9 FEBRUARY 2024, VOL 383 ISSUE 6683 
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

Help wanted, scientists need apply

T

he world is continuously being transformed by science and technology (S&T), but to deliver equitable benefits to the public, scientists must be embedded in influential sectors of society—policy, diplomacy, journalism, law, business, education, and more. This means injecting PhD-level experts at every stage of research and development, from ideation, investigation, and investment to manufacture, deployment, regulation, and after-market evaluation. At first glance, it seems unlikely that scientists would be welcome outside of science itself. Trust in science was severely politicized and eroded during the pandemic, so why would scientists be invited into nonscience spheres of influence? Moreover, familiarizing graduate students with “nonresearch” pursuits and demonstrating how they can apply their skills to ensure that S&T addresses societal challenges are not standard in PhD science curricula. Although trust in science has declined in the United States, trust in scientists has not. A S&T Action Committee survey comprising 48% Republicans, 38% Democrats, and 13% Independent Americans showed that nearly 80% of respondents across political affiliations found politicization and distrust of science to be concerning, and 80% wanted scientists to contribute more to shaping public policy. In fact, The Pew Research Center found that public confidence in scientists increased in the US, from 76 to 87%, during the pandemic. Why? Perhaps, as O’Connor and Weatherall suggest in The Misinformation Age, it is because the public accords scientists broad trust as experts in evidence-based deliberations and finding solutions to complex problems. Scientists generally know little about the nonscience parts of governance, communication, and the economic and procedural systems that affect whether and how science discoveries advance through development and into the public domain. Here, academia, funding agencies, and professional societies and foundations bear crucial responsibility. Organizations that fund research and training should require that universities provide substantial career exploration courses to graduate trainees. Optional 1-day anecdotal presentations are insufficient, and deferring such offerings to the postdoc stage is too late. In parallel, professional societies and advocacy organizations should offer internships or fellowships that place newly minted science PhDs into influential outside-of-science environments.

Such programs do work. For example, in 2014, the US National Institutes of Health awarded pilot grants to 17 universities to devise training programs aimed at furnishing graduate trainees with working knowledge of careers in which their scientific expertise would bring great value. The University of California, San Francisco, created Motivating Informed Decisions, three 4-hour workshops followed by a dozen biweekly “peer mentoring” meetings, in which each member of six- to eight-person teams defines and progressively interrogates potential career goals. Together with the university’s Internships for Career Exploration program, which supports 10-week internships in diverse career settings, graduate students have reported confidence in embarking on diverse career paths. Some internship hosts were so impressed by the participating students that they offered their interns jobs. Another example is the 50-yearold Science and Technology Policy Fellows program of the American Association for the Advancement of Science (AAAS, the publisher of Science). Members of Congress and heads of federal agencies bolster their staffs with these PhD fellows, and the outcomes speak for themselves: Among 4000 graduates of this program, 50% remained in policy, 25% returned to science careers, and 25% moved to other fields. Several states have now established similar programs to populate state government offices with scientists. Clearly, the need for supporting career exploration is not limited to the US. Royal Society programs in the United Kingdom, for example, provide senior researchers and their graduate students and postdocs opportunities to learn about, and contribute to, law and the judiciary, business, management, finance, insurance, and other disciplines. Globally, we’re blessed with a robust cohort of talented S&T doctoral trainees who go on to make breakthrough contributions in academia or the private sector. But in today’s complex world, we also need scientists resolving problems and rendering decisions across society to ensure that S&T adds value to lives and livelihoods, sustains a safe environment, and contributes to security and economic well-being for all. We must illuminate those career pathways for all trainees.

PHOTO: UNIVERSITY OF CALIFORNIA, SAN FRANCISCO

“…scientists must be embedded in influential sectors of society…”

Keith R. Yamamoto is a professor in the Department of Cellular and Molecular Pharmacology at the University of California, San Francisco (UCSF), and vice chancellor for Science Policy and Strategy and director of Precision Medicine, UCSF, San Franciso, CA, USA. He is also the president of AAAS, Washington, DC, USA. yamamoto@ ucsf.edu

–Keith R. Yamamoto

10.1126/science.ado4539

SCIENCE science.org

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

57 1

NEWS IN BRIEF

142%

Estimated increase by 2050 of annual new cancer cases in lowincome countries, compared with 2022’s level, according to the World Health Organization. Causes include increased tobacco and alcohol use, obesity, and air pollution.

Edited by Jeffrey Brainard

CLIMATE POLICY

EU mulls bigger carbon emissions cut

T

he European Union should aim to reduce net carbon emissions by 90% by 2040, one of the most aggressive targets of any bloc or nation, the European Commission proposed this week. The EU has already committed to cuts of 55% by 2030 and zero net emissions by 2050 to help avoid the worst effects of climate change. Some commentators say the 2040 milestone proposed by the Commission does not go far enough, noting that the 90% target was the minimum recommended by an EU scientific advisory panel. The proposal also omits a politically sensitive statement in an earlier, leaked draft that suggested it would be possible to reduce some kinds of carbon emissions from agriculture by 30% by 2040, compared with 2015 levels. The Commission also released a new carbon management strategy, including plans for scaling up carbon capture and storage. The proposed target requires approval by the European Parliament; some members have voiced opposition.

Rare fossil reveals weird early tree | The earliest trees, from nearly 400 million years ago, are known mostly from fossils of their trunks; their leaves and canopy shapes have remained a mystery. A newly reported, 350-million-yearold tree found in Canada provides a vivid answer for one such primordial species: As if having a perpetual bad hair day, a thick PA L E O N T O L O GY

1m

crown of spiky leaves stuck out perpendicularly from the trunk. Scientists named the tree Sanfordiacaulis densifolia, after the owner of the New Brunswick quarry where they found five specimens. The fossils, among the few showing trees with attached leaves and branches, probably were preserved when a landslide buried them in a lake. These trees stood at least 2.6 meters tall, with each of their more than 200 leaves extending about 1.7 meters, the research team reported last week in Current Biology.

COVID-19 poll backs natural cause | In a global survey released last week, most scientists said the COVID-19 pandemic probably began when a natural virus jumped from an animal to a human, not because of an accident in a research lab studying coronaviruses. Organized by the Global Catastrophic Risk Institute, a U.S. think tank, the poll set out to gauge opinions among a larger and more diverse set of researchers than is usually quoted about the controversy. (The

P U B L I C H E A LT H

572

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

pollsters excluded scientists in 60 countries and territories—including China, where evidence indicates the pandemic started— deemed “not free.”) On average, the 168 respondents, most of them virologists and epidemiologists, assigned a 77% likelihood to the natural origin scenario and a 21% likelihood to a lab accident. Still, one in five said there was at least a 50% chance that COVID-19 resulted from a lab mishap. The results quickly triggered a fierce debate on social media about the study’s methodology and seemed unlikely to settle the question among scientists: Only 12% of respondents said no further studies into the pandemic’s origin are necessary.

CERN maps path for big collider | The future of the European laboratory CERN became clearer this week as officials updated their plan to build two enormous particle colliders in a tunnel that would cross the border of Switzerland and France, circling beneath both countries. The tunnel’s path has been decided, yielding a circumference of 91 kilometers, slightly shorter than the initial goal of 100 kilometers, lab officials said. The first of the colliders, to be built by the mid-2040s, would be an electron-positron collider to produce copious amounts of a known particle, the Higgs boson, and study it in detail. To generate new types of particles, CERN would then build a proton-proton collider with an energy seven times that of CERN’s current Large Hadron Collider, which enabled researchers to discover the Higgs boson in 2012. The tunnel and first collider would cost roughly €15 billion. CERN hopes to decide by 2028 whether to proceed with it. PA R T I C L E P H YS I C S

U.S. seeks plan for wolf recovery | The U.S. Fish and Wildlife Service (FWS) will create its first national conservation plan for gray wolves, which may help harmonize regional efforts to preserve the controversial species. Its populations—and people’s willingness to save them—vary widely across the lower 48 states. FWS protects gray wolves in several northeastern states C O N S E R VAT I O N

science.org SCIENCE

ENERGY

Simpler heating method raises hopes for laser fusion

R

PHOTO: EUGENE KOWALUK/UNIVERSITY OF ROCHESTER LABORATORY FOR LASER ENERGETICS

esearchers this week reported that a simpler way to use lasers to spark nuclear fusion may make it more viable as an energy source. Fusion, which promises carbonfree energy generation, works by forcing nuclei of heavy hydrogen isotopes to merge—the process that powers the Sun. In 2022, the powerful laser at Lawrence Livermore National Laboratory’s National Ignition Facility (NIF) achieved a world first by crushing a pellet of fuel and sparking a fusion

under the U.S. Endangered Species Act because of small populations. The predators are more plentiful in western states, but hunters are killing growing numbers in Idaho and Montana as permitted by new laws there. In contrast, Colorado’s wildlife managers began to reintroduce wolves in December 2023, as required by a state ballot initiative. The inconsistent picture has forestalled FWS efforts to delist certain populations because judges have ruled the agency did not consider the species’ recovery nationwide. FWS aims to finish its plan by the end of 2025.

GI virus gets monkey model B I O M E D I C I N E | Rhesus macaques are susceptible to infections by human noroviruses, researchers report this week, providing the first animal model for studies to develop drugs and vaccines for the viruses. Infamous for ruining holiday cruises and large catered events, noroviruses cause an estimated 700 million gastrointestinal infections each year, prompting diarrhea and vomiting. Most people recover within days, but infections can become chronic or fatal in the elderly and immunocompromised. Rhesus monkeys given oral norovirus doses didn’t experience those symptoms, a team from the U.S. National Institute of Allergy and

SCIENCE science.org

reaction that generated more energy than the lasers put in. Now, a team using the University of Rochester’s Omega laser system (pictured) says it can achieve equivalent energy output more cost effectively by using an alternative approach, “direct drive.” It eliminates a step used by NIF that converts laser light into x-rays to trigger the fusion reaction—a change that allows the fuel targets to be simpler and cheaper, the research team reports this week in Nature Physics.

Infectious Diseases found, but they did develop an infection resembling a human one. For example, the virus replicated primarily in the jejunum, a part of the small intestine, as in humans. The animals also failed to develop a durable immune response, making them susceptible to a repeat infection, the researchers write in Nature Microbiology.

Burned scroll muses on pleasure A R C H A E O L O GY | Using a combination of x-rays and artificial intelligence, scientists have reached a new milestone in efforts to read hundreds of 2000-year-old papyrus scrolls incinerated by the eruption of Italy’s Mount Vesuvius in 79 C.E. Physically unfurling the scrolls, found in what was a library near Pompeii, would destroy them. This week, a trio of researchers won the $700,000 grand prize offered by the Vesuvius Challenge organization after they digitally peered inside a scroll and recovered 11 columns of text containing more than 2000 characters written in ancient Greek. (In October 2023, two of the team members identified only a single word—porphyras, or purple.) Preliminary analysis has revealed the text contains musings about pleasure and may have been penned by an Epicurean philosopher, Philodemus.

IN OTHER NEWS

UP AND AWAY? A German company, Messer, last month submitted winning bids totaling $364 million in an auction of the U.S. government’s helium reserve and production facilities, which supply 9% of global demand. Scientists remain worried about how the sale will affect access and cost for the gas, needed for instruments such as nuclear magnetic resonance spectroscopy systems. NEWBORN SCREENING A U.S. advisory panel last week added Krabbe disease, a rare but fatal brain disorder, to the recommended screening list for U.S. infants. The 10-3 approval reverses a split vote in 2023 by the same group. Increasing evidence shows that stem cell transplants can help (although not cure) affected children if given early. SNOW LEOPARD COUNT India has completed its first systematic census of the elusive snow leopard, tallying 718 across areas that included prime habitat in the Himalayas. The Wildlife Institute of India and its partners based the number on sampling and camera traps. The International Union for Conservation of Nature classifies the species, native to Central and South Asia, as vulnerable. 9 FEBRUARY 2024 • VOL 383 ISSUE 6683

573

NE WS

IN DEP TH

In altermagnets, neighboring atoms are rotated and their magnetic spins are flipped.

PHYSICS

Researchers discover new kind of magnetism More than 200 materials could be “altermagnets,” predicted just a few years ago By Zack Savitsky

F

or thousands of years, people have been drawn to the apparent magic of magnets. Ancient Greek philosophers believed dark rocks called lodestones had souls because of their ability to move iron flakes. Physicists now know that magnetic materials glean their power from the behavior of the atoms inside them. But magnetism still holds secrets. Researchers have recently found signs of a wholly new class of magnetism, one with characteristics of each of the two conventional kinds, ferromagnetism and antiferromagnetism. More than 200 materials should exhibit the newfound phenomenon, according to theoretical predictions, and physicists are closing in on direct experimental evidence for it, which could lead to more efficient electronic devices. Already they have found a handful of materials that seem to exhibit this “fundamentally new type of magnetism,” says Paul McClarty, a physicist at the Léon Brillouin Laboratory. “It’s expanding our understanding of the ways that matter can work.” Inside solid materials, atoms are surrounded by electrons that all have a prop-

574

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

erty called spin, which endows each atom with its own tiny magnetic field. The total spin for each atom is represented by an arrow that can point in different directions. In ferromagnets, all the spins inside the material are aligned, resulting in a net magnetic field. In addition to sticking photos to the fridge, ferromagnets are useful because their spins can easily be flipped around by applying another magnetic field, creating distinct states that can be used as computer memory. This technique birthed the emerging technology of spintronics, in which information is encoded via electron spin rather than charge. In the 1930s, scientists realized it’s much more common for the spins of neighboring atoms to point in opposite directions so their net magnetization cancels out. Because the staggered arrangement is much more stable than the uniform one, these antiferromagnets are nearly impossible to magnetize with applied magnetic or electric fields. When French physicist Louis Néel won a Nobel Prize in 1970 for his pioneering work on antiferromagnetism, he described the phenomenon as “interesting but useless.” Nevertheless, the concept has proved handy: During World War II, electric coils were used to make ship hulls

behave like antiferromagnets and evade magnet-seeking mines. More recently, scientists have begun to devise strategies for building spintronic devices out of antiferromagnets. Although their rigid spins are harder to manipulate, they can in principle flip 1000 times faster than those in ferromagnets, allowing for more energy efficient information storage and processing. A few years ago, Libor Šmejkal, a physicist at the Johannes Gutenberg University of Mainz, was hunting for a possible antiferromagnetic spintronic material. He stumbled across a compound called ruthenium dioxide that seemed promising—but odd. His calculations suggested it should have no net magnetization, like a normal antiferromagnet. But he also predicted that when subjected to an electric current, the material would behave like a ferromagnet: Magnetic forces in the material would deflect the electrons in the current, leading to a strong voltage in the perpendicular direction. In 2020, a team in China experimentally confirmed ruthenium dioxide’s paradoxical properties. The following year, Šmejkal and colleagues laid out a proposal explaining how materials like ruthenium dioxide could be part ferromagnet and part antiscience.org SCIENCE

N E WS

ferromagnet. They called them altermagnets. In most materials, electron spin arrows align with the orientation of their host atoms within the crystal lattice. But in some materials, spin arrows can rotate independently of the atoms, and Šmejkal and colleagues considered one in which every other atom was rotated by 90° and its spin flipped by 180°. Altermagnets would combine the most prized features of ferromagnets and antiferromagnets. With zero net magnetization, they are graced with the stability and fast spin-flipping speeds of an antiferromagnet. But the spins in an altermagnet, like those in a ferromagnet, can be readily ushered into distinct up and down states, allowing for easier memory writing. “You can have your cake and eat it, too,” says Jairo Sinova, another physicist in the Mainz group. Whereas ferromagnetic spins are typically flipped with magnetic fields, spins in an altermagnet could be manipulated by applying currents in different directions. Theorists were quick to accept Šmejkal’s description because of its mathematical elegance, but many are surprised the phenomenon went unnoticed for so long. “It’s one of those theoretical constructs which are unquestionable,” says Igor Mazin, a physicist at George Mason University. “Yet it has never been discussed before.” More than 200 materials are predicted to be altermagnetic—more than double the number of known ferromagnetic materials. Researchers are now beginning to look for the property by shining laser light on a material to coax it to eject electrons. By measuring the properties of those electrons, scientists can look for a hallmark of altermagnetism: energy levels that fall within two distinct bands, reflecting both spin-up

and spin-down electrons. (Antiferromagnets also have spin-up and spin-down electrons, but they sit at the same energy levels.) Last month, a team in South Korea found the predicted split in electron energies in the material manganese telluride. Two additional recent studies identify similar signals in manganese telluride and ruthenium dioxide, and also attempt to tie the energy bands to specific spin polarities. “Crystalclear proof is really hard to achieve experimentally,” says Suyoung Lee, a Ph.D. student at Seoul National University who led one of the latest studies. “But I would say that we now have sufficient experimental evidence … that altermagnetism is really a thing.” McClarty says the new experiments are “consistent with altermagnetism,” but only reveal the spin behavior through a slice of the material’s magnetic landscape. Until experimentalists capture the behavior across an entire 3D structure, “I wouldn’t hang up my coat,” he says. Also, before altermagnets can be exploited in electronic devices, scientists must learn to synthesize materials that have a consistent altermagnetic orientation rather than a patchwork of shifting configurations. Mazin says confirmation of materials’ altermagnetic makeup is all but certain. “There’s no way in nature that they wouldn’t be,” he says. He sees the verification effort akin to “an experiment that proves two times two is four.” But for Lee, the hunt promises other payoffs: an opportunity to explore emergent complex phenomena that may lead to practical applications. “I think this is the starting point for a whole new field of altermagnetism,” she says. “I am happy to be part of it.” j Zack Savitsky is a science journalist in Berlin.

Magnetism’s new twist The properties of most magnetic materials depend on whether each atom’s magnetic field—denoted by its spin—is pointing up (pink) or down (blue). In altermagnets, the atoms and their spins rotate independently, giving them properties of both ferromagnets and antiferromagnets.

GRAPHIC: A. MASTIN/SCIENCE

Ferromagnetic

Atom SCIENCE science.org

Antiferromagnetic

Electron spins

Altermagnetic

Rotated atom

CLIMATE CHANGE

NASA mission will zoom in on ocean life and bright hazes PACE satellite will study climate effects of carbon-storing plankton and reflective particles By Paul Voosen

I

t isn’t easy seeing green. The ocean’s mix of plankton, algae, and bacteria absorbs vast amounts of carbon dioxide while producing 50% of Earth’s oxygen. But for decades, Earth-observing satellites could not tease apart the many species making up the green goop. That hampered attempts to study how the floating plants influence climate—and how global warming is affecting this foundational component of the climate system. The $964 million Plankton, Aerosol, Cloud, Ocean Ecosystem (PACE) satellite is about to change that. After surviving pandemic delays and four cancellation attempts by former President Donald Trump’s administration, it was due to launch this week from Cape Canaveral, Florida, aboard a Falcon 9 rocket. Once operational, it will probe the plants floating in the ocean as well as the hazes of fine particles above it, another major influence on climate. “We’re not going to be limited technologically anymore,” says Collin Roesler, an optical oceanographer at Bowdoin College. “We’re going to be limited by our ideas.” The satellite’s primary instrument is NASA’s first “hyperspectral” imager to fly on a major geoscience mission. Rather than collecting reflected light in just a few discrete channels, like the red, green, and blue cones of a human eye, the instrument divides light into more than 200 channels, including many shades of green, which can distinguish different pigments used by phytoplankton depending on their species and habitat. “Right now, what we have on orbit is a box of crayons with only eight crayons,” says Jeremy Werdell, the mission’s principal investigator at NASA’s Goddard Space Flight Center. 9 FEBRUARY 2024 • VOL 383 ISSUE 6683

575

NE WS | I N D E P T H

“With PACE, we are getting a box of crayons with 200 different colors.” By studying the mix of phytoplankton species, PACE should help answer a key climate question, Roesler says. Large phytoplankton, which dominate in colder waters, are bulwarks of carbon storage. “These really big cells get eaten by big zooplankton and then they poop really big poop,” she says. The carbon in the waste is more likely to reach the ocean floor and stay safely sequestered. In the warmer waters at lower latitudes, smaller phytoplankton thrive— but the carbon in their cells tends to be gobbled up by microbes and ultimately emitted back into the atmosphere. Models have suggested these smaller plankton could expand toward the poles with global warming; PACE could see whether these fears are true. PACE’s sharp color vision should also help it track human impacts on lakes, rivers, and coastlines, by separating naturally occurring species from bacteria or dissolved organic matter associated with

lose any polarization. But because aerosol particles are discrete and compact, they reflect polarized light and pop out vividly. Brian Cairns, PACE’s deputy project scientist at NASA’s Goddard Institute for Space Studies, who also led development of Glory’s doomed instrument, says the new satellite carries two “petite but powerful” polarimeters, which “get almost all of what you wanted in a much more compact package.” By identifying the type, size, and abundance of particles, the two polarimeters should fill out the inventory of aerosols, including elusive ones such as sea spray over the ocean. The instruments should also show how aerosols influence the growth and life span of clouds, Cairns says, which can affect the rate of warming. Researchers are currently debating the extent to which cleaner shipping fuels have suppressed the formation of reflective clouds over the ocean, thereby adding to global warming (Science, 4 August 2023, p. 467). James Hansen, the famed former NASA climate scientist who is now at Columbia

NASA’s latest satellite will scan the planet every 2 days, tracking shifts in ocean plankton and particles in the sky.

sewage or fertilizer runoff. And it will explore the role that eddies, kilometers-wide swirls of turbulent water, play in driving the movements of plankton species and the nutrients they depend on. The satellite will restore NASA’s ability to study another key factor in climate: the small aerosol particles and clouds that reflect some 90% of the Sun’s light into space. NASA is still reeling from the loss of the $424 million Glory satellite, which crashed soon after launch in 2011. It carried an advanced polarimeter designed to capture polarized light, in which the waves all vibrate in the same plane. Sunlight bouncing chaotically off the ground and clouds tends to 576

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

University, welcomes the new attention to aerosols. Hansen recently published work suggesting that global warming is accelerating as pollution wanes. “As the world realizes the climate predicament that we face, they will rue not having a monitor of aerosol climate forcing.” The polarimeters, built with off-the-shelf components, may only last a few years, but PACE’s main instrument could run for a decade or more. It’s been a long wait for Roesler, who has spent her career preparing for this intimate green view from space. “The achievement of PACE is that it brings the observing capabilities almost to what I can do in the lab. That’s really astounding.” j

CLIMATE CHANGE

Florida coral restoration in hot water After unprecedented heat wave killed transplanted coral, reef experts are charting a new strategy By Warren Cornwall

F

our years ago, the U.S. National Oceanic and Atmospheric Administration (NOAA) unveiled a $100 million coral moonshot. Over 2 decades, nearly half a million hand-reared coral colonies would be planted on seven ailing reefs in southern Florida, in a bid to revive them. Misson: Iconic Reefs represented “one of the largest ever investments in coral restoration,” Pat Montanio, then head of the agency’s habitat conservation program, said at the time. Today, the project looks as ailing as the coral it was meant to save. A recordbreaking underwater heat wave that swept the Caribbean and southern Florida in 2023 killed most of the transplanted colonies. Elkhorn coral, with its sprawling, flat branches, was to be the cornerstone of the initiative’s first phase, creating a foundation on which other corals could grow. Instead it proved to be one of the most heat sensitive species. Many elkhorn corals, both wild and hand planted, are dead, their blanched skeletons coated in algae. “The picture is not good,” says Jennifer Moore, a coral recovery coordinator at NOAA who helped lead a January meeting in Marathon, Florida, not far from one of the Iconic Reefs, to assess the damage and plot a path forward. The results are provoking a crisis of confidence in decades-old methods of reef restoration pioneered in Florida and adopted around the world. “Is it responsible to grow these species and then just put them back out to die?” asks Ian Enochs, a marine biologist who heads NOAA’s reef monitoring program in the Atlantic Ocean and Caribbean Sea. “We need to have a shift in how we conduct restoration.” When the project began, Florida’s reefs were already in desperate need of help. A string of diseases, coupled with water pollution, human disturbance, and successive heat waves, had all taken a toll. Elkhorn science.org SCIENCE

CREDITS: (GRAPHIC) M. HERSHER/SCIENCE; (DATA) NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION CORAL REEF WATCH; (PHOTO) JASON GULLEY

and staghorn corals—two species in the cess story, and found no living elkhorn or He and some others think sexual reproAcropora genus—were among the worst staghorn. Similar damage was observed duction—mixing eggs and sperm gathered hit, having declined more than 97% since through much of the Caribbean at both from spawning coral to grow new colonies— the early 1980s. natural and restored reefs. “It’s been devcould yield genetic combinations that would The rescue effort, like nearly all reef resastating everywhere,” says Margaret Miller, make some of the offspring more heat tolertorations in Florida, relied largely on breaka coral ecologist and research director ant. The strategy, which a few Florida-based ing off chunks of existing coral and coaxing for Secore International, a Miami-based labs have already begun to use, could also them to build new colonies—all clones of the nonprofit that conducts coral restoration make it easier to treat the new colonies with original coral. Restoration groups can swiftly research and collaborates with groups heat-resistant algal symbionts, which many make thousands of new corals each year by throughout the Caribbean region. corals absorb during their larval stage. suspending coral fragments in open water The destruction has led Nedimyer to Erinn Muller, an ecologist leading coral renurseries until they are large enough to atconclude that, at least for the Acropora spesearch at the Mote Marine Laboratory, says tach to the sea floor. cies in Florida, it’s foolhardy to continue to some elkhorn and staghorn corals sexually Elkhorn and staghorn have long been rely on clonally produced coral. “One of the bred in the lab there survived the hot water favorites. “They’re ridiculously easy to things I learned is we don’t have any stagconditions last summer. At an offshore nurspropagate,” says Ken Nedimyer, technical horn or elkhorn corals that can handle the ery near the southwestern tip of the Keys, director of the nonprofit Reef Renewal. But conditions in the middle and lower Keys,” they were the sole survivors of 24,000 coral their decline means the corals off Florida he says. “It’s a game changer.” fragments. “Seeing this unique sign of resilretain very little of the genetic diience within the sexually produced versity that might have enabled offspring allows us a unique oppornew colonies to cope with the offtunity to dig into what is going on,” Unbearably hot the-charts hot water they faced Muller says. Corals in Florida and the Caribbean faced a record-setting underwater last year. Miller, however, warns that the heat wave in 2023. By 15 September, many areas had endured more than Ocean temperatures rose to unstrategy risks introducing new ge20 weeks with ocean temperatures 1°C above the average during the precedented levels starting in July netic problems such as inbreeding, hottest month of the year, causing a mass coral die-off. 2023, fueled by calm, hot weather because the population of wild coral is Degree heating weeks tied to El Niño and given an added so depleted. Importing sperm or eggs 0 4 8 12 16 >20 boost by climate change. Until last from corals elsewhere could help. She year, NOAA’s scale of heat wave was part of a 2019 experiment that threats to coral reached its highproduced elkhorn coral with eggs est level when water temperatures from the southern Caribbean island hit 12 “degree heating weeks”—the of Curaçao fertilized by Florida-origin equivalent of 12 weeks of temperasperm. Today, the offspring are stuck tures 1°C above the average during in a laboratory, she says, because state Florida the hottest month of the year. But officials worry about introducing outThe last year, when sea-surface temperaside coral genes into Florida waters. Bahamas tures reached as high as 32.8°C, the To her, the current situation warrants Mission: agency added three new categories, such experiments. “I definitely think Iconic Reefs topping out at more than 20 degree additional risky measures are now heating weeks. “This is analogous to called for, unless we just want to give a Category 5 hurricane,” says Derek up,” Miller says. Manzello, a coral reef ecologist and Phanor Montoya-Maya, a marine Cuba head of NOAA’s Coral Reef Watch, biologist and head of the restorawhich devised the warning system. tion program at the Coral RestoraIn the Keys, reefs endured up to 22 tion Foundation, cautions that sexual 0 250 degree heating weeks last year. Many production alone won’t create makm species bleached, turning bone white ture corals quickly enough to prop as the coral polyps expelled symbiotic up the reefs because it can take years algae from their bodies. Some died for larvae to reach reproductive age. straight away, sheets of tissue peeling To breed corals at the scale required from their skeletons. Elkhorn and would also take more lab facilities. staghorn suffered the biggest losses. “The answer is not to stop asexual Artificially planted Acropora propagation,” he says. fared little better. There is no final Mission: Iconic Reefs isn’t over. estimate of casualties at the seven This year it will plant bulky brain corIconic Reefs; a team is currently als, produced asexually, which grow visiting them to take stock. But a more slowly but also weathered the survey in August 2023 showed 30% heat wave better, Moore says. It will of the planted staghorn and 45% also continue to plant some elkhorn of the planted elkhorn was already and staghorn to test approaches that dead, and about 90% was bleachmight be more successful. “We have ing. In late September, Nedimyer to be a little more intentional,” Moore visited Pickles Reef, a focus of In southern Florida, scientists documented widespread bleaching says. “We don’t just want to put them planting he had viewed as a sucand coral death. out to slaughter.” j SCIENCE science.org

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

577

The sweet smell of evening blooms attracts a hawk moth, a major nighttime pollinator.

ECOLOGY

Pollution disrupts nighttime pollination by degrading scents Nitrate radicals, a common pollutant, break down the cues nocturnal insects follow to find nectar sources By Elizabeth Pennisi

U

nder the cover of darkness, countless moths and other insects furiously dart around woodlands and deserts, seeking nectar from night-blooming plants—and pollinating them in the process. But the scents the insects home in on have grown fainter. Nitrate radicals, a common pollutant, break them down before they can travel far, a research team reports on p. 607. The team thinks the olfactory disruption goes as far back as the Industrial Revolution 200 years ago. The research, involving field studies, wind tunnel experiments, and the latest atmospheric models, has worrisome implications. For pollinators and the crops and ecosystems dependent on their skills, nitrate radicals are “potentially a very significant threat,” says Robbie Girling, an applied ecologist at the University of Southern Queensland. The same probably goes for other pollutants. The work demonstrates “the ‘fragility’ of these [scents],” particularly at night, adds Maryse Vanderplanck, a chemical ecologist at CNRS, the French national research agency. She, Girling, and others previously showed that ozone, another pollutant, interferes with the ability of bees and other daytime pollinators to find flowers. But impacts at night are especially concerning. “For every bee and butterfly we see by day visiting flowers, [we] know that many more 578

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

may be visiting after nightfall,” says David Wagner, an entomologist at the University of Connecticut. Nocturnal insects including moths such as hawk moths (Hyles lineata) and tobacco hornworms (Manduca sexta) depend on the pleasant scents of gardenias, honeysuckles, lilacs, jasmine, and most night bloomers to guide them long distances to nectar sources. Some crops, too, appear to lure pollinators at night, says Alison Scott-Brown, a plant biologist at the University of Cambridge. Jeff Riffell, a neuroscientist at the University of Washington (UW), and his colleagues were the first to identify a possible pollutant challenge for nighttime pollinators. In 2014, they showed that certain airborne chemicals made it more difficult for tobacco hornworms to find flowers. But early studies tended to examine high concentrations of pollutants in rather artificial situations. So, in 2016, Riffell teamed up with UW atmospheric chemist Joel Thornton and their graduate student Jeremy Chan to probe the issue under more natural conditions. In a sagebrush patch about 280 kilometers from Seattle, Chan recorded pollinator visits to pale evening primrose (Oenothera pallida). He logged a diverse array of nighttime visitors, including bees, antlions, and moths. Underscoring their importance, the plants produced few to no seeds when Chan covered blooms for the duration of flowering with a mesh bag to prevent pollination.

The researchers suspected nitrate radicals, produced when nitrogen oxides from exhaust and industrial smokestacks react with ozone, might also interfere with pollination, as they build up at night and are highly reactive. Chan, now a chemical ecologist at the University of Naples, did lab studies showing the pollutant breaks down compounds called monoterpenes, which are common in many plant scents. He then synthesized the degraded odors and, in wind tunnel experiments and field studies, showed they are less effective than the original scents at luring moths. “This paper had it all,” says Clinton Francis, a sensory ecologist at California Polytechnic State University (Cal Poly). “It was really thorough and very impressive.” The field studies showed that on average, moths visited each flower twice a night. But when Chan released the degraded scent into the sagebrush patch, the visits dropped to about one every two nights. He also placed moths in a wind tunnel together with paper cones emitting the degraded scent. “Some moths couldn’t smell the ‘flowers’ at all,” Riffell says. “It’s almost like the moths had COVID.” Chan worked with Thornton to understand where, globally, nighttime levels of this pollutant are likely high enough to disrupt pollination, finding hot spots across the Northern Hemisphere. The group also calculated that, beginning in preindustrial times, nitrogen radicals have shortened the distance scents travel by a factor of five, on average. In and near big cities, pollution degrades floral scents so quickly they are probably ineffective, the team reports. “I hope [the work] can help elevate chemical pollution to the same recognition that we give artificial light,” says Sarah Jennings, a chemical ecologist at Cal Poly, referring to another major disrupter of insect behavior. Riffell agrees, adding that nitrogen radicals and other pollutants may also disrupt the chemical cues that many animals use for mating and other purposes. Wagner is more circumspect, suggesting pollinators face greater threats, such as invasive plants. But Ilaria Negri, a zoologist at the Catholic University of the Sacred Heart, says pollution is another significant blow to insects already hammered by improper use of pesticides, diseases, climate change, and development. “Pollinators [are] literally starving,” she says. “That pollinators are unable to find already scarce food due to pollution is challenging for many species.” And that can only be bad news for the plants that depend on them. j science.org SCIENCE

N E WS | I N D E P T H

PUBLIC HEALTH

Does fluoride in drinking water risk IQ loss? A U.S. federal court is examining a controversial link between fluoride and neurotoxicity By Erik Stokstad

that determination after two reviews by the National Academies of Sciences, Engineering, and Medicine, in 2020 and 2021, found long-simmering scientific battle took the evidence was lacking. One flaw, the reon new life last week, as experts clashed views said, was that the assessment didn’t foin a San Francisco courtroom over cus on how people respond to various doses whether the U.S. Environmental Proof fluoride. tection Agency (EPA) should ban fluoThe report has yet to be formally reridation of drinking water to protect leased, and FAN charges that the repeated fetuses and children from the risk of neuroreviews have been designed to delay it—a developmental problems. critique supported by some scientists and The case, being heard in a federal disformer NTP leaders. According to emails trict court, “is precedent setting,” says Lynn released to FAN under the Freedom of InBergeson, a managing partner of Bergeson & formation Act, a senior official Campbell who focuses on chemiat the Department of Health and cal toxicity. Rarely have judges had Human Services intervened in to “manage the enormity of this 2022 to halt publication just days record of scientific evidence. … before the report was scheduled That’s why there’s a lot of attention to be released. But in March 2023, focused on this right now.” as a result of litigation, NTP did Adding fluoride, a common release a 393-page draft. And mineral, to drinking water lessens that May, NTP’s science advisory tooth decay in children and adults board signed off on the report by 25%. The Centers for Disease and sent it to NTP leadership. Control and Prevention calls fluoriDental health advocates say acdation, which began in 1946 in the tivists are already spinning the reUnited States and is decided by loport to their advantage by claiming cal water districts, one of 10 “great NTP found no safe dose (a claim public health achievements” of the not directly stated in the report). 20th century. But from the getThe American Dental Association go, some activist groups worried has strongly urged NTP to add a about potential harm. And over disclaimer to the report highlightthe past few decades, studies of Fluoride is added to drinking water at a facility in Burlington, Vermont. ing its scientific limitations. laboratory animals and of commuIn the meantime, Chen has alnities where drinking water natuEPA and others argue there is little strong lowed the lawsuit to resume. He has sugrally contains fluoride have hinted that high evidence that the current recommended congested that EPA must regulate fluoride if levels might affect brain development. The current case has put the spotlight on centration of fluoride in drinking water— FAN proves that fluoride poses an “unreaan unpublished assessment by the federal 0.7 milligrams per liter—poses a threat. (That sonable risk” to pregnant women and chilgovernment’s National Toxicology Program level is set to avoid discoloration of teeth in dren. Under TSCA, EPA cannot consider a (NTP). It reported “moderate confidence” children.) “The dose makes the poison,” said chemical’s benefits, such as oral health. that drinking water containing fluoride at Paul Caintic, a Department of Justice attorBergeson describes Chen as “a very dislevels at least twice as high as those recomney. “Given the current state of science, the ciplined jurist and a very demanding taskmended by the federal government is associcourt cannot conclude that community water master,” adding that “presumably the data ated with lower IQ in children. The Fluoride fluoridation presents an unreasonable risk.” will be looked at correctly and weighed Action Network (FAN) and other groups The fluoride lawsuit is the first to reach appropriately.” argue that such data indicate EPA should trial under a 2016 TSCA provision allowing Lynn Goldman, an epidemiologist at be regulating fluoride under the Toxic Subcitizens to ask a court to assess a chemical’s George Washington University and former stances Control Act (TSCA). risk. FAN turned to that strategy after EPA EPA official, doubts that Chen will order EPA Last week’s testimony marked the resumpsaid its request that the agency ban fluoridato ban fluoride in drinking water. “I can’t tion, after a yearslong pause, of litigation that tion lacked a “scientifically defensible basis.” imagine a court coming back and saying, ‘You began in 2017. Over 9 days, Judge Edward The trial began in June 2020 but Chen susdid a risk assessment wrong, and here’s how Chen of the U.S. District Court for the Northpended it months later to wait for publication you should have done it,’” she says. But the ern District of California is scheduled to hear of the NTP report. That assessment, based on court might order EPA to consider fluoride as the views of seven experts on a range of techwork conducted by NTP experts from 2016 to a toxic substance under TSCA and make it a nical topics, including whether there is a safe 2019, initially classified fluoride as a cognihigh priority for a new evaluation to establish threshold for fluoride exposure. tive developmental hazard. But NTP removed safe exposure limits. j

PHOTO: ALDEN PELLETT/AP

A

SCIENCE science.org

During opening arguments on 31 January, the attorney for the plaintiffs, Michael Connett, highlighted babies fed formula made with tap water as a “critical vulnerable group being exposed to the highest dose of fluoride of any age group in the population. That is a major cause for concern.” He argued that if EPA was aware of a different compound that posed a similar potential threat to newborns, the agency wouldn’t hesitate to regulate it. “I don’t think we need to speculate about what EPA would do with that scenario,” he said.

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

579

NE WS

FEATURES

A THOUSAND YEARS OF

SOLITUDE How did the first human settlers of the Canary Islands survive a millennium of isolation?

M

ore than 1000 years ago, a young man stood on the northern shore of the island now known as El Hierro. Across the wave-swept Atlantic Ocean, he could see the silhouettes of other islands, a volcanic peak on one soaring toward the clouds only 90 kilometers away. Yet, for him, those islands were as unreachable as the Moon. His body betrayed the rigors of life on his arid volcanic outcrop. His molars were

580

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

By Warren Cornwall, in the Canary Islands worn almost to the gums from grinding fibrous wild fern roots. His ancestors here had farmed wheat, but he and his contemporaries grew only barley and raised livestock such as goats. His genes held evidence that his parents were closely related, like many of the roughly 1000 people on the island, who had not mingled with outsiders for centuries. Also like many of

his fellow islanders, he bore signs of an old head injury, likely sustained in a fight. “This population faced a lot of challenges,” says archaeologist Jonathan Santana of the University of Las Palmas de Gran Canaria (ULPGC). “Survival on this island was a challenge every day.” Yet the first Canarians, who arrived from North Africa roughly 1800 years ago, survived and even thrived on this arid, windswept archipelago for 1000 years. They numbered in the tens of thousands when Europeans arrived at the start of the science.org SCIENCE

N E WS

since the 19th century. For instance, how did people with no apparent seafaring skills reach and survive on the archipelago? Why did their crops and cultures differ from island to island despite their common origin? The answers offer insights into how human societies cope with—and respond to—challenging environments, says Scott Fitzpatrick, a University of Oregon archaeologist who studies island cultures. “The Canaries have been sort of an enigma.”

PHOTO: W. CORNWALL/SCIENCE

Archaeobotanist Jacob Morales found ancient granaries high on a cliff on Gran Canaria.

14th century. Not long after, conquest and genocide had largely erased them as a people. But their DNA lives on in many islanders today, and traces of their lives remain, in granaries, cliff dwellings, ceramic figurines, and hundreds of human remains like those of the man on El Hierro—all remarkably well preserved by the dry climate. By applying the latest archaeological tools to this trove of material, Santana and other home-grown archaeologists are unearthing their stories, shedding light on puzzles that have mystified archaeologists SCIENCE science.org

Africa was home to a variety of Amazigh societies, from pastoral herders in rural tribes to kingdoms with large urban centers heavily influenced by Roman culture. In the latest genetic study, published in Nature Communications in 2023, a team led by geneticist Rosa Fregel of the University of La Laguna sequenced the whole genomes of 49 ancient individuals spanning all major islands and dated from about 300 to 1500 C.E. Her team found that distinct genetic groups, each tied to Amazigh populaTHE CANARY archipelago sits like a small curvtions, had settled the western and eastern ing tilde 100 kilometers off the North Africlusters of islands. can coast (see map, p. 582). Its major islands But DNA doesn’t answer how the first range in size from El Hierro, the smallest at settlers crossed the watery gulf from the 269 square kilometers, to mainland. Peter Mitchell, Tenerife, more than seven an archaeologist at the times larger, roughly the University of Oxford, dissize of Maui. The islands’ misses speculation that volcanoes, some of which they were deported to are still active, reach up the islands by Romans to 3718 meters. Vegetaexpanding into Africa tion ranges from parched in the first century C.E. cacti to shrubland with Archaeological and writwild olives and juniper. ten evidence shows the On a few islands, everRomans knew of the green forests of pine and islands and briefly oplaurel thrive on moisture erated a dye factory on brought by northeasterly a tiny outcrop there, exAncient people on El Hierro, the smallest trade winds. tracting a prized purple of the Canary Islands, wore down When the first Eurohue from a marine snail. their teeth by eating tough wild plants. peans arrived in the 14th But Mitchell says the Rocentury, they found all seven major islands mans had no practice of wholesale deportaoccupied, with up to tens of thousands of tions of communities, usually opting to kill people each on Gran Canaria and neighboror enslave troublesome groups. ing Tenerife. What the Europeans didn’t find Instead, Mitchell suggests the Amazigh was any evidence of seaworthy craft. The issettlers took to the sea to escape conflicts land dwellers “have no ships or other means sparked by a drying climate or the Roman to get from one [island] to the other, unless expansion. Some islands might have seemed they swim,” wrote Nicoloso da Recco, a Genolike an oasis compared with the desert from ese navigator who visited the islands in 1341 which those settlers came. “People might on behalf of the Portuguese monarchy. think going over the Atlantic to get here, European archaeologists were fasci‘This is quite worth it. We’ve never seen so nated with the early Canarians. The French many trees,’” Mitchell says. thought the first settlers were Cro-Magnon, Santana hypothesizes that people’s sealike prehistoric people in France; German faring skills declined over time because they archaeologists thought they must have been had little incentive to traverse the ocean for Aryan; the Spanish thought they were Stone trade. The islands’ volcanic rock contains Age relatives of the same North Africans who no valuable metal ore, and the dry ground settled the Iberian Peninsula. More recently, yielded little food to spare. “It wasn’t worth archaeologists from the Canaries themselves it,” Santana says. have begun to lead investigations. They have tapped into evidence including anA DARK CAVE entrance beckons from the cient bones, the genes of living people, and middle of a 140-meter-tall cliff flanking 1000-year-old grains of barley. a narrow valley on Gran Canaria. Jacob By analyzing ancient DNA from Morales, an archaeobotanist at ULPGC, radiocarbon-dated bones, archaeologists crouches at the end of a ledge, the wind in the past 15 to 20 years have found that whipping his face. Below him lies a landthe first islanders had the strongest genetic scape of stunted shrubs, dry grasses, and ties to the Amazigh cultures of northwestprecipitous, rocky ridges that Spanish phiern Africa, also known as Berbers. Rock inlosopher and poet Miguel de Unamuno scriptions on the islands also echo Amazigh once described as “una tempestad petrialphabets. Two thousand years ago, North ficada” (a petrified storm). A 2-meter gap 9 FEBRUARY 2024 • VOL 383 ISSUE 6683

581

NE WS | F E AT U R E S

of blank sheer rock separates Morales from the cavern. Morales doesn’t plan to cross that blank wall today. But in 2011, he choked back his fear of heights and made his way up to the cave with the aid of ropes set by a rock climber. Researchers at the local archaeological museum wondered why he bothered, he recalls. Other archaeologists had entered the cave as early as the 1980s and toted away the obvious artifacts—bits of basketry and pottery. But Morales was hunting smaller quarry. Sifting through piles of sediment, he picked out tiny fragments of seeds representing a rich variety of foods: barley, durum wheat, lentils, fava beans, and figs. Radiocarbon dating revealed the seeds were as much as 1000 years

The granaries and their variety of foods also suggest that Gran Canaria, where central highlands capture up to 1 meter a year of rainfall compared with 10 to 20 centimeters elsewhere, was one of the most productive islands in the archipelago. Even today, the mesa topping the cliff above the granary “is the best place for cultivating barley in this area,” Morales says. The barley grown there today traces its lineage back to the ancient granaries. Morales and Swedish researchers analyzed DNA in barley seeds found in dwellings and granaries from three islands dating between 1000 and 600 years ago. Each island had a different strain descended from a single shared variety, which itself separated from

The lonely islands People from North Africa settled the Canary Islands, an arid volcanic archipelago, in the early centuries C.E. when the Roman Empire was at its largest. Evidence suggests little or no contact with the mainland for the next 1000 years. Extent of Roman Empire 117 C.E.

Canary Islands

Lanzarote

La Palma Tenerife

Fuerteventura

La Gomera Gran Canaria

El Hierro

Morocco 0

75

Atlantic Ocean

km

old; all predated the arrival of Europeans. “It’s amazing preservation,” he says. Morales’s discoveries showed that some of the ancient Canarians stored food in cliffside granaries. They also indicated that each island’s society proceeded down its own path. People on Gran Canaria—but only on that island—chipped small alcoves from volcanic rock, lined them with plaster, and added wooden doors to protect their food. Each of the alcoves, which were found in more than 50 sites, contained a medley of different seeds, suggesting to Morales that they belonged to separate families who stashed diverse crops in the same space. 582

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

African barley about 2000 years ago, according to genetic estimates. All the strains were genetically distinct from today’s European and North African barley. The team also analyzed other crops; seed by seed, they pieced together evidence that a similar “packet” of domesticated African crops arrived with the first settlers of each island. The inhabitants of some other islands were less fortunate than those on Gran Canaria. On those islands, many crops vanish in more recent layers of sediment, perhaps because of the harsh environment or climatic shifts. On Fuerteventura, for example, signs of barley, wheat, and lentils end after

the eighth century. European colonists in the 15th century noted the island’s residents lived chiefly on livestock meat and seafood. On El Hierro, wheat vanishes from the record, but barley remains. “At the beginning there is exchange of seeds between islands,” Morales says. “Then there is isolation.” The genetic differentiation in barley is mirrored in the archipelago’s people, painting a picture of profound separation from both other islands and the mainland. For example, the genetic split between people colonizing the western and eastern islands persisted even in remains dated to hundreds of years after the islands were settled. “If migration were something that was super common between islands, we shouldn’t see two clusters [persist],” Fregel says. Still, some archaeologists do see signs of new arrivals, at least on Gran Canaria, in changes in how the dead were treated. The first burials, around 300 C.E., were interred communally in caves, says Verónica Alberto-Barroso, an archaeologist at ULPGC. Around 600 C.E., some people began interring bodies in individual open-air tombs of piled rocks, a trend that lasted about 500 years. Then around 1200 C.E., stonelined pits and graves grew popular. Alberto-Barroso and her colleagues also note simultaneous shifts in where people lived on the island, and the kinds of homes and ceramics they made. They conclude new North African groups with new cultural practices might have arrived during these periods. But the changes could simply mark cultural evolution, Santana notes. Other islands don’t show such cultural shifts and some display even stronger signals of genetic isolation. Fregel found that in ancient human remains from four islands— Fuerteventura, Lanzarote, La Gomera, and El Hierro—the chromosomes inherited from each parent were closely related, a sign of a small, inbred population. At least one individual on each island had chromosomes more similar than if their parents had been first cousins. On El Hierro, Fregel analyzed shared genetic variants in four individuals who died between the 13th and 15th centuries. None of these people were directly related, but the similarities in their DNA suggested a recent common ancestor—a sign of a population bottleneck. The population apparently crashed around the ninth century, well after the islands were settled. “It was super drastic,” Fregel says. The apparent ninth century crisis coincides with the onset of warmer than average global temperatures, in a time known as the Medieval Warm Period. Fregel wonders whether the warmth might have brought climate change and famine. But ancient science.org SCIENCE

DNA from 34 people on more populous Tenerife and Gran Canaria shows no signs of a population collapse. “To me that’s the most interesting thing,” Fregel says. “We imagined that all the islands would be the same. We are realizing that each island had a different scenario.” THE PEOPLE OF all the islands do exhibit some

similar signs of hardship. In Santana’s basement lab on Gran Canaria, the skull of the unnamed young man from El Hierro rests on a table in a plastic tray alongside nine others. Many skulls bear small dents, some the size of nickels, others of silver dollars—the marks of blows. “The rate of interpersonal violence is very, very high,” Santana says.

On Gran Canaria, Alberto-Barroso and colleagues reported in 2018 that 27% of 347 adult skulls collected from burial caves show signs of trauma, usually long before the person died. Roughly one-third of male skulls were damaged, and nearly 20% of the female skulls. Most of the injuries were from something like a club or stone on the left side of the front of the skull, consistent with face-to-face fighting. The injury rate is far higher than in other ancient burial sites, including ones in the Iberian Peninsula, New Guinea, and the Solomon Islands, the researchers reported. Other islands are even more extreme, according to unpublished research by Hemmamuthé Goudiaby, an archaeologist

at Archaïos, an archaeological company. For instance, of 82 skulls from El Hierro, 50% of male skulls and 28% of female skulls show signs of trauma. “These populations that seem to have a higher incidence of interpersonal violence are also the ones that live on islands with less available resources,” Goudiaby notes. He and Santana suggest ritualized violence—orchestrated fights or duels— might have served as a way to address conflicts in communities where food was in short supply and there were few options to move away. It’s easy to begin to see the lives of ancient Canarians as ones of deprivation. But Morales cautions against applying modern sentiments to a different time and place, or

CREDITS: (PHOTO) ERNESTO MARTIN; (GRAPHIC) M. HERSHER/SCIENCE

Archaeologists have found more than 50 ancient granaries in the cliffs of Gran Canaria, some still containing seeds of barley, wheat, and figs.

Chronicle of a vanished people After the first settlers reached the Canary Islands, they were apparently mostly isolated from the wider world as well as from other islands of the archipelago for a millennium. Scientists are tracing how the Indigenous islanders sustained themselves and comparing the trajectory of societies on each island. Exploration

Single tombs

Graves

European arrival Conquest begins

Expedition by Mauritanian King Juba II lands on the archipelago, according to Roman author Pliny the Elder.

Some Gran Canarians are buried in individual open-air tombs of piled rocks.

Gran Canarians are increasingly buried in the ground.

Italian Lancelotto Malocello settles on the islands.

0

200

400

600

800

1200

1000

Spanish military expedition invades Lanzarote.

1400

Settlement

Granaries

El Hierro crisis

Largest art

Conquest complete

Domesticated plants and animals including barley, wheat, goats, and pigs signal the arrival of humans on larger islands.

First evidence of Gran Canarians storing grain in caves

Genetics suggest population crash on this island.

Geometric mural, perhaps a calendar, painted on a cave wall on Gran Canaria

Final defeat of Indigenous people on Tenerife

SCIENCE science.org

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

583

NE WS | F E AT U R E S

Geneticist Rosa Fregel collects a sample from the remains of a woman who lived on Gran Canaria between 600 and 900 C.E.

were told all of the Indigenous people disto viewing the Indigenous people as trapped. appeared,” Santana says. “It’s not true. We “I don’t agree with this view. I imagine the know that because of the DNA results.” lives of those people as profoundly linked In a string of papers beginning in the to the natural world, and especially to the early 2000s, Fregel and colleagues found plants,” he says. that, on average, between 15% and 20% of He stresses that early Canarian societies current islanders’ DNA comes from Indigeshowed remarkable adaptability. The first nous sources. Local newspapers trumpeted settlers almost certainly came from places the findings with pride, as they did the diswith metallurgy, yet, faced with a world covery by Morales and colleagues that towithout metal ore, they reinvented tools of day’s Canarian barley is genetically similar stone, wood, and bone. And they enjoyed to the ancient island variety. culturally rich lives, as shown by intricately But Fregel is wary of using genetics to woven tapestries, clay figurines, stone etchconstruct identity. She notes that her surings, and—on the wall of at least one Gran vey of male Y chromosomes, which turned Canaria cave—elaborate geometric paintup less than 10% Indigenous DNA, received ings that may represent a calendar system. a chillier reception. The dominance of EuThen, the Europeans arrived. The vioropean genetics in the male chromosomes lence of traditional Canarian life pales in comparison. The first known contact happened in the early 1300s, when Italian navigator Lancelotto Malocello settled on the island now known as Lanzarote. In 1402, soldiers of Spain’s Castilian monarchy landed on Lanzarote in force, beginning a century of conquest that ended in 1496 with Spanish victory over Indigenous warriors on Tenerife. The Indigenous people were killed, enslaved, forcibly assimilated, or deported, save a scattered few. On Gran Canaria, an Indigenous population estimated at between 10,000 and 60,000 was slashed to as few as 2000. The genocide was almost complete. No Indigenous Canarian community survives today. Only scraps of language, such as names for places, foods, Barley grown today on Gran Canaria is related to seeds or famous leaders, remain. And yet, that Jacob Morales found in ancient granaries on the island. traces of the early settlers live on. “We 584

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

suggests most Indigenous men were killed or kept at the margins of society and left fewer offspring, whereas women were more likely to have children through intermarriage and sexual violence. “Your genome is telling you a story about your ancestors. But your identity is defined by your experience, your family, your friends,” says Fregel, who was born and raised on Tenerife. When she tested her DNA, it was about 20% Indigenous. “I wouldn’t feel any less Canarian if my DNA was from a whole other region.” Still, her findings and those from Santana and Morales have bolstered a modern-day Canarian identity that emerged as the country moved beyond the nationalist conformity enforced under Spanish dictator Francisco Franco, who died in 1975. A year later, when Morales was born, his uncle paid his mother the equivalent of 60 euros to give Morales a middle name evoking an ancient Canarian leader: Bentejui. Morales says his work can help “build memory, and thus identity, for the people who had very little historical documentation.” In one example, Canarian chefs and beermakers have reached out to the scientists about using barley with DNA traced to the ancient Canarian stock. “So much knowledge was really lost when the Spanish conquered the islands,” says Jenny Hagenblad, a plant evolution geneticist at Linköping University who works with Morales. “It’s pretty cool that we can break through this loss of information and we can, in a sense, eat the same things. It is a very living link to the past.” j science.org SCIENCE

INSIGHTS PERSPECTIVES

10 µm

MATERIALS SCIENCE

Stabilizing 3D-printed metal alloys A design strategy overcomes the strength-ductility trade-off in alloy manufacturing By Lai-Chang Zhang1 and Jincheng Wang1,2

T

hree-dimensional (3D) printing, also known as additive manufacturing, continues to reshape industries, including metals production. Among its advantages are decreasing the time and costs for creating intricate metal parts and increasing customization. However, the technology still faces challenges in achieving uniform mechanical properties in 3D-printed metallic alloys. On page 639 of this issue, 586

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

Zhang et al. (1) report a design strategy for printing a robust titanium alloy. The authors show that the addition of molybdenum (Mo) to the powder metal mixture enhances phase stability and improves the strength, ductility, and uniformity of tensile properties of the 3D-printed alloy. The approach could potentially be applied to other powder mixtures and enable the customization of different alloys with enhanced properties. In the layer-by-layer 3D printing process (typically with a high cooling rate of ~103 to

108 K/s), a substantial thermal gradient forms near the edge and bottom of the melt pool, where metal powder has been melted by a laser beam. The thermal gradient induces epitaxial grain growth along the interface between the freshly melted material and the underlying solid material, with grains growing toward the melt pool center. Cycles of heating and partial remelting during the printing of multiple layers ultimately results in the formation of large columnar grains and heterogeneously distributed phases, science.org SCIENCE

A titanium alloy powder decorated with molybdenum particles in this scanning electron microscope image exhibits uniform tensile properties when 3D printed.

both of which are undesirable because they can lead to nonuniform (anisotropic) and compromised mechanical properties (2, 3). Titanium alloys are among the strongest metallic materials. In engineering applications at ambient temperature, a suitable titanium alloy typically exhibits a tensile elongation (the maximum stretch or deformation that a material can withstand before breaking) ranging from ~10 to ~25%, which reflects good material reliability. Although greater elongation (ductility) facilitates easier formability and holds priority in certain applications, an increased strength within this elongation range is preferred for enduring mechanical loads. In both conventional and additive manufacturing techniques for processing metallic materials, a trade-off between strength and ductility has been prevalent (4, 5). There are several strategies generally used to improve the strength and ductility of 3D-printed alloys. These include optimizing alloy design, process control, fine-grain boundary strength, and grain microstructure modification. They also encompass inhibiting undesirable (brittle) phases, introducing second phases, and adding postprocessing treatments. Recent research that addresses the problems of columnar crystals and undesirable phases has focused on the in situ incorporation of elements to modify microstructures and phase compositions. This approach also promotes the formation of equiaxed crystals—structures with grains of roughly equal dimensions along longitudinal and transverse axes. This distinctive in situ alloying ability of 3D printing offers a promising avenue to overcome the enduring compromise between strength and ductility, particularly in techniques such as powder bed fusion and directed energy deposition (5–8). The powder bed fusion process uses thermal power to selectively fuse regions of a metal powder applied as a thin layer on a platform. The directed energy deposition method involves the coaxial feed of metal powder by inert gas from a nozzle while simultaneously undergoing melting by a focused energy source before being deposited onto a surface. There has been exploration of grain morphologies and mechanical properties when diverse elements are added to alloys for 3D printing (7–9). For example, nanoc1

Centre for Advanced Materials and Manufacturing, School of Engineering, Edith Cowan University, Perth, WA, Australia. 2School of Engineering, The University of Western Australia, Perth, WA, Australia. Email: [email protected]; [email protected] SCIENCE science.org

eramic zirconium hydride particles that were incorporated into nonprintable aluminum alloys yielded materials that were printable and noncracking, with a refined equiaxed grain microstructure and tensile properties comparable to those of wrought counterparts (10). However, for titanium alloys, commercially available grain refiners often have limited effectiveness on grain structure. The refinement mechanism of titanium alloys, especially the columnar-toequiaxed transition during solidification in 3D printing, has been studied extensively, but efficiency limitations persist. Attempts to overcome this obstacle include variations in processing parameters (5), high-intensity ultrasound application (11), the introduction of a desired heterostructure through alloy design (7), the addition of solutes as

“The approach could…enable the customization of different alloys with enhanced properties.” grain refiners for heterogeneous nucleation sites (6), and the incorporation of solutes with a high supercooling capacity, such as b-eutectoid stabilizer elements [copper (Cu), iron (Fe), chromium (Cr), cobalt (Co), and nickel (Ni)], which have limited solubility in titanium (12). Instead of using b-eutectoid stabilizer elements, which may lead to the formation of brittle intermetallic eutectoids in titanium alloys, Zhang et al. used Mo from the b-isomorphous family [which includes niobium (Nb), tantalum (Ta), and vanadium (V)] for in situ alloying of a titaniumaluminum-molybdenum-vanadium-chromium alloy (Ti-5Al-5Mo-5V-3Cr+5Mo). In situ alloying precisely delivered Mo into the melt pool, acting as seed nuclei for crystal formation and refinement during each layer scan. The Mo additive facilitated the transition from large columnar grains to fine equiaxed and narrow columnar grain structures. Mo also stabilized the desired b phase and inhibited the formation of phase heterogeneities during thermal cycling. The strategy yielded exceptional uniformity with a concurrent ~926-MPa yield strength and ~26% ductility. Earlier studies described the use of additive elements that were distinct from the constituents of the parent prealloyed powder. However, in the case of Ti-5Al-5Mo5V-3Cr+xMo (where x = 5 wt %), Zhang et al. added extra Mo to the parent powder, achieving a total Mo weight percentage of 10%. This minimized chemical complexity, achieved columnar-to-equiaxed transition

in grain morphology, enhanced desired phase stability, and achieved isotropy in the printed metal. Notably, using the prealloyed powder alone with the same composition did not yield comparable printing results. This is unsurprising given earlier observations. For example, Ti-Nb alloys processed through in situ alloying yielded distinct microstructures and properties compared with using prealloyed powder alone (13). Similarly, a 3D-printed alloy with prealloyed Ti-6Al-4V-5Cu powder showed acicular a´ grains within the columnar b grains, resulting in deteriorated ductility (14). Zhang et al. did, however, identify undissolved Mo particles in the microstructure, and their potential effects are unknown. Indeed, the random presence of undissolved particles in the in situ alloying strategy has raised concerns related to mechanical and corrosion properties (15). For example, the full melting of in situ alloyed additive particles may require higher energy, and overheating could lead to an altered microstructure and poorer mechanical properties. Also, the dynamic fatigue and corrosion performance resulting from undissolved additive Mo particles are unknown. Although postprinting heat treatment can eliminate undissolved particles, it might alter the microstructure, potentially jeopardizing the mechanical properties (9). Zhang et al.’s proposed design strategy opens an avenue for exploring different metal powder feedstocks, diverse printable alloy systems, and different 3D printing techniques as well as for advancing multimaterial printing. It also enables the inhibition of columnar grain formation and prevents undesirable phase heterogeneities. These problems arise as the result of varied heat profiles, which are influenced by printing parameters for each powder. The strategy also overcomes the strength-ductility trade-off in the as-printed state, minimizing the need for postprinting treatment. These advantages will undoubtedly trigger ripples in the realm of 3D printing. j RE FE REN CES A ND N OT ES

1. J. Zhang et al., Science 383, 639 (2024). 2. J. Wang, R. Zhu, Y. Liu, L. Zhang, Adv. Powder Mater. 2, 100137 (2023). 3. D. Gu et al., Science 372, eabg1487 (2021). 4. T. Yang et al., Science 362, 933 (2018). 5. H. Y. Ma et al., J. Mater. Sci. Technol. 183, 32 (2024). 6. B. Vrancken, L. Thijs, J.-P. Kruth, J. Van Humbeeck, Acta Mater. 68, 150 (2014). 7. T. Zhang et al., Science 374, 478 (2021). 8. P. Kürnsteiner et al., Nature 582, 515 (2020). 9. J. Wang et al., J. Mater. Sci. Technol. 61, 221 (2021). 10. J. H. Martin et al., Nature 549, 365 (2017). 11. C. J. Todaro et al., Nat. Commun. 11, 142 (2020). 12. M. S. K. K. Y. Nartu et al., Nat. Commun. 14, 3288 (2023). 13. J. C. Wang et al., J. Mater. Sci. Technol. 105, 1 (2022). 14. K. Li et al., Acta Mater. 256, 119112 (2023). 15. J. C. Wang et al., Mater. Sci. Eng. A 760, 214 (2019). 10.1126/science.adn6566 9 FEBRUARY 2024 • VOL 383 ISSUE 6683

587

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

PHYSIOLOGY

Fibroblasts enable penile erection Perivascular fibroblasts may underlie erectile dysfunction By Ji-Kan Ryu1 and Gou Young Koh2

S

exual well-being in men depends on the ability to attain penile erection, which can be compromised by aging and chronic conditions, including diabetes and atherosclerosis (1). Penile erection results from acetylcholine- and nitric oxide–induced dilation of the supplying arteries and sponge-like corpora cavernosa tissue and relaxation of smooth muscle (2). The extent and duration of penile erection is determined by the balance between these vasodilators and the vasoconstrictor norepinephrine (3, 4). On page 604 of this issue, Linck Guimaraes et al. (5) report that perivascular fibroblasts expressing the high-affinity amino acid transporter SLC1A3 (solute carrier family 1, member 3) are essential for penile erection in mice because they reduce norepinephrine availability, thereby promoting dilation of the corpora cavernosa. The number of SLC1A3+ perivascular fibroblasts decreased in aged mice, which reduced penile blood flow. Because the mechanisms of penile erection are similar in mice and humans (4), these findings may be relevant to erectile dysfunction in aged men. Linck Guimaraes et al. report that in their transgenic mouse model, SLC1A3+ fibroblast proliferation increased by inhibiting Notch signaling during recurrent erection, which led, in extreme cases, to ischemic priapism (a condition characterized by persistent and painful erection with high rigidity of the corpora cavernosa). By contrast, continuous Notch activation reduced SLC1A3+ fibroblast numbers and penile blood flow. These fibroblasts are mainly distributed along blood vessels and smooth muscles in the corpora cavernosa. The authors show that SLC1A3+ fibroblasts take up norepinephrine through SLC6A2 and degrade it through the activity of intercellular monoamine oxidase A, which reduces norepinephrine availability. Most treatments for erectile dysfunction promote nitric oxide–mediated relaxation of vascular smooth muscle in the corpora cavernosa (6). Sildenafil, tadalafil, vardenafil, and other inhibitors of phosphodiesterase type 5 (PDE5) augment the smooth muscle–

1

Department of Urology and Medical Research Center for Controlling Intercellular Communications, Inha University School of Medicine, Incheon, Republic of Korea. 2Center for Vascular Research, Institute for Basic Science, Daejeon, Republic of Korea. Email: [email protected]

588

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

relaxing effects of nitric oxide by blocking the degradation of the downstream signaling messenger cyclic guanosine monophosphate (cGMP) into GMP (6, 7). However, despite their widespread use, PDE5 inhibitors fail to restore an erection in up to 30% of men, leaving limited treatment options (7). In search of effective alternatives, clinical trials using melanocortin or dopamine receptor agonists to target central pathways or soluble guanylate cyclase activators, Rhokinase inhibitors, or potassium channel openers to target peripheral pathways have not shown sufficient benefit to gain approval for this use. This failure is largely due to lack of greater efficacy or fewer side effects than already approved PDE5 inhibitors (8). Although Linck Guimaraes et al. did not examine humans, their study reveals a new therapeutic paradigm of creating conditions that increase norepinephrine uptake or decrease Notch signaling in penile perivascular fibroblasts, which could

translate into treating erectile dysfunction in patients who are unresponsive to PDE5 inhibitors. Thus, multiple strategies could be explored by further preclinical testing and, if promising, evaluated clinically. One approach would be to inhibit Notch signaling in SLC1A3+ fibroblasts of the corpora cavernosa (see the figure) with available Notch inhibitors. A potential limiting factor is that local administration of such inhibitors would also affect endothelial cells, smooth muscle cells, neurons, and other cell types in which Notch signaling is involved in penile erection but differs mechanistically from Notch signaling in fibroblasts. Other approaches could include increasing expression of the norepinephrine transporter SLC6A2 in SLC1A3+ perivascular fibroblasts of the corpora cavernosa. However, the poorly understood regulatory mechanisms of SLC6A2 expression must be characterized. Another approach would be to implant autologous SLC1A3+ fibroblasts in the corpora cavernosa. However, still unknown is whether the number of implanted SLC1A3+ fibroblasts that become associated with the corpora cavernosa would be adequate to have sustained SLC6A2 norepinephrine uptake activity and be functionally sufficient to correct erectile dysfunction. Furthermore, a neural or mechanical stimulator that

New therapeutic prospects for intractable erectile dysfunction Fibroblasts expressing solute carrier family 1, member 3 (SLC1A3+) take up norepinephrine through the SLC6A2 transporter, which results in its degradation and promotes penile erection in mice. Thus, various potential approaches for treating refractory erectile dysfunction in humans are revealed. 1 Inducing SLC6A2 expression

Cross section of human penis

Corpora cavernosa SLC6A2 transporter

Cavernous artery

2 Notch inhibition Corpus spongiosum Proliferation

Urethra Inhibited Notch

3 Implantation of autologous SLC1A3+ fibroblasts Vascular smooth muscle

4 Mechanical stimulation Endothelial cell SLC1A3+ fibroblast

Proliferation

science.org SCIENCE

expands the number of perivascular SLC1A3+ fibroblasts by increasing erection frequency could be developed. Efficacy would depend on determining the yet-to-be-identified stimulation parameters for promoting the proliferation of SLC1A3+ fibroblasts that express SLC6A2. These new approaches will require rigorous preclinical and clinical testing to translate observations made in transgenic mice into therapies that are safe and effective in men. Additional studies would also be necessary to characterize the heterogeneity of corpora cavernosa fibroblasts in diverse patients with erectile dysfunction, where penile fibrosis, endothelial cell injury, smooth muscle transdifferentiation, and other abnormalities can be contributing factors (9, 10). These challenges are, nonetheless, balanced by the drawbacks of available alternatives that involve invasive procedures such as injection of vasodilators into the corpora cavernosa or surgical implantation of a penile prosthesis. Possible new interventions should not overshadow evidence that regular physical activity can improve erectile function through increased expression and activity of nitric oxide synthase, strengthened endothelial cell function, and reduced stress and anxiety (11). Exercise and lifestyle changes could increase the number of perivascular SLC1A3+ fibroblasts and their expression of SLC6A2. These changes are a natural consequence of frequent penile erections but are suppressed when erections are infrequent. Although erections could be necessary to maintain perivascular SLC1A3+ fibroblasts, it is unclear whether this requires regular sexual activity or if nocturnal penile tumescence is sufficient. If sedentary behavior and low sexual activity leads to fewer of these fibroblasts and greater corpora cavernosa sensitivity to norepinephrine, then regular exercise, sexual activity, and other lifestyle changes should be helpful for sexual dysfunction. j RE F ER E NC ES AND NOTES

1. I. Goldstein et al., Int. J. Clin. Pract. 72, e13078 (2018). 2. K. J. Hurt et al., Proc. Natl. Acad. Sci. U.S.A. 99, 4061 (2002). 3. S. M. MacDonald, A. L. Burnett, Urol. Clin. North Am. 48, 513 (2021). 4. K. E. Andersson, Pharmacol. Rev. 63, 811 (2011). 5. E. Linck Guimaraes et al., Science 383, eade8064 (2024). 6. I. Goldstein et al., N. Engl. J. Med. 338, 1397 (1998). 7. A. Samidurai et al., Annu. Rev. Pharmacol. Toxicol. 63, 585 (2023). 8. J. K. Ryu et al., Transl. Androl. Urol. 6, 207 (2017). 9. L. Zhao et al., Nat. Commun. 13, 4302 (2022). 10. D. Fang et al., Front. Endocrinol. 13, 874915 (2022). 11. M. S. Allen, Nat. Rev. Urol. 16, 553 (2019). AC KN OW LE D GMENTS

The authors receive support from the Republic of Korea Ministry of Science and Technology to the Medical Research Center (NRF-2021R1A5A2031612 to J.-K.R.) and the Institute of Basic Science (IBS-R025-D1-2015 to G.Y.K.). 10.1126/science.adn5182 SCIENCE science.org

PLANT SCIENCE

Time for growth Plants measure the duration of metabolic activity to promote rapid growth in long days By Christopher R. Buckley and Michael J. Haydon

D

aylength varies across the year and provides a cue for plants to respond to changing seasons. Photoperiodism, the ability to measure daylength, is critical for fitness and survival because of its influence on plant physiology and behavior. The timing of numerous developmental processes is controlled by daylength, including flowering, growth, dormancy, and senescence (1). Thus, photoperiodism has important implications for the regional adaptation of crops and migration of plant populations in response to climate change (2). Seasonal control of flowering occurs by a well-defined molecular pathway that measures photoperiod as the duration of light, as detected by photoreceptors. On page 605 of this issue, Wang et al. (3) describe a new mechanism of daylength measurement that detects the duration of photosynthetic output and is required for rapid growth during long summer days. The study demonstrates that plants use multiple, independent mechanisms to measure daylength that can differentially influence distinct developmental processes. The transcription factor CONSTANS (CO) plays a key role in photoperiodic flowering. It activates the expression of FLOWERING LOCUS T (FT), which encodes a mobile protein that promotes flowering in the shoot apical meristem (1). Daily oscillations of CO transcription are controlled by the circadian clock, a biological timekeeper that is set by light signals at dawn in a process called entrainment. In Arabidopsis thaliana, flowering is induced in long days because CO protein is stabilized by light in the afternoon. By contrast, in flowering plants such as rice, CO activity is promoted by a short photoperiod. The CO-FT module also controls other photoperiodic processes, including tuberization in potatoes, bud dormancy in perennials, and seed size in annuals (4–6). However, shifting daylength leads to substantial changes in the profile of transcribed genes, indicating that there might be many more processes under photoperiodic control (7). Because relatively School of BioSciences, University of Melbourne, Parkville, VIC, Australia. Email: [email protected]

few genes are regulated by CO, this suggests that the other genes are controlled by COindependent mechanisms. Metabolic signals have been implicated in biological timekeeping in several ways (8). Sugars produced from photosynthesis contribute to entrainment of the Arabidopsis circadian clock in a manner different from light, acting as a “metabolic dawn” (9). Snf1 RELATED PROTEIN KINASE 1 (SnRK1) and TARGET OF RAPAMYCIN (TOR) are major metabolic signaling complexes in plants that influence the pace of circadian rhythms (10, 11), and metabolic signals such as reactive oxygen species modify circadian gene expression in a time-of-day–specific manner (12, 13). Metabolic signals also measure and communicate changes in daylength. A metabolic daylength measuring system has been described that is specifically required to control growth in the short days of winter (14). Expression of the circadian-regulated gene PHLOEM PROTEIN 2-A13 (PP2-A13) is induced after dusk in short days and is required to maintain winter growth but has no effect on growth in long days. The daylength-dependent regulation of PP2-A13 expression is controlled by the timing of photosynthetic activity. Supplying sugar can suppress its expression in short days, whereas inhibiting photosynthesis can activate its expression in long days. Wang et al. describe a distinct mechanism of metabolic daylength measurement that operates in long days and is required to promote rapid growth in these conditions. In Arabidopsis, this mechanism functions in parallel to, but independently of, canonical measurement of absolute daylength by photoreceptors and the CO-FT module (see the figure). MYO-INOSITOL1-PHOSPHATE SYNTHASE 1 (MIPS1) has been identified as a necessary functional component of this pathway. MIPS1 expression is induced specifically during long days and, opposite to pp2-a13 mutants, mips1-2 mutants grow slowly in long days but are unaffected in short days. By modifying the timing, duration, and intensity of light, Wang et al. observed that MIPS1 expression and rapid growth depend on photosynthetic activity in the afternoons of long summer days. Because photoreceptors are activated at much lower light quantity 9 FEBRUARY 2024 • VOL 383 ISSUE 6683

589

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

than is required for net carbon fixation by photosynthesis, plants simultaneously detect a different daylength by these two mechanisms. Thus, the authors propose the concept of a photosynthetic period. How the photosynthetic period is measured is unclear but could involve either metabolic signaling mechanisms that are already implicated in timekeeping or otherwise unknown pathways (8). Moreover, it is not clear whether the mechanisms that control MIPS1 and PP2-A13 expression are the same or distinct. The enzymatic function of MIPS1 is required for the metabolic daylength measuring system because myoinositol can rescue the mutant phenotype but is not sufficient because alone it cannot induce rapid growth in short days (3). Although other factors are apparently required, the discovery of MIPS1 as a marker and essential component of metabolic daylength measurement presents an exciting opportunity to identify upstream and downstream components of the pathway. The existence of parallel photoperiodic pathways that detect different signals raises interesting questions about how processes are regulated in natural conditions, which can be highly variable and unpredictable. Furthermore, it now seems likely that there are numerous mechanisms that could measure different aspects of daylength. Because photoperiodic flowering and seasonal growth are controlled by independent pathways, this opens up the possibility of specifically targeting photoperiodic growth for crop improvement. Although long days induce both flowering and growth in Arabidopsis and temperate crops such as wheat and barley, tropical crops such as rice and soybean preferentially flower in short days. The existence of independent pathways could facilitate different photoperiod

responsiveness for these processes in different plant species. Photoperiod-insensitive flowering has been frequently selected for during the domestication and expansion of crops into different spatial and temporal environments (15). It therefore should be possible to engineer photoperiod-insensitive growth in crops, which could be a useful trait for rapid biomass production in short growing seasons, especially as climate change shifts present environmental optima (2). Furthermore, defining critical time windows to trigger photoperiodic growth could enable energy-efficient regimes to maximize growth in artificial cropping environments. Therefore, future progress in understanding metabolic daylength measurement in crops presents opportunities for agriculture and aligns with the emerging concept of chronoculture (15). j R E F E R E N C ES A N D N OT ES

1. J. M. Gendron, D. Staiger, Annu. Rev. Plant Biol. 74, 481 (2023). 2. L. L. Sloat et al., Nat. Commun. 11, 1243 (2020). 3. Q. Wang, W. Liu, C. C. Leung, D. A. Tarté, J. M. Gendron, Science 383, eadg9196 (2024). 4. C. Navarro et al., Nature 478, 119 (2011). 5. H. Böhlenius et al., Science 312, 1040 (2006). 6 B. Yu et al., Nat. Plants 9, 343 (2023). 7. C. C. Leung et al., PLOS Biol. 21, e3002283 (2023). 8. C. R. Buckley et al., Curr. Opin. Plant Biol. 73, 102333 (2023). 9. M. J. Haydon et al., Nature 502, 689 (2013). 10. A. Frank et al., Curr. Biol. 28, 2597 (2018). 11. N. Zhang et al., Proc. Natl. Acad. Sci. U.S.A. 116, 25395 (2019). 12. A. G. Lai et al., Proc. Natl. Acad. Sci. U.S.A. 109, 17129 (2012). 13. Á. Román et al., Proc. Natl. Acad. Sci. U.S.A. 118, e2020646118 (2021). 14. W. Liu et al., Dev. Cell 56, 2501 (2021). 15. G. Steed et al., Science 372, eabc9141 (2021). AC K N OW L E D G M E N TS

M.J.H. receives research funding from the Grains Research and Development Corporation. 10.1126/science.adn5189

Multiple parallel daylength measurement systems in plants Absolute photoperiod is the time between dawn and dusk, is detected by photoreceptors, and controls timing of flowering. Photosynthetic period is the duration of light required to drive net carbon fixation, does not precisely overlap with absolute photoperiod, and controls seasonal growth. Dawn

Absolute photoperiod

Dusk

CANCER

Transforming lung cancer types Lung cancer cells can escape targeted therapy by switching oncogenic drivers and cell identity By Anton Berns

T

here are two main classes of lung cancer: non–small-cell lung cancer, which is predominantly the lung adenocarcinoma (LUAD) subtype that originates from alveolar type 2 (AT2) cells, and small cell lung cancer (SCLC), for which pulmonary neuroendocrine cells (PNECs) are the main cell of origin. SCLC has the poorest prognosis, killing over 200,000 patients each year worldwide. Remarkably, ~10% of patients that present with LUAD driven by mutated epidermal growth factor receptor (EGFR) and receiving EGFR targeted therapy become refractory to the treatment through histological transformation (HT) into SCLC (1). A similar conversion to neuroendocrine tumors is occasionally observed in patients with prostate adenocarcinoma receiving androgen deprivation therapy (2). The drivers of these HTs and the transition stages involved have remained an enigma. On page 603 of this issue, Gardner et al. (3) identified the consecutive stages LUAD cells go through in HT to SCLC in mice, which helps to uncover this difficult-tostudy conversion in human cancer. Gardner et al. generated ERPMT mouse strains with various configurations of conditional alleles of the retinoblastoma Rb (R) and Trp53 (P) genes, which are both implicated in nearly all SCLCs, with an oncogenic MycThr58 Ala (MycT58A; M) and a doxycyclineinducible EGFR-Leu858 Arg (EGFRL858R; E) oncogene, supplemented with a cell-trace (T) marker. Cell-type–specific Cre drivers were used to switch on these alleles in the various lung cell lineages, enabling the authors to assess the contribution of these alleles to LUAD and SCLC as well as their involvement in HT from LUAD to SCLC (T-SCLC). Single-cell RNA sequencing was used to characterize the

¢

Cloud cover Light intensity

Photosynthetic period

¢

Division of Molecular Genetics, Netherlands Cancer Institute, Amsterdam, Netherlands. Email: [email protected]

590

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

science.org SCIENCE

GRAPHIC: A. FISHER/SCIENCE

cell types emerging during this process (see the figure). It is well recognized that a limited set of oncogenic drivers are usually associated with particular tumors. This is also the case for transformation of AT2 cells into LUAD and PNECs into SCLC. However, Gardner et al. now show that AT2 cells and PNECs do not tolerate particular oncogenic drivers and neither do the neuroendocrine tumors that are derived from them. The HT from LUAD to T-SCLC is notable in this respect: This transformation not only necessitates the acquisition of new oncogenic driver lesions (e.g., loss of the tumor suppressors RB and p53, gain of MYC hyperactivation) that are largely absent in LUAD, it also requires the downregulation of an oncogenic driver that is indispensable for LUAD (mutant EGFR). The authors found that the underlying lineage switch that accompanies this HT appears to proceed through a bottleneck consisting of phenotypically different cell types including undifferentiated basal cell–like cells expressing target genes of the transcription factors MYC and SOX2, reminiscent of epigenetic reprogramming. Apparently, dedifferentiation to a progenitor basal cell–like intermediate is required for this HT. This aligns well with the observation of Gardner et al. that keratin 5 (KRT5)–expressing lung basal cells—which already exhibit progenitor features—can be easily transformed into T-SCLC by inactivation of RB and p53. The sensitivity of a cell lineage to specific oncogenic drivers is not absolute but depends on the state the cell is in. For example, AT2 cells do not tolerate expression of MYCT58A without concomitant EGFRL858R or phosphatidylinositol 3-kinase (PI3K) pathway activation [such as loss of phosphatase and tensin homolog (Pten)], whereas PNECs and even SCLCs do not tolerate EGFRL858R either with or without MYCT58A. This lineage-specific intolerance is a widely observed phenomenon. To drive tumorigenesis, both the nature and flux (strength) of oncogenic signaling have to be in tune with the specific gene expression configuration of a cell. Otherwise, oncogenic signals might become toxic to cells, e.g., by inducing replication stress. High MYC expression is not only toxic to AT2 cells, it is also detrimental for mouse embryo fibroblasts in vitro, causing apoptosis (4). This can be relieved by coexpression of oncogenic (mutant) KRAS, similar to the blunting of MYCT58A toxicity by EGFRL858R or Pten loss in AT2 cells. Although MYCT58A expression appears toxic to AT2 cells, EGFRL858Rinduced LUAD from AT2 cells is accelerated by MYCT58A. Indeed, it appears that LUAD development even requires MYC and can be eliminated by repressing MYC’s function (5), consistent with the observation that onSCIENCE science.org

Lung cancer cell lineages Cell tolerance to specific oncogenic drivers controls the conversion of lung adenocarcinoma (LUAD) to small cell lung cancer (SCLC). In mice, multiple cell types can give rise to tumors that resemble SCLC (T-SCLC). Loss of tumor suppressors Rb and Trp53 (RP) is required, but strong EGFR signaling is incompatible with SCLC development, independent of the cell of origin. Oncogenic MYC is a potent driver of T-SCLC development. PI3K signaling (loss of Pten) and loss of Rb facilitate AT2 cell transition to T-SCLC and to tumors with basal cell–like features. Oncogenic MYC without EGFR or PI3K signaling is not tolerated by AT2 cells. EGFR signaling blocks T-SCLC development from both AT2 and pulmonary endocrine cells. Cell of origin Toxic

AT2 cell

LUAD MYC

MYC

MYC

–EGFR

+EGFR

–EGFR MYC –RP

MYC –Pten Basal cell–like

Lung cancer type

Residual disease

+EGFR MYC

Pulmonary neuroendocrine cell

Cancer

LUAD

T-SCLC

Basal cell–like MYC –RP –Pten

T-SCLC + LCNEC

? (MYC) –RP

SCLC

AT2, alveolar type 2; EGFR, epidermal growth factor receptor; LCNEC, large cell neuroendocrine carcinoma; PI3K, phosphatidylinositol 3-kinase; Pten, phosphatase and tensin homolog; Rb, retinoblastoma

cogenic KRAS-induced LUAD development is impaired by Myc haploinsufficiency (6). Additionally, AT2 cells do not tolerate well oncogenic KRAS signaling on its own. The activation of oncogenic KRAS in AT2 cells in vivo leads to a widespread senescence-like arrest with only a small fraction of AT2 cells that are able to express oncogenic KRAS and develop into LUADs in mice (7). The intolerance of PNECs to the expression of EGFRL858R is even more striking because Gardner et al. found that its expression in ERPMT mice blocks SCLC development from PNECs. The presence of EGFRL858R is also unable to force the reverse HT from SCLC to LUAD. This aligns with previous work showing that established mouse SCLC cell lines respond to forced oncogenic KRAS expression by entering into crisis with the spurious emergence of cells with a LUAD-like phenotype (8). In this context, the observation that inhibition of the mitogen-activated protein kinase (MAPK) pathway, which is downstream of KRAS, with trametinib does not relieve the inhibitory effect of EGFRL858R expression on PNECs is unexpected. This suggests that critical EGFR signaling beyond the MAPK pathway is not compatible with the PNEC lineage. It is unlikely that this entails PI3K signaling because its activation accelerates SCLC development. The observation that fibroblast growth factor (FGF), also a strong inducer of MAPK signaling, enhances SCLC growth (9) and potentiates its metastatic dissemination (10), as well as facilitating the HT of LUAD to SCLC (11), adds to the confusion, although this could also relate to signaling flux. However, an oncogenic FGF receptor 2 (FGFR2) mu-

tant impaired SCLC development induced by loss of RB and p53 in PNECs, whereas it promoted development of T-SCLC when targeted to KRT14+ bronchial epithelial cells (12). These seemingly discordant observations might be reconciled by postulating that different cells of origin, as a result of a different epigenome, require or tolerate dissimilar oncogenic driver configurations for T-SCLC development. This could also help explain the actual diversity seen in lung neuroendocrine tumors (13) with their diverse proliferation and metastatic characteristics (14). This can be tested experimentally using the approaches described by Gardner et al. The predominant driver role of MYC in SCLC and T-SCLC, as well as the dependence of LUADs on MYC, again emphasizes the promise of effective MYC inhibitors for treating lung tumors and other cancers (5, 15). The study by Gardner et al. is in many respects highly illuminating, showing how thorough mouse studies can reveal new cancer biology. j RE F E REN C ES AN D N OT ES

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

S. Sivakumar et al., Cancer Discov. 13, 1572 (2023). J. O. Mori et al., Prostate 82, 1005 (2022). E. Gardner et al., Science 383, eadj1415 (2024). G. I. Evan et al., Cell 69, 119 (1992). L. Soucek et al., Nature 455, 679 (2008). N. M. Sodir et al., Nat. Commun. 13, 6782 (2022). C. Guerra et al., Cancer Cell 4, 111 (2003). J. Calbo et al., Cancer Cell 19, 244 (2011). D. W. Shia et al., Oncogene 42, 434 (2023). M. C. Kwon et al., Genes Dev. 29, 1587 (2015). K. Ishioka et al., Cancer Res. 81, 3916 (2021). G. Ferone et al., Cell Rep. 30, 3837 (2020). J. W. Park et al., Science 362, 91 (2018). D. Yang et al., Cancer Discov. 10, 1317 (2018). T. E. Speltz et al., Nat. Biotechnol. 41, 541 (2023). 10.1126/science.adn5218 9 FEBRUARY 2024 • VOL 383 ISSUE 6683

591

I NS I GHTS

P OLICY FORUM EDUCATION

Teach Indigenous knowledge alongside science Evidence supports the teaching of Indigenous knowledge alongside sciences in the classroom By Amanda Black1,2 and Jason M. Tylianakis2,3

C

onflict has grown around Indigenous knowledge in education policy. There has been growing acceptance of the value of Indigenous knowledge for promoting ecological resilience, transformational approaches in stewardship, and cultural renewal within global fora such as the Intergovernmental Panel on Climate Change. However, despite increasing acceptance at a strategic high level in science-informed policy, there is often a lack of wider acceptance, application, and policy protections of Indigenous knowledge transmission in more local settings, including opposition by some scientists. We argue that Indigenous knowledge can complement and enhance science teachings, benefitting students and society in a time of considerable global challenges. We do not argue that Indigenous knowledge should usurp the role of, or be called, science. But to step from “not science” to “therefore not as (or at all) valuable and worthy of learning” is a non sequitur, based on personal values and not a scientifically defensible position. The current state of global systems in an uncertain risk landscape creates an urgent need for many knowledges and approaches to build resilience and prosperity of communities. One attempt to provide policy protections and opportunities for Indigenous knowledge is the Aotearoa–New Zealand government’s decision to ensure that Indigenous knowledge (Ma¯tauranga Ma¯ori) has equal value with other bodies of knowledge in the school curriculum, after lengthy advocacy from Ma¯ori educators to honor the Treaty of Waitangi, Aotearoa– New Zealand’s founding document. This policy has precipitated a battle of rhetoric among researchers and scientists; initial condemnation of the policy by a group of academics argued that unlike science, Indigenous knowledge is inadequately equipped to provide empirical evidence of universal truths (1), which resulted in

1Faulty of Agriculture and Life Sciences, Lincoln University,

Lincoln, New Zealand. 2Bioprotection Aotearoa, Canterbury, New Zealand. 3School of Biological Sciences, University of Canterbury, Christchurch, New Zealand. Email: [email protected]

592

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

a signed open letter by 2000 academics and public figures in support of the policy that includes Ma¯tauranga Ma¯ori. Since then, the Royal Society of New Zealand–Te Apa¯rangi (the premiere advocacy and advisory body for science and humanities) has been drawn into the debate, which continues from both sides. Those that support this policy have largely argued on the grounds of ethical responsibilities and moral viewpoints, whereas those that oppose it cite dilution or, at worst, abandonment of science that will lead to poor societal outcomes (1). The considerable research effort on innovation, Indigenous knowledge’s relationship with science, and its pedagogy have not (to our knowledge) been synthesized to address this discussion, which is also pertinent to efforts beyond New Zealand [such as recent investment into Indigenous knowledge by the US National Science Foundation (2)]. We suggest that many of the arguments used to “defend” science by presenting Indigenous knowledge as inferior are themselves rooted in logical fallacies. We also argue that the treatment of all Indigenous knowledge as myth is at odds with the literature, which emphasizes a continuum from empirical and science-like aspects of Indigenous knowledge to philosophical and metaphysical ones (3). Teaching sociocultural themes of a Ma¯ori worldview is already encouraged in curriculum guidelines, suggesting that objections are not to having these aspects taught at school but rather to giving them value in the context of knowledge. Yet school curricula already include a range of subjects across the arts and humanities that do not meet criteria of science, and it would be senseless to argue that they do not have “equal value” with science. Moreover, we argue that there is a cost to rejecting Indigenous knowledge, in that framing it with simplistic caricatures misses the potential for complementarity between science and Indigenous knowledge. Concomitantly, we highlight learning benefits that emerge when students are well versed in multiple knowledge systems. Last, we provide evidence that science innovation may be stifled if mainstream science is granted sole dominion over knowledge generation.

COMPLEMENTARY EXPLORATION Indigenous knowledge is often generated empirically and drawn from local context, complementing and challenging scientifically derived universal “truths.” These empirical aspects of Indigenous knowledge (which have been called the “know-how versions of knowledge”) that emphasize the method or workability [(3), p. 103] align most closely with science but still differ sufficiently from science in their context and underpinning worldview that many (but not all) scholars argue that science and Indigenous knowledge are not the same (3). In addition to the more “sciencelike” empirical components and their framings that we discuss below, there are components of Indigenous knowledge that are metaphysical or philosophical and entirely unrelated to science. These latter components alone have been emphasized by those who challenge the validity of Indigenous knowledge, with their criticisms likely fueled by recent examples of Creationism being taught in schools in place of evolutionary theory. Yet this framing of Indigenous knowledge as entirely “myth” is an “appeal to extremes” argument, which would benefit from a more mature and nuanced view of Indigenous knowledge. A key issue is that Western culture discretizes knowledge generation into disciplines, in which science cannot be contaminated by nonscience, whereas such divisions between methods of inquiry are often absent from Indigenous knowledge systems. However, we believe this only becomes a problem when teaching Indigenous knowledge “as” science, but not “alongside” it, as articulated in the Aotearoa–New Zealand policy. By analogy, philosophy is not science and includes inquiries around existence, knowledge, and the self. Both the questions asked and the methods of enquiry frequently depart from what could be called science. Moreover, philosophy can reflect on the nature of science in a way that science is unable to reflect on itself. For this reason, students who study philosophy alongside science can improve science.org SCIENCE

PHOTO: DR. LAUREN WALLER

their learning both of science concepts and concepts about science (4). Attributing greater “value” to science than philosophy (or other nonscience endeavors such as art or the humanities) would make little sense and be an opinion based on values rather than scientific evidence. Similarly, we argue that teaching Indigenous knowledge alongside science should not seek to usurp science (in the way that, for example, creationism seeks to undermine evolutionary theory because they are incompatible with one another), but rather it “provokes science, and can act as a mirror for science to see itself more clearly, reflected in a philosophically different form of knowledge” [(5), p. 87]. A parallel understanding of science and Indigenous knowledge systems would be complementary, emphasizing their similarities and cultural differences; the separation versus connection of empirical and philosophical subjects would be one example of those differences. Another example specific to Aotearoa–New Zealand would be that Te Ao Māori (Māori worldview) uses an intergenerational lens inclusive of the observer that gives cultural integrity to questions and generated outcomes, whereas the scientific method strives to be disconnected from that which it observes. The timescales of knowledge generation are also complementary. For example, short-duration scientific research funding cycles can create institutional barriers to long-term data acquisition and study of large-scale (such as environmental) problems. By contrast, Indigenous knowledge can and has contributed empirically generated, intergenerational knowledge, making it an increasingly valuable tool in environmental management, particularly around rare but increasingly frequent natural events such as large-scale deadly bush fires that plague Australia and parts of North America. For at least 40,000 years, Indigenous Australians have been managing the landscape, leaving a deep human imprint, one that has been nearly erased from living memory. However, in parts of Australia, local authorities, scientists, and Indigenous communities are now coming together to revisit Indigenous fire management and reframing science through Indigenous knowledge to better understand these modern environmental dilemmas (6). This example highlights how knowledge and its cultural context have a place in education because local context matters, particularly when Indigenous communities with their knowledge drive questions or request the support of science tools such as genomics to generate SCIENCE science.org

codeveloped conservation solutions. We hope that viewing Indigenous knowledge as complementary to science, without replacing nor being science, may lead to more nuanced and fruitful conversations around policy in this space and to maximizing the benefits of such policy. Yet despite all of this, the false dichotomy between the validity of Indigenous knowledge and science-generated knowledge persists and is frequently based on a straw person. Science and Indigenous knowledge systems comprise distinct perspectives of understanding the world

the basis that they are corrected as part of the scientific process, in which knowledge is updated as new information becomes available. Yet although Indigenous knowledge is also well known to be dynamic and continuously updated (7), critics do not afford it an equal right to correct itself. For example, “pity the moas were all eaten” (1) is commonly used rhetoric to imply the failure of Māori knowledge around conservation of a giant endemic New Zealand bird in the 15th century. Yet this reasoning mistakenly conflates the validity

Taxonomic plant identification is taught alongside Indigenous knowledge of the use of these plants to Indigenous students from various tribes around Tamaki Mākarau (Auckland) New Zealand.

because they differ in methodologies, philosophies, worldview, and modes of transmission. The knowledge produced through traditional science methods has resulted in many game-changing outcomes, such as the eradication of smallpox and the production of life-saving vaccines. However, it has also proven itself wrong (for example, phlogiston, aether, and phrenology) and produced catastrophic outcomes for humanity (such as the atomic bomb), while failing thus far to solve the most pressing challenges of our time (such as climate change). As scientists, we accept such scientific shortcomings on

of present-day Indigenous knowledge with 15th-century knowledge and decisionmaking. By comparison, this extinction was two centuries before British colonization would produce such mass environmental devastation in its colonies that the Western conservation paradigm would be born. In fact, evidence has shown some present-day Indigenous managed lands to have much higher biodiversity than some Western Conservation managed lands (8), and this can likely be attributed in part to the nuanced relationships that are encoded within Indigenous knowledge. Thus, the argument that Indigenous knowledge 9 FEBRUARY 2024 • VOL 383 ISSUE 6683

593

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

only includes historical, precolonization learnings, whereas mainstream science can continuously learn from its mistakes, is both a straw person and a circular argument because it defines Indigenous knowledge using the exact criteria (outdatedness) for which it is criticized. Some societies, such as many Indigenous groups, lack traditional written communication and thus transmit knowledge within memorable framings, such as stories or myths, to ensure their longevity (9). A superficial interpretation of these framings is often used to depict Indigenous knowledge as purely metaphysical—an example of the appeal to extremes fallacy—justifying its displacement by science (9). Yet this false dichotomy ignores evidence that, like Indigenous knowledge, science also uses abstractions and stories (such as models) to facilitate knowledge transmission and illustrate concepts or key messages. For example, both simulation and statistical models can require simplifications that are known to be false. Bohr’s model of the atom and Newtonian physics are still widely taught in schools as easily understood approximations, despite their limitations with respect to quantum mechanics. By analogy, dismissing Indigenous narratives on their verbatim interpretation risks missing considerable opportunity to learn from the knowledge and experience encoded within them (10, 11). The argument “True science is evidencebased not tradition-based” (1) ignores considerable research demonstrating that false representations of both science and Indigenous knowledge have unnecessarily polarized this debate. We argue that it would be more fruitful to undertake it in an informed and nuanced way. KNOWLEDGE TRANSMISSION In addition to a suite of known benefits to Indigenous students (12), we see the potential for all students to benefit from exposure to Indigenous knowledge, alongside a science curriculum, as a way of fostering sustainability and environmental integrity (13). For science learning, connecting science with student values and fostering understanding of the role of social and cultural context can lead to the production of ethically sourced scientific knowledge (14). In addition, the generation and transmission of Indigenous knowledge are both closely connected to practice: experiencing and doing. Such experiential learning is known to benefit learning in general (14), and the broader range of contexts provided by place-based Indigenous knowledge allows students to connect learning 594

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

with their local environment, which may result in more affective and authentic engagement, leading to greater acceptance and uptake of new knowledge (14). Given the societal and environmental issues facing the planet, providing an intercultural understanding that leads to a more balanced and connected worldview can result in positive outcomes, including effective science education. INNOVATION DRAWS FROM DIVERSITY Innovation, like evolution, draws from diversity, so that diversity of knowledge sources and transfer among them are known to positively influence innovation (15). This value is exemplified by the move toward cross-disciplinarity, in which science can draw on inductive fields of research for hypothesis generation. Given this value of diversity, global challenges faced by humanity could benefit from inclusive science and maintenance of knowledge diversity more generally rather than insisting on assimilation into a single culture of knowledge generation. One path to preventing the extinction of Indigenous knowledge is its dissemination in classrooms, under Indigenous governance and management (supported by the International Bill of Rights and, specifically in New Zealand, the Treaty of Waitangi Act 1975 and the Waitangi Tribunal). Not only will this help to protect Indigenous knowl-

edge holders and their culture, it has the potential to generate innovation more broadly. EVIDENCE, NOT CARICATURES Indigenous knowledge can complement science-generated knowledge in the pedagogy landscape by providing acceptance and understanding and by contributing to the addressing of global challenges. We urge both education policy analysts and scientists engaging in this debate to draw on evidence rather than caricatures of Indigenous knowledge and a partisan approach to knowledge generation. Knowledge is produced in many traditions. The scientific method is one of those, Indigenous approaches are others, and these are not necessarily mutually exclusive. We need to respect Indigenous knowledge for its inherent value and the philosophical reflections it can provide science to improve outcomes, irrespective of how Indigenous knowledge is contextualized. Much of our time as researchers is spent challenging scientifically derived universal truths through work in local contexts, and Indigenous knowledge does the same but with a higher degree of connectivity between the researcher and what is “researched.” Arguably, the ignorance toward Indigenous knowledge and its application is only slightly greater than ignorance to science methodology. We think this is the strongest rationale for teaching them both in schools. j R E F E R E N C ES A N D N OT ES

1. R. Dawkins, “Why I’m sticking up for science,” The Spectator, https://www.spectator.co.uk/article/whyim-sticking-up-for-science (2023). 2. J. Tollefson, Nature 621, 454 (2023). 3. C. Mika, J. World Philos. 7, 101 (2022). 4. M. Monk, J. Osborne, Sci. Educ. 81, 405 (1998). 5. G. Stewart, New Zealand J. Teach. Work 19, 84 (2022). 6. M. S. Fletcher, T. Hall, A. N. Alexandra, Ambio 50, 138 (2021). 7. R. Tang, M. C. Gavin, Conserv. Soc. 14, 57 (2016). 8. S. T. Garnett et al., Nat. Sustain. 1, 369 (2018). 9. G. Snively, J. Corsiglia, Sci. Educ. 85, 6 (2001). 10. S. M. Sidik, Nature 601, 285 (2022). 11. C. H. Trisos, J. Auerbach, M. Katti, Nat. Ecol. Evol. 5, 1205 (2021). 12. F. Duckworth, M. Gibson, S. Macfarlane, A. Macfarlane, Alternative 17, 3 (2021). 13. R. Zidny, J. Sjöström, I. Eilks, Sci. Educ. 29, 145 (2020). 14. S. Jose, P. G. Patrick, C. Moseley, Int. J. Sci. Educ. 7, 269 (2016). 15. J. De Beer, J. New Gener. Sci. 14, 34 (2016). AC K N OW L E D G M E N TS

Learning about the biological structures of plants used in tradtional practices

The authors acknowledge colleagues S. Lambert, S. Nissen, P. Dearden, P. Hulme, and two anonymous referees who provided feedback on various drafts; S. Scott for reference checking; and M. Duley, F. Bulman, and Z. Marshall for images provided. Funding was provided by the New Zealand government, Tertiary Education Commission Centres of Research Excellence Fund for Bioprotection Aotearoa contract 2021–2028, and Biological Heritage National Science Challenge 1920-44-021 A. 10.1126/science. adi9606

science.org SCIENCE

A worker surveys equipment at the National Ignition Facility at Lawrence Livermore Laboratory.

B O OKS et al . NUCLEAR WEAPONS

In search of nuclear stewards A journalist probes the culture and convictions of researchers at US national labs By Jacob Darwin Hamblin

can do social good at the margins of weapons work. A few simply find nuclear science uclear weapons have been the highest exciting and seem to have few qualms about strategic priority of the United States its implications. Others offer rationalizations: for nearly 80 years. Of the more than One noted that if she could not live in a world 2000 nuclear detonations that ocwithout nuclear weapons, she wanted to encurred in the half-century after 1945, sure their safety, security, and reliability. about half were US weapons tests. ToScoles sees dual-use science as fundamenday, the country’s scientists work to maintain tal to the moral universe of the national labs the viability of the arsenal—planand as NNSA’s strategy for keepning subcritical explosions, using ing highly qualified people on computers to make complex calthe payroll. Some employees do culations, and engaging in myrnot work on weapons at all, and iad projects to act as “stockpile most imagine themselves contribstewards.” In Countdown, science uting to nonmilitary domains. At journalist Sarah Scoles tells the Lawrence Livermore, the National stories of the people on the front Ignition Facility focuses on fulines of nuclear research and desion reactions, for example. It velopment at sites managed by exists because of thermonuclear Countdown: The Blinding Future the National Nuclear Security weapons, but the experiments of Nuclear Weapons Administration (NNSA), focusing conducted there might also lead Sarah Scoles specifically on the national labs at Bold Type Books, 2024. to fusion power for electricity Los Alamos, Lawrence Livermore, generation. Similarly, its National 272 pp. Sandia, and Savannah River. Atmospheric Release Advisory Much of the work conducted in these Center models potential casualties from exlabs is secret “mission science” related to plosions or radiological attacks, but it also adweapons, but scientists at national labs also vises about civilian reactor accidents. Many publish research in academic journals. Scoles of the lab’s experts pursue technologies that probes their feelings about the future of numay prevent the diversion of nuclear matericlear weapons and the roles they are playing als away from peaceful uses or detect a forin it. Some believe in deterrence or think they eign nation’s clandestine nuclear test. Others perform nuclear forensics to help determine the source of bomb materials. The reviewer is at the School of History, Philosophy, and If the lines separating nonmilitary work Religion, Oregon State University, Corvallis, OR 97331, USA. Email: [email protected] from the national labs’ weapons mission

PHOTO: CORBIS VIA GETTY IMAGES

N

SCIENCE science.org

seem hazy, this is because they prefer it that way, not least because the ambiguity helps bring young people into the fold. An uncomfortably large proportion of NNSA employees are at retirement age, and the agency is keen to recruit. It scouts new talent from universities across the country, convenes nuclear “bootcamps,” provides paid internships, and offers myriad funding opportunities. Some recruits enjoy these opportunities and then move on, while others stay to establish careers in the national lab system, becoming stewards of the US arsenal. As with military funding in general, the grant money and career opportunities are hard to resist. At the heart of Countdown is a recent policy change by the United States to resume large-scale manufacture of plutonium pits— the part of a warhead that causes a nuclear explosion. Such pits were once produced at the Rocky Flats plant in Colorado, but production was halted there permanently when the Federal Bureau of Investigation found the site in violation of safety and environmental regulations during a 1989 raid. The US is now on track to produce them again at Los Alamos and Savannah River. It is possible that the old pits work fine, but testing them would open a can of worms. The US and other countries agreed to limit nuclear tests to underground ones in 1963, and the 1996 Comprehensive Test Ban Treaty banned nuclear explosions altogether. (The US signed the treaty but never ratified it.) Officially, pit production will ensure that the bombs are reliable, but skeptics argue that new pit production will enable modernized weapon designs. Whether it is necessary or not, it will mean jobs and research dollars for the production facilities and labs. Since 2006, Los Alamos has been operated by a profit-driven conglomerate called Los Alamos National Security, and Scoles sees it subtly changing the lab’s priorities. Some scientists have departed in frustration, management fees have skyrocketed, and corporate culture seems intent on advocating—in the name of national security—a weapons modernization program that also happens to be quite profitable. Countdown offers readers an accessible introduction to some of the frontiers of US nuclear research, from quantum computing to laser-induced fusion. But its great strength is in letting scientists who do this work— many of them women—speak for themselves, as they wrestle with moral quandaries and imagine the consequences of their stewardship of the US nuclear arsenal. j 10.1126/science.adn2591 9 FEBRUARY 2024 • VOL 383 ISSUE 6683

595

I NS I GHTS | B O O K S

SCIENCE LIVES

A Black spaceflight legacy

The Space Race Lisa Cortés and Diego Hurtado de Mendoza, directors National Geographic, 91 min, premiering 12 February 2024.

An aspiring Cold War–era astronaut and those who followed in his footsteps take center stage in a new documentary By Matthew Shindell

Melvin and Victor Glover. This reliance on individuals’ memories over archival history has its drawbacks, as noted above, but it he story of Ed Dwight—a young Air allows the importance of Dwight’s story to Force captain who might have been come through more clearly. America’s first Black astronaut in Although their selection to the astronaut the 1960s—has rarely been treated corps in 1978 came little more than a decade in much depth and has never been after Dwight’s resignation, neither Bluford adequately incorporated into our unnor Gregory knew that Dwight had once derstanding of the early history of human believed himself to be a contender for the spaceflight. Rather, it has emerged multiple title of first Black astronaut. Nor did they times in newspaper articles and documenknow the story of Air Force pilot Robert tary films, especially in the past decade, as Lawrence Jr., who died in a training accident an untold story in the history of the Cold after being selected for the Air Force’s clasWar space race. National Geographic’s new sified space project, the Manned Orbiting documentary, The Space Race, is the latest to Laboratory. They learned of these stories tell Dwight’s story. Although it covers some only after their time in space. familiar territory, it manages to do In their own successes, as well so in new ways and for the first time as those of younger astronauts who examines Dwight’s legacy in the conhave followed in their footsteps, text of the ongoing contributions of Bluford and Gregory are able to African Americans to spaceflight. frame Dwight’s and Lawrence’s expeThis is the story as Dwight tells riences as milestones on the road to it: His astronaut candidacy resulted equality in space. It is clear to them from a promise President John F. that this journey is not yet comKennedy made to the National plete—a fact that is brought home Urban League’s Whitney Young Jr. poignantly by the experience of Young asked Kennedy to deliver a Glover, who was on the International Black astronaut for the civil rights Space Station during the trial of movement—a symbol for Black George Floyd’s murderer, police offiequality and a role model for Black cer Derek Chauvin. Glover, who had students who might pursue military brought with him to space a painting or engineering careers. At the behest of Floyd, struggled emotionally durof the Kennedy White House, Air ing the trial. But he was able to find Force Chief of Staff General Curtis fellowship in a support network of LeMay enrolled Dwight, already an accomplished pilot, in the Aerospace Black astronauts past and present—a Research Pilot School, the program group that we learn has come to call Astronaut Charles Bolden is photographed on the shuttle Columbia. that had trained some members of itself “the Afronauts.” NASA’s first astronaut class. Despite resisAdministrator, making him the first Black Human spaceflight is famously aspiratance, as Dwight remembers it, from the man appointed to this position. During tional—not only do we want astronauts to school’s commandant, Colonel Chuck Yeager, his confirmation hearing, Bolden drew the represent the best of humanity, we also call Dwight did well in the program. Senate’s attention to Dwight, whom he had on them and the technologies that send But Dwight was never named to NASA’s invited to be in the audience, and credited them to space to show us what we are caastronaut corps. NASA promoted his candihim for paving the way for African American pable of and who we can be. As The Space dacy to the Black press and sent him around astronauts. Dwight’s story forms the backRace reminds us, the possibilities of space the country to talk about space exploration bone of this documentary, but it is Bolden are always constrained by our present hopes, to Black audiences. Whether Dwight would who connects Dwight’s story to the more restruggles, and prejudices. But occasionally, have only been used to deliver pro-NASA cent history of human spaceflight. space does show us a glimpse of how we propaganda or whether he would have been Rather than privileging the perspectives might transcend our limitations. j sent to space had Kennedy not been assasof historians, the history of Black contribuRE FE REN CES A ND N OT ES sinated in 1963 is unclear. What is certain is tions to NASA is here navigated by Dwight 1. R. Paul, S. Moss, We Could Not Fail: The First African that Kennedy’s death put an end to Dwight’s and Bolden, whose voices are joined by the Americans in the Space Program (Univ. of Texas Press, first Black astronaut, Guion Bluford; the 2015). The reviewer is a space history curator at the Smithsonian’s third Black astronaut, Frederick Gregory; National Air and Space Museum, Washington, DC, USA. and more recent astronauts, such as Leland 10.1126/science.adn2127 Email: [email protected]

T

596

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

hopes. He resigned from the Air Force in 1966, feeling defeated. Despite all the fanfare NASA and the White House had drummed up around Dwight, his space journey ended unceremoniously. It must be noted that historian Richard Paul has previously scrutinized Dwight’s story and found it to be less clear-cut than Dwight’s retelling; however, none dispute that Dwight was used as a PR corrective for the fact that NASA’s all-white astronaut corps did not reflect the progress of the civil rights movement (1). In 2009, President Barack Obama named highly accomplished astronaut Major General Charles F. Bolden Jr. as NASA

science.org SCIENCE

LET TERS

Decreased water flow in China’s Yellow River, along with other threats, has led to a decline in native fish.

Edited by Jennifer Sills

Protect native fish in China’s Yellow River Since 2021, Chinese President Xi Jinping has underscored the importance of wetland restoration to safeguard the ecosystems of the Yellow River (1), China’s second-longest river and a cradle of its cultural heritage (2). The administration’s “Outline of ecological protection and high-quality development in the Yellow River Basin” identifies measures to preserve vital wildlife habitats along the river (1). However, the plan focuses solely on the flora and fauna in the adjacent terrestrial ecosystems and overlooks the river’s critically endangered indigenous fish. The Yellow River basin has undergone a stark decline in fish diversity. During the 1980s, more than 120 fish species lived in the river’s main stream, but only 47 remained by 2013 (3). This rapid reduction has pushed iconic species such as the Chinese paddlefish and bronze gudgeon to the brink of extinction (4). The “Red list of Chinese vertebrates” highlights the precarious state of 24 native fish species in the Yellow River, with 4 classified as Critically Endangered, 10 as Endangered, and 10 as Vulnerable (5). Even species that were once common are now at risk of extinction. Overfishing is the primary threat to the Yellow River’s fish (3), but other factors have also contributed to the decline. Discharge from industrial activity along the lower and middle reaches of the Yellow River has led to deteriorating water quality (6). Invasive nonnative fish species and changes in the genetic purity of native fish also pose a 598

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

threat. In 2013, more than 25 nonnative species were found in the Yellow River, including sunfish, Brazilian turtles, and crocodiles (3). The Yellow River carp has almost disappeared (4) as a result of poor water quality and the incursion of nonnative carp since the late 1990s (6). Climate change has also contributed to the decline of fish populations. Between 1919 and 2018, and especially since 1986, the Yellow River’s water flow has decreased, with a staggering 64% reduction in the volume of water that reaches the sea (7). Primarily driven by climate change, this reduction in flow exacerbates the challenges faced by the river’s ecosystems (8). A comprehensive fishing ban might help to preserve fish diversity in the Yellow River basin, but the river’s extensive network of reservoirs presents a formidable challenge (9). Spanning three major terrains, the Yellow River features 219 interconnected reservoir facilities that cannot be easily altered (9). Even with a comprehensive fishing ban, numerous fish species might remain confined to specific areas, offering minimal assistance in terms of genetic exchange and reproduction (10). However, local fishing bans could offer protective measures for fish species in specific areas. In addition, China should implement stringent environmental policies, control invasive species, and conduct climate-adaptive ecological restoration. By signing the Kunming-Montreal Global Biodiversity Framework (11) China has committed to fortifying biodiversity preservation efforts. In addition to southern China’s well-known biodiversity hotspots, the country must protect northern regions like the Yellow River basin. The wetland restoration

plan, combined with efforts to restore the Yellow River’s fish species, would represent a good start. Xu Guo1, Qianhui Lin2, Xiangcheng Zheng3, Shuo Wang2, Qiao Li4, Chao Shi2* 1

Geotechnical and Structural Engineering Research Center, Shandong University, Jinan 250061, China. 2College of Marine Science and Biological Engineering, Qingdao University of Science and Technology, Qingdao 266042, China. 3 School of Mathematics, Shandong University, Jinan 250100, China. 4School of Engineering Medicine, Beihang University, Beijing 100083, China. *Corresponding author. Email: [email protected] RE FE REN CES A ND N OT ES

1. Central People’s Government of the People’s Republic of China, “Outline of ecological protection and highquality development plan in the Yellow River Basin” (2023); https://www.gov.cn/zhengce/2021-10/08/ content_5641438.htm [in Chinese]. 2. A. Lawler, Science 325, 8 (2009). 3. X. Shan et al., Mar. Coast. Fish. 5, 1 (2013). 4. W. Tang, D. He, J. Lake Sci. 25, 4 (2013). 5. Z. Jiang et al., Biodivers. Sci. 24, 5 (2016). 6. C. Liu, J. Xia, Hydrol. Process. 18, 12 (2004). 7. Y. Ru et al., Mar. Sci. 30, 3 (2006). 8. W. W. Immerzeel, L. P. Vanbeek, M. F. Bierkens, Science 328, 6 (2010). 9. L. Ran, X. X. Lishan, Z. Xin, X. Yang, Glob. Planet. Change 100, 1 (2013). 10. L. Ran, X. X. Lu, Hydrol. Process. 26, 8 (2012). 11. Convention on Biological Diversity, “Kunming-Montreal Global Biodiversity Framework” (2023); https://www. cbd.int/doc/decisions/cop-15/cop-15-dec-04-en.pdf. 10.1126/science.adn7432

India’s services-driven growth excludes women In the past two decades, India has been developing its business services sector, especially information and communication technology, to drive economic growth (1, 2). With annual growth of 8.2%, high-skilled science.org SCIENCE

I N SI G H T S

service jobs have contributed to economywide surge in productivity (2, 3). However, there are not enough qualified individuals to fill available positions (4). The low representation of women in these industries exacerbates labor shortages (4). Women workforce participation in India has stagnated at less than 30% (5, 6). Of the women who do work, 85% have jobs in agriculture or industries such as textiles, construction, and food (4–6). As India’s business services sector has grown and high-paying job opportunities have increased, women have been excluded. Jobs in the business services sector require information and communication technology skills. Women in India are less likely than men to be able to afford a cell phone or computer (7). Only 30% of women know how to use the Internet to find necessary information, operate mobile money applications, or use e-services (7). Because women are less likely to use technology in their daily lives, they are at a disadvantage in labor and financial markets (8). To address the gender gap in technological skills, India should provide digital skills training for women, tailored to the job requirements of the information and communication technology industry. The nationwide Digital India program, which aims to improve access to technology throughout the country (9), could add such training efforts to its goals. The training should be supplemented with support groups for women in the technology sector and online job-matching services that include peer networks [e.g., (10)]. These efforts, implemented at the local level, could help women overcome the obstacles that prevent them from accessing and using technology. For example, support groups for 4.3 million women learning digital skills across municipalities in Kerala have led to computer use in village offices, the introduction of a digital accounting system, and the deployment of digital technology for information and public services (10). Policy efforts directed toward closing digital gender gaps would have far-reaching benefits. Women would gain increased control over financial resources, providing them with access to additional training programs and opportunities. Giving women financial independence would create cascading effects at the household level as they pass that stability and access to knowledge to their children (11, 12). Isha Gupta Department of Economics, Ramanujan College, University of Delhi, Delhi, India. Email: [email protected] SCIENCE science.org

R EFER ENCES AN D N OT ES

1. B. Eichengreen, P. Gupta, “The service sector as India’s road to economic growth,” Working Paper w16757, National Bureau of Economic Research (2011). 2. K. L. Krishna et al., “Productivity dynamics in India’s service sector: An industry-level perspective,” Working Paper 261, Centre for Development Economics (2016). 3. D. W. Jorgenson, M. P. Timmer, Scand. J. Econ. 113, 1 (2011). 4. B. Bhandari et al., “Skilling India–No time to lose,” Technical Report, National Council of Applied Economic Research (2018). 5. Employment Statistics in Focus, “Female labour utilization in India,” Ministry of Labour and Employment (2023). 6. E. Fletcher, R. Pande, C. M. T. Moore, “Women and work in India: Descriptive evidence and a review of potential policies,” CID Working Papers 339 (2017). 7. “The Mobile Gender Gap Report 2023,” Global System for Mobile Communications (2023); https://www.gsma. com/r/wp-content/uploads/2023/07/The-MobileGender-Gap-Report-2023.pdf. 8. “Bridging the digital gender divide: Include, upskill, innovate,” Organization for Economic Co-operation and Development (2018). 9. Ministry of Electronics and Information Technology, “Digital India” (2023); https://csc.gov.in/digitalIndia. 10. A. Gurumurthy, N. Chami, “Digital India through a gender lens” (Research Report, Heinrich Bohl Stiftung, 2018). 11. T. Suri, W. Jack, Science 354, 1288 (2016). 12. P. Choudhuri, “India’s employment challenges and the demand for skills,” Working Paper 121, National Council of Applied Economic Research (2021). 10.1126/science.adn7738

Oil pollution threatens Persian Gulf marine life The Persian Gulf is a shallow sea located between Iran, Iraq, Bahrain, Kuwait, Oman, Saudi Arabia, Qatar, the United Arab Emirates, and Yemen. These nine countries possess 50% of the world’s oil reserves (1) and produce about 25% of the oil consumed globally (2). The extensive oil and industrial activities, including numerous oil spills between 1980 and 2020 (1, 2), have led to substantial pollution and severely damaged marine life in the Persian Gulf (3). Given that preserving the quality of the marine environment is essential to the social fabric of life and to the economic success in the region (2), governments must prioritize environmental protection and provide support to environmental organizations in the region. Aromatic hydrocarbons are commonly found in oil spills. These pollutants have high melting and boiling points but low solubility and vapor pressure, and they are toxic and mutagenic (4). The high levels of ultraviolet radiation and elevated temperature in the Persian Gulf can substantially increase the toxicity of hydrocarbons (1), which subsequently cause damage to coral reefs, fish, phytoplankton, and zooplankton (1, 5). Aromatic hydrocarbons harm corals and rocky organisms by damaging tissues (4),

causing bleaching (6), and preventing larval settlement (1). Fish and other aquatic organisms can absorb pollutants in water directly and through the food they consume (7). The Gulf War and oil pollution in 1991 caused a reduction in the Saudi prawn population from 4000 t to 25 t, as well as a decrease to 1/10 the abundance of eggs and larvae within a year (2). On Kharg Island, where much of Iran’s crude oil is exported (1), elevated levels of polycyclic aromatic hydrocarbons and their derivatives were found in the liver and muscle tissues of three edible fish species in 2018 (7). Hydrocarbon pollutants also threaten phytoplankton and zooplankton (4), which can then harm higher-level species in the food chain. All governments in the region must prioritize the establishment and enforcement of effective management practices, sustainable methods, and strict regulations to mitigate oil pollution and safeguard the marine environment in the Persian Gulf (2). It is imperative that these governments remove any constraints that hinder the continuous monitoring and assessment of pollution in industrial zones. By implementing monitoring programs, governments can aid in the recovery of damaged ecosystems and ensure the safety of seafood (2, 3). In addition to national government action, regional management programs should be established under international supervision. Independent research knowledge should be shared transparently, and problems caused by regional and local pollution should be identified. The Regional Organization for the Protection of the Marine Environment (8), the sole Sea Forum in the Persian Gulf, plays a crucial role in managing and conserving the area and could facilitate collaborations among oil-producing countries in the region (2). Hossein Yarahmadi Department of Chemical Engineering, Sirjan University of Technology, Sirjan, Iran, Email: [email protected] RE FE REN CES A ND N OT ES

1. M. Oladi, M. R. Shokri, J. Hazard. Mater. 409, 124993 (2021). 2. A. M. Freije, J. Assoc. Arab Univ. Basic Appl. Sci. 17, 90 (2015). 3. S. Uddin, et al., Mar. Pollut. Bull. 173, 112913 (2021). 4. F. Khaksar et al., Mar. Pollut. Bull. 140, 35 (2019). 5. S. L. Coles, Atoll Res. Bull. 507, 1 (2003). 6. “Bleaching; a threat to coral reefs,” IRNA (2021); www. irna.ir/news/84519319, [in Farsi]. 7. A. Ranjbar Jafarabadi et al., Chemosphere 215, 835 (2018). 8. UN Environment Programme, “ROPME sea area” (2024); https://www.unep.org/explore-topics/ oceans-seas/what-we-do/working-regional-seas/ regional-seas-programmes/ropme-sea-area. 10.1126/science.adn5624 9 FEBRUARY 2024 • VOL 383 ISSUE 6683

599

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

NEUROSCIENCE

Moving mitochondria in neurons

M

itochondria must be trafficked over long distances in neurons to support energy-intensive processes. Izquierdo-Villalba et al. found that mitochondrial trafficking in neurons required the interaction of the G protein Gαq with the mitochondrial protein Alex3. Gαq enhanced the binding of Alex3 to proteins that mediate anterograde trafficking of mitochondria. Expression of a constitutively active form of Gαq in neurons from mice with a central nervous system deficiency in Alex3 revealed that Gαq was necessary for the functioning of Alex3 in dendritic arborization and in mitochondrial trafficking and distribution. —WW Sci. Signal. (2024) 10.1126/scisignal.abq1007

Neuronal mitochondria (light blue) are moved away from the cell body in a mechanism that involves G protein signaling.

PLANT SCIENCE

Oil production in citrus plants Secretory cavities serve as reservoirs for a variety of secondary metabolites, such as essential oils and chemicals that defend against pathogens and herbivores. Wang et al. describe a molecular pathway for the development of the secretory cavity (oil glands) in Citrus species. Using genetic mapping and genome editing, the authors identified two transcription factors involved in the development of the secretory cavity and cisregulatory elements within a leaf shape gene that enhance gene 600

activation. A regulatory cascade activates the MYC5 gene, which regulates oil gland sheath cell differentiation and cavity formation. This work lays a foundation for understanding secretory cavity development and the synthesis of essential oils. —AWa Science p. 659, 10.1126/science.adl2953

QUANTUM GASES

Seeing second sound In superfluids, heat propagates like a wave in a phenomenon dubbed “second sound.” However, observing this propagation directly is tricky. Yan et al. used a quantum gas of strongly interacting fermionic lithium atoms held

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

in a box potential to visualize second sound. The researchers used radiofrequency spectroscopy to map out local changes to the temperature. Above the superfluid transition, heat propagated diffusively, but below the transition, wave-like propagation characteristic of second sound was observed. —JS Science p. 629, 10.1126/science.adg3430

POLLINATION

Pollution disrupts scent Chemical pollutants not only reduce animal survival and reproduction, but can also disrupt their senses, changing their behavior and interactions

with other species. Common air pollutants such as ozone degrade floral scents, potentially affecting insects’ ability to locate and pollinate flowers. Chan et al. tested the effects of ozone and nitrate radicals (NO3) on nocturnal hawkmoth pollination of a desert plant. Both pollutants decreased concentrations of monoterpenes, but NO3 had a stronger effect. Degradation of scent molecules by NO3 led to a decline in hawkmoth visitations in wind tunnel and field experiments, with predicted reductions in fruit set and plant fitness, with predicted reductions in fruit set and plant fitness. A global model suggests science.org SCIENCE

that many urban areas have sufficient pollution to significantly reduce the distances at which pollinators can sense flowers. —BEL

total enzymes in tobacco to demonstrate a minimal pathway to a key intermediate product: baccatin III. —MAF Science p. 622, 10.1126/science.adj3484

Science p. 607, 10.1126/science.adi0858

GEOLOGY MASS SPECTROMETRY

Free hydrogen

A collision course breaks the mirror

Hydrogen is an attractive alternative to traditional fossil fuels because it can be used to produce energy without generating carbon dioxide as a by-product. However, natural sources of hydrogen are rare and producing it is energy intensive. Truche et al. found a large natural source of hydrogen gas emitting from deep within the Bulqizë chromite mine. This large hydrogen flux is likely from long-term accumulation within a faulted reservoir. Places with similar geology should be good targets for finding other natural sources of hydrogen. —BG

Because proteins, carbohydrates, and nucleic acids occur in just one of two mirror-image geometries in nature, it is likewise essential for chemists to make just one of the mirror images, or enantiomers, of many pharmaceutical compounds. Currently, the most common means of distinguishing enantiomers is by separating them with chromatography. Zhou et al. report a technique to determine enantiomeric ratios by mass spectrometry alone. The method relies on alternating current excitations of the ions to induce rotational trajectories in which each enantiomer experiences distinct collision behavior, leading to sufficient separation for highresolution quantification. —JSY Science p. 612, 10.1126/science.adj8342

BIOSYNTHESIS

PHOTO: DAVID AUBREY/SCIENCE SOURCE

Enzymes for paclitaxel production Chemical total synthesis of paclitaxel, an important anticancer drug originally derived from yew bark, has been an important challenge for organic chemists, but the material used for drug production comes from isolation of the natural product or semisynthesis from precursors. It is therefore important to fully understand the biosynthetic pathway in plants. Jiang et al. identified a cytochrome P450 enzyme that generates a key chemical feature in the final products, a four-membered oxetane ring. The authors found that a parallel epoxidation is off pathway and not an essential intermediate. Along the way, they characterized another oxidation enzyme and ultimately reconstituted nine SCIENCE science.org

IN OTHER JOURNALS

Edited by Caroline Ash and Jesse Smith

Lichens attached to trees make a substantial, and often overlooked, contribution to carbon cycling in boreal forests.

Science p. 618, 10.1126/science.adk9099

GENE THERAPY

A CRISPR fix for dysfunctional T cells Familial hemophagocytic lymphohistiocytosis (FHL) is a spectrum of inherited, potentially fatal inflammatory diseases caused by loss-of-function mutations in the natural killer cell and T cell cytotoxic machinery. Although FHL can be treated with hematopoietic stem cell transplantation, mortality is still high, and more effective therapeutic approaches are needed. Li et al. developed an adeno-associated virus–based CRISPR-Cas9 system combined with nonhomologous end joining inhibition to repair T cells from a perforin-deficient Epstein-Barr virus–triggered mouse model of FHL. Repaired T cells were able to prevent and cure disease in this model. Moreover, this CRISPR-Cas9 system could be used to mend T cells from pediatric patients with FHL diseases. Gene repair of autologous T cells is thus a potentially promising treatment for human FHL. —STS Sci. Immunol. (2024) 10.1126/sciimmunol.adi0042

LICHENS

Cloaked in life

M

ost static outdoor surfaces are blanketed by microorganisms and, in boreal regions, noticeably by the symbiotic algal and fungal assemblages of lichens. Lichens can photosynthesize at freezing temperatures and grow when covered by snow, but little is known about their contribution to forest photosynthesis or respiration. Matvienko et al. examined the role of epiphytic lichens in the carbon cycle in Siberian forests, where they are estimated to amount to 3 kilograms per hectare of forest. In the laboratory, hydrated lichens were shown to emit methane for several weeks, and carbon dioxide flux was found to be species specific. The findings raise many questions about the roles of lichen diversity and abundance, season, drought, and aspect on forest carbon budgets. —CA Forests (2024) 10.3390/f15010107

MICROBIOME

Tonic signaling for tolerance Type I interferons (IFNs) are a type of chemical messenger known as cytokines that the immune system uses to initiate immune responses. IFNs are well recognized for their ability to protect against viral infections, but they also have a less well understood role in response

to bacteria. Ayala et al. investigated whether commensal bacteria within the intestine elicit IFN responses that can affect immune tolerance, a safeguarding process that limits excessive and damaging immune reactions. Within the intestinal lumen, commensal bacteria induced a subtle type I IFN response referred to as tonic signaling. The tonic IFN signals promoted dendritic cells to release IL-27, which in turn shaped

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

601

R E S E A RC H

ALSO IN SCIENCE JOURNALS CANCER

Changing lung cancer subtypes Acquired drug resistance to certain targeted cancer therapies can lead to histological transformation (HT) from one tumor type to a more aggressive type. HT is a poorly understood process, so Gardner et al. designed experimental models to study the molecular events that underlie the conversion of mutant epidermal growth factor receptor (EGFR) lung adenocarcinoma (LUAD) to neuroendocrine small cell lung cancer (SCLC) (see the Perspective by Berns). Switching certain genes on and off in either alveolar type 2 cells (the precursors of LUAD) or in pulmonary neuroendocrine cells (precursors of SCLC) led to HT that mimicked that observed in human cancer. Neuroendocrine SCLC transformation was only observed when tumor cells reworked their oncogenic driver program from EGFR to the Myc oncogene. —PNK Science p. 603, 10.1126/science.adj1415; see also p. 590, 10.1126/science.adn5218

REPRODUCTION

Helpful fibroblasts The corpora cavernosa are masses of vascular tissue that can fill with blood and thereby enlarge upon stimulation, creating the structure needed for penile erection. By studying the underlying mechanism for this process in mice, Guimaraes et al. determined that perivascular fibroblasts in the corpora cavernosa play a key role in erection physiology (see the Perspective by Ryu and Koh). Norepinephrine is a vasoconstrictor that restricts penile blood flow at baseline, whereas vasodilators released by sexual arousal counteract its effects, allowing an erection to take place. Recurrent erectile activity down-regulates Notch signaling, thereby increasing the numbers of perivascular fibroblasts, and SCIENCE science.org

Edited by Michael Funk

these fibroblasts then suppress vasoconstrictive norepinephrine signaling. Conversely, aging is associated with a decrease in these fibroblasts, contributing to the risk of erectile dysfunction. —YN Science p. 604, 10.1126/science.ade8064; see also p. 588, 10.1126/science.adn5182

PLANT SCIENCE

Photoperiodic growth Plants are highly responsive to photoperiodic cues, and in many parts of the world, daylength varies considerably throughout the year. Complex signaling networks have been identified that regulate the timing of flowering in response to daylength. Wang et al. found that a parallel mechanism exists to regulate vegetative growth (see the Perspective by Buckley and Haydon). Whereas flowering time primarily depends on light signaling, vegetative growth is controlled by photosynthetic and metabolic cues that change according to photoperiod. The authors found that photosynthetic control of vegetative growth is partially dependent on the production of myo-inositol, a precursor for many other molecules involved in plant growth. —MRS Science p. 605, 10.1126/science.adg9196; see also p. 589, 10.1126/science.adn5189

HEALTH ECONOMICS

Understanding providers’ motives Diarrhea is a leading cause of child mortality in India. It becomes deadly when excretions exacerbate severe dehydration and loss of electrolytes. Most health care providers in India know that oral rehydration salts (ORS) are an inexpensive, lifesaving treatment for child diarrhea, yet they are widely underused. Wagner et al. undertook randomized controlled trials involving standardized patients (actors trained

to seek care for a child’s diarrhea) who visited 2282 private health care providers in India. Trials were designed to identify three barriers driving underutilization: assuming patients lack interest in ORS, incentives to prescribe more lucrative (but inappropriate) medicines, and incentives to sell non-ORS medicines in stock when ORS are unavailable. The dominant barrier was assuming that patients were uninterested, showing that simple interventions could save many lives. —EEU Science p. 606, 10.1126/science.adj9986

antiferromagnetic iron telluride. Neither of these materials is superconducting, but iron telluride is a parent compound for a family of iron-based superconductors. Interfacing the layers led to the appearance of superconductivity in the presence of ferromagnetism and topological band structure. This combination of properties makes the heterostructure a promising, although not yet proven, platform for observing chiral topological superconductivity. —JS Science p. 634, 10.1126/science.adk1270

METALLURGY

PLANT REPRODUCTION

Printing preferable parts

Parental coordination

Laser powder bed fusion provides the opportunity to make custom-built metal structures, but these objects can have undesirable variability in mechanical properties. Zhang et al. addressed the origin of this issue, unwanted metastable phases and columnar-shaped crystals, by adding molybdenum nanoparticles to a common aluminum alloy (see the Perspective by Zhang and Wang). The nanoparticles both encouraged the growth of symmetric grains and suppressed the formation of unwanted phases. The resulting samples made from laser powder bed fusion have much better mechanical properties, showing the promise of this design strategy. —BG

During reproduction, flowering plants undergo two fertilization events to form the embryo and the nutritive endosperm. After gametes fuse, they must then resume cell cycle progression to allow them to divide and develop. Simonini et al. found that a protein called RBR1 holds the female endosperm precursor cell in stasis, preventing full DNA replication. During fertilization, the sperm cell delivers a D-type cyclin to the female cell to reactivate the cell cycle. These results give insight into the communication between parental cells to enable coordination of subsequent development. —MRS

Science p. 639, 10.1126/science.adj0141; see also p. 586, 10.1126/science.adn6566

SOLID-STATE PHYSICS

Mixing the right ingredients One of the recipes for realizing topological superconductivity calls for interfacing a topological insulator with a superconductor. In a variant of that approach, Yi et al. grew a heterostructure consisting of layers of a magnetic topological insulator, (Bi,Sb)2Te3 doped with chromium, and

Science p. 646, 10.1126/science.adj4996

BIOGEOGRAPHY

Evolution of landscapes and plants Madagascar is a hotspot of biodiversity and unique species, both because of its long isolation from the mainland and its high rates of speciation. Liu et al. hypothesized that landscape evolution contributed to plant diversification by creating barriers and habitat heterogeneity, providing opportunities for allopatric speciation. They tested this hypothesis by modeling topographic dynamics on

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

602-B

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

Madagascar over the past 45 million years and found high plant species richness and diversification rates along the landward-shifting continental rift escarpment, with low north-south connectivity and high rates of local endemism. This work shows how relatively small-scale, transient landscape evolution can shape long-term biodiversity patterns. —BEL

improved B cell production. The approach also overcame the T cell differentiation block seen with RAG1 mutations in artificial thymic organoids. These findings support further study of HDRmediated gene editing for the correction of RAG1 mutations. —MLN Sci. Transl. Med. (2024) 10.1126/scitranslmed.adh8162

Science p. 653, 10.1126/science.adi0833

PALEOCLIMATE

Clay and climate How were atmospheric carbon dioxide levels regulated to allow habitable surface conditions since the Archean, and what were temperatures really like over that interval? Isson and Rauzi present an ensemble of oxygen isotope measurements from shale, iron oxide, carbonate, and chert over the past two billion years. These data show that the Proterozoic climate was temperate until temperatures began to fall during the early Paleozoic, around 500 million years ago. This drop in temperature occurred in concert with a decline in clay authigenesis and the rise of siliceous life, both of which would have contributed to declining levels of atmospheric carbon dioxide. —HJS Science p. 666, 10.1126/science.adg1366

GENE THERAPY

Knock it out (and in) Gene therapy is a promising treatment for patients with mutations in recombination activating gene 1 (RAG1), a cause of severe combined immunodeficiency, but there are safety concerns due to the resultant constitutive RAG1 expression. Castiello et al. used homology-directed repair (HDR)–mediated gene editing to correct RAG1 while preserving its physiological regulation. A combined knock-out and knock-in gene-editing strategy corrected RAG1 in patient hematopoietic stem and progenitor cells (HSPCs), leading to rescue of protein expression. When transplanted into humanized mice, these edited HSPCs led to 602-C

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

science.org SCIENCE

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

UPCYCLING

Nitrogen stitches for plastic mixtures

N

otwithstanding the well-known 1 through 7 labels in their characteristic triangles, separating plastic waste mixtures into their pure constituents is often impractical. There is particular interest in finding ways of better processing mixtures of polyethylene and polypropylene because they are both so abundant and are often used in tandem. Vialon et al. report that a nitrogen-rich azidotriazine molecule can decompose into reactive nitrenes at the melting temperatures of the polyolefin mixtures, stitching them together. The resultant extruded plastic has much improved ductility and creep resistance. —JSY J. Am. Chem. Soc. (2024) 10.1021/jacs.3c12303

A simple azidotriazine-based grafting agent enables direct upcycling of blends of polyethylene and polypropylene waste into high-performance materials.

regulatory T cell responses to maintain a tolerogenic intestinal milieu. —PNK

combine with other components to make cytoskeletal switches or logic gates. —LBR

J. Exp. Med. (2024)

Proc. Natl. Acad. Sci. U.S.A. (2024)

10.1084/jem.20230063

10.1073/pnas.2315992121

BIOENGINEERING

SCIENCE EDUCATION

Microtubule tracks on a chip

Hidden effects of science competitions

Microtubules are cytoskeletal structures in cells that form the mitotic spindle to allow chromosome separation during cell division. These filamentous structures also form tracks for transporting molecules within cells. Because studies in cells had previously revealed mechanisms that control microtubule subunit nucleation and branching, Zaferani et al. explored methods to build synthetic circuits of microtubular biopolymers. The authors used extracts of meiotic eggs from the African clawed toad (Xenopus laevis) to make patterns by controlling nucleation and physically affecting growth. For example, a chip decorated with microtubular curves and bifurcations could be used to form “microtubule diodes.” Microtubule chips could guide molecular motors or even

Science competitions are meant to encourage student interest in science and science careers, and evaluations typically focus on positive outcomes and successful competitors. Garrecht et al. took the opposite approach, using a latent change score model to examine the effects of early elimination on participants in the German Biology Olympiad. Results showed that most participants’ biology-related study and career task value remained stable from the first to the second round of the competition, and their expectancy of success in biology and biology careers developed positively. However, for participants placing high importance on advancing in the competition, elimination interfered with the development of study and career expectations. The authors encourage science competitions

602

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

to address participants’ values about science and science careers, especially in earlier rounds, and to develop innovative ways of providing feedback to improve post-elimination performance. —MMc J. Res. Sci. Teach. (2023)

in some heavily trawled regions. Limiting trawling will not only protect marine ecosystems, but could also be part of the solution for reducing carbon emissions. —SNV Front. Mar. Sci. (2024) 10.3389/fmars.2023.1125137

10.1002/tea.21901

PROTEIN CRYSTALS CARBON EMISSIONS

Trawling for problems It is well known that trawling, a fisheries practice in which nets are dragged across the sea floor, has disastrous impacts on marine ecosystems. Recent work has revealed that disturbance of seafloor sediments also releases carbon that contributes to climate-altering emissions. Atwood et al. used data-based predictions of global trawling activity from 1996 to 2020, in conjunction with a series of models, and estimated that about half the carbon stirred up was released into the atmosphere over the following 7 to 9 years. This translates into the equivalent of about 10% of the global emissions resulting from land-use change in 2020. The authors also predicted that trawling-related carbon release could lead to marine pH changes

Crystalline immunity In the laboratory, many proteins can be coaxed into forming small, delicate crystals. Natural protein crystallization is a much rarer phenomenon, but there are a few examples in which functional crystals form within cells or tissues. Heyndrickx et al. studied the proteins Ym1 and Ym2 from mice, which are produced by certain immune cells, and found that administering a crystalline form of Ym1, but not the soluble form, stimulated both innate and adaptive immune responses. These proteins are highly expressed in response to allergens and readily form crystals in vivo, but the physiological mechanisms and evolutionary origins of this phenomenon need more study. —MAF eLife (2023) 10.7554/eLife.90676

science.org SCIENCE

RES EARCH

RESEARCH ARTICLE SUMMARY



gated whether the number of fibroblasts is altered by aging and how a reduction in fibroblast number affects penile blood flow.

REPRODUCTION

Eduardo Linck Guimaraes, David Oliveira Dias, Wing Fung Hau, Anais Julien, Daniel Holl, Maria Garcia-Collado, Soniya Savant, Evelina Vågesjö, Mia Phillipson, Lars Jakobsson, Christian Göritz*

INTRODUCTION: Penile erection, a physiological process crucial for sexual function, relies on the intricate regulation of blood flow within the sponge-like vascular bed of the corpora cavernosa (CC). In the flaccid penis, sympathetic release of the vasoconstrictor norepinephrine maintains vascular smooth muscle cells tonically contracted, restricting penile blood flow. Upon sexual arousal, nitric oxide and acetylcholine are released from parasympathetic nerves, mediating vasodilation through the relaxation of vascular smooth muscle cells. The incoming blood fills the CC, leading to penile erection. Despite the recognized importance of endothelial and vascular smooth muscle cells in the erectile process, the vast population of fibroblasts in the CC has been largely overlooked.

RATIONALE: Fibroblasts constitute the largest cell population in the human CC, but their physiological functions remain largely unexplored. Our aim was to elucidate the contribution of CC fibroblasts to the regulation of penile blood flow. Characterization of CC fibroblasts through single-cell gene expression profiling and histological analysis in cleared tissue revealed their heterogeneity and integration in the erectile tissue. Using genetic targeting and optogenetic-induced fibroblast depolarization, we found that fibroblasts actively participate in the regulation of penile blood flow. Furthermore, by chronically altering erection frequency using chemogenetic modulation of brain regions responsible for arousal, we addressed the impact of erection recurrency on fibroblast number and blood flow regulation. Finally, we investi-

Notch activity

Erection frequency

Norepinephrine availability

Fibroblast number

NE NO

NE NO

Flaccid

Erected

Vasodilation

vSMC contraction stimulation via free NE

NE scavenging via fibroblasts

Fibroblasts mediate erectile activity–dependent modulation of penile blood flow. Notch signaling in CC fibroblasts decreases upon erection. Increased erection frequency leads to proliferation and higher number of penile fibroblasts, which in turn reduces the availability of the vasoconstrictor norepinephrine (NE), supporting penile erection. A chronically low erection frequency reduces fibroblast numbers and lowers penile blood flow. NO, nitric oxide; vSMC, vascular smooth muscle cell. Guimaraes et al., Science 383, 604 (2024)

9 February 2024

RESULTS: Our study revealed that fibroblasts in the CC play a pivotal role in supporting vasodilation by modulating norepinephrine availability. The efficacy of this process depends on the number of fibroblasts, which is regulated by erectile activity. Penile erection temporarily alters the spatial arrangement of cells throughout the CC, leading to downregulation of Notch signaling in fibroblasts. Inhibition of Notch signaling in fibroblasts leads to a substantial increase in fibroblast numbers, which can cause long-lasting erections characteristic of priapism. Constitutively active Notch signaling decreases fibroblast numbers and lowers penile blood perfusion. Boosting the frequency of erections reduces Notch signaling, increasing fibroblast numbers and promoting vasodilation. Conversely, a reduction in erection recurrency increases Notch signaling, decreasing the number of fibroblasts and diminishing penile blood perfusion. Aging, one of the major risk factors for erectile dysfunction, reduces the number of penile fibroblasts and limits penile blood perfusion. A reduction of penile fibroblasts in young animals mimics the penile blood flow phenotype of aged animals. CONCLUSION: Fibroblasts, previously regarded as static and homogeneous cells, are emerging as a dynamic and heterogeneous cell population. We discovered that CC fibroblasts act as key blood flow regulators, shifting the balance between vasodilators and the vasoconstrictor norepinephrine toward vasodilation. Notch signaling serves as a central hub coordinating fibroblast turnover, norepinephrine sensitivity, and, ultimately, the erectile process. The dynamic regulation of fibroblast numbers coupled to erection recurrency underscores the plasticity of erectile function. This positive feedback loop may exacerbate erectile dysfunction in chronic conditions such as aging or diabetes. Indeed, our observations in aged animals suggest a potential link between reduced fibroblast number and erectile dysfunction, highlighting the clinical relevance of understanding the cellular mechanisms of erection. Overall, this study provides a mechanism for modulation of penile erection and establishes a foundation for further research in the field of sexual health.



The list of author affiliations is available in the full article online. *Corresponding author. Email: [email protected] Cite this article as E. Linck Guimaraes et al., Science 383, eade8064 (2024). DOI: 10.1126/science.ade8064

READ THE FULL ARTICLE AT https://doi.org/10.1126/science.ade8064 1 of 1

Ä

Corpora cavernosa fibroblasts mediate penile erection

RES EARCH

RESEARCH ARTICLE



REPRODUCTION

Corpora cavernosa fibroblasts mediate penile erection Eduardo Linck Guimaraes1, David Oliveira Dias1, Wing Fung Hau1, Anais Julien1, Daniel Holl1, Maria Garcia-Collado2, Soniya Savant1, Evelina Vågesjö3, Mia Phillipson3, Lars Jakobsson2, Christian Göritz1* Penile erection is mediated by the corpora cavernosa, a trabecular-like vascular bed that enlarges upon vasodilation, but its regulation is not completely understood. Here, we show that perivascular fibroblasts in the corpora cavernosa support vasodilation by reducing norepinephrine availability. The effect on penile blood flow depends on the number of fibroblasts, which is regulated by erectile activity. Erection dynamically alters the positional arrangement of fibroblasts, temporarily downregulating Notch signaling. Inhibition of Notch increases fibroblast numbers and consequently raises penile blood flow. Continuous Notch activation lowers fibroblast numbers and reduces penile blood perfusion. Recurrent erections stimulate fibroblast proliferation and limit vasoconstriction, whereas aging reduces the number of fibroblasts and lowers penile blood flow. Our findings reveal adaptive, erectile activity-dependent modulation of penile blood flow by fibroblasts.

I

n addition to reproduction, sexual activity has several physiological and mental benefits that affect the health of adult humans. For men, sexual well-being largely depends on the ability to attain penile erections, which is compromised by aging and other risk factors (1, 2). Penile erection is mediated through vasodilation of the corpora cavernosa (CC), a trabecularlike vascular bed that extends from the base of the penis throughout the penile body into the glans penis (3). In the flaccid penis, sympathetic release of norepinephrine maintains arterial and trabecular vascular smooth muscle cells (vSMCs) tonically contracted, restricting penile blood flow to a basal level. Upon sexual arousal, nitric oxide (NO) and acetylcholine are released from parasympathetic nerves and mediate vasodilation through relaxation of vSMCs along the central arteries and within the CC. The incoming blood fills the CC, leading to an expansion of the organ. Shear stress– induced endothelial NO maintains the erection through continuous relaxation of vSMCs in the CC (4). The balance between vasodilators and the vasoconstrictor norepinephrine is decisive for the establishment and maintenance of penile erection. Recently, fibroblasts have been recognized in the human CC (5, 6), but whether they have a role in blood flow regulation has not been addressed. 1

Department of Cell and Molecular Biology, Karolinska Institutet, 171 77 Stockholm, Sweden. 2Department of Medical Biochemistry and Biophysics, Division of Vascular Biology, Karolinska Institutet, 171 77 Stockholm, Sweden. 3 Department of Medical Cell Biology, Division of Integrative Physiology, Uppsala University, 751 23 Uppsala, Sweden. *Corresponding author. Email: [email protected]

Guimaraes et al., Science 383, eade8064 (2024)

The CC contains two subsets of perivascular fibroblasts

Here, we identified a large population of cells throughout the C C in mice that express the transporter solute carrier family 1, member 3 (SLC1A3, synonym GLAST; gene name Slc1a3) (fig. S1, A to C), a marker that has been previously used to label perivascular cells in the central nervous system (7–9). To mark SLC1A3+ cells, we used Slc1a3-CreERT2 transgenic mice (10) carrying a Rosa26-tdTomato reporter allele (fig. S1D). Tamoxifen-induced genetic recombination in adult Slc1a3-CreERT2;Rosa26-tdTomato mice led to tdTomato expression throughout the CC (Fig. 1, A and B, and fig. S1, E to G). We rarely found tdTomato+ cells close to the dorsal vein and arteries (Fig. 1B and fig. S1, H to L). To characterize the gene expression profile of SLC1A3+ cells from the adult mouse penis, we isolated tdTomato+ cells from recombined Slc1a3CreERT2;Rosa26-tdTomato mice by fluorescentactivated cell sorting (FACS) and performed single-cell RNA sequencing (fig. S2A). Sorted cells expressed Slc1a3 and fibroblasts markers such as Col1a1, Pdgfra, and S100a4, but little to no vSMC and endothelial cell markers (Fig. 1, C and D, and fig. S2, B to D). We confirmed the expression of the fibroblast markers plateletderived growth factor receptor a (PDGFRa) and S100 calcium binding protein A4 (S100A4) in tdTomato+ cells at the protein level (Fig. 1, E to G). Using unbiased clustering, we showed that SLC1A3+ fibroblasts encompass three subpopulations, which can be identified by the expression of Dnah5 for cluster 1, Cxcl12 for cluster 2, and Gja1 for cluster 3 (Fig. 1, C and D, and fig. S2, E to G). Cells from clusters 1 and 2 distributed throughout the subtunica and trabecular regions of the CC close to the vascula-

9 February 2024

ture, whereas cells from cluster 3 localized mainly within the tunica albuginea, the largely avascular outer region of the penis (fig. S1, H to L, and fig. S2, H to J). Intravital confocal and multiphoton imaging of mice injected intravenously with fluorescein isothiocyanate (FITC)–labeled dextran highlighted the high density of tdTomato+ cells and their proximity to circulating blood throughout the CC under physiological conditions (Fig. 1, H to J, and movie S1). Surface rendering of optically cleared tissue and immunohistochemical analyses confirmed proximity to the vasculature and clear distinction of tdTomato+ fibroblasts from transgelin+ (TAGLN) vSMCs and podocalyxin+ (PODXL) endothelial cells in the CC (Fig. 1, K to M, and fig. S3, A to L). Compared with both vSMCs and endothelial cells, the number of tdTomato+ fibroblasts was three times higher in the CC (Fig. 1N). Altogether, we identified two SLC1A3-expressing perivascular fibroblast subsets in the CC. CC fibroblasts mediate vasodilation

Optogenetic-induced membrane depolarization of mural cells has been used to study their involvement in blood flow regulation in the brain (11, 12). We used this strategy to investigate the potential role of SLC1A3+ fibroblasts in penile blood flow modulation. For comparison, we applied the same approach to vSMCs, which are known to constrict blood vessels upon optogenetic-induced membrane depolarization (13). To this end, we generated doubletransgenic mice expressing the light-gated cation channel channelrhodopsin-2 (ChR2) fused to tdTomato fluorescent protein (14) specifically in penile fibroblasts (Slc1a3-CreERT2;Rosa26ChR2-tdTomato) or vSMCs (Tagln-CreERT2; Rosa26-ChR2-tdTomato) after tamoxifen-induced recombination (fig. S4A). As control, we used Cre– animals from the same lines, which also received tamoxifen without leading to ChR2tdTomato expression. Blood flow was assessed noninvasively using high-resolution laser speckle contrast imaging, which detects red blood cell movement (fig. S4B) (15). Focal exposure of the flaccid glans penis of Tagln-CreERT2;Rosa26ChR2-tdTomato mice to blue light (473 nm, 4.5 mW, 100 Hz, 8 ms pulse width, 8 s on/3 s off) activated vSMCs and resulted in immediate blood flow decrease in the exposed area. Termination of light exposure rapidly reestablished basal blood flow (Fig. 1O). These results are consistent with a previous report showing that ChR2-induced depolarization of vSMCs led to calcium-mediated vasoconstriction (13), validating our approach. By contrast, exposure of the flaccid glans penis or penile body of Slc1a3CreERT2;Rosa26-ChR2-tdTomato mice with the same regimen of blue light activated fibroblasts and induced a steady rise in blood flow compared with light-exposed control mice. After termination of photoactivation, blood flow 1 of 11

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

Fig. 1. Perivascular fibroblasts of the CC mediate vasodilation in the penis. (A) Schematics depicting anatomical structures of the mouse penis. (B) Transverse section of the adult mouse penis showing vSMCs (TAGLN+ in green), endothelium (PODXL+ in cyan), and recombined perivascular fibroblasts (tdTomato+ in red) in the C C of Slc1a3-CreERT2;R26R-tdTomato mice. (C) Uniform manifold approximation and projection (UMAP) representation of sorted tdTomato+ cells from Slc1a3-CreERT2;R26R-tdTomato adult mouse penis. (D) Violin plots depicting the expression of lineage markers of fibroblasts, vSMCs, and endothelial cells. (E to G) TdTomato+ cells expressing the fibroblast markers S100A4 (E) and PDGFRa (F). (G) Merged image showing coexpression of TdTomato, S100A4, and PDGFRa. (H and I) Intravital microscopy time-lapse images showing the penile vasculature (H) and the high density of tdTomato+ cells in the native C C and limited presence around the dorsal vasculature (DV) (I). (J) Intravital multiphoton microscopy image showing the proximity of tdTomato+ fibroblasts to circulating blood (FITC-dextran, i.v.). (K to M) Surface rendering and threedimensional (3D) visualization of the cleared C C showing tdTomato+ perivascular Guimaraes et al., Science 383, eade8064 (2024)

9 February 2024

cells [(K), in red], vSMCs [(L), TAGLN+ cells in green and endothelial cells (PODXL+) cells in cyan], and (M) the proximity between the three cell types. (N) Quantification of tdTomato+ cells, vSMCs, and endothelial cells in the C C. (O and P) Basal blood flow of the glans penis in response to optogenetic-induced depolarization of vSMCs [(O), Tagln;R26R-ChR2 mice (n = 3)] and perivascular fibroblasts [(P), Slc1a3;R26R-ChR2 mice (n = 3)]. R26R-ChR2 mice were used as a control (both n = 3). Blue light (473 nm) was delivered focally on the glans penis (4.5 mW, 100 Hz, 8 ms pulse width, 8 s on/3 s off). (Q) Changes in FITCdextran lumen cross-sectional area in response to intravital two-photon optogenetic activation of tdTomato+ fibroblasts in the C C of Slc1a3;R26R-ChR2 (n = 4) and R26R-ChR2 control mice (n = 3). Data are shown as mean ± SEM. For (N), one-way ANOVA followed by Holm-Šidák multiple-comparisons test was used. For (O) to (Q), two-way ANOVA was used. DA, dorsal artery; TA, tunica albuginea; U, urethra; Trb, trabecula. Cell nuclei are labeled with DAPI. Scale bars: (B), 200 mm; (E) to (G) and (J), 25 mm; (H) and (I), 100 mm; (K) and (L), 50 mm; (M), 10 mm. 2 of 11

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

declined slowly (Fig. 1P and fig. S4C). To improve spatial precision in photoactivation and gain higher resolution, we performed multiphoton imaging and optogenetic stimulation of CC fibroblasts in the penile body (fig. S4D). Focal optogenetic activation of ChR2-tdTomato+ fibroblasts in the trabecular and subtunica regions of the CC increased vasodilation in the exposed area compared with control mice without ChR2 expression, as determined by the FITCdextran–filled trabecular lumen area (Fig. 1Q and fig. S4, E and F). Compared with the slow and steady increase in blood flow mediated by fibroblast depolarization, exposure of the flaccid glans penis to NO or acetylcholine, the main parasympathetic vasodilatory effectors, caused a strong and immediate increase in blood flow (fig. S5, A and B). The distinct kinetics suggest that optogenetic-induced membrane depolarization of fibroblasts does not cause parasympathetic stimulation but rather continuously reduces basal vasoconstriction. CC fibroblasts mediate vasodilation by reduction of norepinephrine availability

In the flaccid penis, blood flow is maintained at a basal level through the presence of the vasoconstrictor norepinephrine (16). A steady reduction of norepinephrine levels in response to continuous fibroblast depolarization could

explain the observed increase in blood flow. Indeed, we found that both tdTomato+ fibroblast subsets in the CC expressed the transporter solute carrier family 6, member 2 (SLC6A2) (fig. S6, A to D), which actively transports norepinephrine into cells, preventing receptor binding (17). In addition, tdTomato+ cells in the CC expressed monoamine oxidase A (MAO-A), an essential enzyme for norepinephrine degradation (18) (fig. S6, E and F). SLC6A2-mediated norepinephrine uptake requires cell membrane localization of the transporter, which is induced by membrane depolarization (19, 20). Fibroblasts isolated from the CC maintained SLC6A2 expression in primary cultures (Fig. 2A) and, in response to depolarization, significantly (P < 0.0001) increased plasma membrane localization of SLC6A2 (fig. S7, A to G). Consistent with this, optogenetic-induced depolarization of primary cultured CC fibroblasts stimulated the uptake of the fluorescent norepinephrine analog FFN270 (21) (Fig. 2, B and C). Application of nisoxetine (NES), a specific SLC6A2 inhibitor (17), prevented depolarization-induced FFN270 uptake (Fig. 2, B and C). To investigate whether SLC6A2 plays a role in fibroblast-mediated vasodilation in vivo, we repeated the optogeneticinduced depolarization of penile fibroblasts in the presence of NES. NES injection into the

Fig. 2. Penile fibroblasts mediate vasodilation through the norepinephrine transporter SLC6A2. (A) Expression of the norepinephrine transporter SLC6A2 in primary cultures of tdTomato+ penile fibroblasts. (B and C) Representative images (B) and quantification (C) of the uptake of the norepinephrine fluorescent analog FFN270 by tdTomato+ penile fibroblasts isolated from Slc1a3;R26R-tdTomato (n = 3) and Slc1a3;R26R-ChR2 (n = 3) mice and Slc1a3;R26R-ChR2 fibroblasts treated with the SLC6A2 inhibitor NES (Slc1a3;R26RChR2+NES, n = 3) after exposure to blue light (470 nm) in vitro. (D) TdTomato+ fibroblasts in the adult mouse CC express the norepinephrine transporter SLC6A2. (E) Penile blood flow in response to focal illumination of the glans penis with blue light in R26R-ChR2 (control; n = 3), Slc1a3;R26R-ChR2 (n = 4), and (Slc1a3;R26R-ChR2+NES) (n = 4) mice, represented as percentage of baseline blood flow. (F) Penile blood flow in response to focal illumination of the glans penis with blue light and administration of the vasoconstrictor agent phenylephrine in R26R-ChR2 (n = 4), Slc1a3;R26R-ChR2 (n = 5), and Slc1a3; R26R-ChR2+NES (n = 4) mice after NO-induced vasodilatation. For (E) and (F), blue light: 473 nm, 4.5 mW, 100 Hz, 8 ms pulse width, 8 s on/3 s off). Data are shown Guimaraes et al., Science 383, eade8064 (2024)

9 February 2024

glans penis completely abrogated the rise in blood flow induced by fibroblast photoactivation (Fig. 2, D and E, and fig. S8A). To directly test the regulation of norepinephrine availability by SLC1A3+ fibroblasts and its impact on blood flow, we pretreated the penis with a NO donor to induce robust blood flow and assessed the vasoconstriction potency of the norepinephrine analog phenylephrine, which is also transported by SLC6A2 (22). After stabilization of NO-induced blood flow, the penis was simultaneously exposed to blue light and phenylephrine. In control animals with no ChR2 expression, phenylephrine induced immediate vasoconstriction, which persisted for the entire duration of the experiment. By contrast, animals with ChR2-expressing fibroblasts showed a steady increase in blood flow upon light exposure, reflecting reduced bioavailability of phenylephrine. This effect was completely abolished in the presence of NES (Fig. 2F and fig. S8B), suggesting that penile fibroblasts facilitate vasodilation by SLC6A2mediated uptake of norepinephrine. Penile erection dynamically alters the spatial arrangement of fibroblasts and vSMCs

Next, we used optical tissue clearing and compared the cellular arrangement of SLC1A3+ fibroblasts, vSMCs, and endothelial cells in the

as mean ± SEM. For (C), one-way ANOVA followed by Holm-Šidák multiplecomparisons test was used. For (E) and (F), two-way ANOVA with Bonferroni’s multiple-comparisons post hoc test was used. P values represent comparison of the last time point between different groups. Cell nuclei are labeled with DAPI. Scale bars: (A) and (B), 10 mm; (D), 25 mm. 3 of 11

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

CC of the flaccid and erected penis (Fig. 3, A to F, and fig. S9, A to F). In the flaccid state, tdTomato+ fibroblasts and endothelial cells showed a high cellular density, prominent in the subtunica area along the outer rim of the CC (Fig. 3, A and C, and fig. S9, A, C, and D). vSMC fibers polarized in various directions and had large contact areas with fibroblasts (Fig. 3, A, C, and D). Erection was induced by cavernous nerve stimulation (23) and, compared with the flaccid state, the dilated penis exhibited a largely expanded trabecular lumen, whereas the tunica albuginea was compressed (Fig. 3B and fig. S9B). Fibroblasts, vSMC fibers, and endothelial cells appeared more elongated and aligned radially (Fig. 3, B, E, and F, and fig. S9, B, E, and F). Although the contact area between endothelial cells and fibroblasts did not change (P = 0.52) (fig. S9G),

the contact between vSMCs and fibroblasts was significantly (P = 0.021) reduced during erection (Fig. 3G), indicating that penile erection dynamically alters proximity between perivascular cell types. Notch signaling is transiently active in CC fibroblasts

The Notch signaling pathway orchestrates cell turnover and blood flow through dynamic changes in cell-to-cell contact in the bone (24, 25) and could serve as a readout for the transition between flaccid and erect penile states. We found expression of the Notch1 and Notch2 receptors, the Notch ligand Jagged 1 (Jag1), as well as the Notch downstream effector recombination signal binding protein for immunoglobulin kappa J region (RBP-Jk, synonym CBF1; gene name Rbpj), in Cxcl12+

(cluster 2) fibroblasts throughout the CC (fig. S10, A to P), indicating that CC Cxcl12+ fibroblasts can both receive and transmit Notch signals. Active Notch signaling can be assessed at single-cell resolution in vivo using transgenic CBF-H2B-Venus reporter mice. In these mice, binding of the Notch intracellular domain (NICD), RBP-Jk, and mastermind-like (MAML) DNA binding complex to cis-regulatory regions leads to expression of the nuclear localized Venus reporter protein (fig. S11A) (26). We detected recently active or ongoing Notch signaling in SLC1A3+ fibroblasts, endothelial cells, and vSMCs throughout the CC (Fig. 3H and fig. S11, B to R). When analyzing the penis of adult Slc1a3CreERT2;Rosa26-tdTomato;CBF-H2B-Venus mice and assessing Venus mRNA expression in combination with Venus protein expression, we found that 10 ± 2.15% of tdTomato+ cells

Fig. 3. Penile erection causes dynamic changes in cell-to-cell contact and reduces Notch signaling in fibroblasts of the CC. (A and B) Cellular arrangement of vSMCs (TAGLN+ in green) and fibroblasts (tdTomato+ in red) in the flaccid (A) and erected (B) penis (optically cleared 200-mm-thick transverse sections). (C to F) Representative images showing the spatial arrangement of fibroblasts (C) in relation to vSMCs (D) in the flaccid CC and fibroblasts (E) and vSMCs (F) in the erected CC. Yellow represents areas of fibroblastvSMC contact. (G) Quantification of tdTomato+ fibroblasts associated with vSMCs in the CC in the flaccid and erected state (flaccid, n = 4; erected, n = 4). (H) Expression of the Notch reporter CBF:H2BVenus (white arrowheads) in SLC1A3+ CC fibroblasts. (I) Expression of H2B-Venus mRNA and protein in tdTomato+ fibroblasts as percentage of the total tdTomato+ fibroblast population in the CC. (J and K) H2BVenus mRNA (J) or Hes1 mRNA (K) expression in tdTomato+ fibroblasts of control and cavernous nerve (CN)–stimulated penises, represented as the number of mRNA puncta in tdTomato+Venus mRNA+ or tdTomato+ cells, respectively. Data are shown as mean ± SEM. For (G), (J), and (K), Mann-Whitney test was used. Cell nuclei are labeled with DAPI. Scale bars: (A) and (B), 200 mm; (C) to (F) and (H), 25 mm. Guimaraes et al., Science 383, eade8064 (2024)

9 February 2024

4 of 11

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

expressed the Venus protein, 16 ± 0.9% expressed the Venus mRNA, whereas only 2 ± 0.4% expressed both (Fig. 3I). These results indicate that Notch signaling is transiently activated in fibroblasts of the CC. Frequent erections lead to decreased Notch signaling in CC fibroblasts

To investigate whether penile erection modulates Notch signaling in fibroblasts of the CC, we performed cavernous nerve stimulation in recombined Slc1a3-CreERT2;Rosa26-tdTomato; CBF-H2B-Venus mice. After 1 hour of repeated stimulation of the cavernous nerve (fig. S12A), Venus mRNA expression was significantly (P = 0.019) reduced in tdTomato+ fibroblasts of the CC (Fig. 3J and fig. S12B). Consistent with this, we also detected down-regulation of Notch2, Jag1, and the downstream Notch effector Hes1 in tdTomato+ fibroblasts after cavernous nerve stimulation, whereas genes unrelated to Notch signaling remained unaltered (Fig. 3K and fig. S12, C to K). Our results show that Notch signaling is dynamically regulated in fibroblasts of the CC and reduced upon penile erection. Inhibition of Notch signaling in CC fibroblasts increases their cell number and raises penile blood flow

To investigate whether modulation of Notch signaling affects penile blood flow, we experimentally reduced Notch signaling in fibroblasts by conditional deletion of RBP-Jk, an obligatory DNA-binding protein for canonical Notch signaling (27), using Slc1a3-CreERT2; Rosa26-tdTomato mice homozygous for conditional Rbp-Jk null alleles (RBP-Jkfl/fl) (28). In a complementary experiment, we constitutively activated Notch signaling in penile fibroblasts using Slc1a3-CreERT2;Rosa26-tdTomato mice crossed to transgenic Rosa26Notch mice (29), which allow Cre-mediated expression of NICD upon tamoxifen administration (fig. S13A). Two months after RBP-Jk deletion, basal blood flow in the penile body and glans penis were significantly (P = 0.0003 and P = 0.004, respectively) increased compared with control animals with unaltered Notch signaling. In turn, NICD animals with constitutively active Notch signaling had low basal blood flow compared with RBP-Jk fl/fl animals (Fig. 4, A and B, and fig. S13, B and C). Tissue analyses revealed that deletion of RBP-Jk significantly (P = 0.0063) increased the number of tdTomato+ fibroblasts in the CC, whereas NICD expression reduced the number of tdTomato+ cells compared with control mice (Fig. 4, C and D). The number, cell surface, and cell volume of endothelial cells and vSMCs were not altered by Notch signaling modulation in fibroblasts (fig. S13, D to K). Brain-specific deletion of RBP-Jk in SLC1A3-expressing perivascular cells (fig. S14, A to E) did not alter basal penile blood flow nor response to phenylephrine (fig. S14, Guimaraes et al., Science 383, eade8064 (2024)

F to H). Altogether, these results show that reduced Notch signaling in fibroblasts of the CC increased their number and led to increased blood perfusion. In turn, constitutive activation of Notch signaling reduced the number of fibroblasts in the CC and decreased blood perfusion. Modulation of Notch signaling did not change the average expression level of SLC6A2 in tdTomato+ fibroblasts (fig. S15, A and B) but did alter the overall abundance of SLC6A2 in the CC through the change in fibroblast numbers (Fig. 4E and fig. S15C). The number of CC fibroblasts regulates penile blood flow

To test whether the number of SLC1A3+ fibroblasts in the CC affects norepinephrine-mediated blood flow regulation, we deleted RBP-Jk or induced constitutive expression of NICD in penile fibroblasts. Two months after recombination, we exposed the penis to phenylephrine after pretreatment with a NO donor. In control mice with unaltered Notch signaling, phenylephrine sharply reduced penile blood flow, efficiently counteracting NO-induced vasodilation for the duration of the experiment. By contrast, phenylephrine-induced blood flow reduction was less effective in RBP-Jkfl/fl animals, which present an increased number of CC fibroblasts. Indeed, we noted that the vasoconstrictor effect of phenylephrine was completely abolished after 5 min, and blood flow steadily increased thereafter. The effectiveness of phenylephrine in reducing blood flow was completely reestablished in RBP-Jkfl/fl mice in the presence of the SLC6A2 inhibitor NES. In NICD-expressing animals, which present fewer CC fibroblasts, phenylephrine induced immediate, long-lasting, vasoconstriction (Fig. 4F). Norepinephrine induces vSMC contraction through Rho-associated protein kinase (ROCK) activation and consequent phosphorylation of myosin phosphatase target subunit-1 (MYPT-1) and C-kinase potentiated protein phosphatase-1 inhibitor (CPI-17) (30–33). We observed that the levels of MYPT-1 and CPI-17 phosphorylation were decreased in RBP-Jk fl/fl mice but increased in the NICD group and were inversely correlated with the number of tdTomato+ cells (fig. S16, A to F), further indicating that CC fibroblasts can limit norepinephrine availability. Consistent with these findings, we observed that a subset of RBP-Jk deleted animals presented atypical long-lasting erections, featuring a largely increased glans penis and reduced blood perfusion, characteristic of ischemic priapism (Fig. 4, G to I) (34). This phenotype was consistently observed across several experiments from 2 months after RBP-Jk deletion. Because the efficiency of tamoxifeninducible Cre-mediated recombination can vary (8), it may explain why this phenotype only appeared in a subset of mice. Indeed, animals presenting priapism contained a higher number of tdTomato+ fibroblasts in the CC compared

9 February 2024

with nonpriapic animals of the same genotype and age (Fig. 4J), indicating that penile fibroblasts can effectively diminish vasoconstriction. The frequency of erections determines fibroblast proliferation

Because inhibition of Notch signaling increases the fibroblast cell population (Fig. 4, C and D), and repeated stimulation of erection can actively modulate Notch signaling (Fig. 3, J and K, and fig. S12, A to G), we tested whether the frequency of erections affects the number of fibroblasts in the CC and penile blood flow. For that, we induced or inhibited penile erection in a physiological context by chemogenetic activation or inhibition of brain neurons in areas known to control erection, such as the medial preoptic area (mPOA) and paraventricular nucleus of the hypothalamus (PVN) (Fig. 5A), (35–37). Neurons in the mPOA project to oxytocin (OXT)–expressing neurons in the PVN (38) that are vital for sexual behavior and induce penile erection when activated by glutamate and analogs (37, 39, 40). Using viral constructs, we expressed hM3D(Gq)-mCherry or hM4D(Gi)mCherry, an excitatory or an inhibitory designer receptor exclusively activated by designer drugs (eDREADD or iDREADD, respectively), or the reporter gene mCherry alone (mock), in calcium/ calmodulin–dependent protein kinase IIa–positive (CaMKIIa+) neurons of the mPOA and PVN (Fig. 5A and fig. S17, A to C). CaMKIIa+ neurons showed a similar level of mCherry expression in animals transduced with mock, eDREADD, and iDREADD viral constructs (fig. S17D). Neurons expressing mCherry or OXT were found intermingled in the same region but did not overlap (fig. S17, B and E to G). We confirmed that, compared with the mock-transduced group, activation of eDREADD or iDREADD by intraperitoneal injection of the DREADD ligand clozapine N-oxide (CNO) increased or decreased neuronal activity in mCherry+ and OXT+ neurons, respectively, as shown by c-fos expression (fig. S17, E to K). A single injection of CNO induced a fourfold increase or a fourfold decrease in the number of erections per hour in freely moving animals expressing eDREADD or iDREADD, compared with mock-transduced animals, respectively (Fig. 5B). To determine whether chronic increase or decrease in the frequency of erections modulates the number of fibroblasts in the CC, we administrated CNO daily (5 mg/kg in saline) for 4 weeks to recombined Slc1a3CreERT2;Rosa26-tdTomato mice transduced with eDREADD, iDREADD, or mock viral constructs in the mPOA and PVN while supplying the thymidine analog 5-ethynyl-2′-deoxyuridine (EdU) in the drinking water (Fig. 5A). We observed that boosting the frequency of penile erections significantly increased the generation of tdTomato+ fibroblasts, as measured by EdU incorporation, as well as the total number 5 of 11

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

B

150 100 50

P=0.006

800

400

0

fl/fl

tdTomato DAPI

0 l

20

0

2

4

6

8

D IC ;N

a3

10

minutes

G

H Flaccid

Priapic

Flaccid

Priapic

P=0.045

150

J 1800 1200

100

50

600

0 l

c ia pi Sl

c1

a3 ;

Pr

Fl ac c

id

0

Blood flow

P=0.0054

R bp Sl -J c1 κ fl/f a3 l +p ;R ria b pi p-J sm κ fl /f

Sl

I

No. tdTomato+ cells / section

c1

;R

a3

;N

on

tro l

fl/ fl

-J κ ;R bp

60 40

IC

tro

D

l

κ fl/f

a3

80

0

C

D

100 PMyc mice were dying of cancer localized to the thyroid (fig. S7B). To date, we have been unsuccessful in activating the Ascl1CreERT2 allele specifically in the lungs, while sparing the trachea and thyroid. We therefore isolated tdTom+ cells from the airways of four distinct genotypes of mice combining Ascl1 lineage–traced Myc with loss of Rb1, Trp53, or both tumor supGardner et al., Science 383, eadj1415 (2024)

102

50 25

Myc 0 100 200 300 Days (post-TAM)

10

0

10-2

SpcCreERT2

SpcCreERT2 EGFR WT Myc

10-4

75 50 25

EGFR 0 100 200 300 Days (post-TAM)

(circles) on AT2 (SpcCreERT2; red) or PNEC (Ascl1CreERT2; blue) cells in the airway over time; n = 3 mice per time point. Control traces (tdTomato only; WT) are shown as gray squares; **P < 0.01, ****P < 0.0001. (D) Long-term survival for cohorts shown in (C); (top) Ascl1CreERT2 > EGFRL858R (n = 13) or MycT58A (n = 15) and (bottom) SpcCreERT2 > EGFRL858R (n = 14) or MycT58A (n = 12). Mice having a single copy of Rosa26LSL-tdTom and Rosa26LSL-rtTA3 were maintained on DOX chow throughout studies investigating lineage-trace allele-mediated expression of Tg.TetO-EGFRL858R.

pressor genes. We expanded cells ex vivo using organotypic culture conditions and engrafted equivalent cell numbers into the flanks of athymic mice. Combined loss of Rb1 and Trp53 accelerated the growth of these Mycdriven tumors, but all genotypes were sufficient to form transplantable cancers. Moreover, all tumors had a similar histologic appearance consistent with high-grade neuroendocrine cancer (fig. S7C). These data suggest that PNECs could be transformed by Myc (alone) if expanded ex vivo, but this experiment did not demonstrate that Myc was required for tumor maintenance. To test whether PNEC-derived tumors are dependent on Myc, we sorted tdTom+, rtTA3expressing PNECs from the airway and infected cells with lentiviruses containing tetracycline promoter–driven Myc constructs. Coupling rtTA3 expression to the lineage trace (Ascl1CreERT2) removed the likelihood of infecting lineagenegative cells present as contaminants. Although these cells were sparse (fig. S8), we could consistently generate organoid cultures from as few as 50 cells when DOX was present in the culture to drive the transcription of Myc. Removal of DOX resulted in near complete growth suppression of organoids expressing MycWT as compared with MycT58A (fig. S9, A to C), a long-lived version of Myc (42) (fig. S9C). Engrafting PNECs expressing inducible MycWT into athymic mice demonstrated DOX-dependent tumor growth. Removing DOX from tumor-

9 February 2024

Myc

100

0

0 2 4 6 Weeks (post-TAM)

ee

ks ee

ee

w 4

w 2

EGFR

75

Ad5.Spc-Cre

6

10

-3

100

0

****

10

EGFR

-5

-1

-2

WT

Ascl1CreERT2

0 2 4 6 Weeks (post-TAM)

** 10

Myc

10-1

Survival (%)

10

0

ks

tdTomato % (viable)

PNEC

D

Ascl1CreERT2

**

tdTomato % (viable)

RPMT

C

Survival (%)

A

bearing mice dramatically reduced tumor volumes over the course of several weeks (fig. S9D), with residual fibrotic tumor tissue comprising noncycling cells (fig. S9E). Consistent with their rapid proliferation, Myc-driven PNEC organoids were sensitive to compounds that exacerbated replication stress, such as topoisomerase inhibitors (etoposide), as well as inhibitors of enzymes required for cell-cycle progression, including Cdk4/6 (palbociclib) and Wee1 (adavosertib) (fig. S9, F and G). Additionally, direct inhibition of the Myc-Max interface using the small molecule MYCi975 (43) provided similar reduction in organoid growth as compared with reduction resulting from DOX removal (fig. S9F). AT2 cells are refractory to transformation by Myc

Although Myc alone may be sufficient to drive transformation and expansion of PNECs, our earlier results suggested that it was insufficient to transform the AT2 lineage (Fig. 3C). To further investigate what underlies this bottleneck, we established AT2 organoid cultures using recently published methods (44) from lineage-traced, WT, or MycT58A-expressing cells. These organoids expanded rapidly as compared with their WT counterparts but were unsustainable beyond three passages (fig. S10A). In early time points after the expression of MycT58A in AT2 cells in vivo, we noted incorporation of EdU in tdTom+ cells. However, 5 of 11

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

1 year after the initiation of the trace, tdTom+ cells in the airway failed to incorporate EdU, suggesting that they were no longer proliferative or were eliminated (fig. S10B). Ex vivo, AT2 organoids expressing MycT58A demonstrated increased DNA damage sensing, replication stress, and markers of programmed cell death as compared with WT (fig. S10C). It is unlikely that this resulted from an excess of Myc protein, because levels were lower in AT2 cells compared with levels tolerated in PNEC organoids (45) (fig. S10D). Lastly, consistent with the observation that Ras signaling through phosphatidylinositol-3-kinase (PI3K) relieves Myc-induced apoptosis (46), we likewise observed that Myc significantly accelerated oncogenic EGFR-driven LUAD, implying that enhanced signaling via the EGFR>Ras>Mek>Erk pathway can relieve intolerance to Myc in the AT2 lineage (fig. S10, E and F). Although our data suggest that the different oncogenes driving AT2 and PNEC lineages are in stark contrast, it remained unclear whether other epithelial airway cells can transform in response to Myc alone. To address this, we performed a generalized, conditional trace using an Nkx2.1CreERT2 allele that will drive Cremediated recombination in a broad range of cell types derived from the anterior foregut endoderm, including the trachea, pituitary, thyroid, and most of the lung (47). At early time points, lineage-labeled tissues within the thyroid (both Ascl1+ and Ascl1–) expanded after Myc expression as compared with tissues in control mice (fig. S11A); however, at later time points there was outgrowth of tdTom+/Ascl1+ cells in bronchioles not observed in WT controls (fig. S11B). Together, these data suggest that the Ascl1+ PNEC is distinct in its tolerance to Myc, but we had not yet explained how intolerance to Myc could be overcome in the AT2 lineage during HT. Deletion of Pten removes a barrier to Myc transformation

Genes up-regulated in tumor cells as they escaped the bottleneck to HT in our ERPMT model (termed “breakout”; fig. S12A) were notably associated with PI3K signaling (fig. S12, B and C) as compared with cells found in the bottleneck and not yet adopting neuroendocrine fate (48). Thus, we asked whether increasing PI3K-dependent Akt signaling through deletion of Pten would enable MycT58A-driven transformation in an AT2 lineage trace. Notably, we observed fully penetrant Myc-driven transformation in an AT2 cell (using SpcCreERT2) when one copy of Pten was inactivated (Spc>Pten;Myc; Fig. 4A). At the median period when animals in the Spc>Pten;Myc cohort were moribund, lungs from Spc>Myc mice (PtenWT/WT) showed no evidence of macroscopic disease (Fig. 4B). Similar observations were made when combining Myc expression with Gardner et al., Science 383, eadj1415 (2024)

a conditionally active, mutant PI3K allele (E545K) or inactivating both copies of Pten (PtenFl/Fl; Fig. 4B). Examining Spc>Pten;Myc animals 3 months after recombination reinforced the observation that lesions were variable in their frequency, size, and histologic appearance (fig. S13). However, they were Ascl1-negative and glandular in appearance, suggesting that they were adenocarcinoma-like and excluding the likelihood of SCLC, squamous, or mixed histology (fig. S13). AT2 lineage markers, including prosurfactant protein C (pro-SPC) and MHC Class II, were low or absent, and basal epithelial markers such as Sox2 and keratin 18 were variably expressed (fig. S13). Using scRNA-seq, we also observed a notable increase in the highly undifferentiated, basal stem-like state that followed combined loss of Pten and expression of Myc in AT2 cells—not seen with either genetic perturbation alone (Fig. 4C). Expectedly, this enrichment in the undifferentiated state was also associated with greater Myc transcriptional output (Fig. 4D). Pulmonary basal cells efficiently generate SCLC

The basal stem-like program associated with AT2 cells capable of adaptation to Myc supported the possibility that an intermediate state during HT may be basal-like. More generally, this raised the possibility that the basal cell may serve as a cell of origin for SCLC, as speculated by others (49). However, targeting basal cells in mouse models is limited by their anatomic location, noted to be refractory to viral infections delivering Cre (50). Indeed, we observed an absence of tdTom+ cells in the lungs of mice 1 month after labeling when using a Krt5CreERT2 allele to target basal cells (fig. S14A). However, following regeneration in the proximal lung induced by naphthalene damage of secretory cells (51, 52), tdTom+ cells were detected within the lungs of mice, without a significant change in the trachea or thymus (fig. S14B). As basal cells are known to serve as multipotent progenitors (fig. S14C) (51–53), we then crossed mice to delete Rb1 and/or Trp53 from the basal lineage trace and found that Rb1 loss alone was sufficient to skew cells toward a neuroendocrine fate (fig. S14D). Consistent with this, we observed fully penetrant neuroendocrine tumorigenesis in mice that have lost both Rb1 and Trp53 with (Krt5>RPMT) or without (Krt5>RPT) expression of Myc from the basal lineage (fig. S14E); however, tumors arose with significantly shorter latency in mice expressing Myc. Expression of the conditional tdTomato reporter allele was easily observed throughout the keratinized epithelium of mice after recombination, but tumorigenesis was restricted to the proximal airway (fig. S14, F and H). Histologically, both models produced SCLC-like tumors; however, we noted that most tumors were adjacent to or within the thymus and did not invade the lungs unless we damaged

9 February 2024

lungs with naphthalene (fig. S14, G to I). These data suggest that basal cells may serve not only as cells of origin for SCLC but also that basal-cell progeny and anatomic location can be influenced by genotype and injury. An efficient model of AT2-derived SCLC after the loss of Pten

Deletion of Pten was sufficient to lower the barrier to transformation by Myc in the AT2 lineage; however, the resulting tumors were not neuroendocrine (fig. S13). We suspected that Rb1 loss was required (10) —in addition to adaptation to Myc—for neuroendocrine transformation. Thus, to recapitulate bona fide transformation from an AT2 cell to SCLC efficiently, we generated a model in which we could delete Rb1, Trp53, and a copy of Pten, and express Myc and tdTom (RPPtenMT). If tumorigenesis was initiated in PNECs, there was no difference in latency between the RPMT and RPPtenMT models, but a significant difference emerged if the model was initiated in AT2 cells (Fig. 4E). AT2-derived RPPtenMT tumor cells displayed neuroendocrine expression profiles most similar to de novo SCLC (Fig. 4F and fig. S15, A and B). We also noted two subpopulations of cells stratified by expression of Ascl1 (fig. S15B). Analysis of differential gene expression in the Ascl1low group was notable for expression of neuronal genes, including Stnm2, Nfix, and Mapt. The Ascl1low group also expressed high levels of NeuroD1, which is consistent with prior work showing Myc driving subtype plasticity from an Ascl1high to Ascl1low/NeuroD1high transcriptional profile in multiple models of SCLC (54–58) (fig. S15C). Pathology revealed extensive heterogeneity with mixing of classic SCLC and large-cell neuroendocrine (LCNEC) tumor features, marked by variable expression of neuronal and neuroendocrine markers (fig. S15D). Rb1 loss is necessary but insufficient for fully penetrant LUAD to SCLC transformation

The ERPMT model provided an efficient system to study HT, but we had not addressed a core requirement for Myc in this process. We generated another model in the absence of Myc overexpression (ERPT) and found that although some animals had recurrent disease after the removal of DOX (ERPT on > off), the penetrance of the phenotype was incomplete (Fig. 5A). Moreover, whereas the de novo ERPT LUAD showed an absence of Ascl1 and Myc protein, only some recurrent tumors appeared to be histologically consistent with SCLC, further limiting the utility of the ERPT model in recapitulating HT (Fig. 5, B to D). Although recurrence was incomplete, the ERPT tumors recurring as SCLC did show an increase in Myc protein (Fig. 5C). We analyzed the transcriptomes of ERPT and ERPMT LUAD tumor cells compared with those of normal AT2 cells and found that both the ERPT and ERPMT 6 of 11

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

A

B

C

D

E

F

Fig. 4. Loss of Pten in the AT2 lineage removes the barrier to Myc transformation. (A) Survival of mice where Myc (n = 12), PtenFl/WT (n = 5), or the combination of these alleles (n = 23) are initiated in AT2 cells by using a single copy of SpcCreERT2. Data were censored between 250 and 325 days in the nonlethal arms. (B) Comparative histology of whole lungs from representative Spc>Myc, Spc>PtenFl/+;Myc, Spc>PtenFl/Fl;Myc, or Spc>PI3KLSL-E545K;Myc mice at ~3 months after labeling. Higher-magnification regions (boxed) are provided at right; scalebars, 200 mm. (C) Bar plot showing fraction of each epithelial lineage archetype detected per sample as in Fig. 2G. (D) Bar plot of hallmark gene sets

models transcriptionally maintained expression of AT2 genes. However, Myc and embryonic stem cell target genes were elevated in the ERPT model and increased in the ERPMT Gardner et al., Science 383, eadj1415 (2024)

significantly enriched [false discovery rate (FDR) < 0.01] within the undifferentiated cell state of the Spc>PtenFl/+;Myc sample pool. (E) Effect of Pten deletion combined with deletion of p53 and Rb1 and expression of Myc and tdTomato (RPPtenMT or RPMT) in neuroendocrine (blue) or AT2 (red) cells; n = 10 per arm with x axis split for clarity, ****P < 0.0001. (F) Clustered heatmap of z-normalized imputed expression of AT2 and PNEC signature genes and model oncogenic drivers (Myc and Tg.EGFR) for all tumor-epithelial cells from the RPPtenMT model (green) and the de novo LUAD (red) and SCLC (blue) models described in Fig. 1. Genes and cells are clustered by using the average Euclidean distance method.

model, further supporting the role Myc overexpression has in HT after EGFR withdrawal (Fig. 5E). True paired cases of human LUAD before and after HT are limited, but some

9 February 2024

studies have compared unrelated de novo LUAD with transformed SCLC (T-SCLC) in an attempt to understand this phenomenon as it relates to de novo SCLC tumorigenesis 7 of 11

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

A

B

C

E

D

F

G

H

I

J

Fig. 5. Rb1 loss cooperates with Myc expression to facilitate HT. (A) Survival of mice with the ERPT genotype (no MycT58A transgene) initiated with Ad5.Spc-Cre on DOX (ERPT on; dark green, n = 23), or on and then off DOX once mice developed advanced disease (ERPT on > off; light green, n = 20); ***P < 0.001. As before, DOX diet was removed when individual mice exhibited signs of labored breathing and/or substantial weight loss with a hunched appearance. For each model, cohorts of mice were followed until approximately three times the median latency elapsed, at which point lungs were collected and the study was ended. (B) Histologic appearance of “ERPT on” lungs performed with H&E and IHC staining for Ascl1 and Myc (scalebar, 100 mm). Gardner et al., Science 383, eadj1415 (2024)

9 February 2024

(C) Similar to (B), now with mice on and then off DOX, representative of a SCLClike tumor. (D) As shown in (B), now with mice on and then off DOX, representative of a LUAD-like tumor. Pathologic interpretation is provided below each representative example. (E) (Left) Dot plot showing frequency of expressing cells (node size) and log-transformed expression (node color) of AT2 marker genes in normal AT2, ERPT LUAD, and ERPMT LUAD models. Genes shown are expressed in at least 25% of cells within at least one condition. (Right) KDE plots showing mean log-transformed expression by condition for select gene signatures. (F) Volcano plot showing differentially expressed genes from bulk RNAseq of human transformed SCLC versus LUAD. Genes from the 8 of 11

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

Hallmark MYC Targets V1 signature are colored according to conditional enrichment, top genes [abs(log2FC) > 1 and FDR < 1 × 10–5, where “abs” is absolute and FC is “fold change”] are labeled, and the pathway normalized enrichment score (NES) and FDR from the gene set enrichment analysis (GSEA) are inset (bottom right). (G) Tumorigenesis initiated with an adenoviral (n = 10; Ad5.Spc-Cre) or AT2 lineage-trace allele (n = 19; SpcCreERT2) in the EPMT model (Rb1WT/WT) produces LUAD that does not relapse following DOX removal (n = 5 per group), ***P < 0.001. (H) Representative sagittal lung H&E sections from each group in (G) on DOX at point of moribund disease or 1 month after the

(1, 3, 59, 60). We reanalyzed a publicly available cohort and could demonstrate that Myc target genes were differentially increased in T-SCLC as compared with LUAD, suggesting that Myc may have a role in facilitating HT in a clinical setting (Fig. 5F). Rb1 loss cooperates with expression of Myc to facilitate neuroendocrine transformation

To compare the relative dependence of Rb1 loss and Myc overexpression for HT, we generated a model in which Rb1 was WT (EPMT) and initiated tumorigenesis in AT2 cells using Ad5. Spc-Cre or a lineage trace (SpcCreERT2>EPMT). If on DOX, then both models produced LUAD with differences in latency likely reflecting the broad initiation achievable when using a lineagetrace allele as compared with sparse infection using inhaled adenovirus. (Fig. 5, G and H). LUAD did not recur in EPMT mice taken off DOX. Through a comparison of de novo LUAD and off-DOX residual cells from tumors of varying Rb1 status, we observed marked upregulation of transcriptional programs associated with HT after DOX removal that were dependent on loss of Rb1 (Fig. 5I). Additionally, the relative abundance of lineage archetypes shifts within these models in response to DOX removal suggests that, in the presence of Rb1 loss, high Myc expression selects for a residual state that is stem-like and capable of full neuroendocrine transformation (Fig. 5J). Thus, these events cooperate, as neither Rb1 loss nor Myc overexpression alone are sufficient for fully penetrant HT. We conclude that the AT2 cell is highly refractory to transformation by oncogenic Myc, as are many other cell types within the lung, the noted exception being the PNEC. Intolerance of AT2 cells to Myc can be relieved through activation of the Akt signaling pathway, such as through the deletion of the tumor suppressor Pten. Although this generates a permissive stemlike state, the full conversion to a Myc-driven, high-grade neuroendocrine cancer requires the additional loss of Rb1. Discussion

Carcinogenesis and related processes, such as tumor progression, therapy resistance, and HT, remain incompletely understood. To understand the mechanism by which HT occurs in lung cancer, we have developed mouse models in Gardner et al., Science 383, eadj1415 (2024)

removal of DOX from an otherwise moribund animal. (I) Clustered heatmap of gene signatures differentially enriched between the tumor-epithelial cells of the EPMT and ERPMT models before and after DOX removal. All pathways are significantly enriched (NES > 0 and FDR < 1 × 10–5) in at least one condition. Less significant signatures (FDR < 0.01) are transparent, and signatures not meeting this threshold are blank. Rows and columns are clustered by using the complete Manhattan distance method and metric. (J) Bar plot showing fraction of each epithelial lineage archetype detected per sample (as shown in Fig. 2G) for the EPMT (left) and ERPMT (right) models before and after removal of DOX.

which lung tumorigenesis can be initiated in different cell lineages to follow the transformation of EGFR-driven LUAD to Myc-driven SCLC. In doing so, we demonstrate that HT can be simplified conceptually to a change in oncogenic drivers as the tumor cells transition between states that resemble cells in the AT2 and the PNEC lineages. The driver oncogene in the PNEC lineage is Myc, and the bottleneck in HT can be relieved by mechanisms that allow an AT2 cell to become stem-like and capable of tolerating Myc as an oncogenic driver. Understanding the events that facilitate HT clinically has been limited by the lack of samples obtained from the same patient before, during, and after HT. Comparisons of de novo SCLC to transformed SCLC have highlighted pathways activated after neuroendocrine differentiation (59). A more recent study relied on putative HT samples in which a LUAD oncogenic driver was detected in a histologically confirmed SCLC. When compared to de novo SCLC, these cases were enriched for mutations that activate the PI3K signaling pathway (61). Consistent with this result, we found that loss of Pten or activation of Pik3ca was sufficient to break a barrier in the AT2 lineage to transformation by Myc but insufficient to lead to neuroendocrine transformation. Previous work has established a combination of genetic events capable of reprogramming human cell types to neuroendocrine cancers (62); however, the intermediate steps in this process remain unclear. We have built upon this foundational work by developing intact genetically engineered animals and sampling tumor cells throughout the lifetime of HT. Moreover, our findings show that although Rb1 loss is necessary for HT to a neuroendocrine cancer, it is the extinction of the driver oncogene transcriptionally that is required before the emergence of neuroendocrine features. Beyond the genetic manipulations and changes in gene expression described here, we speculate that there are likely other mechanisms that allow for such lineage conversion, cellular plasticity, or transdifferentiation. Lineage tracing has demonstrated that basal progenitors can give rise to PNECs (63, 64). More recently, tracheal basal cells have been shown to differentiate toward a PNEC fate under hypoxic conditions, but the intermediate transcriptional steps in this complex process are uncharted (65).

9 February 2024

In our attempts to describe the transition state between LUAD and SCLC, we found that these tumor cells appear “basal-like” on the basis of their transcriptional profile but lack definitive basal lineage markers. Instead, these cells are more accurately described as “lineage negative,” stem-like progenitors. Such cells have been described to arise in the mouse lung following major airway injury (66). Thus, the airway cell most capable of such plasticity may be the pulmonary basal cell or a stem-like cell that has yet to be fully characterized. Lastly, it is unclear whether therapeutic targeting of pathways facilitating HT in patients poised to undergo HT will be efficacious. Our results suggest that noninvasive monitoring for activation of the Akt signaling pathway (such as the appearance of PI3K mutations in circulating tumor DNA) may serve to alert physicians to the likelihood of HT before the emergence of an aggressive, transformed SCLC. It is not yet known whether direct targeting of Myc in such Myc-driven scenarios (including HT or de novo SCLC tumorigenesis) will be fruitful; however, strategies to inhibit the transcriptional activity of Myc proteins have advanced substantially over the last decade and might ultimately have clinical utility in multiple contexts (43, 67, 68). Limitations of the study

There are several key limitations to this study. First, we are attempting to understand HT by modeling a human phenomenon in the mouse, where the complexity of cell types and microenvironments is not identical to that found in humans. Recent descriptions of the human distal airway have found a diversity of cell types in human lungs that are not observed in the mouse, including specialized regenerative cells with hybrid alveolar-secretory character and multiple subtypes of basal cells (19, 69, 70). For example, basal cells are not present in the mouse lung unless the airway is damaged, which we demonstrated using naphthalene. Furthermore, most laboratory mice are housed under pathogenfree conditions. We know that lifetime exposure to various carcinogens and particulates fundamentally alters the microenvironment of the lung, including the likelihood that a cancer may develop (71, 72). Second, there are processes in human cells not found in the mouse that may be critical for HT, including the role for 9 of 11

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

APOBEC-mediated hypermutation. Studies of EGFR-mutated LUAD have found APOBEC mutagenesis signatures after treatment with EGFR-targeted therapies (11, 73, 74), and this mutational signature is enriched after HT (1, 61). It is unclear whether APOBEC mutations are directly responsible for activating the PI3K signaling pathway, (hypothetically) relieving intolerance of the AT2 cell to Myc (1, 59, 61); however, other studies would suggest that this is possible (75, 76). Lastly, we were unable to initiate tumorigenesis in the ERPMT model with the SpcCreERT2 allele. This proved impossible because in the mouse genome, the Rb1 locus is ~3 Mb from the Sftpc locus (GRCm38/mm10). Such proximity precludes generation of the desired genotype with both homozygous floxed copies of Rb1 and the SpcCreERT2 lineage-trace allele. RE FE RENCES AND N OT ES

1. J. K. Lee et al., Clonal History and Genetic Predictors of Transformation Into Small-Cell Carcinomas From Lung Adenocarcinomas. J. Clin. Oncol. 35, 3065–3074 (2017). doi: 10.1200/JCO.2016.71.9096; pmid: 28498782 2. M. F. Zakowski, M. Ladanyi, M. G. Kris; Memorial Sloan-Kettering Cancer Center Lung Cancer OncoGenome Group, EGFR mutations in small-cell lung cancers in patients who have never smoked. N. Engl. J. Med. 355, 213–215 (2006). doi: 10.1056/NEJMc053610; pmid: 16837691 3. L. V. Sequist et al., Genotypic and histological evolution of lung cancers acquiring resistance to EGFR inhibitors. Sci. Transl. Med. 3, 75ra26 (2011). doi: 10.1126/scitranslmed.3002003; pmid: 21430269 4. H. Beltran et al., Divergent clonal evolution of castrationresistant neuroendocrine prostate cancer. Nat. Med. 22, 298–305 (2016). doi: 10.1038/nm.4045; pmid: 26855148 5. D. Hirano, Y. Okada, S. Minei, Y. Takimoto, N. Nemoto, Neuroendocrine differentiation in hormone refractory prostate cancer following androgen deprivation therapy. Eur. Urol. 45, 586–592, discussion 592 (2004). doi: 10.1016/ j.eururo.2003.11.032; pmid: 15082200 6. Á. Quintanal-Villalonga et al., Lineage plasticity in cancer: A shared pathway of therapeutic resistance. Nat. Rev. Clin. Oncol. 17, 360–371 (2020). doi: 10.1038/s41571-020-0340-z; pmid: 32152485 7. J. George et al., Comprehensive genomic profiles of small cell lung cancer. Nature 524, 47–53 (2015). doi: 10.1038/ nature14664; pmid: 26168399 8. C. M. Rudin et al., Molecular subtypes of small cell lung cancer: A synthesis of human and mouse model data. Nat. Rev. Cancer 19, 289–297 (2019). doi: 10.1038/s41568-019-0133-9; pmid: 30926931 9. M. G. Oser, M. J. Niederst, L. V. Sequist, J. A. Engelman, Transformation from non-small-cell lung cancer to small-cell lung cancer: Molecular drivers and cells of origin. Lancet Oncol. 16, e165–e172 (2015). doi: 10.1016/S1470-2045(14)71180-5; pmid: 25846096 10. M. J. Niederst et al., RB loss in resistant EGFR mutant lung adenocarcinomas that transform to small-cell lung cancer. Nat. Commun. 6, 6377 (2015). doi: 10.1038/ncomms7377; pmid: 25758528 11. M. Offin et al., Concurrent RB1 and TP53 Alterations Define a Subset of EGFR-Mutant Lung Cancers at risk for Histologic Transformation and Inferior Clinical Outcomes. J. Thorac. Oncol. 14, 1784–1793 (2019). doi: 10.1016/j.jtho.2019.06.002; pmid: 31228622 12. T. J. Desai, D. G. Brownfield, M. A. Krasnow, Alveolar progenitor and stem cells in lung development, renewal and cancer. Nature 507, 190–194 (2014). doi: 10.1038/nature12930; pmid: 24499815 13. D. B. Frank et al., Early lineage specification defines alveolar epithelial ontogeny in the murine lung. Proc. Natl. Acad. Sci. U.S.A. 116, 4362–4371 (2019). doi: 10.1073/pnas.1813952116; pmid: 30782824

Gardner et al., Science 383, eadj1415 (2024)

14. A. N. Nabhan, D. G. Brownfield, P. B. Harbury, M. A. Krasnow, T. J. Desai, Single-cell Wnt signaling niches maintain stemness of alveolar type 2 cells. Science 359, 1118–1123 (2018). doi: 10.1126/science.aam6603; pmid: 29420258 15. Cancer Genome Atlas Research Network, Comprehensive molecular profiling of lung adenocarcinoma. Nature 511, 543–550 (2014). doi: 10.1038/nature13385; pmid: 25079552 16. Y. Ouadah et al., Rare Pulmonary Neuroendocrine Cells Are Stem Cells Regulated by Rb, p53, and Notch. Cell 179, 403–416.e23 (2019). doi: 10.1016/j.cell.2019.09.010; pmid: 31585080 17. C. S. Kuo, M. A. Krasnow, Formation of a Neurosensory Organ by Epithelial Cell Slithering. Cell 163, 394–405 (2015). doi: 10.1016/j.cell.2015.09.021; pmid: 26435104 18. P. Sui et al., Pulmonary neuroendocrine cells amplify allergic asthma responses. Science 360, eaan8546 (2018). doi: 10.1126/science.aan8546; pmid: 29599193 19. K. J. Travaglini et al., A molecular cell atlas of the human lung from single-cell RNA sequencing. Nature 587, 619–625 (2020). doi: 10.1038/s41586-020-2922-4; pmid: 33208946 20. K. Politi et al., Lung adenocarcinomas induced in mice by mutant EGF receptors found in human lung cancers respond to a tyrosine kinase inhibitor or to down-regulation of the receptors. Genes Dev. 20, 1496–1510 (2006). doi: 10.1101/ gad.1417406; pmid: 16705038 21. K. D. Sutherland et al., Cell of origin of small cell lung cancer: Inactivation of Trp53 and Rb1 in distinct cell types of adult mouse lung. Cancer Cell 19, 754–764 (2011). doi: 10.1016/ j.ccr.2011.04.019; pmid: 21665149 22. C. E. Barkauskas et al., Type 2 alveolar cells are stem cells in adult lung. J. Clin. Invest. 123, 3025–3036 (2013). doi: 10.1172/JCI68782; pmid: 23921127 23. A. M. Laughney et al., Regenerative lineages and immunemediated pruning in lung cancer metastasis. Nat. Med. 26, 259–269 (2020). doi: 10.1038/s41591-019-0750-6; pmid: 32042191 24. Y. He et al., MHC class II expression in lung cancer. Lung Cancer 112, 75–80 (2017). doi: 10.1016/j.lungcan.2017.07.030; pmid: 29191604 25. T. Yazawa et al., Lack of class II transactivator causes severe deficiency of HLA-DR expression in small cell lung cancer. J. Pathol. 187, 191–199 (1999). doi: 10.1002/(SICI)1096-9896 (199901)187:23.0.CO;2-3; pmid: 10365094 26. L. M. Marsh et al., Surface expression of CD74 by type II alveolar epithelial cells: A potential mechanism for macrophage migration inhibitory factor-induced epithelial repair. Am. J. Physiol. Lung Cell. Mol. Physiol. 296, L442–L452 (2009). doi: 10.1152/ajplung.00525.2007; pmid: 19136583 27. S. Bae, Y. Bessho, M. Hojo, R. Kageyama, The bHLH gene Hes6, an inhibitor of Hes1, promotes neuronal differentiation. Development 127, 2933–2943 (2000). doi: 10.1242/ dev.127.13.2933; pmid: 10851137 28. J. D. Buenrostro, B. Wu, H. Y. Chang, W. J. Greenleaf, ATAC-seq: A Method for Assaying Chromatin Accessibility Genome-Wide. Curr. Protoc. Mol. Biol.109, 21.29.1–21.9.9 (2015). doi: 10.1002/0471142727.mb2129s109; pmid: 25559105 29. P. A. Ewels et al., The nf-core framework for communitycurated bioinformatics pipelines. Nat. Biotechnol. 38, 276–278 (2020). doi: 10.1038/s41587-020-0439-x; pmid: 32055031 30. S. Heinz et al., Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol. Cell 38, 576–589 (2010). doi: 10.1016/j.molcel.2010.05.004; pmid: 20513432 31. Y. Oren et al., Cycling cancer persister cells arise from lineages with distinct programs. Nature 596, 576–582 (2021). doi: 10.1038/s41586-021-03796-6; pmid: 34381210 32. M. D. Borromeo et al., ASCL1 and NEUROD1 Reveal Heterogeneity in Pulmonary Neuroendocrine Tumors and Regulate Distinct Genetic Programs. Cell Rep. 16, 1259–1272 (2016). doi: 10.1016/j.celrep.2016.06.081; pmid: 27452466 33. A. Cutler, L. Breiman, Archetypal Analysis. Technometrics 36, 338–347 (1994). doi: 10.1080/00401706.1994.10485840 34. M. Lange et al., CellRank for directed single-cell fate mapping. Nat. Methods 19, 159–170 (2022). doi: 10.1038/s41592-02101346-6; pmid: 35027767 35. M. C. Wagle et al., A transcriptional MAPK Pathway Activity Score (MPAS) is a clinically relevant biomarker in multiple cancer types. NPJ Precis. Oncol. 2, 7 (2018). doi: 10.1038/ s41698-018-0051-4; pmid: 29872725 36. G. Chen et al., Targeting TM4SF1 exhibits therapeutic potential via inhibition of cancer stem cells. Signal Transduct. Target.

9 February 2024

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

55.

56.

57.

58.

Ther. 7, 350 (2022). doi: 10.1038/s41392-022-01177-7; pmid: 36229443 X. Ouyang et al., WDR72 Enhances the Stemness of Lung Cancer Cells by Activating the AKT/HIF-1a Signaling Pathway. J. Oncol. 2022, 5059588 (2022). doi: 10.1155/2022/5059588; pmid: 36385964 M. S. Kowalczyk et al., Single-cell RNA-seq reveals changes in cell cycle and differentiation programs upon aging of hematopoietic stem cells. Genome Res. 25, 1860–1872 (2015). doi: 10.1101/gr.192237.115; pmid: 26430063 N. Tsuboyama et al., Therapeutic targeting of BAP1/ASXL3 sub-complex in ASCL1-dependent small cell lung cancer. Oncogene 41, 2152–2162 (2022). doi: 10.1038/s41388-02202240-x; pmid: 35194152 Z. Zhao et al., PAX9 Determines Epigenetic State Transition and Cell Fate in Cancer. Cancer Res. 81, 4696–4708 (2021). doi: 10.1158/0008-5472.CAN-21-1114; pmid: 34341073 Y. Xia et al., Targeting CREB Pathway Suppresses Small Cell Lung Cancer. Mol. Cancer Res. 16, 825–832 (2018). doi: 10.1158/1541-7786.MCR-17-0576; pmid: 29523765 M. A. Gregory, Y. Qi, S. R. Hann, Phosphorylation by glycogen synthase kinase-3 controls c-myc proteolysis and subnuclear localization. J. Biol. Chem. 278, 51606–51612 (2003). doi: 10.1074/jbc.M310722200; pmid: 14563837 H. Han et al., Small-Molecule MYC Inhibitors Suppress Tumor Growth and Enhance Immunotherapy. Cancer Cell 36, 483–497.e15 (2019). doi: 10.1016/j.ccell.2019.10.001; pmid: 31679823 S. Naranjo et al., Modeling diverse genetic subtypes of lung adenocarcinoma with a next-generation alveolar type 2 organoid platform. Genes Dev. 36, 936–949 (2022). doi: 10.1101/gad.349659.122; pmid: 36175034 D. J. Murphy et al., Distinct thresholds govern Myc’s biological output in vivo. Cancer Cell 14, 447–457 (2008). doi: 10.1016/ j.ccr.2008.10.018; pmid: 19061836 A. Kauffmann-Zeh et al., Suppression of c-Myc-induced apoptosis by Ras signalling through PI(3)K and PKB. Nature 385, 544–548 (1997). doi: 10.1038/385544a0; pmid: 9020362 H. Taniguchi et al., A resource of Cre driver lines for genetic targeting of GABAergic neurons in cerebral cortex. Neuron 71, 995–1013 (2011). doi: 10.1016/j.neuron.2011.07.026; pmid: 21943598 E. Y. Chen et al., Enrichr: Interactive and collaborative HTML5 gene list enrichment analysis tool. BMC Bioinformatics 14, 128 (2013). doi: 10.1186/1471-2105-14-128; pmid: 23586463 S. Lázaro et al., Differential development of large-cell neuroendocrine or small-cell lung carcinoma upon inactivation of 4 tumor suppressor genes. Proc. Natl. Acad. Sci. U.S.A. 116, 22300–22306 (2019). doi: 10.1073/pnas.1821745116; pmid: 31611390 G. Ferone et al., SOX2 Is the Determining Oncogenic Switch in Promoting Lung Squamous Cell Carcinoma from Different Cells of Origin. Cancer Cell 30, 519–532 (2016). doi: 10.1016/ j.ccell.2016.09.001; pmid: 27728803 J. R. Rock et al., Basal cells as stem cells of the mouse trachea and human airway epithelium. Proc. Natl. Acad. Sci. U.S.A. 106, 12771–12775 (2009). doi: 10.1073/pnas.0906850106; pmid: 19625615 K. U. Hong, S. D. Reynolds, S. Watkins, E. Fuchs, B. R. Stripp, Basal cells are a multipotent progenitor capable of renewing the bronchial epithelium. Am. J. Pathol. 164, 577–588 (2004). doi: 10.1016/S0002-9440(10)63147-1; pmid: 14742263 B. Hogan, P. R. Tata, Cellular organization and biology of the respiratory system. Nat. Cell Biol. (2019). doi: 10.1038/s41556019-0357-7; pmid: 31346286 A. S. Ireland et al., MYC Drives Temporal Evolution of Small Cell Lung Cancer Subtypes by Reprogramming Neuroendocrine Fate. Cancer Cell 38, 60–78.e12 (2020). doi: 10.1016/ j.ccell.2020.05.001; pmid: 32473656 R. R. Olsen et al., ASCL1 represses a SOX9+ neural crest stemlike state in small cell lung cancer. Genes Dev. 35, 847–869 (2021). doi: 10.1101/gad.348295.121; pmid: 34016693 D. Lissa et al., Heterogeneity of neuroendocrine transcriptional states in metastatic small cell lung cancers and patient-derived models. Nat. Commun. 13, 2023 (2022). doi: 10.1038/ s41467-022-29517-9; pmid: 35440132 J. M. Chan et al., Signatures of plasticity, metastasis, and immunosuppression in an atlas of human small cell lung cancer. Cancer Cell 39, 1479–1496.e18 (2021). doi: 10.1016/ j.ccell.2021.09.008; pmid: 34653364 C. M. Gay et al., Patterns of transcription factor programs and immune pathway activation define four major subtypes of SCLC with distinct therapeutic vulnerabilities. Cancer Cell 39,

10 of 11

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

59.

60.

61.

62.

63.

64.

65.

66.

67.

68.

69.

346–360.e7 (2021). doi: 10.1016/j.ccell.2020.12.014; pmid: 33482121 A. Quintanal-Villalonga et al., Multiomic analysis of lung tumors defines pathways activated in neuroendocrine transformation. Cancer Discov. 11, 3028–3047 (2021). doi: 10.1158/2159-8290. CD-20-1863; pmid: 34155000 S. Sivakumar et al., Integrative Analysis of a Large Real-World Cohort of Small Cell Lung Cancer Identifies Distinct Genetic Subtypes and Insights into Histologic Transformation. Cancer Discov. 13, 1572–1591 (2023). doi: 10.1158/2159-8290.CD-220620; pmid: 37062002 S. Sivakumar et al., Integrative analysis of a large real-world cohort of small cell lung cancer identifies distinct genetic subtypes and insights into histological transformation. Cancer Discov. 13, 1572–1591 (2023). doi: 10.1158/2159-8290.CD-220620; pmid: 37062002 J. W. Park et al., Reprogramming normal human epithelial tissues to a common, lethal neuroendocrine cancer lineage. Science 362, 91–95 (2018). doi: 10.1126/science.aat5749; pmid: 30287662 D. T. Montoro et al., A revised airway epithelial hierarchy includes CFTR-expressing ionocytes. Nature 560, 319–324 (2018). doi: 10.1038/s41586-018-0393-7; pmid: 30069044 H. Mou et al., Airway basal stem cells generate distinct subpopulations of PNECs. Cell Rep. 35, 109011 (2021). doi: 10.1016/j.celrep.2021.109011; pmid: 33882306 M. Shivaraju et al., Airway stem cells sense hypoxia and differentiate into protective solitary neuroendocrine cells. Science 371, 52–57 (2021). doi: 10.1126/science.aba0629; pmid: 33384370 A. E. Vaughan et al., Lineage-negative progenitors mobilize to regenerate lung epithelium after major injury. Nature 517, 621–625 (2015). doi: 10.1038/nature14112; pmid: 25533958 M. E. Beaulieu et al., Intrinsic cell-penetrating activity propels Omomyc from proof of concept to viable anti-MYC therapy. Sci. Transl. Med. 11, eaar5012 (2019). doi: 10.1126/ scitranslmed.aar5012; pmid: 30894502 T. E. Speltz et al., Targeting MYC with modular synthetic transcriptional repressors derived from bHLH DNA-binding domains. Nat. Biotechnol. 41, 541–551 (2023). doi: 10.1038/ s41587-022-01504-x; pmid: 36302987 M. C. Basil et al., Human distal airways contain a multipotent secretory cell that can regenerate alveoli. Nature 604, 120–126 (2022). doi: 10.1038/s41586-022-04552-0; pmid: 35355013

Gardner et al., Science 383, eadj1415 (2024)

70. P. Kadur Lakshminarasimha Murthy et al., Human distal lung maps and lineage hierarchies reveal a bipotent progenitor. Nature 604, 111–119 (2022). doi: 10.1038/s41586-022-04541-3; pmid: 35355018 71. W. Hill et al., Lung adenocarcinoma promotion by air pollutants. Nature 616, 159–167 (2023). doi: 10.1038/s41586023-05874-3; pmid: 37020004 72. C. Mascaux et al., Immune evasion before tumour invasion in early lung squamous carcinogenesis. Nature 571, 570–575 (2019). doi: 10.1038/s41586-019-1330-0; pmid: 31243362 73. C. Martínez-Ruiz et al., Genomic-transcriptomic evolution in lung cancer and metastasis. Nature 616, 543–552 (2023). doi: 10.1038/s41586-023-05706-4; pmid: 37046093 74. P. Selenica et al., APOBEC mutagenesis, kataegis, chromothripsis in EGFR-mutant osimertinib-resistant lung adenocarcinomas. Ann. Oncol. 33, 1284–1295 (2022). doi: 10.1016/j.annonc.2022.09.151; pmid: 36089134 75. R. Buisson et al., Passenger hotspot mutations in cancer driven by APOBEC3A and mesoscale genomic features. Science 364, eaaw2872 (2019). doi: 10.1126/science.aaw2872; pmid: 31249028 76. S. Henderson, A. Chakravarthy, X. Su, C. Boshoff, T. R. Fenton, APOBEC-mediated cytosine deamination links PIK3CA helical domain mutations to human papillomavirus-driven tumor development. Cell Rep. 7, 1833–1841 (2014). doi: 10.1016/ j.celrep.2014.05.012; pmid: 24910434 ACKN OWLED GMEN TS

Funding: This work was supported by the National Cancer Institute (NCI) NIH cooperative agreement U01CA224326 (to H.V.); NCI NIH grants P01CA120964 and R35CA197588 (to L.C.C.); NCI NIH grants R01CA256188, R01CA280414, R01CA280572, and R21CA266660-01 (to A.M.L.); the Burroughs Wellcome Fund (to A.M.L.); the Lung Cancer Research Foundation (to A.M.L.); the Damon Runyon Cancer Research Foundation fellowship DRG-2343-18 (to E.E.G.); and NIH training grant T32 GM132083 (to E.M.E). The Varmus lab is supported in part by the Lewis Thomas University Professor endowment. Author contributions: Conceptualization: E.E.G. and H.V. Methodology: E.E.G., E.M.E., M.J.H., K.L., J.T., B.D.S., L.C.C., and A.M.L. Experimental Design: E.E.G., E.M.E., A.M.L., and H.V. Visualization: E.E.G., E.M.E., M.J.H., and A.M.L. Funding acquisition: E.E.G., L.C.C., A.M.L., and H.V. Supervision: H.V. and A.M.L. Pathology review: C.Z. Writing – original draft: E.E.G.

9 February 2024

Writing – review and editing: all authors. Competing interests: L.C.C. is a cofounder and member of the Scientific Advisory Board (SAB) and holds equity in Faeth Therapeutics, Volastra Therapeutics, and Larkspur Therapeutics. He is also a cofounder, former member of the SAB and holds equity in Agios Pharmaceuticals and Petra Pharmaceuticals (now owned by Loxo@Lilly). These companies are developing novel therapies for cancer. L.C.C.’s laboratory has previously received some financial support from Petra Pharmaceuticals. None of these companies are currently providing support for the Cantley laboratory. H.V. is a member of the SABs of Volastra, Dragonfly Therapeutics, and Surrozen. None of these companies are currently providing support for the Varmus laboratory. All other authors declare no competing interests. Data and materials availability: Lead contact: Requests for resources should be directed to and will be fulfilled by Eric E. Gardner ([email protected]) or Ashley M. Laughney ([email protected]). Materials: All mouse models, organoids derived from mice, and plasmids described in this work will be made available to investigators through an institutional or third-party Material Transfer Agreement (MTA) upon reasonable request. Select plasmids will be submitted to Addgene upon manuscript acceptance. Not all mouse strains are currently active in the Varmus Laboratory colony—please contact Eric E. Gardner for specific information and availability. Data and code: The processed single-cell data and relevant code, including Docker environments with Jupyter notebooks demonstrating key analyses, are available on either the Gene Expression Omnibus (GEO no. GSE248207) or Laughney Lab GitHub (https://github.com/ LaughneyLab/Lung_Histological_Transformation). 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.adj1415 Materials and Methods Supplementary Text Figs. S1 to S15 Tables S1 to S4 References (77–115) MDAR Reproducibility Checklist Submitted 23 June 2023; accepted 8 December 2023 10.1126/science.adj1415

11 of 11

RES EARCH

RESEARCH ARTICLE SUMMARY



HEALTH ECONOMICS

What drives poor quality of care for child diarrhea? Experimental evidence from India Zachary Wagner*, Manoj Mohanan, Rushil Zutshi, Arnab Mukherji, Neeraj Sood

INTRODUCTION: Diarrhea is a leading cause of

death in children, with nearly 500,000 young lives lost to diarrhea each year. Almost all these lives could be saved with a low-cost and widely available treatment: oral rehydration salts (ORS). However, at the present time, nearly half of diarrhea cases around the world do not receive ORS. Millions of young lives could be saved if we can find ways to increase ORS use. Even when children seek care from a health care provider for their diarrhea, as most do, they often do not receive ORS. Surprisingly, most health care providers in developing countries know that ORS is a lifesaving and inexpensive treatment for child diarrhea, yet few prescribe it. This know-do gap has puzzled experts for decades and cost millions of lives. RATIONALE: To develop interventions that in-

crease ORS prescribing, we must have a clear understanding of why providers do not prescribe ORS even though they know it is the standard of care. There are several leading explanations. First, providers might think that

Gain in ORS prescribing or dispensing by removing each barrier

patients prefer non-ORS treatments (e.g., antibiotics) or dislike ORS because of poor taste, lack of observable symptom relief, and perceptions that ORS is not a real medicine. Second, providers could be responding to financial incentives to sell more-profitable alternatives. ORS is inexpensive, and antibiotics generate nearly double the profit. Finally, providers often have ORS stock-outs (out-of-stock events) and might prefer to dispense something they have in stock to not lose out on the sale. Each of these potential barriers to ORS prescribing suggests a different solution. We used a randomized controlled trial to simultaneously study the role of these three leading explanations for the underprescribing of ORS. More than 2000 providers across 253 medium-sized towns in the Indian states of Karnataka and Bihar participated in the study. To measure the effect of perceived patient preferences, we had standardized patients (actors trained to act as patients) make unannounced visits where they presented a case of diarrhea for their 2-year-old child, and we randomly assigned whether they expressed

Percent contribution to the problem

a preference for ORS, a preference for antibiotics, or no preference. To measure the effect of financial incentives, some of the standardized patients assigned to the no-preference arm informed the provider that they would purchase medicines from a different location, thereby eliminating the provider’s financial incentive to recommend more-lucrative treatments. Finally, to estimate the effect of ORS stock-outs, we randomly assigned all providers in half of the 253 towns to receive a 6-week supply of ORS. RESULTS: We found that having standardized

patients express a preference for ORS increased ORS prescribing by 27 percentage points compared with no preference; 28% prescribed or dispensed ORS when standardized patients expressed no preference, and 55% prescribed or dispensed ORS when standardized patients expressed an ORS preference (96% increase). We show that this is mainly because providers think only 18% of their patients want ORS on average, when, in reality, ORS was the most preferred treatment reported by patients in household surveys. Eliminating stock-outs increased ORS provision by 7 percentage points on average and by 17 percentage points among clinics that sell, rather than prescribe, medicine. Eliminating financial incentives to sell more-lucrative medicines had no effect on average but did increase ORS prescribing at pharmacies by 9 percentage points. By combining these results with the prevalence of each barrier estimated through provider and household surveys, we estimate that provider misperceptions that patients do not want ORS explain 42% of underprescribing, whereas stock-outs and financial incentives explain only 6 and 5%, respectively.

Share of cases prescribed or dispensed ORS

100%

CONCLUSION: Provider misperceptions that pa42% Provider perceptions of patient preferences

80% No financial incentives No stock-outs 60%

40%

20% 0%

All patients show preference for ORS ORS Status quo: ORS prescribed or dispensed 42% of the time

Stock-outs

47% Other barriers

?

6% 5% Providers’ financial incentives

Provider perceptions that patients do not want ORS are the most important barrier to ORS prescribing. The figure shows that if all patients showed a preference for ORS, it would increase ORS prescribing by 24.5 percentage points, from 42 to 66.5%. Thus, provider perceptions of patient preferences explain 42% of the problem, whereas stock-outs and financial incentives explain only 6.4 and 5%, respectively. Estimates are based on the effects of removing each barrier combined with estimates on the prevalence of each barrier. The parameters used for these calculations are listed in table S21. In the status quo, 16% of patients showed a preference for ORS, 42% of providers had ORS in stock, and all providers had a financial incentive. 606

9 FEBRUARY 2024 • VOL 383 ISSUE 6683

tients do not want ORS play the biggest role in the underprescribing of ORS and are 6 to 10 times more important than financial incentives or stock-outs. These results suggest that interventions to change provider misperceptions of patients’ ORS preferences should be aggressively explored because they have the potential to substantially increase ORS use. These interventions could target patients or caretakers and encourage them to express an ORS preference when they seek care, or they could target providers directly and inform them that ORS preferences are more common than they think.



The list of author affiliations is available in the full article online. *Corresponding author. Email: [email protected] Cite this article as Z. Wagner et al., Science 383, eadj9986 (2024). DOI: 10.1126/science.adj9986

READ THE FULL ARTICLE AT https://doi.org/10.1126/science.adj9986 science.org SCIENCE

RES EARCH

RESEARCH ARTICLE



HEALTH ECONOMICS

What drives poor quality of care for child diarrhea? Experimental evidence from India Zachary Wagner1,2*, Manoj Mohanan3, Rushil Zutshi1,2, Arnab Mukherji4, Neeraj Sood5,6 Most health care providers in developing countries know that oral rehydration salts (ORS) are a lifesaving and inexpensive treatment for child diarrhea, yet few prescribe it. This know-do gap has puzzled experts for decades. Using randomized experiments in India, we estimated the extent to which ORS underprescription is driven by perceptions that patients do not want ORS, provider’s financial incentives, and ORS stock-outs (out-of-stock events). Patients expressing a preference for ORS increased ORS prescribing by 27 percentage points. Eliminating stock-outs increased ORS provision by 7 percentage points. Removing financial incentives did not affect ORS prescribing on average but did increase ORS prescribing at pharmacies. We estimate that perceptions that patients do not want ORS explain 42% of underprescribing, whereas stock-outs and financial incentives explain only 6 and 5%, respectively.

D

iarrhea is the second leading cause of death for children in low- and middleincome countries (LMICs), with nearly half a million children under 5 years old dying each year (1). This is true even though nearly all such deaths could be prevented with a simple and inexpensive medicine: oral rehydration salts (ORS). ORS has been lauded as one of the most important medical advances of the 20th century (2), yet it has been underutilized for decades (3). At present, nearly half of diarrhea cases around the world do not receive ORS (4). Millions of young lives could be saved if we can find ways to increase ORS use. Lack of access to health care in LMICs likely explains some portion of low ORS use. However, most children with diarrhea visit a health care provider for treatment, and many of these children still do not receive ORS (5, 6). Providers often prescribe antibiotics instead of ORS, which do not address dehydration and are not appropriate for common viral diarrhea pathogens. Inappropriate prescribing for diarrhea is particularly egregious among private providers, who treat most childhood illness in LMICs (5–7). In India, where nearly a quarter of all child deaths from diarrhea occur, more than 75% of caretakers seek care for a child’s diarrhea from private providers, and 45% do not receive ORS (8).

1

Department of Economics, Sociology and Statistics, RAND Corporation, Santa Monica, CA, USA. 2Pardee RAND Graduate School, Santa Monica, CA, USA. 3Sanford School of Public Policy, Duke University, Durham, NC, USA. 4Center for Public Policy, Indian Institute of Management Bangalore, Bangalore, Karnataka, India. 5Sol Price School of Public Policy, University of Southern California, Los Angeles, CA, USA. 6Schaeffer Center for Health Policy and Economics, University of Southern California, Los Angeles, CA, USA. *Corresponding author. Email: [email protected]

Wagner et al., Science 383, eadj9986 (2024)

There is little evidence to date documenting why so many children fail to receive ORS when they visit a private health care provider. This is a critical global health question that has been researched for decades (5, 6, 9, 10) and that many global funders have tried to address (11, 12), yet we still know little about the root causes of the problem. This makes it challenging to design effective interventions to increase ORS prescribing in the private sector. Prior work documents that even when private providers know that ORS is the appropriate treatment for diarrhea, which most do, they still fail to prescribe it (9). This implies that lack of knowledge is unlikely to be an important driver, and educating providers about ORS is unlikely to increase ORS dispensing. Several studies that focused on increasing provider knowledge as a means of improving ORS use have been ineffective (12–14). Why is it, then, that so many children who seek care for diarrhea in the private sector do not receive ORS? There are several remaining explanations for why private providers underprescribe ORS. First, providers might think that patients prefer something different than ORS (e.g., an antibiotic), and private providers could be particularly responsive to patient preferences to ensure that patients keep supporting their business. Providers might think that patients prefer non-ORS treatments or dislike ORS because of poor taste, lack of observable symptom relief (it treats and prevents dehydration rather than diarrhea symptoms), and a perception that ORS is not a real medicine (it is a powder that is mixed with water, rather than a pill or injection). Second, private providers could be responding to financial incentives to sell moreprofitable alternatives. ORS is inexpensive, and antibiotics generate nearly double the profit. There is evidence that financial incentives drive

9 February 2024

inappropriate antibiotic prescribing in other settings (15, 16). Finally, private providers often have ORS stock-outs (out-of-stock events) and might prefer to dispense something that they have in stock to not lose out on the sale (5, 17). Each of these potential barriers to ORS prescribing suggests a different solution. It is critical to understand how much each barrier contributes to inappropriate care in order to focus resources on interventions that would be most effective at increasing ORS use. To the best of our knowledge, this is the first study to simultaneously evaluate and quantify the extent to which each of these key barriers contributes to underprescribing of ORS. In this study, we used several randomized controlled trials with 2282 private providers across 253 medium-sized towns in India to answer the following research questions: (i) What is the effect of provider perceptions of patients’ treatment preferences on ORS and antibiotic prescribing? (ii) What is the effect of removing the provider’s financial incentive to sell more lucrative medicines on ORS and antibiotic prescribing? (iii) What is the effect of eliminating stock-outs on ORS and antibiotic prescribing? and (iv) What portion of the problem do each of these barriers contribute? Method Sample

Our study took place in two Indian states: Karnataka and Bihar. We chose states that are very different from one another in socioeconomic status and diarrhea care, seeking to ensure that our results were representative of a broad population. Bihar, which is in the east, is one of the poorest states in India, with 46% of the adult population having no schooling and only 42% having completed at least secondary school (8). By contrast, Karnataka, which is in the south, has above average per capita income, with only 26% of the adult population having no schooling and 62% having completed at least secondary school. Additionally, Bihar has below average ORS use (57% of cases are treated with ORS compared with the national average of 61%), and Karnataka has above average ORS use (70% of cases treated with ORS) (8). We sampled 253 medium-sized towns [see fig. S1 for a map of our study locations and section 1.1 of the supplementary materials (SM) for a full description of our sampling process]. We attempted to enroll all private providers who treated child diarrhea based on a census we created. We were able to contact 59% of those identified in the census, and 69% of those identified consented to participate in the study. Of the 2451 providers we enrolled, we were able to collect prescribing data at follow-up for 2282 (6.9% lost-to-follow-up rate). Table S1 shows the characteristics of the 2282 providers in our sample, which were collected through a 1 of 8

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

baseline provider survey. Ninety-two percent of providers were male, with an average age of 44 years; providers had 18.5 years of experience on average, and 86% said that they would give ORS to a child with diarrhea based on a case vignette presented to the provider. Providers saw an average of 24.7 patients per day and 6.3 diarrhea cases per week; 56% dispensed medications directly to patients, and 52% had ORS available at baseline. Our sample includes four different types of providers: 20% were providers with an MBBS degree (similar to an MD degree in the United States); 37% were providers with a degree in traditional medicine, including Ayurveda, Yoga and Naturopathy, Unani, Siddha and Homeopathy (AYUSH); 21% were rural medical practitioners (RMPs), who typically lack formal training but still practice medicine; and 22% were pharmacies. MBBS, AYUSH, and RMP providers generally own their own practice and run it as a business. Most only have one provider who sees patients. We refer to this grouping of providers (i.e., excluding pharmacies) as “private clinics.” Pharmacies in our sample are stand-alone businesses that sell medicines. Providers who work at pharmacies generally have no medical training but often advise patients on treatment options for child diarrhea and are often the first providers to whom caretakers go to seek care. Many clinics have an attached pharmacy that is part of the same business, and we interpreted these entities as clinics that sell medicine and not pharmacies. Tables S2 and S3 show that the providers who were lost to follow-up had similar characteristics across study arms (only 2 of 28 comparisons of provider characteristics were statistically significant). Table S4 shows characteristics of caretakers in the areas served by the providers in our sample who recently sought care for their child’s diarrhea; 57% of caretakers in our setting sought treatment at a private clinic, 31% sought treatment at a private pharmacy, and only 9% sought treatment in the public sector. Of those who sought care from a private provider, 42% were prescribed ORS and 49% were prescribed antibiotics (caretakers are often not able to distinguish between antibiotics and other medicines). These numbers are similar to estimates from demographic and health surveys (8). In each state, we trained actors to play the role of a father seeking care for their 2-year-old child who had been having diarrhea for 2 days. We recruited 40 actors in each state to go through an extensive 2-week standardized patient (SP) training and selected the top 25 actors to make the SP visits. The extensive 2-week training included memorizing a script as well as responses to common questions and practice visits with real providers. All SPs presented a case of a 2-year-old child who had been having uncomplicated diarrhea for 2 days with clinical presentation of a case Wagner et al., Science 383, eadj9986 (2024)

caused by rotavirus. Half the SPs presented a moderate case (four to five loose stools the previous night) and the other half a severe case (10 to 12 loose stools the previous night and symptoms of dehydration), but both cases were severe enough to require ORS. The standard of care for a case of rotavirus as reported in the Indian Academy of Pediatrics consensus statement is ORS combined with zinc, and no antibiotics (18). Although antibiotics are appropriate for some cases of child diarrhea, we designed the case with clinical presentation (no blood in stool, feces quality not sticky or smelly, and short duration) such that it was clear that antibiotics would not be necessary. The only difference between the SP roles were (i) the opening statement, in which they expressed a preference (four different statements, which are described more below) and (ii) the severity of the diarrhea episode (moderate or severe). Each SP played all eight possible SP roles to help control for individual SP effects. We received institutional review board (IRB) approval from RAND’s Human Subjects Protections Committee in the United States and the Karesa Independent Ethics Committee in India as well as approval from the Indian Council of Medical Research. See section 1.3.1 of the SM for more details on the SP design and overall data collection process. Experimental design

Our SP experiment included four study arms. To estimate the effect of provider perceptions of patient preferences on ORS prescribing, we randomized the treatment preferences that SPs expressed to the provider during the opening statement; some expressed a preference for ORS, some expressed a preference for antibiotics, and some expressed no preference (see table S20 for exact opening statements). To estimate the effect of financial incentives to sell more-lucrative medicines, we randomized half of the SPs who expressed no preference to also inform the provider that they would purchase any prescribed medication from a pharmacy in their hometown, thus eliminating the financial incentive at the point of sale. To estimate the effect of stock-outs, we randomized half of the providers to receive a 6-week supply of ORS before the SP visit. This generated exogenous variation in stock-outs and thus enabled us to isolate the causal effect of stock-outs on ORS dispensing. Section 1.3 of the SM describes the experimental design in more detail. We used these three experiments to test the following main hypotheses: H1: Providers will be more likely to prescribe ORS (antibiotics) when the patient shows a preference for ORS (antibiotics). H2: Providers will be more likely to prescribe ORS and less likely to prescribe antibiotics when there is no financial incentive to sell morelucrative medicines.

9 February 2024

H3: Providers will be more likely to recommend ORS when they have it available to dispense at their clinic. We then combined the causal effects estimated in the randomized controlled trials (RCTs) with population level estimates of the prevalence of each barrier (e.g., the share of providers who had an ORS stock-out) to estimate the effect of eliminating each barrier on ORS prescribing in the population. This allowed us to estimate what ORS prescribing would look like in the private sector if each barrier were removed. Results

In this section, we report all analyses and results according to a preanalysis plan that we registered at ClinicalTrials.gov (NCT04833790) and at the American Economic Association’s RCT registry (AEARCTR-0007276). Our empirical approach to estimate results is described in more detail in section 1.6 of the SM. Preliminary analyses

Table 1 describes status quo prescribing or dispensing for child diarrhea among providers who were not assigned to receive ORS supply and where the SP expressed no preference. With no intervention, 28% of providers prescribed or dispensed ORS to SPs, although 86% said they would prescribe ORS in a provider survey (see table S1), and 69% prescribed or dispensed antibiotics, although only 50% said they would. This shows that the know-do gap in child diarrhea care uncovered by Mohanan et al. (9) persists more than a decade later. This also highlights that a lack of knowledge about the standard of care is likely not the root cause of underprescribing of ORS. On average, providers dispensed 1.8 different medicines, and the visit cost INR 115 (USD 1.44). Providers spent about 4 min on average with our SPs. Providers in Bihar were more likely to prescribe ORS and/or antibiotics, spend more time with the SPs, and dispense medication directly to the patient compared with providers in Karnataka. Figure S2 shows SP treatment outcomes with more granularity. Table S5 shows that pharmacies were least likely to prescribe or dispense ORS (16%) and that MBBS providers were most likely to prescribe or dispense ORS (43%). Table S6 shows that providers in the four SP study arms were balanced on key characteristics, and table S7 shows that the two ORS supply arms were also balanced. Primary analyses

Our primary analyses followed a preregistered analysis plan, and all subgroups were prespecified (19). This section describes the results of our main hypotheses. H1: Showing a preference for ORS nearly doubled ORS prescribing or dispensing

Table 2 shows that when SPs expressed a preference for ORS, providers were 27 percentage 2 of 8

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

Table 1. Prescribing in no-preference control group. Data were recorded from SP visits from SPs who showed no treatment preference and providers who did not receive ORS supply.

Pooled (n = 275)

Outcome

Bihar (n = 146)

Karnataka (n = 129)

Prescribed or dispensed ORS 0.276 0.308 0.240 ..................................................................................................................................................................................................................... Prescribed or dispensed antibiotics 0.691 0.781 0.589 ..................................................................................................................................................................................................................... Prescribed or dispensed zinc 0.091 0.123 0.054 ..................................................................................................................................................................................................................... Prescribed or dispensed ORS and zinc 0.047 0.068 0.023 ..................................................................................................................................................................................................................... Directly dispensed any medicines 0.476 0.521 0.426 ..................................................................................................................................................................................................................... Number of medicines prescribed or dispensed 1.8 2.2 1.5 ..................................................................................................................................................................................................................... Cost of visit (INR) 116 157 69 ..................................................................................................................................................................................................................... Time spent with provider (min) 4.2 4.8 3.5 .....................................................................................................................................................................................................................

points more likely to prescribe or dispense ORS (96% increase); 28.0% prescribed or dispensed ORS when SPs expressed no preference, and 55.1% prescribed or dispensed ORS when SPs expressed an ORS preference (p < 0.01; top panel of Table 2). Providers were also 10 percentage points less likely to prescribe antibiotics when SPs expressed a preference for ORS (14% decrease). When SPs expressed a preference for (inappropriate) antibiotics, providers were 7.3 percentage points more likely to prescribe antibiotics (10% increase); 70.4% prescribed antibiotics when SPs expressed no preference, and 77.7% prescribed antibiotics when SPs expressed a preference for antibiotics (p < 0.01). Expressing a preference for antibiotics did not change ORS prescribing. Table S8 shows that the effect of showing an ORS preference was strongest among providers with no formal training, with a 47 percentage point increase for pharmacies and a 41 percentage point increase for RMPs, and weakest among the providers with the most training, with an 8.9 percentage point increase for MBBS providers (p value of difference in effect was 30% of their patients want ORS are about twice as likely to prescribe or dispense ORS compared with providers who think 5.0

0

With O3 chemistry

-150

-100

-50

0

longitude

50

100

150

Pollinator recognition distance (km)

B

Pollinator recognition distance (km)

A

-150

>5.0

0

With NO3 chemistry -100

-50

0

50

100

150

longitude

C

100

Current % of pre-industrial

NO3 and O3 levels found in either the polluted urban or near-pristine environments (Fig. 3A). For Hyles, NO3 oxidation of the floral odor eliminated their behavioral attraction (Fig. 3B). For Manduca, which was more sensitive to the floral odor (fig. S7), NO3 oxidation resulted in a 50% decrease in Manduca visitation rate (P = 0.047, comparison of population proportions) and to a level that was not significantly different from that of the solvent (no flower odor) control. By contrast, O3 oxidation of the floral odor—at O3 concentrations typical of highly polluted environments (120 ppb)—had no impact on hawkmoth visitation (Fig. 3B). Our scent oxidation experiments showed that a subset of monoterpenes in the floral odor were degraded with NO3 exposure. However, the decline of hawkmoth attraction could be due to the decreased monoterpene concentrations or the moth’s perception of the oxidation products from the odor reacting with NO3. Examination of the oxidation products from our flowtube experiments showed the production of organic nitrates associated with the odor compounds (fig. S6). We created a synthetic mixture of O. pallida scent compounds to simulate the selective depletion of monoterpenes by NO3 (called the NO3-proxy mixture). The NO3-proxy mixture (with 84% less b-pinene and 67% less b-ocimene, but lacking the oxidation products) elicited significantly fewer responses than did the untreated flower odor and the same amount of responses as those with the clean air (no odor) control and the NO3-degraded scent that contained the oxidation products. Controls of the floral odor exposed to nitrogen dioxide (NO2; the nonreactive precursor to NO3) or O3 alone (to control for any physiological effects on the moth) were not significantly different from responses to the untreated floral odor (Fig. 3B). Taken together, our results show that the moth’s inability to navigate and recognize the flower is from NO3 selectively degrading a subset of compounds in the scent and not because of the moth perception of the oxidation products. To determine how oxidation of floral scents may affect plant-pollinator interactions in the field, we conducted experiments to test various scent oxidation treatments in our Grants Pass, Washington, site (Fig. 3C). Treatments included real flowers, artificial flowers emitting the floral untreated or NO3-degraded scent, or artificial flowers not emitting a scent (visual control) (28). Visitation rates to the floral odor—both real floral odor (one flower) and our synthetic floral odor—and to real flowers were not significantly different (Fig. 3D). By contrast, the visitation rate to floral odor exposed to NO3 was significantly less than to the untreated odor (P = 0.027, generalized linear model, Poisson, logistic link) (tables S10 and S11) and not significantly different from the clean air control (Fig. 3D). To establish the effects of scent oxidation on fruit set, we determined the

80 60 40 20 0

Fig. 4. Global impacts of O3 and NO3 on pollinator recognition. (A) Map of floral scent recognition distance by using O3 degradation of the volatiles b-pinene and cis-b-ocimene, with degradation thresholds of 84 and 67%, respectively, and horizontal wind speed from the bottom grid of the GEOS-Chem model. The NO3 and O3 distributions were generated by using GEOS-Chem standard 12.1.0 with the 2013 emissions inventory and 2013 meteorology with a 2°-by-2.5° grid and 72 vertical levels to 0.01 hPa for the monthly average of January 2013 to February 2014. The bottom vertical level and the average concentrations for July 2013 (northern latitudes) and January 2013 (southern latitudes)—summer periods when the pollinators were present—were used. (B) Map of floral scent–recognition distance by using NO3 degradation of the volatiles b-pinene and cis-b-ocimene with the same conditions as in (A), and with data from the same GEOS-Chem model run. (C) Map of 2013 floral scent– recognition distance from (B) divided by preindustrial floral scent–recognition distance as a percentage. The preindustrial NO3 and O3 distributions were generated by using GEOS-Chem 13.2.1 classic with the 2013 nonanthropogenic emissions inventory and 2013 meteorology with a 4°-by-5° grid and 72 vertical levels to 0.01 hPa for January 2013 to December 2013. The bottom vertical level and the average concentrations for July 2013 (northern latitudes) and January 2013 (southern latitudes) were used. The outputs for the plots in (A) to (C) were masked by using a land-ocean mask to remove the values over bodies of water.

9 February 2024

4 of 5

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

and has been evaluated against observations in several studies (28). Using known rate constants for reactions of monoterpenes with O3 and NO3, we calculated the distance for floral scents to be oxidized to a level unrecognizable by hawkmoths, given the simulated concentrations of O3 and NO3 and wind velocities within the lowest grid level in the GEOS-Chem model (34). A scentrecognition distance was computed for each location and for the respective summer months (July for the Northern Hemisphere and January for the Southern Hemisphere) and plotted as a series of global maps that illustrate the differing impacts of NO3 and O3 on pollinator perception of the floral scents (Fig. 4). Results from the model show that NO3 degradation of monoterpenes in the floral scent has a more severe impact on recognition distance than that by O3 alone (Fig. 4, A and B) and that scent-recognition distances are reduced to below 400 m in many populated areas. Regions with the most severe impacts from NO3 include North America, Europe, Central Asia, the Middle East, and southern Africa. In addition, we performed a simulation of the preindustrial atmosphere using GEOS-Chem to assess the percent change in scent-recognition distance that has occurred since the preindustrial era (Fig. 4C). The comparison map shows that in most populated regions of the world, there has been a 75% or more decrease in scent-recognition distances since preindustrial times (Fig. 4C). In certain sparsely populated areas (such as Greenland), NO3-related changes are relatively small and do not affect the scentrecognition distances. In other areas (parts of Southern Africa), scent-recognition distances may be unchanged relative to preindustrial times because of both natural O3 and NOx emissions and chemical feedbacks that stem from their couplings with arboreal VOC emissions (fig. S9 and table S13) (19). Geographic areas may thus differ widely in the impact of NO3 on pollinator recognition of floral scents. Over the past 10 years, annual variation in NO3 may occur, especially in regions with biomass burning or other meteorological effects. However, when comparing across years (2013, 2019, and 2021), the model showed similar global trends in O3 and NO3 concentrations and distributions, with level impacts orders of magnitude greater when compared with preindustrial conditions (table S13) (30), which is consistent with anthropogenic inputs of NO3 around urban areas. Our results demonstrate that atmospheric O3 and NO3 oxidation affects nocturnal pollinator visitations in the field by changing floral scent chemical composition, reducing scent-recognition distances. As a further example of the impacts of NO3 on many common floral volatiles, we compared the oxidation rates of diverse compounds with NO3 and O3 under mean environmental conditions of the northern latitudes (fig. S8B and table S12). Building on past atmoChan et al., Science 383, 607–611 (2024)

spheric chemistry work on the effects of NO3 (20, 21), our results show that most floral compounds had significantly greater reactivity toward typical NO3 concentrations than O3, except for the sesquiterpenes a-humulene and b-caryophyllene (fig. S8B and table S12). The monoterpenes were the most susceptible to O3 and NO3 degradation of the floral scent compounds analyzed, with certain monoterpenes being more reactive than others. Many studies (7, 11, 35–37) have established that monoterpenes are ubiquitous floral volatiles important for scent recognition by pollinators; our results demonstrate that certain subsets of compounds in the scent are more sensitive to anthropogenic pollutants and generalize to other systems that use NO3-reactive monoterpenes as key volatiles for scent recognition. Although other classes of volatiles such as sesquiterpenes, green leaf volatiles, and aromatics have different reactivity profiles, our analytical framework can be used to estimate the impacts of O3 and NO3 on scent-recognition distances in other systems if the relevant volatiles and rate constants can be determined. Olfaction and chemical signaling mediate diverse ecological and evolutionary processes, including predator-prey interactions, host selection, and mate selection (38, 39). At the population level, our results indicate that nitrate radicals, stemming from nitrogen oxide pollution, negatively affect both plants (by decreasing fitness) and insect pollinators (by decreasing their ability to locate nectar resources) and in a regionspecific manner. Future work is needed to determine the community-level response to anthropogenic pollutants and to identify how different ecological processes are affected. Anthropogenic pollutants are temporally and regionally variable, and it will be necessary to characterize these impacts in different geographic locations to understand and ultimately mitigate these effects. RE FERENCES AND NOTES

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19.

9 February 2024

D. M. Dominoni et al., Nat. Ecol. Evol. 4, 502–511 (2020). W. Halfwerk, H. Slabbekoorn, Biol. Lett. 11, 20141051 (2015). B. M. Siemers, A. Schaub, Proc. Biol. Sci. 278, 1646–1652 (2011). B. M. Van Doren et al., Proc. Natl. Acad. Sci. U.S.A. 114, 11175–11180 (2017). A. C. Owens et al., Biol. Conserv. 241, 108259 (2020). E. Agathokleous, Z. Feng, J. Peñuelas, Trends Ecol. Evol. 37, 939–941 (2022). R. D. Girling, I. Lusebrink, E. Farthing, T. A. Newman, G. M. Poppy, Sci. Rep. 3, 2779 (2013). B. Cook et al., J. Chem. Ecol. 46, 987–996 (2020). F. Démares, L. Gibert, P. Creusot, B. Lapeyre, M. Proffit, Sci. Total Environ. 827, 154342 (2022). S. Dötterl, M. Vater, T. Rupp, A. Held, J. Chem. Ecol. 42, 486–489 (2016). G. Farré-Armengol et al., New Phytol. 209, 152–160 (2016). J. D. Fuentes, M. Chamecki, T. Roulston, B. Chen, K. R. Pratt, Atmos. Environ. 141, 361–374 (2016). Q. S. McFrederick, J. C. Kathilankal, J. D. Fuentes, Atmos. Environ. 42, 2336–2348 (2008). J. M. W. Ryalls et al., Environ. Pollut. 297, 118847 (2022). J. M. W. Ryalls et al., Proc. Biol. Sci. 289, 20221692 (2022). J. A. Riffell, L. Abrell, J. G. Hildebrand, J. Chem. Ecol. 34, 837–853 (2008). D. H. Ehhalt, Phys. Chem. Chem. Phys. 1, 5401–5408 (1999). N. L. Ng et al., Atmos. Chem. Phys. 17, 2103–2162 (2017). R. A. Graham, H. S. Johnston, J. Phys. Chem. 82, 254–268 (1978).

20. R. Atkinson, S. M. Aschmann, A. M. Winer, J. N. Pitts, Environ. Sci. Technol. 18, 370–375 (1984). 21. R. Atkinson, S. M. Aschmann, A. M. Winer, J. N. Pitts, Environ. Sci. Technol. 19, 159–163 (1985). 22. D. P. Gregory, Aliso 5, 385–419 (1964). 23. R. E. Stockhouse, Am. Midl. Nat. 96, 241–245 (1976). 24. H. Simon, A. Reff, B. Wells, J. Xing, N. Frank, Environ. Sci. Technol. 49, 186–195 (2015). 25. W. A. Haber, G. W. Frankie, Biotropica 21, 155–172 (1989). 26. A. Lavi, M. P. Vermeuel, G. A. Novak, T. H. Bertram, Atmos. Meas. Tech. 11, 3251–3262 (2018). 27. J. Krechmer et al., Anal. Chem. 90, 12011–12018 (2018). 28. Materials and methods are available as supplementary materials. 29. M. Huang et al., Atmos. Chem. Phys. 17, 5721–5750 (2017). 30. M. Khan et al., Atmos. Res. 164-165, 347–357 (2015). 31. M. Vanderplanck et al., Antioxidants 10, 636 (2021). 32. D. A. Satterfield, T. S. Sillett, J. W. Chapman, S. Altizer, P. P. Marra, Front. Ecol. Environ. 18, 335–344 (2020). 33. R. Gelaro et al., J. Clim. 30, 5419–5454 (2017). 34. K. R. Mylne, Boundary-Layer Meteorol. 60, 15–48 (1992). 35. J. T. Knudsen, R. Eriksson, J. Gershenzon, B. Ståhl, Bot. Rev. 72, 1–120 (2006). 36. G. Farré-Armengol, M. Fernández-Martínez, I. Filella, R. R. Junker, J. Peñuelas, Front. Plant Sci. 11, 1154 (2020). 37. J. A. Riffell, H. Lei, L. Abrell, J. G. Hildebrand, Science 339, 200–204 (2013). 38. T. Eisner, J. Meinwald, Proc. Natl. Acad. Sci. U.S.A. 92, 1 (1995). 39. B. S. Hansson, M. C. Stensmyr, Neuron 72, 698–711 (2011). 40. J. A. Riffell, Olfaction in the Anthropocene: NO3 negatively impacts floral scent and nocturnal pollination [Dataset]. Dryad (2024); https://doi.org/10.5061/dryad.vt4b8gtzk. 41. J. A. Riffell, riffelllab/Chan-et-al: Chan et al - Olfaction in the Anthropocene (version v1). Zenodo (2024); https://doi.org/10.5281/zenodo.8411927. 42. The International GEOS-Chem User Community, geoschem/ geos-chem: GEOS-Chem 12.1.0 (12.1.0). Zenodo (2018); https://doi.org/10.5281/zenodo.1553349. 43. The International GEOS-Chem User Community, geoschem/ GCClassic: GEOS-Chem 13.2.1 (13.2.1). Zenodo (2021); https:// doi.org/10.5281/zenodo.5500717. AC KNOWLED GME NTS

The authors thank T. Daniel and J. Hille Ris Lambers for technical advice and assistance, B. Nguyen for help in the moth insectary, E. Sammeth and R. Mettey for help with the insectary and GC-EAD, and R. Wolf for permission to use his image in Fig. 1B. We thank G. Davidowitz and C. Francois for providing the Hyles lineata and the Washington Department of Fish and Wildlife for access to the field sites. Funding: This work was supported by Air Force Office of Scientific Research grants FA9550-21-1-0101, FA9550-20-1-0422, and AWD004055-G4 a.m.01 (J.A.R.); the National Science Foundation under grants 2121935 (J.A.R.) and 2202287 (B.A.); the National Institutes of Health under grant R01AI148300 (J.A.R.); the Human Frontier Science Program under grant RGP0044/2021 (J.A.R.); the Danish National Research Foundation under grant DNRF168 (J.K.C.); an Endowed Professorship for Excellence in Biology (J.A.R.); and endowments from L. Riddiford, J. Truman, R. T. Paine, J. S. Edwards, and B. Hall (J.K.C.). Author contributions: Conceptualization: J.K.C., J.A.T., and J.A.R. Methodology: J.K.C., R.A., S.P., R.X., U.A.J., B.A., J.M.L., J.A.T., and J.A.R. Investigation: J.K.C., R.A., S.P., R.X., U.A.J., B.A., J.M.L., J.A.T., and J.A.R. Visualization: J.K.C., R.A., J.A.T., and J.A.R. Funding acquisition: J.K.C., R.A., J.A.T., and J.A.R. Project administration: J.A.T. and J.A.R. Supervision: J.K.C., J.A.T., and J.A.R. Writing – original draft: J.K.C., J.A.T., and J.A.R. Writing – review and editing: J.K.C., R.A., S.P., R.X., U.A.J., B.A., J.M.L., J.A.T., and J.A.R. Competing interests: The authors declare that they have no competing interests. Data and materials availability: Data are available at (28) and (40), and custom code is available at (41). GEOS-Chem code is available at (42) and (43). 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.adi0858 Materials and Methods Equation S1 Figs. S1 to S9 Tables S1 to S13 References (44–79) MDAR Reproducibility Checklist Submitted 4 April 2023; accepted 4 January 2024 10.1126/science.adi0858

5 of 5

RES EARCH

MASS SPECTROMETRY

Differentiating enantiomers by directional rotation of ions in a mass spectrometer Xiaoyu Zhou1, Zhuofan Wang1, Shuai Li1, Xianle Rong2, Jiexun Bu3, Qiang Liu2, Zheng Ouyang1,4* Conventional mass spectrometry does not distinguish between enantiomers, or mirror-image isomers. Here we report a technique to break the chiral symmetry and to differentiate enantiomers by inducing directional rotation of chiral gas-phase ions. Dual alternating current excitations were applied to manipulate the motions of trapped ions, including the rotation around the center of mass and macro movement around the center of the trap. Differences in collision cross section were induced, which could be measured by ion cloud profiling at high resolutions above 10,000. High-field ion mobility and tandem mass spectrometry analyses of the enantiomers were combined and implemented by using a miniature ion trap mass spectrometer. The effectiveness of the developed method was demonstrated with a variety of organic compounds including amino acids, sugars, and several drug molecules, as well as a proof-of-principle ligand optimization study for asymmetric hydrogenation.

B

iological systems normally favor one of the two mirror-image forms, or enantiomers, of the amino acids, sugars, and nucleotides, respectively, present in proteins, carbohydrates, DNA, and RNA (1–4). Specifically, chiral amino acids mostly exist as L-enantiomers, whereas carbohydrates are mainly D-enantiomers (5). Owing to the intrinsic chiral environment of biological systems, enantiomeric drug molecules often show different physiological behaviors and pharmacological activities (6, 7). Using the therapeutic drug thalidomide as an example, the R-enantiomer is sedative, whereas the S-enantiomer is teratogenic (8, 9) and its misuse led to birth defects in more than 10,000 newborns (10). Differentiation of enantiomers is crucial for drug development as well as for disease diagnosis (4, 11). Analysis of enantiomers can be achieved based on their differential interaction with electromagnetic fields, as manifested in optical rotation of polarization, circular dichroism (12, 13), or anomalous dispersion in x-ray crystallography (14, 15); or through interactions with chiral selectors, such as the stationary phase in chromatography (16). Modern mass spectrometry (MS) systems serve as a powerful platform for high-throughput analysis of complex mixtures with high structural specificity. Although determination of enantiomeric excess is often done by coupling MS with gas chromatography or high-performance liquid chromatography (HPLC) with chiral stationary phases, direct differentiation of enantiomers by MS has remained as a major challenge. Ion mobility MS (IMMS) could be used to perform 1

State Key Laboratory of Precision Measurement Technology and Instruments, Department of Precision Instrument, Tsinghua University, Beijing 100084, China. 2Department of Chemistry, Tsinghua University, Beijing 100084, China. 3 PURSPEC Technology (Beijing) Ltd., Beijing 100084, China. 4 Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen 518055, China. *Corresponding author. Email: [email protected]

Zhou et al., Science 383, 612–618 (2024)

chiral analysis, whereas chiral references need to be introduced either for chemical interactions prior to IMMS analysis (17–20) or for gasphase collisions during tandem IMMS analysis (17, 18, 21–25). Recently, we developed an IMMS method simply by using an ion trap for both high-field ion mobility separation and mass analysis (26). Ion cloud profiling is performed under high E/N conditions [1 MTd; Td: Townsend number; E, the field strength of the radiofrequency (RF) field; N, the number density of the background gas], which led to super high resolutions above 10,000 (see supplementary materials for definition) for identifying the differences in molecular structures using high-field ion mobility. The motivation for this work was to explore the possibility of directly distinguishing enantiomers using this high-resolution IM method with small differences in collision cross sections that could be induced by electromagnetic fields inside the ion trap. It has been demonstrated that differentiation effects could be induced by electromagnetic fields, often based on the polarity differences of the enantiomers (27–29). Mirrored propeller configurations have been used to illustrate the differences that could be induced for enantiomers owing to the directional rotations (29–31) (Fig. 1A). Hypothetically, manipulating the rotation directions using electromagnetic fields should be more effective for ionic than neutral states of the chiral compounds. In this work, we discovered differences in collision cross sections induced for enantiomeric ions in the ion cloud profiling process and subsequently developed a method to control the directional rotations of the ions using dual alternating current (ac) excitation, which enabled a direct separation and analysis of the enantiomers. A Mini b miniature mass spectrometer [PURSPEC Technology (Beijing) Ltd., Beijing, China] was modified with a dual-LIT (linear ion trap) configuration (32) and a dual-ac

9 February 2024

excitation for analysis of the enantiomeric ions. The ions were trapped in the LIT by an RF field, which set the E/N value above 1 MTd. Two ac resonance excitations were applied in both x and y directions, which induced the complex motions of the trapped ions (Fig. 1B and fig. S1). An excited ion could move around the center of the ion trap (macro motion) while rotating around its center of mass. The ions of both enantiomers had similar macro motions and rotational motions (Fig. 1B and fig. S2). Simulations were carried out using a pair of enantiomers of mass to charge ratio (m/z) 551 (see supplementary materials for detailed information for the setup of initial conditions and the induced differences in collisional cross section). The rotating electric field, ER-motion, drove the rotations of the ions (Fig. 1C), and the rotational directions of the ions could be tuned by varying the phase difference of the ac excitations, Dϕ ¼ ϕy  ϕx , where ϕx and ϕy represent the phases of the ac applied in the x and y directions of the trap, respectively. For instance, for Dϕ ¼ T p=2, R- and S-enantiomer ions had opposite rotational motions. The differences in the directional rotations resulted in differences in the drag force in the macro motion of the ions owing to the collisions with the background gas molecules. This led to differences in the amplitude of the macro motion (Fig. 1C) and the possible differentiation of the trapped enantiomeric ions. Through the scan of the ac voltage, the amplitudes of the enantiomeric ions increased accordingly and the ions were ejected sequentially from the trap (Fig. 1D). Scope and sensitivity of enantiomer differentiation

To experimentally test the directional rotation effect described above for differentiating enantiomeric ions, we first selected a pair of enantiomers, (R)-(-)-1,1′-Bi-2-naphthol bis (trifluoromethanesulfonate) and (S)-(+)-1,1′Bi-2-naphthol bis(trifluoromethanesulfonate), abbreviated as R- and S-binaphthyl-triflate respectively, for verification of the hypothesis and predictions by simulations (Fig. 2A). Binaphthyl-triflates have been previously selected as a model system in testing the propeller concept for enantiomer separation (29–31). The two naphthyl groups of relatively large cross sections serve as two blades of the propeller and potentially could help to enhance the difference in the drag force effect once directional rotations are established. Various types of derivatizations do not necessarily complicate the simulation described above, where the experimental measured differences were used to calibrate the damping coefficients for the simulations (see supplementary materials for detailed information). The simulated dual-ac excitation and the macro-motion amplitudes are shown in Fig. 2, B and C, respectively. The 1 of 7

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

A

C

Distinguishable?

=-

/

= /

ER-motion

No

Mirrored motions

ER-motion 15E4

R S

Ey [V/m]

Ey [V/m]

15E4

0

- 15E4 -15E4

0

R S

0

- 15E4 -15E4

15E4

Ex [V/m] Macro motion 1.0

15E4

Macro motion 1.0

R S

y / r0

y / r0

Yes

0

Ex [V/m]

0.0

R S

0.0

Symmetry-breaking motions

-1.0 -1.0

0.0

- 1.0 -1.0

1.0

D

=-

R-motion 1.00

r / r0

Macro motion y

1.0

x / r0

x / r0 B

0.0

/

Ejection

Ion Trapping Boundary R S

0.50

x 0.00

-300 Enantiomer

0.0

motion patterns of the ions could be tuned by varying the phase difference Dϕ of the dual-ac signals. It was clear that for Dϕ < 0, the macromotion amplitudes of S-binaphthyl-triflate ions were greater than those of the R-enantiomers, and the reverse was true forDϕ > 0. The dual-ac

4.184

4.188

and experienced different drag forces through their macro motions, resulting in differences in amplitudes. (D) Macro-motion amplitudes of the simulated ion trajectories during the ion cloud profiling process with ac amplitude scan. Enantiomeric ions were ejected sequentially, depending on their damping coefficients, when their oscillation amplitudes exceeded the trap boundary, indicated by the dashed line. The simulated R- and S-enantiomers of m/z 551 were set with identical initial conditions of ion motion but different reduced damping coefficient: b′ ¼ 3.230 × 10−4 for the R-form and b′ ¼ 4.560 × 10−4 for the S-form. b′ ¼ 2b=Wm, where W is the angular frequency of the RF field, m is ion mass, and b is the damping coefficient of the R- or S-enantiomers.

excitation with Dϕ ¼ T p=2 should produce the strongest effective rotating E fields and generate the largest directional rotation effect for ion motion separation (Fig. 2D and fig. S3). Experimental measurements performed accordingly by ion cloud profiling of protonated ions of

9 February 2024

5.0

Time [ ms]

[V]

Fig. 1. The principle, instrumental setup, and simulations for macro motions and directional rotations of enantiomeric ions trapped in LIT. (A) Schematics of mirrored (upper) and symmetry-breaking (lower) rotational motions. (B) Schematic of the macro motions with rotational (R-) motions of a pair of enantiomeric ions in a RF trapping field. Two additional ac excitations were applied in the x and y directions to manipulate the directional rotations. (C) Simulations of the rotating E field that drove the rotations (upper), which resulted in a detectable difference in amplitude of macro motion of the enantiomeric ions (lower). The phase differences, Dϕ ¼ p=2 (left) and p=2 (right), of the ac applied in the x and y directions enabled tunable directional rotations; the rotating R- and S-enantiomers had different collision cross sections

Zhou et al., Science 383, 612–618 (2024)

2.5

300

m/z 551, produced by nano-ESI (electrospray ionization), confirmed the detectable differences in macro-motion amplitude when Dϕ ≠ 0, as well as the maximum differences at Dϕ ¼ þp=2 or p=2 with reversed ion-ejection orders (Fig. 2E). The enantiomer mixture solution was 2 of 7

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

Fig. 2. Characterization of directional rotation effect for enantiomer separations. (A) Molecular structures of R- and S-binaphthyl-triflate enantiomers. (B) Rotating dual-ac excitation fields. (C) Lissajous figures of ion macro motion. (D) Measured VAC values of S-binaphthyl-triflate (red, upper) and R-binaphthyl-triflate (blue, upper) as a function of Dϕ. (E) Profiling spectra of R- and S-binaphthyl-triflate

prepared with concentrations of 75 and 25 mM for R- and S-binaphthyl-triflate, respectively. Resolutions >10,000 were obtained for analysis of the ionic enantiomers by ion cloud profiling. Drag forces of the ions could be characterized by the excitation voltage at the ion ejection, Vac. The Vac values of the R- and S-binaphthyl-triflate ions as a function of Dϕ are shown in Fig. 2D, and the corresponding spectra are shown in Fig. 2E and fig. S4. We optimized the E/N conditions by adjusting the gas pressure and the trapping voltage VRF of the RF field for maximizing the enantiomer separation and the analysis performance using ion cloud profiling (fig. S5). Zhou et al., Science 383, 612–618 (2024)

mixture with phase difference, Dϕ. The resolution is defined as VAC =DVAC, where VAC and DVAC are the ac excitation voltage and the full width at half maximum of the peak, respectively. Error bars represent the mean ± SD of five replicates. The curves were fit by B-spline. R- and S-binaphthyl-triflate concentrations were 75 and 25 mM, respectively.

This method can be easily implemented using an ion trap, which is also capable of MS/MS analysis for structural confirmation, and therefore holds potential for a broad range of applications. We further demonstrated the applicability of this method for analyzing enantiomers of a variety of different chemical biological compounds, including amino acids, carbohydrates, small-molecule drugs, and metabolite compounds. For drug compounds, enantiomers with opposite handedness may have completely opposite effects in the human body (8, 9). For instance, R-thalidomide is sedative while S-thalidomide is teratogenic, and D-penicillamine

9 February 2024

is antiarthritic whereas L-penicillamine is toxic. Consequently, a fast and facile means for identifying the chirality of the organic compounds is of considerable interest for drug discovery as well as drug compound synthesis. In this study, R- and S-penicillamine as well as R- and S-thalidomide were analyzed, with the ion cloud profiling spectra recorded as shown in Fig. 3, A and B, respectively. Another example shown in Fig. 3C is the analysis of enantiomeric mixtures of 2-hydroxyglutarate (2HG). The 2HG serves as a metabolite biomarker, but the L/D ratio of enantiomeric forms is as an indication of the cause of disease. Whereas mutations in 3 of 7

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

B

Penicillamine D

D

D

L

L

34

D

S

100

0

E

Glucose

Rel. Int.

0

R

38 42 VAC [mV]

S,R,S,S (L)

R

40

F

S,S,R,R (D)

Rel. Int.

Rel. Int.

0

L

100

L D

G

Leu

Phe

H

Trp

Rel. Int.

L

100 D L

Leu

0 100

Phe D

0 100

L

Trp D

0

D

34

Fig. 3. High-resolution separation of chiral compounds. Structures of molecule with single stereocenter and ion cloud profiling spectra of enantiomer mixtures of L- and D-forms (or R- and S-forms) at Dϕ ¼ p=2 (upper) and p=2 (lower): (A) penicillamine, (B) thalidomide, (C) 2HG. Structures of molecule with multiple stereocenters and ion cloud profiling spectra of enantiomer mixtures of L- and D-forms (or R- and S-forms) at Dϕ ¼ p=2 (upper) and p=2 (lower):

the genes encoding isocitrate dehydrogenase result in an elevation of D-2HG and acute leukemias and brain tumors (11, 33), some other metabolism errors would lead to a higher concentration of L-2HG (34, 35). MS analysis of 2HG for intro-surgery examination has been developed using miniature MS systems with direct sampling ionization (36, 37). With the method reported here, differentiation of the 2HG enantiomeric forms could be readily implemented for point-of-care purposes. This method has been tested for analyzing enantiomers with multiple stereocenters. Monosaccharides with four chiral centers, glucose and Zhou et al., Science 383, 612–618 (2024)

L

I

38 42 VAC [mV]

S,S

100

S,S

36 40 44 VAC [mV]

3 Trp Phe

2 Leu 1 0

140

145

150 [Å]

155

(D) glucose, (E) mannose, and (F) palonosetron. (G) Molecular structures and (H) Ion cloud profiling spectra of amino acid enantiomers: leucine (Leu), phenylalanine (Phe), tryptophan (Trp). (I) Measured peak separation DVAC values of amino acid enantiomers plotted as a function of collision cross section (CCS); CCS values are from reference (47). The chiral center (red) and ionization site (blue) are marked in the molecular structures.

mannose, were analyzed with spectra shown in Fig. 3, D and E, respectively. Palonosetron with two stereocenters was also analyzed (Fig. 3F), and its S enantiomeric form is used to treat nausea and vomiting caused by chemotherapy (38). High resolutions above 10,000 were obtained for analysis of all of these chiral compounds, and the order of the enantiomer ejection could be switched by adjusting Dϕ. In another experiment, we selected a group of three amino acids with increasing sizes for the side chains leucine, phenylalanine, and tryptophan (Fig. 3G). As shown in Fig. 3H, the order of ejection for L- and D-enantiomeric forms is the same at

9 February 2024

0

0

33 37 41 VAC [mV]

R,R,S,S (L)

S,S

R,R

0 33 37 41 VAC [mV]

Peak Separation VAC [mV]

0

40 44 VAC [mV]

R,R

D

R,S,R,R (D)

36

100

0 100

D

Palonosetron

D

D

L

100

R,R

L

100

0

0

D

44 48 VAC [mV]

Mannose

L

100

L

100

R

D

100

2HG

L

S

100

0

0

L

S Rel. Int.

Rel. Int.

100

C

Thalidomide

Rel. Int.

A

the same Dϕ, which is expected because they all have similar configurations with the amine group charged. Further peak separations were observed for amino acids with larger side chains (Fig. 3H). This could also be explained with the propeller analogy (Fig. 1A). The enantiomeric configurations with larger collision cross sections are expected to show a larger difference in terms of colliding with gas-phase molecules with symmetry-breaking motions. The measured peak separation DVAC values for each pair of enantiomers are plotted as a function of collision cross section as shown in Fig. 3I, which confirms this hypothesis. 4 of 7

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

Fig. 4. Determination of enantiomeric ratio. (A) Ion cloud profiling spectra and (B) calibration curve for S- and R-binaphthyl-triflate mixtures. For quantitative analysis of the mixture, the concentration of S-binaphthyl-triflate was 10 mM and the actual e.e. = ðcR  cS Þ=ðcR þ cS Þ varied from 0 to 0.82, where cR and cS are the concentrations of R- and S-binaphthyl-triflate, respectively. Measured e.e. = ðIR  IS Þ=ðIR þ IS Þ, where IR and IS are the intensity of R- and S-binaphthyl-triflate. Each value represents the mean ± SD of 15 replicates. (C) Profiling spectra and (D) calibration curve for S- and R-thalidomide mixtures. For quantitative analysis of the mixture, the concentration of S-thalidomide was 20 mM and the actual e.e. varied from 0 to 0.98. Each value represents the mean ± SD of 10 replicates.

We further explored the possibility of quantifying enantiopurity on the basis of the directional rotation effect using ion cloud profiling. The analysis was performed for a series of mixtures of R- and S-binaphthyl-triflate, with enantiomeric excess (e.e.) values from 0 to 0.82. The ion cloud profiling spectra were recorded as shown in Fig. 4A, and the measured e.e. values were calculated and plotted against the actual values as shown in Fig. 4B. Calibration curves for pure samples are shown in fig. S7. We characterized the quantitation of enantiopurity for the drug thalidomide (Fig. 4, C and D), with e.e. values for the mixture samples ranging from 0 to 0.98, which were confirmed by circular dichroism measurements (fig. S8). Good linearity was obtained for the calibration. Application to asymmetric catalysis optimization

This developed method could be used to facilitate the optimization of reaction conditions for enantioselective synthesis. For demonstration, we selected asymmetric hydrogenation (AH), an important type of reaction owing to Zhou et al., Science 383, 612–618 (2024)

its pronounced efficiency, atom economy, and extensive substrate versatility (39, 40). An essential facet of AH development involves the systematic and high-throughput exploration of reaction conditions to identify the optimal catalysts for specific substrates that enable high product enantioselectivity and yield (41–43). Currently, chiral-phase HPLC serves as the main analysis method for this purpose. The procedures for using chiral-phase HPLC to analyze newly synthesized compounds, however, typically require relatively large (greater than milligram) sample amounts and can be laborious and time-consuming, especially when compared with the method introduced here (fig. S9) (42). In this study, the ion cloud profiling with direction ion rotation was used for screening ligands in manganese-catalyzed asymmetric hydrogenation of quinoline (Fig. 5A and fig. S10) (44–46). A stainless-steel wire was inserted into the crude reaction product solution, adsorbing less than 10 ng on its tip (see supplementary materials for the method of estimation), and then inserted into a nano-ESI capillary containing 10 ml of acetonitrile. A high voltage was then applied to the metal wire to produce ions

9 February 2024

through nano-ESI, which were subsequently analyzed by the miniature dual-LIT mass spectrometer (Fig. 5, B and C, and fig. S9B). Tandem mass spectrometry and ion cloud profiling with directional ion rotation were performed for reaction product compounds at m/z 298, with both MS/MS and chiral-specific IMS spectra obtained for structural analysis (Fig. 5, C to E). A complete procedure for the entire analysis took less than 1 min. The IMS spectra were recorded for hydrogenation products from reactions with four different catalyst ligands— s L1, sL2, sL3, and sL4 (Fig. 5A and fig. S10, sL: ligands with S configuration)—as shown in Fig. 5E. The e.e. values associated with use of the four ligands were calculated and confirmed in separate measurements with chiral-phase HPLC (Figs. S12 and S13), indicating an optimal reaction condition using ligand sL4 with an e.e. obtained at 89.3% ± 0.6% for the enantioselective synthesis. Our study shows that when the enantiomers are trapped as ions, the directional rotational effect can break the chiral symmetry sufficiently to separate them based on differential collision behavior. This helped to physically separate 5 of 7

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

A

S

Catalyst with Ligand sL1, sL2, sL3

+

or sL4

H2 + R

BB

HV nanoESI

< 10 ng

C

D

Products m/z 298.4

100

0 250

300

m/z

266.4

50

E

205.3

298.4

0 150

350

m/z 298.4

205.3

100

Rel. Int.

MS/MS

50

Mini MS

10 µL acetonitrile

Crude products

Rel. Int.

250 m/z

S

266.4

350

R 89%

100

76%

sL4

Chiral analysis sL3

50

34% sL2

e.e. [%]

Fig. 5. Rapid screening of catalysts for asymmetric hydrogenation. (A) Asymmetric hydrogenation of 2-(3,4-dimethoxyphenethyl) quinoline catalyzed by manganese catalysts containing various ligands: sL1, (SC,RFC)-N-2-(1H-imidazol-2-yl)1-(2-bis[3,5-di-phenylphenyl]phosphine)-ferrocenylethylamine (pink); sL2, (SC,RFC)-N-2-(1H-4(5)phenyl-imidazol-2-yl)-1-(2-bisphenylphosphine)ferrocenylethylamine (orange); sL3, (SC, RFC)-N-2(1H-imidazol-2-yl)-1-(2-diphenylphosphino)ferrocenylethylamine (blue); sL4, (SC, RFC)-N-2(1H-benzo[d]imidazol-2-yl)-1-(2-diphenylphosphino)ferrocenylethylamine (purple). Here, C and FC stand for central chirality and planar chirality, respectively. (B) Procedure for direct analysis of crude reaction product using a miniature mass spectrometer. Crude product of less than 10 ng was sampled by the stainless-steel wire, transferred to the nanoESI tip, diluted by 10 ml of acetonitrile, and subjected to ESI for MS/MS analysis and e.e. analysis. (C) MS spectra of the crude reaction product. (D) MS/MS spectra of the isolated hydrogenation product ion (left) with structure shown to indicate the fragmentation (right). (E) Ion cloud profiling spectra (left) and measured e.e. values of reaction products (right) using catalyst with different ligands conditions: sL1(pink); s L2(orange); sL3(blue); sL4(purple). Measured e.e. = ðIS  IR Þ=ðIR þ IS Þ, where IR and IS are the intensity of R- and S-2-(3,4-dimethoxyphenethyl)1,2,3,4-tetrahydroquinoline. Each value represents the mean ± SD of 10 replicates.

6% sL1

34

the enantiomer ions in the trap through collisions with background gas molecules. This work demonstrates the selection of enantiomers without requiring chiral references and simple implementation using an ion trap, which can be readily applied with the use of small instruments and simple procedures for a broad range of applications. Further investigation and development are needed to establish a comprehensive understanding of the correlations between the excitation inducing the directional rotation effect and elucidation of the enantiomeric structures. Determination of the absolute chiral configurations might be challenging using this method alone; however, in combination with structure analysis by MS/ MS, using a single ion trap analyzer might provide viable solutions for fast chiral analysis at high specificity. We have demonstrated a quantitation at e.e. up to 98%, although higher e.e. values for measurement might be desirable Zhou et al., Science 383, 612–618 (2024)

38 42 VAC [mV]

for other applications. This, however, could be intrinsically limited by the dynamic range of ion traps, mostly due to the space charge problem. The use of ion traps of larger trapping capacity, or careful control of the number of trapped ions, could help to overcome this limitation. RE FERENCES AND NOTES

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

9 February 2024

Y. Xu, T. F. Zhu, Science 378, 405–412 (2022). J. S. Richardson, Nature 268, 495–500 (1977). H. Lam et al., Science 325, 1552–1555 (2009). Y. Liu, Z. Wu, D. W. Armstrong, H. Wolosker, Y. Zheng, Nat. Rev. Chem. 7, 355–373 (2023). A. Banreti et al., Nat. Commun. 13, 7059 (2022). S. Lechner et al., Nat. Commun. 14, 3548 (2023). J. F. Beary 3rd, C. R. Eaton, Science 257, 145–145 (1992). D. V. Augusti, R. Augusti, F. Carazza, R. G. Cooks, Chem. Commun. (19): 2242–2243 (2002). E. Dorey, Nat. Biotechnol. 18, 1239–1240 (2000). N. Vargesson, Birth Defects Res. C Embryo Today 105, 140–156 (2015). G. Notarangelo et al., Science 377, 1519–1529 (2022). S. Daly, F. Rosu, V. Gabelica, Science 368, 1465–1468 (2020). R. M. Kim et al., Nature 612, 470–476 (2022).

0 sL1

sL2

sL3

sL4

Ligand

14. K. Harrison, A. S. Mackay, L. Kambanis, J. W. C. Maxwell, R. J. Payne, Nat. Rev. Chem. 7, 383–404 (2023). 15. Y. Ma, P. Oleynikov, O. Terasaki, Nat. Mater. 16, 755–759 (2017). 16. A. Furusho et al., Anal. Chem. 91, 11569–11575 (2019). 17. C. Xie et al., Anal. Chem. 93, 859–867 (2021). 18. H. Wang, F. Wu, F. Xu, Y. Liu, C.-F. Ding, Anal. Chem. 93, 15096–15104 (2021). 19. W. A. Tao, R. G. Cooks, Anal. Chem. 75, 25A–31A (2003). 20. E. Zlibut, J. C. May, J. A. McLean, J. Am. Soc. Mass Spectrom. 33, 996–1002 (2022). 21. J. R. Enders, J. A. McLean, Chirality 21 (suppl 1), E253–E264 (2009). 22. X. Yu, Z.-P. Yao, Anal. Chim. Acta 968, 1–20 (2017). 23. W. A. Tao, D. Zhang, E. N. Nikolaev, R. G. Cooks, J. Am. Chem. Soc. 122, 10598–10609 (2000). 24. D.-Q. Han, Z.-P. Yao, Trends Analyt. Chem. 123, 115763 (2020). 25. P. Dwivedi et al.., Anal. Chem. 78, 8200–8206 (2006). 26. X. Zhou, Z. Wang, J. Fan, Z. Ouyang, Nat. Commun. 14, 1535 (2023). 27. D. Patterson, M. Schnell, J. M. Doyle, Nature 497, 475–477 (2013). 28. M. Avalos et al., Chem. Rev. 98, 2391–2404 (1998). 29. J. B. Clemens, O. Kibar, M. Chachisvilis, Nat. Commun. 6, 7868 (2015). 30. D. Schamel et al., J. Am. Chem. Soc. 135, 12353–12359 (2013). 31. Y. Zhang et al., Nat. Commun. 10, 3742 (2019).

6 of 7

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

32. X. Liu, X. Wang, J. Bu, X. Zhou, Z. Ouyang, Anal. Chem. 91, 1391–1398 (2019). 33. J.-A. Losman et al., Science 339, 1621–1625 (2013). 34. J. Zhao et al., Nat. Metab. 3, 1372–1384 (2021). 35. D. Ye, K.-L. Guan, Y. Xiong, Trends Cancer 4, 151–165 (2018). 36. R. Zou et al., Anal. Chem. 91, 1157–1163 (2019). 37. F. Pu et al., Anal. Bioanal. Chem. 411, 1503–1508 (2019). 38. A. De Leon, Proc. Bayl. Univ. Med. Cent. 19, 413–416 (2006). 39. R. Noyori, Angew. Chem. Int. Ed. 41, 2008–2022 (2002). 40. L. Alig, M. Fritz, S. Schneider, Chem. Rev. 119, 2681–2751 (2019). 41. M. R. Friedfeld, H. Zhong, R. T. Ruck, M. Shevlin, P. J. Chirik, Science 360, 888–893 (2018). 42. C. C. Wagen, S. E. McMinn, E. E. Kwan, E. N. Jacobsen, Nature 610, 680–686 (2022). 43. A. R. Rosales et al., Nat. Catal. 2, 41–45 (2019). 44. C. Liu et al., Angew. Chem. Int. Ed. 60, 5108–5113 (2021).

Zhou et al., Science 383, 612–618 (2024)

45. C. Liu, M. Wang, Y. Xu, Y. Li, Q. Liu, Angew. Chem. Int. Ed. 61, e202202814 (2022). 46. C. Liu, X. Liu, Q. Liu, Chem 9, 2585–2600 (2023). 47. X. Zheng et al., Chem. Sci. 8, 7724–7736 (2017).

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.sciencemag.org/about/science-licensesjournal-article-reuse

ACKN OWLED GMEN TS

We thank Y. Xia, X. Ma, and W. Zhang (Tsinghua University) for helpful discussions, and J. Yang (Tsinghua University) for help with the organic synthesis. Funding: This work was supported by National Natural Science Foundation of China, grant nos. 21627807, 21934003, and 22227807. Author contributions: Conceptualization: X.Z., Z.O. Investigation: X.Z., Z.W., S.L., J.B., X.R., Q.L. Writing – original draft: X.Z., Z.O. Writing – review and editing: X.Z., Z.O. Competing interests: Z.O. is the founder and J.P. is an employee of PURSPEC Technology, which is developing miniature MS systems. Data and materials availability: All data are available in the main text or the supplementary materials.

9 February 2024

SUPPLEMENTARY MATERIALS

science.org/doi/10.1126/science.adj8342 Supplementary Text Figs. S1 to S13 Tables S1 to S4 Equations S1 to S14 References (48–52) Submitted 18 July 2023; resubmitted 22 August 2023 Accepted 3 January 2024 10.1126/science.adj8342

7 of 7

RES EARCH

GEOLOGY

A deep reservoir for hydrogen drives intense degassing in the Bulqizë ophiolite Laurent Truche1*, Frédéric-Victor Donzé1, Edmond Goskolli2, Bardhyl Muceku3, Corinne Loisy4, Christophe Monnin5, Hugo Dutoit1, Adrian Cerepi4 Deep crustal production of hydrogen (H2) is a potential source of primary energy if recoverable accumulations in geological formations are sufficiently large. We report direct measurements of an elevated outgassing rate of 84% (by volume) of H2 from the deep underground Bulqizë chromite mine in Albania. A minimum of 200 tons of H2 is vented annually from the mine's galleries, making it one of the largest recorded H2 flow rates to date. We cannot attribute the flux solely to the release of paleo-fluids trapped within the rocks or to present-day active and pervasive serpentinization of ultramafic rocks; rather, our results demonstrate the presence of a faulted reservoir deeply rooted in the Jurassic ophiolite massif. This discovery suggests that certain ophiolites may host economically useful accumulations of H2 gas.

H

ydrogen (H2), like electricity, is a carbonfree energy carrier that plays an important role in modern industry and in the energy transition; however, most H2 is manufactured using natural gas through a process that consumes energy and releases large amounts of carbon dioxide into the atmosphere. A previously overlooked geologic source of H2 could contribute to diversifying our energy mix and reducing the carbon footprint of our economy. Although geologic H2 plays a central role in the abiotic synthesis of simple organic compounds in the Earth’s crust and in supporting deep microbial communities (1–4), the high mobility of the molecule combined with its high reactivity were thought to prevent accumulation in the subsurface except in rare cases (5–8). This paradigm is called into question by recent findings of H2-rich fluids in surface seeps and underground mines or boreholes in specific geological settings, along with a reanalysis of historical drilling data (9–14). In particular, H2-rich fluids produced by subsurface reactions accompanying the serpentinization of ultramafic rocks such as peridotites are well-documented in uplifted orogenic and ophiolitic bodies (15–21). Despite these observations, our understanding of the processes and settings most conducive to the formation of substantial accumulations of H2 are still evolving. Notwithstanding the possible importance of geologic H2 as a clean fuel or as an energy source for life, the current knowledge of H2 occurrences within the Earth’s lithosphere is limited, with recoverable and economic resources of H2 poorly quantified. The main reason for this lack of fundamental un1

Université Grenoble Alpes, CNRS, IRD, ISTERRE, UMR 5275, Grenoble, France. 2National Agency of Natural Resources, NANR-AKBN, Tirana, Albania. 3Polytechnic University of Tirana, FGJM, Department of Earth Sciences, Tirana, Albania. 4 Université Bordeaux, CNRS, Bordeaux INP, EPOC, UMR 5805, Pessac, France. 5Université Toulouse 3, CNRS, IRD, GET, UMR 5563, Toulouse, France. *Corresponding author. Email: [email protected]

Truche et al., Science 383, 618–621 (2024)

derstanding of the H2 system stems from the inherent challenges in sampling deep geological fluids and the lack of deep infrastructure targeting potentially fertile H2 settings. This leads to the pivotal question of whether there is a H2 geological system comparable to that of petroleum. Our study unveils a high emission rate of almost pure geologic H2, suggesting the potential for a new extractable primary energy source. The Bulqizë mine: geological context and intense H2 outgassing

We conducted an exploration campaign in the deep underground chromite mine of Bulqizë in Albania (Fig. 1), where the presence of flammable gas was first reported in 1992 at a depth of 620 m. After the discovery of the gas, major explosions occurred in 2011, 2017, and 2023. The Bulqizë chromite mine is one of the largest chromium extraction sites in the world, with total recovery of over 20 million tons of high-grade ore averaging 35 wt% Cr2O3. The mine is situated within the Bulqizë Jurassic ultramafic massif (fig. S6), approximately 40 km northeast of Tirana. This massif is a component of the giant Eastern Mediterranean suprasubduction-zone ophiolite belt that extends from Turkey to Slovenia over more than 3000 km and represents one of the largest and most complete preserved segments of oceanic lithosphere on Earth (22–26). The massif spans an area of 370 km2 to a depth of 6 km and hosts numerous folded and faulted concordant chromite ore bodies embedded in the mantle sequence (26–28). We observed very intense outgassing in the deeper levels of the mine, specifically at depths ranging from 500 to 1000 m below the surface (Fig. 1 and Table 1). The most intense gas discharge in the mine galleries is located in a tectonic zone, i.e., a highly faulted domain, where focused and intense bubbling is visible in drainage pools and runoff streams located at level L19 (fig. S1 and movie S1). The water in

9 February 2024

the mine has an origin distinct from that of the gas. The water percolates from the upper levels of the mine, mainly draining through the shafts, whereas the lower levels of the mine are apparently dry as miners reported no evidence of water inflow during the excavation of the deepest galleries (28). However, this does not rule out the involvement of meteoric distal water or water from other origins in serpentinization elsewhere in the geological formation or below. We measured a gas flow rate of 5 ± 1 L/s (at 25°C and 1.031 × 105 Pa) from several vigorous bubbling zones located in a small 30 m2 pool (28). This gas is composed of H2 (84.0 vol%) and CH4 (13.2 vol%) with minor concentrations of N2 (2.7 vol%). Therefore, the amount of H2 discharged in the gallery from this single pool is 11 t/yr (∼30 kg/day) to the minimum, as we did not account for the many minor bubbling points. Since 2017, advancements in monitoring technology have enabled precise measurements of the H2 flow rate throughout the mine (28). This has been made possible by the installation of H2 sensors and flow meters on both the ventilation circuit of inner shaft N9 (at L19) and a dense network of 38 boreholes, extending from L17 to L21 (170 m deep) through the fault zone (figs. S2 and S3). The stale air from L19 to L21, containing 0.40 vol% H2, is discharged through shaft N9 at a flow rate of 840 Nm3/min (29). This leads to a H2 flow rate of 3.4 Nm3/min or 158 t/yr. Notably, the small 30 m2 pool emits by itself 7% of the H2 flow vented through shaft N9, demonstrating the importance of the fault zone as a major drain or reservoir for H2 (28). The 38 interconnected boreholes were drilled to manage the discharge of H2 from the tectonic zone intersecting the mine galleries and the ore body. The boreholes are constantly flushed by 4530 Nm3/hr of air. The exhaust gases, containing 1.20 vol% H2, are channeled outside the mine through a dedicated pipe, thus providing an additional H2 flow rate of 54 Nm3/hr (42 t/yr). This flow rate has remained constant over a six-year observation period. Only a fraction of the total amount of air vented from the mine is monitored for H2. Indeed, 100 Nm3/sec of air is circulated through the mine by the main ventilation system. Thus, a minimum of 200 tons of H2 are released from the mine every year (1.0 × 108 mol/yr). Such a relatively large flow of H2 greatly exceeds the few outgassing rates previously reported from dry seeps and hyperalkaline springs hosted in ophiolites (Table 1). The H2 outgassing rates we report are minimum values that have been accurately measured and are not extrapolated from single point surface measurements and diffusive flux models. Revealing the presence of a deep reservoir

The elevated H2 output raises the question of whether the outgassing results from active 1 of 4

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

Fig. 1. Schematic 3D view of the Bulqizë underground chromite mine. (A) Map of the deeper levels with location of the H2-bearing fault zone indicated. (B) Locations where H2 outgassing rates have been measured. The entrance of the mine is at an altitude of 840 m above mean sea level (amsl). Table 1. Outgassing rates of H2 from different sites. Concentrations of H2 and CH4 in the free gas phase from the Bulqizë mine (N2 and O2 concentrations are given in table S1) and other ophiolite-hosted seeps and bubbling pools.

H2 outgassing site

Ref.

Area (m2)

H2 flow (t/yr)

H2 (vol%)

CH4 (vol%)

H2/CH4 (vol/vol)

Oman, Haylayn pool (bubbles + diffuse) 20 ∼200 0.158 86.4 6.7 12.9 ............................................................................................................................................................................................................................................................................................................................................ Oman, Misfah pool (bubbles + diffuse) 20 ∼1000 0.056 66.9 7.2 9.3 ............................................................................................................................................................................................................................................................................................................................................ Turkey, Chimaera (diffuse dry seeps) 21 2000 3.5 9.9 87.0 0.11 ............................................................................................................................................................................................................................................................................................................................................ Albania, Bulqizë mine, L19 pool (focused bubbling) This study 30 11 84.0 13.2 6.4 ............................................................................................................................................................................................................................................................................................................................................ Albania, Bulqizë mine, L17 tectonic zone (boreholes) This study 400 42 1.20 0.15 8.0 ............................................................................................................................................................................................................................................................................................................................................ Albania, Bulqizë mine, level L19 (shaft N9) This study ∼20,000* 158 0.40 0.05 8.0 ............................................................................................................................................................................................................................................................................................................................................

*The area of level L19 (~20,000 m2) corresponds to the horizontal projection of the mining works’ footprint, and not to the H2 outgassing area, which is unknown but mostly concerns the fault zone. These estimates provide an indication of the scale of the sites, even if the areas of surface and underground outgassing sites are not directly comparable.

serpentinization or from the liberation of fossil H2 entrapped within the rock. Evaluating the volume of rock involved in this H2 gas generation and outgassing is therefore crucial. We explored three scenarios that could account for the observed H2 flow rate of 1.0 × 108 mol/yr, emphasizing the respective roles of the fault zone, the mine’s drainage volume, and the Truche et al., Science 383, 618–621 (2024)

entire Bulqizë massif. Our aim is to differentiate between active (present-day) flows and stocks (i.e., accumulations), with the latter arising from a combination of ancient (possibly fossil) and ongoing H2 generation processes. Although multiple sources of geologic H2 in the subsurface are recognized (6, 8), our analysis concentrates on three scenarios that are

9 February 2024

particularly relevant to the ophiolitic context (5, 8, 17–21, 28, 30). Our first scenario considers both the decrepitation of fluid inclusions and the release of H2 occluded in microporous minerals and microfractures. In this case, H2 is considered a paleo byproduct of the early stages of a serpentinization process that started 165 to 160 Myr ago (26). 2 of 4

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

Fig. 2. Three different scenarios for H2 production. Life span of a constant H2 flow rate of 1.0 × 108 mol/yr as function of the drained volume of rock and porosity, according to three different scenarios. Scenario 1: decrepitation of paleo H2 occluded in the rock (dark blue); Scenario 2: active serpentinization at low temperature (yellow); Scenario 3: release of H2 trapped inside the fault zone which acts as a reservoir (green). The gray area represents the estimated volume range of the fault zone acting as a reservoir.

In our second scenario, we assume that the observed H2 flow within the mine results from present-day low-temperature serpentinization. In our third scenario, we assume that the release of H2 occurs when a previously sealed fault zone is opened during mining operations. In this latter case, H2 is stored within the fractures and connected porosity of the fault zone at a filling flow rate substantially lower than the outgassing rate measured today within the mine. For all three scenarios, we calculated the life span of the measured current H2 flow rate in the mine as a function of the drained volume of rock (Fig. 2). We assume a rock density of 3000 kg/m3 and consider an average temperature and hydrostatic pressure of 100°C and 30 MPa at 3 km depth, respectively. We also use these parameters to estimate an equivalent porosity containing gaseous H2 (28, 31). We also consider a 3-km average depth for the ophiolite layer due to its bowl-shaped geometry, with a maximum depth of 6 km (26). In scenario 1, which involves the release of paleofluids occluded in the rocks (20, 28, 30, 32, 33), we measured a range of H2 content from 7.5 ×10−6 to 39.0 ×10−6 mol/kgrock (average = 15.4 × 10−6 mol/kgrock; n = 9 samples) Truche et al., Science 383, 618–621 (2024)

in the bulk harzburgite, dunite, and chromitite rock samples collected in the deepest levels of the mine (table S2 and fig. S4). Given that the H2 flow rate in the mine has remained constant during the last 6 years, this requires a drained rock volume of 5 to 27 km3. If we assume that the mine is the sole outlet for H2, this implies that the total volume of ophiolite constituting the entire Bulqizë massif (i.e., 370 km2 × 3 km) would be depleted in 250 to 1300 years. Thus, this scenario seems unlikely, as all H2 within the massif would have vanished instantaneously on a geological timescale. In scenario 2, which involves the direct consequences of an active serpentinization process at depth in the presence of water, a volume of 1.2 × 10−4 km3 must be altered every year to account for the observed H2 flow rate of 1.0 × 108 mol/yr, assuming that 0.3 mol of H2 are produced per kg of serpentinized peridotite at temperatures below 100°C (20). If we consider a peridotite area of 1 km2, this would result in a serpentinization front advancing by 0.15 m per year. Notably, this rate is 250 times faster than the regional uplift rate of 0.6 mm/yr for the Bulqizë ophiolite (34). Thus, if the uplift is negligible and the velocity of the serpentini-

9 February 2024

zation front remained constant over time, the reaction front would have advanced beyond the maximum thickness of the ophiolite (6 km) in 40,000 years, implying that the H2 production potential should be exhausted by now. A similar conclusion can be reached by considering a maximum mine drainage volume of 135 km3, corresponding to the effective rainwater catchment area of the mine of 45 km2 (fig. S7) and a rock thickness of 3 km. In this case, the whole drainage volume should have been fully serpentinized in only 13.5 Ma whereas final exhumation of the massif is dated at 15 to 45 Ma (24). Another independent line of evidence to rule out present-day active serpentinization as the source of H2 monitored in the mine comes from the comparison of the observed H2 flow rate with the global estimates of H2 production rate from both the oceanic lithosphere and the Precambrian continental crust. The latter ranges from 1.0 × 1010 to 1.2 × 1012 mol/yr (35–37), meaning that the contribution from the Bulqizë mine alone would be 0.01 to 1% of this global flux. This amount is an unrealistic contribution, considering the small size of the Bulqizë ultramafic massif compared with both the 80,000-km-long oceanic ridge system and the 1.06 × 108 km2 Precambrian shield surface area. Therefore, H2 produced by serpentinization or any other processes must have accumulated over a long period of time in a reservoir rock, the most realistic one being the fault zone as indicated by its intense degassing. This conclusion does not mean that decrepitation (scenario 1) or serpentinization (scenario 2) do not happen, but they alone cannot explain the actual resulting flow rate. In scenario 3, we consider that the flow measured in the mine results from the degassing of H2 gas trapped in the fault zone only after being produced. The mine perforated the top of the fault zone at several points when it reached a depth ranging from 0.5 to 1 km, releasing the gas stored in this sealed volume, which acts as a porous reservoir. Based on in situ observations (28), the fault zone is ~10 m wide, with a length varying from 100 m to 1 km, and a maximum height of 5 km (with the sealed top being around 500 m deep in the mine). These dimensions yield volumes of rock ranging from 5.0 × 10−3 to 5.0 × 10−2 km3 (represented by the gray area in Fig. 2). The total pore space volume generated by the overall amount of fractures present in this highly damaged zone is unknown, so we assume that the equivalent porosity has an average value of 5% at depth as measured in fault zones in Oman ophiolites (38). Thus, a fault volume of 1.3 × 10−3 km3 with a porosity of 5% would be sufficient to sustain the observed H2 flow rate for 6 years. Considering the estimated range of fault volume and the same porosity, the measured flow rate could be sustained for 25 to 238 years. In other words, the total amount of H2 stored in the fault zone would range from 3 of 4

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

5000 to ~50,000 tons. The temperature ranging from 40° to 160°C (28), along with the dry conditions in the deep horizons of the ophiolite may have hampered consumption of H2 by both microbial activity (too hot) and abiotic redox (too cold) reactions, at least in the deepest part of the reservoir. Hydrogen-rich fluids isolated in the crust for billions of years have already been reported in deep fracture networks from underground mines and boreholes in Precambrian shields demonstrating that H2 can accumulate over geological timescales (9, 39). Implications for geologic H2 exploration

What sets our discovery apart is the large flux of almost pure H2 gas we have observed. In the context of energy transition, our findings could substantially affect the ongoing search for new energy resources. We reveal that ophiolites, which are mantle rocks from the oceanic crust obducted onto continents, not only constitute effective source rocks, but also have the potential to host high-quality, H2-rich gas reservoirs. These geological formations, prevalent across all the continents, extend beyond being mere geological anomalies. In fact, many ophiolites worldwide have been found to contain springs or seeps that release H2 gas. Within this geological context, H2 exhibits potential for commercial extraction as it can be focused into fracture zones and trapped. In the past, the oil and gas industry has largely ignored ophiolites, considering them unsuitable for hydrocarbon resource exploitation. However, these onshore formations might offer potential for large-scale H2 accumulations and therefore constitute a promising target for H2 exploration. A key point is the presence of drainage systems such as faults and tectonic zones in these geological settings. As a result, a holistic understanding of the tectonic and petrophysical factors influencing the migration pathways and accumulation of H2 is needed to guide exploration. The configuration and properties of the seal remain uncertain, but the ore body and the necking of the faults may play crucial roles. Chromitite, commonly found

Truche et al., Science 383, 618–621 (2024)

in ophiolites, is noteworthy as several known H2-rich seeps are situated near chromite mines (32, 40, 41). The possible connection between chromitite and H2 emissions, though not yet confirmed, must be ascertained. Finally, in the emerging quest for geologic H2, it is crucial to consider both the nature of these H2 resources— whether they are ancient (fossil) or recent—and the potential impact on deep-seated microbial ecosystems that thrive on H2 (42, 43). A comprehensive understanding of these factors is essential to mitigate risks and ensure a sustainable development of H2 resources. RE FERENCES AND NOTES

1. D. S. Kelley et al., Science 307, 1428–1434 (2005). 2. C. Greening et al., ISME J. 10, 761–777 (2016). 3. L. Truche, T. M. McCollom, I. Martinez, Elements 16, 13–18 (2020). 4. E. P. Reeves, J. Fiebig, Elements 16, 25–31 (2020). 5. N. J. Smith, T. J. Shepherd, M. T. Styles, G. M. Williams, “Hydrogen exploration: A review of global hydrogen accumulations and implications for prospective areas in NW Europe” in Petroleum Geology Conference Series. (The Geological Society, 2005); vol. 6, pp. 349–358. 6. V. Zgonnik, Earth Sci. Rev. 203, 103140 (2020). 7. C. J. Boreham et al., APPEA J. 61, 163 (2021). 8. A. V. Milkov, Earth Sci. Rev. 230, 104063 (2022). 9. B. Sherwood Lollar et al., Astrobiology 7, 971–986 (2007). 10. V. A. Nivin, Appl. Geochem. 74, 44–55 (2016). 11. L. Truche et al., Earth Planet. Sci. Lett. 493, 186–197 (2018). 12. J. Guélard et al., Geochem. Geophys. Geosyst. 18, 1841–1865 (2017). 13. A. Prinzhofer, C. S. Tahara Cissé, A. B. Diallo, Int. J. Hydrogen Energy 43, 19315–19326 (2018). 14. N. Lefeuvre et al., Geochem. Geophys. Geosyst. 22, e2021GC009917 (2021). 15. C. Neal, G. Stanger, Earth Planet. Sci. Lett. 66, 315–320 (1983). 16. T. A. Abrajano et al., Chem. Geol. 71, 211–222 (1988). 17. G. Etiope et al., Appl. Geochem. 84, 286–296 (2017). 18. E. T. Ellison et al., J. Geophys. Res. Solid Earth 126, e2021JB021981 (2021). 19. C. Monnin et al., J. Geophys. Res. Biogeosci. 126, e2021JG006243 (2021). 20. J. A. Leong et al., Geochim. Cosmochim. Acta 347, 1–15 (2023). 21. G. Etiope, Int. J. Hydrogen Energy 48, 9172–9184 (2023). 22. A. H. F. Robertson, Lithos 65, 1–67 (2002). 23. Y. Dilek, H. Furnes, M. Shallo, Lithos 100, 174–209 (2008). 24. B. Muceku et al., Terra Nova. 20, 180–187 (2008), 25. Y. Dilek, H. Furnes, Lithos 113, 1–20 (2009). 26. A. Meshi, F. Boudier, A. Nicolas, I. Milushi, Int. Geol. Rev. 52, 117–141 (2009). 27. L. Hoxha, Mitt. Ges. Geol. Bergbaustud. Osterr. 48, 52–58 (2007).

9 February 2024

28. Supplementary Text, Materials and Methods are available as supplementary materials. 29. Nm3/min is a SI unit for volumetric flow rate of gas normalized at a temperature of 0°C and a pressure of 101.325 kPa. 30. N. G. Grozeva, F. Klein, J. S. Seewald, S. P. Sylva, Philos. Trans. A Math. Phys. Eng. Sci. 378, 20180431 (2020). 31. F.-V. Donzé, L. Truche, Zenodo (2023): 10.5281/zenodo.8308027. 32. G. Etiope et al., Sci. Rep. 8, 8728 (2018). 33. F. Klein, N. G. Grozeva, J. S. Seewald, Proc. Natl. Acad. Sci. U.S.A. 116, 17666–17672 (2019). 34. J. Carcaillet, J. L. Mugnier, R. Koçi, F. Jouanne, Quat. Res. 71, 465–476 (2017). 35. S. L. Worman, L. F. Pratson, J. A. Karson, E. M. Klein, Geophys. Res. Lett. 43, 6435–6443 (2016). 36. A. S. Merdith et al., Geochem. Geophys. Geosyst. 21, e2019GC008869 (2020). 37. B. S. Lollar, T. C. Onstott, G. Lacrampe-Couloume, C. J. Ballentine, Nature 516, 379–382 (2014). 38. I. Katayama et al., Tectonophysics 814, 228978 (2021). 39. G. Holland et al., Nature 497, 357–360 (2013). 40. C. Bohdanowicz, AAPG Bull. 18, 750–760 (1934). 41. G. Etiope, M. Schoell, H. Hosgörmez, Earth Planet. Sci. Lett. 310, 96–104 (2011). 42. W. J. Brazelton et al., PeerJ 5, e2945 (2017). 43. E. M. Fones et al., ISME J. 13, 1750–1762 (2019). AC KNOWLED GME NTS

We are grateful to AlbChrome Ltd. mining company and its CEO, C. Acar; the miners, engineers (B. Raman and V.O. Karatay); and the director of the Bulqizë mine (M. Demçe) for their strong support in the field. We thank P. Kelemen and two anonymous reviewers for their insights and manuscript reviews. Funding: This project has received fundings from the French National Research Agency (grant agreement ANR-20-CE01-0020) and from the CNRS (MITI 2022.2). L.T. acknowledges support from the Institut Universitaire de France. Author contributions: Funding acquisition: A.C. and L.T. Project administration: A.C., B.M., E.G., and L.T. Fieldwork: All authors. Laboratory analysis: L.T., A.C., and C.L. Visualization: L.T., H.D., and F.V.D. Writing – original draft: L.T. and F.V.D. Writing – review and editing: All authors. Competing interests: Authors declare that they have no competing interests. Data and materials availability: All data are available in the main text or the supplementary materials. The MATLAB script used in this study is available at Zenodo (31). 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.sciencemag.org/about/science-licenses-journalarticle-reuse SUPPLEMENTARY MATERIALS

science.org/doi/10.1126/science.adk9099 Materials and Methods Supplementary Text Figs. S1 to S10 Tables S1 to S3 References (44–50) Movie S1 Submitted 18 September 2023; accepted 10 January 2024 10.1126/science.adk9099

4 of 4

RES EARCH

BIOSYNTHESIS

Characterization and heterologous reconstitution of Taxus biosynthetic enzymes leading to baccatin III Bin Jiang1†, Lei Gao2†, Haijun Wang2†, Yaping Sun1,3†, Xiaolin Zhang1, Han Ke2, Shengchao Liu1, Pengchen Ma4, Qinggang Liao1, Yue Wang1, Huan Wang1, Yugeng Liu1, Ran Du1, Torben Rogge4‡, Wei Li1, Yi Shang5, K. N. Houk4, Xingyao Xiong1, Daoxin Xie6, Sanwen Huang1, Xiaoguang Lei2,7*, Jianbin Yan1* Paclitaxel is a well known anticancer compound. Its biosynthesis involves the formation of a highly functionalized diterpenoid core skeleton (baccatin III) and the subsequent assembly of a phenylisoserinoyl side chain. Despite intensive investigation for half a century, the complete biosynthetic pathway of baccatin III remains unknown. In this work, we identified a bifunctional cytochrome P450 enzyme [taxane oxetanase 1 (TOT1)] in Taxus mairei that catalyzes an oxidative rearrangement in paclitaxel oxetane formation, which represents a previously unknown enzyme mechanism for oxetane ring formation. We created a screening strategy based on the taxusin biosynthesis pathway and uncovered the enzyme responsible for the taxane oxidation of the C9 position (T9aH1). Finally, we artificially reconstituted a biosynthetic pathway for the production of baccatin III in tobacco.

P

aclitaxel, which is derived from the secondary metabolism of Taxus genus plants in the Taxaceae family, has been clinically used to treat various cancers (1–3). Paclitaxel contains a structurally complex 6-8-6 tricyclic carbon skeleton that bears nine stereocenters, one notable oxetane ring motif, and one phenylisoserine chain (3, 4). Paclitaxel can pass through nanopores in the microtubule wall and interact with tubulin on the lumen surface of microtubules to disrupt microtubule dynamics, which triggers cytotoxic effects against cancer cells (5, 6). The overall conformational rigidification of the tricyclic carbon skeleton is generated by the oxetane ring of paclitaxel (7). One strategy presently used to produce paclitaxel is chemical semisynthesis, in which paclitaxel precursors, such as baccatin III, are isolated from plant cell cultures or raw tree biomass of Taxus and subsequently chemically 1

Shenzhen Branch, Guangdong Laboratory for Lingnan Modern Agriculture, Key Laboratory of Synthetic Biology, Ministry of Agriculture and Rural Affairs, Agricultural Genomics Institute at Shenzhen, Chinese Academy of Agricultural Sciences, Shenzhen, China. 2Beijing National Laboratory for Molecular Sciences, Peking-Tsinghua Center for Life Sciences, Key Laboratory of Bioorganic Chemistry and Molecular Engineering of Ministry of Education, College of Chemistry and Molecular Engineering, Peking University, Beijing, China. 3State Key Laboratory for Conservation and Utilization of Subtropical Agro-Bioresources, College of Life Sciences, South China Agricultural University, Guangzhou, China. 4Department of Chemistry and Biochemistry, University of California, Los Angeles, Los Angeles, CA, USA. 5Yunnan Key Laboratory of Potato Biology, The CAAS-YNNU-YINMORE Joint Academy of Potato Sciences, Yunnan Normal University, Kunming, China. 6 Peking-Tsinghua Center for Life Sciences, School of Life Sciences, Tsinghua University, Beijing, China. 7Institute for Cancer Research, Shenzhen Bay Laboratory, Shenzhen, China. *Corresponding author. Email: [email protected] (J.Y.); [email protected] (X.L.) †These authors contributed equally to this work. ‡Present address: Institute of Chemistry, Technical University of Berlin, Straße des 17. Juni 115, 10623 Berlin, Germany.

Jiang et al., Science 383, 622–629 (2024)

converted into paclitaxel (8). The semisynthesis strategy relies heavily on natural resources and artificial cultivation of Taxus plants, which cannot meet the growing market demand (9). With the development of synthetic biology, biosynthetic strategies to produce paclitaxel in a green and sustainable manner have become extremely attractive (10, 11). Since the structure of paclitaxel was first identified in 1971 (3), elucidating its biosynthetic pathway has always been a challenging task in natural product research. To date, approximately 20 possible enzymatic reactions have been reported (12). The reaction steps can be divided into three critical processes, including the formation of a taxane skeleton with a 4(20)-ene-5a-ol moiety from the substrate geranylgeranyl pyrophosphate (GGPP), the biosynthesis of baccatin III through the epoxidation of the C4 and C20 double bonds, and the attachment of a phenylisoserine chain to the C13 position of baccatin III to generate paclitaxel (fig. S1). C13 side chain formation and its combination with baccatin III to produce paclitaxel have been well studied (13). However, several essential steps in forming baccatin III remain unknown, especially oxetane ring formation and C9 oxygenation (fig. S1) (14, 15). As a result, the upstream and downstream steps of paclitaxel biosynthesis cannot be linked, and a complete paclitaxel biosynthetic pathway has so far been lacking (16–18). Results Identification of TOT1 for oxetane ring formation

The oxetane ring of paclitaxel is a notable chemical motif and contributes to the molecule’s anticancer mechanism (19–21). The generally accepted proposal for oxetane ring formation in paclitaxel involves the following chemical processes: first, the 4(20)-ene-5a-yl acetate moiety (called the double-bond moiety) is

9 February 2024

epoxidated to yield a 4(20)b-epoxy-5a-acetoxy intermediate (called the epoxide ring), and then ring expansion from the epoxide ring to the oxetane ring occurs via a rearrangement reaction (22) (fig. S2C). Our previous study demonstrated that the CYP725A subfamily, which contains all characterized oxygenases involved in the skeleton tailoring of paclitaxel, occurs exclusively in the Taxus genus (14, 23–28). Given that the oxetane ring is only known to be produced by Taxus plants (29, 30), the oxygenase responsible for forming the oxetane ring is likely derived from this subfamily. Therefore, we isolated the reaction substrate from Taxus mairei, taxa-4(20),11-diene-2a,5a,7b,9a,10b,13a-hexaolhexa-acetate (taxadiene hexa-acetate, 1) (31) and developed a two-step screening strategy to determine if any member of the CYP725A subfamily could catalyze the oxidation of the double-bond moiety. We first divided the genes in the CYP725A subfamily into three groups (groups I, II, and III) based on phylogenetic analysis (Fig. 1A) and simultaneously expressed all genes in each group in tobacco leaves by Agrobacteriummediated transformation. When the expression of exogenous genes reached a high level on the fourth day after Agrobacterium infiltration, we injected substrate (1) into the corresponding leaf region where the infiltration was performed. After a reaction for 1 day, the leaf metabolites were extracted with methanol and analyzed by liquid chromatography mass spectrometry (LC-MS) (Fig. 1B). We found that the substrate peak area in the group II sample was considerably decreased, producing a prominent peak for the oxidation product (2) with the same retention time and MS spectra as 1-dehydroxybaccatin IV, which contains the oxetane ring (fig. S3A); these results indicate that the desired enzymes were present in group II (Fig. 1C). By contrast, only substrate peak (1) could be detected in the samples of groups I and III, without any other peak of substrate oxidation (Fig. 1C). To determine which gene in group II oxidized the substrate, we next expressed all genes of this group in tobacco leaves individually and performed substrate-feeding experiments separately. The reaction catalyzed by Chr9_74725878 exhibited a strongly decreased substrate peak (1) and increased oxetane product peak (2), indicating that the enzyme encoded by Chr9_74725878 could catalyze oxetane ring formation (Fig. 1D). In addition to the oxetane product peak, the enzyme catalyzes the formation of another oxidative product (3) (Fig. 1, C and D). Its retention time and MS spectra were consistent with those of baccatin I (fig. S3B), which contains a C4,C20-epoxide ring. To rule out the possibility that other cytochrome P450 enzymes in tobacco may be involved in the reaction, we 1 of 8

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

Fig. 1. TOT1 is responsible for oxetane formation in Taxus plants. (A) Phylogenetic analysis of the CYP725A subfamily genes (protein sequence identity 0.05, respectively. The thickness of the lines indicates the r value of the correlation coefficient. (F) The genomic location of TOT and T9aH with respect to other genes. The known cytochrome P450s, acyltransferases, and TXS involved in the paclitaxel biosynthesis and functionally uncharacterized CYP725As are indicated with blue and red lines, respectively. (G) Subcellular location of TOT, T9aH, and T5aH. T5aH, an ER-localized cytochrome P450, was used as a positive control. Colocalization analysis between different channels was performed by analyzing the gray value distributions of the white line in the region of interest. Scale bars are 5 mm. (H) Illustration of the biosynthetic process of baccatin III in plant cells.

The spatial organization of enzymes was assumed to benefit metabolite synthesis and regulation (45, 46). To investigate the subcellular location of the six cytochrome P450

9 February 2024

enzymes, we fused the green fluorescent protein (GFP) gene with the genes individually at their C termini and performed a thorough analysis. As shown in Fig. 5G, similar to T5aH 6 of 8

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

(18), TOT and T9aH displayed a comparable distribution pattern to that of the endoplasmic reticulum (ER) marker, demonstrating that TOT and T9aH were located at the ER. Furthermore, we found that other core components responsible for baccatin III synthesis, including T13aH, T7bH, and T2aH, were also localized to the ER (fig. S15A). Combined with the chloroplast localization of taxadiene synthase (fig. S15B) (47) and cytoplasmic localization of acyltransferases (12, 48), the ER localization of the cytochrome P450 enzymes indicated that highly synergistic regulation occurred between the chloroplast, ER, and cytoplasm (Fig. 5H). First, isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP) are generated through the 2-C-methyl-D-erythritol 4-phosphate (MEP) pathway (18) and then condensed by geranylgeranyl diphosphate synthases (GGPPS) to form GGPP (49), the starting substrate for taxane diterpenoids. Next, GGPP is cyclized and transformed into taxadiene by terpene synthase (TXS) in chloroplasts (40, 47). Subsequently, taxadiene is transferred to the cytoplasm probably by contact sites of plastidER (50); synergistically modified by six ERassociated P450s, including T2aH, T5aH, T7bH, T9aH, T13aH, and TOT, as well as two cytoplasmic acyltransferases (TAT and TBT); and eventually converted into baccatin III (Fig. 5H). Discussion

Natural oxetane-containing compounds (OCCs) are structurally characterized by enormous diversity and complexity, thereby conferring particular advantages and opportunities for modern drug discovery and development. More than 600 different OCCs have been found in animals, plants, and microorganisms, although most are produced by plants (30). The following reaction mechanisms have been discovered for the biosynthesis of natural oxetane skeletons: [2+2] cycloaddition (51), ring contraction (52, 53), and ring expansion (54) (fig. S2). [2+2] Cycloaddition and ring contraction mechanisms have been found in animals and microorganisms. By contrast, natural OCCs, which may originate from ring-expansion reactions, are only found in plants. In this study, we identified the cytochrome P450 enzyme TOT1 in Taxus mairei that catalyzes an unusual oxidative rearrangement reaction to produce the oxetane ring (Figs. 1 to 3); this reaction represents a distinctive chemical logic for oxetane ring formation and reshapes our knowledge on the origin of oxetane in nature. The catalytic flexibility or promiscuity of cytochrome P450s contributes to metabolic bifurcation, thereby fostering structural diversity among diterpenoids (55, 56). Our findings show that only nine enzymes are needed to synthesize baccatin III from GGPP, which is far less than the at least 13 or 14 enzymes that were thought to be required (11, 22, 54, 57, 58). Jiang et al., Science 383, 622–629 (2024)

The discovery also highlights that catalytic promiscuity is somewhat common in the cytochrome P450s and acyltransferases of baccatin III biosynthesis. Previous studies have demonstrated that T5aH can catalyze the formation of many unknown oxygenated products other than taxadien-5a-ol (59–61), and recently it was revealed that this enzyme even has weak cyclase activity leading to the oxetane structure (62), which indicates its functional promiscuity. T13aH could catalyze C13 hydroxylation with taxadien-5a-ol and taxadien-5a-yl acetate as substrates, thereby exhibiting a substrate flexibility similar to that of cytochrome P450s in paclitaxel biosynthesis (24). We found that T13aH could oxidize taxadiene to produce many hydroxylated products, one of which was confirmed to be taxadiene5a,10b,13a-triol (fig. S16), which suggests that the function promiscuity of T13aH is far beyond present understanding. Therefore, it is possible that the C1 and C10 hydroxylation activity could be fulfilled by one or more enzymes among T5aH, T13aH, T9aH, T2aH, TOT, and T7bH, presenting promising avenues for further investigation. Owing to the functional promiscuity of these cytochrome P450s, the metabolic pathway constituted by the nine core genes is believed to be a complex network rather than a linear one, which makes it quite challenging to fully decipher the role of each gene in baccatin III biosynthesis. Meanwhile, owing to the possible bifurcation pathway, the overall yield of baccatin III in tobacco is low (~50 ng per g dry weight). Future research will focus on elucidating the specific catalytic orders and substrate promiscuities of these nine enzymes and find the rate-determining step for metabolic engineering optimization. With more effort from a broad field of scientists, including natural product chemists, plant physiologists, and synthetic biologists, green and efficient manufacturing of paclitaxel may be achieved through synthetic biology in the near future. RE FERENCES AND NOTES

1. A. Ahmed Khalil et al., Curr. Pharm. Des. 28, 3363–3373 (2022). 2. S. M. Swain, S. F. Honig, M. C. Tefft, L. Walton, Invest. New Drugs 13, 217–222 (1995). 3. M. C. Wani, H. L. Taylor, M. E. Wall, P. Coggon, A. T. McPhail, J. Am. Chem. Soc. 93, 2325–2327 (1971). 4. E. Baloglu, D. G. Kingston, J. Nat. Prod. 62, 1448–1472 (1999). 5. F. Naaz, M. R. Haider, S. Shafi, M. S. Yar, Eur. J. Med. Chem. 171, 310–331 (2019). 6. H. Freedman, J. T. Huzil, T. Luchko, R. F. Ludueña, J. A. Tuszynski, J. Chem. Inf. Model. 49, 424–436 (2009). 7. Y. Fu et al., Curr. Med. Chem. 16, 3966–3985 (2009). 8. J. C. Borah, J. Boruwa, N. C. Barua, Curr. Org. Synth. 4, 175–199 (2007). 9. A. Sharma et al., Biotechnol. Bioprocess Eng. 27, 706–728 (2022). 10. H. Wang et al., Acta Pharm. Sin. B 11, 3322–3334 (2021). 11. M. Sabzehzari, M. Zeinali, M. R. Naghavi, Biotechnol. Adv. 43, 107569 (2020).

9 February 2024

12. T. Wang et al., Molecules 26, 2855 (2021). 13. R. Sanchez-Muñoz et al., Front. Bioeng. Biotechnol. 8, 410 (2020). 14. X. Xiong et al., Nat. Plants 7, 1026–1036 (2021). 15. R. Croteau, R. E. Ketchum, R. M. Long, R. Kaspera, M. R. Wildung, Phytochem. Rev. 5, 75–97 (2006). 16. Y. Tong, Y. Luo, W. Gao, Phytochem. Rev. 21, 863–877 (2022). 17. I. Mutanda, J. Li, F. Xu, Y. Wang, Front. Bioeng. Biotechnol. 9, 632269 (2021). 18. J. Li et al., Nat. Commun. 10, 4850 (2019). 19. F. Guéritte-Voegelein, D. Guénard, P. Potier, J. Nat. Prod. 50, 9–18 (1987). 20. C. S. Swindell, S. F. Britcher, J. Org. Chem. 51, 793–797 (1986). 21. D. Willenbring, D. J. Tantillo, Russ. J. Gen. Chem. 78, 723–731 (2008). 22. S. Jennewein, M. R. Wildung, M. Chau, K. Walker, R. Croteau, Proc. Natl. Acad. Sci. U.S.A. 101, 9149–9154 (2004). 23. J. Hefner et al., Chem. Biol. 3, 479–489 (1996). 24. S. Jennewein, C. D. Rithner, R. M. Williams, R. B. Croteau, Proc. Natl. Acad. Sci. U.S.A. 98, 13595–13600 (2001). 25. A. Schoendorf, C. D. Rithner, R. M. Williams, R. B. Croteau, Proc. Natl. Acad. Sci. U.S.A. 98, 1501–1506 (2001). 26. S. Jennewein, C. D. Rithner, R. M. Williams, R. Croteau, Arch. Biochem. Biophys. 413, 262–270 (2003). 27. M. Chau, R. Croteau, Arch. Biochem. Biophys. 427, 48–57 (2004). 28. M. Chau, S. Jennewein, K. Walker, R. Croteau, Chem. Biol. 11, 663–672 (2004). 29. S. B. Horwitz, Cell 177, 502–505 (2019). 30. V. Vil et al., Appl. Microbiol. Biotechnol. 103, 2449–2467 (2019). 31. D. D. C. de Marcano, T. Halsall, J. Chem. Soc. D 1969, 1282–1283 (1969). 32. D. D. C. de Marcano, T. Halsall, J. Chem. Soc. D 1970, 1381–1382 (1970). 33. M. M. Francl, G. Hansell, B. P. Patel, C. S. Swindell, J. Am. Chem. Soc. 112, 3535–3539 (1990). 34. J. L. Giner, J. A. Faraldos, Helv. Chim. Acta 86, 3613–3622 (2003). 35. R. Kaspera, J. L. Cape, J. A. Faraldos, R. E. Ketchum, R. B. Croteau, Tetrahedron Lett. 51, 2017–2019 (2010). 36. S. P. de Visser, F. Ogliaro, N. Harris, S. Shaik, J. Am. Chem. Soc. 123, 3037–3047 (2001). 37. J. Soler, S. Gergel, C. Klaus, S. C. Hammer, M. Garcia-Borràs, J. Am. Chem. Soc. 144, 15954–15968 (2022). 38. H. G. Floss, U. Mocek, in TAXOL: Science and Applications, M. Suffness, Ed. (CRC Press, 1995), pp. 191–208. 39. M. Miyazaki, K. Shimizu, H. Mishima, M. Kurabayashi, Chem. Pharm. Bull. 16, 546–548 (1968). 40. M. R. Wildung, R. Croteau, J. Biol. Chem. 271, 9201–9204 (1996). 41. D. Hampel, C. J. D. Mau, R. B. Croteau, Arch. Biochem. Biophys. 487, 91–97 (2009). 42. M. Chau, K. Walker, R. Long, R. Croteau, Arch. Biochem. Biophys. 430, 237–246 (2004). 43. K. Walker, R. Croteau, Proc. Natl. Acad. Sci. U.S.A. 97, 583–587 (2000). 44. Y. Yukimune, H. Tabata, Y. Higashi, Y. Hara, Nat. Biotechnol. 14, 1129–1132 (1996). 45. L. J. Sweetlove, A. R. Fernie, Annu. Rev. Plant Biol. 64, 723–746 (2013). 46. C. H. Abrahamson, B. J. Palmero, N. W. Kennedy, D. Tullman-Ercek, Annu. Rev. Biophys. 52, 553–572 (2023). 47. D. C. Williams et al., Arch. Biochem. Biophys. 379, 137–146 (2000). 48. K. Walker, A. Schoendorf, R. Croteau, Arch. Biochem. Biophys. 374, 371–380 (2000). 49. J. Hefner, R. E. B. Ketchum, R. Croteau, Arch. Biochem. Biophys. 360, 62–74 (1998). 50. P. Mehrshahi, C. Johnny, D. DellaPenna, Trends Plant Sci. 19, 501–507 (2014). 51. Q. Cheng, D. C. Lamb, S. L. Kelly, L. Lei, F. P. Guengerich, J. Am. Chem. Soc. 132, 15173–15175 (2010). 52. J. Bridwell-Rabb, A. Zhong, H. G. Sun, C. L. Drennan, H.-W. Liu, Nature 544, 322–326 (2017). 53. M. Hecker, V. Ullrich, J. Biol. Chem. 264, 141–150 (1989).

7 of 8

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

54. M. Hezari, R. Croteau, Planta Med. 63, 291–295 (1997). 55. U. Bathe, A. Tissier, Phytochemistry 161, 149–162 (2019). 56. W. Gao et al., Org. Lett. 11, 5170–5173 (2009). 57. R. W. Alexander, Trends Biochem. Sci. 26, 152 (2001). 58. Y. Zhang et al., Mol. Plant 16, 1951–1961 (2023). 59. V. G. Yadav, J. Mol. Catal. B Enzym. 110, 154–164 (2014). 60. Q. Wu, Z. Huang, J. Wang, H. Yu, J. Xu, Bioresour. Bioprocess. 9, 82 (2022). 61. J. C.-T. Liu, R. De La Pena, C. Tocol, E. S. Sattely, BioRxiv 2023.09.27.559859 [Preprint] (2023); https://doi.org/ 10.1101/2023.09.27.559859. 62. Y. Zhao et al., J. Am. Chem. Soc. 146, 801–810 (2024). 63. Q. Liao, Contigs of taxus genome. Figshare (2021); https://doi. org/10.6084/m9.figshare.15000672.v1. ACKN OW LEDG MEN TS We thank the staff (P. Lu, T. He, B. Niu, Y. Peng, M. Su, F. Wang, and X. Liao) at the Agricultural Genomics Institute at Shenzhen and J. Zhou at Peking University for help with activity screening and MS analysis. We thank W. Wang (Engineer Center of Pharmaceutical Technology, Tsinghua University) for assistance with target compound detection in cell lines. We acknowledge computational calculation support from the High-performance

Jiang et al., Science 383, 622–629 (2024)

Computing Platform of Peking University. Funding: This work was funded by the National Key R&D Program of China (grants 2023YFA0915800, 2018YFA0903200, 2020YFA0907900, 2021YFC2102900, 2022YFC3401500, 2022YFC2502500, and 2023ZD04076); the National Natural Science Foundation of China (grants 22193073, 22322701, 21661140001, 92253305, and 22101009); the Science Technology and Innovation Commission of Shenzhen Municipality of China (grant ZDSYS20200811142605017); the Innovation Program of Chinese Academy of Agricultural Sciences (CAAS) and the Elite Young Scientists Program of CAAS; the National Institutes of Health National Institute of Allergy and Infectious Diseases (grant R01 AI 141481 to K.N.H.); and the Beijing National Laboratory for Molecular Sciences (grant BNLMS-CXX-202106). Author contributions: J.Y. and X.L. conceived the study. B.J. and Ya.S. cloned the genes and heterogeneously expressed them in tobacco with help from Y.W. and W.L. S.H., Yi.S., and X.X. helped connect Taxus materials. L.G. performed the purification and synthesis of the substrates or standards from Taxus extracts. B.J. and Ya.S. conducted the gene function confirmation. B.J. and X.Z. performed subcellular localization and colocalization analyses. X.Z., S.L., Q.L., and Hu.W. conducted quantitative real-time polymerase chain reaction (PCR) and the transcriptional and bioinformatic analyses, respectively. Ha.W., L.G., and R.D. analyzed the samples using LC-MS. H.K., P.M., Y.L., T.R., and K.N.H. performed the DFT calculations. J.Y., X.L., D.X., S.H., B.J., L.G., Ha.W., Ya.S., and X.Z. wrote the paper with input from all the authors. Competing interests: J.Y., S.H., B.J., and Ya.S. are inventors on the patent applications related to this work (CN202310496624.6,

9 February 2024

CN202310961179.6, and CN202311137424.8) that were submitted by the Agricultural Genomics Institute at Shenzhen. The other authors declare no competing interests. Data and materials availability: The cds and protein sequence files of CYP725As, including TOT (Chr9_74725878, ctg447_gene.1; CYP725A37, named by D. Nelson) and T9 (Chr9_26460669, ctg2120_gene.5; CYP725A55, named by D. Nelson), are available at Figshare (63). Cartesian coordinates of computed structures are provided in data S1. All other data are presented in the text or supplementary materials. Requests for materials should be addressed to corresponding author J.Y. 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.adj3484 Materials and Methods Figs. S1 to S16 Tables S1 and S2 Data S1 References (64–85) MDAR Reproducibility Checklist Submitted 22 June 2023; accepted 10 January 2024 Published online 25 January 2024 10.1126/science.adj3484

8 of 8

RES EARCH

QUANTUM GASES

Thermography of the superfluid transition in a strongly interacting Fermi gas Zhenjie Yan1†, Parth B. Patel1, Biswaroop Mukherjee1, Chris J. Vale2, Richard J. Fletcher1, Martin W. Zwierlein1* Heat transport can serve as a fingerprint identifying different states of matter. In a normal liquid, a hotspot diffuses, whereas in a superfluid, heat propagates as a wave called “second sound.” Direct imaging of heat transport is challenging, and one usually resorts to detecting secondary effects. In this study, we establish thermography of a strongly interacting atomic Fermi gas, whose radio-frequency spectrum provides spatially resolved thermometry with subnanokelvin resolution. The superfluid phase transition was directly observed as the sudden change from thermal diffusion to second-sound propagation and is accompanied by a peak in the second-sound diffusivity. This method yields the full heat and density response of the strongly interacting Fermi gas and therefore all defining properties of Landau’s two-fluid hydrodynamics.

H

eat transport is a ubiquitous phenomenon at work in everything from steam engines to the formation of stars, and it dictates how energy, information, and entropy flow in the system. In conventional materials, heat, mass, and charge are all transported by the motion of (quasi)particles, such as electrons in metals. This common origin of transport results, for example, in the Wiedemann-Franz law, relating thermal and electrical conductivity. However, in strongly correlated systems, such as high-temperature superconductors (1), neutron stars (2), and the quark-gluon plasma of the early universe (3), the notion of a quasiparticle is poorly defined. It is unknown whether there is a common relaxation rate for heat, density, and spin transport (4) or if strong correlations separate these phenomena. Understanding the flow of entropy is at the forefront of current research, with powerful theoretical models connecting thermal flow in quantum systems to gravitational duals (3, 5). Directly measuring thermal transport, as distinct from mass or charge transport, is thus of great relevance for elucidating the origin of heat dissipation in strongly correlated matter. Strongly interacting atomic Fermi gases near a Feshbach resonance provide an ideal platform for quantitative studies of fermion transport (6–10). As a result of scale invariance in resonant Fermi gases (11), measurements performed in one system constrain the equation of state and transport properties of other strongly interacting Fermi systems, including neutron matter at densities 25 orders of mag-

1

MIT-Harvard Center for Ultracold Atoms, Research Laboratory of Electronics, and Department of Physics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA. 2Optical Science Centre and ARC Centre of Excellence in Future Low-Energy Electronics Technologies, Swinburne University of Technology, Melbourne 3122, Australia.

*Corresponding author. Email: [email protected]. †Present address: Department of Physics, University of California, Berkeley, CA 94720.

Yan et al., Science 383, 629–633 (2024)

nitude higher. The system features the largest superfluid transition temperature Tc, relative to its density, of all known fermionic systems (12). In this study, we introduce a thermography method to image heat in interacting quantum gases. The method requires only a temperaturedependent spectral response that can be locally resolved. In the case of the Fermi gas that we studied here, the radio-frequency (rf) spectrum is temperature dependent (13, 14). We spatially resolved this spectral response and directly measured heat transport in the strongly interacting Fermi gas. The nature of heat transport can help distinguish states of matter. In ordinary liquids, heat transport is purely diffusive and governed by thermal conductivity. By contrast, in superfluids, heat propagates as a wave called “second sound.” The two-fluid model of superfluidity introduces normal and superfluid components that can move in and out of phase (15, 16). This gives rise to two distinct sound modes, first and second sound, corresponding to a density and an entropy wave (17). The speed of second sound c2 is a direct measure of the superfluid fraction rS =rN, the ratio of the superfluid density rS to the normal component density rN (18). Its attenuation yields the second-sound diffusivity D2, which involves the thermal conductivity, bulk, and shear viscosities (17, 19). Consequently, we observe a dramatic change in thermal transport as the Fermi gas is cooled belowTc. Simultaneously recording the complete density and heat response of the system to a known external perturbation allows us to completely characterize the two-fluid hydrodynamics of the strongly interacting Fermi gas (19, 20). Previous studies of thermal transport in quantum gases relied on the weak coupling between the density and temperature of the gas (21–24). This allowed the observation of second sound in Bose (25, 26) and Fermi gases (21, 24, 27) but without directly measuring heat propagation. By using a homogeneous box potential formed by

9 February 2024

light sheets, we observed running and standing waves of second sound, demonstrating multiple reflections of entropy waves from the walls of the box. Our thermography works across the superfluid transition, allowing the observation of a pronounced peak in thermal diffusion atTc, characteristic of critical behavior expected near second-order phase transitions. Spectral thermometry

The working principle of our method is sketched in Fig. 1, A to D. In rf spectroscopy, interacting atoms are ejected from the many-body system into an initially unoccupied internal spin state (28). For interacting gases, the resulting spectra depend on temperature. At high temperatures, when the thermal de Broglie wavelength is shorter than both scattering length and interparticle distance, the spectra approach the bare, unshifted response for an isolated atom. Conversely, at low temperatures, the spectra display interaction-induced shifts known as “clock shifts.” In the particular case of attractive twocomponent Fermi gases, at zero temperature the spectral peak is shifted by approximately the pairing energy EB of fermion pairs (13), and at nonzero temperature, broken pairs contribute to the response at lower frequencies (Fig. 1A). For a fixed detuning w0 on the flank of the spectrum, the rf response is sensitive to changes in temperature (Fig. 1B). As the rf response can be spatially resolved, this allows for a direct measurement of the local temperature from a single image of rf-transferred atoms. As an exemplary application of this method, we may detect second sound in the fermionic superfluid, which is a wave in the gas of excitations that, close to Tc , consists predominantly of broken pairs (Fig. 1C). A suitably detuned rf drive can transfer atoms from the gas of excitations, yielding a direct, local measure of heat (Fig. 1D). We stress that the method does not depend on this simplified picture of broken pairs and only relies on the temperature dependence of the rf spectrum. It therefore applies in a wide range of temperatures set by the magnitude of clock shifts, which for the unitary Fermi gas are on the scale of the Fermi temperature (13). Our experiment began with a uniform fermionic superfluid trapped in a cylindrical box potential whose axial direction is defined as the z axis, formed by an equal mixture of resonantly interacting fermions in the first (1) and third (3) hyperfine state of 6Li at a Feshbach resonance (magnetic field, 690 G) (29). The density of n0 ¼ 0:75 mm3 per spin state corresponds to a Fermi energy of EF ¼ h  10:5 kHz and a Fermi temperature of TF ¼ EF =kB ≃ 500 nK, where h is Planck’s constant and kB is the Boltzmann constant. To create temperature gradients in the superfluid gas, we resonantly excited a standing wave of second 1 of 5

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

A

B

C

recording the dependence of nf on temper@nf f ature, @n @T jn, and density, @n jT (18). This method of calibrating spectral responses versus each thermodynamic variable while holding other parameters constant can be applied universally. More generally, all that is required for the observation of thermal transport is access to any local observable that is sensitive to temperature, meaning that it can be achieved even without a calibrated thermometer. Integrating the two-dimensional (2D) temperature profile along the uniform x axis yields a 1D temperature profile, DT ðz Þ, the deviation in temperature from the equilibrium state, with a precision of 500 pK from a single image, as shown in Fig. 1F. The data reveal an essentially flat density in the presence of a ~8-nK temperature difference across the box.

D NF SF z

E

F

x

Observation of heat propagation

z

Fig. 1. Direct local thermography using rf spectroscopy. (A) A sketch of rf spectra at various temperatures for the unitary Fermi gas (13). Blue, gray, and red lines correspond to the rf response IðwÞ at successively higher temperatures. (B) At fixed frequency w0 on the flank of a spectrum [black dotted line in (A)], the rf response is sensitive to temperature and serves as a local thermometer. (C) In a simplified picture, the superfluid component (SF) consists of fermion pairs, whereas the normal fluid (NF) is composed of broken pairs. (D) The unpaired atoms are transferred to a weakly interacting state by an rf pulse and subsequently imaged to determine the spatial distribution of the normal component density. (E and F) In situ observation of a second-sound wave after resonant gradient excitation. Shown are the column density and local temperature, respectively, from simultaneous in situ absorption images of unperturbed (3) and rftransferred (2) atoms, with density n and temperature variation DT, averaged along the x axis, shown below. The vertical dotted line marks the edge of the box potential (half maximum of potential). The black dashed line in (F) is a fit to the fundamental eigenmode in the box [eq. S1 in (18)]. Second sound has a significant effect on the temperature, but not the density.

sound using an oscillating potential gradient along the z axis (Fig. 2A). Our thermography uses rf transfer of atoms from state 1 into the initially unoccupied state f ≡ 2. Simultaneous in situ absorption images of atoms in states 2 and 3 along one of the radial direction (y axis) yield the original gas density nðx; z Þ (Fig. 1E), as well as the density nf ðx; z Þ of rf-transferred atoms, carrying the information on the local temperature (Fig. 1F). The rf thermometer is calibrated on gases in thermal equilibrium by Yan et al., Science 383, 629–633 (2024)

Armed with the ability to spatially resolve temperature in the strongly interacting Fermi gas, we directly observed second sound as the free back-and-forth sloshing of heat after resonant gradient excitation (Fig. 2, B to D). Figure 2B shows the measured temperature variation DT ðx; z; t Þ obtained at various times after secondsound generation. Figure 2C presents the time evolution of the 1D temperature profiles DT ðz; t Þ, and Fig. 2D shows the corresponding evolution of the amplitude DT ðk1 ; t Þ of the first spatial Fourier mode supported by the axial box length L ¼ 91 mm (kj ¼ jp=L), all clearly demonstrating the wave-like propagation of heat. Here, the absolute temperature of the gas in equilibrium, obtained from expansion (14), was T ¼ 63ð2Þ nK ¼ 0:125ð5ÞTF , or T ¼ 0:75ð3Þ Tc when compared with the superfluid transition temperature Tc ¼ 0:167TF reported in (12). A damped sinusoidal fit to DT ðk1 ; t Þ yielded a speed of second sound of c2 ¼ w=k ¼ 3:57 ð2Þmm=s , corresponding to about a tenth of the Fermi velocity c2 ¼ 0:092ð2ÞvF . From the measured damping rate G, we obtained a diffusivity of second sound D2 ¼ G=k2 ¼ 2:44ð11Þℏ=m. As was found for the diffusivities of spin (30, 31), momentum (32), and first sound (33), a natural scale for the diffusivity of second sound is reduced Planck’s constant ℏ = h/2p divided by the particle mass m (27, 34). This scale directly emerges in a strongly interacting quantum fluid from a mean-free path of carriers of approximately one interparticle spacing d, and characteristic speeds of ℏ=md given by Heisenberg’s uncertainty (30). A similar scale of diffusivity is also measured for second sound in the strongly interacting bosonic superfluid 4He (35), whereas the more weakly interacting fermionic 3He in its superfluid A1 and B phases displays much larger values that are many hundreds to thousands of times ℏ=m (36). Thermography provides an unprecedented view of the superfluid transition in the strongly interacting Fermi gas. Figures 2, F and G, show

9 February 2024

the transition from heat diffusion in the normal state to wave-like propagation of heat, second sound, in the superfluid. For these data, we created a local hotspot on one side of the box by locally applying an intensity-modulated optical grating (Fig. 2E). Modulation at ∼2 kHz efficiently creates high-frequency phonons that rapidly decay into heat (33, 37), creating a temperature profile with good overlap with the j ¼ 1 mode. The subsequent evolution of the temperature amplitude DT ðk1 ; T Þ displays a striking change in character from exponential decay above Tc to the damped sinusoid of second sound below Tc . Entropy and density response functions

The full linear response theory of two-fluid hydrodynamics for superfluids was provided over half a century ago by Hohenberg and Martin (19). Under an external potential that acts on the density n with wave vector k and frequencyw, systems respond through changes in their density n as well as their temperature or equivalent entropy densitys. Thermography enables us to obtain the corresponding response functions, not only cn;n ðk; wÞ but alsocs;n ðk; wÞ. These encode all the thermodynamic and twofluid hydrodynamic information of the unitary Fermi gas (18–20). To determine the linear response functions, we apply a potential gradient, oscillating at frequency w. The steady-state temperature change DT ðk1 ; wÞ and density change Dnðk1 ; wÞ, measured after an integer number of oscillation cycles, yield the respective out-of-phase response functions (19, 20). The change in entropy per particle, Ds, is linked to the temperature and density variation by the equation of state. For our scale invariant, unitary Fermi gas, this connection is provided by the specific heat per particle cV at constant density (11, 12)  Ds ¼ cV

 DT 2 Dn  T 3 n0

ð1Þ

Measurements of fractional temperature and density variations thus directly yield the entropy variation in units of cV . Figures 3, A and B, display the entropy and density response of the superfluid in a frequency range that solely excites the lowest spatial mode (j ¼ 1), the sloshing mode. The density reveals a dominant peak attributed to first sound near 90 Hz (33) and a faint signature of second sound at 20 Hz, expected in a gas of nonzero expansivity, where density and temperature are coupled. However, in the entropy channel, whose signal derives predominantly from the rf transfer (18), the strong second-sound peak indicates a large response. This directly demonstrates that second sound in the unitary Fermi gas is predominantly an entropy wave, whereas first sound is essentially isentropic. This is similar to the case in superfluid 4He 2 of 5

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

A

E

Oscillating Gradient

NF

7 µm

SF

B

Oscillating Optical Grating

x z

z

F

G

C

D

Fig. 2. Direct observation of the superfluid transition from heat propagation in a strongly interacting Fermi gas. (A) Generating second sound with an oscillating potential gradient for data shown in (B) to (D) at a temperature of T ¼ 63 nK or 0:75Tc. (B) In situ thermographs at times t ¼ 0; 26, and 54 ms after second-sound excitation. (C) Time evolution of the axial temperature profiles, revealing the wave-like propagation of heat. (D) Amplitude of the first spatial Fourier mode of the temperature profiles DT(k1, t) versus time (gray circles). A fit to a damped sinusoid (dashed line) gives the speed and attenuation

(17) but drastically different from the case in 2D and 3D Bose gases, in which density and entropy are strongly coupled (25, 26, 38). In Figs. 3, C and D, we show the thermal evolution of the entropy and density responses in the first spatial Fourier mode, which serve as a direct measurement of the out-of-phase entropy-density [Imcs;n ðk1 ; wÞ] and densitydensity [Imcn;n ðk1 ; wÞ] response functions (18). The measured response functions completely encode all information about the two-fluid hydrodynamics in a unitary Fermi gas (18–20). The peak positions and widths give the speeds and diffusivities of first and second sound. The speed of first sound is a direct measure of the energy of the gas (33), and the speed of second sound yields the superfluid density. The height of the second-sound peak in the entropy-density response is given by the expansivity ap of the gas, and the weight of the second-sound versus the first-sound response in the density-density response directly equals Yan et al., Science 383, 629–633 (2024)

rate of second sound. (E) Local heating with an intensity-modulated optical grating for data shown in (F) to (G). (F) Time evolution of temperature amplitudes DT(k1, t) (solid circles) and fits (dashed lines) at various gas temperatures. The dotted lines show the DT ¼ 0 line for each temperature. The fitting method used in (D) and (F) is indicated by eq. S24 in (18). (G) Twodimensional interpolation with Gaussian smoothing of temperature amplitudes versus time across the superfluid transition. In (D) and (E), the initial temperature variation for each time trace is normalized to be 1.

g  1 , where g ¼ cp =cV is the ratio of heat capacities at constant pressure and density. The thermodynamic quantities ap and g are related by the isothermal compressibility kT , the heat capacity, and temperature by g  1 ¼ T a2p =ðnkT cV Þ, and in particular for the unitary gas simply by g  1 ¼ 23 ap T . Heat transport across superfluid transition

Figure 4A shows the speed of second sound, measured consistently with our three independent methods: free evolution after resonant excitation of the second-sound mode (yellow squares), local heating (red diamonds), and steady-state response functions (blue circles). The superfluid fraction is obtained fromc2 and the previously measured equation of state (12, 18) and is shown in Fig. 4B. The measurements show a qualitative agreement with Nozières and Schmitt-Rink theory (39, 40) (dotted-dashed line), although their absolute value of Tc differs from experiment. Our super-

9 February 2024

fluid fraction agrees well with the result reconstructed for the homogeneous case from the second-sound measurement in a quasi-1D trapped gas in (21), which relied on the same equation of state from (12). With the local heating method (red diamonds), we are able to observe the continuous evolution of c2 and rS from a finite value in the superfluid phase to zero in the normal phase. The phasetransition temperature Tc obtained from this measurement is consistent with the equilibrium thermodynamic measurement (12) (the vertical gray area) and the onset of pair condensation (7, 13), which we have measured here as well (Fig. 4C). As is expected, there is a clear quantitative difference between the superfluid fraction, which saturates to unity at temperatures T ≲ 0:1TF , and the pair condensate fraction, which remains ≲0:75 . The superfluid density quantifies the portion of the fluid that flows without friction. Formally, it measures the rigidity against phase twists, 3 of 5

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

A

B

A

C

D

B

C

D

Fig. 3. Steady-state entropy and density response of the unitary Fermi superfluid. Shown are the (A) change in entropy per particle (Ds) and (B) density (Dn) after excitation by an integer number of cycles of an oscillating axial potential gradient (Fig. 2E). For frequencies below 50 Hz, the drive duration is 5 cycles at an amplitude of g ¼ h  2:12 Hz=mm; for frequencies above 50 Hz, we drive for 20 cycles at an oscillation amplitude of g ¼ h  0:85 Hz=mm. The gas temperature is T=Tc ¼ 0:75. Amplitudes of the first spatial Fourier mode are shown in (C) and (D) for various temperatures in the superfluid phase. The solid lines are fits using the full entropy- and density-response function from two-fluid hydrodynamics [eqs. S9 and S10 in (18)].

whereas the condensate fraction is a measure for the number of fermion pairs at zero–center of mass momentum. In the zero-temperature limit, the entire system is superfluid, but only a fraction of fermion pairs are condensed, owing to quantum depletion and Pauli blocking (6, 7, 9). A further dramatic signature of the superfluid transition is seen in the temperature dependence of the second-sound diffusivity D2 in the superfluid state, and thermal diffusion in the normal state, shown in Fig. 4D. We observe a striking peak in this transport coefficient within a range DT ≈ 0:1Tc around the critical temperature of superfluidity, rising above a background minimum value of about 2ℏ=m up to nearly three times this value. This behavior echoes that found in liquid 4He (35, 41) near its superfluid transition, associated with classical criticality. Indeed, the order parameters of both the Fermi superfluid and liquid helium belong to the same 3D XY static universality class, and also the same [model F in (42)] dynamic universality class, dictating a behavior D2 ºjTc  T jn=2 near the transition, with critical exponent n ≈ 0:672, as observed in 4He (41). Related critical behavior for the speed of second sound c2 ºðTc  T Þn=2 and rS ºðTc  T Þn is qualitatively consistent with the steep slopes we observe close to Tc in these quantities. For the unitary Fermi gas, the width of the region governed by criticalYan et al., Science 383, 629–633 (2024)

ity is not precisely known but is estimated to be on the order of Tc (43, 44). A quantitative analysis of critical behavior, such as the measurement of critical exponents, is prevented by the residual inhomogeneity of the gas density, giving a variation of DðT =Tc Þ ∼ 5  103, and by the finite size of our system. Indeed, even for the lowest spatial mode j ¼ 1, second sound becomes overdamped (G ≳ 2w) within 3% of Tc. At low temperatures T =Tc < 0:6, D2 is again seen to rise significantly, which we attribute to the diverging mean-free path of phonons, the only remaining contribution at low temperatures once pair-breaking excitations are frozen out. Above the transition temperature, the secondsound mode evolves into a thermal-diffusion mode whose diffusivity is directly given by thermal conductivity k: D2 ¼ k=ncP (19, 20, 45, 46). We therefore find quantum-limited thermal diffusion ∼2ℏ=m (47), similar to prior results for spin (30, 31), momentum (32), and first-sound diffusion (33) in the unitary gas. However, the nonmonotonous behavior of second-sound diffusivity, with steep rise at low temperatures and around Tc, has not been observed in other transport coefficients. The second-sound diffusivity D2 was independently measured with Bragg scattering (27), and a small rise in the second-sound damping rate approaching Tc was observed. However, a peak in D2 near Tc could not be resolved, pre-

9 February 2024

Fig. 4. The speed and diffusivity of second sound. (A) Speed of second sound, normalized by the Fermi velocity, as a function of temperature, determined by fitting the steady-state response functions (blue circles), and the free evolution of second sound after resonant gradient excitation (yellow squares) or after local heating (red diamonds). The first-sound speed measured from the response functions (gray circles) is also shown. The dotteddashed line indicates Nozières-Schmitt-Rink theory (39). (B) The superfluid fraction of the unitary Fermi gas obtained from the speed of second sound [symbols as shown in (A); also see eq. S13 in (18)]. The blue shaded area indicates the uncertainty from the equation of state. Solid green circles indicate the superfluid fraction obtained from quasi-1D experiments (21) that also utilized the MIT equation of state of the unitary Fermi gas (12). (C) Pair condensate fraction measured with the rapid-ramp technique to detect fermion pair condensates (13). (D) Second-sound diffusivity obtained from various methods [symbols as shown in (A)]. The vertical gray area shown in all panels indicates the uncertainty of critical temperature from (12).

sumably because Bragg scattering as a density probe becomes insensitive to heat propagation above Tc . Away from Tc , the values for D2 reported in (27) were about half of what we observed. Given that the experiment in (27) used a much higher wave vector and correspondingly more elevated frequencies, the gas may no longer have been hydrodynamic but instead entered the collisionless regime, which is similar to the behavior for high-momentum first sound in (33, 37). Assuming the hydrodynamic relation G ¼ D2 k2 for such modes will yield 4 of 5

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

too small a value for D2 . By contrast, in the present work using thermography, we verified hydrodynamic scaling by exciting also the second ( j ¼ 2) spatial mode supported by the box, finding within error bars identical values of D2 [fig. S4 in (18)]. In the superfluid regime of the unitary Fermi gas, there are three contributions to secondsound diffusion: thermal conductivity k, shear viscosity h, and bulk viscosity z3 from normalsuperfluid counterflow (36, 48). Although it is known that z3 ¼ 0 for a pure phonon gas with linear dispersion (49), in the range T =Tc ≳ 0:5, the normal fluid is dominated by pair-breaking excitations. In this case, all three contributions are of similar importance (36, 48). Assuming z3 ¼ 0 in this regime, as was done in (27), is not warranted, and obtaining viscosity and thermal conductivity from first- and secondsound diffusion alone is not possible. Outlook

Direct measurement of heat transport has been a long-standing goal in quantum gas experiments. Thermography now opens the door to study a host of intriguing nonequilibrium phenomena, from nonlinear heat waves to quench dynamics (50, 51) and even far-from-equilibrium phenomena such as prethermal states (52, 53). Using tomographic imaging techniques (54), the complete 3D spectral response can be measured, enabling the investigation of transverse entropy transport in anisotropic or inhomogeneous systems. For thermodynamic systems with additional degrees of freedom beyond density and temperature, for example, spin-imbalanced systems, additional independent measurements such as probes of the local spin polarization can be supplemented to fully determine thermodynamic response functions. The spectral response continues to serve as a channel highly sensitive to temperature. Therefore, our spectroscopic thermometry method may be applicable to other quantum gas platforms, including Bose

Yan et al., Science 383, 629–633 (2024)

gases, Bose-Fermi mixtures, impurity systems, and Hubbard quantum simulators (55). RE FERENCES AND NOTES

1. P. A. Lee, N. Nagaosa, X.-G. Wen, Rev. Mod. Phys. 78, 17–85 (2006). 2. M. G. Alford, L. Bovard, M. Hanauske, L. Rezzolla, K. Schwenzer, Phys. Rev. Lett. 120, 041101 (2018). 3. A. Adams, L. D. Carr, T. Schäfer, P. Steinberg, J. E. Thomas, New J. Phys. 14, 115009 (2012). 4. S. A. Hartnoll, Nat. Phys. 11, 54–61 (2015). 5. M. Blake, R. A. Davison, S. Sachdev, Phys. Rev. D 96, 106008 (2017). 6. S. Giorgini, L. P. Pitaevskii, S. Stringari, Rev. Mod. Phys. 80, 1215–1274 (2008). 7. W. Ketterle, M. W. Zwierlein, Riv. Nuovo Cim. 31, 247 (2008). 8. M. W. Zwierlein, Novel Superfluids (Oxford Univ. Press, 2015), pp. 269–422. 9. W. Zwerger, in Quantum Matter at Ultralow Temperatures, M. Inguscio, W. Ketterle, S. Stringari, G. Roati, Eds. (IOS Press, 2016), pp. 63–142. 10. S. Krinner, T. Esslinger, J.-P. Brantut, J. Phys. Condens. Matter 29, 343003 (2017). 11. T.-L. Ho, Phys. Rev. Lett. 92, 090402 (2004). 12. M. J. H. Ku, A. T. Sommer, L. W. Cheuk, M. W. Zwierlein, Science 335, 563–567 (2012). 13. B. Mukherjee et al., Phys. Rev. Lett. 122, 203402 (2019). 14. Z. Yan et al., Phys. Rev. Lett. 122, 093401 (2019). 15. L. Tisza, Nature 141, 913 (1938). 16. L. Landau, Phys. Rev. 60, 356–358 (1941). 17. I. M. Khalatnikov, An Introduction to the Theory of Superfluidity (CRC Press, 1965). 18. See supplementary materials. 19. P. Hohenberg, P. Martin, Ann. Phys. 34, 291–359 (1965). 20. L. P. Kadanoff, P. C. Martin, Ann. Phys. 24, 419–469 (1963). 21. L. A. Sidorenkov et al., Nature 498, 78–81 (2013). 22. J.-P. Brantut et al., Science 342, 713–715 (2013). 23. D. Husmann et al., Proc. Natl. Acad. Sci. U.S.A. 115, 8563–8568 (2018). 24. D. K. Hoffmann et al., Nat. Commun. 12, 7074 (2021). 25. P. Christodoulou et al., Nature 594, 191–194 (2021). 26. T. A. Hilker et al., Phys. Rev. Lett. 128, 223601 (2022). 27. X. Li et al., Science 375, 528–533 (2022). 28. S. Gupta et al., Science 300, 1723–1726 (2003). 29. B. Mukherjee et al., Phys. Rev. Lett. 118, 123401 (2017). 30. A. Sommer, M. Ku, G. Roati, M. W. Zwierlein, Nature 472, 201–204 (2011). 31. C. Luciuk et al., Phys. Rev. Lett. 118, 130405 (2017). 32. C. Cao et al., Science 331, 58–61 (2011). 33. P. B. Patel et al., Science 370, 1222–1226 (2020). 34. Z. Yan et al., Bull. Am. Phys. Soc. 64, Abstract D07.00006 (2019). 35. W. B. Hanson, J. R. Pellam, Phys. Rev. 95, 321–327 (1954). 36. D. Vollhardt, P. Wölfle, The Superfluid Phases of Helium 3 (Taylor & Francis, 1990). 37. C. C. N. Kuhn et al., Phys. Rev. Lett. 124, 150401 (2020). 38. L. Verney, L. Pitaevskii, S. Stringari, Europhys. Lett. 111, 40005 (2015). 39. E. Taylor et al., Phys. Rev. A 80, 053601 (2009). 40. P. Nozières, S. Schmitt-Rink, J. Low Temp. Phys. 59, 195–211 (1985).

9 February 2024

41. L. S. Goldner, N. Mulders, G. Ahlers, J. Low Temp. Phys. 93, 131–182 (1993). 42. P. C. Hohenberg, B. I. Halperin, Rev. Mod. Phys. 49, 435–479 (1977). 43. E. Taylor, Phys. Rev. A 80, 023612 (2009). 44. T. Debelhoir, N. Dupuis, Phys. Rev. A 93, 023642 (2016). 45. B. Frank, W. Zwerger, T. Enss, Phys. Rev. Res. 2, 023301 (2020). 46. P. Patel, “Quantum transport in strongly interacting, ultracold Fermi gases in box potentials,” thesis, Massachusetts Institute of Technology (2022). 47. X. Wang, X. Li, I. Arakelyan, J. E. Thomas, Phys. Rev. Lett. 128, 090402 (2022). 48. H. Smith, H. H. Jensen, Transport Phenomena (Clarendon Press, 1989). 49. M. A. Escobedo, M. Mannarelli, C. Manuel, Phys. Rev. A 79, 063623 (2009). 50. M. W. Zwierlein, C. H. Schunck, C. A. Stan, S. M. F. Raupach, W. Ketterle, Phys. Rev. Lett. 94, 180401 (2005). 51. P. Dyke et al., Phys. Rev. Lett. 127, 100405 (2021). 52. M. Gring et al., Science 337, 1318–1322 (2012). 53. C. Eigen et al., Nature 563, 221–224 (2018). 54. M. J. H. Ku, B. Mukherjee, T. Yefsah, M. W. Zwierlein, Phys. Rev. Lett. 116, 045304 (2016). 55. C. Gross, I. Bloch, Science 357, 995–1001 (2017). 56. Z. Yan et al., Replication Data for: Thermography of the superfluid transition in a strongly interacting Fermi gas, version 1, Harvard Dataverse (2023); https://doi.org/10.7910/DVN/KKLTXW. AC KNOWLED GME NTS

We thank B. Svistunov, N. Prokof’ev, W. Zwerger, and in particular the late L. Pitaevskii for illuminating discussions. Funding: This work was supported by the NSF (Center for Ultracold Atoms Award Nos. PHY-1734011 and PHY-2012110), the Air Force Office of Scientific Research (FA9550-16-1-0324 and MURI Quantum Phases of Matter FA9550-14-1-0035), the Office of Naval Research (N00014-17-1-2257), and the Vannevar Bush Faculty Fellowship (ONR No. N00014-19-1-2631). Author contributions: Z.Y., P.B.P., B.M., and R.J.F. contributed to building the experimental setup. Z.Y., P.B.P., and R.J.F. performed the measurements, and analyzed the data. All authors contributed to the interpretation of the data and the preparation of the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data shown in this work can be found at Harvard Dataverse (56). 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 SUPPLEMENTARY MATERIALS

science.org/doi/10.1126/science.adg3430 Materials and Methods Supplementary Text Figs. S1 to S10 References (57, 58) Submitted 28 December 2022; accepted 10 January 2024 10.1126/science.adg3430

5 of 5

RES EARCH

METALLURGY

Ultrauniform, strong, and ductile 3D-printed titanium alloy through bifunctional alloy design Jingqi Zhang1†, Michael J. Bermingham1†, Joseph Otte2, Yingang Liu1, Ziyong Hou3,4,5, Nan Yang1, Yu Yin1, Mohamad Bayat6, Weikang Lin1, Xiaoxu Huang3,4, David H. StJohn1, Matthew S. Dargusch1* Coarse columnar grains and heterogeneously distributed phases commonly form in metallic alloys produced by three-dimensional (3D) printing and are often considered undesirable because they can impart nonuniform and inferior mechanical properties. We demonstrate a design strategy to unlock consistent and enhanced properties directly from 3D printing. Using Ti−5Al−5Mo−5V−3Cr as a model alloy, we show that adding molybdenum (Mo) nanoparticles promotes grain refinement during solidification and suppresses the formation of phase heterogeneities during solid-state thermal cycling. The microstructural change because of the bifunctional additive results in uniform mechanical properties and simultaneous enhancement of both strength and ductility. We demonstrate how this alloy can be modified by a single component to address unfavorable microstructures, providing a pathway to achieve desirable mechanical characteristics directly from 3D printing.

T

hree-dimensional (3D) printing or additive manufacturing (AM) of metals and alloys typically involves multiple physical and metallurgical phenomena that impart complex microstructures and varied mechanical properties in the fabricated products (1–7). For metallic alloys, which solidify with a cubic crystal structure, columnar grains often prevail in the 3D-printed part (8, 9) because grains with the easy growth directions tend to align closely to the maximum temperature gradient of the melt pool and grow epitaxially from the partially melted layers (10). Although this highly textured columnar grain structure can be beneficial for certain applications, in most cases it is undesirable because it degrades the mechanical performance and results in mechanical property anisotropy (11, 12). Accordingly, extensive effort has been devoted to transforming the coarse columnar grain structure into fine equiaxed grains to achieve superior and isotropic mechanical properties (13–17). Generally, the columnar-to-equiaxed transition (CET) and grain refinement can be promoted through process control and by adding grainrefining inoculants (12, 18). The former typically includes manipulating the 3D printing processing parameters or introducing external interferences (16, 17). However, the effectiveness and practicalities of this approach are limited to specific alloys or 3D printing technologies. 1

School of Mechanical and Mining Engineering, The University of Queensland, St. Lucia, Brisbane, QLD, Australia. Centre for Microscopy and Microanalysis, The University of Queensland, St. Lucia, Brisbane, QLD, Australia. 3 International Joint Laboratory for Light Alloys (Ministry of Education), College of Materials Science and Engineering, Chongqing University, Chongqing, China. 4Shenyang National Laboratory for Materials Science, Chongqing University, Chongqing, China. 5Department of Materials Science and Engineering, KTH Royal Institute of Technology, Stockholm, Sweden. 6Department of Mechanical Engineering, Technical University of Denmark, Lyngby, Denmark. 2

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

Zhang et al., Science 383, 639–645 (2024)

Alternatively, the metallurgical approach through additives has proven highly effective but often results in an undesirable loss in ductility owing to the formation of brittle second phases (9). Therefore, simultaneously addressing the coarse columnar grains and eliminating property anisotropy without adversely affecting the ductility is highly desirable. A further complication is that many allotropic alloy systems, including titanium alloys, are also susceptible to the heterogeneous distribution of phases associated with the solid-state thermal cycling experienced during the 3D printing process (19–22). This poses an additional challenge to achieve uniform mechanical properties of 3D-printed parts made from these alloys (23, 24). The localized heating, cooling, and reheating nature of 3D printing effectively creates dynamic in situ heat treatments that encourage the decomposition of initially formed phases and/or the precipitation of new phases through solid-state phase transformations (20, 25). Because thermal cycles are spatially variable, the associated heat treatments can produce an inhomogeneous distribution of phase along the building direction of the part, thereby resulting in the spatial variation of mechanical properties (21, 22). Postprinting heat treatments can be effective in mitigating these phase heterogeneities but introduce delays and additional costs and are not effective in refining textured columnar grains (26). The confluence of these issues has made it extremely challenging to achieve uniform and superior mechanical properties in the as-fabricated state. We demonstrate a design strategy to address this challenge by simultaneously controlling the grain structure and constituent phases in products manufactured by laser powder bed fusion (L-PBF). We selected the Ti−5Al−5Mo− 5V−3Cr (Ti-5553) metastable b titanium alloy as a model alloy because it shows the coex-

9 February 2024

istence of coarse columnar b grains and a heterogeneous distribution of phases (Fig. 1, A to C). This results in highly nonuniform, position-dependent tensile properties from L-PBF, as we demonstrate (Fig. 1, D and E) and as has been demonstrated in other studies across multiple 3D printing technologies (27–29). We show that the single addition (up to 5.0 wt %) of elements from the b-isomorphous group [in this case, we selected Mo; see section on alloy design in the materials and methods (30)] into Ti-5553 powder to form a composite blend achieves bifunctionality: (i) During 3D printing, some of the Mo particles partially melt, but the core survives to nucleate fine grains during solidification and prevent coarse columnar grains from forming. (ii) The dissolved Mo solute stabilizes the b phase and suppresses the formation of isothermal w phases and a phases under solid-state thermal cycling. As a result, the microstructural change because of the bifunctional Mo addition not only results in an improved uniformity in mechanical properties, but simultaneously enhances the strength and ductility. In general, the design strategy provides a pathway for achieving uniform and enhanced mechanical properties in parts produced by 3D printing. Mechanical properties

We added 2.5 wt % and 5.0 wt % Mo to Ti-5553 (which we refer to as Ti-5553+2.5Mo and Ti5553+5Mo, respectively) by mechanical mixing and produced the Ti-5553 parts with and without Mo additions using refined L-PBF processing parameters for titanium alloys (30). Given the fact that the part dimension and size can affect the thermal history and consequently the mechanical properties of the part (31), we adopted two types of part geometries (dog-bone– and cuboid-shaped parts; fig. S2) to evaluate the effect of Mo additions on mechanical properties. For simplicity, we discuss the representative data of Ti-5553+5Mo unless otherwise stated. We compare the tensile engineering stress-strain curves of Ti-5553 and Ti-5553+5Mo specimens (Fig. 2A), which were machined from the cuboid parts. In marked contrast to Ti-5553, which exhibits inferior tensile properties and substantial property variation throughout the parts, Ti-5553+5Mo shows enhanced and more uniform mechanical properties (see fig. S3 for fracture analysis). We also found the high consistency in mechanical properties of Ti5553+5Mo in the dog-bone–shaped parts (fig. S4), regardless of the change in part geometry. To evaluate the degree of anisotropy of the tensile ductility, we compared the tensile ductility data of Ti-5553+5Mo with those of Ti-5553 and similar alloys that have a small deviation from the chemical composition of Ti5553—that is, Ti−5Al−5Mo−5V−3Cr−1Zr (Ti-55531) and Ti−5Al−5Mo−5V−1Cr−1Fe (Ti-55511) (Fig. 2B) (27, 28, 32–34). Generally, a substantial deviation of the ductility data point from the blue dashed 1 of 7

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

Fig. 1. Microstructures and mechanical properties of Ti-5553 produced by L-PBF. (A) The coexistence of coarse columnar b grains and spatially dependent phases in Ti-5553 produced by L-PBF. (1) Schematic illustration of the L-PBF process. (2) EBSD IPF map showing coarse columnar b grains along the building direction (BD). (3) SEM-BSE micrographs showing the phase distribution along the BD. The yellow arrows point out a phases with a darker contrast in the b-Ti matrix. (B) Schematic illustration of the microstructure heterogeneity in terms of columnar b grains and heterogeneously distributed phases on the cross-section S−S (the yz-plane), as indicated in (A). (C) TEM micrographs of Ti-5553. (1) Dark-field TEM image showing a phases. (2) TEM SAED pattern from

line indicates a high degree of anisotropy in the tensile ductility. Ti-5553+5Mo clearly shows higher and more isotropic ductility compared with Ti-5553 and similar alloys. We also compared the yield strength and the elongation to failure of Ti-5553+5Mo with those of Ti-5553 (as well as those of Ti-55531 and Ti55511) produced in the as-fabricated L-PBF state and under postprinting heat treatment conditions (Fig. 2C) (29, 32, 34–38). Compared with Ti-5553 and its similar alloys in the asfabricated state, Ti-5553+5Mo shows comparable yield strength but notably higher ductility. Postprinting heat treatment is commonly used to balance the mechanical properties of Ti5553 produced by L-PBF. Although a high yield strength (>1100 MPa) can be achieved under certain heat treatment conditions, the ductility often deteriorates substantially, with an elongation to failure of 5.0 wt % Mo) may generate further constitutional supercooling and help contribute toward the grain refinement in concert with other mechanisms. Mo-rich particles

To gain a deeper insight into the role of Mo addition in solidification, we then focused on the topmost layers of Ti-5553 and Ti-5553+5Mo. Unlike the lower region of the fabricated part, where the microstructure can be substantially altered owing to the deposition of subsequent layers, the topmost layer only undergoes one or several thermal cycles because of the deposition of neighboring tracks. In principle, such 3 of 7

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

Fig. 3. Microstructure characterization of Ti-5553 and Ti-5553+5Mo. (A) EBSD IPF maps (overlaid with the image quality map and high-angle grain boundaries with the misorientation angle ≥10°) of the xy-plane (along the scanning direction) (1) and the xz-plane (along the building direction) (2) of the Ti-5553 sample. (B) EBSD IPF maps (overlaid with image quality map and high-angle grain boundaries with the misorientation angle ≥10°) of the xy-plane

thermal cycling is much weaker compared with that experienced by the lower region, and hence the topmost layer can provide a window for understanding the role of Mo in solidification. We show the EBSD inverse pole figure (IPF) and kernel average misorientation (KAM) maps of the microstructure close to the top surfaces of Ti-5553 (Fig. 3C) and Ti-5553+5Mo (Fig. 3D), respectively. Ti-5553 exhibits coarse columnar grains in the upper layer (Fig. 3C), whereas Ti5553+5Mo is dominated by fine equiaxed grains (Fig. 3D). Further characterization of Ti-5553+ 5Mo by scanning electron microscopy (SEM)– backscattered electrons (BSE) shows that there are some particles located at the center of the equiaxed dendrite grains (Fig. 3E and fig. S8), which is a typical characteristic of a nucleating particle seeding a grain. Such particles are enriched in Mo, as evidenced by SEM–energy-dispersive x-ray spectroscopy (EDX) mapping (Fig. 3F). To further explore these Mo-rich particles, we performed SEM-EDX and transmission electron microscopy (TEM)–EDX line-scanning across the interface between the Mo-rich particles and the titanium matrix (fig. S9 and Fig. 4A). The concentration profile obtained using SEM-EDX shows that there is a gradual Zhang et al., Science 383, 639–645 (2024)

(1) and xz-plane (2) of the Ti-5553+5Mo sample. (C) EBSD maps of the top surface layers of Ti-5553. (1) IPF map overlaid with high-angle grain boundaries. (2) KAM map. (D) EBSD maps of the top surface layers of Ti-5553+5Mo. (1) IPF map overlaid with high-angle grain boundaries. (2) KAM map. (E) SEM-BSE micrograph of the top surface of Ti-5553+5Mo. (F) SEM-EDX maps showing the elemental distribution in the region in (E).

change from the Mo particle to the titanium matrix (fig. S9, B to D), which is also confirmed by TEM-EDX line-scanning (Fig. 4B). Additionally, the high-resolution TEM imaging shows that the Mo-rich particle matches perfectly with the titanium matrix with a coherent interface (Fig. 4C). Our SEM and TEM observations suggest that the transitional region across the Mo particle and titanium matrix plays an important role in grain refinement. Because these characterizations were performed on the samples that have undergone both solidification and solidstate transformations during L-PBF, it remains unclear whether the transitional region at the interface develops in the melt pool or results from solid-state diffusion after solidification. Unfortunately, this process cannot readily be characterized by current experimental techniques. To help understand the dissolution of Mo particles during solidification, we performed diffusion simulation using DICTRA (diffusion-controlled transformation) (Fig. 4D). The simulation is based on the assumption that an individual Mo particle with a radius of 2.5 mm is presented in the Ti-5553 melt pool (with a simulation cell of 100 mm) and then evolves during the heating and the subsequent cooling

9 February 2024

processes before solidification. It is evident that during the heating process (from time = 0 s to time = 7.68 × 10−4 s) and the cooling process (from time = 7.68 × 10−4 s to time = 1.54 × 10−3 s), significant diffusion of Mo solute from the Mo particle to the Ti-5553 melt leads to a Mo-rich region surrounding the particle. Careful examination of the composition profiles of Mo in the DICTRA simulation and the SEM-EDX characterization reveals that the dissolution region from simulation is slightly larger than that from experimental characterization. This observation can be explained by the fact that, under realistic conditions, the fluid flow within the melt pool encourages the Mo-rich solute to rapidly diffuse and mix with the bulk liquid, thereby leading to a narrower Mo-rich region. Nevertheless, the DICTRA simulation supports our hypothesis that the transitional region across the Mo particle and titanium matrix can occur during the moments before and during solidification, although it would be further facilitated by subsequent solid-state diffusion during cooling after solidification. On the basis of the experimental and simulation results, it appears clear that the contribution of Mo additions to grain refinement 4 of 7

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

Fig. 4. TEM characterization and DICTRA simulation of the interface between Mo particles and the titanium matrix. (A) Bright-field TEM micrograph showing the interface of a Mo particle and the titanium matrix. Insets show the SAEDs from the [011] zone axis of Mo and b-Ti. (B) TEM-EDX linescanning showing the compositional profile across the interface of the Mo particle and titanium matrix. (C) High-resolution TEM (HRTEM) image showing

works in two ways: (i) The solid cores that survive in partially dissolved Mo particles act as heterogeneous nucleation sites for the new grains. (ii) The Mo solute associated with the dissolution of Mo particles surrounding the solid core potentially generates a supercooled region that enhances heterogeneous nucleation efficiency of the partially melted Mo particles. We schematically illustrate the mechanism of grain refinement in fig. S10. Upon being heated to above the melting point of Ti-5553, some Mo particles undergo partial dissolution during L-PBF (fig. S10, A and B, stages a and b). Mo has a very high melting point of 2783 K—much higher than the 1877 to 1933 K of the b-Ti phase in Ti-5553 (43)—and is thermodynamically quasistable in the titanium melt. Given the high reactivity of liquid titanium, such characteristics are essential to ensure that some solid Mo particles survive in the melt long enough for grain nucleation. During the laser melting process, the partial dissolution of Mo particles creates an enveloping liquid enriched in Mo solute (fig. S10, A and B, stage b). Although the Mo-rich solute in this enveloping liquid is expected to rapidly diffuse and mix with the bulk liquid, the continued dissolution (or melt-back) Zhang et al., Science 383, 639–645 (2024)

the coherent interface between the Mo particle and titanium matrix (1), along with the SAEDs of Mo (2) and Ti matrix (3) taken from the ½113 zone axis. (D) DICTRA simulation showing the diffusion across the Mo particle and titanium melt during the heating and cooling processes of L-PBF. The inset shows a higher magnification of the composition profile of Mo and the boundary movement resulting from the dissolution process.

of the particle provides a steady flux of Mo solute. The local Mo composition in this region may reach very high levels well beyond the nominal bulk alloy composition. The DICTRA simulation supports concentrations of ~20 wt % Mo extending a few micrometers into the melt; however, a very steep concentration gradient closer to the particle-liquid interface exists, so the local concentration may exceed this. The local Q and DTFR values increase to ~18.5 K at 20 wt % Mo but would increase further with the concentration of Mo solute. This Mo-rich liquid surrounding the surviving solid Mo core may provide a supercooled zone if the temperature of the melt pool drops rapidly (fig. S10B, stages c and d). Although this supercooled zone arises as a result of localized composition variations, its generation is conceptually different from the traditional notion of constitutional supercooling, which develops as a result of the solute rejection into the melt ahead of a growing solid-liquid interface, thereby restricting the growth of the solid phase (39). In the current case, the surrounding enriched Mo solute has developed from Mo particle dissolution, creating a rich transient Ti-Mo liquid (richer in Mo compared with the nominal alloy

9 February 2024

composition) that facilitates the development of a supercooled zone ahead of the particle as the melt temperature falls (fig. S10B, stages c and f). Consequently, the confluence of these factors and the presence of multiple Mo particles spatially distributed throughout the melt results in the fine equiaxed grain structure of Ti-55553+5Mo. Constituent phases

In addition to refining the grain structure, the Mo addition also alters the constituent phases in the microstructure in the solid state. Although some of the larger Mo powders evidently survive to nucleate equiaxed grains, many of the smaller particles added are likely fully dissolved in the melt pool and therefore increase the overall Mo solute concentration in the Ti-5553 alloy [see (30) for particle size distribution]. This inevitably affects the phase stability and microstructural evolution. We show the phase distribution along the building direction of the Ti-5553 part (Fig. 5, A and C), and, as expected, there is a clear variation in the distribution of a phase from the bottom to the top of the part. The presence of a phase is further confirmed by x-ray diffraction (XRD) (Fig. 5 of 7

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

5B). The TEM selected-area electron diffraction (SAED) pattern in Fig. 1C has some discrete diffraction spots, which are the signature for the isothermal w phase. This observation is consistent with the coexistence of the isothermal w phase and the a phase in Ti-5553 upon being heat treated in the temperature range of 523 to 773 K (44, 45). Our differential scanning calorimeter (DSC) measurement further supports the formation of these phases in this temperature range (fig. S11). The addition of Mo tends to stabilize the b phase and suppress the precipitation of the a phase. We observed that the addition of Mo up to 5.0 wt % results in an intensity reduction of the a phase in the XRD spectra (Fig. 5B). Although other phases possibly may be present in Ti-5553+5Mo that we could not detect within the sensitivity of our XRD, if other phases do exist, they have no notable effect on the mechanical properties or uniformity. In contrast to Ti-5553, Ti-5553+5Mo shows solidification cellular structures along the building direction (Fig. 5, D and E) without any evidence of the needle-like a phase. Further SEM-EDX of the cellular structures reveals that the boundary (dark region) of the cellular structures is enriched in Ti solute (Fig. 5E and fig. S12).

This observation is expected when Ti is alloyed with b-isomorphous elements such as Ta, W, Nb, and Mo because these elements have partition coefficients greater than unity (18). The formation of cellular structures has also been reported in high-solute–containing alloys produced by L-PBF—for example, 316L stainless steel (3), Ni-based superalloy (46), and Ti−42Nb alloy (47)—and is commonly associated with the microsegregation of solute in solidification. By promoting grain refinement during solidification and stabilizing the b phase under solid-state thermal cycling, the single, small Mo addition results in two notable changes in the mechanical property of Ti-5553: (i) a simultaneously improved yield strength and ductility and (ii) an enhanced uniformity in mechanical properties (Fig. 2A and table S2). The strength increase is attributed primarily to the combined effect of solid solution strengthening of Mo solute and grain refinement, based on the estimation of strengthening contributions (30) (fig. S16). Unlike the addition of b-eutectoid elements, which increase the strength at the expense of ductility (48), the Mo addition simultaneously enhances the strength and ductility.

Fig. 5. Phase analysis of Ti-5553 and Mo-doped Ti-5553. (A) Schematic illustration showing microstructural examinations of different locations (1, 2, and 3) in a Ti-5553 cubic part. (B) XRD spectra showing the effect of Mo additions on the phase change. The inset shows the b(110) peak shifts toward lower angles. arb. units, arbitrary units. (C) SEM-BSE images showing the heterogeneous phase distribution along the building direction of the Ti-5553 part. The numbers in the upper right corners indicate the locations where the SEM-BSE was Zhang et al., Science 383, 639–645 (2024)

9 February 2024

The ductility improvement is mainly attributed to the suppression of precipitation of metastable phases by Mo solute as opposed to other potential effects, such as the transformationinduced plasticity effect (30). The coherent interface between the Mo particle and Ti alloy may not adversely affect the fatigue performance (30). Furthermore, the nanoparticles contribute to the high uniformity of mechanical properties of Ti-5553+5Mo because of their bifunctional role in the elimination of both textured columnar grain and phase heterogeneities. The primary aim of our design strategy is to demonstrate that we can eliminate the coexistence of columnar grains and phase heterogeneities through a single additive. Because we have not fully explored the compositional space, we do not know whether 5.0 wt % Mo is the optimal addition level, so further finetuning may lead to greater improvement in mechanical properties. Conclusion

We explored how to simultaneously address the formation of columnar grains and heterogeneously distributed phases in the Ti-5553 alloy produced by L-PBF. We have shown that the

performed, as schematically illustrated in (A). These SEM-BSE images were taken using a concentric backscattered detector at a low accelerating voltage (3 kV) and a small working distance (6 mm). The phase contrast stems partially from electron channeling (25). (D) SEM-BSE images showing solidification cellular structures in Ti-5553+5Mo. The numbers indicate the locations where the SEM-BSE was performed, as schematically illustrated in (A). (E) SEM-EDX maps showing elemental segregation in cellular structures. 6 of 7

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

addition of Mo up to 5.0 wt % to Ti-5553 results in substantial CET and grain refinement, which we attributed to heterogeneous nucleation on partially unmelted Mo particles, with the dissolved Mo solute developing a supercooling zone that enhances the grain refinement efficiency. Additionally, the dissolved Mo solute enables the elimination of phase heterogeneities in Ti-5553 by stabilizing the b phase. Compared with Ti-5553, Ti-5553+5Mo exhibits higher strength, higher ductility, and more uniform tensile properties. More broadly, we anticipate that the design strategy is likely to be applicable beyond the titanium alloy considered here and could guide the design of other alloys because both columnar grains and heterogeneously distributed phases are frequently reported in a wide range of metallic alloys produced by 3D printing (10, 23, 24, 49–52). RE FE RENCES AND N OT ES

1. D. Herzog, V. Seyda, E. Wycisk, C. Emmelmann, Acta Mater. 117, 371–392 (2016). 2. T. DebRoy et al., Prog. Mater. Sci. 92, 112–224 (2018). 3. Y. M. Wang et al., Nat. Mater. 17, 63–71 (2018). 4. R. Cunningham et al., Science 363, 849–852 (2019). 5. P. Kürnsteiner et al., Nature 582, 515–519 (2020). 6. C. Panwisawas, Y. T. Tang, R. C. Reed, Nat. Commun. 11, 2327 (2020). 7. T. Zhang et al., Science 374, 478–482 (2021). 8. J. H. Martin et al., Nature 549, 365–369 (2017). 9. D. Zhang et al., Nature 576, 91–95 (2019). 10. T. M. Pollock, A. J. Clarke, S. S. Babu, Metall. Mater. Trans. A 51, 6000–6019 (2020). 11. B. E. Carroll, T. A. Palmer, A. M. Beese, Acta Mater. 87, 309–320 (2015). 12. T. DebRoy, T. Mukherjee, H. L. Wei, J. W. Elmer, J. O. Milewski, Nat. Rev. Mater. 6, 48–68 (2021). 13. M. J. Bermingham, D. H. StJohn, J. Krynen, S. Tedman-Jones, M. S. Dargusch, Acta Mater. 168, 261–274 (2019). 14. Q. Tan et al., Acta Mater. 196, 1–16 (2020). 15. Y. Liu et al., Acta Mater. 220, 117311 (2021). 16. P. Kontis et al., Acta Mater. 177, 209–221 (2019). 17. C. J. Todaro et al., Nat. Commun. 11, 142 (2020). 18. D. Zhang et al., Metall. Mater. Trans. A 51, 4341–4359 (2020).

Zhang et al., Science 383, 639–645 (2024)

19. M. M. Kirka, P. Nandwana, Y. Lee, R. Dehoff, Scr. Mater. 135, 130–134 (2017). 20. W. Xu et al., Acta Mater. 85, 74–84 (2015). 21. T. Mukherjee, T. DebRoy, T. J. Lienert, S. A. Maloy, P. Hosemann, Acta Mater. 209, 116775 (2021). 22. J. Zhang et al., Nat. Commun. 13, 4660 (2022). 23. P. C. Collins, D. A. Brice, P. Samimi, I. Ghamarian, H. L. Fraser, Annu. Rev. Mater. Res. 46, 63–91 (2016). 24. J. J. Lewandowski, M. Seifi, Annu. Rev. Mater. Res. 46, 151–186 (2016). 25. J. Haubrich et al., Acta Mater. 167, 136–148 (2019). 26. M. Laleh et al., Prog. Mater. Sci. 133, 101051 (2023). 27. C. Hicks, G. Sivaswamy, T. Konkova, P. Blackwell, Mater. Sci. Eng. A 825, 141928 (2021). 28. A. Caballero et al., Philos. Mag. 102, 2256–2281 (2022). 29. J. Zhang, M. Bermingham, J. Otte, Y. Liu, M. Dargusch, Scr. Mater. 223, 115066 (2023). 30. Materials and methods are available as supplementary materials. 31. S. Huang, P. Kumar, W. Y. Yeong, R. L. Narayan, U. Ramamurty, Acta Mater. 225, 117593 (2022). 32. H. Deng et al., J. Alloys Compd. 810, 151792 (2019). 33. Z. Liu et al., Mater. Sci. Eng. A 798, 140093 (2020). 34. A. Zafari, E. W. C. Lui, M. Li, K. Xia, J. Mater. Sci. Technol. 105, 131–141 (2022). 35. H. Schwab, F. Palm, U. Kühn, J. Eckert, Mater. Des. 105, 75–80 (2016). 36. H. D. Carlton, K. D. Klein, J. W. Elmer, Sci. Technol. Weld. Join. 24, 465–473 (2019). 37. T. Zhang et al., Mater. Sci. Eng. A 835, 142624 (2022). 38. B. He, J. Li, X. Cheng, H.-M. Wang, Mater. Sci. Eng. A 699, 229–238 (2017). 39. D. H. StJohn, M. Qian, M. A. Easton, P. Cao, Acta Mater. 59, 4907–4921 (2011). 40. M. J. Bermingham, D. Kent, H. Zhan, D. H. Stjohn, M. S. Dargusch, Acta Mater. 91, 289–303 (2015). 41. M. A. Easton, D. H. StJohn, Acta Mater. 49, 1867–1878 (2001). 42. M. S. K. K. Y. Nartu et al., Nat. Commun. 14, 3288 (2023). 43. C. Leyens, M. Peters, Eds., Titanium and Titanium Alloys: Fundamentals and Applications (Wiley, 2005). 44. N. G. Jones, R. J. Dashwood, M. Jackson, D. Dye, Acta Mater. 57, 3830–3839 (2009). 45. Y. Zheng et al., Acta Mater. 103, 850–858 (2016). 46. T. Keller et al., Acta Mater. 139, 244–253 (2017). 47. S. Tekumalla, M. Seita, S. Zaefferer, Acta Mater. 262, 119413 (2024). 48. Z. Xiong, X. Pang, S. Liu, Z. Li, R. D. K. Misra, Scr. Mater. 195, 113727 (2021).

9 February 2024

49. S. A. Mantri, R. Banerjee, Addit. Manuf. 23, 86–98 (2018). 50. M. Duport et al., Addit. Manuf. 61, 103340 (2023). 51. Y. Tian et al., Metall. Mater. Trans. A 45, 4470–4483 (2014). 52. S. H. Sun et al., Acta Mater. 64, 154–168 (2014). 53. J. Zhang et al., Data from: Ultra-uniform, strong and ductile 3D printed titanium alloy through bifunctional alloy design, dataset, Dryad (2024); https://datadryad.org/stash/dataset/ doi:10.5061/dryad.6t1g1jx50. AC KNOWLED GME NTS

We thank M. Y. Lu for providing the nanoindentation equipment. We acknowledge the Australian Microscopy and Microanalysis Research Facility at the Centre for Microscopy and Microanalysis (CMM), The University of Queensland, for the facilities and technical assistance. Funding: This study received support from Australian Research Council Research Hub for Advanced Manufacturing of Medical Devices grant IH150100024 (M.S.D.); Australian Research Council Discovery Project grant DP220102748 (M.J.B.); National Key Research and Development Program of China grant 2021YFB3702101 (X.H. and Z.H.); and the “111” Project from the Ministry of Education and the State Administration of Foreign Experts Affairs of China, grant B16007 (X.H. and Z.H.). Author contributions: Conceptualization: J.Z., M.J.B., X.H., and M.S.D. Methodology: J.Z., J.O., Y.L., N.Y., Y.Y., and W.L. Investigation: J.Z., M.J.B., J.O., Z.H., and M.B. Funding acquisition: M.S.D., D.H.S., and M.J.B. Project administration: M.S.D. Supervision: M.J.B., X.H., D.H.S., and M.S.D. Writing – original draft: J.Z., M.J.B., and M.S.D. Writing – review & editing: J.Z., M.J.B., J.O., Y.L., Z.H., N.Y., Y.Y., M.B., W.L., X.H., D.H.S., and M.S.D. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data are available in the main text, in the supplementary materials, or on Dryad (53). 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.adj0141 Materials and Methods Supplementary Text Figs. S1 to S18 Tables S1 to S5 References (54–73) Submitted 1 June 2023; resubmitted 11 October 2023 Accepted 18 December 2023 10.1126/science.adj0141

7 of 7

RES EARCH

PLANT REPRODUCTION

A paternal signal induces endosperm proliferation upon fertilization in Arabidopsis Sara Simonini*, Stefano Bencivenga, Ueli Grossniklaus* In multicellular organisms, sexual reproduction relies on the formation of highly differentiated cells, the gametes, which await fertilization in a quiescent state. Upon fertilization, the cell cycle resumes. Successful development requires that male and female gametes are in the same phase of the cell cycle. The molecular mechanisms that reinstate cell division in a fertilization-dependent manner are poorly understood in both animals and plants. Using Arabidopsis, we show that a spermderived signal induces the proliferation of a female gamete, the central cell, precisely upon fertilization. The central cell is arrested in S phase by the activity of the RETINOBLASTOMA RELATED1 (RBR1) protein. Upon fertilization, delivery of the core cell cycle component CYCD7;1 causes RBR1 degradation and thus S phase progression, ensuring the formation of functional endosperm and, consequently, viable seeds.

S

exual reproduction entails specification of the germline and the formation of male and female gametes, which fuse during fertilization (1). In flowering plants, fertilization involves two pairs of gametes that are produced by multicellular gametophytes. The male gametophyte (pollen) germinates to form a pollen tube that transports two sperm cells to the female gametophyte. The latter develops within the ovule and contains two female gametes: the egg cell (EC) and the central cell (CC). During double fertilization, one sperm fuses with the EC and the other with the CC (2). The fertilized EC develops into the embryo, whereas the fertilized CC forms the endosperm, a placenta-like tissue sustaining embryonic growth (3). Male and female gametes are in a quiescent state, which ensures that they do not divide in the absence of fertilization and mediates synchronization of their cell cycles when their nuclei fuse upon fertilization. In plants, depending on the species, sperm are arrested in the G1 (4–10) or G2 phase (11, 12), whereas it is unclear at which cell cycle stage the female gametes are arrested (10–14). More than 100 factors govern cell cycle arrest and progression (15, 16), and their deregulation often leads to reduced fertility because of aberrant gametogenesis and/or embryogenesis (17–21). In the model plant Arabidopsis, mutations in some cell cycle genes affect only the CC: some cause proliferation in the absence of fertilization, whereas others cause a lack of division upon nuclear fusion (22–26). These observations have fueled the hypothesis that a mechanism preventing cell division operates in the CC, which is lifted by a fertilizationdependent signal (13, 27, 28). Institute of Plant and Microbial Biology and Zurich-Basel Plant Science Center, University of Zurich, CH-8008 Zurich, Switzerland. *Corresponding author. Email: [email protected] (U.G.); [email protected] (S.S.)

Simonini et al., Science 383, 646–653 (2024)

Quiescent ECs and CCs arrest at different cell cycle stages

We assessed the stage at which mature female gametes arrest by analyzing the expression of the cell cycle machinery using available transcriptome datasets (29–31). We found that the CC expressed high levels of virtually all transcripts encoding factors involved in and necessary for DNA replication, whereas this was not the case for the egg and pollen (Fig. 1A and table S1). This suggested that the CC might be arrested in S phase. To test this, we measured the DNA content of the EC and CC by staining ovules with propidium iodide (PI) (32) at 0 to 10 hours after pollination (HAP) (Fig. 1, B to E). We quantified fluorescence intensity in the nuclei as a proxy for the DNA amount in a given nucleus (C-value) (Fig. 1, B to F). Sporophytic cells of the ovule are diploid (2n) and may be at any stage of the cell cycle. When they are in G1 or G2 phase, their ploidy/C-value is 2n/2C or 2n/4C, respectively, whereas cells in S phase have Cvalues between 2C and 4C. Indeed, sporophytic cells showed a range of ploidy/C-values from 2n/2C to 2n/4C (Fig. 1F). The CC is homodiploid at maturity, so we expected a ploidy/ C-value of 2n/2C if it was in G1 phase and 2n/4C if it was in G2 phase. However, unfertilized CCs (0 HAP) had an average ploidy/C-value of 2n/3C (Fig. 1F and fig. S1A), indicating that they had entered but not completed S phase. By contrast, unfertilized ECs, which are haploid (1n), had an average ploidy/C-value of 1n/2C (Fig. 1F, fig. S1A). Thus, their DNA had been replicated (2C) and they were in G2 phase. We also quantified the DNA content of ECs and CCs by scoring the fluorescence intensity of histone H2B fused to the fluorescent protein tdTomato expressed under the control of two different promoters, pGRP1::H2B-tdTomato (33) and pRPS5a::H2B-tdTomato (34) (fig. S1, B to F). We found that the chromatin content of the CC was ~1.5 times (mean = 1.73, n = 58) greater than that of the EC, but not two times greater, as

9 February 2024

would be expected if they were at the same cell cycle stage (fig. S1, E and F). From these two independent analyses, we conclude that before fertilization, the EC is in G2 phase, whereas the CC has entered but not completed S phase. The fact that the CC did enter but not complete S phase suggested that DNA replication would resume upon fertilization. To test this hypothesis, we incubated inflorescences of Arabidopsis with the nucleotide analog 5ethynyl-2′-deoxyuridine (EdU) to visualize DNA synthesis in situ before fertilization and at 4, 6, and 8 HAP (Fig. 1, G to I, and fig. S1G). The unfertilized CC incorporated no EdU, indicating the absence of DNA synthesis (Fig. 1I). This observation was supported by the presence, in the EC and CC, of transcripts encoding the replication licensing factor CTD1a, correlating with the absence of DNA synthesis (35, 36) (table S1). We also detected a green fluorescent protein (GFP)–tagged version of CTD1a in immature (Fig. 1J) and unfertilized CCs (Fig. 1, K to N). These data suggest that the unfertilized CC is arrested in S phase at maturity and, unlike the EC, does not reach the G2 phase (Fig. 1O). The CC completes S phase after fertilization

After fertilization, we observed changes in the CC. First, the DNA level increased from 2n/3C to about 3n/5C, indicating successful fertilization (Fig. 1F and fig. S1A) and supporting previous observations that sperm has a DNA content of 1n/2C (11). Second, at ~8 HAP, the ploidy/C-value of the fertilized CC had increased to 3n/6C, indicating that it was now in G2 phase (Fig. 1F and fig. S1A) and thus that DNA synthesis occurred after fertilization. Third, in some ovules, the EC had a ploidy/C-value of 2n/4C, whereas that of the CC remained 2n/3C (fig. S1A) with a sperm nucleus visible in its proximity (fig. S1A). This was consistent with nonsynchronous nuclear fusion (karyogamy) in ECs and CCs (28, 37–40) and indicates that sperm delivery to the female gametophyte is insufficient to trigger cell cycle reactivation in the CC. Rather, karyogamy is required, because an increase in the ploidy/C-value occurred only in CCs with a ploidy/C-value of 3n/5C. Consistent with this increase in the ploidy/C-value, we also observed EdU staining, indicative of active DNA synthesis, specifically in the CCs of fertilized ovules (Fig. 1, H and I), and depletion of CTD1a-GFP, typically occurring when the S phase is initiated, exclusively in the CC (Fig. 1, L to N). The EC behaved very differently upon fertilization: its ploidy/C-value increased from 1n/2C to 2n/4C (Fig. 1F and fig. S1A) without detectable EdU incorporation (Fig. 1I) or persistent CTD1a-GFP fluorescence (Fig. 1, K to N). These observations confirm that mature ECs and sperm cells are arrested in G2 phase (11) and reveal that replication of the CC genome is completed only after karyogamy (Fig. 1O). 1 of 8

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

Fig. 1. Quiescent ECs are in the G2 phase of the cell cycle and CCs arrest in S phase, which is completed only upon fertilization. (A) Circular heatmap of the relative expression levels of transcripts encoding cell cycle components in the EC, CC, and pollen (PO). Genes are organized according to their expression at the different phases of the cell cycle from G1 to M, as indicated. A single gene may be represented in more than one phase according to its function. (B to E) Representative images of PI-stained ovules showing the female gametophyte at various times (in HAP), as indicated, and after the first endosperm division. Endo, endosperm; Zyg, zygote; SC, sperm cell. (F) Box plots showing quantification of PI staining, as in (B) to (E), indicating the ploidy/C-values of sporophytic (Sp) cells, CC (dark green), EC (light green), endosperm (EN), and zygote (ZY) at the indicated times (in HAP). (G and H) Representative images of EdU staining of ovules at the indicated time points after pollination. (I) Box plots showing quantification of EdU signals as in (G) and (H) in Sp Simonini et al., Science 383, 646–653 (2024)

9 February 2024

cells, CCs (orange), and ECs (red) at the indicated time points (in HAP). (J to M) Confocal fluorescence microscopy analyses of CTD1a-GFP (green) in the female gametophyte before (J) and after (K) fusion of the polar nuclei (Pn) in the CC before fertilization and in the primary endosperm nucleus after fertilization [(L) and (M)]. Cell walls were stained with PI (purple). Fertilized ovules are identified by accumulation of PI in the degenerated synergid cell (Syn). (N) Quantification of CTD1a-GFP signal intensity in the nuclei of Sp cells, ECs (purple), and CCs (pink) in unfertilized and fertilized ovules. Values for each category are normalized to the averaged signal intensity value of 10 Sp cells surrounding the female gametophyte. (O) Schematic representation summarizing the cell cycle phases of the EC and CC relative to fertilization. Unfertilized mature ECs are in G2 phase, unfertilized CCs are arrested in S phase, and fertilized CCs have completed S phase. Scale bars, 20 mm. n.s., not significant; *P < 0.01; **P < 0.001; and ***P < 0.0001 according to a t test. 2 of 8

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

Degradation of RBR1 in the CC is required for cell cycle progression

To investigate the mechanism underlying karyogamy-dependent DNA synthesis in the CC, we focused on RETINOBLASTOMA RELATED1 (RBR1) (41–44), an evolutionarily conserved, potent inhibitor of entry into and progression through the S phase (45–47). In Arabidopsis, rbr1 mutant CCs undergo uncontrolled divisions to produce an endosperm-like structure (23) (Fig.

2, A and B), suggesting that RBR1 may be involved in preventing S phase completion in the CC. We investigated RBR1 dynamics in the CC around fertilization by analyzing expression of a yellow fluorescent protein (YFP)–tagged version of the protein, RBR1-YFP (48). Indeed, RBR1 accumulated in unfertilized CCs, as previously described (49) (Fig. 2, C and D). At 7 to 8 HAP, however, CCs exhibited a weaker or undetectable fluorescent signal, before RBR1-

YFP reappeared in both endosperm nuclei after the first division (Fig. 2, C to G). To determine whether RBR1 degradation in the CC is mediated by the 26S proteasome, as it is in animals and other plant tissues (48, 50), we treated pistils with the 26S proteasome inhibitors epoxomicin, syringolin-A (SylA), and MG-132 and imaged RBR1-YFP around fertilization. When treated with epoxomicin or SylA, the RBR1-YFP signal persisted in the fertilized

Fig. 2. Degradation of RBR1 in the CC is required for cell cycle progression. (A and B) Cleared, unfertilized ovules of RBR1/rbr1-3 plants showing a wild-type (WT) female gametophyte (A) and a mutant female gametophyte (B), in which the CC has undergone divisions in absence of fertilization. The CC nucleolus (A) and the endosperm-like nucleoli (B) are artificially colored cyan. Scale bars, 20 mm. (C) Quantification of RBR1-YFP signal intensity in Sp cells and CCs of unfertilized and fertilized ovules. Values for each category are normalized to the averaged signal intensity of 10 Sp cells surrounding the female gametophyte. Scale bars, 20 mm. (D to G) Representative images of ovules expressing RBR1-YFP (yellow) at various time points (in HAP), as indicated. The contrast of the image in (F) was increased to reveal the faint RBR1-YFP signal in the nucleus of the synergid cell (white arrowhead). Cell walls were stained with PI (purple). Scale bars, 20 mm. (H) Quantification of RBR1-YFP signal in Sp cells, unfertilized (Unfe) CCs (light green), and fertilized (Fert) CCs (dark green) in inflorescences treated with a mock solution or with the proteasome inhibitors epoxomicin, syringolin-A (Syl-A), and MG-132. Values for each category are normalized to the averaged signal intensity value of 10 Sp cells surrounding the female gametophyte. (I to K) Cleared seeds of inflorescences treated with a mock solution (I), epoxomicin (J), and Syl-A (K). At the top right corner is the percentage of seeds showing the corresponding phenotype. Embryo and endosperm nucleoli are artificially highlighted in yellow and cyan, respectively. Scale bars, 50 mm. n.s., not significant and ***P < 0.0001 according to a t test.

Simonini et al., Science 383, 646–653 (2024)

9 February 2024

3 of 8

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

CC at ~8 HAP, whereas it disappeared from mock-treated pistils (Fig. 2H), indicating that these inhibitors prevented or slowed down RBR1YFP degradation in the CC nucleus. Although MG-132 is reported to inhibit RBR1 degradation (48), it had no statistically significant effect on RBR1-YFP stability in our experiments (Fig. 2H). This might be explained by the chemical properties of the inhibitors: MG-132 is a reversible inhibitor that is highly unstable in water, whereas epoxomicin and Syl-A are high-affinity, water-stable, irreversible inhibitors. We repeated the epoxomicin and Syl-A treatments and left pollinated pistils to develop for another 4 days without inhibitors. A single treatment of ~2 hours led to the formation of seeds with apparently normal, globular-stage embryos that were surrounded by a few, massively enlarged endosperm nuclei (Fig. 2, I to K). Similar endosperm phenotypes were previously observed in plants lacking components of the DNA replication machinery (25–27) or CULLIN4 (51), an E3 ubiquitin-ligase involved in protein degradation by the 26S proteasome. The similarities between the phenotype of these mutants and our inhibitor treatments support the conclusion that protein degradation and DNA synthesis are required for normal endosperm development. CCs express several CYCLIN-DEPENDENT KINASES (CDKs) (19) that, by dimerizing with cyclins, phosphorylate RBR1 and provoke its degradation (fig. S2, A to E). To prevent RBR1 degradation before fertilization, the CC is equipped with KIP-RELATED PROTEINS (KRPs) that bind to CDKs and inhibit their function. In ick1,2,6,7 mutant plants, which lack four KRPs (52), we observed that 13.1% of unfertilized ovules contained two to four endosperm-like nuclei, indicating division of the CC in the absence of fertilization (fig. S2, A to D). Together with the data described above (fig. S2E), this indicates that RBR1 persistence in the CC results in cell cycle arrest, preventing progression to the G2 phase. Moreover, RBR1 degradation is a prerequisite for S phase completion, ultimately ensuring that synchronized maternal and paternal genomes initiate proper endosperm development. Sperm cell cycle components are delivered to the female gametes upon fertilization

To understand how RBR1 degradation is induced upon fertilization in the CC, we isolated CCs by laser-assisted microdissection at various time points after pollination and analyzed their transcriptomes (Fig. 3A; table S2; and fig. S3, A to D). RBR1 protein degradation in the fertilized CC occurs at ~8 HAP (Fig. 2, C to G). We therefore searched for transcripts that were the most abundant at this time. Such a peak in abundance might be caused by fertilizationdependent de novo transcription or by delivery of paternal transcripts from the sperm to the Simonini et al., Science 383, 646–653 (2024)

CC upon gamete fusion. Because RBR1 degradation occurs through a cell cycle–regulated mechanism (45, 53–55), we focused on transcripts encoding cell cycle components (Fig. 3B). Transcripts of the CYCD7;1 gene, encoding a cyclin D-type family member, were absent from unfertilized CCs but highly abundant at 8 HAP. D-type cyclins mediate entry into the S phase (56, 57). CYCD7;1 interacts with RBR1 and promotes its degradation (58, 59) and is the only D-type cyclin that is highly expressed in pollen (31, 58). Moreover, ectopic expression of CYCD7;1 induces CC proliferation in the absence of fertilization (60). We thus hypothesized that CYCD7;1 is a paternal signal that is stored in the sperm and delivered to the CC upon fertilization, thus provoking RBR1 degradation and S phase completion. To test whether CYCD7;1 is a paternal signal, we looked for CYCD7;1 mRNA in sperm and unfertilized ovules by in situ hybridization on WT ovules fertilized with pollen expressing a pCYCD7;1::CYCD7;1-YFP transgene (58). A YFP-specific antisense probe (Fig. 3, C to J) allowed us to distinguish maternal and paternal transcripts in fertilized CCs. In pCYCD7;1::CYCD7;1-YFP plants, signal was detected during pollen development (Fig. 3D) and in sperm nuclei of mature pollen grains and elongating pollen tubes (Fig. 3, E and F). At ~4 HAP, a punctate signal was observed in WT ovules, indicative of CYCD7;1-YFP transcripts in delivered sperm nuclei (Fig. 3H). In some fertilized ovules at 4 HAP, we detected signals confined to the nuclei of CCs after karyogamy (Fig. 3I), whereas at 8 HAP, they were in the cytoplasm of ECs and CCs (Fig. 3J). The presence of paternal CYCD7;1 transcripts in fertilized female gametes was confirmed in reciprocal crosses between WT and cycd7;11 mutant plants using a probe specific to the WT CYCD7;1 allele (fig. S4, E to J). We then analyzed the expression of a nuclearlocalized YFP protein under control of the pCYCD7;1 promoter (58), allowing us to monitor its spatiotemporal activity. In ovules, no YFP signal was detected in unfertilized or fertilized ECs or in CCs (Fig. 3L). These data indicate that CYCD7;1 is transcribed in pollen, where its mRNA is stored in the sperm nuclei and delivered to the female gametes upon fertilization. The CYCD7;1 protein is present in mature pollen (58, 61), so we wondered whether it is also stored in sperm nuclei and delivered to female gametes. We fertilized WT plants with pollen from pCYCD7;1::CYCD7;1-YFP transgenic plants and analyzed CYCD7;1-YFP fluorescence. CYCD7;1-YFP was detected in sperm nuclei of mature pollen (Fig. 3M), in growing pollen tubes (Fig. 3N), and in fertilized EC and CC nuclei (Fig. 3, O and P). These data demonstrate that CYCD7;1 is a paternally derived factor, with both transcripts and proteins being delivered to the female gametes. The restricted

9 February 2024

nuclear localization of mRNA and protein likely ensures that CYCD7;1 is active only upon successful karyogamy and not simply after sperm delivery. Paternal CYCD7;1 promotes cell cycle progression of the CC

To investigate whether CYCD7;1 promotes RBR1 degradation in the CC, we studied the dynamics of RBR1 degradation upon pollination by crossing pRBR1::mCherry-RBR1 plants (62) with pCYCD7;1::CYCD7;1-YFP pollen. We found that RBR1 depletion in the CC occurred after CYCD7;1 delivery (fig. S4, A to C), consistent with the notion that CYCD7;1 promotes RBR1 degradation. Moreover, CC-specific (Fig. 4, A and B) or ubiquitous (Fig. 4C) ectopic expression of CYCD7;1 induced the development of endosperm-like structures in unfertilized ovules, consistent with a previous report (60). The interaction between RBR1 and CYCD7;1, which is required for phosphorylation and thus degradation of RBR1, is mediated by a Leu-x-Cys-xGlu (LxCxE) motif (59, 63). To determine whether this interaction is necessary for RBR1 degradation, we ubiquitously expressed a CYCD7;1 LxCxE mutant variant (pRPL18:: CYCD7;1mut) and found that it did not induce CC proliferation in unfertilized ovules (Fig. 4, A to D), indicating that the CYCD7;1-RBR1 interaction is required for the S phase to proceed in the CC upon fertilization. In plants expressing RBR1-YFP and CYCD7;1 in unfertilized CCs (pMEA::CYCD7;1), the RBR1YFP signal was very weak or undetectable, and endosperm-like structures formed (Fig. 4, E and F). These findings indicate that CYCD7;1 alone is sufficient to induce RBR1 degradation in the CC and, consequently, to stimulate endosperm proliferation. To investigate whether CYCD7;1 exerts paternal control over cell division in fertilized CCs, we used the T-DNA insertion mutants cycd7;11 (58), cycd7;1-2 (58), and cycd7;1-3 and a new mutant allele created by CRISPR-Cas9, cycd7;1CRISPR (fig. S4D), in reciprocal crosses with WT plants. If paternal delivery of CYCD7;1 were required for S phase completion in the CC, then its absence would delay division only when cycd7;1 mutants were used as a male parent, but not as a female parent, in these crosses. We scored the percentage of fertilized ovules with undivided CCs (Fig. 4G) or two, four, or eight endosperm nuclei (Fig. 4, H to J). In crosses of the WT with cycd7;1 mutant pollen, we observed a significant delay in the first CC division (Fig. 4L and fig. S4E). At 10 HAP, 94% of mammals and reptiles and 82% of vascular plants being endemic to the island (1). However, isolation alone does not explain the high rate of in situ speciation of so many taxa, and within-island geological processes and their impacts on habitat distributions in space and time could help explain how this diversity has evolved (1, 2, 6). The primary drivers of diversification in Madagascar have been suggested to be the diverse climate and changes in physical geography (6). The wet, resource-rich rainforest in eastern Madagascar is starkly different from the temperate forests and subarid and arid bioclimatic zones of the west, and these bioclimatic zones have existed since at least the Oligocene (7), providing opportunities for adaptive speciation to the variety of habitats along this environmental gradient (8, 9). The highrelief topography of Madagascar has aided in the formation of refugia during Quaternary climate cycles (6) and provided mountain and river barriers to the dispersion of vertebrates, driving microendemism (10). Although these and other mechanisms for diversification have been proposed, no single process has been iden1

Swiss Federal Research Institute (WSL), 8903 Birmensdorf, Switzerland. 2Department of Environmental Systems Science, ETH Zürich, 8092 Zürich, Switzerland. 3Department of Earth Sciences, ETH Zürich, 8092 Zürich, Switzerland.

*Corresponding author. Email: [email protected] (Y.L.); [email protected] (Y.W.) †These authors contributed equally to this work. ‡These authors contributed equally to this work.

Liu et al., Science 383, 653–658 (2024)

tified that can explain why Madagascar has such high biodiversity relative to other tropical forest regions around the world (2). Biodiversity developed through vicariant speciation is often associated with tectonically active regions, where the processes of mountain building lead to complex topography, creating new habitats, generating environmental gradients, and fragmenting existing habitats (11–15). Madagascar does not fit this hypothesis well because tectonic activity since the rifting of Gondwanaland has been minimal and localized (3, 4). However, Madagascar is surrounded by passive continental margins formed during the rifting from Africa, Australia, and, most recently, India (16), and these margins include physiographic escarpments, particularly on the east coast, which represents the youngest margin (17). The main water divide of the island marks the westward limit of the eastern escarpment, which is characterized by an increase in elevation of up to 2 km over a distance of