (2024)

16 February 2024

-Flag-IP

BN-PAGE

-Flag SDS-PAGE

-HA

- 130

- 130

10/300 GL column. (D) Western blot demonstrating the coelution of sensor CHS3 and executor CSA1 from clade 2 Per-0. Western blot analysis of the distribution of CHS3 and CSA1 from the SEC fractions in (C). (E and F) The CHS3-CSA1 pair forms a hetero-oligomeric complex. (E) Co-IP–BN-PAGE results for the clade 2 Per-0 CHS3-CSA1 pair. (F) Results for the clade 3 Ws-2 pair. 3 of 11

,

-Flag-IP

y

BN-PAGE

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

determine whether sensor CHS3 and executor CSA1 form separate homo-oligomers or a heterooligomeric complex. When we immunoprecipitated either HF-tagged sensor CHS3 or executor CSA1, we detected the oligomer consistently, whether we were blotting for the sensor or its executor partner in both clade 2 Per-0 and clade 3 Ws-2 (Fig. 2, E and F). The P-loop mutant of CSA1, which was extremely weakly coimmunoprecipitated with CHS3, was used as a negative control. These data indicate that CHS3 and CSA1 oligomerize and form a hetero-oligomeric complex—and not separate homo-oligomers—to function. Conserved and clade-specific TIR domain features are required for CHS3-CSA1 function

y g y ,

4 of 11

y

16 February 2024

g

Yang et al., Science 383, eadk3468 (2024)

the AE interface of clade 3 Ws-2 CHS3 had no obvious effect on heterodimer formation, but the double mutation strongly decreased heterodimer formation (Fig. 3D). The H-to-A single mutation in the clade 3 Ws-2 CHS3TIR AE interface slightly decreased CHS3-CSA1 oligomerization (Fig. 3F and fig. S6), whereas GH-to-AA double mutation almost abolished oligomerization (Fig. 3F). Thus, the TIR domain AE interfaces of both clade 2 and clade 3 CSA1CHS3 pair are uniformly required for function, heterodimer formation, and oligomerization. We further investigated the effects of BB-loops on heterodimer formation and oligomerization. We found that the mutants of the clade 2 Per-0 CSA1TIR BB-loop and the CHS3TIR BB-loop retained CHS3-CSA1 heterodimer formation and oligomerization (Fig. 3, C and E), despite our observation that the CSA1TIR BB-loop mutant failed to induce cell death (Fig. 3A). Thus, these TIR domain features are required for clade 2 CSA1 function but are dispensable for heterodimer formation and oligomerization. The IDT-to-EAA mutation in the clade 3 Ws-2 CSA1TIR BB-loop retained heterodimer formation (Fig. 3D) but lost oligomerization (Fig. 3F and fig. S6), consistent with its full loss of cell death phenotype. Thus, the CSA1 clade 3 BB-loop is uniquely required for both function and oligomerization. The G-to-A mutation in both the clade 2 Per-0 and clade 3 Ws-2 CSA1TIR BBloops retained oligomerization (fig. S5, D and E). Mutation of the clade 3 Ws-2 CHS3TIR BBloop nearly abolished heterodimer formation and oligomerization (Fig. 3, D and F, and fig. S6), consistent with this mutant’s effect on cell death induction (Fig. 3B). Thus, the requirement for the CHS3TIR BB-loop in function, protein interaction, and oligomerization in clade 3 Ws-2 contrasts with clade 2 Per-0, which does not require the CHS3TIR BB-loop. To confirm differential requirements for function and oligomerization between the CHS3TIR BB-loops from clade 2 and clade 3, we made more mutations in CHS3TIR BB-loops to assess cell death phenotype and oligomerization of clade 2 and clade 3 pairs (table S1). We generated four-residue (FADT-to-GGGG), five-residue (YLDYR-to-GGGG), and nine-residue (FADTYLDYR-to-GGGGGGGGG) mutations in the clade 2 Per-0 CHS3TIR BB-loop and then coexpressed these CHS3 mutants with Per-0 CSA1. Meanwhile, we mutated nine residues (FTNGISRDQ-to-GGGGGGGGG) in the clade 3 Ws-2 CHS3TIR BB-loop at the equivalent position to the clade 2 Per-0 CHS3TIR BB-loop and coexpressed it with Ws-2 CSA1. Consistent with results noted above, the CHS3TIR BB-loop is required for cell death induction and oligomerization of the clade 3 Ws-2 pair but not that of the clade 2 Per-0 pair (Fig. 4, A to D). We also transiently expressed wild-type or mutant CHS3-CSA1 pairs from either clade 2 Per-0 or clade 3 Ws-2 in Arabidopsis leaves to

p

TNLs, such as RPP1 and ROQ1, oligomerize and form a tetramer upon activation (21, 22). The structures of RPP1 and ROQ1 showed that the TIR domains engage in a head-to-head symmetric interaction involving the alpha helices aA and aE, called the AE interface, after which further head-to-tail asymmetric interaction is induced, reorganizing the asymmetric BB-loop (21, 22). These two interactions in TIR domains are essential to align the TIR domains in a conformation conducive to NADase active-site function (21, 22). Several previous studies have defined requirements of the AE interface and BB-loop in TIR domain function (45–49). On the basis of the co-IP and BN-PAGE results (Fig. 2), we hypothesized that the majority of TNL CHS3 and CSA1 proteins first associate to form heterodimers, which then dimerize to a heterotetrameric oligomer. If our hypothesis is correct, then the AE interface in the TIR domains of both sensor CHS3 and executor CSA1 should be essential to mediate head-tohead interaction, but only the BB-loop in the TIR domain of executor CSA1, which contains conserved catalytic residue, should be required for head-to-tail interaction (fig. S4A). We aligned the full-length protein sequences of CHS3 and CSA1 from different clades with other plant TNLs to identify conserved residues within the AE interface and BB-loop. We then introduced mutations in the AE interfaces and BB-loops of CHS3TIR and CSA1TIR (fig. S4, B and C). We made these mutations based on either previous studies reported or the charge and side chain of amino acid (50, 51) (table S1). Consistent with published studies on plant TNLs, mutating two conserved residues (SH to AA) in the aA helix of CSA1TIR from clade 2 Per-0 or clade 3 Ws-2 completely abolished CHS3-CSA1 pair– mediated cell death (Fig. 3, A and B; cell death scale is shown in fig. S5A). The protein sequence alignments showed that the sensor CHS3 from clade 2 Per-0 contains two conserved residues (SH) in the aA helix of the TIR domain, whereas clade 3 Ws-2 CHS3 only contains one conserved residue (H) (fig. S4B). Mutating SH to AA in clade 2 Per-0 CHS3TIR abrogated

cell death (Fig. 3A). By contrast, an H-to-A single mutation in clade 3 Ws-2 CHS3TIR only slightly decreased cell death (Fig. 3B). We then generated a GH-to-AA double mutation in Ws-2 CHS3TIR and found that the cell death phenotype triggered by the clade 3 Ws-2 CHS3-CSA1 pair was largely suppressed (Fig. 3B). Thus, as expected, the AE interfaces of both CSA1TIR and CHS3TIR are required for function. We then investigated the functional roles of the BB-loops of CSA1TIR and CHS3TIR. The residues of the CSA1TIR BB-loop are conserved across all clades, but those of the CHS3TIR BB-loop are not (fig. S4C). Therefore, we mutated several conserved residues in the CSA1TIR BBloops and some residues at equivalent positions in CHS3TIR BB-loops (table S1). We found that an IDT-to-EAA mutation in the CSA1TIR BB-loop abolished cell death elicited by the clade 2 Per-0 or clade 3 Ws-2 CHS3-CSA1 pair (Fig. 3, A and B). In addition, we generated a G-to-A mutation in the CSA1TIR BB-loop, a highly conserved BB-loop residue that is essential for ROQ1-mediated cell death (22) (fig. S4C). This mutation also resulted in the loss of CHS3CSA1 pair–mediated cell death in both clade 2 and clade 3 pairs (fig. S5, B and C). These results are consistent with the formation of a CHS3-CSA1 TNL pair resembling the RPP1- or ROQ1-like heterotetramer, where the CSA1TIR BB-loop is required for NADase active-site opening. We mutated four residues (FADT-toDEAA) in the clade 2 Per-0 CHS3TIR BB-loop and coexpressed this CHS3 mutant with Per-0 CSA1 to assess cell death phenotype. We found that this BB-loop mutant of clade 2 Per-0 CHS3 with CSA1 retained strong cell death induction (Fig. 3A). However, when we introduced a fourresidue (DVFT-to-AEDA) mutation into the clade 3 Ws-2 CHS3TIR BB-loop and coexpressed it with Ws-2 CSA1, the cell death phenotype was strongly decreased (Fig. 3B). These data suggest that, despite the close relatedness of these two CHS3-CSA1 TNL pairs, the BB-loop of clade 2 Per-0 CHS3 is dispensable for cell death induction, whereas that of clade 3 Ws-2 CHS3 is essential. Next, we performed co-IP and BN-PAGE experiments comparing wild-type CHS3 and CSA1 with these mutant alleles from both clade 2 and clade 3 CHS3-CSA1 pairs to further explore TIR domain structural requirements for heterodimer formation and oligomerization. The SHto-AA mutation in the TIR domain AE interface of either clade 2 Per-0 CSA1 or CHS3 led to loss of CHS3-CSA1 interaction and oligomerization (Fig. 3, C and E), consistent with this mutant’s loss of cell death phenotype (Fig. 3A). Similarly, mutation in the AE interface of clade 3 Ws-2 CSA1 resulted in loss of protein interaction and oligomerization (Fig. 3, D and F, and fig. S6), consistent with a requirement of these residues for function and parallel to the corresponding clade 2 Per-0 mutant. The single mutation in

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

D

Per-0 + + -

+ + -

+ + -

+ +

+ -

CHS3_F39D/A40E/D41A/T42A-V5 EV

+

-

-

-

-

+ - kDa

Cell death Score

5

Ws-2 + + -

CSA1-HA CSA1_S33A/H34A-HA CSA1_I48E/D49A/T50A-HA CHS3-V5 CHS3_H29A-V5 CHS3_G28A/H29A-V5 CHS3_D42A/V43E/F44D/T45A-V5

+ + -

+ + -

+ + -

+ + -

+ +

kDa

-V5

- 130

-HA

- 130

-HA

- 130

-V5

- 130

-V5

- 130

-HA

- 130

-HA

- 130

-HA-IP

Input

- 130

+

+ - -

+ -

-

-

-

- + + - -

+ -

+ -

- + + - -

-

-

-

-

-

-

+ -

+ -

-

-

-

-

+ kDa

- 1236 - 1048

BN-PAGE

-

+ kDa

Ws-2 CSA1-HA CSA1_S33A/H34A-HA CSA1_I48E/D49A /T50A-HA CHS3-V5 CHS3_H29A-V5 CHS3_G28A/H29A-V5 CHS3_D42A/V43E /F44D/T45A-V5

- 720 - 480 - 242 -V5

-V5

- 130 PS

16 February 2024

-

+ -

+ -

+

+ + - + - - - -

-

-

-

-

+

-

-

-

-

+ -

+ -

- + - +

+ -

-

+ -

+ -

- - + + - - - + - -

-

-

-

-

-

+ kDa

-

-

-

- 1236 - 1048

-HA

-

-

+ -

- + kDa

- 1236 - 1048

- 720 - 480 - 242 -HA

Fig. 3. CHS3-CSA1 TNL pair function and oligomerization require both conserved and clade-specific TIR domain features. (A and B) In planta (N. tabacum) cell death phenotypes (left) and the corresponding percentage representation of cell death scores (right). Cell death scoring is described in fig. S5A. Images were photographed under UV light at 4 to 5 dpi. Dead tissue is shown in cyan, and live tissue is shown in gray (here and in subsequent figures). Conserved residues in the TIR domain AE interface in both CSA1 and CHS3 and the function of the CSA1TIR BB-loop are required for cell death induction in clade 2 (A) and clade 3 (B). The CHS3TIR BB-loop of clade 2 Per-0 is dispensable for cell death induction (A), but the CHS3TIR BB-loop of clade 3 Ws-2 is required for cell death induction (B). Stacked bars are color-coded showing the proportions (in percentage) of each cell death score (0 to 5). Fifteen leaves were scored for each stacked bar. Clade 2 accession and proteins are in Yang et al., Science 383, eadk3468 (2024)

+ + - -

- 720 - 480 - 242 -V5

- 130 PS

-V5

- 130 PS

blue, and clade 3 accession and proteins are in gray. (C and D) Co-IP assays show the effects of mutations in the AE interfaces and BB-loops of CHS3TIR and CSA1TIR on protein interaction. (C) Conserved residues in the TIR domain AE interfaces of CSA1 and CHS3 are required for interaction of the clade 2 Per-0 pair, but residues in the BB-loop are dispensable. (D) By contrast, although conserved residues in the AE interface are also required for interaction of the Ws-2 clade 3 pair, the CHS3TIR BB-loop is also essential. (E and F) BN-PAGE assays show the effects of TIR domain AE interface and BB-loop mutations of CSA1 and CHS3 on oligomerization. (E) Conserved residues in the TIR domain AE interfaces of CSA1 and CHS3 are required for oligomerization of the clade 2 Per-0 pair, but residues in the BB-loops are dispensable. (F) By contrast, residues in both the AE interfaces and the BB-loops of CSA1TIR and CHS3TIR are required for oligomerization of the clade 3 Ws-2 pair. 5 of 11

,

PS

+

y

- 130

+ -

y g

-HA

+ -

- 1236 - 1048

- 720 - 480 - 242 -HA

+ -

SDS-PAGE

+ - -

BN-PAGE

+ -

+

y

F

+ -

+ -

g

-V5

Per-0

CSA1-HA CSA1_S30A/H31A-HA CSA1_I45E/D46A /T47A-HA CHS3-V5 CHS3_S25A/H26A-V5 CHS3_F39D/A40E /D41A/T42A-V5

4

p

+ -

Input

2 3

1

CSA1-HA +CHS3_D42A/V43E/F44D/T45A-V5

CSA1-HA CSA1_S30A/H31A-HA CSA1_I45E/D46A/T47A-HA CHS3-V5 CHS3_S25A/H26A-V5

-HA-IP

CSA1_S33A/H34A-HA +CHS3-V5

CSA1_I48E/D49A/T50A-HA +CHS3-V5

CSA1-HA +CHS3_F39D/A40E/D41A/T42A-V5

SDS-PAGE

0

CSA1-HA +CHS3_G28A/H29A-V5

CSA1_I45E/D46A/T47A-HA +CHS3-V5

E

CSA1-HA +CHS3-V5

CSA1-HA +CHS3_H29A-V5

CSA1-HA +CHS3_S25A/H26A-V5

C

60

CSA1_S30A/H31A-HA +CHS3-V5

Ws-2

40

100 80

60

40 20 0

Per-0 CSA1-HA +CHS3-V5

B

100 80

Cell death Score (%)

Cell death Score (%)

20 0

A

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

N. benthamiana and N. tabacum were grown in a growth chamber with 24°C/20°C and 6 of 11

,

16 February 2024

Materials and methods Plant material and growth condition

y

Oligomerization of NLRs plays an essential role in defense activation and cell death induction (18–31). To date, the oligomeric state and structural requirements for the formation of activated oligomers have not been addressed for paired NLRs, a major subtype of plant intracellular innate immune receptors. The TNL pair RRS1-RPS4 (RESISTANCE TO RALSTONIA SOLANACEARUM1–RESISTANCETOP.SYRINGAE 4) (evolutionarily related to the CHS3-CSA1 pair) was thought to be unable to form a RPP1or ROQ1-like heterotetramer because the sensor RRS1 does not contain a conserved catalytic Glu residue in its TIR domain, and its BBloop was thought to be too short to support RPP1TIR-like asymmetric homodimer formation (21). However, we found that activated CHS3-CSA1 TNL pairs from two closely related clades oligomerize and form a hetero-oligomeric complex (