Strewn across seabeds worldwide are hard, pinkish nodules, some as large as a soccer ball, built over centuries by tiny calcareous algae. Known as rhodoliths or “rolling stone reefs,” these stony expanses don’t look as luxuriant as other marine habitats such as coral reefs. But they harbor an astonishing array of marine life, according to a new study that used environmental DNA (eDNA) techniques to detect more than 1000 species, some probably new to science, in rhodolith beds off Brazil—the largest in the world.

“These beds certainly harbor far more diversity than currently recognized,” says study author Guilherme Pereira-Filho, a marine biologist at the Federal University of São Paulo (UNIFESP). “Worldwide, rhodolith biodiversity could rival or even exceed that of coral reefs.” And like corals they are threatened by climate change, fishing, and mineral and oil exploitation.

Rhodoliths are considered “habitat engineers” because, like corals, they form reeflike structures that serve as home to other algae, small invertebrates, and fish. Although these beds cover about 4 million square kilometers worldwide, an area larger than the combined extent of tropical coral reefs, kelp forests, and seagrass meadows, few studies have inventoried their full suite of species, notes marine biologist Fernando Morais, a rhodolith expert at the Brazilian National Museum who was not part of the new study. “Studying this fauna is very important because many of these organisms serve as food for several fish species,” some of which are economically important, he says.

In Brazil, such studies took on new urgency recently, after the Brazilian government announced plans to begin oil exploration near the mouth of the Amazon River, in an area containing a recently discovered 9500-square-kilometer rhodolith bed. Scientists and environmentalists argue that if drilling and spills damage the beds, they could take decades or more to recover because the calcareous nodules grow very slowly—often less than 1 millimeter per year.

“We are deeply worried,” says marine biologist Paulo Horta at the Federal University of Santa Catarina. The Brazilian beds are already dredged to harvest for their calcium carbonate, used as crop fertilizer. Trawling and ocean acidification, which hampers the calcification process , are also taking a toll. “Oil exploration in new areas will only exacerbate these issues,” Horta says.

The oil company Petrobras, which was granted a license to conduct exploratory research near the mouth of the Amazon, has said it has the technical capacity to avoid environmental impacts. But the reservations of skeptics grew stronger in January after a drilling fluid spill occurred, temporarily pausing exploration.

The new research, published this month in Biological Conservation , focused on rhodolith beds in two spots: around the Queimada Grande Islands, off the coast of São Paulo, and in the Abrolhos archipelago, home to the largest rhodolith bed worldwide, covering more than 20,000 square kilometers. The research team didn’t sample in the area slated for drilling. But they say their findings likely apply to rhodolith beds throughout the country.

In each location, divers collected rhodoliths from depths of between 10 and 60 meters and brought them to the surface for analysis. The team inspected the nodules and recorded the physical traits of algae and small invertebrates living on and inside of them. They also crushed rhodolith nodules in the lab and looked for genetic traces of organisms using metabarcoding, which can identify multiple species within a single mixed sample by sequencing short, specific eDNA markers.

Their analysis found DNA sequence variants representing 1800 species. When the team combined the DNA information with the morphological observations, they were able to confirm the names of at least 450 species, including 21 that had never been recorded in the southwestern Atlantic Ocean, such as the pink-spotted sea anemone Aiptasiogeton eruptaurantia , previously only found in U.S. waters and the Caribbean Sea, and the marine sponge Hooperia anfractuosa , never seen before in the entire Atlantic.

Many sequences could not be linked to a known species in international genomic databases, and some could represent organisms new to science. Among them is an alga from the genus Dissimularia , previously only observed in the Pacific Ocean. The DNA data showed the genus was present, but the sequence did not match any of its four currently recognized species.

Pereira-Filho, who has been studying rhodoliths for more than a decade, says the team would have been able to detect more taxa had species from the Global South been better represented in genomics databases. Although more than 16,000 marine species are known to occur in Brazil, for example, such databases only include DNA sequences from 3700.

The study authors estimate that just the two Brazilian rhodolith beds could hold as many as 1% of all currently known marine species worldwide, suggesting rhodoliths may be among the planet’s most biodiverse ecosystems. “We expected high biodiversity, but the estimated number we found was far above expectations,” says lead author Pedro Longo, a marine biologist and Ph.D. student at UNIFESP.

The eDNA technique also uncovered some surprises about known species, such as the presence of the jellyfish Alatina alata , likely in its microscopic juvenile stages. Although this species has been recorded in its medusa form in deep waters, the polyp stage, believed to inhabit shallower waters, had never been directly observed in nature.

“To the untrained eye, rhodoliths might look more like small pink stones than living algae forming an important habitat,” says marine biologist Nadine Schubert at the University of Algarve, whose studies have shown that rhodoliths might also be important carbon sinks, by trapping atmospheric carbon dioxide in calcium carbonate deposits. “But they provide very important services and must be further studied and protected.”

Of all the myriad animal senses, the most mysterious and controversial is the perception of magnetism. Somehow, migratory songbirds, sea turtles, and other creatures detect Earth’s magnetic field and use its directionality to help them navigate. Now, a paper published in Science has found a surprising mechanism : Iron-rich immune cells within homing pigeons’ livers seem to give the birds their magnetic compass.

“The concept … is just mind-blowing,” says Catherine Lohmann, a sensory ecologist at the University of North Carolina at Chapel Hill who was not involved. “It’s truly a new direction and a very fresh take on a controversy that’s been in the literature for a long time.”

Many animals have magnetically informed senses of direction, including birds, turtles , sharks , and dogs . A few researchers even think humans might have a vestigial magnetic sense . Exactly how this sense works has been hotly debated. An early hypothesis was that minute crystals of magnetite embedded within the animals’ tissues somehow act like compass needles. A more recent idea is that proteins in the retina, called cryptochromes, react to magnetic fields; this would allow migrating songbirds to fly in the right direction even in the dim glow of twilight. Last year, researchers studying homing pigeons discovered another mechanism. Laboratory experiments revealed that varying magnetic fields induces electric currents in their inner ears , stimulating nerves that lead to the brain.

The new study’s discovery in pigeons began with a serendipitous meeting. While attending a scientific conference, ornithologist Martin Wikelski, who studies migratory species at the Max Planck Institute of Animal Behavior, struck up a conversation with immunologist Christian Kurts of the University of Bonn. After Wikelski described the role of magnetism in animal navigation, Kurts mentioned he had found that immune cells called macrophages recovered from the spleens of mice and humans contained tiny magnetic iron particles that formed when the macrophages broke down old red blood cells and sequestered their iron atoms. Could similar immune cells be playing a role in homing pigeon navigation?

Kurts had an idea how to test the hypothesis, so he brought in Bonn postdoctoral researcher Clivia Lisowski—who had long been fascinated by how cells sense their environment—to lead the investigation. “I was hooked,” she recalls.

First, Lisowski checked whether various pigeon tissues displayed the same tiny magnetic particles as mice immune cells. She and her colleagues expected to find a hot spot in the spleen, which in mammals is the primary location where macrophages recycle red blood cells. Instead, a sensitive magnetometer showed the liver had the strongest signal of all the tissues tested. It was relatively faint but still more than 20 times the instrument’s background noise level.

Careful staining of thin slices of homing pigeon tissue confirmed that a form of iron called ferritin abounded in liver macrophages but was scarce in the spleen and absent in the beak or brain. A closer look with an electron microscope also showed many of the pigeons’ liver macrophages were right next to neurons. In mammals, neurons in the spleen can communicate with macrophages, and in both mammals and birds, these neurons also connect to the central nervous system.

Next, the researchers tested whether these iron-rich macrophages act as magnetic compasses for the pigeons through a simple, elegant experiment: knocking out the macrophages with a drug called clodronate liposomes. The team trained 34 homing pigeons, a variety bred for their skilled wayfinding, to fly a 19-kilometer route due east. During the day, pigeons use the position of the Sun to orient themselves. But when it’s cloudy and completely overcast, they rely on their magnetic sense to get their bearings. Near Lake Constance, the team injected 18 birds with clodronate and, 24 hours later, released them one by one when dense clouds completely blocked the Sun. The birds were outfitted with GPS transmitters, so the team could track the birds in real time.

All 18 birds got hopelessly lost, only returning home after the skies had cleared. In contrast, 16 birds released after getting sham injections immediately flew straight home. “For me that was the first indication that there’s something really exciting going on,” Wikelski says. To rule out the possibility that the drug had disoriented the birds or caused other side effects to make them lose their way, the researchers released drugged birds on sunny days. They flew home just fine.

“This is an extraordinarily exciting finding,” says Verner Bingman, a neuroethologist at Bowling Green State University. Still, he views the study as “a proof of concept” that should be investigated further. Rather than knocking out the macrophages, Bingman would like to see an experiment in which magnetic information from the liver is manipulated. Similar interventions were made during experiments in the 1970s, when researchers outfitted homing pigeons with metal coils that altered the magnetic fields around their heads, causing them to fly in the opposite direction.

Susanne Åkesson, an animal ecologist at Lund University, says several questions remain about the potential role of the liver in navigation, such as how the macrophages might be passing magnetic information to nearby neurons. One idea is that as the bird shifts its position relative to Earth’s magnetic field lines, the ferritin changes orientation and tugs on the web of fibers within a macrophage, possibly triggering the release of signaling molecules.

If the ferritin mechanism is confirmed, Wikelski says, it “could be very general from bees to bats to whales to sharks.” But Lohmann remains cautious. Unlike a homing pigeon flying home, the study of animals’ magnetic sense has rarely taken the straight, narrow path. “I think time will tell whether it’s correct or not,” Lohmann says, “but it’s intriguing.”

A 6-month regimen of an experimental drug for the hepatitis B virus (HBV) added to standard antivirals has “functionally cured” 19% of people in two efficacy trials, meaning they can naturally control that virus without any further treatments. The results , published today in The New England Journal of Medicine ( NEJM ) and presented at Europe’s largest meeting on liver health, come from people whose chronic HBV infections were already relatively well controlled with the existing drugs, so its effectiveness in other populations that are more challenging to treat remains unknown.

The findings are “remarkable” and “a major step” forward for the field, hepatologist Anna Lok of the University of Michigan Medical School wrote in an editorial accompanying the NEJM editorial—although she cautioned it was far from a solution to a major global problem.

GSK in January had announced that its drug, called bepirovirsen (bepi), had positive results in two phase 3 efficacy trials that involved more than 1800 participants in 29 countries, but the company did not report any details. Many researchers anticipated that about 10% of trial participants would achieve natural suppression of HBV—what the field calls a functional cure—as that’s what was found in the earlier phase 2 bepi trials. The even more impressive result is “exciting news,” says Nick Walsh, an epidemiologist at Monash University who specializes in hepatitis diseases and was unconnected to the study. “This will certainly bring much needed attention to HBV and should accelerate further efforts to find a cure, which indeed needs to be urgently rolled out,” he says.

But Walsh and other HBV researchers stress that bepi’s unprecedented powers may well have a limited impact for most of the 240 million people worldwide living with chronic HBV infections, which kill 1.1 million each year. The World Health Organization estimates only 27% of infected people have been diagnosed, and of those, fewer than 5% receive treatment.

Chronic HBV infections can destroy the liver and ultimately kill if left unchecked. They currently require lifelong treatment, and less than 1% of patients who receive existing drugs can control the virus after stopping treatments. The most commonly used HBV drugs disrupt the ability of the virus to make copies of its genes by introducing into infected cells defective analogs of the nucleosides or nucleotides—nukes—that make up its DNA. Bepi, in contrast, is an antisense oligonucleotide that derails viral replication by binding to HBV’s messenger RNA, preventing it from making needed proteins and triggering its destruction. The drug separately stimulates immune responses against the virus.

The new phase 3 studies focused on the subset of the treated population for whom the standard drugs were working. Participants had relatively low levels of the viral surface protein, or antigen, in their blood and did not have HIV infections or liver scarring known as cirrhosis.

Two-thirds of participants added bepi shots for 24 weeks to their daily nuke pills, while a control group added placebo shots. Everyone then continued with nukes for another 24 weeks. Six months after stopping all treatment, 233 of 1220 people who had received bepi had functional cures—both undetectable HBV DNA and surface antigen—versus zero of 614 participants in the placebo arm. The functional cure rate went up to 26% in participants who entered the study with the lowest levels of the viral surface antigen. Side effects were common, but rarely severe.

When HBV infects adults, typically through sex or sharing needles during drug use, the immune system clears the virus about 90% of the time. In infants who become infected, mainly by mother-to-child transmission, their less mature immune systems falter more often—90% develop chronic infections. An infection becomes chronic when HBV integrates its DNA into human chromosomes and also forms an intractable miniature chromosome inside cells called cccDNA.

Like the nukes, bepi doesn’t eliminate the embedded cccDNA—a “sterilizing” cure—but it helps suppress HBV levels to low enough levels for a long enough period to reinvigorate immune responses against the virus. Many researchers see the functional cure bepi provides as a more realistic goal than a sterilizing cure. The new results are “a long-awaited milestone in the journey toward curing HBV infection,” says Fabien Zoulim, a hepatologist at Claude Bernard University Lyon 1.

Whether a bepi-induced functional cure is lifelong remains a key issue. Zoulim notes that the durability of surface antigen loss “requires further assessment through extended follow-up.” Data from the much smaller phase 2 studies of the drug have shown viral suppression for up to 3 years in more than 90% of the functionally cured patients, according to Melanie Paff, a GSK scientist who spoke at a press briefing last week.

GSK has already submitted the new phase 3 data to drug regulatory agencies in Europe, the United States, Canada, Japan, and China. The company expects at least some approvals to occur later this year. It has yet to announce how much the drug will sell for in wealthy countries and resource-limited ones. “We are committed to pricing which balances the value of innovation and patient access, and we’ll continue to work constructively with payers around the world to achieve this,” a GSK spokesperson told Science .

Despite the promise of bepi, some hepatitis B researchers remain frustrated that pharmaceutical companies have not made more of an effort to design even better, truly curative drugs. “There have not been enough collaborations within industry—each company tries to work on its own,” Lok says, “and we expect that a cure, for many patients, will require a multiprong approach.”

About 98 million years ago, a large dinosaur with a magnificent sail on its back prowled the brackish waters of what is today eastern Morocco. Named Spinosaurus for the elongated vertebra supporting its sail, this dinosaur’s lifeways have long perplexed paleontologists. For instance, its anatomy hints it was semiaquatic, but scientists debate how much of its life was spent in the water versus on land—and what kind of watery environment it preferred.

According to research published last week in Historical Biology , at least some spinosaurs appear to have spent their days in salty estuaries and marshes , as evidenced by special glands above their eyes that would have allowed them to “cry” salty tears to filter it from their blood.

“It’s a really cool bit of work and a really intriguing set of results,” says David Hone, a paleontologist at Queen Mary University of London who was not involved with the research. “It’s impressive that they were able to draw out this data from the very limited fossil remains.”

One major mystery surrounding Spinosaurus centers on its aquatic lifestyle. Some think the dinosaurs were amphibious, spending most of their lives swimming and diving. “Spinosaurs were probably very, very good swimmers,” says David Martill, a paleontologist at the University of Portsmouth who was not involved with the work. They had flat phalanges that helped them paddle and crocodile-shaped snouts, tails, and teeth that helped them swim and grip slippery prey.

But others argue spinosaurs were in fact terrestrial shoreline stalkers, wading in shallow waters and scooping up fish such as herons. Many spinosaur specimens, including the recently announced Spinosaurus mirabilis from Niger , have been found in inland terrestrial environments, which supports this model. Isotope analyses of their fossilized bones suggest they primarily ate fish but not as much as crocodiles , indicating they were not as dependent on the water as crocodiles are today.

With only a few known spinosaur fossils for researchers to analyze and argue over, the debate appeared to have stalled. The discovery that some of these dinosaurs might have had salt glands could inject new life into scientists’ arguments.

To keep excess salt from accumulating in their blood, animals that frequently swim in saltwater or eat lots of marine prey have evolved glands to efficiently remove salt from the blood and expel it from the body as a briny ooze. In modern birds and reptiles, the glands appear in one of three places: on the top of the skull, around the eye, or on the tongue. Sharks and rays have them in the rectum. Penguins and seagulls have them just above the eye socket. Bird lineages that live in marine habitats have independently evolved salt glands at least 40 times.

“In high-salinity environments, these glands help the animals solve the problem of excreting salt,” says Andrea Cau, a paleontologist at the OPHIS Paleontological Museum and Herpetological Center and lead author of the study.

In the new research, Cau and his colleagues uncovered evidence of potential salt glands after looking at several spinosaur specimens, including Baryonyx walkeri from the United Kingdom, Irritator challengeri from Brazil, and different Spinosaurus species from Morocco. Their analysis—based on up-close inspections, high-resolution photographs, and CT scans—revealed evidence of a conspicuous depression above the eye in certain specimens. This depression, the researchers say, might have been home to a salt gland and its supporting vasculature.

“We did not start [by] assuming some gland,” says Cau, who began the project to better understand idiosyncratic bumps and grooves on spinosaur snouts. But eventually, he says he and his team “realized the hypothesis provided a good explanation for other elements,” such as the notch above the eye and the different ancient environments where spinosaurs roamed. For one, the spinosaur species exhibiting salt glands were found in what were once saltwater environments, whereas those that lacked them were more common to freshwater habitats.

“It’s really strong evidence that spinosaurs were evolving to adapt to a wider range of environments,” Hone says. “I’m actually more surprised that this isn’t more common across the spinosaurs.”

“This discovery of a salt gland supports those people who think that the skeletal evidence [illustrates] an aquatic or semiaquatic existence,” Martill says. “If they’re in the water a long time, they need a salt gland.” But Cau is agnostic about how the discovery contributes to the debate, stating a salt gland also would benefit spinosaurs if they were only heronlike waders.

However, not all scientists are convinced the gland is there in the first place. Other Spinosaurus skull fragments from Morocco and elsewhere don’t appear to show this pocket for the salt gland, indicating it may be an artifact rather than a feature spanning multiple species, says Paul Sereno, a paleontologist at the University of Chicago. “Solid supporting evidence is lacking among available specimens,” he says.

Although he concedes that the new study does not settle how Spinosaurus and its kin made their living, Cau wants people to appreciate something undebatable: the full grandeur of the animal. “The biology of this dinosaur,” he says, “is more than just the debate about its aquatic adaptations.”

The textbook view of mantle plumes—the long-lived columns of hot rock that rise from deep in the planet—makes them seem like giant blowtorches, steadily searing the crust from below. But new studies of the sea floor around Iceland are pointing to a long-debated alternative: that plumes are actually more like blobs rising in a lava lamp, delivering intermittent pulses of heat.

The work, presented this month at the general assembly of the European Geosciences Union (EGU) and published in March in Geochemistry, Geophysics, Geosystems , combines seismic imaging and offshore drilling data collected over the past 5 years. “This is the smoking gun,” said Stephen Jones, a geophysicist at the University of Birmingham, during the EGU talk.

The studies could also help explain an abiding puzzle. Mantle plumes are thought to have caused several of the planet’s mass extinctions by driving massive volcanic eruptions that belched carbon dioxide and other greenhouse gases, disrupting the climate. But such eruptions can last for millions of years, whereas the climate shocks tied to extinctions can unfold within thousands. Pulsing plumes could drive bursts of volcanic activity better matched to the climatic upheavals.

The new evidence comes from mysterious V-shaped ridges and canyons that stretch southward from Iceland along the Mid-Atlantic Ridge—a submarine mountain range formed where the Atlantic Ocean’s crust spreads apart. In 1971, Peter Vogt, then a geologist at the U.S. Naval Oceanographic Office, proposed that the smaller ridges were the handiwork of a plume pulsing beneath Iceland. As waves of heat from the plume pulses rippled southward along the base of the crust, away from Iceland along the Mid-Atlantic Ridge, they periodically generated thicker bands of crust. Seafloor spreading then pulled the older crust outward on each side of the ridge axis, leaving behind chevron-shaped tracks that open toward Iceland.

The idea remained controversial for decades, with some geologists arguing that shifts in the crust caused by tectonic faulting created the features, rather than plume pulsing. But 2 years ago, Jones and colleagues conducted a seismic imaging campaign off the southwestern coast of Iceland on the Marcus G. Langseth , a U.S. research ship. They laid dozens of ocean-bottom seismo-meters across the ridges and fired air guns into the water to generate seismic waves that rippled through the crust to the sensors.

The surveys revealed stark, repeating changes in wave speed—and therefore structure and composition—in the crust below the ridges, all the way down to the mantle. The pattern suggests the ridges formed from periodic surges in melt rather than from faulting, which would shift rather than transform the rock. “These are robust features of the data,” Jones says. “It’s definitely a melt supply process.”

The seismic data also address a criticism of the plume-pulse idea: that the ridges are not symmetrical enough to reflect pulses radiating outward from Iceland. But Jones’s team detected a broad symmetry in the upper crust on either side of the Mid-Atlantic Ridge, consistent with pulsing.

Around the same time, researchers on the JOIDES Resolution , the now-retired U.S. drill ship, recovered volcanic rocks from ridge-trough pairs in the same region. In the March paper, Nicky White, a geologist at the University of Cambridge, and his co-authors examined the rocks’ composition, which showed the ridges were created by a mantle source that was hotter by some 30°C than the source of the rock in the troughs. The extra heat would be enough to trigger bursts of melting in the upper mantle, creating the thicker crust of the ridges. “There’s extra heat associated with the V-shaped ridges,” he says. “We’ve nailed our colors to the mast.”

The two data sets strongly reinforce each other, says Sverre Planke, a marine geophysicist at the University of Oslo, with the crustal thicknesses inferred from the seismic imaging matching those predicted by White’s geochemical analyses. “It’s an important step forward.”

The findings may help explain how the massive volcanic bursts fed by mantle plumes, sometimes called large igneous provinces, can release climate-altering amounts of carbon quickly enough to trigger extinctions. A blob of hot mantle hitting an area where the crust is thin—such as the Rockall Trough northwest of the United Kingdom—might do the trick. “This is kind of the fundamental mechanism by which we can rapidly generate carbon from large igneous provinces,” Hazel Knight, a Ph.D. student working with Jones, said in a second EGU talk .

The new picture could also help explain the Paleocene-Eocene thermal maximum (PETM), an event some 56 million years ago in which global temperatures shot up by at least 5°C over several thousand years. Researchers have long suspected volcanism in the North Atlantic played a role but struggled to explain how it could produce so much carbon so quickly.

Knight and her co-authors collected new seismic images of the Eriador Ridge, an undersea hill near the Rockall Trough that formed at the time of the PETM. They found evidence of a sudden surge of magma—enough to supply “a substantial proportion of the carbon needed to drive the PETM,” Knight said.

Thomas Gernon, a geologist at the University of Southampton, says carbon released by a pulsing plume could have added to releases triggered by a tectonic split in the North Atlantic. Gernon proposed in 2022 that the split liberated additional carbon from rocks underneath Greenland and Europe. “You can have both things going on,” he says.

Iceland’s pulsing plume may not be unique; others, beneath the Canary Islands and Afar in East Africa, show hints of similar behavior. But whether most large igneous provinces dance to the same beat, Planke adds, is “a debate that’s not easy to resolve.”

When cancer researcher Jiang Yang joined Sun Yat-sen University Cancer Center in 2019 after spending 15 years at leading U.S. universities, he maintained collaborations with some of his former colleagues. But all would be off-limits under a bill introduced this month in the U.S. Congress.

The measure, called the Securing Innovation and Research from Adversaries (SIRA) Act, would prohibit U.S. scientists from using federal funding “to enter into, support, or carry out any research collaboration” with any Chinese scientist “associated with” Chinese entities on one of several U.S. government blacklists. The bill’s sweeping definition of collaboration includes co-authorship, sharing data, material transfers, and any joint supervision of students.

The blacklists are equally broad. They include any university, laboratory, or hospital considered part of the country’s “military-civilian fusion strategy” to boost the country’s standing as a science and technology superpower. Yang’s employer, a top cancer research center in Guangzhou, China, would presumably fall under the ban.

The fate of the bill is not clear given that this Congress has a lot on its plate and a dismal record of passing any legislation. But the research community is still worried about its appearance. “I think it’s more an attempt to intimidate universities” into dropping any existing collaborations and blocking new ones, speculates Caroline Wagner, a science policy guru at Ohio State University.

The SIRA Act is the latest effort by conservatives to halt all research interactions with China on the grounds that they pose an imminent threat to U.S. national security. “The CCP [Chinese Communist Party] actively exploits the openness of academia, research partnerships, and scientific collaboration to acquire sensitive technologies and advance its military capabilities,” said the bill’s co-sponsor, Representative John Moolenaar (R–MI), who is also chair of the House of Representatives’s Select Committee on the CCP. “The CCP should not get a single dime, directly or indirectly, of American research funding,” added Senator Jim Banks (R–IN), who introduced an identical bill in the Senate.

Critics say cutting off all joint research with China would deprive U.S. scientists of an important source of talent and slow progress. “The United States achieved its technological prowess by cultivating and attracting the best and brightest minds from at home and around the world,” said Representative Ro Khanna (CA), the top Democrat on the House select committee, in a statement to Science . “Overly broad legislation risks chilling legitimate, nonsensitive research and actually harms our ability to outinnovate China.”

Yang, who trained at the University of Wisconsin, Stanford University, and the Memorial Sloan Kettering Cancer Center before returning to his native China in 2019, agrees. “I feel sad about losing the chance to work with my friends and mentors,” he says. “And I also feel bad about what it could mean for ordinary people. The cessation of collaborative cancer research [between the U.S. and China] could eliminate their last hope for new therapies against such a deadly disease.”

Even if the bill does not pass, Congress could fold the language into other legislation with a better chance of being enacted, such as annual policy guidance for the Department of Defense or a spending bill. (Moolenaar tried unsuccessfully to attach an earlier version of the SIRA Act to last year’s defense bill.) Other moves aimed at punishing China, such as restrictions on importing electric vehicles, have won bipartisan support, and science advocates worry a ban on collaborative research could be equally popular.

Wagner and other research advocates are hoping Moolenaar’s bill dies and that Democrats, if they claim the majority in one or both chambers of Congress in the November midterm elections, will propose a more targeted approach to protecting U.S. interests. Her preference, which she calls “smart openness,” would make clear the distinction between suitable and inappropriate collaborations.

“I think the community supports research security,” Wagner says. “But they don’t know what fields they can work in and exactly what’s allowed. What they want from Washington is clarity.”

Yang thinks the anti-China sentiment driving the bill is already harming U.S. science. “Even though many people still believe U.S. universities are the best in the world, many Chinese students are no longer willing to pursue a postdoc in the U.S.,” Yang says. “The vibe is shifting toward other countries for collaboration.”

Women have long been underrepresented in science—and on Wikipedia. But one corner of academia may have quietly reversed part of that trend. Among biology faculty at top U.S. research universities, women are now more likely than men to have Wikipedia biographies , according to a paper published today in the Proceedings of the Royal Society B . The latest data likely reflect, at least in part, the work of organized editing campaigns aiming to include more women on the website.

“It definitely speaks to all of the amazing crowdsourcing and outreach that’s come out of our community around the gender gap,” says Kelly Doyle Kim, who studies Wikipedia’s gender gap at the Wiki Education Foundation and was not involved with the work. She says she was surprised by the findings, particularly because previous studies in other STEM fields found women academics were less likely to appear on Wikipedia than men with similar publication records.

The study authors were also surprised. “I thought that women were going to be underrepresented,” says David Alvarez-Ponce, an evolutionary genomicist at the University of Nevada, Reno and co-author of the study. “But it turns out that we found the opposite.”

Alvarez-Ponce and his colleague embarked on the study after seeing the news 2 years ago that women had finally reached 20% of biography subjects on the English-language Wikipedia, a number editors and volunteers had spent years trying to raise. The scientists wondered whether the statistic held true for women in their field, too. They manually searched for Wikipedia entries for all 5825 tenure-track and tenured faculty who were affiliated with biology departments at 146 universities as of 2024, collecting data including page length, number of edits, and annual page views. The gender of the faculty members was surmised using listed pronouns or photographs.

The team found that 9.4% of women in the data set had Wikipedia biographies, compared with 7.5% of men. The gap widened among more senior faculty—women who are full professors were almost 7% more likely than men who are full professors to appear on the site.

These trends are recent, though. By analyzing when the Wikipedia pages were created, the team found that men biologists were more likely to have biographies until 2018. Between 2019 and 2021, women and men had similar chances. Then, in 2022 the pattern reversed and women were more likely to have a Wikipedia page than men.

The researchers suspect that organized editing campaigns likely helped drive the shift. Nearly half of the women’s biographies created since 2015 were written by editors affiliated with Women in Red , a volunteer effort aimed at addressing Wikipedia’s gender imbalance.

The new study also found that women’s biographies tended to be longer than men’s, even after normalizing for publication output and career stage. But women’s pages were viewed and edited at similar rates to men’s pages once those factors were taken into account.

The findings don’t mean broader inequities in academia have been solved, Alvarez-Ponce cautions. Women remain underrepresented in senior STEM positions and often face barriers in funding, recognition, and promotion. Plus, these results for the field of biology may not extend to other disciplines.

Still, he says, “Wikipedia is a very important source of information for many people across the world, especially young people. It’s a way in which people can be exposed to role models.” For researchers interested in representation, he adds, the platform offers something rare: a massive, publicly accessible record of whom society chooses to document and whom it overlooks.

Every fall, Culex pipiens mosquitoes, the primary carriers of West Nile virus in the United States, are supposed to take their cues from the waning daylight and go dormant for the winter. Now, a new backyard study finds that the glow of a floodlight may be enough to delay that shutdown, giving the mosquitoes more opportunities to bite.

The research, published last week in the Journal of Insect Physiology , suggests artificial light at night powerfully disrupts mosquitoes’ ability to go dormant —raising the possibility that as cities grow brighter, disease transmission seasons could extend. “It’s a study that needed to be done,” says Dina Fonseca, a vector biologist at Rutgers University who was not involved in the work. “Their findings are robust.”

As fall arrives, mosquito larvae hatch into adults that are primed to fatten themselves up and wait out the winter in cool, dark spaces such as basements and caves. Scientists have long known that mosquitoes rely on shortening daylight as the main cue to enter this dormant state, known as diapause.

Earlier lab work showed that low levels of artificial light could confuse mosquitoes and delay the onset of diapause. The question was whether the same would hold in the messier environment of a real city.

To find out, researchers asked homeowners in Columbus, Ohio, to host small containers of mosquito larvae in their yards: some placed directly under existing outdoor lights, others tucked into naturally dark corners of the same property. After allowing the larvae to mature into adults, researchers collected the containers to test whether the mosquitoes within them had entered diapause or were still primed to blood-feed and reproduce.

Mosquitoes raised under light in September entered diapause at roughly one-quarter of the rate of those kept in the dark. By October, the contrast was stark: Every mosquito in a dark enclosure went dormant, while 59% of those exposed to light stayed active. “Light pollution was a much stronger inhibitor of overwintering dormancy than temperature,” said Lydia Fyie, entomologist at the University of Maine and the study’s lead author. Even lights measuring just 0.87 lux, roughly the brightness of a starlit night, were enough to trigger mosquito activity.

If mosquitoes stay active longer, that means more chances for the biting insects to pick up and transmit disease. It also means the potential for an additional generation of mosquitoes before winter, which could lead to more individuals entering the following spring and a larger population building through summer. “The longer the season, the more generations of Culex you get,” Fonseca said. “Compound interest.”

Fonseca noted one key constraint of the study: It didn’t use wild-caught mosquitoes but instead used a laboratory colony of Culex mosquitoes, which may respond differently after generations under artificial conditions. “To truly nail it, you really should do this with field-collected mosquitoes,” she said. Then again, she acknowledges that wild Culex are notoriously uncooperative in captivity, making it hard to do any wild-caught studies of the species in the lab.

The next step, researchers say, is long-term seasonal monitoring of wild mosquito populations at high- and low-light sites—tracking when diapause begins and ends in the field across multiple years.

Katie Westby, a vector biologist at Washington University in St. Louis’s Tyson Research Center, sees the findings as part of a pattern still coming into focus. “There is mounting evidence that [light at night] is a significant force on mosquito biology and behavior,” she said. But “how many mosquitoes stay active longer … and what that means for their winter survival,” Westby added, “is a bit of an open question.”

In 2015, Amanda Rojek, then a young physician who had just started work on her Ph.D. at the University of Oxford, was pulled away to respond to the largest Ebola outbreak the world has ever seen. She helped run a clinical trial in Sierra Leone called RAPIDE, part of a scramble to test potential treatments as the virus surged through West Africa. But researchers had little experience and no consensus on how to do this well. RAPIDE tested just one experimental Ebola treatment, there was no control group, and setting everything up took months. By the time it was clear the drug did not improve patients’ survival, the outbreak was fading and there was no chance to test something else. “I was despondent,” she recalls.

Over a decade later, Rojek will be among those running a clinical trial amid another large Ebola outbreak in the Democratic Republic of the Congo (DRC). This time, things are different. Drawing on past Ebola outbreaks and the COVID-19 pandemic, the World Health Organization (WHO) and researchers from around the world have developed a protocol for a quick, adaptable randomized clinical trial. And there is precedent for success: In 2019 the PALM trial, conducted during another devastating Ebola outbreak in the DRC, found that two of the four drugs it evaluated saved patients’ lives .

So on 15 May, when an epidemic of the rare Ebola Bundibugyo variant was confirmed in the DRC, WHO and the Africa Centres for Disease Control and Prevention convened researchers that evening by video call. They decided to prioritize two drugs, the broad-spectrum antiviral remdesivir and an experimental antibody called MBP134. Their trial protocol already has a green light from the ethics board in the DRC and is awaiting regulatory approval.

But although it’s now easier to spin up a rigorous trial, researchers and health workers on the ground still face major hurdles. With more than 100 Ebola cases confirmed, and more than 850 suspected, they’ll need more treatment centers and diagnostics—and will have to contend with strong mistrust of the health care system in a country in the midst of armed conflict. And with no Bundibugyo-specific vaccines available, researchers are still rushing to hash out the details of parallel prevention trials, where they hope to protect people exposed to the virus using obeldesivir, an oral drug closely related to remdesivir.

Oxford researchers who set up the lauded Recovery trial to test experimental COVID-19 treatments are part of the new Ebola effort. Recovery showed the value of “simplifying the trial design as much as possible, so we’re only really collecting absolutely essential data,” says Peter Horby, one of the principal investigators of that trial, who is working with Rojek.

Rojek and her colleagues attempted something similar in response to a 2022 Ebola outbreak in Sudan—but the outbreak fizzled before a treatment trial could launch. When another filovirus called Marburg caused an outbreak in Rwanda in 2024, they designed a trial that could outlast short outbreaks, enrolling additional patients whenever there was another epidemic somewhere.

That trial, known as PARTNERS, is now being reactivated in the DRC and Uganda. For the moment, it is intended to test two drugs. One is MBP134, a combination of two antibodies that were isolated from a survivor of the West African Ebola outbreak that started in Guinea in late 2013. In lab studies , the antibody cocktail helped ferrets and cynomolgus monkeys survive infections with different species of Ebola virus, including Bundibugyo. “This is by far the strongest preclinical data for any of the treatments or therapies,” says virologist Thomas Geisbert of the University of Texas Medical Branch.

The other is remdesivir, a licensed antiviral that was originally developed to treat hepatitis C but has also been used to treat COVID-19 with mixed results. Remdesivir was one of the two drugs that disappointed in the 2019 PALM trial, but in vitro it is more potent against Bundibugyo than the viral species that caused that outbreak, Ebola Zaire, Geisbert says. “A case can be made, I think, for considering remdesivir with this information,” he says. Trial participants will all receive supportive care and be randomly assigned to get MBP134, remdesivir, both drugs, or neither.

At a meeting last week, a WHO technical advisory group suggested testing another drug candidate: maftivimab, one of the three antibodies in a cocktail that reduced mortality in the PALM trial. Some scientists have expressed concern about the drug, however. “It did not work at all against [the Sudan strain of] Ebola in nonhuman primates,” Geisbert says—and in vitro data were better for Sudan than Bundibugyo, he notes. WHO has said it will hold a follow-up meeting on Thursday to consider additional data before any final recommendation.

Any successful trial will need treatment centers staffed with health care workers trained to support the study. “That’s going to take a little while,” says Armand Sprecher of Doctors Without Borders. Mistrust of the health care system will likely complicate the effort. People in the region had tried to alert authorities for weeks about an epidemic but had been told it was a false alarm, says Jules Villa, a medical anthropologist at the Pasteur Institute who has worked near the outbreak’s epicenter. “A week later, they're told it's Ebola, everyone is in space suits, and they are expected to agree to do expedited funerals,” he says, in which the bodies of loved ones are rapidly buried in disinfected coffins by teams wearing full hazmat suits, cutting short the mourning period.

Treatment centers will also need plenty of reliable diagnostic tests. An automated system using polymerase chain reaction technology can detect Ebola Bundibugyo, but it’s unclear just how many of the reagent-laden cartridges its manufacturer in South Korea will be able to provide, and the machines are not conveniently located near treatment units, Sprecher says. “Diagnostics are the crisis right now.”

Once the trial is running, the biggest factor for whether any drug ends up working may be how quickly those infected seek treatment. In a study in nonhuman primates published last year, Geisbert and colleagues found the four treatments from the PALM trial worked about equally well—yet only two had shown benefit in patients. “It really came down to the viral load at the time of treatment,” Geisbert says. When PARTNERS investigators have data on levels of virus in patients, they will try to distribute trial participants with different viral loads evenly among the randomized groups, Horby says.

Getting people to enroll at the first sign of symptoms is a hard sell, Sprecher acknowledges. An Ebola treatment unit “is a scary place with health care workers who dress in odd clothing and from where many people do not return alive.” But the earlier antivirals are given, “the more effective they will be, so fewer [trial participants] will be needed,” Sprecher says, “and the sooner you can declare victory and start using the treatments systematically for everyone.”

One important question is what happens if one of the drugs does show efficacy, says Craig Spencer, a public health expert at Brown University and an Ebola survivor himself. “I think there's been a great improvement in us getting trials set up,” he says. “I think there's been hardly any improvement at all in ensuring access to these treatments after the trials have been done.” The United States has bankrolled the development of MBP134 with the aim of stockpiling the drug, and it owns the existing doses. But the drug was developed using the blood of someone who survived Ebola, and its efficacy can only be tested in countries like the DRC that experience outbreaks, Spencer notes. “It’s a moral failure and a horrible shame … [to] then take that learning and put it into a stockpile in the United States, so that one day, if there's an Ebola outbreak here, people would have ready access to it in New York City.”

After her experience in Sierra Leone, Rojek changed her Ph.D. thesis topic to “Improving patient centered research during infectious disease outbreaks.” She notes that when RAPIDE launched, there were debates about whether conducting an Ebola drug trial would detract from efforts to provide basic supportive care at clinics with few resources. That’s another way the field has evolved: trials are now seen as necessary “to provide an evidence basis for care,” she says—like what’s available to “any patient I treat in a high-income country.”

Particle physicists in Europe intend to build a 91-kilometer-long circular collider—the largest accelerator ever—to smash electrons into positrons, officials at the European particle physics laboratory, CERN, announced today in an online press conference. The Future Circular Collider (FCC) would be completed by the mid-2040s, after CERN’s current atom smasher, the 27-kilometer-long Large Hadron Collider (LHC), winds down. It would cost 15 billion Swiss francs, or about $19 billion—and it might pave the way for a much more powerful, and expensive, successor.

The choice of the FCC is the result of a 2-year-long community planning exercise, and it’s hardly a surprise, as CERN has been studying the concept for years. The announcement does not commit CERN to build the massive machine, which faces funding hurdles. But its inclusion in the so-called European particle physics strategy does mark the evolution of the FCC from an idea into an official plan. (The strategy’s most immediate goal is a major upgrade of the LHC that is scheduled to start later this year.)

“It’s a monumental decision,” says Costas Fountas, a physicist at the University of Ioannina and president of the CERN Council, which consists of delegates from CERN’s 25 member nations and sets policy for the lab. “The community lined up behind the FCC as their preferred solution for the next large project at CERN, and what is even more pleasing for me is that the member states support this.”

The new machine, officially the FCC-ee, would actually be the first of two new accelerators. It would occupy a huge new tunnel at CERN and smash electrons into positrons at energies up to 0.365 tera-electron volts (TeV), generating, among other things, large numbers of Higgs bosons. The Higgs, discovered in 2012 by the LHC , anchors physicists’ explanation of how fundamental particles get their mass. Although the FCC-ee’s collision energy would be lower than the LHC’s 13.6 TeV, its electron-positron collisions would be cleaner than the LHC’s proton-proton collisions, enabling physicists to study the Higgs in unprecedented detail.

Once the FCC-ee has accomplished its mission, physicists would rip it out and replace it with a far more powerful proton collider called the FCC-hh, reusing the costly tunnel, which would account for nearly half the FCC-ee’s cost. The FCC-hh would smash protons at an energy of 100 TeV, seven times that of the LHC. It would cost much more than the FCC-ee and likely wouldn’t come on line until the 2070s.

With this two-step strategy, CERN hopes to retrace the path that led to the LHC. In 1983, the lab started to dig its current 27-kilometer-long tunnel on the outskirts of Geneva to house the Large Electron-Positron Collider (LEP). It ran from 1989 to 2000 and scrutinized particles called the W and Z bosons, which convey the weak nuclear force and were discovered using a smaller proton-antiproton collider at CERN. Then the lab replaced the LEP with the LHC, which started to take data in earnest in 2010.

Results from the LHC have spurred physicists to think big, notes Young-Kee Kim, a particle physicist at the University of Chicago. They initially hoped the LHC would find not just the Higgs, but other new particles as well, which meant the LHC’s successor could be a smaller, straight-shot linear collider that would study those particles in detail. But because the LHC has produced only the Higgs, enthusiasm has shifted from a linear collider to a scheme that could eventually lead to a much larger proton collider, which might deliver the hoped-for discoveries.

The updated European strategy jibes well with the plans of particle physicists in the United States , notes Hitoshi Murayama, a theorist at the University of California, Berkeley. In 2023, they formulated their own long-range plan, which called for them to participate in a “Higgs factory” like the FCC-ee wherever it was built. U.S. physicists also hope to build an exotic accelerator called a muon collider , but that machine would be a smaller, presumably cheaper competitor to the FCC-hh, Murayama says.

Transforming the European plan into an actual project will require a decision by the CERN Council, which would come in 2028 at the earliest, says Mark Thomson, CERN’s director-general. Before then, researchers must refine the machine’s design. “My job in the next 2 years is to take the project forward and get it to the point where we can bring it to council for a hopefully positive decision,” Thomson says.

A bigger challenge may be figuring out how to pay for the collider, as the CERN budget will only cover about half of it. The European Union might contribute 20% of the total cost, Thomson says, and philanthropists might add about $1 billion . The U.S. particle physics plan also calls for a contribution to a Higgs factory of between $1 billion and $3 billion, Murayama notes. “I’m definitely concerned about the financial model,” he says. “They still have to raise at least 2 or 3 billion Swiss francs.”

Still, Kim says she’s gratified to see the concept move forward. “Seeing the progress gives a lot of hope to the community.”

For most graduate students in science, a teaching assistantship is mainly a way to pay the bills as they pursue a research career. But for Jasmine Clark, that role allowed her to find her true calling.

“Being a science educator really, really spoke to me,” says Clark, who in 2013 received her Ph.D. in microbiology from Emory University. “I found that my niche was in the classroom, helping people to better understand science.”

Clark’s audience has only grown since then. For more than a decade she has been an instructor at Emory’s nursing school. In 2018, she was elected to the Georgia state legislature, where she has been an advocate for improving health care access and services. And come January 2027, Clark expects to begin to educate the other 434 members of the U.S. House of Representatives about the importance of using science to set policy after winning a Democratic primary last week in a deep-blue suburban Atlanta district. When seated, she will be the first Black woman in Congress to hold a science Ph.D.

Clark, 43, spoke this week with Science about her background, the roots of her political activism, and what she hopes to accomplish in Washington, D.C. Here are excerpts from that interview.

Q: When did you decide to be a microbiologist?

A: Both of my parents have medical backgrounds—my father was a Kaiser [Permanente] physician in Atlanta for many years and my mother is a nurse—and I absolutely was supernerdy in high school. So, I grew up assuming I was going to go to med school until I did a research project as an [undergraduate] honors student at the University of Tennessee. That diverted my path into graduate school.

Q: How did you get into politics?

A: I was in the classroom helping people in the [Emory] nursing school with their microbiology projects when Donald Trump got elected [in 2016]. I like to say I went to bed on November 8 a scientist and woke up the next morning a mad scientist.

Q: Five months later you led Atlanta’s March for Science, one of the biggest of the pro-science rallies that took place on 22 April 2017 around the world . How did that happen?

A: We had scientists who were interested, political advocates and activists who were interested, and we just opened the door for everyone to come together and collaborate. And because I was the only Ph.D. scientist in the room at the time, I ended up being named the director. I had no idea how many people would participate, and [I was amazed] when 10,000 people showed up. Organizing the Atlanta March for Science is probably the most difficult thing I’ve ever done until I decided to run for Congress.

Q: How do the challenges facing the country now compare with what was happening in 2017?

A: When I look at what propelled me to run for [state] office in 2017 and for Congress in 2025, a lot of the reasons are exactly the same. We’re in a situation where we are absolutely gutting our public health. We’ve cut funding for research, thousands of people at CDC [the Centers for Disease Control and Prevention] have been fired, and I worry that we’re not ready for the next pandemic. And the person in charge [Health and Human Services Secretary Robert F. Kennedy Jr.] is someone that doesn’t really believe that germs are that bad.

Q: Both parties say they are “following the science” in setting national health care policy. And a majority of members of Congress with medical backgrounds are Republicans. So how do you explain the sharp differences between the two parties?

A: Unfortunately, there are a lot of Republicans in positions of power who are Republican first and then whatever their profession is second, and they check their knowledge at the door. The difference is, I don’t have to reconcile being on the side of science and being a Democrat. I don’t have to say I’m going to leave my science part behind when I go to vote with my caucus.

Q: Where would you put yourself on the political spectrum among Democrats?

A: I’m definitely left of center, but I don’t know if I’m left enough for some people. I believe we need to move forward as a country, so in that way I’m progressive. But I’m also pragmatic, and willing to work with those on the other side of the aisle to get things done.

Q: To the extent that committees still matter in setting policy, what House committees would best allow you to advance your goals?

A: I definitely want to be on the education committee. I would love to be on oversight, because I think that it’s important to hold people like RFK Jr. accountable. I’m also concerned about health care for veterans. And I think it would be fun—and important—to be on a committee that focuses on science and technology.

Q: Although your district voted almost three-to-one in 2024 for Kamala Harris, you must still win the general election before taking office. How much will you be campaigning between now and November?

A: My district is pretty safely Democratic, so winning the primary is pretty much winning the election. But I’m not going to treat it that way. I’m going to still campaign as if I have a competitive election. But yeah, we are headed to the United States Congress, and I’m ready to inject it with a little science.

A U.S. federal mandate to make scholarly papers free to read could triple the government’s bill for publishing fees to $937 million by 2030—an increase most research agencies aren’t prepared for, says a report released yesterday by Congress’s spending-watchdog agency. But some question the report’s worrying forecast and say it should have focused more attention on exploring creative solutions to the challenge of paying for rising publishing costs amid increasingly stretched research budgets.

The analysis by the Government Accountability Office (GAO) examined nine agencies that fund research, including the National Institutes of Health (NIH), the National Science Foundation, and the departments of Energy and Agriculture. The report estimated the potential financial consequences of implementing the so-called Nelson Memo—guidance issued in 2022 under then-President Joe Biden that requires immediate and free public access to taxpayer-funded research papers and data. Most of the agencies had enacted policies to meet this policy by the end of 2025. But only NIH has analyzed and planned for how the cost would affect its operations, the report notes—a process that last year resulted in NIH issuing a controversial proposal to limit how much the agency’s grantees can spend on publication fees .

Publishers typically charge authors an average of more than $3000 for each article they make immediately free to read, and agencies allow their grantees to charge these costs to their federal grants, GAO said. Publishers say these fees help cover the cost of quality checks and other production steps that journals traditionally funded through subscriptions—though some outsiders argue the fees greatly exceed actual expenses and that publishers are simply charging what the market will bear. Publishing fees have increased in recent years; the Nelson Memo indirectly incentivized paying the fees because it disallowed a previous practice that aimed to protect publishers’ subscription revenues by allowing them to keep federally funded studies behind a paywall for 6 months or more.

The GAO report estimates that in 2024, researchers used funds from the nine agencies to pay for publishing 46% of all papers produced with agency funding, to the tune of $295 million. That cost would rise to as much as $937 million by 2030 if all papers funded by those agencies were published public access under the “author pays” business model. (GAO did not analyze the effects of NIH’s proposal to limit its spending on publishing costs, which that agency has not yet finalized.)

The increased spending for publishing anticipated by GAO also reflects inflation in public access fees, which GAO extrapolated to future years based on increases in recent years. Agencies need to plan for these increases—and balance them against funding other priorities such as equipment and staffing—because President Donald Trump’s budget proposal for the 2027 fiscal year calls for cuts in research spending, the report suggests. (Trump’s proposal also calls for ending payments for “prohibitively high publishing costs,” without defining them.)

Planning for future publication costs, however, could be complicated by ongoing delays at key science agencies in approving new grants, says Eric Schares, a librarian at Iowa State University who studies the costs of academic publishing. “Trying to project [spending on publishing costs] into the future with this current administration and funding problems, I think it’s going to be very difficult.” GAO’s report does acknowledge that changing practices could affect its cost estimate.

Publishers and universities interviewed by GAO voiced several criticisms of the author-pays business model, including that some authors lacking government funding cannot afford it. The fees also incentivize publishers to produce more papers that may lack quality and can cost institutions more than subscriptions. The report acknowledges other business models, but adds, “a variety of stakeholders we spoke with stated that these models would not effectively support large-scale, sustainable publication of federally funded research publications.”

Still, some observers fault GAO for assuming the author-pays publishing model would predominate over others that might not cost agencies so much, such as encouraging authors to deposit articles in public repositories or supporting journals that make articles free to read without charging fees or subscriptions. “It is not the federal government’s responsibility to prop up a particular business model,” says Christopher Steven Marcum, a consultant who as a White House official during the Biden administration helped write the Nelson Memo. (That memo did not endorse a particular business model.) “The report seems to say, publishers control all the power, so federal agencies just have to cough up more money for them. I’m a little bit mystified by that.”

And GAO missed an opportunity to examine how federal spending could foster the development of scholarly publishing tools that support free public access, says Alison Mudditt, CEO of PLOS, a nonprofit journal publisher that 2 decades ago helped pioneer the author-pays model. Those tools include preprint servers, platforms for sharing data, and systems for connecting metadata attached to research outputs, says Mudditt, whose organization was among the stakeholders interviewed by GAO. But currently “a large share of that [federal spending on publication costs] is flowing to a very small number of commercial publishers,” she says, “while the shared infrastructure that makes science discoverable and reusable remains chronically underfunded.”

The U.S. research community is asking the Senate to do something it hasn’t done in more than 3 decades: Hold a hearing on the president’s nominee to lead the National Science Foundation (NSF) before voting on whether to confirm him.

A hearing would give critics of President Donald Trump’s research agenda a chance to express their doubts about whether James O’Neill, who lacks an advanced science degree or experience as a researcher, is qualified to lead the $9 billion research agency. The request also highlights a dilemma facing the Senate’s Republican majority: whether it should place loyalty to Trump ahead of its constitutional duty to assess O’Neill’s competence to become NSF’s 16th director. The agency has lacked a permanent director since Sethuraman Panchanathan, a Trump appointee, resigned more than 1 year ago after the White House ordered NSF to freeze and cancel grants and shrink its staff.

“The person occupying this role matters enormously,” more than 100 scientific organizations asserted in a 14 May letter to Senator Bill Cassidy (R–LA), who chairs the committee with jurisdiction over the nomination. “We are eager to hear how Mr. O’Neill plans to keep America at the forefront of international scientific leadership.”

A second letter to Cassidy, sent today from AAAS , which publishes Science , expresses its concerns about O’Neill more explicitly. “While an unconventional background is not necessarily disqualifying, it does require greater scrutiny of his nomination by Congress,” wrote AAAS CEO Sudip Parikh. AAAS also signed onto the earlier letter.

Although Trump nominated O’Neill on 2 March, the Committee on Health, Education, Labor, and Pensions (HELP) that Cassidy chairs has yet to set a date for a hearing. He and Senate Majority Leader John Thune (R–SD) risk angering Trump by agreeing to hold one. But recent events may make them more inclined to defy the president.

Last week, Cassidy lost a chance to retain his seat after Trump backed Representative Julia Letlow (LA), who finished first in the Republican primary. Cassidy drew the president’s wrath after voting to impeach him. Thune is reportedly angry that Trump is supporting Texas’s attorney general, Ken Paxton, against an incumbent Republican senator, John Cornyn, in a primary runoff in that state next week. Giving Democrats a public forum to question Trump’s judgment could be seen as political payback.

The last NSF director to go through such a public vetting before being confirmed was Walter Massey in February 1991. It’s not something nominees look forward to, says former NSF Director Neal Lane.

“My reaction was, ‘Oh, boy, no hearing,’” says Lane, who was confirmed in 1993 without a hearing but faced a grilling by senators 5 years later prior to being confirmed as director of the White House Office of Science and Technology Policy. “Preparing for a confirmation hearing is a lot of work.”

Even if the HELP committee forgoes a hearing, it could approve O’Neill’s nomination at a business meeting, without any substantive discussion of his qualifications. The full Senate could also choose to vote on the nomination without any input from the HELP committee. But Trump would have to renominate O’Neill next year if the Senate fails to take action before the current Congress adjourns at the end of December.

Jeffery Taubenberger, an influenza researcher named last year as acting director of the U.S. National Institute of Allergy and Infectious Diseases (NIAID), has “stepped down from his position,” Senator Tammy Baldwin (D–WI) revealed at a Senate hearing today on the budget for its parent institution, the National Institutes of Health (NIH).

The departure of Taubenberger, who is best known for resurrecting the sequence of the 1918 flu virus, comes amid a broad leadership shake-up at NIAID, which has long been under attack from President Donald Trump’s administration and Republicans in Congress who think it played a role in sparking the COVID-19 pandemic. Those changes included the removal last year of NIAID’s director and the reassignments last week of three of its senior leaders —all of whom served under former NIAID Director Anthony Fauci, who became villainized by conservatives for his actions during the pandemic.

At the hearing held by a Senate appropriations panel, however, NIH Director Jay Bhattacharya cited other reasons for Taubenberger’s unexpected exit. He was asked by Senator Patty Murray (D–WA) about Taubenberger and the recent departures or reassignments of seven other top NIAID officials, at a time when there have been outbreaks of Ebola and hantavirus. “We need to have those people on the job right now, and this is, I think, deeply concerning to all of us,” said Murray, noting the Trump administration has also cut 1000 infectious disease research grants.

Bhattacharya replied that NIAID, which has a $6.6 billion budget and is NIH’s second largest institute, was moving away from biodefense work to prioritize “conditions that people actually have,” a plan he, Taubenberger, and Bhattacharya’s principal senior adviser, John Powers III, described in Nature Medicine in January. The NIH director added “that shift means that we need some new leadership.” He further said the leaders Murray mentioned “are still at the NIH, but they’ve been assigned to places where they can help with the change of mission of the NIAID to focus on infectious disease and on allergy and immunology.” (In fact, several were fired, resigned, or retired, in at least one case rather than accept a demotion.)

Murray warned that the tumult at NIAID will have reverberations. “It just seems to me we have dismantled our infectious disease research and development pipeline, and we will pay the price,” she said.

Taubenberger’s appointment as acting NIAID director in April 2025 surprised some onlookers because his experiments with the 1918 influenza virus 2 decades ago had sparked controversy and today might generate even more concern amid debate about risky “gain-of-function” (GOF) work. Some scientists and others claim research of this type, involving manipulation of a coronavirus, sparked the COVID-19 pandemic and that Taubenberger was not appropriately against GOF studies.

But he had the support of NIAID influenza researcher and collaborator Matthew Memoli, who had bonded with Bhattacharya during the pandemic over their opposition to vaccine mandates. Memoli at one point was named acting NIH director, and now serves as principal deputy director. According to a knowledgeable NIH source, Taubenberger recently interviewed for the full NIAID director job with the Department of Health and Human Services (HHS).

Some in the scientific community echo Murray’s NIAID worries. “NIAID has lost a lot of experience and talent. I am skeptical that this is necessary for the realignment that Director Bhattacharya discussed,” says Jeremy Berg, former director of NIH’s basic research institute, who has been highly critical of the Trump administration’s NIH changes. Berg, also a former Science editor-in-chief, adds: “NIAID just did not have that many people involved in civilian biodefense compared to those working on the topics that he claims are the ‘new’ priorities.”

Taubenberger’s departure “leaves [NIAID] without a hand at the helm at the worst possible moment,” adds epidemiologist Jennifer Nuzzo, who directs the Pandemic Center at the Brown University School of Public Health. “As the nation simultaneously battles multiple deadly outbreaks—hantavirus, Ebola, and measles—this vacancy is part of a chilling pattern of leadership gaps across our entire federal health infrastructure. We are essentially asking our front-line defenses to fight a multifront biological war without a permanent general in the war room or anyone to lead the research response we need to develop tools to protect Americans from these threats.”

NIH insiders tell Science they expect Powers, an infectious disease researcher who took a position at NIAID last year as a contractor, will replace Taubenberger. They say Powers and Michael Allen, an administrator with a military background who joined NIAID last year in the newly created position of chief operating officer, have been deeply involved in running the institute. Taubenberger “was not calling the shots,” says one NIAID staffer.

Taubenberger became acting director after the Trump administration forced out Jeanne Marrazzo, an infectious disease specialist who replaced Fauci when he retired. Beyond the furor over Taubenberger’s 1918 flu work, he was more recently embroiled in a controversy over a decision last year to cut $500 million in messenger RNA (mRNA) research to improve COVID-19 vaccines, which had been funded by a different HHS agency. Some of that money was instead funneled into a universal influenza vaccine being developed by Memoli and Taubenberger that relies on inactivating viruses.

Baldwin at the Senate hearing said mRNA vaccine technology could be useful during the current Ebola outbreak, and the older flu technology is considered by many experts to be “outdated and inferior.” Asked whether the new funding to Memoli and Taubenberger for inactivated virus vaccines “circumvented” peer review and was a decision handed down by HHS, Bhattacharya insisted that “is just not accurate” and that the allocation “followed the normal process” for funding intramural, or in-house, research programs.

It is unclear whether Taubenberger will stay on at NIAID and continue to run his lab. HHS did not reply to a request for an interview with him, and NIH did not respond to a request for comment.

When Philip Kranzusch puzzled out the structure of an enzyme from the cholera bacterium in 2013, the biochemist got a jolt. The folds, the active site, the overall architecture were unmistakable: This was a bacterial cousin of a human protein that acts as a sentry for invading viruses.

“A lightning bolt ran through my mind,” says Kranzusch, then a young postdoc in the laboratory of structural biologist Jennifer Doudna at the University of California (UC), Berkeley. “Immune proteins in human cells could be far more ancient than we’d thought.”

That idea was heretical, says Rotem Sorek, a microbial genomicist at the Weizmann Institute of Science. After all, “Antibodies, a superimportant aspect of our immunity, were invented in vertebrates.” At the same time, bacteria’s own defenses against the viruses that plague them—bacteriophages—appeared to have no counterparts in higher organisms. Absent in animals and plants, for example, are CRISPR—an antiphage system that Doudna and others harnessed as a gene editor—and restriction enzymes, which chop up DNA of invading phages.

“It was natural to assume that immune defenses across kingdoms of life would be unique,” says Kranzusch, now at Harvard Medical School. But over years of follow-up work, Kranzusch and Sorek showed the enzyme that surprised Kranzusch in 2013 is part of a phage defense system that mirrors key features of human immune antiviral signaling.

“It was a startling discovery,” says Aaron Whiteley, a bacteriologist at the University of Colorado Boulder, and it marked a major paradigm shift. “We convinced the field,” Kranzusch says, “that the rules of engagement in this arms race have been the same since the dawn of life.”

Around that time, researchers were uncovering a vast arsenal of immune systems in bacteria and archaea, a distinct domain of microbes. Nearly 300 systems are now known, up from just a handful a decade ago, with roles that include detecting viral infection, transmitting alarm signals, and sacrificing infected cells to save the population. “These systems do anything you can imagine … and things you cannot imagine,” says Eugene Koonin, an evolutionary biologist at the U.S. National Center for Biotechnology Information (NCBI). “Nature has essentially explored every logically possible way to cope with infection.”

Parallels with immune systems in plants and animals have multiplied as well. More than a dozen bacterial systems are known to be built from the same molecular parts as those in more complex organisms. The implication is profound: Some core elements of the human innate immune response—our first line of defense against bacteria, viruses, and fungi—may trace back to battles between microbes and phages billions of years ago. “It’s just as Jacques Monod said: ‘What is true for E. coli is true for the elephant,’” says Aude Bernheim, a microbiologist at the Pasteur Institute.

At the Symposium on the Immune System of Bacteria at Rockefeller University this month, researchers reported that some human antiviral proteins can function in bacteria to fight phages, underscoring how deeply conserved antiviral defenses may be. Scientists are already using insights from microbes to predict new players in human immunity. “We’re solving molecular mechanisms of immunity across domains of life,” says Bernheim, who, with immunologist Enzo Poirier of the Curie Institute, reported a striking example in Science last year: They showed that SIRa1, a human protein related to bacterial antiphage proteins, helps regulate innate immune signaling.

Some biologists are also turning to microbes as stripped-down systems to probe how our own immunity works, and fails. “It’s much easier to work with bacteria, so we can learn much faster,” Sorek says.

Insights into how bacteria and archaea defend themselves are beginning to reshape how scientists approach viral disease in people. These microbial systems point to more precise targets for antiviral drugs and may expand biology’s molecular toolkit, much as CRISPR did a decade ago. The pace of discovery is so rapid that “it’s like drinking from a fire hose,” says Joseph Bondy-Denomy, a microbial immunologist at UC San Francisco. “The possibilities feel unlimited.”

Bacteriophages, the threat that drove all this molecular innovation, are thought to outnumber every living entity on Earth combined and kill off prodigious numbers of microbes each day. (Phage is Greek for “devour.”) With tail structures reminiscent of lunar landers, the viruses latch onto a bacterium or an archaeon and inject their DNA, hijacking the host’s machinery to churn out new phages until the microbe bursts and releases them.

Scientists have long known that microbes can mount a vigorous defense. In the 1970s, lab studies showed infected cells often trigger their own deaths before new phages are released. Two decades later, researchers identified individual bacterial genes that block phage replication. But those observations were largely treated as curiosities—one reason why, as Bernheim puts it, “It took the field 40 years to discover CRISPR.” Even then, that system—in which an RNA guide directs a protein, Cas, to cut the DNA of phage genes—was initially viewed as an outlier rather than evidence of a broader microbial immune arsenal.

A conceptual shift took root in the early 2010s, when Koonin and Kira Makarova of NCBI noticed certain immune-related genes in bacteria and archaea cluster in “defense islands.” They argued that by searching near the clusters , biologists could reveal additional defense systems even when their functions were unknown.

That was a career-defining moment for Sorek, who had just pivoted from studying the human genome to bacteria. With so many pairs of eyes poring over the roughly 20,000 human genes, he wondered what chance remained of discovering something. By contrast, Sorek says, the millions of bacterial genomes—with billions of enigmatic genes—were like “a candy store.”

Building on the defense island concept, Sorek’s group in 2014 identified the bacteriophage exclusion system—a six-gene collection that confers resistance to phages in Bacillus cereus , a bacterium that causes food poisoning. In 2018, Sorek and his then-postdoc Shany Doron widened the search, identifying 10 more defenses .

“This was a seminal paper. It catalyzed the field,” says Michael Laub, a molecular biologist at the Massachusetts Institute of Technology. The Israeli team showed each system was protective, “but we didn’t know the function of any of them,” Sorek says.

Later that year, Doron and Adi Millman, then a Ph.D. student in Sorek’s group, identified roughly 2000 additional candidate antiphage systems. Still more emerged when Koonin and CRISPR pioneer Feng Zhang, a molecular biologist at the Broad Institute, undertook a sweeping analysis of bacterial and archaeal genomes in the GenBank database. The discoveries overwhelmed Sorek’s lab.

“We were looking at systems upon systems, deciding which to work on. It was completely addictive,” says Bernheim, who had just joined the lab as a postdoc.

But racking up dozens of new systems without fathoming how they work felt hollow to Sorek. “I transformed the lab into one that was more biochemistry-centric and we started deciphering mechanisms,” he says. Parallels with defensive systems in higher organisms came into focus.

Among the antiphage defenses Sorek’s group explored was Thoeris, which they named after an Egyptian fertility deity. Thoeris protects against phages that include T4, a lab workhorse that helped illuminate gene structure and regulation. One Thoeris protein contains a segment that looked startlingly familiar, closely resembling the Toll/interleukin-1 receptor (TIR) domain, a hallmark of immune signaling in animals and plants.

“This immediately turned our attention to the possibility of shared evolutionary roots for immunity across kingdoms of life,” Sorek says.

TIR signaling in plants has since been shown to rely on small molecules first discovered in Thoeris systems—powerful evidence that the pathway was inherited from microbes, not reinvented. In a 22 April bioRxiv preprint , Sorek’s team reported that the TIR domain of a human immune protein involved in inflammatory signaling can generate similar molecules. The finding suggests a biochemical strategy once thought unique to microbes and plants also operates in human immunity—and could point to new drug targets for inflammatory disease.

The Vibrio cholerae system that first set the field on fire revealed another parallel. When the system’s sentry enzyme detects infection—often by sensing phage proteins such as proteases—it triggers a cascade of signals that can drive the infected cell to kill itself (see graphic, below). In 2020, Sorek and Kranzusch showed this defense system, dubbed CBASS, resembles a pathway in human cells in which the enzyme cGAS detects viral DNA and activates immune signaling.

1 Detection

A cGAS-like enzyme senses infection, although exactly how remains a matter of debate.

2 Activation

Alarm signals such as cGAMP, a loop of nucleotides, activate immune proteins.

3 Annihilation

The infected bacterium self-destructs before new phages can escape.

Sorek’s group soon found two other bacterial defense mechanisms with counterparts in human cells: viperins, which make nucleotides that gum up phage transcription, and gasdermins, which infected bacteria use to self-destruct by punching holes in their plasma membrane.

Meanwhile, Zhang’s team at Broad uncovered an antiphage system unlike any other. It “blew our minds,” says biochemist Alex Gao, now at Stanford University. The system challenged the very definition of a gene, and the group took 4 years to assemble the evidence needed to get it work published.

The team had deciphered the mechanism of an unusual defense system called DRT2, built around a reverse transcriptase—an enzyme that copies RNA “backwards” into DNA. Found in Klebsiella pneumoniae , a bacterium notorious for hospital-acquired antibiotic resistance, DRT2 at first appeared to consist only of the enzyme and a mysterious strand of RNA that did not encode a protein.

Publishing simultaneously in Science with a group led by biochemist Samuel Sternberg of Columbia University, the Broad team showed that during phage infection, the enzyme converts the RNA into DNA that encodes a functional gene. That gene ultimately produces Neo, a protein highly toxic to phages. “There could be a whole other layer of genetic information and coding that has escaped detection,” Sternberg says.

“Just absolutely wild biology,” Kranzusch says. “The assumption was that bacterial defense systems would be simple. It turns out they’re just as complex as in any other kingdom.”

No version of DRT2 has been found in plants or animals. But in a preprint posted to bioRxiv in October 2025 , Sternberg’s group identified in Escherichia coli a related antiphage defense system, DRT10, that the team suggests could be an evolutionary relative of telomerase—an enzyme that has a very different purpose in eukaryotes, the domain of life that includes animals, plants, fungi, and some other microbes. Telomerase maintains protective caps for chromosomes, which would otherwise shorten during cell division.

In an “eerily similar” mechanism to telomerase, Sternberg says, DRT10 uses a noncoding RNA template to synthesize long, repeating stretches of DNA that thwart phages. “The functional and conceptual link is compelling,” Gao says.

Other defenses in bacteria and humans take aim at protein synthesis, a cellular function viruses hijack to reproduce. In a reversal of the usual path of discovery in the field, molecular biologist Artem Nemudryi of the University of Florida and colleagues started with human Schlafen proteins—known to inhibit viruses such as HIV—and traced their roots back into bacteria. There, they found Schlafen-like immune proteins that, once activated by phage infection, cleave transfer RNAs—the molecular shuttles that ferry amino acids to growing proteins—as reported in March in Nature Microbiology . One system detects a phage tail assembly protein before turning on its RNA-cutting activity.

Besides sensing viral DNA, bacteria often recognize phages through direct protein-protein interactions—sometimes targeting specific viral structures with remarkable precision. At the Rockefeller symposium, Laub’s team reported that a bacterial enzyme called KNOCK—a cousin of enzymes central to antiviral defense in eukaryotes—adds a phosphate group to a protein in the phage tail fiber. The infected cell still bursts, but the phages it releases are crippled and unable to invade new cells.

By contrast, human antiviral immunity relies largely on sniffing out foreign nucleic acids, with protein-based recognition thought to play a less prominent role, Kranzusch says. But with dozens of such protein interactions tallied in bacterial defenses thus far, he says, “we should be looking for these in human immunity.”

In some cases human and bacterial immune mechanisms are so similar they can be swapped and still work. At the Rockefeller symposium, microbiologist Kevin Forsberg of the University of Texas Southwestern Medical Center reported that a human antiviral protein dubbed ISG20, for interferon-stimulated gene 20, can defend E. coli against phage attack. It does so by destroying the viral RNA—the same function it serves in human cells.

Kranzusch expanded on that theme at the conference, describing other human ISGs that can protect bacteria from phage infection. One latched onto a structural feature in a phage protein also found in human viruses—evidence that immune defenses target aspects of viral replication preserved across billions of years of evolution.

“The findings shocked a lot of people,” Sternberg says. “Who would have thought that you could put human ISGs in a bacterium and learn something relevant about their roles in humans?”

Prokaryotic	Bacteria	Archaea	Eukaryotic	Animals	Plants	Fungi
CRISPR-Cas
Restriction enzymes
CBASS			cGAS-STING
Viperinlike SAM enzymes			Viperin
Gasdermins			Gasdermins
DRT10			Telomerase*
Thoeris			Toll/interleukin-1 receptor domains
Argonaute			RNA interference
NLR-like proteins			NLR immune receptors
Retrons
Schlafen-like proteins			Schlafen proteins

The trail of these ancient microbial defenses leads deep into the origins of complex life. Eukaryotes likely evolved from archaeal ancestors, and some human immune pathways may trace back to those roots. The viperins, for example, are widespread in archaea, as Bernheim and colleagues discovered, and eukaryotes may have inherited them directly from those ancestors.

But related systems are also found in bacteria. Exactly which domain of life first evolved viperins remains unclear, in part because microbes swap genes through horizontal transfer, often obscuring evolutionary relationships. Certain immune pathways are found in bacteria and eukaryotes but not in archaea, suggesting they were picked up via horizonal transfer. “You can imagine early eukaryotes coming along and plucking one from bacteria that would become, for instance, cGAS,” Whiteley says.

Not all bacterial systems made the leap to more complex organisms. The human genome appears to contain molecular remnants of CRISPR-related machinery, for example, but not the full system, Bondy-Denomy says. “It’s reasonable to infer that CRISPR was tried in eukaryotes, but selection didn’t keep it.”

One reason may be that immune defenses are inherently dangerous. “They are weapons,” Koonin says. “And as we all know, weapons sometimes shoot in the wrong direction and hurt their owner. Think of autoimmunity.” Broad-spectrum systems such as CRISPR may also impose heavy metabolic costs. In eukaryotes, Koonin says, narrower and less energetically expensive defenses “ultimately took center stage in immunity.”

This ancient evolutionary history promises present-day payoffs for medicine. For example, inflammasomes—molecular complexes that trigger inflammation and can drive autoinflammatory syndromes—include so-called NLR proteins that have counterparts in bacteria. Researchers have found introducing into bacterial proteins mutations resembling those linked to autoinflammatory disease hyperactivates the microbes’ immune response—suggesting they could serve as a model for human autoimmunity that sheds light on its triggers.

The human cGAS-STING pathway, which can help rally immune cells against tumors but also contribute to inflammatory disease, is a major target for drug developers. Studies of the simpler CBASS counterpart in bacteria helped resolve the atomic structure of human cGAS and clarify how STING is activated—key advances for designing therapeutics, Kranzusch says. Meanwhile, bacterial and archaeal viperins generate small molecules that defend against a broad range of viruses. “Their mechanism of action suggests these molecules can also work against viruses infecting humans,” Sorek says.

Bacterial defense systems are also a rich source of new laboratory tools. Immune systems across microbes, plants, and animals have long been a wellspring of game-changing technologies thanks to their specificity—the ability to zero in on foreign molecules, then cleave or remove them. “That’s exactly the kind of tool molecular biologists need,” Sorek says, pointing to restriction enzymes, whose ability to cut DNA at precise sequences helped launch genetic engineering, and CRISPR, which revolutionized gene editing.

Poised to join that pantheon are retrons, bacterial defenses that produce DNA-RNA hybrid molecules and can trigger phage-infected cells to self-destruct. A team led by Seth Shipman, a neuroscientist at the Gladstone Institute of Data Science and Biotechnology, has cooked up modified retrons—dubbed “recombitrons”—that efficiently splice new DNA sequences into phage genomes, potentially creating a powerful tool for engineering viruses, the team reported in Nature Biotechnology in April.

“The war between bacteria and phages has given us an immense repository of high-value enzymes and biosynthetic capabilities,” Sternberg says.

Researchers are just beginning to grasp the scope of bacterial defenses. In Science last month, Bernheim and colleagues reported using artificial intelligence to scan bacterial genomes for additional defense systems. “We found tons of domains that have never been linked to immune defense,” she says. “It’s probably hundreds of thousands of protein families. We’re sure to discover crazy molecular functions.”

In a second Science paper , Laub and colleagues unveiled DefensePredictor , a machine learning model for tracking down antiphage systems that lie outside known defense islands. Across 1000 bacterial genomes, the team identified nearly 3000 protein clusters with no evolutionary ties to known systems, revealing what they describe as a “vast, uncharacterized defense repertoire.”

They and others are eager to probe that repertoire—and discover how ancient microbial defenses might aid us in our own struggles within a viral world.

Parents often fight about how to raise their children, but a mouse study suggests an even more basic form of parental conflict starts soon after conception when the chromosomes from each side come together. The new research, published yesterday in Nature Genetics , indicates genes from one parent can sometimes unexpectedly silence those inherited from the other, and that the effect could persist for generations. If the findings extend to humans, they could help explain why some people who carry a disease-causing genetic variant never develop the condition.

The phenomenon, named paramutation, challenges the conventional view of inheritance and had not been shown to naturally occur in mammals. It’s a form of epigenetics, in which chemical modifications to DNA sequences or proteins they wrap around turn genes on or off. In paramutation, through a process scientists don’t fully understand, one copy of a gene can silence the dominant version inherited from the other parent by coating it with methyl groups.

Paramutation’s discovery in mice marks a major shift in how scientists think about epigenetic inheritance in mammals, says Andrew Pospisilik, an epigeneticist at the Van Andel Institute who was not involved in the work. Rather than “a collection of rare exceptions,” he says, these epigenetic effects are “pervasive and ubiquitous.”

Since its discovery in maize in 1950s, paramutation has been observed in other plants, fruit flies, and nematodes. The effects can be visible. For example, maize plants can inherit a “dominant” gene that turns their stalks and leaves purple. Yet some of these plants appear light green because the version, or allele, of the gene inherited from the other parent interacts with—and silences—its dominant counterpart. Paramutation has even appeared in gene-edited mice , although some researchers have dismissed the findings as an artifact of the manipulation. It was seen as “one of those strange things that happens in transgenic animals,” says physiologist Anna Krook of the Karolinska Institute.

Still, some scientists suspected paramutation happened naturally in mammals, including humans. The problem was evidence. It “had largely remained at the level of theory,” says Philippe Arnaud, an epigeneticist at Clermont Auvergne University.

Until recently, researchers lacked the tools to reliably trace which parent has passed along a particular version of a gene, called an allele. Traditional DNA sequencing didn’t read long sequences of DNA to see the variation needed to distinguish maternal and paternal alleles—the short stretches were often identical, obscuring their origin. But the arrival of long-read sequencing changed that, as researchers could examine enough DNA from a gene at once to link individual alleles to a specific parent.

To examine how epigenetics plays out in mammals, epigeneticist Andrew Feinberg of Johns Hopkins University and his collaborators crossed two strains of heavily inbred mice and tracked the inheritance and methylation states of their alleles in the liver and muscle across two generations. The team found that 7% of the inherited epigenetic marks influence gene activity in a way that contradicted Gregor Mendel’s law of dominance. Instead of a dominant allele overpowering the other copy as expected, which allele was expressed was shaped by methylation. In the liver, for example, 304 genes were heavily methylated—and likely repressed—in female mice, but not in males. If similar patterns of epigenetic inheritance are found in humans, the findings have important clinical implications, Krook says. Because drugs are detoxified in the liver, “this could be one of the main reasons why certain drugs work one way in one sex and one in another.”

Some of this gene regulation came from more established epigenetic mechanisms such as imprinting—in which certain genes are suppressed, depending on which parent they originated from—and broader sex-specific gene silencing. But Feinberg’s team also found three cases of paramutation quieting a dominant mouse gene, one on a gene thought to regulate rodent sperm function.

Among the inherited epigenetic traits, the researchers also uncovered how entirely new epigenetic patterns can emerge in mice through interactions between genes inherited from each parent. Such changes could fuel evolution, Feinberg argues, because they could infuse populations with even more variation than is possible through DNA mutations alone.

The findings could be the tip of an epigenetic iceberg. Because the study examined just two tissues and two mouse strains, the true extent of non-Mendelian inheritance in mammals is likely far more extensive, Pospisilik says.

Feinberg and his colleagues next plan to use long-read sequencing to search for paramutation in humans, as well as other forms of epigenetic inheritance. They further want to investigate how these processes are shaped by diet and other factors, which reveal the extent to which environmental influences are passed from one generation to the next.

Since its launch more than 4 years ago, NASA’s JWST space observatory has discovered an unexpected wealth of bright galaxies in the early universe . But one tiny red blob reported last year stands out as an extreme outlier: Its discoverers suggested it might be a galaxy shining less than 100 million years after the Big Bang, far earlier than theorists thought possible. Now, another team argues the object—dubbed Capotauro—may instead be something even more exotic: the long-sought explosion of a primordial, first-generation star.

The team also notes a more mundane possibility: that observers have stumbled on an unusually cool brown dwarf—a failed star the size of Jupiter—glowing faintly in the infrared as it drifts through the Milky Way. Each explanation would be significant, says Andrea Ferrara of the Scuola Normale Superiore, lead author of the paper published last week in The Open Journal of Astrophysics . “It’s a win-win situation.”

Although a galaxy so early in the universe’s history would be revolutionary because the odds of JWST seeing it “should be zero,” he says, a supernova could provide the clearest evidence yet for the nature of the universe’s first stars. Even if it’s a brown dwarf, it would be among the coldest yet detected and the farthest from Earth. Future observations could distinguish among the possibilities: Motion would favor a brown dwarf, fading brightness a supernova, and steady light a distant galaxy.

Giovanni Gandolfi of the Astronomical Observatory of Rome led the team that first flagged Capotauro as a distant galaxy candidate. He named it after a mountain in his home region of Emilia-Romagna. Follow-up JWST observations to pin down its redshift—the stretching of light by cosmic expansion that reveals distance—weren’t conclusive. Reported in a paper posted to arXiv in September 2025 , they suggested it was either an extremely distant galaxy or a brown dwarf only a few hundred light-years from Earth that had cooled to almost room temperature over billions of years, “like a witness to the history of the Milky Way,” Gandolfi says.

But Ferrara noticed something odd about Capotauro. Between the first JWST image and the follow-up more than 2 years later, the source had brightened by 20%—behavior not expected from either a galaxy or a brown dwarf. Ferrara and his colleagues instead suggest it is a pair-instability supernova, a type of blast thought to occur only in very massive stars composed almost entirely of primordial hydrogen and helium, without the heavier elements that accumulate later in cosmic history. In such stars, energetic gamma rays inside the core can spontaneously transform into electron-positron pairs, robbing the star of radiation pressure to support its outer layers and triggering a collapse that leads to a runaway thermonuclear explosion.

To support their case, the researchers took theorists’ predictions of a pair-instability supernova’s light curve—how its brightness should change over time—and compared them with the two data points for Capotauro. They match well, but to be truly convincing, the team needs a third data point in a year or more, when the brightness is predicted to drop off steeply. “If real, it could be truly something that is changing our views of the early universe,” Ferrara says. “But we want to be sure.”

Finding a pair-instability supernova in the early universe would not only prove such blasts can occur, but it could also provide astronomers’ first glimpse of a galaxy where giant first-generation stars are burning gas that came directly from the Big Bang. The best fitting light curve points to a star at least 250 times the mass of our Sun exploding less than 300 million years after the Big Bang, Ferrara and his colleagues say.

“It’d be great if it’s true, and it might be,” says Stan Woosley, a theorist at the University of California, Santa Cruz who has played a key role in developing models of pair-instability supernovae. Detecting even one such bright example corresponding to a star at the heavy end of the mass range would imply that many more primordial stars exist at lower masses, he says. And if the host galaxy is still visible after the supernova fades, he adds, its properties could be probed to learn about the size and brightness of primordial stars. But distinguishing such a supernova from other kinds of extra-bright explosions at these distances is “going to be hard,” Woosley warns.

Other astronomers lean toward a brown dwarf that’s not very distant at all. In April, Maruša Bradač of the University of Ljubljana and colleagues reported that two other extremely distant JWST galaxy candidates were actually brown dwarfs, after observing their positions shifting, they reported in a paper posted to arXiv .

But brown dwarfs are interesting in themselves, as they fill in the gap between planets and stars, Bradač adds. Because they’re so small and dim, most brown dwarfs are detected in Earth’s near neighborhood. JWST’s surveys are turning up many brown dwarfs at greater distances, revealing where in the Milky Way they form and reside, and—from their temperatures—how they age.

Like her two recent discoveries, Capotauro may also turn out to be a brown dwarf, Bradač says. “I think the jury is still out.” Astronomers won’t have long to wait: JWST may observe Capotauro again in the next few months. Its position could shift, its brightness could fade—or it could remain steady. Any of the results would excite Bradač. “I’m looking forward to those data coming.”

Grants managers at two of the U.S. government’s largest funders of scientific research have recently placed unprecedented limitations on the ability of U.S. scientists to publish with co-authors from other countries, researchers say. Units of the National Institutes of Health (NIH) are privately directing grantees to request permission in advance for any co-authorship with a scholar affiliated with a foreign institution, even if all the work was done in the United States. NASA, meanwhile, is reportedly telling some grantees that papers co-authored with researchers in China may have violated its rules.

Neither agency has publicly issued new formal guidance describing these requirements. Instead, officials are informing grantees individually, leaving researchers confused and concerned. In several cases, NIH grantees say they have been asked to remove published papers with foreign co-authors from annual progress reports to the agency. Observers say the policy creates an incentive to preemptively remove foreign co-authors from forthcoming papers.

At NIH, co-authorship by scientists with foreign affiliations—including ones working at U.S. institutions—has historically been accepted, and relatively common: According to the most recent analysis available, 30% of papers produced with NIH funding in 2017 had both U.S. and non-U.S. authors . Some oversight of these collaborations for national security considerations is reasonable, says Tobin Smith, senior vice president at the Association of American Universities, a group of leading research institutions. “You’ve got to assess the risk in each collaboration.” But, he says, “I worry, based upon what we’re hearing, that agencies are now shifting to a blanket mode, and it’s more about who you publish with than what science you are actually publishing. And that will hurt science.”

Since at least 2003, NIH has required U.S.-based investigators to obtain agency approval before publishing a paper with a “foreign component,” defined as “performance of any significant scientific element” of the research outside of the U.S. But now, NIH managers appear to have changed the definition of foreign component to include any co-authorship with a scientist affiliated with a foreign institution, even if all work for the project occurred in the U.S., says Kristin West, director of research ethics and compliance at COGR, a nonprofit that represents research universities on regulatory matters.

NIH officials are telling grantees who submitted annual progress reports for this fiscal year—which NIH reviews when deciding whether to continue funding for multiyear grants—to remove papers that name co-authors affiliated with foreign institutions if NIH had not previously approved a foreign component for the grant. Those co-authors could include visiting colleagues, students, or postdoctoral researchers temporarily working in the U.S.; overseas researchers who donated research material but didn’t take part in the research; and scientists who moved abroad after conducting the work in the U.S.

“NIH is flagging just the fact that this co-authorship was there as evidence of a foreign component without looking further,” West says. “Everybody’s very confused by this interpretation right now.”

One of those affected is Iain Drummond, who studies kidney development and regeneration at the MDI Biological Laboratory in Maine and directs its NIH-funded Centers of Biomedical Research Excellence. He became aware of NIH’s new requirements when the agency asked a different research center at MDI to remove papers with foreign co-authors from its annual progress report. So, when Drummond’s center was preparing its report, it omitted 16 of the 22 papers it had planned to list because they had co-authors affiliated with non-U.S. institutions. All had conducted the work in the U.S., including MDI’s president, Hermann Haller, who also holds an appointment at Hannover Medical School in Germany.

After removing the 16 papers, “I said, well, Jesus, we’re not reporting anything. It’s very frustrating,” Drummond says. “I don’t know how they’re going to evaluate our productivity.” He says NIH’s new approach also creates an incentive for grantees to remove foreign-affiliated authors from papers before they are submitted for publication—contrary to standard publication ethics.

Some policy directives go further. An email from NIH’s National Institute of General Medical Sciences (NIGMS) to research-center grantees, seen by Science , requires their institutions to promise that U.S. authors of papers NIH has flagged because of foreign co-authors will not collaborate with them in the future. The email does suggest institutions can ask NIH’s permission to continue those collaborations. According to a statement from a spokesperson for the Department of Health and Human Services, NIH’s parent agency, the NIGMS email is “a clarification of longstanding policy, not a new directive.” The funding mechanism in question, Institutional Development Awards, “has always been restricted to U.S.-based institutions and entities.” The statement does not address questions from Science about other NIH institutes’ actions regarding co-authors with foreign affiliations.

Drummond said he understands NIH’s interest in ensuring that funds do not flow to foreign organizations without authorization, but he doesn’t like where the enforcement move seems headed. “We are engaging in science internationally to promote science,” he says. “The easy route for us would be just to cut off foreign involvement entirely and not include foreign authors. And that to us is a concession to some form of xenophobia.”

NASA, meanwhile, appears to be building on a 2011 law, the Wolf Amendment, that bars using the agency’s money for bilateral collaborations with entities in China. NASA guidance has suggested grantees can still engage in collaborations as long as the co-author in China didn’t receive NASA funding.

But in recent months, NASA has told some grantee institutions they may be in violation of the Wolf Amendment because their researchers co-authored papers with scientists affiliated with institutions in China, even if no NASA funding went overseas, West says. And, she says, NASA has informed some grantees that violators may be subject to a lawsuit under the federal False Claims Act, which prevents willful, fraudulent use of government grants and contracts. NASA’s moves come as some members of Congress in recent months have pushed it to more closely monitor and stop violations of the Wolf Amendment.

A statement provided by a NASA spokesperson says, “The agency does not regulate independent, bilateral research conducted by U.S. scientists using non-NASA funding” and “has not adopted a new interpretation of the Wolf Amendment.”

The “patchwork” interpretations from both agencies create ambiguities that leave researchers unsure how to proceed, says Mark Barnes, a lawyer at Ropes & Gray who represents universities on regulatory matters. “If the U.S. government wants to ban all scientific contact with China, it could try to do it so. But it hasn’t. People who are trying to do the right thing often can’t figure out what that is.”

Only a handful of members of Congress have doctoral-level scientific training, and even fewer highlight those academic credentials on the campaign trail. But that’s what Jasmine Clark did in winning a Democratic primary election yesterday in Georgia—all but ensuring that, in January 2027, she will become the first Black woman with a science Ph.D. to serve in the U.S. House of Representatives.

“The majority of her ads showed her in a lab coat and described her as a scientist,” says Emory University microbiologist Eric Hunter, her former Ph.D. adviser. “That approach could have backfired, but instead it resonated with voters.”

Clark, 43, who earned her microbiology degree in 2013 from Emory, is no stranger to taking political risks. In June 2025, while serving her fourth, 2-year term as a member of the Georgia state legislature, she decided to challenge longtime U.S. Representative David Scott (D), a revered figure in his heavily Democratic metro Atlanta district. In announcing her candidacy, she promised to be “a voice for science and truth in the face of Republican disinformation.”

Along with a war chest that far exceeded Scott’s and a crowded field of newcomers, Clark’s campaign benefited from Scott’s death on 22 April at the age of 80. Yesterday, with Scott’s name still on the ballot, she garnered 56% of the vote against her five remaining opponents.

“She was a bright, smart woman who was excited about science,” recalls Hunter about Clark, who joined his lab in 2007 after doing a first-year rotation in the school’s microbiology department. Although she did well, Clark recalls, “she was upfront about not seeing herself pursuing a career as a researcher. Instead, she said she wanted to share her knowledge with others and become a health educator.” She’s done that since 2014 as an instructor at Emory’s nursing school, teaching anatomy and microbiology to students hoping to enter the program.

Her own political education began when she headed up the Atlanta chapter of the nationwide March for Science in April 2017 to protest the policies of the then–newly elected President Donald Trump. Drawing on the concentration of research institutions in and around Atlanta, Clark assumed a role that “propelled me into a whole new space,” she told The Emory Wheel , the university’s student newspaper. In November 2018, she defeated a Republican incumbent to win a seat in Georgia’s state legislature.

“I have a Ph.D. in microbiology, which makes me very different from my colleagues at the statehouse,” she told an Atlanta radio station shortly before this year’s primary election. “And I’ve been using my scientific background to fight for policies that make sense for Georgia.”

She hopes to do the same at a national level when she gets to Washington, D.C., she added. “What RFK Jr. [Health and Human Services Secretary Robert F. Kennedy Jr.] has been doing is very concerning to me. So I’m running to protect our public health system.”

Although she must win the general election in November, the odds are heavily in her favor. Clark faces Republican Jonathan Chavez, who lost to Scott in 2024 by almost 45 percentage points. Failed Democratic presidential candidate Kamala Harris won the district by a similar margin in 2024, as did Democratic Senator Raphael Warnock in 2022.

Her ability to influence national policy will be shaped by whether Democrats regain control of the House in November. But whatever the outcome of that election, Hunter thinks Clark is well-equipped to do battle with the Trump administration and congressional Republicans.

“This is a person who’s served in a Republican-dominated [Georgia] House and is well aware of what she’ll be getting into,” he notes. “Somebody has to be a voice of clarity and authority on science in Congress, and she’s quite dynamic.”

NEW HAVEN, CONNECTICUT— Just a day ago, the brain was in a living person. Now, hours after its owner died, it sits on a cart draped in tubes that quiver as they pump liters of blood substitute and other fluids through the organ, supplying oxygen and removing waste. With most of its key functions intact but its electrical activity quenched by anesthesia, the brain hovers between life and death. As it metabolizes experimental drugs, sensors record its reactions, capturing hundreds of data points on its cells, proteins, and physiology. Then, after 24 hours in this state, it will be sliced into hundreds of pieces for more detailed study.

The brain is one of more than 700 that the 5-year-old biotech startup Bexorg has nurtured and studied using a set of proprietary brain-sustaining machines it calls BrainEx. The platform grants researchers an intimate look into how potential therapies might work inside brains with neurodegenerative diseases such as Parkinson’s, Alzheimer’s, or amyotrophic lateral sclerosis. Bexorg can biopsy the brains and discover how long a drug stays in cells, whether it hits its molecular target, and any hints of side effects.

The system promises far more realistic conditions for testing drugs than lab animals or cells in a dish, its developers say. Whole brains come with decades of environmental exposures, histories of drug treatments, and unique genetics that can affect responses to experimental medicines, says physician Zvonimir Vrselja, one of Bexorg’s founders and CEO. “You get cells that have been there for 60 to 80 years.”

Early results, including a 2025 poster reporting that the preserved brains match living brains in their response to certain therapies, are encouraging, says Bruna Bellaver, who studies neurodegeneration at the University of Pittsburgh. “It’s a huge step up from mouse models,” she says of the platform.

Bexorg has largely stayed under the radar. Its founders have published no papers on its work other than early experiments with pig brains, although Vrselja says they are preparing their first paper on human ones. Now, the company is scaling up and inviting new attention. Its new lab space, which Science visited last week, will house a 1.2-meter-tall robotic arm to automate the process of slicing up to 1600 brains per year and analyzing 11,000 proteins in each. At a media event today, the company will showcase Bexorg’s assembly line process—and seek to reassure the public that its disembodied brains don’t cross ethical lines or risk regaining consciousness.

Insights from these brains will soon be put to the test: At least one of Bexorg’s collaborators, the pharmaceutical company Biohaven, is launching a clinical trial of a drug meant to boost energy supplies in ailing brains based, in part, on data gathered in Bexorg brains. Biohaven’s chief science officer, Bruce Car, thinks that by predicting a therapy’s safety and efficacy better than animals or cell models can, the system could shave years and millions of dollars off traditional drug development processes. So far, he says, “The technology has been everything it’s been promised to be.”

The BrainEx machines sit in six plexiglass cubicles in an office overlooking Yale University, where Vrselja, neuroscientist Nenad Sestan, and their colleagues conceived of the idea 10 years ago. Before each brain is placed in a machine, surgeons examine it through a loupe, then suture four plastic ports into the vessels that once supplied it with blood so it can start reacting to drugs and generating data. Once the brain is attached to the BrainEx machines, an artificial lung and kidney oxygenate and filter fluids as they flow through the organ.

Vrselja and Sestan first used the approach to restore function to the brains of pigs obtained from a local slaughterhouse. They reported the results in a 2019 Nature paper—and promptly faced concerns that the brains might preserve traces of consciousness, feel pain, or retain memories.

The brains are already almost devoid of the coordinated neural firing necessary even for minimal consciousness, says Brendan Parent, a bioethicist at New York University Langone Health and one of six ethicists on Bexorg’s advisory board. But the company also forestalls any electrical activity with the anesthetic propofol, among other measures. Bexorg obtains brains in partnership with organizations that procure donated organs for transplantation, and Vrselja says once families understand the company’s process and goals, their response is overwhelmingly positive.

Animal models have clear shortcomings, especially when it comes to testing drugs in the brain. There’s no guarantee that a drug that passes easily into a mouse’s brain will do the same in a human’s, and a harmful overdose or ineffective underdose can stop a promising therapy in its tracks. “This is threading a needle at the best of times,” Car says. “Sometimes you get it right from your [animal model] and sometimes you miss entirely.”

Recent efforts by the U.S. government to push researchers and drugmakers away from animal testing in favor of human-based systems or computer models also represent “a huge tailwind for us,” Vrselja says.

The company has raised $42.5 million to date, not including several grants and partnerships with biotech companies and universities whose amounts Bexorg declined to disclose. Beyond testing drugs—including some developed by Bexorg itself—the brains might reveal new markers of disease processes, such as neurodegeneration in Alzheimer’s, that could be useful to physicians for diagnosis and monitoring.

The approach is especially well-suited for studying neurodegenerative disorders because these don’t generally involve brain electrical activity, Vrselja says, and because donors’ brains often have more than one such condition —a phenomenon that’s been difficult to re-create and study in the lab.

Car’s team at Biohaven has used about 130 of Bexorg’s brains to test several drugs, including one intended to prevent toxic proteins from building up in the brain in diseases like Parkinson’s. The drug didn’t interact with its target in a mouse, but Car says it worked in human brains at a dose 20 times lower than the company had initially calculated—saving the company a year of development and potentially preventing the risk of serious side effects.

Biohaven is also developing a compound called BHV-8100, which interacts with metabolic enzymes to increase the brain’s energy and allows neurons to use glucose more efficiently. These metabolic pathways are damaged in many neurodegenerative conditions. The U.S. Food and Drug Administration has approved Biohaven’s application to begin clinical trials with BHV-8100 supported by data from the Bexorg brains; later this month the company will announce which disease it is targeting.

Already, “It’s a remarkable brain bank,” says Li-Huei Tsai, a neuroscientist at the Massachusetts Institute of Technology who uses brain organoids grown from human stem cells and chips that mimic the blood-brain barrier to study neurodegeneration. She points out that BrainEx brains may not be perfect models for living ones—the systems that drain fluid from the brain might work differently in a full human body, for instance, and preventing neurons from firing can affect how blood flows through the organ. Car adds that because the brains lack electrical activity, they may not indicate whether a drug will cause seizures, although the company plans to eventually remove the anesthesia from some brain slices. Car says other models can fill in the gaps.

As Bexorg expands, it may take on other disease areas, including psychiatric disorders and cancer, Vrselja says. Eventually, the team hopes to maintain brains in BrainEx for up to 2 weeks in an effort to gather far more data about long-term processes such as brain plasticity in response to treatments.

The company is also developing a machine learning model called NeuroLens that acts as a “virtual brain,” trained on the brain readouts, donors’ medical records, and protein and microscopy data from brain tissue. The model could eventually allow researchers to test new drug molecules before even going into a physical brain. In this virtual form, the pampered brains in Bexorg’s lab will live on even after their life support is withdrawn.