Nature - The International Journal of Science / 29 February 2024

Table of contents :
Ending US–China science pact would be a dangerous folly.
What I’ve learnt from teaching in Kyiv amid a war.
RECYCLING SYSTEM KEEPS EGG CELLS SQUEAKY CLEAN.
BUBBLE PRINTING: SOAP FILMS ETCHED WITH LASER PULSES.
HAND WRITING: RELIC POINTS TO ORIGINS OF BASQUE LANGUAGE.
MOVE OVER, CRISPR: RNA-EDITING THERAPIES PICK UP STEAM.
WHY VOLUNTEERS WILL GATHER DNA FROM HUNDREDS OF LAKES.
200 YEARS OF NAMING DINOSAURS: SCIENTISTS CALL FOR BETTER RULES.
THE LIFE AND DEATH OF A BOG MAN REVEALED AFTER 5,000 YEARS.
What the EU’s tough AI law means for research and ChatGPT.
JUST 5 WOMEN HAVE WON A TOP MATHS PRIZE IN THE PAST 90 YEARS.
MEGA-CRISPR TOOL GIVES A POWER BOOST TO CANCER-FIGHTING CELLS.
WHAT A TREATMENT FOR ‘SUPER GONORRHOEA’ MEANS FOR FUTURE DRUG DEVELOPMENT.
How our love of pets grew from a clash of world views.
To unravel the origin of life, treat findings as pieces of a bigger puzzle.
Save lives in the next pandemic: ensure vaccine equity now.
Why humans reciprocate but animals usually do not.
Cold war lessons for Arctic diplomacy.
Long COVID needs novel clinical trials.
Train taxonomists to save biodiversity.
Liquid-like droplets of supramolecular polymers.
Mobile DNA explains why humans don’t have tails.
Online images are more gender-biased than text.
Light can restore a heart’s rhythm.
Ion and lipid orchestration of secondary active transport.
Most of the photons that reionized the Universe came from dwarf galaxies.
Sulfur dioxide in the mid-infrared transmission spectrum of WASP-39b.
Light-driven nanoscale vectorial currents.
Monolithic silicon for high spatiotemporal translational photostimulation.
High fatigue resistance in a titanium alloy via near-void-free 3D printing.
Site-specific reactivity of stepped Pt surfaces driven by stress release.
Supramolecular polymers form tactoids through liquid–liquid phase separation.
Identifying general reaction conditions by bandit optimization.
Super-additive cooperation.
On the genetic basis of tail-loss evolution in humans and apes.
Online images amplify gender bias.
Protracted neuronal recruitment in the temporal lobes of young children.
A distinct cortical code for socially learned threat.
A single-cell time-lapse of mouse prenatal development from gastrula to birth.
Prevalence of persistent SARS-CoV-2 in a large community surveillance study.
Circulating myeloid-derived MMP8 in stress susceptibility and depression.
Autonomous transposons tune their sequences to ensure somatic suppression.
A new family of bacterial ribosome hibernation factors.
Allosteric modulation and G-protein selectivity of the Ca 2+ -sensing receptor.
NEW TOOLS DETAIL THE DYNAMIC WORLD OF CELL ORGANIZATION.
Corrections.

Citation preview

https:/ www.foxitsoftware.cn/editmac/?MD=Mshanchu

The international journal of science / 29 February 2024

Ending US–China science pact would be a dangerous folly

NATURE AND SCIENCE/ALAMY

With a renewal of the two countries’ collaborative agreement still on hold, there’s too much talk about the risks of working together — and too little about the benefits.

T

wo things can be said of the continuing delay to renewing the US–China Science and Technology Cooperation Agreement. The good news is that the two sides are still talking about continuing with the landmark 45-year-old agreement, which has yielded historic levels of research collaboration and student exchanges between the two countries. The bad news is that one or both sides could still walk away. This would be catastrophic. Wisdom and forward thinking must prevail. Before China and the United States established diplomatic relations on 1 January 1979, there was little or no formal relationship between the two nations, and high levels of mistrust. Science cooperation was identified as offering a relatively swift way to break the ice and begin establishing people-to-people contacts. Then-US president Jimmy Carter and China’s premier at the time, Deng Xiaoping, signed the science agreement before the month was out, on 31 January. Admittedly, the two countries’ motivations for pursuing scientific cooperation were different. For China, the decision was development-led. The nation was far from the research-driven power that it is now. Today, it boasts some 3,000 higher-education institutions; back then, its annual per-capita income stood at less than US$200. China’s leaders wanted to learn how to build a world-class higher education system, as well as how they might use research to boost economic growth, and, by extension, living standards, as Julian Gewirtz, a historian of China–US economic-research ties, writes in Unlikely Partners (2017). The United States also had a political goal: to steer China away from the orbit of the Soviet Union during the ongoing cold war. Today, although the two countries can hardly be described as ‘best friends forever’, the fruits of their collaboration are clear. Some 3 million Chinese students have studied at universities in the United States since the agreement was brokered. In 2021, US universities awarded more than 8,000 doctorates to students from China, out of a total of around 25,000 international doctorates. Each country is the other’s biggest research partner, by a considerable margin. Relations took a negative turn during Donald Trump’s US presidency, from 2017 to 2021. After the start of the COVID-19 pandemic, rhetoric harshened significantly, and

Wan Gang and John Holdren hold an image of Deng Xiaoping and Jimmy Carter.

A narrative took hold that there is something inherently suspicious about cooperation between US and Chinese researchers.”

this was followed by an erosion of trade and diplomatic links. In the realms of research and higher education, a narrative took hold that there is something inherently suspicious about cooperation between US and Chinese researchers — with an emphasis on known threats such as spying and intellectual-property theft. This has clearly affected collaborations, but has also had a broader reach. There has been surveillance of some innocent researchers, as we report this week (see page 1149). And Florida’s decision to stop universities hiring researchers from China (as well as from Iran and a handful of other countries) would not have seemed out of place during the cold war (see go.nature.com/3tachvy). The United States has not been alone in initiating decoupling efforts. From March 2020, China’s government adopted a policy whereby its researchers would no longer be encouraged to publish in international journals. China’s leadership has also taken to talking more and more about self-reliance, one implication of which is less and less need for collaborative effort.

Mutual benefits John Holdren, a physicist at Harvard University in Cambridge, Massachusetts, was science adviser to former US president Barack Obama and, in 2011, he and Wan Gang, China’s then minister of science and technology, renewed the US–China science pact. That agreement was designed to ensure that the benefits would be mutual, Holdren tells Nature. Those benefits are both national and global. Collaboration between the two countries on environmental protection includes projects to monitor and Nature | Vol 626 | 29 February 2024 | 927

XINHUA/CAI YANG/ALAMY

Editorials

China and the United States cooperated to ensure that miniature neutron source reactors can run on low-enriched uranium.

improve air and water quality, as well as watershed protection, and projects to reduce electronic waste — benefiting both countries in different ways. The US Environmental Protection Agency has called its relationship with China “one of its most significant”. When it comes to global challenges, researchers in China, the United States and Europe are cooperating extensively on studying the role of nature in human prosperity (A. D. Guerry et al. Proc. Natl Acad. Sci. USA 112, 7348–7355 (2015); Z. Ouyang et al. Science 352, 1455–1459; 2016). This evolving body of work is foundational to ongoing efforts to incorporate nature into how economies are valued. Another notable but little-known project aims to reduce the risk of nuclear proliferation. Since 2009, China and the United States have been working together to convert a type of nuclear research reactor called a miniature neutron source reactor so that instead of using highly enriched, weapons-grade uranium as fuel, it runs on low-enriched uranium — which cannot be used in nuclear weapons. China has supplied this type of reactor to a number of countries, including Iran, Nigeria and Pakistan. In a small way, this cooperation has contributed to a safer world. And then there’s climate change. After a period of silence that began in 2022, the two countries began talking again last year, thanks in no small measure to the long-standing relationship between their then climate envoys, John Kerry and Xie Zhenhua. Last year, California made an agreement with China pledging to cut carbon emissions and transition away from using fossil fuels. Both Kerry and Zhenhua are moving on to new roles, and the legacy of their diplomatic efforts risks being undermined if scientists in the two 928 | Nature | Vol 626 | 29 February 2024

The answer to handling risks is to assess them, manage them and mitigate them.”

nations cannot maintain their research ties. Germany’s handling of its research relations with China could offer lessons. Last month, the German Academic Exchange Service published some sensible recommendations that balance the risks of such collaborations with the benefits. The document acknowledges the benefits that have come from closer ties, while advocating what it calls a “realpolitik approach” to future links — one based on practical objectives, rather than ideology. Ultimately, it says that universities should be the ones to decide what is mutually beneficial in this regard, while taking the necessary precautions to protect against possible harm.

Risk management There are, of course, always risks when researchers from different political systems collaborate. And it shouldn’t surprise anyone that big powers spy on each other, says Holdren. But, as with most applications of science in public affairs, from nanotechnology to nuclear energy, the answer to handling risks is to assess them, manage them and mitigate them — always using rigorously tested scientific knowledge. After 45 years of scientific cooperation, the United States and China risk veering off course. It would be a dangerous folly to bring an end to research cooperation that has such potential to help meet the many challenges faced by China, the United States and the world. In 1979, scientists broke the ice at a time of great tension. As tensions rise once again, researchers could be the foot in the door that keeps communications open.

A personal take on science and society

World view

By Inna Makhovych

What I’ve learnt from teaching in Kyiv amid a war Ukraine’s universities have adapted by blending innovative forms of remote learning. Lessons from this experiment are relevant to the rest of the world.

INNA MAKHOVYCH

I

n September 2022, seven months after Russia invaded Ukraine, I enrolled in a doctoral degree programme at the National Academy of Educational Sciences in Kyiv. Like most Ukrainians, I assumed that the war would end in a few more months. But this week marked the second anniversary of the invasion. For two years now, Ukrainians have lived through power cuts and air-raid alerts. The academic year has had to stretch into the summer, when heating costs are lower. Despite these challenges, education and research have managed to keep going. And we’ve learnt a great deal about how to adapt our universities to constantly changing circumstances. As well as being a student, I am a senior lecturer at the Kyiv National University of Technologies and Design. Using a combination of online and offline methods, I teach English to more than 100 undergraduate students. Our interactions feed into my doctoral dissertation, which is on the use of gamification — incorporating game-like elements to increase participation — to motivate students and individualize the learning process. Ukraine is a live laboratory for testing innovations in education. Each student faces a different set of challenges. Some are not always able to attend in-person classes; many cannot fit a regular academic schedule into their lives. My classroom has become hyper-individualized as I try to cater to the differing needs of each student. Oleh, for instance (students’ names have been changed for privacy), arrived in Kyiv from his home town only at the tail end of the autumn semester, so he had to catch up on all the class material he’d missed. His parents were anxious about letting him move to the city after an intense spell of missile strikes on Kyiv during the summer. Another student, Ivan, has moved to Finland — he is among the estimated 650,000 men who have left Ukraine to escape the war. Owing to language barriers in their host countries, many exiled students continue to take classes at Ukrainian universities. Ivan sends me videos of his diction and pronunciation using messaging apps such as Telegram and Viber. Self-learning is another skill each student must develop. I regularly use gamified platforms such as Quizlet and Kahoot, which allow students to work on assignments at different times and at their own pace. I’ve created pods of students on these platforms and assigned them specific tasks. Working in a group environment online gives students a sense of a cohesive classroom, because they can compare their performance with that of their peers on the

Ukrainian academics are studying the effects of online education and a stressful environment on the quality of learning.”

Inna Makhovych is senior lecturer at the Kyiv National University of Technologies and Design and a PhD student at the National Academy of Educational Sciences in Kyiv, Ukraine. e-mail: innaumis@ gmail.com

leader board. We also engage in real-time collaborative play on Quizlet Live, which lets students showcase their language skills. Students collaborate in teams to achieve shared goals and compete against other teams. Ukraine’s academic adaptations are relevant to the world because learning apps and educational technology platforms have made big inroads globally in the aftermath of the COVID-19 pandemic. Serious research is needed into what works and what doesn’t. Ukrainian academics, including me, are trying to systematically study the effects of online education and a stressful environment on the quality of learning. We deserve support and offers of collaboration from institutions abroad. Research spending in Ukraine was in decline before the war, dropping from 0.7% of gross domestic product in 2011 to 0.3% in 2021, according to the World Bank. The war has stretched public finances even further. But it is important to sustain research and academic work even during conflict. Education can switch one’s attention away from anxiety and stress. Educational institutions offer a semblance of normalcy. They are spaces where young Ukrainians can engage with their peers in a safe environment. And students are very happy to be in a classroom. Take Katia, a fourth-year undergraduate in my class, originally from Avdiivka — a city that fell to the Russians this month. Katia lost her home and had to move to Kyiv with her mother. To help pay the rent, she has been juggling her university classes with a part-time job for a delivery company. Although she misses classes occasionally, she is unwilling to give up on education. Other students in my class have had to deal with the death of a close family member or extended periods of separation from their parents. Compared to their struggles, the difficulties I encounter as a doctoral student are modest. Because of the threat of power cuts, my house has a car battery rigged up as an emergency power source for the Internet router and the phones. In Kyiv, we have invented shorthand vocabulary to discuss the severity of air strikes. Prylit, which means arrival, is a way to say that a missile has evaded the air-defence system and reached its target. When this happens, we evacuate to a shelter. Otherwise, I stay at home and focus on my dissertation. Ukraine’s experience over the past two years provides a template for how to organize teaching and learning in the middle of a war. We have shown that education is possible in any situation. Although there seems to be no end in sight for the war, I have faith in the future. Many young people who have chosen to stay in Ukraine during this difficult period are so incredibly smart. Their brilliance is hard to miss in classroom discussions. I hope the world will invest in them and their future.

Nature | Vol 626 | 29 February 2024 | 929

Selections from the scientific literature

L TO R: YU ZHAO, HAITAO XU, YU ZHAO; M. AIESTARAN ET AL./ANTIQUITY (CC BY 4.0); SAIYNA BASHIR/REUTERS

Research highlights RECYCLING SYSTEM KEEPS EGG CELLS SQUEAKY CLEAN

HAND WRITING: RELIC POINTS TO ORIGINS OF BASQUE LANGUAGE

Mouse egg cells contain specialized ‘recycling bins’ that help to keep the cells tidy before fertilization. All cells must cope with misfolded and clumped proteins that can pollute the cellular interior, or cytoplasm. But this is a particular problem for immature egg cells, called oocytes, which must survive for long periods awaiting fertilization. To see how mouse eggs deal with their rubbish, Gabriele Zaffagnini at the Centre for Genomic Regulation in Barcelona, Spain, and his colleagues looked for protein clumps in oocytes as well as in mature eggs and developing embryos. In oocytes, the researchers discovered specialized structures containing protein clumps as well as cellular machinery that recycles proteins. The authors observed that this machinery kicks into gear when eggs mature, just before ovulation and fertilization. Blocking this process in oocytes led to defective eggs and, in earlystage embryos, to severely disrupted development. Chromosomal factors account for much of the agerelated decline in fertility, but defects in protein degradation could also have a role, the researchers say.

Inscriptions carved on a 2,100-year-old bronze hand might be the earliest written example of the language that gave rise to modern Basque. Basque — one the oldest living languages — is thought to be descended from a language spoken by the Vascones, an Iron-Age people who inhabited parts of northern Spain before the Romans arrived in the region in the first century BC. Mattin Aiestaran at the Aranzadi Science Society in Donostia– San Sebastián, Spain, and his colleagues analysed the symbols engraved on a hand-shaped bronze plate (pictured) that was unearthed at an ancient Vasconic village. The authors found that one of the inscribed words is similar to the Basque word zorioneko, which means ‘of good fortune’. The inscriptions on the hand, combined with the fact that it was probably designed to hang on a door, suggest that it was dedicated to a deity of fortune and used as a good-luck charm. Archaeologists have long thought that the Vascones lacked a writing system other than that used on coins, but the findings show that these ancestors of modern Basque people already knew and used writing in the first century BC.

Cell https://doi.org/mhdt (2024)

BUBBLE PRINTING: SOAP FILMS ETCHED WITH LASER PULSES Soap films — thin, delicate and seemingly a blink away from popping — can be engraved using lasers under the right conditions. Such films comprise thin layers of liquid sandwiched between walls of detergent molecules, or micelles. When a film is perturbed, any excess micelles in the liquid layer rush to reinforce the film’s walls, restoring its smooth surface. Haitao Xu and Yu Zhao at Tsinghua University in Beijing found that if they increased a moving soap film’s detergent concentration beyond a critical point, they could carve longlasting grooves into its surface using a laser. That’s because a film that has very high quantities of micelles has a vanishingly low elasticity, preventing its surface from recovering. The pulses created a series of pits in the film; these pits elongated as the film flowed. The elongated etchings, each less than a millimetre long, collectively resemble the dashed lines of road markings. By moving the laser laterally across the direction of flow, the researchers could inscribe wave-like patterns onto the film (pictured). The engravings were unaffected by swirling vortices created in the wake of an obstacle inserted into the film. Phys. Rev. Fluids 9, L022001 (2024)

Antiquity 397, 66–84 (2024)

THERAPY COURSE HOLDS POSTPARTUM DEPRESSION AT BAY One-to-one therapy sessions during pregnancy can slash the odds of a person developing postpartum depression or serious anxiety, a study of 755 women in Pakistan has found. People commonly experience mental-health disorders such as depression and anxiety in the weeks after giving birth, particularly in resource-poor settings. Pamela Surkan at the Johns Hopkins Bloomberg School of Public Health in Baltimore, Maryland, and her colleagues ran a randomized controlled trial to test whether cognitive behaviour therapy — in which counsellors challenge a person’s negative thoughts — can reduce an individual’s odds of developing these disorders when given during pregnancy. The team assigned 380 women to receive six one-to-one sessions with nonspecialist counsellors, and 375 women to receive standard care along with interventions to make that care easier to access. All participants had at least mild anxiety at the start of the study. Compared with the control group, those who received counselling were 81% less likely to be experiencing a major depressive episode and 74% less likely to have developed moderate-to-severe anxiety six weeks after giving birth. Nature Med. https://doi.org/ gthwnx (2024)

Nature | Vol 626 | 29 February 2024 | 931

The world this week

CHRISTOPH BURGSTEDT/SPL

News in focus

Tools to edit messenger RNA (artist’s illustration) are said to be safer than the CRISPR–Cas9 system, which changes the genome itself.

MOVE OVER, CRISPR: RNA-EDITING THERAPIES PICK UP STEAM Two RNA-editing therapies for genetic diseases have in the past few months gained approval for clinical trials, raising hopes for safer treatments. By Mariana Lenharo

R

NA editing is gaining momentum. After decades of basic research into how to alter this complex molecule, at least three therapies based on RNA editing have either entered clinical trials or received approval to do so. They are the first to reach this milestone. Proponents of RNA editing have long argued that it could be a safer, more flexible alternative to genome-editing techniques such as CRISPR, but it poses substantial technical problems. The launch of human trials signals the growing maturity and acceptance of the

field, scientists say. “There’s a much greater understanding of RNA technology, and that’s been partially enhanced by the RNA vaccine and the COVID pandemic,” says Andrew Lever, a biologist at the University of Cambridge, UK. “RNA is now seen as a very important therapeutic molecule.”

Temp job RNA has a key role in protein synthesis: the genetic information encoded in DNA is transcribed into messenger RNA (mRNA) and then translated into proteins. RNA molecules are composed of units called nucleotides, each containing one of four bases, or letters.

RNA-editing techniques aim to compensate for harmful mutations by changing the sequence of RNA, allowing normal proteins to be synthesized. RNA editing can also increase the production of beneficial proteins. Unlike CRISPR genome editing, RNA editing doesn’t change genes. Nor does it introduce permanent changes, because RNA molecules are transient. This means that the duration of the therapeutic effect could be shorter. But that transience could offer safety advantages. One risk of CRISPR therapies is off-target effects, or unintended changes outside the target genomic region, notes Joshua Rosenthal, a neurobiologist at the Nature | Vol 626 | 29 February 2024 | 933

News in focus Marine Biological Laboratory in Woods Hole, Massachusetts. “An off-target effect in DNA is potentially quite dangerous. In RNA, it’s less so, because it’s going to turn over.”

One letter at a time One common RNA-editing approach, singlebase editing, harnesses an enzyme that is already found in cells: adenosine deaminase acting on RNA (ADAR). This enzyme swaps a base called adenine in the RNA sequence for a base called an inosine. Wave Life Sciences in Cambridge, Massachusetts, is exploring single-base editing to treat a genetic disorder called alpha-1 antitrypsin deficiency (AATD), which can damage the lungs and the liver. The disease reduces the production of AAT, a protein made in liver cells that protects lungs from damage caused by inhaling polluted air or other irritants. Wave’s product is a short chain of nucleotides that directs naturally occurring ADAR enzymes to change a specific letter in each mRNA molecule to correct the mutation that affects AAT production. “By using the cell’s endogenous machinery to edit that single base, you now make a normal protein. And we’ve shown that the normal protein can be expressed at high levels,” says Paul Bolno, Wave’s president and chief executive. In mice, the drug edited around 50% of the target mRNA in liver cells, which is enough to produce therapeutic effects, Bolno says. The company’s clinical trial of the drug began last December in the United Kingdom and Australia, and will evaluate the drug’s safety and other features.

Editing whole paragraphs Another approach, called RNA exon editing, changes thousands of genetic letters in an RNA molecule at once, as opposed to changing just one letter. Exon editing is akin to editing a whole paragraph instead of correcting one typo, says Lever. This technology is particularly important for disorders caused by multiple mutations in a person’s genome; such arrays of mutations are difficult to address with single-base changes, he adds. The technique targets pre-mRNA, which is transcribed from DNA and then processed to make mRNA. Pre-mRNA includes both exons — parts of the RNA transcript that contain instructions for making proteins — and introns, which don’t contain such instructions. Through a mechanism called RNA splicing, the introns are cut out of the pre-mRNA, and the exons are stitched together to form the final mRNA, which is translated into protein. Companies such as Ascidian Therapeutics in Boston, Massachusetts, are leveraging the RNA-splicing process to remove mutation-containing exons and replace them with healthy ones. Last month, Ascidian received approval from the US Food and Drug 934 | Nature | Vol 626 | 29 February 2024

Administration for a clinical trial of an exon editor to treat Stargardt disease, which causes vision loss. People with the disease have several mutations in a single gene, leading to the production of a defective version of a protein that usually protects the retina. Ascidian’s therapy relies on an engineered DNA segment that is delivered into cells and produces normal RNA exons. These replace

“RNA is now seen as a very important therapeutic molecule.” the mutated ones during the splicing process, resulting in functional proteins. The DNA also produces RNA sequences that facilitate exon editing. “With one molecule, [the therapy] is able to replace 22 exons at one time,” says biologist Robert Bell, head of research at Ascidian.

Cancer-quashing RNA The potential of RNA-based therapies is not limited to genetic diseases. Rznomics, a biopharmaceutical company in Seongnam, South

Korea, is testing an RNA editor to treat hepatocellular carcinoma, the most common type of liver cancer. In September 2022, the company started a clinical trial in South Korea, which it intends to expand internationally. Rznomics’s approach involves mRNA splicing — but, unlike Ascidian’s method, it doesn’t use the cell’s own splicing machinery. Instead, the company co-opted a naturally occurring ribozyme, an RNA molecule that can induce splicing in target regions of mRNA. Researchers engineered the ribozymes to cut open mRNAs in tumour cells and insert a lethal cargo: an RNA sequence that is translated into a protein that generates a toxin that induces cell death. When surrounding cancer cells come into contact with these cells, the toxin spreads, promoting their death as well. The therapeutic molecule replaces an RNA sequence that is associated with tumour growth. The use of the splicing approach against more than one disease is very exciting, says Lever, who is also the chief medical officer of Spliceor in Cambridge, UK, a firm that is working on RNA-splicing therapies. “It opens up a whole new range of possibilities of treatment for things which otherwise can’t be treated.”

WHY VOLUNTEERS WILL GATHER DNA FROM HUNDREDS OF LAKES Massive environmental DNA project aims to take a record-setting snapshot of biodiversity worldwide. By Lydia Larsen

I

n a first-of-its-kind project, researchers are tapping into the power of citizen science to collect DNA samples from hundreds of lakes around the world. Not only will the resulting cache of environmental DNA (eDNA) be the largest ever gathered from an aquatic setting in a single day — it could also yield a fuller picture of the state of biodiversity around the globe and improve scientists’ understanding of how species move about over time. Scientists are increasingly using eDNA — which is shed by all organisms — to evaluate the presence of species in a given environment. Researchers have shown that it can be cheaply and efficiently extracted from water1, soil2, ice cores3 and filters from air-monitoring stations4. It has even been used to detect endangered species that haven’t been spotted for years, including a Brazilian frog species (putatively assigned to Megaelosia

bocainensis) that researchers had thought went extinct in the 1960s(ref. 5). Kristy Deiner, an environmental scientist at the Swiss Federal Institute of Technology (ETH) in Zurich who is leading the massive lake project, says that eDNA represents a “paradigm shift” in how scientists monitor biodiversity. Deiner’s research group has already received applications from more than 500 people across 101 countries to participate in collecting eDNA from their local lakes and shipping the samples to ETH Zurich. These global-scale projects are “really what the eDNA community needs”, says Philip Francis Thomsen, an environmental scientist at Aarhus University in Denmark and a volunteer for the lake project. “By involving citizens, we not only increase the geographical scope of our sampling but also foster a sense of public ownership and awareness regarding global biodiversity issues,” says Cátia Lúcio Pereira, the project’s coordinator,

K. DEINER

who works with Deiner at ETH Zurich. Although eDNA is generally considered to be a boon for biodiversity monitoring, researchers recognize that it’s not perfect. For instance, DNA from a particular site might come from a species that just briefly passed through the region, rather than living there. And researchers don’t have a clear understanding of how factors such as microbial ingestion of the DNA, high temperatures and ultraviolet radiation degrade the genetic material once it has been shed, or how those factors might alter the list of species detected. Deiner acknowledges the limitations, but says that eDNA-monitoring technology has come a long way since it was first used decades ago. She and her team have a plan to carefully handle the samples they receive, extract the genetic material and amplify the plant and animal DNA to detect the presence of species. “We’re more fine-tuning things now,” she says. Deiner also doesn’t necessarily see the transfer of eDNA from one region to another as a negative thing — she could even use it to her advantage. She began studying how eDNA moves in rivers about ten years ago. The genetic material, she suggests, could flow from soil, down rivers and into lakes, making these watery pools the ideal location to sample from to get an idea of the species diversity of an entire region, or catchment.

Local sampling Her project — called LeDNA, which stands for lake eDNA — aims to prove that the eDNA in a lake represents not just lake-dwelling species, but also terrestrial animals that live along the rivers that feed into the lake and around the lake itself. It will also examine the differences in species richness between geographical

SAMPLING SITES

Volunteers will gather environmental DNA from the world’s lakes for the LeDNA project.

regions, and try to decipher how species in various habitats might be interacting with one another. Deiner’s research group recruited volunteers for LeDNA through a combination of social media, networking with other eDNA researchers and reaching out to citizen-science groups. The recruits will be assigned a lake near them from a curated list of 5,000 around the globe. “We really worked hard to try and reach a lot of these areas so that the sample is truly a global effort,” Deiner says. Although the team hasn’t finalized the lakes that it will sample, it hopes to include about 800, says Lúcio Pereira (see ‘Sampling sites’). The researchers also say that they have mostly

The LeDNA, or lake eDNA, project has already recruited more than 500 citizen scientists from around the world to collect environmental DNA from their local lakes. Although the researchers behind the project have mostly finished recruiting, they would still like more volunteers in Asia, North Africa and the Middle East, to better assess the biodiversity in those regions. Potential sampling lakes

finished the recruitment, although they still want more volunteers in Asia, North Africa and the Middle East. Once assigned a lake, volunteers will receive instructions and a water-sampling filter. They will all aim to gather their samples on the same day — 22 May, which is the International Day for Biological Diversity — although there is a flexible two-week window for collection if they need it. Francis Thomsen points out that hundreds of people taking samples might lead to issues with data quality, depending on how closely they each follow the set protocols sent to them. Sampling eDNA, however, is easier to standardize than other biodiversity-monitoring methods, in which surveyors typically have to locate and identify individual species in person, he says. Lúcio Pereira says that the team recognizes the possible threat to data quality, but that the volunteers will all have identical kits and in-depth training on the sampling protocol. A perk of participating in the project, particularly for eDNA scientists, is that local partners will be able to use the data in their own research, as well as contribute to LeDNA publications. “What’s cool about this is it’s participatory,” says Rachel Meyer, director of the California eDNA programme, which is run by University of California researchers and matches volunteers with scientists for the collection of eDNA samples across the state. The data are there “if people want it”, she says, “and there’s plenty of incentive to want it”.

LEDNA

1. 2. 3. 4. 5.

Goldberg, C. S. et al. Methods Ecol. Evol. 7, 1299–1307 (2016). Allen, M. C. et al. Sci. Rep. 13, 180 (2023). Varotto, C. et al. Sci. Rep. 11, 1208 (2021). Littlefair, J. E. et al. Curr. Biol. 33, R426–R428 (2023). Lopes, C. M. et al. Mol. Ecol. 30, 3289–3298 (2021).

Nature | Vol 626 | 29 February 2024 | 935

An 1862 illustration of Megalosaurus, the first dinosaur to be named.

200 YEARS OF NAMING DINOSAURS: SCIENTISTS CALL FOR BETTER RULES Some palaeontologists want more rigorous guidelines for naming species. By Katharine Sanderson

I

t’s been 200 years since researchers named the first dinosaur: Megalosaurus. In the centuries since, hundreds of other dinosaur species have been discovered and catalogued — their names inspired by everything from their physical characteristics to the scientists who first described them. Now, some researchers are calling for the introduction of a more robust system, which they say would ensure species names are more inclusive and representative of where and how fossils are discovered. Megalosaurus was named by William Buckland, a geologist who discovered the enormous reptile’s fossilized remains in a field in Stonesfield, UK, in 1824. Buckland chose the name Megalosaurus on account of the immense size of the bones he and others had excavated. “It was a sensation — the first gigantic extinct land reptile ever discovered,” says Paul Barrett, a palaeontologist at the Natural History Museum in London. “Such an animal had never been conceived of before.” The word dinosaur — from the Greek meaning ‘fearfully great lizard’ — was introduced in 1841. 936 | Nature | Vol 626 | 29 February 2024

Unlike in other scientific disciplines — such as chemistry, in which strict rules govern a molecule’s name — zoologists have relatively free rein over the naming of new species. Usually, the scientist or group that first publishes work about an organism gets to pick its name, with few restrictions. There is a set of guidelines for

“The problem in terms of numbers is insignificant. But it is significant in terms of importance.” species naming overseen by the International Commission on Zoological Nomenclature (ICZN). These include the requirements that the name is unique, that it is announced in a publication and that, for dinosaurs, it is linked to a single specimen.

Problems abound To explore how dinosaur naming has changed over the past 200 years, Emma Dunne, a palaeobiologist at Friedrich-Alexander University in Erlangen–Nuremberg, Germany,

and her colleagues analysed the names of all of the dinosaur fossils from the Mesozoic Era (251.9 million to 66 million years ago) that have been described, around 1,500 in total. The authors wanted to know how much effort it would take to address what they saw as problematic names, which they describe as those “emanating racism, sexism, named under (neo)colonial contexts or after controversial figures”. They found several such names, equating to less than 3% of the dinosaurs they looked at. Some of the names the team identified derive from the colonial names for lands where species have been discovered. Indigenous-language names of places or researchers are often not used or are mistranslated, the authors say. For example, many of the dinosaurs discovered during a series of expeditions run between 1908 and 1920 by German explorers in Tendaguru in Tanzania, which was then part of German East Africa, were named after German people rather than local expedition members, and the samples remain in Germany. “The problem in terms of numbers is really insignificant. But it is significant in terms of importance,” says Evangelos Vlachos, a palaeontologist at the Museum of Paleontology Egidio Feruglio in Trelew, Argentina, who also worked on the study. He wants future naming systems to be more rigorous. “We don’t say that tomorrow we need to change everything. But we need to critically revise what we have done, see what we have done well and what we have not done well, and try to correct it in the future.” The use of eponyms — naming a species after a person or people — has become much more common in recent years, with just over half of names that are eponyms having been given in the past 20 years, the authors say. They found that in instances in which a species has a gendered name ending, the majority are masculine. They suggest that to avoid perpetuating stereotypes, names could focus on physical descriptions, such as Stegosaurus (from ‘roof lizard’ in Greek, referring to the animal’s platelike spines) or Triceratops (‘three-hornedface’). This also adds to the usefulness of the name for communication, they say. The team’s analysis has not yet been published or peer-reviewed.

No name changes The ICZN is firmly against going back and renaming species whose names might now be considered offensive, and would not consider banning eponyms, says ICZN president Thomas Pape, a taxonomist at the Natural History Museum of Denmark in Copenhagen. “We do not recommend renaming unless there are what we would call formal nomenclatural reasons,” he adds. This is because the organization places great importance on preserving

PAUL D. STEWART/SPL

News in focus

the ‘stability’ of names, and this could be threatened if they are changed retrospectively, he says. The ICZN would consider introducing different naming systems, Pape says, perhaps including repositories for names to be peerreviewed, or insisting that names can be considered official only if they are first published in a certain set of journals. But no formal changes are currently planned. Meanwhile, Barrett says, some palaeontologists are trying to drive change in the community. “There’s been a marked change in the desire to credit formerly overlooked figures when naming new dinosaurs and to ensure that issues of patrimony are faced and accounted

for,” he says. He adds that Indigenous collaborators and colleagues are now more often recognized, “whereas previously most eponyms reflected the roles of scientists in the global north”. Many researchers also try to use names derived from the languages, interests and traditions in countries where dinosaur remains are discovered, helping to foster community engagement and to reflect the historical context of the material. Dunne says that although she would like to see change, she doesn’t want to add further unpaid work to the burdens facing academics. “But there does need to be something,” she says, adding that the ICZN “could do better and be more representative of the community”.

THE LIFE AND DEATH OF A BOG MAN REVEALED AFTER 5,000 YEARS Vittrup Man was a Scandinavian wanderer who settled down between 3300 and 3100 BC. By Ewen Callaway

STEPHEN FREIHEIT VIA A. FISCHER ET AL./PLOS ONE

B

efore he was bludgeoned to death and left in a Danish bog, an ancient individual now known as Vittrup Man was an emblem of past and future ways of living. He was born more than 5,000 years ago into a community of Mesolithic hunter-gatherers who probably lived in northern Scandinavia, as their ancestors had for millennia. But Vittrup Man spent his adult life across the sea in Denmark among farming communities, whose ancestors came from the Middle East. It’s impossible to know the lives that Vittrup Man touched during his lifetime; it was his death that caught people’s imagination thousands of years later. His remains — ankle and shin bones, a jawbone and a skull fractured by at least eight heavy blows — were discovered in the early twentieth century in a peat bog near a town called Vittrup in northern Denmark, alongside a wooden club that was probably the murder weapon. His “unusually violent” death distinguished Vittrup Man from other similarly aged remains found in bogs, says Karl-Göran Sjögren, an archaeologist at the University of Gothenburg, Sweden, who co-led a team that charted Vittrup Man’s life in a study this month (A. Fischer et al. PLoS ONE 19, e0297032; 2024). But nothing else about Vittrup Man stood out until researchers examined his ancestry for a study that came out earlier this year

(M. E. Allentoft et al. Nature 625, 301–311; 2024). Vittrup Man, they learnt, was related to hunter-gatherers from what is now Norway and Sweden, and not to the farming communities with Middle Eastern roots that had arrived in Denmark hundreds of years before his death. “This is an indication that his origin may be a bit further north,” says Sjögren, possibly near the Arctic Circle where people still lived by fishing, hunting and gathering. Carbon and nitrogen isotope levels in bones and teeth, which can reveal aspects of diet, suggest that Vittrup Man got his calories from the ocean as a child,

Vittrup Man’s skull was shattered by blows.

before transitioning to freshwater fish and wild game as a teenager and a diet including cereals, dairy and meat typical of farming communities starting as a young adult. Incorporated into his teeth, the researchers found protein fragments from seals, whales and fish as well as sheep or goats. A childhood among northern Scandinavian hunter-fisher-gatherers might have prepared Vittrup Man for a long open-sea voyage to Denmark. What’s not clear is why he left the familiar to live among farmers. Some archaeologists, including some of Sjögren’s co-authors, surmise that Vittrup Man was taken captive and enslaved before being killed — a fate not uncommon in early Neolithic Scandinavia, when numerous social groups coexisted. Sjögren favours the idea that Vittrup Man lived like a foreign merchant, mediating the exchange of goods between farmers and hunter-gatherers. Flint axes made of high-quality Danish stone, which have been identified along the Norwegian coast, could have been traded for materials from northern Scandinavia such as basalt. “Maybe once he came of age, his role in society was to establish connections with farmers that lived across the sea,” says Thomas Booth, a bioarchaeologist at the Francis Crick Institute in London. He lived with the farmers for the last decades of his life, but it’s not inconceivable that he voyaged back and forth between homes old and new, adds Sjögren.

Ritual sacrifice? What, then, of Vittrup Man’s violent death, probably in his early thirties? Dozens of Neolithic human remains — many of them young males, like Vittrup Man — have been discovered in bogs, and archaeologists think that ritual sacrifice explains many of these deaths. These people often had bone malformations that would have marked them out among their peers, but not Vittrup Man, says Sjögren. Genome analysis suggests that Vittrup Man was blue-eyed and his skin might have been darker than that of typical Neolithic farmers, but his dark hair colour and height wouldn’t have stood out. “Why they chose to sacrifice some people, it’s really hard to say,” says Sjögren. Vittrup Man’s hunter-gatherer ancestry more or less vanished from Scandinavian in the centuries after his death, and it’s not clear if any close relatives survived him. Researchers sequencing ancient human genomes by the hundreds have begun to build genealogies of ancient families, and it’s not inconceivable that a relative could one day be found. The life — and death — of Vittrup Man goes to the heart of one Europe’s biggest transitions, says Booth, when hunter-gathering communities like his sat on the edge of a new way of life. “It gives you a sense of the worlds that these people are inhabiting.” Nature | Vol 626 | 29 February 2024 | 937

News in focus

What the EU’s tough AI law means for research and ChatGPT

European Union countries are poised to adopt the world’s first comprehensive set of laws to regulate artificial intelligence (AI). The EU AI Act puts its toughest rules on the riskiest AI models, and is designed to ensure that AI systems are safe and respect fundamental rights and EU values. “The act is enormously consequential, in terms of shaping how we think about AI regulation and setting a precedent,” says Rishi Bommasani, who researches the societal impact of AI at Stanford University in California. The legislation comes as AI develops apace. This year is expected to see the launch of new versions of generative AI models — such as GPT, which powers ChatGPT, developed by OpenAI in San Francisco, California — and existing systems are being used in scams and to propagate misinformation. China already has a patchwork of laws to guide commercial use of AI, and US regulation is under way: last October, President Joe Biden signed the nation’s first AI executive order, requiring federal agencies to take action to manage the risks of AI. EU nations’ governments approved the legislation on 2 February, and the law now needs final sign-off from the European Parliament, one of the EU’s three legislative branches; this is expected to happen in April. If the text remains unchanged, as policy watchers expect, the law will enter into force in 2026. Some researchers have welcomed the act for its potential to encourage open science, whereas others worry that it could stifle innovation. Nature examines how the law will affect research. What is the EU’s approach? The EU has chosen to regulate AI models on the basis of their potential risk, by applying stricter rules to riskier applications and outlining separate regulations for general-purpose AI models, such as GPT, which have broad and unpredictable uses. The law bans AI systems that carry

938 | Nature | Vol 626 | 29 February 2024

JEAN-FRANCOIS BADIAS/AP VIA ALAMY

The EU AI Act is the world’s first major legislation on artificial intelligence and strictly regulates general-purpose models.

The European Parliament must give the final green light to the law. A vote is expected in April. ‘unacceptable risk’, for example those that use biometric data to infer sensitive characteristics, such as people’s sexual orientation. High-risk applications, such as using AI in hiring and law enforcement, must fulfil certain obligations. For example, developers must show that their models are safe, transparent and explainable to users, and that they adhere to privacy regulations and do not discriminate. For lower-risk AI tools, developers will still have to tell users when they are interacting with AI-generated content. The law applies to models operating in the EU, and any firm that violates the rules risks a fine of up to 7% of its annual global profits. “I think it’s a good approach,” says Dirk Hovy, a computer scientist at Bocconi University in Milan, Italy. AI has quickly become powerful and ubiquitous, he says. “Putting a framework up to guide its use and development makes absolute sense.” Some don’t think the laws go far enough, leaving “gaping” exemptions for military and national-security purposes, as well as loopholes for AI use in law enforcement and migration, says Kilian Vieth-Ditlmann, a

political scientist at AlgorithmWatch, a Berlinbased non-profit organization that studies the effects of automation on society. How much will it affect researchers? In theory, very little. Last year, the European Parliament added a clause to the draft act that would exempt AI models developed purely for research, development or prototyping. The EU has worked hard to make sure that the act doesn’t affect research negatively, says Joanna Bryson, who studies AI and its regulation at the Hertie School in Berlin. “They really don’t want to cut off innovation, so I’d be astounded if this is going to be a problem.” But the act is still likely to have an effect, by making researchers think about transparency, how they report on their models and potential biases, says Hovy. “I think it will filter down and foster good practice,” he says. Robert Kaczmarczyk, a physician at the Technical University of Munich in Germany and co-founder of LAION (Large-scale Artificial Intelligence Open Network), a non-profit organization aimed at democratizing machine learning, worries that the law could hinder small companies that drive research, and

News explainer

which might need to establish internal structures to adhere to the laws. “To adapt as a small company is really hard,” he says. What does it mean for powerful models such as GPT? After heated debate, policymakers chose to regulate powerful general-purpose models — such as the generative models that create images, code and video — in their own two-tier category. The first tier covers all general-purpose models, except those used only in research or published under an open-source licence. These will be subject to transparency requirements, including detailing their training methodologies and energy consumption, and must show that they respect copyright laws. The second, much stricter, tier will cover general-purpose models deemed to have “high-impact capabilities”, which pose a higher “systemic risk”. These models will be subject to “some pretty significant obligations”, says Bommasani, including stringent safety testing and cybersecurity checks. Developers will be made to release details of their architecture and data sources. For the EU, ‘big’ effectively equals dangerous: any model that uses more than 1025 FLOPs (the number of computer operations) in training qualifies as high impact. Training a model with that amount of computing power costs between US$50 million and $100 million — so it is a high bar, says Bommasani. It should capture models such as GPT-4, OpenAI’s current model, and could include future iterations of Meta’s open-source rival, LLaMA. Open-source models in this tier are subject to regulation, although research-only models are exempt. Some scientists are against regulating AI models, preferring to focus on how they’re used. “Smarter and more capable does not mean more harm,” says Jenia Jitsev, an AI researcher at the Jülich Supercomputing Centre in Germany and another co-founder of LAION. Basing regulation on any measure of capability has no scientific basis, adds Jitsev. They use the analogy of defining as dangerous all chemistry that uses a certain number of person-hours. “It’s as unproductive as this.” By Elizabeth Gibney

JUST 5 WOMEN HAVE WON A TOP MATHS PRIZE IN THE PAST 90 YEARS Prestigious awards such as the Fields Medal and Abel Prize have been awarded predominantly to men. By Sarah Wild

O

n 30 January, mathematician Claire Voisin became the first woman to win the Crafoord Prize — one of the pre-eminent awards in mathematics. The win highlights an ongoing issue in the discipline: a lack of gender diversity among the winners of its most prestigious awards. Voisin is one of just 5 women to have won a top maths prize in the past 90 years — all in the past decade (see ‘Mostly men’). “Awards are one mechanism by which work and thinkers are promoted in the broader community,” says Kathryn Leonard, a mathematician at Occidental College in Los Angeles, California, and former president of the

Association for Women in Mathematics (AMS), based in Providence, Rhode Island. “If women and people from other excluded groups continue to be excluded, their work is not being celebrated and shared.” Six of the world’s top maths honours — the Fields Medal and the Abel, Shaw, Wolf, Crafoord and Breakthrough prizes — have been awarded a total of 217 times, but only 7  times to women. Two women, Voisin and Maryam Mirzakhani, have claimed two of those awards each. Voisin shared the Shaw Prize in 2017; Mirzakhani won the Fields Medal in 2014, and was posthumously awarded the  Breakthrough Prize in 2020, mainly for her theoretical work in understanding the symmetry of curved surfaces.

MOSTLY MEN

The most prestigious mathematics prizes have been awarded to women just seven times — all in the past decade.

Women Men

1

10

20

30

1935–39 1940–44 1945–49 1950–54 1955–59 1960–64 1965–69 1970–74 1975–79 1980–84 1985–89 1990–94 1995–99 2000–04 2005–09 2010–14

28

1

2015–19

35

2

2020–24

30

4

Abel Prize

Breakthrough Prize

Crafoord Prize

Fields Medal

Shaw Prize

Wolf Prize

Total

Nature | Vol 626 | 29 February 2024 | 939

News in focus

Gender gap “The situation has improved tremendously during my professional lifetime,” Leonard acknowledges. But she adds that, in an ideal world, award winners should represent the community they are part of. It is difficult to gauge how many women earn maths degrees each year, or to assess female representation in mathematics professions worldwide. The International Mathematical Union (IMU), a global body based in Berlin that promotes cooperation in maths, has member organizations in more than 80 countries, and does not keep statistics on its member

demographics. According to a 2018 survey by the AMS (see go.nature.com/49jvrvh2), women account for between 25% and 30% of mathematics PhDs in the United States. “While there has been increasing awareness about the gender gap and important progress in recent years, some aspects remain unchanged,” says Carolina Araujo, chair of the IMU’s Committee for Women in Mathematics. “Statistics show that, while there has been a steady increase of the proportion of women authors of scientific papers in mathematics in the last decades, the proportion of women authors in ‘top journals’ in mathematics remains below 10%.” Ways to close this gender gap include actively promoting the visibility of female researchers and diversifying committees that make awarding decisions, Araujo adds. Another way, which might help people who take career breaks to care for children, would be to add 18 months per child to age limits where these apply to awards, she says.

MEGA-CRISPR TOOL GIVES A POWER BOOST TO CANCER-FIGHTING CELLS A system that edits RNA rather than DNA can give new life to exhausted CAR T cells. By Sara Reardon

T

he CRISPR–Cas9 gene-editing system excels at altering and disrupting genes. But the changes it makes are permanent, which can be a big problem if the system goes awry. Now, a CRISPR-based system that targets a cell’s short-lived messenger RNA instead of its DNA could provide a more precise and reversible way of designing cell therapies — and even help scientists to discover how different genes work together. The results were published on 21 February (V. Tieu et al. Cell https://doi.org/ mhp4; 2024).

tumour cells. But DNA-editing CRISPR systems can pose safety problems and are relatively inefficient in these cells. Bioengineer Stanley Qi and immunologist Crystal Mackall, both at Stanford University in California, and their colleagues developed an alternative system, called MEGA (multiplexed effector guide arrays). It uses CRISPR guide RNA but swaps the DNA-cutting Cas9 for the

RNA gets its turn Engineered CRISPR systems generally have two main components: a DNA-cutting enzyme, often Cas9, and a piece of ‘guide’ RNA that directs the enzyme to the stretch of DNA to be edited. One of the system’s most promising medical applications has been its potential use in producing chimeric antigen receptor (CAR) T cells. These are made by engineering the immune foot soldiers called T cells to attack specific proteins on the surfaces of 940 | Nature | Vol 626 | 29 February 2024

Engineered T cells (red) on cancer cells.

RNA-cutting enzyme Cas13d. The CRISPR half of the duo directs Cas13d to a target mRNA, which is produced from a DNA template. “We are not really touching any DNA,” Qi says. This avoids the risk of inducing permanent changes or, worse, cutting DNA in places other than the designated target. The mRNA doesn’t last long in a cell, so any mistakes will quickly disappear. Active cells such as T cells produce a constantly changing variety of mRNA molecules, each directing the production of a specific protein. Cas13d cuts the target mRNA, destroying it and preventing it from churning out its specific protein. This has the same effect as turning off the associated gene. MEGA allowed the researchers to create ‘multiplex’ CRISPR–Cas13d systems that can shut down the production of multiple proteins, effectively turning off up to ten genes at a time.

Rejuvenating exhausted cells The team used the system to address a phenomenon called T-cell exhaustion. If CAR T cells are activated too many times by a longterm tumour, they become less effective. To give a jolt to tired T cells, the researchers designed CRISPR systems that target mRNA molecules involved in functions including energy production and sugar metabolism. T cells treated with some MEGA combinations stopped expressing molecular signals of exhaustion and became better at shrinking tumours in mice. Qi, Mackall and their colleagues also created a version of Cas13d that is switched on only when the CAR T cells are treated with the antibiotic trimethoprim. By varying the doses of trimethoprim, the researchers could ‘tune’ mRNA levels up and down, giving the team precise control over when and how molecular pathways were activated, rather than just shutting them down. “It’s always thrilling to see how the RNA CRISPR toolbox is applied,” says systems biologist Jonathan Gootenberg at the Massachusetts Institute of Technology in Cambridge. The ability to tune the collection of RNA transcripts, he says, will be especially useful for cell therapies. Joseph Fraietta, an immunologist at the University of Pennsylvania in Philadelphia, agrees. In his experience with CRISPR, he says, his group can edit only about three genes in CAR T cells at a time before the cells become unhealthy. “This will open more avenues,” he says. But he cautions that the system requires continuously high levels of Cas13d, which might trigger an immune response. Mackall and Qi say that MEGA’s ability to tune gene expression allows scientists to vary the levels of a wide array of mRNAs at one time, revealing how different amounts of mRNA from various combinations of genes work together to carry out cellular functions.

STEVE GSCHMEISSNER/SPL

In 2023, physicist and mathematician Ingrid Daubechies, best known for using wavelets in image compression, received the Wolf Prize. Karen Keskulla Uhlenbeck secured the Abel Prize in 2019 for her pioneering efforts in modern geometric analysis, and number theorist Maryna Viazovska won the Fields Medal in 2022.

SPL

Feature

The bacterium Neisseria gonorrhoeae infects nearly 100 million people worldwide each year.

WHAT A TREATMENT FOR ‘SUPER GONORRHOEA’ MEANS FOR FUTURE DRUG DEVELOPMENT

Effective, affordable antimicrobial drugs aren’t moneymakers. Can non-profit organizations pick up the slack? By Maryn McKenna 942 | Nature | Vol 626 | 29 February 2024

SOURCE: D. THOMAS & C. WESSEL THE STATE OF INNOVATION IN ANTIBACTERIAL THERAPEUTICS (BIO, 2022)

WE NEED MORE ANTIBIOTICS, AND WE NEED TO EXPLOIT ALL THE DIFFERENT ROUTES AVAILABLE.”

L

ast November, a clinical trial offered a glimmer of hope in the often gloomy fight against antimicrobial resistance. An oral antibiotic, called zoliflodacin, was shown to be effective against the bacterium that causes the sexually transmitted disease gonorrhoea. And because it is the first of a new class of antibiotics, the drug also offers hope of stopping the spread of ‘super gonorrhoea’,which is resistant to most standard treatments. That month brought good news on another front, too. An international research team reported that a new antifungal drug, fosravuconazole, was safe and effective at treating a devastating disease called fungal mycetoma, which scars and damages the skin and can lead to amputation if left untreated. Antifungal drugs are difficult to develop and are scarce. The only existing mycetoma treatment requires taking expensive pills for a long time:

These organizations hope to fill a big gap in the development and testing of drugs at a time when most legacy pharmaceutical firms have withdrawn from antimicrobial drug discovery, and many of the small biotechnology companies that picked up the torch have gone bankrupt (see ‘Stagnant investment’). These two latest achievements suggest that non-profits could help to solve the problem of drug access, while fending off the rise of drug-resistant microbes, which contribute to almost five million deaths per year. “For someone like me, a clinician on the front lines, this is good news,” says Helen Boucher, an infectious-diseases physician and dean of the Tufts University School of Medicine in Boston, Massachusetts, who has testified before the US Congress about the challenge of finding such drugs. “We need more antibiotics, and we need to exploit all the different routes available to develop them.”

Pass-along value Both drugs followed a complicated path to the market. They were originally developed by conventional pharmaceutical companies: fosravuconazole by Eisai in Tokyo, and zoliflodacin by Entasis Therapeutics in Waltham, Massachusetts, which is now part of Innoviva in Burlingame, California. Under the agreements these companies have with the non-profit organizations, the original firms retain some rights to the drugs, either for manufacturing and distribution or for sales in

some high-income countries. But the goal has been to get both of these drugs to low-income countries at affordable prices. For the antifungal, fosravuconazole, that is already happening. On the basis of the trial results reported last year, Sudan’s Ministry of Health has allowed people to receive the drug while the country’s National Medicines and Poisons Board evaluates its registration. The trial, which recruited 104 people in Sudan between 2017 and 2020, compared treatment using fosravuconazole over 12 months with the standard of care, a drug called itraconazole (see go.nature.com/3t635ro). The trial found no statistically significant clinical difference, meaning that treatment for mycetoma could be made in a less expensive and less complex way. “Most of the itraconazole that is available in the African endemic countries is very expensive; it can cost over US$2,000 per year,” says Borna Nyaoke-Anoke, a Nairobi-based physician and head of the DNDi’s mycetoma disease programme, who managed the trial. “We’re looking at a neglected tropical disease affecting the most vulnerable and neglected patients, who are barely able to make $1 a day.” Eisai first developed ravuconazole, the predecessor to fosravuconazole, as part of a search for treatments for skin infections. But in 2003, a Venezuelan research project found that the drug was effective in mice against Chagas disease1, a parasitic infection that affects several million people in Latin America. The study caught the attention of the DNDi, which was searching for treatments to address the neglected disease at the time. Eisai and the DNDi jointly launched a phase II trial in Bolivia to explore the drug’s efficacy against Chagas disease. Although it was unsuccessful, an unrelated research project at the Mycetoma Research Center at the University of Khartoum in Sudan found that ravuconazole was effective against fungal mycetoma2. This offered the compound a second chance at approval, and validated a strategy

STAGNANT INVESTMENT

US companies developing antibiotic drugs have seen relatively flat investment in the past decade, particularly compared with companies working on cancer drugs. Cancer drugs

Antibiotics

7 Venture investment (US$ billions)

two pills per dose, twice per day, for several months. The new drug requires only taking two pills once a week, which could reduce stress and expense for tens of thousands of people in South Asia and sub-Saharan Africa. What is especially notable about the success of these two drugs, however, is that they followed a new path to get this far. Both trials were conducted by non-profit organizations that were founded specifically to bring new drugs to the market: zoliflodacin by the Global Antibiotic Research and Development Partnership (GARDP) based in Geneva, Switzerland, and fosravuconazole by the Drugs for Neglected Diseases Initiative (DNDi), also based in Geneva.

6 5 4 3 2 1 0 2011

2013

Nature | Vol 626 | 29 February 2024 | 943

2015

2017

2019

ANTIMICROBIAL MARKET FAILURE

In 2022, there were 64 antimicrobial drugs in clinical testing, of which only 8 were being developed by large companies (more than US$1 billion in revenue). Smaller companies have been pursuing these kinds of drug more rigorously, but many have gone bankrupt in the process. Small companies Drugs in development: 51 Large companies 8 Non-profit organizations 5

that the DNDi has pursued since 2003, when the organization was founded by members of Médecins Sans Frontières (MSF, also known as Doctors Without Borders): breathing life into drugs that have been abandoned or underused. “We call it repurposing: taking a drug that was developed for one indication, that was sort of left on the shelf for whatever reason, that we then repurpose for another indication,” says Rachel Cohen, the DNDi’s senior adviser for global policy advocacy and access. (The organization has since developed its own drug discovery programme, but it does not operate laboratories or manufacturing facilities.) The development of zoliflodacin was bumpier; its route to market encapsulates the decline of antibiotics research. The compound was first identified at Pharmacia, a pharmaceutical company based in Kalamazoo, Michigan. Pharmacia was acquired by Pfizer, based in New York City, in 2003. When Pfizer exited antibiotics research in 2011, zoliflodacin went to AstraZeneca, a drug firm in Cambridge, UK. Then, when AstraZeneca began to withdraw from antibiotic research, the drug ended up at the newly founded Entasis in 2015. Then the US National Institute of Allergy and Infectious Diseases funded a clinical phase II trial, which tested the safety and efficacy of zoliflodacin in 179 people3. Entasis also managed to acquire an agreement from the US Food and Drug Administration (FDA) that will earn it regulatory approval once the company has completed one successful phase III trial. The FDA typically requires two such trials before granting approval. During this time, the incidence of gonorrhoea was rising around the world3; in the United States, it increased by two-thirds between 2013 and 2017. And the bacterium was steadily gaining resistance to the several families of antibiotics used to cure it, turning what had been an easily handled infection into an almost untreatable menace. The World Health Organization (WHO) was also turning its attention to the problem of antibiotic resistance; it proposed a global action plan that was adopted in 2015. The following year, the WHO helped to create the GARDP. Funded with €2 million (US$2.16 million) from several governments and MSF, the 944 | Nature | Vol 626 | 29 February 2024

GARDP started with administrative support from the DNDi and followed its model of identifying low-cost opportunities to spin off useful drugs. “DNDi is set up to focus on neglected diseases, and we were set up to address antimicrobial resistance,” says microbiologist Laura Piddock, who is the GARDP’s scientific director. In 2019, the GARDP became an independent organization with a focus on specific bacterial infections in low-income nations that lack full access to new drugs. At around the same time, Entasis decided to focus its limited funding on a different drug from its AstraZeneca programme, a new combination antibiotic aimed at drug-resistant pneumonia. “We were a small biotech at the time, and our focus was the US and Europe,” says John Mueller, who was one of Entasis’s founders and is now chief development officer at Innoviva Specialty Therapeutics, a subsidiary of Innoviva. (Innoviva bought Entasis in 2022.) “GARDP’s interest was low- and middle-income countries, and those usually are later down in your commercialization path,” Mueller adds. Innoviva licensed zoliflodacin to the GARDP,

WE WERE SET UP TO ADDRESS ANTIMICROBIAL RESISTANCE.” and agreed to collaborate with the non-profit organization on the phase III trial — which expanded to 16 sites in 5 countries — and the subsequent FDA registration.

Push and pull As large pharmaceutical companies continue to back out of the antibiotics business, drug developers have been trying to stimulate public conversation about creating better incentives for small biotechnology firms (see ‘Antimicrobial market failure’). There are two basic strategies: push and pull. Push incentives support research by getting new compounds through trials and up to the point of approval; pull incentives take over afterwards, supporting companies so that they can survive until their earnings start to flow. Push incentives have had some success. The best known example is the non-profit organization CARB-X (Combating Antibiotic-Resistant Bacteria Biopharmaceutical Accelerator), based at Boston University, which has committed US$452 million to early-stage research, contributed by governments and big philanthropic organizations.

Pull incentives, however, have faced political headwinds. Although many options have been debated, only one has launched: a subscription-style arrangement in which drug companies receive grants that constitute advance down payments on future sales. The idea is that by removing the incentive for a company to sell a new drug aggressively, and thereby reducing the use of the drug, it will delay the inevitable arrival of resistance to it. Only one entity has committed to that subscription programme so far: the UK government. In 2022, the UK National Institute for Health and Care Excellence (NICE) decreed that the National Health Service (NHS) could justify paying £10 million (US$12.6 million) per year to two companies for future purchases of two drugs that will be used against severely drug-resistant, hospital-acquired infections. In mid-2023, the NHS endorsed the concept and announced plans to expand it, potentially raising the payments to £20 million. By contrast, in the United States, a piece of legislation known as the PASTEUR Act, which would create a similar programme, has not made it through Congress despite repeated tries. But, by funding the registration and launch of drugs in countries that cannot afford to pay commercial prices for them, both the DNDi and the GARDP are effectively providing pull incentives — which is something that these organizations can afford to do because, as non-profits with external funders, they do not need to earn the level of income that a company would require. That means they are providing a model not only for how muchneeded drugs can be rescued in the development process, but also for how companies can be supported when their compounds are released onto the market. Specialists who have been watching this field for a while say that it’s important not to forget the part that pharmaceutical companies have played in developing fosravuconazole and zoliflodacin, which both emerged from conventional discovery programmes. Drug discovery is extremely expensive; one estimate from 2016 calculates the cost from the initial identification of a compound through to FDA approval at $1.4 billion. Although the DNDi and the GARDP have each gathered millions of dollars from funders, neither has had to bear that kind of bill. That makes the non-profits “a crucial add-on”, says Kevin Outterson, the executive director of CARB-X. “What GARDP and DNDi do is unique in the world and absolutely necessary,” he adds. But, he says, they’re “not a replacement”. Maryn McKenna is an independent journalist based in Atlanta, Georgia. 1.

Urbina, J. A., Payares, G., Sanoja, C., Lira, R. & Romanha, A. J. Int. J. Antimicrob. Agents 21, 27–38 (2003). 2. Ahmed, S. A. et al. PLoS Negl. Trop. Dis. 8, e2942 (2014). 3. Taylor, S. N. et al. N. Engl. J. Med. 379, 1835–1845 (2018).

SOURCE: D. THOMAS & C. WESSEL THE STATE OF INNOVATION IN ANTIBACTERIAL THERAPEUTICS (BIO, 2022)

Feature

Science in culture

RAPP HALOUR/ALAMY

Books & arts

Indigenous peoples of the Americas saw all animals, including humans, as interconnected.

How our love of pets grew from a clash of world views Indigenous Americans’ relationships with and knowledge of animals have influenced how Europeans have thought about animals since 1492. By Surekha Davies

I

t’s an enduring myth that the development of livestock husbandry is an essential step on the path of human progress. Many books have emphasized its importance, including historian Alfred W. Crosby’s Columbian Exchange (1972) and geographer Jared Diamond’s Guns, Germs and Steel (1997). But not all cultures have seen animals as creatures to be penned and farmed. Indigenous peoples in the Americas, for example, recognized that humans and animals have much in common. The clash between differing views of human–animal relationships still resonates today.

In The Tame and the Wild, historian Marcy Norton explores the history and lasting importance of this clash, which began in the late fifteenth century when Europeans arrived on The Tame and the Wild: People and Animals after 1492 Marcy Norton Harvard Univ. Press (2024)

the shores of the Americas, including in the Caribbean. Norton draws on a rich array of sources, including treatises on hunting and natural history, Indigenous books (known as amoxtli), accounts from soldiers and missionaries, trial records from the Spanish Inquisition, dictionaries and paintings. Her fascinating and scholarly account reveals how these encounters transformed Europe and the Americas. Relationships between humans and animals that emerged from these meetings of different peoples planted the seeds of many of today’s ethical and environmental challenges  — from colonial wealth and the dispossession

Nature | Vol 626 | 29 February 2024 | 945

Books & arts of Indigenous peoples to the modern meat industry. They even explain people’s bonds with their pets.

Norton analyses human–animal relationships in Europe, Greater Amazonia (the Caribbean and lowland South America) and Mesoamerica. Many Indigenous peoples of the Americas consider all beings to be interconnected and permeable. By attempting to think like the animals they hunted, and by wearing the creatures’ pelts and consuming the meat, Indigenous people could take on some of the “beauty and power” of these living beings. By contrast, in Judeo-Christian thought, humans are distinct from and superior to animals. Norton identifies four ways in which people interacted with animals. In Europe, through hunting and husbandry, and in the Americas, through predation and ‘familiarization’ — a process of feeding and taming individual animals that came and went freely. Familiarized animals were never eaten in Greater Amazonia, but were sometimes consumed during Mesoamerican rituals. Each way of life shaped how people categorized animals and the extent to which animals were considered “fellow subjects with desires, emotions, and even reason”. In Europe, hunters distinguished vassal animals, such as hunting dogs, horses and falcons, from prey animals, particularly deer and boar. Nonetheless, for a hunt to be successful, hunters had to recognize that their prey had minds with needs, feelings, experiences and motives. Killing prey animals did not require their objectification. But the Christian view of a human–animal divide did provide a basis for livestock husbandry, a practice that requires animals to be seen merely as objects. The importance of husbandry went beyond nutrition, livelihoods and products, such as clothing. By objectifying animals, people created a “distance between those who owned and managed living animals and those who taxed and consumed their corpses”. With the rise of slaughterhouses — separate from butcher’s shops and required to be outside city limits — consumers in the fifteenth century were disconnected from animal rearing and killing. But Europeans weren’t able to distance themselves entirely from livestock husbandry during the fifteenth and early sixteenth centuries. The suspicion that animals might have thoughts and feelings beyond those of hunger or sleep found an outlet in fears about devils and witchcraft. Theologians and inquisitors associated animal features, such as horns, hooves, claws and tails, with the Devil and his human servants, witches. Paintings of Hell depicted the carnage associated with livestock husbandry: reptilian demons led people to slaughter or cut their victims up before roasting them on a spit or boiling them in a cauldron. Witches were considered to have 946 | Nature | Vol 626 | 29 February 2024

ALAMY

Food and fear

Animal features, such as horns, were associated with the Devil in the fifteenth century.

unnaturally close relationships with animals, to engage in bestiality and to have powers — such as the ability to fly — associated with animals. Well into the seventeenth century, people who had caring relationships with animals that did not have a ‘job’ raised suspicions of witchcraft. Such suspicions led to the execution of some 50,000 people in Europe between the late fifteenth and late eighteenth century. Missionaries, some of whom had persecuted alleged witches in Spain, carried “their predisposition to see idolaters as shape-shifting sorcerers” to Mesoamerica. Faced with what they saw as an idolatrous culture that “did not uphold a species divide”, they interpreted

ritual specialists or “knowledge manipulators” as witches. In so doing, they invented the colonial concept of the nahual, or animal double, out of the Indigenous concept of the nahualli, or knowledge manipulator. Colonial inquisitors tried Indigenous people suspected of having an animal form as sorcerers.

Global links In contrast to Christians, the peoples of Greater Amazonia understood personhood as something that everything from rocks to humans possessed. Predation transformed prey and predator: people’s bodies were changed by assimilating what they ate, or

by what they absorbed from the skins they prepared and wore. Even poison on arrow tips had “vegetative agency”. People had emotional bonds with animals that didn’t require those animals to perform a service. Individual animals were tamed from the wild, not produced through breeding regimes. Norton argues that “the emergence of the modern pet was, at least in part, a result of this entanglement of European and Indigenous modes of interaction”. In the sixteenth century, most animals shipped to Europe were those that Indigenous people had familiarized. For example, the peoples of the Caribbean Islands and lowland Brazil offered parrots and monkeys to Europeans as gifts and trades — in an attempt to “familiarize” these strangers. Aristocrats in Europe initially acquired these animals as status symbols. These tame animals transferred the nurturing they had received to their new human companions, who were surprised to find that such relationships were meaningful and desirable. Norton’s findings contribute to work in the history of science that reveals how modern science — popularly assumed to be Western in origin — has global roots. It was not just discussions with Indigenous peoples of the Americas that European naturalists absorbed into their zoological treatises, but also concepts and practices derived from Indigenous modes of relating to non-human animals. Norton’s analysis also re-configures histories of conquest. Scholars of Atlantic-Ocean-facing regions of Africa, the Americas and Europe have revealed how Spanish conquistadors joined inter-ethnic conflicts rather than defeating empires such as that of the Aztecs with their bugs, bullets and bigotry alone. As Norton highlights, conquistadors’ expansive use of livestock husbandry “deprived Indigenous people of their labor, their land, and, not infrequently, their lives”. Mines required workers and livestock, and animal husbandry became an alternative industry when mines were exhausted. On the island of Hispaniola, ranchers exported livestock — raised by people held in slavery and Indigenous labourers dispossessed of their lands and coerced into the work — to other colonies in exchange for enslaved people and commodities. The Tame and the Wild is a meticulous and profound reckoning with human–animal relationships. Illuminating for anthropologists, ecologists, biologists and historians alike, it should be read as widely as The Raw and the Cooked (1964), French anthropologist Claude Lévi-Strauss’s classic study of the myths and world views of peoples of eastern Brazil — whose culinary habits are alluded to in the title. Surekha Davies is a historian of science and the author of Renaissance Ethnography and the Invention of the Human. She is currently based in the Netherlands.

Books in brief In the Shadow of Quetzalcoatl Merilee Grindle Belknap/Harvard Univ. Press (2023) Archaeologist and anthropologist Zelia Nuttall focused on the preColumbian cultures of Mexico (those that existed before 1492). Born in 1857, she didn’t attend university and struggled to be recognized for her many achievements. Yet she learnt the languages of the Aztecs and their Mixtec predecessors; decoded their calendar; and taught herself to decipher their pictographic histories and legends. This biography of Nuttall, by Latin American political specialist Merilee Grindle, does justice to a remarkable but forgotten scholar.

Economics in America Angus Deaton Princeton Univ. Press (2023) On his first US visit, British economist Angus Deaton — half believing “the place was infested with gangsters” — thought he saw a man in a restaurant bleeding from a gunshot wound. Later, Deaton understood that he himself had fired the ‘shot’, by stepping on a ketchup packet. This incident chimes with his thought-provoking assessment of US economics. At first glance the field seems driven by politics and “devoid of scientific content”, notes the Nobel laureate. But some economists “do everything that good scientists should do”.

Our Ancient Lakes Jeffrey McKinnon MIT Press (2023) Most lakes are, at most, a mere 10,000 years old. They formed after the last glacial period. But a few — such as lakes Titicaca in South America, Tanganyika in Africa and Baikal in Asia — are millions of years old, formed by tectonic processes. Ancient lakes contain a lot more biodiversity, some of it unique, than do younger ones. This fascinates biologist Jeffrey McKinnon, whose intriguing book explains how these lakes are altering our understanding of “the formation of new species and how life diversifies”.

Ocean Life in the Time of the Dinosaurs Nathalie Bardet et al. Princeton Univ. Press (2023) Giant marine reptiles such as ichthyosaurs, which flourished in the Mesozoic era (252 million to 66 million years ago), are often wrongly called dinosaurs. But that’s like calling a whale a pachyderm because both whales and elephants are large mammals, note four palaeontologists. Their fine book about these extinct marine reptiles, illustrated by Alain Bénéteau, was first published in French in 2021. It details the anatomical, physiological and behavioural adaptations that land-dwelling reptiles needed to flourish in the oceans.

Women in Science Now Lisa M. Munoz Columbia Univ. Press (2023) A project launched in the 1960s asked school children to “draw a scientist”. By 1983, it had collected 5,000 drawings, of which only 28 depicted a female scientist. By 2018, 33% of the more than 20,000 gathered drawings, showed women. In science, too, there has been a shift towards gender equity. But serious obstacles remain, says science writer Lisa Munoz in this practical analysis, complementing it with female scientists’ vivid career stories. “No single intervention, policy, or law is enough,” she rightly notes. Andrew Robinson

Nature | Vol 626 | 29 February 2024 | 947

Setting the agenda in research

JEFF VANUGA/NATURE PICTURE LIBRARY

Comment

One scenario suggests that life began in geothermal pools on land, such as this hot spring in Yellowstone National Park in Wyoming.

To unravel the origin of life, treat findings as pieces of a bigger puzzle Nick Lane & Joana C. Xavier

Explaining isolated steps on the road from simple chemicals to complex living organisms is not enough. Looking at the big picture could help to bridge rifts in this fractured research field.

948 | Nature | Vol 626 | 29 February 2024

T

he origin of life is one of the greatest challenges in science. It transcends conventional disciplinary boundaries, yet has been approached from within those confines for generations. Not surprisingly, these traditions have emphasized different aspects of the question. Or rather, questions. The origin of life is really an extended continuum from the simplest prebiotic chemistry to the first reproducing cells, with molecular machines encoded by genes — machines such as ribosomes, the protein-building factories found in all cells.

Most scientists agree that these nanomachines are a product of selection — but selection for what, where and how? There is no consensus about what to look for, or where. Nor is there even agreement on whether all life must be carbon-based — although all known life on Earth is. Did meteorites deliver cells or organic material from outer space? Did life start on Earth in the hot waters of hydrothermal systems on land or in deep seas? Observations alone cannot constrain these possibilities. The few geological traces that hint at early life are enigmatic. Is a bacterium-like

imprint really a fossil, or some geochemical structure? Is a weak carbon isotope signature on the surface of a mineral a fingerprint of life (which accumulates the lighter carbon-12) or the result of another type of chemical activity? Genes are not directly helpful either. Comparing gene sequences in modern organisms allows researchers to reconstruct a ‘tree of life’ going back to some of the earliest cells that have genes. Although the exact genetic make-up of this ancestral population is disputed, by definition it already had genes and proteins and so can tell us little about how they arose. None of this precludes understanding the origin of life, but it does make competing hypotheses hard to prove or disprove unambiguously. Combine that with the overarching importance of the question and it’s clear why the field is beset with over-claims and counter-claims, which in turn warp funding, attention and recognition. This context has splintered the field. Strongly opposed viewpoints have coexisted for decades over basic questions such as the source of energy and carbon, the need for light and whether selection acts on genes, chemical networks or cells. To understand how life might have begun, researchers must stop cherry-picking the most beautiful bits of data or the most apparently convincing isolated steps, and explore the implications of these deep differences in context. Depending on the starting point, each hypothesis has different testable predictions. For example, if life started in a warm pond on land, the succession of steps leading from prebiotic chemistry to cells with genes is surprisingly different from those that must be posited if the first cells emerged in deep-sea hydrothermal vents. Building coherent frameworks — in which all the steps in the continuum fit together — is essential to making real progress. To see why, here we highlight two of the most prominent frameworks, which propose radically distinct environments for the origin of life.

Prebiotic soup Most people have heard of prebiotic soup. That’s in part because the hypothesis is grounded in the chemistry that works best for making many of the building blocks of living things. In the modern version of this idea, the synthesis of organic molecules begins with derivatives of cyanide, energized by ultraviolet radiation. This chemistry can produce relevant products, such as the nucleotide building blocks of genes, in high yields — although different reactions occur in distinct environments,

ranging from laboratory equivalents of the atmosphere to geothermal ponds and streams1. Where did all this cyanide come from? Meteorite impacts might be one source, but there is little agreement about that among geologists. Nor does this approach explain just how these “reservoirs of material … come to life when conditions change”2. That is, how compounds that formed under disparate conditions could persist for long periods (potentially millions of years) before somehow coming together and self-assembling into growing cells. This framework posits that nucleotides are concentrated in a small pond. To form RNA, the simplest and most versatile genetic material, nucleotides must polymerize. That is most easily achieved by drying them out (polymerization is a type of dehydration reaction). Proponents imagine a succession of wet–dry cycles, in which the pond dries out to form polymers of RNA, then fills again with water containing more nucleotides and so on, cycle after cycle, making more and more RNA3. But this concept raises some difficult questions. It places the onus on an ‘RNA world’, in

“The two frameworks have different advantages and disadvantages, and it is premature to dismiss either.” which RNA acts both as a catalyst (in a similar way to enzymes) and as a genetic template that can be copied. The problems are that there is little evidence that RNA can catalyse many of the reactions attributed to it (such as those required for metabolism); and copying ‘naked’ RNA (that is not enclosed in compartments such as cells) favours the RNA strands that replicate the fastest. Far from building complexity, these tend to get smaller and simpler over time. Worse, by regularly drying everything out, wet–dry cycles keep forming random groupings of RNA (in effect, randomized genomes). The best combinations, which happen to encode multiple useful catalysts, are immediately lost again by re-randomization in the next generation, precluding the ‘vertical inheritance’ that is needed for evolution to build novelty. If selection on RNA in drying ponds could somehow be made to generate greater complexity, what must it achieve? To make cells that grow and reproduce, RNA must encode metabolism: the network of hundreds of reactions that keeps all cells alive. Modern-day metabolic reactions bear no resemblance to

the cyanide chemistry that makes nucleotides in this model. Evolution would therefore need to replace each and every step in metabolism, and there is no evidence that such a wholesale replacement is possible. Unlike evolving an eye, a process in which intermediates have function, encoding only half the steps of a metabolic pathway (or half the pathways needed for a free-living cell) has little, if any, benefit. Can genes that encode multiple metabolic pathways have arisen at once? The odds against this are so great that the astrophysicist Fred Hoyle once compared it to a tornado blowing through a junkyard and assembling a jumbo jet. It is not good enough to counter that evolution will find a way: a real explanation needs to specify how. On balance, we would say that prebiotic chemistry starting with cyanide can produce the building blocks of life, but most of the downstream steps predicted by this framework remain problematic.

Hydrothermal systems Our own favoured scenario is that the chemistry of life reflects the conditions under which life began, in deep-sea hydrothermal systems on the early Earth4. In broad brush strokes, this means that gases such as carbon dioxide (the near-universal source of carbon in cells today) and hydrogen feed a network of reactions with a topology resembling metabolism. Genes and proteins arise within this spontaneous protometabolism and promote the flux of materials through the network, leading to cell growth and reproduction. There are plenty of problems here, too, but they differ from those in the prebiotic soup framework. The first problem is that H2 and CO2 are not particularly reactive — indeed, their chemistry was largely ignored for decades, although rising interest in green chemistry is changing that. But deep-sea vents are labyrinths of interconnected pores, which have a topology resembling cells — acidic outside and alkaline inside. The flow of protons from the outside to the inside of these pores can drive work in much the same way that the inward flow of protons can drive CO2 fixation in cells today5. Research in the past few years shows that these conditions can drive the synthesis of carboxylic acids6 and long-chain fatty acids7, which can self-assemble into cell-like structures bounded by lipid bilayer membranes5. But many chemists are troubled by the idea that, in the absence of enzymes to serve as catalysts, hydrothermal flow could drive scores of reactions through a network that prefigures metabolism, from CO2 right up to

Nature | Vol 626 | 29 February 2024 | 949

Comment

DEBORAH KELLEY AND MITCH ELEND, UNIV. WASHINGTON

nucleotides. The chemist Leslie Orgel once dismissed this scenario as an “appeal to magic”. Certainly, further data are required, supporting or otherwise. Multiple steps have now been shown to occur spontaneously in core metabolic pathways (such as the Krebs cycle and amino-acid biosynthesis) without being driven by enzymes8, but this is still far from demonstrating flux through the entire network. Polymerization is another stumbling block. Nucleotides have been polymerized in water on mineral surfaces9, but this raises similar questions to those noted for wet–dry cycles about how selection could act. If the problem is solved by polymerizing nucleotides inside growing protocells, mineral surfaces would not have been available. Polymerization would then have needed to happen in cell-like (aqueous gel) conditions, but without enzymes. If serious attempts to synthesize RNA under those conditions fail, the overall framework would need to be modified. Conversely, if these difficult problems are resolved, then the hydrothermal scenario offers a promising route to the emergence of genetic information, overcoming Hoyle’s jumbo-jet argument. Patterns in the genetic code suggest direct physical interactions between amino acids and the nucleotides that encode them, especially for those formed most easily by metabolism5. Such associations mean that random RNA sequences could act as templates for non-random peptides that have a function in growing protocells. The first genes wouldn’t have had to encode metabolism, but just enhance flux through a spontaneous protometabolism — for example, by enabling the reaction between H2 and CO2. Thus, in short, the two frameworks have different advantages and disadvantages, and it is premature to dismiss either.

Findings can be true but irrelevant Similarly probing questions apply to other origins-of-life scenarios. If organic molecules were delivered from space — for instance, in carbonaceous chondrites such as the Murchison meteorite10 — then how and where did they come together, how did they polymerize, and so on? The delivery of organics from space simply stocks a soup and doesn’t solve most of the downstream problems — with the further issue that such a delivery method is unlikely to have been reliable and consistent at specific locations. If life started out as droplets known as coacervates, in which immiscible liquids separate into distinct phases that promote different types of chemistry, then one must ask where all the precursors to feed their growth came from. And how did these phase-separated droplets morph into cells with different topology, in which these distinct chemistries now mostly occur under aqueous-gel conditions? Similar questions can be asked about 950 | Nature | Vol 626 | 29 February 2024

A 13-metre-tall carbonate chimney in the Lost City hydrothermal field in the Atlantic Ocean.

‘eutectic freezing’ (in which growing ice crystals concentrate the surrounding soup) and layered minerals or pores in volcanic rocks, such as basalt or floating pumice, that catalyse organic synthesis. All of these fragments of scenarios are ‘true’, in that there is empirical evidence supporting each snapshot moment. But the fact that it is possible to make amino acids by passing electrical discharges through a Jovian mixture of gases, as the US chemist Stanley Miller

famously did 70 years ago, does not mean that is how life began — merely that this chemistry is possible. Likewise, the fact that analogous chemistry can occur in hydrothermal systems, or from cyanide in terrestrial geothermal systems, or in interstellar space, does not mean that all of these environments were required for life to start, just that this chemistry is favoured under many conditions. The question is always: what happens next? If none of these scenarios is ‘wrong’, then

there is space in the field to pursue multiple frameworks. No one needs to abandon their favoured positions (yet). But brash claims for a breakthrough on the origin of life are unhelpful noise if they do not come in the context of a wider framework. The problem is ultimately answerable only if the whole question is taken seriously.

Look for convergence points An important feature of these competing frameworks is that they must ultimately converge on cells with genes and proteins — on life as we know it on Earth. This convergence offers new possibilities for collaboration, because any answer will probably feature aspects of more than one framework. Exactly where these convergences occur will depend on which hypothetical steps are disproved. Cofactors offer a possible convergence point. They got their name because they work together with an enzyme to catalyse a reaction. But from an origins-of-life perspective, the term is misleading because cofactors usually catalyse the same reaction on their own, albeit more slowly. Many cofactors derive from nucleotides, such as nicotinamide adenine dinucleotide. These might prove hard to make when starting with CO2. Could it be that cofactors were initially synthesized from cyanide, but, once in circulation, tended to catalyse CO2 chemistry, now driving a lifelike protometabolism that included their own synthesis11? Perhaps, but this idea also shows how important it is to test predictions within a specific framework first. In the simplest scenario, all of biochemistry begins from CO2 in a hydrothermal system, whereas the alternative scenario calls for at least two places and two types of chemistry — adding up to much more uncertainty. Occam’s razor says that the simplest scenario should be tested thoroughly first. If the simplest chemistry is shown not to work — that is, if it is not possible to synthesize cofactors from CO2 without cofactors — then the alternative can be taken seriously. This question could be approached experimentally or using modern computational chemistry tools, but either way, the best way to make progress is to test the simplest idea to destruction first. If it can be shown not to work, then the convergence point might be real, and should be explored seriously.

Towards an answer The origins-of-life field faces the same problems with culture and incentives that afflict all of science — overselling ideas towards publication and funding, too little common ground between competing groups and perhaps too much pride: too strong an attachment to favoured scenarios, and too little willingness to be proved wrong. These incentives are amplified by the difficulty of disproving complex interrelated hypotheses involving different

disciplines when there is so little direct evidence — no ‘smoking gun’ to be discovered. Changing this culture will take some work, given the political reality of science — the relentless pressure to publish, to secure funding, tenure or promotion — but it is necessary if the field wishes to continue attracting students. This requires that scientists, but also editors and funders, are aware of the issues that fragmented the field and work to overcome them. We highlight four priorities to begin to move in the right direction. Train interdisciplinary scientists. Pursuing hypotheses across conventional disciplinary boundaries calls for a new generation of scientists — PhD students, postdoctoral researchers and early-career principal investigators (PIs) — with wide-ranging expertise and a willingness to test specific hypotheses within coherent wider frameworks. The field will clearly benefit from doctoral training that stresses collegiality, interdisciplinarity and the rigorous, openminded testing of competing hypotheses. Foster good communication. To promote such a culture, one of us ( J.C.X.) co-founded the Origin of Life Early-career Network (OoLEN) in 2020, which has grown to more than 200 international researchers, from students

“It is too soon to aim for consensus or unity, and the question is too big; the field needs constructive disunity.” to early-career PIs. It is run by volunteers and has no institutional ties, financial or otherwise. Members engage in debates through regular meetings (online or in-person), disseminate research and write articles together. There is still no shortage of disagreements, but that is part of scientific research and OoLEN promotes a healthy approach to them12. For later-career researchers, conferences could help to reach across divides in similar ways. Physics meetings have provided examples. In one, proponents of loop quantum gravity and string theory switched sides in a debate, framing good-humoured but strong arguments against their own position in a constructive form of ‘steel manning’.

valuable is a more systematic use of open-access, community-driven knowledge bases that would host and curate data. These would help to collate experimental conditions, highlight genuine gaps in empirical evidence and enable analysis of large data sets through machine-learning studies. Improve publishing practices. Researchers should aspire to contextualize their findings in cover letters, papers and press releases, to give a sense of how the work fits into a wider framework. Refraining from hype might seem unrealistic but could work if researchers implemented this practice in their roles as peer reviewers for papers and grants as well as authors. Journal editors and grant-awarding bodies should also consider how polarized the field is to ensure fair reviews. One way to improve the peer-review process would be to enlist more early-career researchers, who tend to be less entrenched in their positions. Transparent peer review (in which anonymous reports are published with a paper) could also curb bias, because it enables constructive criticism without concealing prejudice. It is too soon to aim for consensus or unity, and the question is too big; the field needs constructive disunity. Embracing multiple rigorous frameworks for the origin of life, as we advocate here, will promote objectivity, cooperation and falsifiability — good science — while still enabling researchers to focus on what they care most about. Without that, science loses its sparkle and creativity, never more important than here. With it, the field might one day get close to an answer.

The authors Nick Lane is professor of evolutionary biochemistry in the Division of Biosciences, University College London, UK. Joana C. Xavier is a scientist in the Department of Chemistry, Imperial College London, UK. e-mails: [email protected]; [email protected] 1. 2. 3. 4. 5.

Embrace open science. Accepting that specific hypotheses will be disproved and that frameworks will be reshaped requires the publication of negative results — too often undervalued and unpublished. But it is clearly important for the field to know whether, for example, attempts to synthesize cofactors from CO2 fail — and, specifically, under what conditions. Dissemination of negative data could be promoted in several ways. Most

6. 7. 8. 9. 10. 11. 12.

Patel, B. H., Percivalle, C., Ritson, D. J., Duffy, C. D. & Sutherland J. D. Nature Chem. 7, 301–307 (2015). Szostak, J. Nature 557, S13–S15 (2018). Hassenkam, T. & Deamer, D. Sci. Rep. 12, 10098 (2022). Martin, W., Baross, J., Kelley, D. & Russell, M. J. Nature Rev. Microbiol. 6, 805–814 (2008). Harrison, S. A. et al. Annu. Rev. Ecol. Evol. System. 54, 327–350 (2023). Preiner, M. et al. Nature Ecol. Evol. 4, 534–542 (2020). Purvis, G. et al. Commun. Earth Environ. 5, 30 (2024). Muchowska, K. B., Varma, S. J. & Moran, J. Chem. Rev. 120, 7708–7744 (2020). Jerome, C. A., Kim, H.-J., Mojzsis, S. J., Benner, S. A. & Biondi, E. Astrobiology 22, 629–636 (2022). Martins, Z. et al. Earth Planet. Sci. Lett. 270, 130–136 (2008). Xavier, J. C., Hordijk, W., Kauffman, S., Steel, M. & Martin, W. F. Proc. R. Soc. B 287, 20192377 (2020). Preiner, M. et al. Life 10, 20 (2020).

The authors declare no competing interests.

Nature | Vol 626 | 29 February 2024 | 951

SAUMYA KHANDELWAL/REDUX/EYEVINE

Comment

Nearly one-third of the world’s population has still not received a single dose of vaccine for COVID-19.

Save lives in the next pandemic: ensure vaccine equity now Colin Carlson, Daniel Becker, Christian Happi, Zoe O’Donoghue, Tulio de Oliveira, Samuel O. Oyola, Timothée Poisot, Stephanie Seifert & Alexandra Phelan

The proposed Pandemic Agreement must ensure that COVID-19 vaccine nationalism is never repeated; 290 scientists call for action.

952 | Nature | Vol 626 | 29 February 2024

S

ince 2022, member states of the World Health Organization (WHO) have been negotiating a new treaty — provisionally termed the Pandemic Agreement. If adopted, it would transform how the world handles pandemic prevention, preparedness and response. Opinions differ on what negotiators should prioritize. But no issue has captivated public attention as much as vaccine equity — or done more to bring countries to the negotiating table. During the COVID-19 pandemic, scientists began to design vaccine candidates only a few hours after the first SARS-CoV-2 genome sequence was shared. By the end of 2020, mass vaccination had begun in the United States and Europe. High-income countries promised to share vaccines through the voluntary WHO COVID-19 Vaccines Global Access (COVAX)

programme, but failed to meet their commitments. When South Africa and India appealed to the World Trade Organization for an emergency waiver of intellectual-property rights related to COVID-19 vaccines, so that every country could start their own manufacturing, high-income countries blocked the proposal for months. The refusal of wealthier nations to cooperate had cost between 200,000 and 1.3 million lives by the end of 2021 in low- and middle-income countries1,2. Today, nearly one-third of the world’s population has still not received a single dose, and the death toll resulting from vaccine nationalism continues to grow. The Pandemic Agreement could be the last chance to fix this problem before the next COVID-19 arrives. Yet the proposed solution — the Pathogen Access and BenefitSharing (PABS) System, which was outlined in

Article 12 of the latest treaty draft — still hangs in the balance. The second-to-last session of the treaty’s Intergovernmental Negotiating Body is now under way. So far, countries have been unable to agree on this part of the text. As time runs out, we urge WHO member states to agree on a ‘science-for-science’ mechanism that ensures vaccine equity in the next pandemic.

The road to PABS Across all fields, scientists from the global north have frequently extracted data and samples from the global south without the permission of the people there, without collaborating meaningfully — if at all — with local scientists, and without providing any benefit to the countries where they conduct their work. In 1993, the Convention on Biological Diversity recognized parties’ sovereign rights to their ‘genetic resources’. Since 2014, under the Nagoya Protocol on Access and Benefit-sharing, countries have developed their own legislation to ensure that they receive benefits (such as financial compensation or scientific collaboration) when scientists and others from outside the country access their genetic resources. Discussions on access and benefit-sharing in global health began in earnest in 2007, when the Indonesian government refused to share avian influenza samples with the rest of the world, on the grounds that such samples were often used to make vaccines that were never made available in most places3. Sparked by this conflict — and the 2009 H1N1 flu pandemic — WHO member states developed the 2011 Pandemic Influenza Preparedness (PIP) Framework to streamline the sharing of influenza viruses with pandemic potential, as well as vaccines and other benefits. Under the PIP Framework, 14 manufacturers have promised that when the next influenza pandemic starts, they will share up to 10% of the vaccines that they make (around 420 million doses) with the WHO. In exchange, these companies have access to a global network of laboratories and their flu samples. The PIP model shows significant promise, but is so far untested and applies only to influenza. The proposed PABS System in the Pandemic Agreement would take lessons from the PIP Framework and apply an access and benefit-sharing scheme to any pathogen with pandemic potential, such as SARS-CoV-2. Under the PABS System, scientists would share pathogen samples and data through a global network of laboratories and sequence data repositories. In exchange for access to samples and data, manufacturers of vaccines or therapeutics would give at least 20% of their products to the WHO (half for free, and half at affordable prices). The WHO would then distribute these on the basis of public-health risk and needs. Users of the PABS System would also contribute to a capacity-development fund, and be encouraged to explore other

kinds of benefit-sharing, such as scientific collaborations and technology transfer.

Science-for-science With regard to physical samples, the Nagoya Protocol and its national implementing legislation can be cumbersome to navigate4. Some scientists are apprehensive about the idea of introducing similar barriers into work with genetic sequence data, especially during outbreaks. In relation to the Nagoya Protocol, several professional societies, including the American Society for Microbiology, have endorsed a group of US scientists that opposes “any restriction or control of access and/or use” of any genetic sequences (see go.nature.com/3i5ds). Comments from sessions indicate that such concerns are increasingly being echoed by representatives of global north countries in the current Pandemic Agreement negotiations. Some critics have even argued that the proposals for PABS would block progress towards open science, in favour of a transactional approach5. As a collective of 290 scientists from 36 countries, we argue that a pandemic treaty cannot succeed unless it ensures that everyone will benefit from pandemic science. Under the new treaty, should it be adopted with the current vision of the PABS System, countries will still be expected to ensure that their scientists share lifesaving data openly and rapidly. Scientists will still be able to share their data freely outside of PABS platforms, and widely used databases could enter into the PABS System — meaning that most researchers would never experience any disruptions to their workflow. The WHO could also establish its own repository or clearinghouse for genetic sequence data and samples, which would potentially provide scientists with more transparent management of these resources and the guarantee of continued access. Financing committed largely by pharmaceutical firms using these platforms (which sometimes directly derive profits from publicly funded science) would, in turn, go towards expanding sequencing capacity and scientific research in low-resource settings. It would also help to support other priorities, such as pandemic prevention6. What’s more, scientists everywhere, but especially in the global south, would benefit from a system that creates opportunities for international collaboration — and that ensures that people receive credit for sharing their data.

Hold the course Access and benefit-sharing could just as easily be called ‘science for science’: the PABS System will support more pandemic science, and ensure that scientists’ contributions result in their communities having access to lifesaving advancements. Last week, the Intergovernmental Negotiating

Body for the Pandemic Agreement reconvened for its penultimate session. If Article 12 is weakened or dismantled, it will be a monumental setback for global health justice — and for the global scientific community. Although today’s scientific community has embraced the ideals of open data sharing, the world is no closer to a fair system for sharing vaccines and therapeutics. Intellectual property, not benefit-sharing, is the antithesis of open science. We dream of a world in which such barriers are dismantled for lifesaving medicines. Until that day, the Pandemic Agreement offers the last best chance to avoid repeating the mistakes made during the COVID-19 pandemic.

The authors Colin Carlson is assistant research professor and director of the Verena Institute at Georgetown University, Washington DC, USA. Daniel Becker is assistant professor at the University of Oklahoma, Norman, Oklahoma, USA. Christian Happi is professor of molecular biology and genomics and director at the African Center of Excellence for Genomics of Infectious Diseases, Redeemer’s University, Osun, Nigeria. Zoe O’Donoghue is programme manager for the Verena Institute at Georgetown University, Washington DC, USA. Tulio de Oliveira is professor of bioinformatics and director, Centre for Epidemic Response and Innovation, Stellenbosch University, Stellenbosch, South Africa, and fractional professor and director, KwaZulu-Natal Research and Innovation Sequencing Platform, University of KwaZulu-Natal, Durban, South Africa. Samuel O. Oyola is senior scientist in molecular biology and head of genomics, International Livestock Research Institute, Nairobi, Kenya. Timothée Poisot is associate professor, Université de Montréal, Montreal, Canada. Stephanie Seifert is assistant professor, Paul G. Allen School for Global Health, Washington State University, Pullman, Washington. Alexandra Phelan is associate professor and senior scholar, Center for Health Security, Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA. e-mail: [email protected] 1. Moore, S., Hill, E. M., Dyson, L., Tildesley, M. J. & Keeling, M. J. Nature Med. 28, 2416–2423 (2022). 2. Watson, O. J. et al. Lancet Infect. Dis. 22, 1293–1302 (2022). 3. Sedyaningsih, E. R., Isfandari, S., Soendoro, T. & Supari, S. F. Ann. Acad. Med. Singap. 37, 482–488 (2008). 4. WHO Emerging Tech, Research Prioritisation & Support Team. Implementation of the Nagoya Protocol and Pathogen Sharing: Public Health Implications (World Health Organization, 2017). 5. Hampton, A.-R., Eccleston-Turner, M., Rourke, M. & Switzer, S. J. Law Med. Ethics 51, 217–220 (2023). 6. Vora, N. M. et al. Nature 605, 419–422 (2022). The authors declare competing interests; see go.nature. com/42qkf for details. A list of co-signatories accompanies this article online; see go.nature.com/42qkf

Nature | Vol 626 | 29 Feburary 2024 | 953

Readers respond

Correspondence Cold war lessons for Arctic diplomacy

Long COVID needs novel clinical trials

Russia’s threat to withdraw from the Arctic Council is a matter for global concern, with burning cold-war security issues becoming hot again. Since 1996, the council has been the highlevel forum dealing with common Arctic issues through science and dialogue. But, as stipulated in its founding Ottawa Declaration, it “should not deal with matters related to military security”. This wisdom was abandoned nine days after the full-scale Russian invasion of Ukraine in February 2022, when the seven other Arctic Council states issued a joint statement “pausing participation in all meetings of the Council and its subsidiary bodies”. This pause in dialogue is becoming permanent, undermining open science along with climate and other research in the Arctic. But more than that, the continuing lack of dialogue among allies and adversaries alike is the beginning of conflict. Lessons from after the Second World War should be heeded now. The third International Polar Year (IPY), which became the International Geophysical Year (IGY) 1957–58, led directly to cooperation between the United States and Soviet Union in Antarctica as well as space throughout the cold war. The IGY facilitated the 1959 Antarctic Treaty, which became the first nuclear-arms agreement and template for the Arctic Council, with continuous cooperation among superpower adversaries. The fifth IPY, in 2032–33, offers a practical time horizon to reverse the deterioration of East–West relations, again with science diplomacy and commoninterest building.

Diagnostic biomarkers and effective therapies are urgently needed for the millions of people living with long COVID. But the challenges of designing and conducting clinical trials mean that only large, well-funded academic centres can engage with the problem. We propose an alternative approach, based on interactions between clinician–patient pairs and researchers. Before clinical trials, an online platform could enable the peer review of trial designs and plans for statistical analyses. After recruitment, the focus would shift to clinician- and patient-reported outcomes and biomarker read-outs, ideally from wearable technologies. During treatment, a cloudbased system could be used to report adverse events and real-time biomarker read-outs, with general practitioners providing an untapped source of data. After treatment, the peer-review system could make data accessible to all relevant researchers. We are confident that this ‘grassroots’ system would avoid long COVID problems that can plague clinical trials: low enrolment, late or missing trial reporting and faked or fatally flawed results.

Paul Arthur Berkman Science Diplomacy Center, Falmouth, Massachusetts, USA. [email protected]

Marc Jamoulle University of Liège, Liège, Belgium. Elena Louazon Université Libre de Bruxelles, Brussels, Belgium. Tomaso Antonacci Belgian Association of Long COVID Patients, Brussels, Belgium.

Europe must join forces to monitor its forests Last November, the European Commission proposed a regulation to establish a coordinated monitoring framework for resilient forests using a combination of imagery from the European Union’s Copernicus Earth-observation satellites and in situ data, mainly from national forest inventories. The proposal is based on the premise that forest monitoring in Europe is “fragmented and patchy”, with no fully developed “consistent, transnational datagathering approach”. This premise, however, is misleading. In 1986, the commission launched a coordinated forest monitoring scheme, which evolved in cooperation with the International Co-operative Programme on Assessment and Monitoring of Air Pollution Effects on Forests (ICP Forests), which I currently chair. The programme now covers 37 European countries and has a comprehensive portfolio of harmonized, quality-assured methodologies, databases and governance. Such infrastructures can provide essential data to explain changes in forest conditions and to understand processes, both key aspects when aiming to build resilient forests. At a time of increased signals of forest vulnerability, it would be a missed opportunity not to take advantage of all the available resources for the future European forest monitoring system.

Johan Van Weyenbergh KU Leuven, Leuven, Belgium. johan.vanweyenbergh@kuleuven. be

Marco Ferretti Swiss Federal Institute for Forest, Snow and Landscape Research, Birmensdorf, Switzerland. [email protected]

The authors declare competing interests; see go.nature.com/42rxxnk for details.

The author declares competing interests; see go.nature.com/49gphse for details.

954 | Nature | Vol 626 | 29 February 2024

Train taxonomists to save biodiversity Species extinctions are speeding up worldwide. Biodiversity monitoring and assessment must underpin efforts to tackle this crisis (E. Tekwa et al. Phil. Trans. R. Soc. B 378, 20220181; 2023). Yet expertise in taxonomy, the scientific basis for biodiversity research and management, has been in decline. University credit hours in taxonomy that have been reallocated to fields such as molecular biology or biotechnology should be reinstated. Training in technologies such as digital and virtual-reality herbaria, wildlife camera traps and environmental-DNA analysis should be combined with schooling in empirical research practices. Community scientists and Indigenous people play an important part in conservation, and trained members of local groups could bolster volunteer efforts to monitor biodiversity. Artificial intelligence can also help: trained on large taxonomic data sets, it could be used to recognize plant morphologies or animal audio recordings to aid species identification, for example. Such initiatives could fill gaps in expertise and help to achieve the United Nations Sustainable Development Goals for biodiversity conservation by 2030. Dasheng Liu Ecological Society of Shandong, Jinan, China. [email protected] Julian R. Thompson University College London, London, UK. Jun Gao Nanjing Institute of Environmental Science, Ministry of Ecology and Environment, Nanjing, China. Henglun Shen Zaozhuang University, Zaozhuang, China.

Expert insight into current research

NATURE PICTURE LIBRARY/ALAMY

News & views

Figure 1 | Hair braiding in Namibia. Efferson et al.2 shed light on how reciprocal behaviour (such as this type of cooperative activity) might have evolved in human societies.

Evolution

Why humans reciprocate but animals usually do not Sarah Mathew

Reciprocal cooperation can be advantageous, but why it is more common in humans than in other social animals is a puzzle. A modelling and experimental study pinpoints the conditions needed for reciprocity to evolve. See p.1034 Reciprocity is so intuitive to humans that its evolutionary logic can seem self-evident. If there is a high chance that individuals will interact again, it pays to be nice to those who might return the favour. A rich body of theoretical work1 has confirmed this idea, showing that — as long as there is a high probability of

interacting with the same person again, and individuals preferentially help those who have previously helped them — reciprocal cooperation is advantageous despite its short-term cost. However, on page 1034, Efferson et al.2 report evidence suggesting that the evolutionary path to reciprocity is treacherous at

best, and impossible at worst — unless natural selection favours not only individuals, but also groups, that cooperate more. As counterintuitive as the finding might be, it could clarify a key paradox about reciprocity theory. Despite the long-term gains that reciprocal cooperation offers, most animals do not cooperate with individuals that are not related to them, even when they have many opportunities for future interactions3. By contrast, humans exchange a wide variety of goods and services with unrelated individuals during daily life (Fig. 1) in a manner that is consistent with reciprocity theory4. A satisfactory theory should explain not just why humans cooperate, but also why other animals do not when the conditions for cooperation to evolve seem to be met. Despite theoretical shortcomings in explaining why most animals do not behave reciprocally, researchers continue to ascribe cooperation in repeated interactions to reciprocity, and have turned their gaze to a more unusual phenomenon — cooperation in one-off interactions. In a conspicuous deviation from

Nature | Vol 626 | 29 February 2024 | 955

News & views reciprocal cooperation, humans help strangers, even in transient interactions. If the recipient cannot return the favour later, then how does the helper recoup the cost of helping? Some think that cooperation in such one-off interactions is a ‘misfiring’ in modern settings, resulting from a reciprocity-driven psychology that evolved when humans lived in smaller groups in which interactions between individuals were almost always repeated5. Others think that such cooperation arises through group selection, in which competition between groups of people with different cooperative behaviours and norms favours those with high levels of in-group cooperation, including in one-off interactions6. Teasing apart which of these explanations is correct is difficult, fuelling a debate over the roots of our altruistic disposition5,7. To try to settle this argument, Efferson et al. developed a model that simulates which cooperation strategies evolve in populations over time when interactions are repeated and when group selection occurs, to derive precise predictions of the scenarios that give rise to one-off cooperation. Contrary to both sides of the debate, neither repeated interactions nor group selection consistently produced oneoff cooperation. More surprisingly, repeated interactions did not yield the most obvious outcome, reciprocal cooperation. These results emerged because the model was constructed in a way that did not make standard simplifying assumptions, thus ensuring that it did not sidestep certain realities of the natural world that have profound effects. First, rather than having the model consider cooperation and non-cooperation as two discrete options, individuals could cooperate to any extent along a continuous scale. Second, any conceivable cooperation strategy could arise through chance and compete with existing strategies in the population. In standard models, the strategies that can arise are predetermined to make the model more straightforward, so existing strategies in the population are artificially protected from their full range of competing strategies. These two decisions exposed a fundamental weakness of repeated interactions as a mechanism by which cooperation can evolve — when cooperative reciprocity gains a foothold, it opens the door for less-cooperative strategies in which individuals reciprocate by giving a little less than they receive. Over time, cooperation slides down to negligible levels. With reciprocal cooperation on a slippery slope, why didn’t group selection claim victory instead? To encompass a wide range of realistic scenarios in their model, Efferson et al. varied the timing of when cooperation events take place relative to when individuals disperse from their original group, which affects whom individuals will cooperate with, as well as whom they will compete with for 956 | Nature | Vol 626 | 29 February 2024

resources. In the majority of the resulting scenarios, the advantage that cooperators get by being around other cooperators, is cancelled out by cooperators competing with other cooperators for resources. Researchers have long known how this ‘cancellation effect’ plays out at the individual level8, but only in the past four years has its detrimental effects been described at the group level9. Efferson and colleagues’ study assessed the cancellation effect at both the individual and group level. Remarkably, although neither repeated interactions nor group selection work in isolation, they almost always generate reciprocal cooperation when they act in concert. Moreover, the resulting cooperation is much bigger than the sum of what either reciprocity or group selection can generate alone, which the authors refer to as super-additive cooperation. To test the theoretical findings, the authors then conducted a social-dilemma experiment with participants from the Ngenika and Perepka groups in Papua New Guinea. The influence of state institutions at the location of these groups is weak, so it is easier to observe how individual social strategies influence cooperation. In the experiment, individuals could transfer any amount of their endowed cash to their

“Having gained a finer awareness of the limitations of the prevailing theories, we can now ask new questions.” partner, who then received double the amount transferred. When paired with an in-group member, first-movers transferred high amounts and second-movers reciprocated more than they received; when paired with an out-group partner, first-movers made low transfers, and second-movers reciprocated less than they received. This pattern of escalating and de-escalating reciprocity in in-group and out-group interactions respectively, was observed in the theoretical model only when repeated interactions and group selection operate simultaneously. The finding that repeated interaction needs group selection to yield super-additive reciprocal cooperation could be the long-awaited answer to why reciprocity is pervasive in humans and includes high-stakes interactions (such as proactive sharing of food), but is rare in other social animals and usually restricted to low-stakes interactions (such as tolerating an individual at a feeding site). Repeated interaction needs group selection to yield reciprocal cooperation, but the conditions for group selection to occur are nearly universally absent in the natural world, because there is insufficient genetic variation between groups. However, cultural characteristics, which influence

human behaviour, do differ between groups10. Therefore, groups can differ in their success when competing with other groups, making group-level selection an important force in shaping the evolution of human societies. Having gained a finer awareness of the limitations of the prevailing theories, we can now ask new questions. Given that reciprocity models produce qualitatively different results when traits are modelled as continuous versus discrete characteristics, are there other theories based on models of discrete traits that should be re-examined? For instance, researchers have shown that the intuitive idea that arbitrary conventions can persist owing to the social pressure to do what others do, which applies to discrete norms (such as which side of the road to drive on), does not hold up for continuous norms (for example, how much to tip at a restaurant)11. Another intriguing question is whether cultural norms can mitigate the cancellation effect by manipulating the social scale at which people cooperate and compete. For example, norms can dissuade people from waging war with culturally similar groups, but not from going to war with culturally dissimilar people10. Perhaps such norms reoccur in different societies because they expand and redraw group boundaries in ways that dampen the effect of cancellation. What other combinations of mechanisms can yield super-additive cooperation? Will the reciprocation strategies observed in the Ngenika and Perepka extend to other cultural contexts? Evidently, the case regarding cooperation isn’t closed, but we can undoubtedly make faster progress if we stop assuming reciprocity to be the baseline from which deviations in cooperation are assessed — because reciprocity in theory, as in real life, cannot be taken for granted. Sarah Mathew is at the Institute of Human Origins and at the School of Human Evolution and Social Change, Arizona State University, Tempe, Arizona 85287, USA. e-mail: [email protected] 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

van Veelen, M., García, J., Rand, D. G. & Nowak, M. A. Proc. Natl Acad. Sci. USA 109, 9929–9934 (2012). Efferson, C., Bernhard, H., Fischbacher, U. & Fehr, E. Nature 626, 1034–1041 (2024). Clutton-Brock, T. Nature 462, 51–57 (2009). Phelps, J. R., Pitogo, K. M. E., Emit, A. T. & Hill, K. PLoS ONE 18, e0290270 (2023). Delton, A. W., Krasnow, M. M., Cosmides, L. & Tooby, J. Proc. Natl Acad. Sci. USA 108, 13335–13340 (2009). Boyd, R., Richerson, P. J. & Henrich, J. Behav. Ecol. Sociobiol. 65, 431–444 (2011). Zefferman, M. R. Evol. Hum. Behav. 35, 358–367 (2014). Taylor, P. D. Evol. Ecol. 6, 352–356 (1992). Akdeniz, A. & van Veelen, M. Evolution 74, 1246–1254 (2020). Handley, C. & Mathew, S. Nature Commun. 11, 702 (2020). Yan, M., Mathew, S. & Boyd, R. PNAS Nexus 2, pgad054 (2023). 

The author declares no competing interests. This article was published online on 21 February 2024.

Polymer chemistry

Liquid-like droplets of supramolecular polymers Jennifer L. Ross

The molecules of liquid crystals and proteins can form liquid-like condensates, but such a phenomenon had not been observed for supramolecular polymers, which are held together by non-covalent bonds — until now. See p.1011 The separation of liquids from each other is such a common occurrence that there is an idiom about oil and water not mixing. It is more surprising when a substance that is dissolved in water spontaneously separates. If that substance forms a liquid-like state on separation, rather than aggregating into a solid, the resulting droplets can be used to confine or organize objects from the nanometre to the macroscopic scale. In the past few years, scientists have observed that such liquid–liquid phase separation (LLPS) of proteins, RNA and DNA occurs in live cells and has crucial roles in many biological processes, including gene regulation and RNA processing and degradation1–6. On page 1011, Fu et al.7 report the first demonstration that synthetic supramolecular polymers — macromolecules that self-assemble as a result of non-covalent interactions between the constituent monomer molecules — undergo LLPS. The authors made the serendipitous observation that a compound known as ureidopyrimidinone glycine (UPy–Gly) polymerizes in solution to form short fibres, which then promote a condensation process

that produces high-density phases containing supramolecular fibrils. Because the fibrils have a high aspect ratio (they are nanometres in diameter, and micrometres in length), these condensed assemblies of fibrils were not spherical, but instead were tactoids — that is, spindle-shaped with pointed ends (Fig. 1). Fu and colleagues went on to rigorously characterize the shape, structure, mechanics and dynamics of the condensed phase. Using a technique called confocal laser scanning microscopy, the authors found that the tactoids were liquid-like, acting as droplets that can merge together, and that the fibrils could diffuse within the tactoids. As time advanced, the tactoids grew larger both through merging events and through continuous elongation of the fibrils. The authors also found that a minimum fibril length is needed for the condensation to occur. Much like conventional LLPS, the condensation process observed by Fu et al. is driven by an increase in entropy (a measure of the number of states a system can adopt). It might seem counter-intuitive that a process in which supramolecular polymer molecules become ‘more organized’ than they were before leads to

a

b

larger entropy, but this is the result of smaller molecules in the system being ‘freed up’ during tactoid formation. One indicator of the role of entropy is that the process is controlled by temperature, with higher temperatures driving faster condensation and producing smaller tactoids. Another indicator is that the tactoids continue to grow over time, increasing in length as the fibrils in them get longer. Fu et al. also observed that the addition of crowding agents (inert compounds, such as polymers, the role of which is to occupy space) sped up the formation of the tactoids, increased the number that formed and reduced their size. These effects occur because crowding agents reduce the volume of solvent available for the UPy–Gly molecules, thereby increasing the effective concentration of those molecules — which then gain entropy when larger molecules in the system are pushed together. Increasing the amount of crowding agent also changed the shape of the tactoids, from bipolar (wide, spindle-shaped) condensates in which the fibrils are oriented along arcs that connect the tips of the spindle, to longer and thinner homogeneous (rod-shaped) tactoids, in which the fibrils all align with the straight axis of the tactoid. Using an X-ray scattering technique, the authors determined that the fibrils in the tactoids were hexagonally packed, and that they packed together more densely over time and at high concentrations of crowding agents. Not only did individual fibrils in the tactoids continue to grow, but they also became less mobile, causing the tactoids to become more viscous. Similar gelation has been observed in many condensed systems that undergo LLPS, including protein systems8–10. Such gelation is often reversible, but irreversible protein gelation can lead to aggregation of the condensed phases, and is often observed in neurodegenerative and neuromuscular diseases3,11,12. Studies of how synthetic polymer systems can form gels (and of the putative c Liquid droplet Crowding agent

Short fibres in solution

Insoluble tactoid Irregular hexagonal packing

Insoluble rods

Glass surface

Denser, regular hexagonal packing

Figure 1 | Supramolecular polymers form tactoid assemblies. a, Supramolecular polymers self-assemble from subunits in solution and are held together by non-covalent bonds. This produces fibres that increase in length over time. Fu et al.7 report that fibres of a supramolecular polymer act as micrometre-scale versions of liquid-crystal molecules. At high concentration or in the presence of crowding agents (inert compounds that take up space in solution; not shown), the fibres condense into

Immiscible liquid polymer

spindle-shaped, liquid-like droplets known as tactoids, in which the fibres are packed imperfectly into a hexagonal arrangement. Over time (or with crowding agents), the fibres in the tactoids grow even longer and align, forming more densely packed rods with regular hexagonal packing. b, The tactoids form forest-like arrays on a glass surface. c, They also form crown-like arrangements at the surfaces of liquid droplets suspended in an immiscible liquid polymer.

Nature | Vol 626 | 29 February 2024 | 957

News & views reverse process) might shed light on how protein aggregation can be reversed, given that synthetic polymers can easily be engineered to mimic the properties of proteins. Another intriguing attribute of Fu and colleagues’ tactoids is that they can self-assemble into larger-scale structures. For example, the authors observed that tactoids on a glass coverslip aligned perpendicularly to the surface, forming an array. At first, the tactoids could diffuse laterally on the surface and coalesce. But as they aged and grew longer, they stopped coalescing and tilted, creating a forest of leaning spindles. This indicates that supramolecular polymers can undergo hierarchical assembly processes, from the nanoscale (supramolecular polymer subunits) to the millimetre scale (assemblies of tactoids), which could potentially be used to fabricate microscale bristles that, in turn, serve as scaffolds for larger structures. By contrast, tactoids on a liquid surface formed distorted, crown-like shapes. Fu et al. observed this by dissolving UPy–Gly molecules in a system in which droplets of one polymer (dextran) form in another immiscible liquid polymer (polyethylene glycol; PEG). The supramolecular polymer formed from UPy–Gly at first dissolved into the PEG, but then formed condensates at the surface of the dextran droplets, eventually encapsulating the droplets. The authors found that continuous or discrete supramolecular networks could be formed at the surface of the droplets by varying the pH and concentrations of salts in the system. Similar condensation behaviour to that of UPy–Gly was observed when other supramolecular polymers were tested in the PEG–dextran system, indicating that LLPS is a general phenomenon for supramolecular polymers. Fu and colleagues’ study shows that supramolecular polymers are not immune to the laws of physics that prevail in other systems at different scales. For example, the entropic driving force that causes the condensation of the tactoids is also responsible for the condensation of liquid-crystal mesogens — nanometre-scale molecules that form liquid crystals. The authors’ findings therefore open up the possibility of using supramolecular polymers as models of liquid-crystal behaviour. Although biological fibres and rods, such as virus particles, microtubules and actin filaments, have also been shown to act like liquid-crystal mesogens at the micrometre scale13–18, they are difficult to modify. By contrast, there are many ways to modify the aspect ratio, chemical nature and chirality (handedness) of tactoids formed from synthetic supramolecular polymers, enabling them to be used as micrometre-scale models that closely relate to the molecular-scale liquid-crystal mesogens that are of interest to researchers. Future studies of the self-assembly of liquid crystals formed from supramolecular polymers 958 | Nature | Vol 626 | 29 February 2024

might reveal the hierarchical organization of the world, from the nano- to the macroscale. Jennifer L. Ross is in the Department of Physics, Syracuse University, Syracuse, New York 13244, USA. e-mail: [email protected] 1.

2. 3. 4. 5.

McManus, J. J., Charbonneau, P., Zaccarelli, E. & Asherie, N. Curr. Opin. Colloid Interface Sci. 22, 73–79 (2016). Peng, P.-H., Hsu, K.-W. & Wu, K.-J. Am. J. Cancer Res. 11, 3766–3776 (2021). Wang, B. et al. Sig. Transduct. Target Ther. 6, 290 (2021). Yoshizawa, T., Nozawa, R.-S., Jia, T. Z., Saio, T. & Mori, E. Biophys. Rev. 12, 519–539 (2020). Hyman, A. A., Weber, C. A. & Jülicher, F. Annu. Rev. Cell Dev. Biol. 30, 39–58 (2014).

6. Alberti, S., Gladfelter, A. & Mittag, T. Cell 176, 419–434 (2019). 7. Fu, H. et al. Nature 626, 1011–1018 (2024). 8. Boeynaems, S. et al. Trends Cell Biol. 28, 420–435 (2018). 9. Sahu, S. et al. PNAS Nexus 2, pgad231 (2023). 10. Xu, Y. et al. Adv. Mater. 33, 2008670 (2021). 11. Zbinden, A., Pérez-Berlanga, M., De Rossi, P. & Polymenidou, M. Dev. Cell 55, 45–68 (2020). 12. Ray, S. et al. Nature Chem. 12, 705–716 (2020). 13. Edozie, B. et al. Soft Matter 15, 4797–4807 (2019). 14. Weirich, K. L. et al. Proc. Natl Acad. Sci. USA 114, 2131–2136 (2017). 15. Oakes, P. W., Viamontes, J. & Tang, J. X. Phys. Rev. E 75, 061902 (2007). 16. Nyström, G., Arcari, M. & Mezzenga, R. Nature Nanotechnol. 13, 330–336 (2018). 17. Maeda, H. Langmuir 13, 4150–4161 (1997). 18. Dogic, Z. Phys. Rev. Lett. 91, 165701 (2003). The author declares no competing interests.

Evolution

Mobile DNA explains why humans don’t have tails Miriam K. Konkel & Emily L. Casanova

The lack of a tail is one thing that separates apes — including humans — from other primates. Insertion of a short DNA sequence into a gene that controls tail development could explain tail loss in the common ancestor of apes. See p.1042 Tails are a common feature in the animal kingdom, and all mammals have a tail at some point during embryonic development1. In humans, the tail disappears at the end of the embryonic phase — approximately eight weeks in utero2 — although internal parts remain in the form of the tailbone. The loss of the tail is considered a distinctive characteristic of apes and might have influenced our own upright walking style. On page 1042, Xia et al.3 report that the insertion of a type of mobile genetic sequence that moved around the genome during evolution, known as a transposable element, could be associated with the loss of the tail. Most monkeys have a tail, and tails were present at the origin of the primate lineage more than 65 million years ago4. In fact, the absence of a tail is one way to distinguish apes from monkeys. This characteristic, or phenotype, is shared across all apes, suggesting that tail loss coincided with, or occurred shortly after, the rise of apes after they diverged from their last common ancestor with monkeys around 25 million years ago. With this knowledge, Xia and colleagues compared prime candidate genes for tail loss across the genomes of several primate species, initially focusing on exons (the regions of DNA that code for proteins). When this approach turned out not to be fruitful, the authors extended their investigation to non-protein-coding regions that were

upstream or downstream of the genes, or in the genes themselves. The latter regions are known as introns, and they commonly interrupt protein-coding sequences. Xia and colleagues found that a type of primate-specific transposable element called an Alu element5 was inserted in an intron of the TBXT gene — but only in apes, and not in other primate lineages. TBXT is also known as Brachyury (meaning ‘short tail’) because mutations in the gene have been associated with short tails in several species, including Algerian mice (Mus spretus)6 and Manx cats (Felis catus)7. But how can the insertion of a short, roughly 300-base-pair Alu element into the non-coding sequence of a gene contribute to a tailless phenotype? To answer this, Xia and colleagues further scrutinized the entire TBXT gene and identified another Alu element oriented in the opposite direction (inverted) in intron 5. Because inverted Alu sequences in close proximity can pair up and create double-stranded RNA structures, the authors proposed that exon 6, which resides between the two Alu sequences, might be removed from the RNA transcript straight after transcription in a process called splicing (Fig. 1). To determine whether these two Alu sequences create alternatively spliced versions of TBXT RNA transcripts, Xia and colleagues took human and mouse embryonic stem cells and differentiated them so that

Alu element 4

Introns

Alu

5 Exons

Ape-specific Alu element 6

7

Alu

TBXT DNA

Transcription

Al

u

Alu

6

5 7

4

7

4

Alu

5

Alu

6

TBXT RNA

Removal of introns by splicing Mature TBXT RNA without introns

4

5

6

7

Full-length TBXT

4

5

7

Alternatively spliced TBXT without exon 6

Figure 1 | An ape-specific alternative RNA transcript of a tail-development gene. The TBXT gene is involved in tail development. Xia et al.3 compared TBXT sequences across primates, and identified a short mobile DNA sequence, known as an Alu element, that is inserted into a non-protein-coding region (intron) of TBXT in apes but not in other primates. When DNA is transcribed into RNA, the interaction of this Alu element with another nearby Alu element, which is not specific to apes and is oriented in the opposite direction, can lead to the removal of a protein-coding region (exon) during splicing, resulting in two possible versions of mature RNA — one that is full-length and one in which exon 6 is missing. Expression of this exonskipped TBXT might have contributed to tail loss as early apes evolved.

they mimicked the developmental state in which TBXT is expressed and implicated in tail development. Mouse cells, the genomes of which do not contain the primate-specific Alu insertions, expressed only the full-length Tbxt transcripts, but human cells expressed both the full-length transcript and a shorter transcript that did not include exon 6. Removing either Alu element using the gene-editing tool CRISPR–Cas9 resulted in an almost-complete loss of the transcript that lacked exon 6. To tie the altered transcripts to the tailless phenotype observed in apes, Xia et al. created several mouse lines. One mouse model simply had exon 6 of Tbxt deleted. To confirm that highly similar sequences oriented in opposing directions (reverse complementary sequences) could result in alternative splicing and therefore the skipping of exon 6, they also created ‘humanized’ mouse models by modifying the intron sequences flanking exon 6 of Tbxt. They did so by integrating either a pair of reverse complementary Alu sequences or a pair of reverse complementary mouse-specific sequences. The authors confirmed that mice lacking the functional, full-length Tbxt RNA transcripts did not survive to birth6,7, and found that mice with one intact copy of the gene and one altered copy had variable phenotypes, ranging from being tailless to having a full-length tail. However, the humanized mouse model with reverse complementary Alu sequences flanking exon 6 did not have a tailless or shortened-tail phenotype. This raises the question of whether tail loss is indeed solely driven by

the ape-specific Alu insertion, or whether other contributing factors are required. Intriguingly, inserting mouse-specific reverse complementary sequences did result in a shortened-tail phenotype in mice. Xia et al. made another exciting finding: all mice with one copy of Tbxt in which exon 6 was deleted and one copy with the insertion of the reverse complementary mouse-specific sequences lacked a tail. Together, the data support the role of the ape-specific Alu insertion in contributing to the tailless phenotype in apes. Furthermore, the authors observed that mice that expressed high levels of the exon-skipped gene transcript had an increased risk of defects in the embryonic structure that later forms the brain and spinal cord, known as the neural tube. Thus, the authors raise the possibility that tail loss in our ape ancestors might have come with the price of having an increased risk of neural tube defects. So, why did early apes lose their tails? Some researchers interpret the loss as adaptive, meaning it would have provided an evolutionary advantage. This is an idea that Xia et al. also engage with, echoing previous ideas that the loss of the tail contributed to improved two-legged (bipedal) locomotion. Research in human transitional species, such as Ardipithecus ramidus, suggests that bipedalism initially evolved in our tree-dwelling ancestors and was later used for a ground-dwelling lifestyle8. Scientists have tended to focus on adaptive explanations of tail loss and how it might enable human mobility, but several

lines of evidence suggest that having a tail does not hinder bipedalism, and could in fact support it. For example, tails seem to help to maintain posture in capuchin monkeys (Sapajus libidinosus) when they are transporting stone tools and walking bipedally9. Although humans regularly transport loads bipedally, robotics research suggests that a waist-mounted ‘tail’ can increase stability10, meaning that a tail could be adaptive even for modern humans. Physical isolation of primate populations offers an alternative explanation. Tectonic activity that began around 25 million years ago in East Africa, accompanied by volcanism, lake development and the reorganization of river networks, led to notable shifts in climate and the availability of resources. These geographical changes might have happened at the same time as the early apes started to evolve11. Early ape ancestors could have become isolated because of climate upheaval. With small population sizes, random genetic drift — such as the fixation of the Alu insertion reported by Xia et al. — could have played a larger part than did selection in tail loss12. Thus, altered function of the TBXT gene in early apes could have resulted from genetic drift in a small, reproductively isolated population, as an adaptive response, or both. Although the ultimate causality might remain unknowable, Xia and colleagues’ results offer a deeply compelling new chapter in the tale of our tail, and identify ways by which transposable elements can contribute to the diversification of the human repertoire of gene expression and, ultimately, typical human features. Miriam K. Konkel is in the Department of Genetics and Biochemistry, Clemson Center for Human Genetics, Clemson University, Clemson, South Carolina 29634, USA. Emily L. Casanova is in the Department of Psychology, Loyola University, New Orleans, Louisiana 70118, USA. e-mails: [email protected]; [email protected] 1. Mallo, M. Cell. Mol. Life Sci. 77, 1021–1030 (2020). 2. Gasser, R. F. Atlas of Human Embryos (Harper and Row, 1975). 3. Xia, B. et al. Nature 626, 1042–1048 (2024). 4. Chester, S. G. B., Williamson, T. E., Bloch, J. I., Silcox, M. T. & Sargis, E. J. R. Soc. Open Sci. 4, 170329 (2017). 5. Deininger, P. Genome Biol. 12, 236 (2011). 6. Herrmann, B. G., Labeit, S., Poustka, A., King, T. R. & Lehrach, H. Nature 343, 617–622 (1990). 7. Buckingham, K. J. et al. Mamm. Genome 24, 400–408 (2013). 8. Pattison, K. Fossil Men: The Quest for the Oldest Skeleton and the Origins of Humankind (William Morrow, 2020). 9. Massaro, L., Massa, F., Simpson, K., Fragaszy, D. & Visalberghi, E. Primates 57, 231–239 (2016). 10. Nabeshima, J., Sariji, M. Y. & Minamizawa, K. ACM SIGGRAPH 2019 Posters, 52 (2019). 11. Stevens, N. J. et al. Nature 497, 611–614 (2013). 12. Whitlock, M. C. Evolution 54, 1855–1861 (2000). The authors declare no competing interests.

Nature | Vol 626 | 29 February 2024 | 959

News & views Sociology

Online images are more gender-biased than text Bas Hofstra & Anne Maaike Mulders

A big-data analysis shows that men are starkly over-represented in online images, and that gender bias is stronger in images compared with text. Such images could influence enduring gender biases in our offline lives. See p.1049 Images are an increasingly important vehicle for circulating information online and for grabbing and keeping people’s attention1. On page 1049, Guilbeault and colleagues2 detail how online images reflect and exaggerate existing gender bias found in the offline world. The authors googled a wide variety of social categories (such as ‘lawyer’, ‘doctor’, ‘neighbour’, ‘friend’ and ‘cousin’) and documented that, across these social categories, the number of images representing men was larger than the number representing women compared with similar social categories in other media. Furthermore, the association of these categories with a particular gender (for example, women and ‘art teacher’ versus men and ‘carpenter’) is much stronger in online images compared with online texts, and exposure to these online images can affect individuals’ measures of unconscious (implicit) bias for several days. The findings support the idea that production of online content, in combination with the dynamics of people’s online behaviours, might reproduce and even exacerbate offline patterns related to inequity and a lack of diversity. Because people spend a significant portion of their lives online, these dynamics seem striking. Guilbeault and colleagues compiled a large data set of 349,500 images — 100 images extracted for each of the 3,495 social categories they studied — and employed thousands of human coders to capture whether those images leaned towards male or female representation (excluding images found of non-binary people). Men seemed to be starkly over-represented in images compared with their representation in a range of other sources (such as in texts, in public opinion and in a census describing employment statistics). What makes the results compelling is that they were not affected by the country from which the images were searched for, suggesting that male over-representation is not unique to Google searches performed in specific countries. Furthermore, the authors replicated their results using images extracted from 960 | Nature | Vol 626 | 29 February 2024

different sources, such as Wikipedia and the Internet Movie Database (IMDb). This suggests that the observed male over-representation and gender bias in images are consistent and generalizable across other widely popular online platforms. The authors also compared the magnitude of gender bias in images and text by assessing to what extent 2,986 social categories co-occur with references to women or men in various online texts. Using several models to analyse text from a range of sources, they observed a similar association between gender and social categories in text as in images. The gender associations — that is, how each gender is represented in each social category — correlate strongly across both types of media. This suggests that gender associations for the social

Cosmetologist Beauty consultant Model Hairstylist Nurse practitioner Interior decorator Art teacher Flight attendant Cook Legal assistant Programmer Mechanic Military officer Heart surgeon Mathematician Investment banker Police chief Football player Blacksmith Plumber

category ‘art teacher’, for example, are comparable in both images and texts. Yet, the authors found that gender associations with certain categories are much more extreme in images than in text (Fig. 1). For example, observing female ‘art teachers’ in online images was more likely than reading about female ‘art teachers’. Perhaps this finding is not wholly unexpected because, as the authors note, images portraying people naturally signal some demographic information, whereas text can be written in a way that minimizes information on specific demographics. So why does all of this matter? Guilbeault et al. argue that people process images more quickly and images are more memorable than text. Therefore, if online images convey such strong gender associations, they could influence people’s beliefs about gender. To test this conjecture, Guilbeault and colleagues conducted a preregistered experiment in which 450 participants from the United States were organized randomly into three groups: participants searched for specific occupations using either Google News or Google Images, or for phrases that do not describe occupations (such as ‘guitar’) using Google (the control group). Afterwards, the participants completed tasks designed to measure unconscious bias in relation to gender (an ‘implicit association test’). The authors found that the participants who searched for images exhibited stronger gender bias compared with those who searched text or who were in the control group. Most strikingly, gender bias seemed to endure for up to three days after the experiment, but only in those

Text

Images

100% female

0 Gender association

100% male

Figure 1 | Gender associations with social categories in online images compared with online text. Guilbeault et al.2 searched for 2,986 social categories, such as occupations, using Google Images (for images) or Google News (for text) and measured the frequency with which each social category was associated with one or another gender in the search results. They found that gender association was more extreme for online images than for online texts. The graph shows a selection of occupations across the range of gender association scores. (Figure adapted from Fig. 1a of ref. 2.)

who searched for images. These results suggest that online images, and hence the online realm, are not only highly gendered, but that this gendered nature might also influence further gender bias in everyday life. Previous work shows that exposure to stereotype-confirming images negatively affects women’s self-esteem and hampers their leadership aspirations, suggesting that gender-biased images can establish and reinforce gendered career choices3. Repeating the current study’s measurements of unconscious bias using social categories other than occupations (for example, ‘cousin’) would enable a further exploration of the consequences of strong gender associations in online images. The authors’ conclusions might also be strengthened by conducting the implicit association test with more participants and in different countries, or by further examining conscious (explicit) gender bias. There are several lingering questions that are essential for future studies to address. What are the exact mechanisms that cause the Internet to become such a gendered environment with respect to online images? Could it be related to particular populations of Internet

users, certain design choices or the transfer of existing offline imagery to websites? Once the exact mechanisms are known, what interventions could be put in place to ameliorate those dynamics? Answering these questions is imperative in an age in which images generated by artificial intelligence (AI) will probably become highly prevalent and widespread on the Internet. If these AI-generated images are based on online images that are already gendered, imagery found on the Internet might spiral into becoming increasingly gender-biased. Bas Hofstra and Anne Maaike Mulders are in the Department of Sociology, Radboud University, Nijmegen 6525 GD, the Netherlands. e-mail: [email protected]; [email protected] 1.

Zhang, L. & Rui, Y. ACM Trans. Multimedia Comput. Commun. App. 9, 36 (2013). 2. Guilbeault, D. et al. Nature 626, 1049–1055 (2024). 3. Simon, S. & Hoyt, C. L. Group Proc. Intergroup Rel. 16, 232–245 (2013). The authors declare no competing interests. This article was published online on 14 February 2024.

Biomedical engineering

Light can restore a heart’s rhythm

From the archive Stephen Hawking proposes that black holes can explode, and William H. Bragg reflects on the perseverance of scientists.

50 years ago Quantum gravitational effects are usually ignored in calculations of the formation and evolution of black holes … Even though quantum effects may be small locally, they may still … add up to produce a significant effect over the lifetime of the Universe …  [I]t seems that any black hole will create and emit particles such as neutrinos or photons … As a black hole emits this thermal radiation one would expect it to lose mass. This in turn would increase the surface gravity and so increase the rate of emission. The black hole would therefore have a finite life … For a black hole of solar mass this is much longer than the age of the Universe. There might, however, be much smaller black holes which were formed by fluctuations in the early Universe. Any such black hole … would have evaporated by now. Near the end of its life the rate of emission would be very high and about 1030 erg would be released in the last 0.1 s. This is a fairly small explosion by astronomical standards but it is equivalent to about 1 million 1 Mton hydrogen bombs. From Nature 1 March 1974

Igor R. Efimov

100 years ago

Implantable electric pacemakers save millions of lives worldwide, but they aren’t perfect. A proof-of-concept study shows that using light to regulate a heartbeat might be a better option than existing strategies. See p.990

In what way do we hope to benefit by research? … There is so much … work to be done before the good observations come; it may be that weeks are spent in preparation and five minutes in making the actual measurement. It is all very humiliating; and the blunders one makes are very foolish … [O]ne redeeming feature is that … there is always the hope … every student … who strives to understand the workings of Nature by experiment … is paid by … discovery of a richer world. There is a fellowship between all who have tried to understand … [W]e must research, and with all our energy … [T]he spirit of research is like the movement of running water, and the absence of it like the stagnation of a pool. Scientific research, in its widest sense, … is an act of faith in the immensity of things. From Nature 1 March 1924

Life starts with a heartbeat and ends without it. This regular rhythm is set by the body’s natural pacemaker: a collection of cells known as the sinus node1. When this node fails, cardiologists can implant an electric pacemaker to stimulate a person’s heart back to a normal rate2. But standard pacemakers are powered by electrochemical batteries that have a limited life, and the devices are prone to electrode failure and interference from external electromagnetic fields3. On page 990, Li et al.4 present a technique that uses the energy from light to stimulate the heart, which could offer a solution to these problems. Heart muscle consists of cells that interact through chemical, mechanical and electrical signalling systems. The electrical coupling allows the entire muscle to be excited by enabling an electrical signal to spread rapidly from a single stimulation point. The other types of

coupling have offered inspiration for alternatives to the standard electric pacemaker. Researchers have investigated the possibility of targeting specific proteins that can transmit the required signal mechanically or through changes in temperature or light. These sensors could be triggered non-invasively by light or ultrasound5,6. However, despite considerable efforts, these approaches are yet to yield clinically viable therapies. One obstacle is that both natural and genetically engineered molecular sensors are not sufficiently sensitive. Another problem is that existing devices are not sophisticated enough to interface well with human tissue. It has also proved difficult to achieve targeted and stable delivery of the genetically engineered molecular sensors to the heart. An alternative tactic involves implanting biocompatible photoelectrochemical devices,

Nature | Vol 626 | 29 February 2024 | 961

News & views a Conventional pacemaker

b Photoelectric pacemaker Pacemaker

Laser

Right atrium Left ventricle Right ventricle

Silicon device

Electrode

Figure 1 | Improving pacemaker design with light. a, In the best conventional pacemakers currently available, the device is connected to electrodes that pace contraction of the right atrium or the right and left ventricles using electrical stimuli. This strategy can lead to asynchronous contraction, and can otherwise fail owing to a limited battery life and the unreliability of electrodes. b, Li et al.4 propose an alternative pacemaker that can stimulate the heart at several sites non-invasively using silicon, which is photoelectric, meaning it can convert light energy from a laser or a light-emitting diode into an electric current. The authors’ proof-of-concept design successfully paced a pig’s heart in vivo.

which convert energy from light into an electric current. More than a decade ago, a silicon-based photoelectrochemical device was proposed as a prosthesis for people with damaged retinas7. Members of the same research group as Li et al. then came up with a set of design principles for interfacing such devices with various biological targets8. In the present work, Li et al. produced a proof-of-concept device that uses laser light to generate an electric current in a silicon device that is designed to be implanted at the surface of the heart muscle. The authors tested the device on isolated heart cells, as well as on an intact heart that had been removed from a rat. They then showed that they could use the device to stimulate a mouse’s heart in vivo. Finally, Li et al. demonstrated that their device could reliably pace a pig’s heart, either during open-heart surgery or after an endoscopic operation. As part of the characterization of their device, Li et al. mapped the 3D distribution of the electric current that was generated below the surface of the heart muscle. The current’s effect on the heart can be understood by quantifying an ‘activating function’ that stimulates the heart muscle through a set of virtual electrodes9–11. The authors’ 3D map enables calculation of this activating function, which can then be used to optimize the energy and waveform characteristics of the light required for clinical application. One of the key challenges of implantable pacemakers is that they can cause the different parts of the heart to contract out of sync. Unlike a healthy heart, which is stimulated by the body’s conduction system, a heart that is controlled by a pacemaker is stimulated by 962 | Nature | Vol 626 | 29 February 2024

electrodes that are usually implanted in the right atrium or ventricles (Fig. 1a). Although the electrical signal travels rapidly through the whole organ, this localized stimulation can result in asynchronous excitation, leading to out-of-sync contractions. Various resynchronization devices add a left ventricular electrode, preventing and mitigating heart failure in such cases12. However, these devices either target only a few sites in the heart or they require precise implantation of one electrode at a specific point in the conduction system. Li et al. say that this problem can be overcome by stimulating the heart at several sites using a network of their silicon devices (Fig. 1b). However, it is unclear how these devices will be joined together, and how they might be implanted in a beating heart.

“Implantable pacemakers can cause the different parts of the heart to contract out of sync.” There are other engineering challenges to be surmounted. First, the heart is surrounded by a layer of fat. Will implantation involve penetrating this fat and anchoring the device to the muscle? Or will the device and its light source be delivered through a vein to the heart’s interior? Second, the implantation of a foreign body might induce a physiological response, which could lead to Li and colleagues’ devices being encapsulated with fibrous tissue, as was the case for early implantable electric

pacemakers13. Third, and perhaps most importantly, it is not yet clear how the authors plan to deliver light to the heart, to which access is complicated by the rib cage and lungs. In the authors’ proof-of-concept experiments, the hearts were exposed, providing easy optical access to the heart surface. This makes it possible to stimulate several silicon devices with light. But a heart is usually covered with many layers of tissue, which scatter light. For this reason, the light used in Li and colleagues’ strategy would either have to be generated by light-emitting diodes in the devices themselves or be directed to the heart from a remote source through light guides14. Both approaches have the same problem as conventional electric pacemakers: the devices can be rejected by the body, or otherwise fail. Ideally, if there is a way of stimulating Li and colleagues’ devices through the rib cage, external light sources could be incorporated into a wearable device. However, the device’s sensitivity might prove insufficient for this solution. With all of these hurdles still to be cleared, it is difficult to predict when or whether photoelectric stimulators, such as the one designed by Li and colleagues, will prove to be more robust than conventional electric pacemakers. However, the authors’ exciting proof of concept shows the enormous potential that the technology holds, and suggests that photoelectric devices could eventually transform a range of therapies, including those requiring neural, muscular and cardiac stimulation. Igor R. Efimov is in the Department of Biomedical Engineering and the Department of Medicine (Cardiology), Northwestern University, Chicago, Illinois 60611, USA. e-mail: [email protected]

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

Fedorov, V. V. et al. J. Am. Coll. Cardiol. 56, 1386–1394 (2010). Parsonnet, V., Zucker, I. R., Gilbert, L. & Asa, M. M. Am. J. Cardiol. 10, 261–265 (1962). White, W. B. & Berberian, J. G. Emerg. Med. Clin. North Am. 40, 679–691 (2022). Li, P. et al. Nature 626, 990–998 (2024). Hsueh, B. et al. Nature 615, 292–299 (2023). Lee, K. L. et al. J. Am. Coll. Cardiol. 50, 877–883 (2007). Mathieson, K. et al. Nature Photon. 6, 391–397 (2012). Jiang, Y. et al. Nature Biomed. Eng. 2, 508–521 (2018). Sobie, E. A., Susil, R. C. & Tung, L. Biophys. J. 73, 1410–1423 (1997). Sepulveda, N. G., Roth, B. J. & Wikswo, J. P. Biophys. J. 55, 987–999 (1989). Efimov, I. R., Cheng, Y., Van Wagoner, D. R., Mazgalev, T. & Tchou, P. J. Circ. Res. 82, 918–925 (1998). Chung, M. K. et al. Heart Rhythm 20, e17–e91 (2023). Parsonnet, V., Zucker, I. R., Kannerstein, M. L., Gilbert, L. & Alvares, J. F. J. Surg. Res. 6, 285–292 (1966). Gutruf, P. et al. Nature Commun. 10, 5742 (2019).

The author declares no competing interests. This article was published online on 21 February 2024.

Review

Ion and lipid orchestration of secondary active transport https://doi.org/10.1038/s41586-024-07062-3

David Drew1 ✉ & Olga Boudker2,3 ✉

Received: 27 June 2023 Accepted: 12 January 2024 Published online: 28 February 2024 Check for updates

Transporting small molecules across cell membranes is an essential process in cell physiology. Many structurally diverse, secondary active transporters harness transmembrane electrochemical gradients of ions to power the uptake or efflux of nutrients, signalling molecules, drugs and other ions across cell membranes. Transporters reside in lipid bilayers on the interface between two aqueous compartments, where they are energized and regulated by symported, antiported and allosteric ions on both sides of the membrane and the membrane bilayer itself. Here we outline the mechanisms by which transporters couple ion and solute fluxes and discuss how structural and mechanistic variations enable them to meet specific physiological needs and adapt to environmental conditions. We then consider how general bilayer properties and specific lipid binding modulate transporter activity. Together, ion gradients and lipid properties ensure the effective transport, regulation and distribution of small molecules across cell membranes.

Small molecule transporters catalyse the translocation of diverse solutes from ions to large organic molecules across cell membranes. Many small molecule transporters belong to the solute carrier (SLC) transporter superfamily, which in human constitutes the secondlargest fraction of the membrane proteome after G-protein-coupled receptors1,2. Here we focus on the ion-coupled SLC transporters, also known as secondary active transporters, which utilize established ion gradients to catalyse concentrative transport. This is in contrast to passive SLCs or primary active transporters, which are driven mostly by ATP hydrolysis. In general, SLCs are essential for cell homeostasis—more than a third of SLCs are associated with inherited diseases—and they can influence drug pharmacokinetics or are drug targets themselves1,2. Historically, SLCs have received less attention from pharmaceutical companies than G-protein-coupled receptors or membrane channels. However, this is changing in the era of precision medicine, and recent success stories in targeting SLCs, such as drugs for type 2 diabetes that inhibit the kidney sodium-coupled glucose transporter 2 (SGLT2)3, have increased interest in this area. Transporters can be viewed as enzymes that follow Michaelis– Menten kinetics4. They cycle through multiple conformational states to translocate their substrate molecules by providing alternating access to the substrate-binding site from one side of the membrane at a time5,6. This complexity contrasts with ion channels, which allow solutes to diffuse rapidly through open pores. Transporter working cycles are defined by multiple conformational states, perhaps explaining why the structural and mechanistic understanding of SLCs has lagged behind that of channels and receptors. The rise of single-particle cryo-electron microscopy has facilitated the structural description of SLCs and, combined with detailed kinetic, structural dynamics and computational analyses, has revealed common mechanistic themes. Here we discuss and compare the Na+- and H+-coupling mechanisms that drive substrate translocation. We highlight examples of strict and flexible ion coupling

and allosteric ion-mediated regulation with interesting physiological ramifications. We outline how lipids shape oligomerization, switch transport activity on and off, and regulate conformational dynamics. We conclude that ion coupling and lipid regulation in transporters is highly varied and intricate, and that unravelling these mechanisms is critical for understanding their physiological roles and guiding the development of effective drugs.

Ion coupling in transporters The basics The functional units of secondary active transporters are almost always formed by two domains. The domains either rock around a centrally located substrate in ‘rocker-switch’ and ‘rocking-bundle’ proteins, or one of the domains carries the substrate across the membrane, as in ‘elevator’ proteins7 (Fig. 1a). In rocker-switch proteins, the two domains are structurally symmetric7, whereas in rocking-bundle proteins, they are asymmetric. Alternating-access models refer to global conformational rearrangements between the outward-facing and inward-facing states6. Intermediate occluded states with no solvent access to the substrate-binding site occur on paths between open outward-facing and inward-facing states7 (Fig. 2). Transmembrane helices and loops proximal to the substrate usually restructure to form the occlusion, serving as transporter gates. Many secondary active transporters have transmembrane segments with non-helical breaks, which go halfway across the membrane before turning back (re-entrant loops and helical hairpins) or continue across as helix–break–helix segments. Non-helical breaks expose backbone carbonyl oxygen atoms, ideal for cation coordination. Secondary active transporters probably evolved from smaller proteins with fewer transmembrane segments, which oligomerize to form functional units8. In most modern transporters, these domains have fused into single polypeptides, yet they contain structural repeats

Science for Life Laboratory, Department of Biochemistry and Biophysics, Stockholm University, Stockholm, Sweden. 2Department of Physiology and Biophysics, Weill Cornell Medicine, New York, NY, USA. 3Howard Hughes Medical Institute, Weill Cornell Medicine, New York, NY, USA. ✉e-mail: [email protected]; [email protected]

1

Nature | Vol 626 | 29 February 2024 | 963

Review a

Rocker-switch

Substrate

Outward facing

Inward facing

Rocking-bundle Outward facing

Inward facing

Elevator Outward facing

b

Inward facing

Antiport

Symport Driving ion

Substrate

[S]in [S]out

=

Uniport

n [ion]out

[ion]inn

(

× exp – zFV RT

)

[S]in [S]out

=

n [ion]in n [ion]out

(

× exp – zFV RT

)

[S]in [S]out

(

= exp – zFV RT

)

Fig. 1 | General principles of ion-coupled alternating-access mechanisms. a. Secondary active transporters are classified on the basis of the global transitions that provide alternating access to the substrate-binding site. Top, rocker-switch transporters comprise two structurally similar domains (light and dark grey ellipsoids). The substrate binds between the domains, catalysing their rocking between outward-facing and inward-facing conformations. Middle, in rocking-bundle transporters, the domains are structurally dissimilar. The bundle domain (dark grey ellipsoid) rearranges around the substrate against a more rigid scaffold domain (light grey ellipsoid). Bottom, elevator transporters show the most structural divergence between the domains. The transport domain (dark grey ellipsoid) translocates the substrate by moving against the rigid scaffold domain (light grey ellipsoid), which is often shorter than the transport domain and stabilized by oligomerization. Intermediate occluded states between the outward-facing and inward-facing states are not shown. The gating structural elements rearrange upon binding of

ions and substrates and contribute to their collective coordination and occlusion (grey rectangle). Gating elements are more extensive in rockingbundle and elevator transporters than in rocker-switch transporters. Part a was adapted from ref. 7. b. Ions moving down their electrochemical gradient drive concentrative substrate uptake or export according to the Nernst equations shown below the panels. S is the substrate, n is the number of transported ions, z is the total charge moving across the membrane, V is transmembrane voltage, F and R are the Faraday and universal gas constants, respectively, and T is the absolute temperature. Left, in symporters, the driving ion(s), usually Na+ or H+, move in the same direction as the substrate. After the substrate and ions are released, the transporter spontaneously resets to its outward-facing state. Middle, in antiporters, binding of either solute or ion(s) is required to catalyse transitions between outward-facing and inward-facing states, ensuring their stoichiometric exchange. Right, in uniporters, the substrate equilibrates according to its electrochemical gradient.

composed of several transmembrane helices with a parallel or inverted topology9. Moreover, transmembrane helices that coordinate and gate substrate and ions in outward-facing and inward-facing states are often related by inverted symmetry9.

Secondary active transporters mediate vectorial substrate transport using the transmembrane ion concentration gradient and electrical potential according to the Nernst equation (Fig. 1b). If the transport reaction results in a net transmembrane charge movement, the

964 | Nature | Vol 626 | 29 February 2024

a

Outward-open

Outward-occluded

Occluded (substrate)

Occluded (empty)

Inward-open

b

Substrate H+

Outward-occluded (empty)

Inward-occluded

Outward-open

Outward-occluded Substrate Na+

Occluded (empty)

Occluded (substrate)

Inward-occluded (empty)

Inward-open

Inward-occluded

Fig. 2 | The major conformations of rocker-switch and elevator symporters. a, In this hypothetical rocker-switch symporter protein, H+ binding to the proton carrier residue in the outward-open state allows the substrate to bind. Accompanying rearrangements of the gating helices occlude the substrate from the bulk solvent in the outward-occluded state. The binding of the ion and substrate catalyses the rocker-switch transition to a substrate-occluded and inward-occluded state in a process that typically breaks the inter-domain salt bridges on the cytoplasmic side of the transporter. In the inward-occluded state, the often lower substrate affinity and decreased proton carrier residue pKa favour substrate and H+ release, and coupled rearrangement to the inward-open state. The symporters spontaneously reset to the outward-open state via an empty-occluded intermediate, frequently driven by the reformation

of the inter-domain salt bridge. In rocker-switch proteins, empty inwardoccluded and empty outward-occluded states are seldom resolved structurally, as they are probably too transient. b, A hypothetical elevator transporter undergoes similar conformational transitions to the rocker-switch scheme in a, and in this example it is coupled to the binding of two sodium ions and substrate. One difference is that the transporter gates—the structural elements that coordinate and occlude the substrate and ions—are structurally more pronounced. They need to close in the unloaded transporter to allow the resetting elevator motions after the substrate and ions are released. Empty inward-occluded and empty outward-occluded states are added to the elevator transport cycle because they can be long-lived and observed experimentally.

transport is electrogenic and is driven by both concentration differences and electrical potential; if there is no net charge movement, the electroneutral transport is driven solely by concentration gradients. Coupled ions are either co-transported with the substrate or counter-transported—referred to as symport and antiport, respectively. Uniporters are passive transporters driven solely by the electrochemical gradient of their substrate. H+ and Na+ ions are best suited to drive transport because they usually show inwardly directed concentration gradients and experience favourable, negative-inside cell membrane potentials. They provide energy in the form of the proton-motive force (PMF) and the sodium-motive force (SMF) as they move down their concentration gradient in the electric field. Other ions, such as K+ and Cl−, are used less frequently as the primary driving ions, probably because the resting membrane potentials are closer to their thermodynamic equilibria across cell membranes, and therefore their electrochemical gradients are smaller. Cells use constitutively open and regulated

ion channels and primary active transporters to maintain these ion distributions and electrical potentials. It has been suggested that natural H+ gradients in deep-sea vents might have energized primordial living cells that emerged around 3.8 billion years ago10. However, the membranes of these cells were composed of fatty acids rather than more modern lipids, and would have been leaky to H+, necessitating conversion of PMF to SMF as a ‘battery’, because these membranes would have been less permeable to Na+ ions. Consistent with this idea, bioinformatic analysis suggests that Na+/H+ antiporters are among the most ancient proteins, predating the evolutionary split between bacteria and archaea11. Modern prokaryotes evolved to use either SMF or PMF, depending on their cellular environment. Those living in high-salinity or alkalinity habitats favour a SMF, whereas those encountering scarce environmental Na+ are more likely to rely on a PMF. Fungi and plants use PMF12,13 exclusively, whereas animal cells use both PMF and SMF. Nature | Vol 626 | 29 February 2024 | 965

Review Substrate translocation in most rocker-switch proteins is protoncoupled14, with H+ binding typically being a pre-requisite for the local rearrangement of gating helices, which mediate substrate occlusion and release (Fig. 2a). However, some rocker-switch proteins, including members of the major facilitator superfamily (MFS) and the architecturally distinct multidrug and toxic compound extrusion (MATE) transporters, use Na+ or have promiscuous ion-binding sites that can use either ion15–17. Rocking-bundle proteins and elevator transporters often feature more complex ion- and substrate-gated rearrangements underlying the symport and antiport of multiple ions, and highly specific ion-binding sites (Fig. 2b). These transporters exhibit structural asymmetry, with one labile domain—usually termed ‘bundle’ in rocking-bundle proteins and ‘transport’ in elevator proteins— that moves against a less labile domain, referred to as ‘scaffold’7,18 (Fig. 1a). Such specialized bundle and transport domains may enable more elaborate coupling mechanisms than the simpler symmetric domains of rocker-switch proteins.

Symport mechanisms Simplistically, H+ coupling only requires an ionizable residue to bind and release H+ at a near-physiological pH (Fig. 3a). Although histidine, with an intrinsic pKa of 6.0, seems to be the only suitable residue, other charged residues can have markedly shifted pKa values when buried in protein interiors. NMR studies have shown that the pKa of buried lysine19, for example, can be shifted from 10.4 down to 5.3, and the pKa of glutamate can be in the 7 to 8.5 range, far above its intrinsic pKa of 4.1 (refs. 20,21). Arginine is the only charged amino acid whose pKa is always too high for H+ coupling22. Na+ is usually coordinated by four to six electronegative oxygen atoms in the side chains of aspartate, glutamate, serine, threonine, asparagine, glutamine and tyrosine, the sulfur of methionine, and main-chain carbonyl oxygens23. Conserved ionizable residues in the middle of transmembrane segments are obvious candidates for ion coupling, because the high energetic costs of their placement into the hydrophobic membrane environment suggests their functional necessity24. For efficient symport, substrate and ions must be transported together and never, or almost never, alone (Fig. 1b). One symport mechanism relies on the thermodynamically coupled binding of ions and solutes. In the H+-coupled symporters, ionizable groups can be part of the substrate-binding site or allosterically linked to it. They must be protonated before substrates can bind25–28 (Fig. 3a). For instance, structural studies, biochemistry and molecular dynamics simulations, have concluded direct H+–substrate interactions in the plant sucrose transporter SUC1 and H+-coupled oligopeptide transporters, in which protonation of an acidic residue is essential for substrate binding and uptake29,30. In addition, H+ coupling often entails the reorganization of connected, local electrostatic networks27,31. Thus, in some oligopeptide transporters, protonation of the substrate-binding glutamate25 requires the protonation of an extracellular histidine first; however, this proton is not transported. In many cases, H+-binding site residues form salt bridges, and protonation and substrate binding trigger their breakage and formation of new salt bridges27,30,32–36. Direct interactions between the substrate and coupled Na+ ions appear infrequent but occur, for example, in a prokaryotic neurotransmitter: Na+ symporter (NSS) homologue LeuT at the Na1 site37 and in the sodium-driven chloride/bicarbonate exchanger38 (NDCBE), in which the substrates coordinate Na+ ions together with the transporter side chains (Fig. 3b). Structural studies have shown that allosteric interactions between ion-binding and substrate-binding sites are more common than direct interactions (Fig. 3a,b). For example, in the rocker-switch H+-sugar symporter LacY, the main proton-binding site (E325) does not directly coordinate the sugar. Instead, it interacts with and properly positions a nearby histidine residue (H322), which binds the sugar27,39. In the H+-coupled sugar symporter XylE, a close homologue of the human sugar uniporter GLUT1 (refs. 40,41), no ionizable groups coordinate the 966 | Nature | Vol 626 | 29 February 2024

substrate40, and the proton carrier aspartate (D27) in transmembrane helix 1 (TM1) is located 10 Å from the binding site31. Deprotonated D27 is salt-bridged to an arginine residue, which restricts the mobility of TM1 (ref. 14). Protonation of D27 releases this latch and allows TM1 to move towards TM7b on the opposite side of the substrate-binding vestibule14. TM7b is critical for sugar binding and forms a lid over the substrate-binding site40, which closes upon interaction with TM1 (refs. 14,42). Thus, occluding the substrate cavity only when H+ binds in the gating helix and sugar binds in the cavity ensures their coupled symport (Fig. 3a). A similar H+-coupling mechanism, which relies on a protonation-dependent salt bridge to release a substrate-coordinating arginine, has been proposed for the galactonate transporter DgoT, a homologue of human vesicular glutamate (VGLUT) transporters32. Further, in structurally unrelated elevator glutamate transporters termed excitatory amino acid transporters (EAATs), protonation of a glutamate residue in TM7 helps release and position an arginine in TM8 for substrate coordination43,44. The protonated glutamate hydrogen bonds to a main-chain carbonyl oxygen in the helical hairpin of the transporter gate, promoting its closure, substrate occlusion and elevator transitions43,44. Allosterically coupled binding also underlies symport in archaeal glutamate transporter homologues, the aspartate-Na+ symporters GltPh and GltTk, which bind aspartate and three symported Na+ ions in a cooperative manner45–47, with amino acid affinity scaling approximately with the cube of the Na+ concentration. Na+ binding alters the local network of hydrogen bonds and salt bridges to yield a high-affinity binding site configuration44,48–50. The transporters exhibit high affinity in the outward-facing state, where Na+ concentration is high, and low affinity in the inward-facing state, where Na+ concentration is low. Thus, the transporter is more likely to bind aspartate in the outward-facing state and release it in the inward-facing state, and the probability of leaking aspartate—that is, transporting it without Na+—is low. Kinetic studies on EAATs using electrophysiology51 and purified GltPh (ref. 52) have shown that two Na+ ions bind before the substrate, whereas the third Na+ ion (Na2) is in rapid equilibrium between being bound and free. Only when substrate and Na2 are bound does the transporter gate close and elevator transition occurs. Coupled ions bind weakly in the absence of substrates, but because their concentration in the environment can be high, they might substantially occupy these sites without substrate. However, their binding is allosterically coupled to the opening of the transporter gate, which creates a large energetic barrier for translocation53–55. Similar allosteric coupling between Na+ and transporter gates exists in other families, such as the rocking-bundle SGLTs56 and LeuT57, and the rocker-switch MelB58. Ion-bound outward-facing conformations are frequently open and readily accessible for substrate binding that catalyses gate closure and transition to intermediate, occluded states58. Coupled binding is not a universal mechanism of symport as, for instance, serotonin binds to the serotonin transporter SERT in a Na+-independent manner59. In this case, binding the full complement of symported serotonin, two Na+ ions and a Cl− ion catalyse the transition from the outward-facing to inward-facing conformation (Fig. 3b). The H+-sugar symporter LacY is an interesting case, in which lactose binding to the transporter is essentially pH-independent because the pKa of its proton carrier glutamate is around 10.5, and is therefore immediately protonated in the outward-facing state60. Sugar binding to the protonated transporter catalyses the rocking-switch transition, enabling translocation, and the proton must be released in the inward-facing state before spontaneous resetting to the outward-facing conformation61. In addition to variations in the order of ion and substrate binding and release during the transport cycle, the tripartite ATP-independent periplasmic (TRAP) transporters are an interesting deviation in secondary active transporters, as they depend on a soluble solute-binding periplasmic (P)-subunit 62. The membrane component of TRAP is topologically and mechanistically similar to those of elevator

a Proton-coupled symporters Direct substrate–ion interactions Ionizable residue

Distant residues

Substratebinding site

PepTso

SUC1

Substrate H+

Allosteric coupling

XylE

DgoT

Proton-coupled antiporters Ping-pong mechanism

Allosteric competition

MdfA, NapA

b

NorM LeuT

SERT E290

S372 D98 Cl Na2 Na1

Leu Na1 Na2 5HT

Scaffold

c

GltPh apo

EAAT3 apo

Q318 R397

Core

E374 R447

Scaffold EAAT3

R447 K+

Core K+

E374

EAAT3 3Na+, H+, Glu–

R447 H+

E374

Glu

Na2 Na1

Na3

Fig. 3 | Examples of ion-coupling mechanisms. a, Top left, in some H+-coupled symporters, the substrate-binding site contains an ionizable residue(s) that directly coordinates the substrate when protonated. The protons can reach the proton carrier residue directly (left) or are relayed by other charged groups or waters (right). Top right, H+ binding to distant proton carrier residue(s) can be allosterically coupled to substrate transport by triggering movement of gating helices. Protons reach the distal proton carrier directly (left) or are relayed by charged residues or waters (right). Bottom left, in some H+-coupled antiporters, proton binding catalyses the transition from outward-facing to inward-facing states. The protonated residue releases H+ to bind the substrate, catalysing the transition from inward-facing to outward-facing states. In some cases, the direct competition between substrates and H+ for the same site is termed a ping-pong mechanism. Bottom right, in others, H+ and the substrate bind at distant allosterically coupled sites. When one binds, the affinity for the other decreases, triggering its release. In the multidrug transporter NorM, H+ or Na+ bind to a promiscuous site in one domain, and another H+ can bind to the other domain: the final ion-coupled stoichiometry is flexible and depends on the

charged state of the drug. b, Left, coupling to Na+ symport can occur through direct interactions with substrates. For example, the substrate carboxyl in LeuT (Protein Data Bank (PDB) ID: 2A65) coordinates Na1. Na2 is at a distant site, where the ion restructures protein for substrate binding and transport. Right, in the homologous SERT (PDB: 7AIL), serotonin (5HT) does not coordinate Na1; instead, this is achieved via the carboxylate of D98. A serine, instead of E290, completes a Cl−-binding site. Thus, the Na+ sites are conserved, and coupling is expanded to include a Cl−. c, Left, as a symporter coupled to three Na+ ions, apo GltPh (PDB: 4P3J) spontaneously resets from inward-facing to outward-facing states after releasing its cargo. R397 and Q318 stabilize the occlusion by the gating helical hairpin 2 (red). In the homologous human EAAT3, E374 replaces Q318. E374 forms a salt bridge with R447 in apo EAAT3 (PDB: 8CUI) (second left), preventing proper hairpin occlusion until K+ binds (PDB: 8CUA) (third left). Right, the Q318 to E374 mutation expands coupling from three Na+ to an additional symported H+, carried by E374 (PDB: 8CTC) and an antiported K+.

dicarboxylate-Na+ symporters63,64. The difference is that TRAP transporters cannot bind their substrates directly, even when Na+ ions are bound in the inward-facing state63. Instead, the substrate can only be accepted from the soluble P-subunit, and only once the P-subunit has attached to the scaffold domain of the elevator transporter63,64. It is unclear whether the binding of the P-subunit triggers the transition from inward-facing to outward-facing state, or if it rearranges the outward-facing state to accept the substrate63. The P-subunit shows very high substrate affinity in the nanomolar range, and accumulates

nutrients that are scarce in the environment65. Its involvement abrogates the need for tight binding to the transporter itself, which would have otherwise resulted in a slow off-rate, which is incompatible with fast transport.

Antiport mechanisms In symporters, spontaneous resetting of the transporter from the inward-facing to the outward-facing state after the release of ions and substrate, completes the cycle. Antiporters have evolved so that they Nature | Vol 626 | 29 February 2024 | 967

Review must bind the counter-transported substrate or ion to reset, thus ensuring coupling (Fig. 1b). In simple systems, such as Na+/H+ antiporters, the ions bind at the same sites, coordinated by a conserved aspartate66,67, and the transport kinetics fit a simple, competitive equilibrium68. This mechanism is sometimes referred to as a ‘ping-pong’ mechanism69 (Fig. 3a). In these elevator proteins, the aspartate residue in the transport domain faces the hydrophobic scaffold domain, which is likely to prevent elevator movements unless the neutralizing H+ or Na+ ions have bound70. In H+ antiport-coupled multidrug exporters, such as E. coli MdfA, protonation of the TM1 aspartate (D34) appears to be sufficient to shift the transporter from an outward-facing to inward-facing state through rearrangements of labile helices71–73. Although the binding residues vary when accommodating different substrates in these promiscuous transporters, all drugs interact with deprotonated D34 in the inward-facing state, facilitating H+ release14. Competition between H+ and substrates for the same charged residue forms the basis of drug-H+ antiport, similar to the multidrug-H+ antiporter EmrE74. Some Na+-coupled symporters have evolved supplementary coupling to the antiport of a K+ ion (Fig. 3c). In human EAATs, for example, K+ binding is obligatory for the transporter to reset from the inward-facing to the outward-facing state, and the transporter shows stoichiometric coupling to the symport of three Na+ ions, a proton, and the counter-transport of one K+ ion75. The K+ ion binds in the substrate glutamate binding site, utilizing some of the same residues that coordinate the substrate and a Na+ ion44. K+ binding closes the inner gate of the transporter43 and catalyses elevator transitions, perhaps by favouring an occluded intermediate44. By contrast, archaeal homologues do not rely on K+ binding for their reset, since their gates can shut spontaneously. In other systems, including the serotine transporter SERT76, the closely related dopamine transporter DAT77, and the prokaryotic homologue LeuT78, K+ counter-transport increases their concentrative capacity. In all these cases, K+ ions competitively decrease Na+ and substrate affinity, as would be expected for a ping-pong antiport mechanism, and increase the transport rate79. Mechanisms other than competitive binding to the shared sites by counter-transported solutes have also been observed. The starkest example is the CLC family of H+−2Cl− exchangers, which show simultaneous synergistic binding of Cl+ ions and protons80 that is prohibited by the ping-pong mechanism, suggesting a different, so far incompletely understood, coupling mechanism81.

Conserved and varied ion coupling Whereas most secondary active transporters display unvarying ioncoupled stoichiometries, others have evolved to be flexible14. For instance, in H+-oligopeptide symporters, H+-multidrug antiporters, H+-metal DraNramp symporters, and Na+-iodide symporters, the ion:substrate stoichiometries vary depending on the charge, size and chemical nature of the substrate36,48,82–85. These changes can lead to a switch in net transport from being electrically neutral to electrogenic or vice versa84,86. Proton coupling in rocker-switch transporters is especially malleable, enabling broad alterations and flexibility in energy transduction. For example, the multidrug transporter MdfA can catalyse either electrically neutral or electrogenic transport depending on the charge of the drug87. Neutralizing proton-binding acidic residues in the multidrug transporter LmrP only abolishes binding and transport of drugs with two positive charges but not singly charged drugs88. Moreover, the proton-coupling residues can be shifted to different helices without the loss of H+ coupling in both MdfA and LmrP89,90. Although three acidic residues are required for strict and maximal H+-coupling in the yeast maltose transporter Mal11, the transport ceases only when all three are neutralized91. Protons are unique among cations because they can travel via the Grotthus shuttling mechanism along water wires—that is, the chemical bonds between protons and oxygen atoms can rearrange to move the excess positive charge along the wire92. CLC proton-chloride 968 | Nature | Vol 626 | 29 February 2024

exchangers, peptide and phosphate transporters are examples of H+-coupled transporters that rely on water wires to move protons through protein25,93,94. Thus, protons can move to their binding sites through highly confined hydrophobic pathways, perhaps explaining why protons and coupled transported solutes have different ways of reaching their binding sites even if they are adjacent95,96. Conversely, Na+, K+ and Cl− ions are similar to larger solutes in that they require aqueous vestibules to diffuse toward their binding sites in the protein interior, where they are partially or completely dehydrated when they bind. Na+-binding sites in rocking-bundle and elevator transporters are remarkably conserved. For example, all three Na+ sites in archaeal glutamate transporter homologues and EAATs are fully conserved, and the structural changes upon ion binding, although different in detail, are similar overall49. Similarly, two Na+-binding sites are conserved in LeuT and human NSS, even though the amino acid sequence identity between them is only 20–30%. An aspartate in TM1 of SERT97 and DAT98 may compensate for the loss of Na1 coordination by substrates in LeuT and the γ-aminobutyric acid (GABA) transporter GAT1 (refs. 99,100) (Fig. 3b). Moreover, in TM7, the replacement of glutamate with serine generates a Cl−-binding site adjacent to Na1 (ref. 101). In this way, these transporters preserve the Na1 site and generate a novel Cl− site for additional coupling or regulation. The LeuT Na2 site is broadly conserved among Na+-coupled transporters of the amino acid, polyamine, and organocation (APC) superfamily despite sequence divergence. Notably, even in transporters that are not Na+-coupled, the placement of the cation-binding sites can be conserved. For example, elevator H+-coupled UraA102 and PurTCP103 of the nucleobase/ascorbate transporter family have their proposed proton carrier residues at the site equivalent to the Na+-binding site of a distantly related sodium-driven chloride–bicarbonate exchanger38 (NCBD). In ApcT, a H+-coupled sodium-independent amino acid transporter with a LeuT fold, the amino group of a lysine residue may replace Na+ in the Na2 site and serve as the proton carrier104. In the LeuT-fold carnitine/γ-butyrobetaine antiporter CaiT, an arginine residue guanidinium group occupies the Na2 site, rendering the transporter cation-independent105. A recent structural comparison of elevator transporters noted their marked architectural similarity, including ion-binding sites, even though they vary in their topology, suggesting either a deep evolutionary relatedness or evolutionary convergence106. Their Na+-binding sites are formed by the equivalent secondary structure elements. The persistence of Na+-binding sites in evolution, owing to either conservation or convergence, might reflect their reliance on the helix–turn–helix and helix–break–helix motifs, which are prevalent in the rocking-bundle and elevator transporters for Na+ binding and substrate coordination (Fig. 3b,c). Main-chain carbonyl oxygen atoms of discontinuous helices contribute to Na+ coordination, making binding less dependent on the exact amino acid sequence. Moreover, flexible non-helical regions seem to be especially suitable for Na+-coupled reshaping, leading to the formation of high-affinity substrate-binding sites and the closure of transporter gates.

Imperfect ion coupling and slippage Similar to other biochemical processes, especially those that are established through allosteric mechanisms, ion coupling is probabilistic, and is determined by kinetic constants and pathways107–109. Although transporters mostly avoid futile transport cycles running down substrate and ionic gradients109, ion and solute movements are not always strictly coupled. The coupling efficiency can vary between proteins within the same family25,110, mutants of the same protein111,112, or between different substrates82,86,96,113. For example, careful measurements of an E. coli CLC protein114 showed that its Cl−:H+ exchange stoichiometry is 2.2 ± 0.3:1. However, proton coupling decreases when transporting NO3−, whereas SCN− transport is decoupled from

protons altogether113. Imperfect coupling may result in the substrate ‘slipping’ through the transporter without movement of the coupled ions. In bacteria and yeast with potent PMF, solute leaks can alleviate excessive internal accumulation and buildup of aberrant osmotic pressure, preventing ‘substrate-accelerated death’109. Ion transport without coupled solutes produces ion leaks. Na+ leaks in eukaryotic transporters are commonly measured by electrophysiology in the absence of substrates115–120. However, when substrates are present, most transporters show no leaks. In some cases, kinetic mechanisms that involve partial uncoupling might enhance transporter selectivity. Many transporters—including the bacterial Na+-coupled glucose transporter vSGLT—show low, millimolar-range affinities for their cognate substrates, yet select them among potentially toxic competing molecules. In a proposed mechanism, the substrates slip back out of the transporter into the extracellular milieu, while the ions reach the cytoplasm121. The ‘proofreading’ resulting from the more probable escape of non-cognate substrates occurs at the expense of dissipating the ionic gradient. Simultaneous opening of the outside and the inside gates—prohibited by the classical alternating-access mechanism122—results in glucose slipping about 10% of the time in molecular dynamics simulations. Permeation of water and urea through SGLT1 occurs through the same pathway123, with up to 300 waters slipping through per transport cycle124. The water flux through SGLT1 mediates absorption of around 5 l a day in an average human adult, and is the rationale for salt–sugar solutions used in rehydration therapy124. Water permeation through substrate pathways and partially decoupled gates have been observed in molecular dynamics simulations of several transporter families125, consistent with experimental data123. Uncoupled movements of the external and internal gates have been observed by single-molecule fluorescence resonance energy transfer in LeuT126, suggesting that this might be a more common phenomenon than previously appreciated. Members of the rocking-bundle NRAMP family of metal ion transporters show different H+:substrate stoichiometries depending on the membrane potential, pH gradient and identity of the metal127. E. coli EmrE typically mediates antibiotic efflux coupled to H+ antiport. However, extensive NMR studies of its dynamics have shown that H+ coupling is mailable, and H+ symport or uncoupled H+ and drug fluxes are also possible under certain conditions128. This flexibility makes E. coli vulnerable to substrate-like compounds, which trigger uncoupled proton flux through EmrE, killing the bacterium128. Although the physiological relevance of cation leaks remains unclear in some cases, some ion-coupled transporters have evolved to be leaky for anions—typically Cl−—with a clearer phenotypic effect. In the case of mammalian EAATs, glutamate and Na+ binding activates a Cl− conductance, which can be significant for some subtypes. For example, the retinal transporter EAAT5 may be a bone fide glutamate-gated chloride channel, tuning membrane excitability129,130. Chloride permeation is conserved, even in archaeal homologues131, and occurs through a channel that opens transiently between the transport and scaffold domains during the elevator transition132. In the rocker-switch VGLUT in synaptic vesicles, the positive-inside potential generated by the V-type ATPase drives uptake of the negatively charged neurotransmitter glutamate133. To balance the osmotic pressure due to the accumulating glutamate, VGLUT permits Cl− efflux, possibly through the permeation pathway shared with glutamate134,135.

Lipid regulation of transport Mammalian membranes contain hundreds of different lipids, and their composition varies widely between different cell types, organelles and membrane microdomains136. The outer and inner leaflets of plasma membranes are also highly asymmetric, with lipids differing in their headgroups and the length and unsaturation of their acyl chains between the leaflets137. Energy-taxing transport systems maintain this

asymmetry, and its disruption by lipid scramblases is an important signalling mechanism, with roles in blood coagulation, immune cell activation, phagocytosis and apoptosis138. The lipid bilayer modulates membrane protein folding, assembly and activity through effects of bilayer bulk properties, annular lipids in the immediate protein vicinity, and specific lipid binding107 (Fig. 4a). Membrane proteins adapt to the bulk properties of their resident membrane, with thickness being especially important. For example, endoplasmic reticulum and Golgi membrane proteins have shorter transmembrane helices than plasma membrane proteins, reflecting their thinner organellar membranes139. Some membrane proteins show bell-shaped activity dependence on membrane thickness140, including MelB, which demonstrates the fastest melibiose uptake in liposomes with bilayer thickness matching its transmembrane helix length141. Hydrophobic mismatch occurs when the transmembrane helix length and the membrane thickness differ, potentially leading to membrane deformation and resulting in significant energetic costs140 (Fig. 4b). Minimizing these costs drives membrane protein folding142 and oligomerization143,144, and shapes the energy landscape of the transport cycle145. Annular lipids form a belt around transporters, influencing their immediate environment and establishing multiple transient interactions with proteins137. Extracting transporters from yeast using styrene maleic acid polymers revealed that protein-proximal lipids differ from the bulk lipid146 (Fig. 4a). Finally, proteins can form tight and specific interactions with non-annular lipids137.

Lipid-mediated oligomerization The scaffold-domain helices in elevator transporters are often anomalously short or highly tilted7. In Na+/H+ exchangers, for example, scaffold transmembrane helices 2 and 9 are only 12–14 residues long66,147,148, and molecular dynamics simulations show membrane bilayer thinning around the scaffold domain of a Na+/H+ exchanger NapA from Thermus thermophilus7,70. This energetically costly mismatch can drive transporter oligomerization, minimizing the protein–lipid interface. Consistent with this notion, elevator proteins are almost always obligate oligomers7. Specific lipid binding can further stabilize elevator oligomers. For example, the Na+/H+ antiporter NhaA forms dimers that are stabilized by cardiolipin binding at the interface, with the lipid phosphate group bridging cationic amino acids from each protomer144–146,149,150 (Fig. 4c). Such lipid-assisted assembly might provide a regulatory modality—salt stress increases cardiolipin synthesis in E. coli151—that favours the assembly of functional NhaA dimers required to alleviate salt stress152. The role of lipids, however, is diminished in oligomers with large interfaces. For example, the Na+/H+ antiporter NapA has an additional N-terminal helix, which contributes to a large (approximately 1,800 Å2) interface, which makes its dimerization lipid-independent149,153. The mammalian Na+/H+ exchanger NHA2 has one more N-terminal helix than NapA, which instead makes most of the oligomerization contacts147. Because of this additional helix, there is a gap of around 20 Å gap between the delipidated NHA2 protomers, which closes in the presence of phosphatidylinositol lipids147, suggesting lipid-regulated dimer stability. The late endosomal Na+/H+ exchanger NHE9 has a unique β-hairpin lipid domain, which contributes to the oligomerization interface and the binding of a phosphatidylinositol-3,5-bisphosphate (PtdIns(3,5)P2) lipid that is specific to the organelle150. Purified NHE9 monomerizes when the lipid-coordinating lysine residues are mutated. The addition of PtdIns(3,5)P2, but not the plasma membrane lipid PtdIns(4,5) P2, increases the Na+ affinity of the transporter148,150. Collectively, this elevator transporter family highlights how specific lipids can fine-tune oligomerization, which is functionally important. Similarly, the dimerization and activity of the purine transporter UapA depend on phosphatidylinositol and phosphatidylethanolamine, and mutating lipid-coordinating residues results in monomeric protein that is inactive154. Nature | Vol 626 | 29 February 2024 | 969

Review a

b Hydrophobic mismatch Drives oligomerization ΔGdef

ΔGdef

Affects conformational transitions ΔGdef Bulk lipids Second-shell lipids

c

Annular lipids

Annular lipids Non-annular lipids

Non-annular lipid modes of transporter regulation (1) Lipid-mediated oligomerization

(2) Extrinsic on–off regulation

(3) Conformational dynamics

Out

In (–) GLUT glucose transporter

NhaA Na+/H+ exchanger

ClC-7 Cl–/H+ exchanger

SERT Serotonin transporters

Fig. 4 | Lipid properties regulate transporters. a, Extracellular view of a transporter (yellow surface) in a lipid environment. Lipid regulatory mechanisms can be conceptualized as varying from tight binding of non-annular lipids (purple sphere) to weak interactions with annular lipids (green and grey spheres) and influences of the bulk lipid properties (blue spheres). b, Hydrophobic mismatch can drive transporter oligomerization and modulate conformational preference. Top, the membrane deformation due to the hydrophobic mismatch between the protein and membrane incurs an energetic penalty, ∆Gdef. Protein oligomerization involving regions of the mismatch relieves the penalty. Bottom, the extent of the hydrophobic mismatch varies in different conformational states of the transporter, and the bilayer favours the state with the least membrane deformation, which has the smallest ∆Gdef. The extent to which the membrane modulates the oligomerization and

state distributions depends on membrane thickness, elasticity and other factors that affect ∆Gdef. c, Structures of transporters, illustrating lipid–protein interactions that influence their assembly and activity. Annular lipids differ in composition from bulk lipids, making multiple weak interactions with the transporter, as illustrated here for phosphatidylserine (yellow spheres) and the fructose transporter GLUT5 (ref. 178) (PDB: 4YB9). Non-annular lipids (yellow spheres) can: (1) mediate oligomerization, as shown here for cardiolipin binding between NhaA protomers (light blue and brown) (PDB:8PS0); (2) modulate interactions with regulatory domains, as shown here for PtdIns(3,5)P2 (ref. 165) binding between the cytoplasmic regulatory CBS domain (aqua) and membrane domain (light pink) of human ClC-7 (PDB: 7JM7); and (3) stabilize conformational states, as shown for cholesterol binding to SERT171 (PDB: 8DE3) (scaffold domain, light blue; bundle domain, green).

Rocker-switch and rocking-bundle proteins do not often feature anomalously short transmembrane helices7, but can still form lipiddependent oligomers149. Thus, although LeuT is functional as a monomer155, it forms a cardiolipin-mediated dimer149. Mammalian NSS transporters also form dimers and multimeric species in cells, often stabilized by phosphatidylinositol bisphosphate lipids156–159. The APC superfamily cation–chloride cotransporters also form obligatory dimers through extensive interactions of their cytosolic domains and the lipid-mediated interface between the transmembrane domains160. Notably, their two protomers are tilted relative to each other and warp the bilayer, suggesting that the membrane might modulate their assembly or activity.

mediate interactions between the transporter catalytic domains and their regulatory regions. The trimeric rocking-bundle bacterial betaine–sodium symporter BetP is a notable example of lipid-mediated regulation in response to osmotic stress163. Anionic lipids surround BetP, and phosphatidylglycerol and cardiolipin bind at the trimer interface. The interactions of the charged C-terminal regulatory domain with the bilayer stabilize a down-regulated BetP state with high substrate affinity and reduced dynamics. The increase in cytoplasmic K+ and changes in membrane packing during osmotic stress cause partial unfolding of the C-terminal domain, its detachment from the membrane, and reorganization of the interprotomer interfaces and bound lipids. Collectively, these changes release the conformational constraints on the transporter, leading to its activation. Although monomeric mutants of BetP are functional, they no longer respond to osmotic stress164. In an interesting analogy, PtdIns(3,5)P2 inhibits the lysosomal Cl−/H+ exchanger ClC-7 (ref. 165), probably by binding on the interface. The lipid appears to stabilize interactions between the regulatory cytoplasmic cystathionine-β-synthase (CBS) domain and the transmembrane transporter domain, restricting its dynamics (Fig. 4c). By contrast, PtdIns(4,5)P2 binds to the C-terminal regulatory tail of the plasma membrane Na+/H+ exchanger NHE1, and is likely to facilitate its detachment and activate the transporter166. PtdIns(4,5)P2 also activates SPNS2, a sphingosine-1-phosphate transporter with an MFS fold, where the

Lipids modulate transport cycle dynamics Hydrophobic mismatch and the associated energetic penalty can vary between different states in the transport cycle140 (Fig. 3b). For example, significant and different membrane deformations were observed in the inward-facing states of GltPh and GltTk transporters54,145,161. Thus, bulk lipid properties, such as thickness and elasticity, should influence the relative energies of the states and therefore transport kinetics. Surprisingly, however, there is a relative lack of data showing direct lipid-mediated regulation, through either bilayer effects or lipids binding directly to the transporters, even though such mechanisms are well documented for channels162. In those known cases, lipids seem to 970 | Nature | Vol 626 | 29 February 2024

intrinsically disordered regulatory region binds to the lipid, altering transporter dynamics167. Cholesterol is an abundant lipid in eukaryotic plasma membranes, with concentrations reaching 30–40% in some cell types136. Cholesterol binds, stabilizes, modulates activity and ensures localization to microdomains of many solute transporters168–170. For example, SERT and DAT transporters share a conserved cholesterol-binding site in bundle domains171. Cholesterol stabilizes the outward-facing conformation of these transporters, activates transport and affects the potency of antagonist binding172,173 (Fig. 4c). The zwitterionic head group of phosphatidylethanolamine often competes with salt bridges that are involved in ion coupling and modulates the energies of the conformational sates14,174. In LacY, for instance, phosphatidylethanolamine helps to deprotonate E325 following sugar release into the cytoplasm175 and is required for the H+-coupled sugar uptake but not for uncoupled counterflow (transmembrane sugar exchange)176. Further, MelB shows diminished Na+- and H+-coupled melibiose uptake in an E. coli strain that lacks phosphatidylethanolamine177, whereas only H+ (but not Na+) symport was affected in a strain lacking phosphatidylglycerol and cardiolipin. Phosphatidylethanolamine is also important for the activities of GLUT and Lyp1 transporters, but unsaturated lipids also contribute by increasing membrane fluidity178,179. Anionic phosphatidylserine, which is enriched in the inner leaflet of the plasma membrane of eukaryotic cells, may facilitate breaking of intra-bundle salt bridges in the rocker-switch GLUT glucose transporters178,180. Anionic lipids are also essential for activity of the rocking-bundle H+-coupled lysine transporter Lyp1 (ref. 179).

Outlook Ion coupling is a fundamental aspect of secondary active transport. Structural studies have revealed how ions are coordinated in transporter cores and suggest how they are coupled to the binding and transport of solutes. The wealth of structural information on these proteins provides an illusionary sense that the ion-coupling mechanisms are well understood. However, deliberate manipulations of coupling residues that aim to alter the number or identity of the coupled ions via mutation have succeeded only occasionally101 and frequently fail14, suggesting that our understanding remains incomplete. Although we now have an inkling of what makes a good ion-binding site and, to an extent, the molecular basis of selectivity—for example, between H+ or Na+—a further level of understanding is required before we can re-wire or engineer transporters with desired ion coupling. A recent study based on the analysis of ion coupling evolution in glutamate transporters demonstrated that amino acid variations distant from the ion-binding sites can also determine whether the transporter couples to ions or is ion-independent181. Thus, we propose that more diffuse protein properties, such as accessible conformational ensembles, global and local dynamics, and allosteric networks, are essential determinants of ion coupling. The rapid expansion of novel experimental and computational methods to study protein conformational ensembles and their modulation by ion and substrate binding is likely to increase our understanding in future. One challenge that remains, is the ability to accurately measure the kinetics of purified transporters reconstituted into liposomes to test mutant or re-engineered transporters. This is particularly difficult for mammalian transporters, perhaps owing to strict lipid composition requirements178. Transport assays with single-transporter resolution might be particularly informative and show that some transporters have heterogeneous kinetics, alternating between different activity modes182–185. Characterizing this complexity would be important as we seek drugs that allosterically inhibit, activate, or alter the ion coupling of target transporters. Although we have accumulated data on individual examples of lipid-mediated regulation of transporters, only a few systematic studies compare multiple transporters under similar conditions, making

it difficult to reach general conclusions. For example, mammalian GLUT transporters that natively expressed in different tissues exhibit their highest activities in the same brain-like lipid mixture, whereas E. coli XylE was most active in E. coli-like lipids178. The structures of XylE and GLUT transporters are highly similar14, yet there must be evolved differences to account for these lipid preferences. Indeed, although the thermostabilities of mammalian GLUT and XylE transporters are similar before purification, the GLUT transporters are far less stable after purification in detergent, indicating they are stabilized by specific lipids152. The minor structural differences that fine-tune transporters to specific lipids and the extent to which transporters have evolved to harmonize with lipid compositions in different tissues remain poorly understood. In cells, SLCs are also regulated by partner proteins, subcellular localization and post-translational modifications. For example, phosphorylation can regulate transporters and influence oligomerization186 and the association of extramembraneous regions166. Little is known about the underlying mechanisms, and structural validation is mostly lacking. Isolating transporters from native tissues in vesicles without detergent might preserve native modifications, complexes and bound lipids187. Imaging of such preparations by single-particle cryo-electron microscopy or cryo-electron tomography could yield structural insights under more physiologically relevant conditions. Interestingly, transporter activity may even be directly coupled with signalling pathways—for example, the Na+/H+ exchanger SLC9C1 is regulated by fused voltage sensors and cyclic nucleotide-binding domains188. The physiological substrates of many mammalian transporters still remain unknown. Biochemical and biophysical studies are essential for validating substrate–transporter pairing and ion coupling, and for providing a mechanistic rationale for apparent redundancies in transporter activities. Structures of the Na+-coupled glucose transporter SGLT2 have shown how drugs for type 2 diabetes interact with extracellular gates59. Given the differences in the extracellular gates and Na+ stoichiometries between SGLT1 and SGLT2 transporters, such information could enable the development of drugs with narrower or broader specificities. As demonstrated by the multidrug exporter EmrE128, understanding ion coupling has led to the discovery of H+-uncoupling compounds that could be exploited as new antibiotics. The uncoupling protein UCP1 is a mitochondrial transporter that couples fatty acid transport with a proton leak, generating heat in brown adipose tissues. Recent cryo-electron microscopy structures of human UCP1 (refs. 189,190) have revealed details of its transport mechanism that might aid the design of novel drugs targeting obesity. How ions and lipids orchestrate small molecule transport remains of utmost importance to human health and disease. 1. 2. 3. 4. 5. 6.

7.

8. 9. 10. 11.

Lin, L., Yee, S. W., Kim, R. B. & Giacomini, K. M. SLC transporters as therapeutic targets: emerging opportunities. Nat. Rev. Drug Discov. 14, 543–560 (2015). Pizzagalli, M. D., Bensimon, A. & Superti-Furga, G. A guide to plasma membrane solute carrier proteins. FEBS J. 288, 2784–2835 (2021). Wright, E. M. SGLT2 inhibitors: physiology and pharmacology. Kidney360 2, 2027–2037 (2021). Klingenberg, M. Ligand–protein interaction in biomembrane carriers. The induced transition fit of transport catalysis. Biochemistry 44, 8563–8570 (2005). Mitchell, P. A general theory of membrane transport from studies of bacteria. Nature 180, 134–136 (1957). Jardetzky, O. Simple allosteric model for membrane pumps. Nature 211, 969–970 (1966). The terminology ‘alternating-access’ was first coined in this study of small molecule transporters. Drew, D. & Boudker, O. Shared molecular mechanisms of membrane transporters. Annu. Rev. Biochem. 85, 543–572 (2016). This review defined the elevator alternating-access mechanism, which was first observed in the sodium-coupled glutamate transporter GltPh. Keller, R., Ziegler, C. & Schneider, D. When two turn into one: evolution of membrane transporters from half modules. Biol. Chem. 395, 1379–1388 (2014). Forrest, L. R. Structural symmetry in membrane proteins. Annu. Rev. Biophys. 44, 311–337 (2015). Lane, N. & Martin, W. F. The origin of membrane bioenergetics. Cell 151, 1406–1416 (2012). Weiss, M. C. et al. The physiology and habitat of the last universal common ancestor. Nat. Microbiol. 1, 16116 (2016).

Nature | Vol 626 | 29 February 2024 | 971

Review 12. 13. 14.

15. 16. 17. 18. 19.

20. 21. 22. 23. 24. 25. 26.

27. 28.

29. 30. 31. 32. 33.

34.

35. 36.

37.

38. 39. 40. 41. 42. 43. 44.

45. 46.

Andersen, C. G., Bavnhoj, L. & Pedersen, B. P. May the proton motive force be with you: a plant transporter review. Curr. Opin. Struct. Biol. 79, 102535 (2023). Bianchi, F., Van’t Klooster, J. S., Ruiz, S. J. & Poolman, B. Regulation of amino acid transport in Saccharomyces cerevisiae. Microbiol. Mol. Biol. Rev. 83, e00024–19 (2019). Drew, D., North, R. A., Nagarathinam, K. & Tanabe, M. Structures and general transport mechanisms by the major facilitator superfamily (MFS). Chem. Rev. 121, 5289–5335 (2021). This recent review comprehensively explores the structure, function, regulation, dynamics, oligomerization and complexes of the major facilitator superfamily (MFS). Ethayathulla, A. S. et al. Structure-based mechanism for Na+/melibiose symport by MelB. Nat. Commun. 5, 3009 (2014). Claxton, D. P., Jagessar, K. L. & McHaourab, H. S. Principles of alternating access in multidrug and toxin extrusion (MATE) transporters. J. Mol. Biol. 433, 166959 (2021). Castellano, S. et al. Conserved binding site in the N-lobe of prokaryotic MATE transporters suggests a role for Na+ in ion-coupled drug efflux. J. Biol. Chem. 296, 100262 (2021). Yernool, D., Boudker, O., Jin, Y. & Gouaux, E. Structure of a glutamate transporter homologue from Pyrococcus horikoshii. Nature 431, 811–818 (2004). Isom, D. G., Castaneda, C. A., Cannon, B. R. & Garcia-Moreno, B. Large shifts in pKa values of lysine residues buried inside a protein. Proc. Natl Acad. Sci. USA 108, 5260–5265 (2011). Morrison, E. A., Robinson, A. E., Liu, Y. & Henzler-Wildman, K. A. Asymmetric protonation of EmrE. J. Gen. Physiol. 146, 445–461 (2015). Gayen, A., Leninger, M. & Traaseth, N. J. Protonation of a glutamate residue modulates the dynamics of the drug transporter EmrE. Nat. Chem. Biol. 12, 141–145 (2016). Fitch, C. A., Platzer, G., Okon, M., Garcia-Moreno, B. E. & McIntosh, L. P. Arginine: its pKa value revisited. Protein Sci. 24, 752–761 (2015). Lev, B., Roux, B. & Noskov, S. Y. in Encyclopedia of Metalloproteins (eds Kretsinger, R. H. et al.) (Springer, 2013); https://doi.org/10.1007/978-1-4614-1533-6_242. Jaud, S. et al. Insertion of short transmembrane helices by the Sec61 translocon. Proc. Natl Acad. Sci. USA 106, 11588–11593 (2009). Parker, J. L. et al. Proton movement and coupling in the POT family of peptide transporters. Proc. Natl Acad. Sci. USA 114, 13182–13187 (2017). Smirnova, I. N., Kasho, V. & Kaback, H. R. Protonation and sugar binding to LacY. Proc. Natl Acad. Sci. USA 105, 8896–8901 (2008). This study uses fluourescent-probe-based analysis and kinetics to conclusively demonstrate that LacY is always protonated prior to sugar binding across all physiologically relevant pH ranges. Kaback, H. R. & Guan, L. It takes two to tango: the dance of the permease. J. Gen. Physiol. 151, 878–886 (2019). Lolkema, J. S. & Poolman, B. Uncoupling in secondary transport proteins. A mechanistic explanation for mutants of lac permease with an uncoupled phenotype. J. Biol. Chem. 270, 12670–12676 (1995). Bavnhoj, L. et al. Structure and sucrose binding mechanism of the plant SUC1 sucrose transporter. Nat. Plants 9, 938–950 (2023). Solcan, N. et al. Alternating access mechanism in the POT family of oligopeptide transporters. EMBO J. 31, 3411–3421 (2012). Madej, M. G., Sun, L., Yan, N. & Kaback, H. R. Functional architecture of MFS D-glucose transporters. Proc. Natl Acad. Sci. USA 111, E719–E727 (2014). Leano, J. B. et al. Structures suggest a mechanism for energy coupling by a family of organic anion transporters. PLoS Biol. 17, e3000260 (2019). Geistlinger, K., Schmidt, J. D. R. & Beitz, E. Human monocarboxylate transporters accept and relay protons via the bound substrate for selectivity and activity at physiological pH. PNAS Nexus 2, pgad007 (2023). Jia, R. et al. Hydrogen–deuterium exchange mass spectrometry captures distinct dynamics upon substrate and inhibitor binding to a transporter. Nat. Commun. 11, 6162 (2020). Parker, J. L. et al. Structural basis of antifolate recognition and transport by PCFT. Nature 595, 130–134 (2021). Bozzi, A. T., Bane, L. B., Zimanyi, C. M. & Gaudet, R. Unique structural features in an Nramp metal transporter impart substrate-specific proton cotransport and a kinetic bias to favor import. J. Gen. Physiol. 151, 1413–1429 (2019). Transport kinetics of an bacterial metal transporter elegantly show that some transition metals are proton-coupled whereas others are not, and that uncoupled proton uniport is possible in the presence of a membrane potential. Yamashita, A., Singh, S. K., Kawate, T., Jin, Y. & Gouaux, E. Crystal structure of a bacterial homologue of Na+/Cl-dependent neurotransmitter transporters. Nature 437, 215–223 (2005). Wang, W. et al. Cryo-EM structure of the sodium-driven chloride/bicarbonate exchanger NDCBE. Nat. Commun. 12, 5690 (2021). Abramson, J. et al. Structure and mechanism of the lactose permease of Escherichia coli. Science 301, 610–615 (2003). Sun, L. et al. Crystal structure of a bacterial homologue of glucose transporters GLUT1–4. Nature 490, 361–366 (2012). Deng, D. et al. Crystal structure of the human glucose transporter GLUT1. Nature 510, 121–125 (2014). Mitrovic, D. et al. Reconstructing the transport cycle in the sugar porter superfamily using coevolution-powered machine learning. eLife 12, e84805 (2023). Canul-Tec, J. C. et al. The ion-coupling mechanism of human excitatory amino acid transporters. EMBO J. 41, e108341 (2022). Qiu, B. & Boudker, O. Symport and antiport mechanisms of human glutamate transporters. Nat. Commun. 14, 2579 (2023). This cryo-EM study of a human glutamate transporter reveals the detailed structural mechanism of coupled symport of sodium ions and protons and potassium antiport. Reyes, N., Oh, S. & Boudker, O. Binding thermodynamics of a glutamate transporter homolog. Nat. Struct. Mol. Biol. 20, 634–640 (2013). Verdon, G., Oh, S., Serio, R. N. & Boudker, O. Coupled ion binding and structural transitions along the transport cycle of glutamate transporters. eLife 3, e02283 (2014).

972 | Nature | Vol 626 | 29 February 2024

47.

48. 49.

50. 51.

52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62.

63. 64. 65.

66. 67. 68.

69. 70. 71.

72. 73. 74. 75. 76. 77. 78. 79.

80.

81. 82.

Guskov, A., Jensen, S., Faustino, I., Marrink, S. J. & Slotboom, D. J. Coupled binding mechanism of three sodium ions and aspartate in the glutamate transporter homologue GltTk. Nat. Commun. 7, 13420 (2016). Ravera, S. et al. Structural insights into the mechanism of the sodium/iodide symporter. Nature 612, 795–801 (2022). Qiu, B., Matthies, D., Fortea, E., Yu, Z. & Boudker, O. Cryo-EM structures of excitatory amino acid transporter 3 visualize coupled substrate, sodium, and proton binding and transport. Sci. Adv. 7, eabf5814 (2021). Jensen, S., Guskov, A., Rempel, S., Hanelt, I. & Slotboom, D. J. Crystal structure of a substrate-free aspartate transporter. Nat. Struct. Mol. Biol. 20, 1224–1226 (2013). Koch, H. P., Hubbard, J. M. & Larsson, H. P. Voltage-independent sodium-binding events reported by the 4B–4C loop in the human glutamate transporter excitatory amino acid transporter 3. J. Biol. Chem. 282, 24547–24553 (2007). Oh, S. & Boudker, O. Kinetic mechanism of coupled binding in sodium-aspartate symporter GltPh. eLife 7, e37291 (2018). Garaeva, A. A., Guskov, A., Slotboom, D. J. & Paulino, C. A one-gate elevator mechanism for the human neutral amino acid transporter ASCT2. Nat. Commun. 10, 3427 (2019). Wang, X. & Boudker, O. Large domain movements through the lipid bilayer mediate substrate release and inhibition of glutamate transporters. eLife 9, e58417 (2020). Alleva, C. et al. Na+-dependent gate dynamics and electrostatic attraction ensure substrate coupling in glutamate transporters. Sci. Adv. 6, eaba9854 (2020). Bisignano, P. et al. Inhibitor binding mode and allosteric regulation of Na+-glucose symporters. Nat. Commun. 9, 5245 (2018). Krishnamurthy, H. & Gouaux, E. X-ray structures of LeuT in substrate-free outward-open and apo inward-open states. Nature 481, 469–474 (2012). Hariharan, P. & Guan, L. Cooperative binding ensures the obligatory melibiose/Na+ cotransport in MelB. J. Gen. Physiol. 153, e202012710 (2021). Niu, Y. et al. Structural basis of inhibition of the human SGLT2–MAP17 glucose transporter. Nature 601, 280–284 (2022). Grytsyk, N., Sugihara, J., Kaback, H. R. & Hellwig, P. pKa of Glu325 in LacY. Proc. Natl Acad. Sci. USA 114, 1530–1535 (2017). Guan, L. & Kaback, H. R. Lessons from lactose permease. Annu. Rev. Biophys. Biomol. Struct. 35, 67–91 (2006). Rosa, L. T., Bianconi, M. E., Thomas, G. H. & Kelly, D. J. Tripartite ATP-independent periplasmic (TRAP) transporters and tripartite tricarboxylate transporters (TTT): from uptake to pathogenicity. Front. Cell Infect. Microbiol. 8, 33 (2018). Davies, J. S. et al. Structure and mechanism of a tripartite ATP-independent periplasmic TRAP transporter. Nat. Commun. 14, 1120 (2023). Peter, M. F. et al. Structural and mechanistic analysis of a tripartite ATP-independent periplasmic TRAP transporter. Nat. Commun. 13, 4471 (2022). Mulligan, C., Leech, A. P., Kelly, D. J. & Thomas, G. H. The membrane proteins SiaQ and SiaM form an essential stoichiometric complex in the sialic acid tripartite ATP-independent periplasmic (TRAP) transporter SiaPQM (VC1777-1779) from Vibrio cholerae. J. Biol. Chem. 287, 3598–3608 (2012). Lee, C. et al. A two-domain elevator mechanism for sodium/proton antiport. Nature 501, 573–577 (2013). Hunte, C. et al. Structure of a Na+/H+ antiporter and insights into mechanism of action and regulation by pH. Nature 435, 1197–1202 (2005). Mager, T., Rimon, A., Padan, E. & Fendler, K. Transport mechanism and pH regulation of the Na+/H+ antiporter NhaA from Escherichia coli: an electrophysiological study. J. Biol. Chem. 286, 23570–23581 (2011). Schuldiner, S. Competition as a way of life for H+-coupled antiporters. J. Mol. Biol. 426, 2539–2546 (2014). Coincon, M. et al. Crystal structures reveal the molecular basis of ion translocation in sodium/proton antiporters. Nat. Struct. Mol. Biol. 23, 248–255 (2016). Nagarathinam, K. et al. Outward open conformation of a major facilitator superfamily multidrug/H+ antiporter provides insights into switching mechanism. Nat. Commun. 9, 4005 (2018). Heng, J. et al. Substrate-bound structure of the E. coli multidrug resistance transporter MdfA. Cell Res. 25, 1060–1073 (2015). Wu, H. H., Symersky, J. & Lu, M. Structure of an engineered multidrug transporter MdfA reveals the molecular basis for substrate recognition. Commun. Biol. 2, 210 (2019). Yerushalmi, H. & Schuldiner, S. A common binding site for substrates and protons in EmrE, an ion-coupled multidrug transporter. FEBS Lett. 476, 93–97 (2000). Zerangue, N. & Kavanaugh, M. P. Flux coupling in a neuronal glutamate transporter. Nature 383, 634–637 (1996). Nelson, P. J. & Rudnick, G. Coupling between platelet 5-hydroxytryptamine and potassium transport. J. Biol. Chem. 254, 10084–10089 (1979). Schmidt, S. G. et al. The dopamine transporter antiports potassium to increase the uptake of dopamine. Nat. Commun. 13, 2446 (2022). Billesbolle, C. B. et al. Transition metal ion FRET uncovers K+ regulation of a neurotransmitter/sodium symporter. Nat. Commun. 7, 12755 (2016). Schmidt, S. G., Nygaard, A., Mindell, J. A. & Loland, C. J. Exploring the K+ binding site and its coupling to transport in the neurotransmitter:sodium symporter LeuT. eLife 12, RP87985 (2023). Picollo, A., Xu, Y., Johner, N., Berneche, S. & Accardi, A. Synergistic substrate binding determines the stoichiometry of transport of a prokaryotic H+/Cl− exchanger. Nat. Struct. Mol. Biol. 19, 525–531 (2012). This elegant study uses isothermal titration calorimetry to uncover an unexpected coupling between protonation and Cl− binding in a bacterial H+/Cl− exchanger. Accardi, A. Structure and gating of CLC channels and exchangers. J. Physiol. 593, 4129–4138 (2015). Parker, J. L., Mindell, J. A. & Newstead, S. Thermodynamic evidence for a dual transport mechanism in a POT peptide transporter. eLife 3, e04273 (2014). In this study, a reconstituted proteoliposome system is used to demonstrate that the proton:substrate stoichiometry is different between di- and tri-peptides, highlighting flexibility in substrate–H+ coupling.

83. Fluman, N. & Bibi, E. Bacterial multidrug transport through the lens of the major facilitator superfamily. Biochim. Biophys. Acta 1794, 738–747 (2009). 84. Tirosh, O. et al. Manipulating the drug/proton antiport stoichiometry of the secondary multidrug transporter MdfA. Proc. Natl Acad. Sci. USA 109, 12473–12478 (2012). 85. Edgar, R. & Bibi, E. A single membrane-embedded negative charge is critical for recognizing positively charged drugs by the Escherichia coli multidrug resistance protein MdfA. EMBO J. 18, 822–832 (1999). 86. Dohan, O. et al. The Na+/I symporter (NIS) mediates electroneutral active transport of the environmental pollutant perchlorate. Proc. Natl Acad. Sci. USA 104, 20250–20255 (2007). This study revealed that iodide and perchlorate are co-transported by the LeuT-fold protein NIS with distinct numbers of Na+ ions. 87. Lewinson, O. et al. The Escherichia coli multidrug transporter MdfA catalyzes both electrogenic and electroneutral transport reactions. Proc. Natl Acad. Sci. USA 100, 1667–1672 (2003). 88. Schaedler, T. A. & van Veen, H. W. A flexible cation binding site in the multidrug major facilitator superfamily transporter LmrP is associated with variable proton coupling. FASEB J. 24, 3653–3661 (2010). 89. Sigal, N., Fluman, N., Siemion, S. & Bibi, E. The secondary multidrug/proton antiporter MdfA tolerates displacements of an essential negatively charged side chain. J. Biol. Chem. 284, 6966–6971 (2009). This study showed that the H+-coupling residue can be shifted to a different location in the bacterial MFS protein while retaining its ability to use export drugs, demonstrating that H+-coupled transport can be flexible. 90. Debruycker, V. et al. An embedded lipid in the multidrug transporter LmrP suggests a mechanism for polyspecificity. Nat. Struct. Mol. Biol. 27, 829–835 (2020). 91. Henderson, R. & Poolman, B. Proton-solute coupling mechanism of the maltose transporter from Saccharomyces cerevisiae. Sci Rep. 7, 14375 (2017). 92. Li, C. & Voth, G. A. A quantitative paradigm for water-assisted proton transport through proteins and other confined spaces. Proc. Natl Acad. Sci. USA 118, e2113141118 (2021). 93. Han, W., Cheng, R. C., Maduke, M. C. & Tajkhorshid, E. Water access points and hydration pathways in CLC H+/Cl− transporters. Proc. Natl Acad. Sci. USA 111, 1819–1824 (2014). 94. Liu, Y. et al. Key computational findings reveal proton transfer as driving the functional cycle in the phosphate transporter PiPT. Proc. Natl Acad. Sci. USA 118, e2101932118 (2021). 95. Lee, S., Mayes, H. B., Swanson, J. M. & Voth, G. A. The origin of coupled chloride and proton transport in a Cl−/H+ antiporter. J. Am. Chem. Soc. 138, 14923–14930 (2016). 96. Bozzi, A. T. et al. Structures in multiple conformations reveal distinct transition metal and proton pathways in an Nramp transporter. eLife 8, e41124 (2019). 97. Yang, D. & Gouaux, E. Illumination of serotonin transporter mechanism and role of the allosteric site. Sci. Adv. 7, eabl3857 (2021). 98. Wang, K. H., Penmatsa, A. & Gouaux, E. Neurotransmitter and psychostimulant recognition by the dopamine transporter. Nature 521, 322–327 (2015). 99. Nayak, S. R. et al. Cryo-EM structure of GABA transporter 1 reveals substrate recognition and transport mechanism. Nat. Struct. Mol. Biol. 30, 1023–1032 (2023). 100. Zhu, A. et al. Molecular basis for substrate recognition and transport of human GABA transporter GAT1. Nat. Struct. Mol. Biol. 30, 1012–1022 (2023). 101. Zomot, E. et al. Mechanism of chloride interaction with neurotransmitter:sodium symporters. Nature 449, 726–730 (2007). This article showed that charged neutralization with either a negatively charged residue or a choloride ion is evolutionary-conserved to such an extent that just a single mutation introduces Cl− coupling to the bacterial NSS homologue LeuT. 102. Yu, X. et al. Dimeric structure of the uracil:proton symporter UraA provides mechanistic insights into the SLC4/23/26 transporters. Cell Res. 27, 1020–1033 (2017). 103. Weng, J. et al. Insight into the mechanism of H+-coupled nucleobase transport. Proc. Natl Acad. Sci. USA 120, e2302799120 (2023). 104. Shaffer, P. L., Goehring, A., Shankaranarayanan, A. & Gouaux, E. Structure and mechanism of a Na+-independent amino acid transporter. Science 325, 1010–1014 (2009). 105. Kalayil, S., Schulze, S. & Kuhlbrandt, W. Arginine oscillation explains Na+ independence in the substrate/product antiporter CaiT. Proc. Natl Acad. Sci. USA 110, 17296–17301 (2013). 106. Trebesch, N. & Tajkhorshid, E. Structure reveals homology in elevator transporters. Preprint at bioRxiv https://doi.org/10.1101/2023.06.14.544989 (2023). 107. LeVine, M. V., Cuendet, M. A., Khelashvili, G. & Weinstein, H. Allosteric mechanisms of molecular machines at the membrane: transport by sodium-coupled symporters. Chem. Rev. 116, 6552–6587 (2016). 108. Swanson, J. M. Multiscale kinetic analysis of proteins. Curr. Opin. Struct. Biol. 72, 169–175 (2022). 109. Henderson, R. K., Fendler, K. & Poolman, B. Coupling efficiency of secondary active transporters. Curr. Opin. Biotechnol. 58, 62–71 (2019). This review highlights examples of imperfect ion coupling in secondary-active transporters and where these ion leaks may benefit the host organism. 110. Bazzone, A., Zabadne, A. J., Salisowski, A., Madej, M. G. & Fendler, K. A loose relationship: incomplete H+/sugar coupling in the MFS sugar transporter GlcP. Biophys. J. 113, 2736–2749 (2017). 111. Walden, M. et al. Uncoupling and turnover in a Cl−/H+ exchange transporter. J. Gen. Physiol. 129, 317–329 (2007). 112. Lim, H. H. & Miller, C. Intracellular proton-transfer mutants in a CLC Cl−/H+ exchanger. J. Gen. Physiol. 133, 131–138 (2009). 113. Nguitragool, W. & Miller, C. Uncoupling of a CLC Cl−/H+ exchange transporter by polyatomic anions. J. Mol. Biol. 362, 682–690 (2006). 114. Miller, C. & Nguitragool, W. A provisional transport mechanism for a chloride channel-type Cl−/H+ exchanger. Philos. Trans. R. Soc. Lond. B 364, 175–180 (2009). 115. Panayotova-Heiermann, M., Loo, D. D. & Wright, E. M. Kinetics of steady-state currents and charge movements associated with the rat Na+/glucose cotransporter. J. Biol. Chem. 270, 27099–27105 (1995).

116. Galli, A., DeFelice, L. J., Duke, B. J., Moore, K. R. & Blakely, R. D. Sodium-dependent norepinephrine-induced currents in norepinephrine-transporter-transfected HEK-293 cells blocked by cocaine and antidepressants. J. Exp. Biol. 198, 2197–2212 (1995). 117. Vandenberg, R. J., Arriza, J. L., Amara, S. G. & Kavanaugh, M. P. Constitutive ion fluxes and substrate binding domains of human glutamate transporters. J. Biol. Chem. 270, 17668–17671 (1995). 118. Cammack, J. N., Rakhilin, S. V. & Schwartz, E. A. A GABA transporter operates asymmetrically and with variable stoichiometry. Neuron 13, 949–960 (1994). 119. Borre, L., Andreassen, T. F., Shi, L., Weinstein, H. & Gether, U. The second sodium site in the dopamine transporter controls cation permeation and is regulated by chloride. J. Biol. Chem. 289, 25764–25773 (2014). 120. Mager, S. et al. Conducting states of a mammalian serotonin transporter. Neuron 12, 845–859 (1994). 121. Bisignano, P. et al. A kinetic mechanism for enhanced selectivity of membrane transport. PLoS Comput. Biol. 16, e1007789 (2020). 122. Forrest, L. R. & Rudnick, G. The rocking bundle: a mechanism for ion-coupled solute flux by symmetrical transporters. Physiology 24, 377–386 (2009). 123. Zeuthen, T., Gorraitz, E., Her, K., Wright, E. M. & Loo, D. D. Structural and functional significance of water permeation through cotransporters. Proc. Natl Acad. Sci. USA 113, E6887–E6894 (2016). This pivotal study demonstrated that water is co-transported together with glucose across the apical membrane of the small intestine by SGLT1 rather than osmosis, a pathway of physiological significance in rehydration therapy. 124. Loo, D. D., Zeuthen, T., Chandy, G. & Wright, E. M. Cotransport of water by the Na+/glucose cotransporter. Proc. Natl Acad. Sci. USA 93, 13367–13370 (1996). 125. Li, J. et al. Transient formation of water-conducting states in membrane transporters. Proc. Natl Acad. Sci. USA 110, 7696–7701 (2013). 126. Terry, D. S. et al. A partially-open inward-facing intermediate conformation of LeuT is associated with Na+ release and substrate transport. Nat. Commun. 9, 230 (2018). 127. Bozzi, A. T. & Gaudet, R. Molecular mechanism of Nramp-family transition metal transport. J. Mol. Biol. 433, 166991 (2021). 128. Spreacker, P. J. et al. Activating alternative transport modes in a multidrug resistance efflux pump to confer chemical susceptibility. Nat. Commun. 13, 7655 (2022). 129. Vandenberg, R. J., Huang, S. & Ryan, R. M. Slips, leaks and channels in glutamate transporters. Channels 2, 51–58 (2008). 130. Wadiche, J. I., Amara, S. G. & Kavanaugh, M. P. Ion fluxes associated with excitatory amino acid transport. Neuron 15, 721–728 (1995). 131. Ryan, R. M. & Mindell, J. A. The uncoupled chloride conductance of a bacterial glutamate transporter homolog. Nat. Struct. Mol. Biol. 14, 365–371 (2007). 132. Chen, I. et al. Glutamate transporters have a chloride channel with two hydrophobic gates. Nature 591, 327–331 (2021). In this elegant paper, the authors combine cross-linking, electrophysiology and cryo-EM to capture the chloride-conducting state of a Na+-coupled glutamate transporter. 133. Chang, R., Eriksen, J. & Edwards, R. H. The dual role of chloride in synaptic vesicle glutamate transport. eLife 7, e34896 (2018). 134. Li, F. et al. Ion transport and regulation in a synaptic vesicle glutamate transporter. Science 368, 893–897 (2020). 135. Han, L. et al. Structure and mechanism of the SGLT family of glucose transporters. Nature 601, 274–279 (2022). 136. Harayama, T. & Riezman, H. Understanding the diversity of membrane lipid composition. Nat. Rev. Mol. Cell Biol. 19, 281–296 (2018). 137. Levental, I. & Lyman, E. Regulation of membrane protein structure and function by their lipid nano-environment. Nat. Rev. Mol. Cell Biol. 24, 107–122 (2023). This recent review surveys how specific lipids and lipid bilayer properties adjust to regulate the activity of membrane proteins. 138. Kobayashi, T. & Menon, A. K. Transbilayer lipid asymmetry. Curr. Biol. 28, R386–R391 (2018). 139. Sharpe, H. J., Stevens, T. J. & Munro, S. A comprehensive comparison of transmembrane domains reveals organelle-specific properties. Cell 142, 158–169 (2010). 140. Andersen, O. S. & Koeppe, R. E. 2nd Bilayer thickness and membrane protein function: an energetic perspective. Annu. Rev. Biophys. Biomol. Struct. 36, 107–130 (2007). 141. Dumas, F., Tocanne, J. F., Leblanc, G. & Lebrun, M. C. Consequences of hydrophobic mismatch between lipids and melibiose permease on melibiose transport. Biochemistry 39, 4846–4854 (2000). 142. Corin, K. & Bowie, J. U. How physical forces drive the process of helical membrane protein folding. EMBO Rep. 23, e53025 (2022). 143. Chadda, R. et al. Membrane transporter dimerization driven by differential lipid solvation energetics of dissociated and associated states. eLife 10, e63288 (2021). 144. Jiang, Y. et al. Membrane-mediated protein interactions drive membrane protein organization. Nat. Commun. 13, 7373 (2022). 145. Zhou, W. et al. Large-scale state-dependent membrane remodeling by a transporter protein. eLife 8, e50576 (2019). 146. van ‘t Klooster, J. S. et al. Periprotein lipidomes of Saccharomyces cerevisiae provide a flexible environment for conformational changes of membrane proteins. eLife 9, e57003 (2020). 147. Matsuoka, R. et al. Structure, mechanism and lipid-mediated remodeling of the mammalian Na+/H+ exchanger NHA2. Nat. Struct. Mol. Biol. 29, 108–120 (2022). Cryo-EM structures reveal a surprisingly dynamic oligomeric interface in the elevator Na+/H+ exchanger NHA2, which could be remodelled by the binding of specific lipids. 148. Winklemann, I. et al. Structure and elevator mechanism of the mammalian sodium/ proton exchanger NHE9. EMBO J. 39, e105908 (2020). 149. Gupta, K. et al. The role of interfacial lipids in stabilizing membrane protein oligomers. Nature 541, 421–424 (2017). 150. Kokane S, M. P. et al. PI-(3,5)P2-mediated oligomerization of the endosomal sodium/ proton exchanger NHE9. Preprint at bioRxiv https://doi.org/10.1101/2023.09.10.557050 (2023).

Nature | Vol 626 | 29 February 2024 | 973

Review 151. Romantsov, T., Guan, Z. & Wood, J. M. Cardiolipin and the osmotic stress responses of bacteria. Biochim. Biophys. Acta 1788, 2092–2100 (2009). 152. Nji, E., Chatzikyriakidou, Y., Landreh, M. & Drew, D. An engineered thermal-shift screen reveals specific lipid preferences of eukaryotic and prokaryotic membrane proteins. Nat. Commun. 9, 4253 (2018). 153. Landreh, M. et al. Integrating mass spectrometry with MD simulations reveals the role of lipids in Na+/H+ antiporters. Nat. Commun. 8, 13993 (2017). 154. Pyle, E. et al. Structural lipids enable the formation of functional oligomers of the eukaryotic purine symporter UapA. Cell Chem. Biol. 25, 840–848.e844 (2018). 155. Kawate, T. & Gouaux, E. Fluorescence-detection size-exclusion chromatography for precrystallization screening of integral membrane proteins. Structure 14, 673–681 (2006). 156. Anderluh, A. et al. Direct PIP2 binding mediates stable oligomer formation of the serotonin transporter. Nat. Commun. 8, 14089 (2017). 157. Luethi, D. et al. Phosphatidylinositol 4,5-bisphosphate (PIP2) facilitates norepinephrine transporter dimerization and modulates substrate efflux. Commun. Biol. 5, 1259 (2022). 158. Anderluh, A. et al. Single molecule analysis reveals coexistence of stable serotonin transporter monomers and oligomers in the live cell plasma membrane. J. Biol. Chem. 289, 4387–4394 (2014). 159. Das, A. K. et al. Dopamine transporter forms stable dimers in the live cell plasma membrane in a phosphatidylinositol 4,5-bisphosphate-independent manner. J. Biol. Chem. 294, 5632–5642 (2019). 160. Chew, T. A., Zhang, J. & Feng, L. High-resolution views and transport mechanisms of the NKCC1 and KCC transporters. J. Mol. Biol. 433, 167056 (2021). 161. Arkhipova, V., Guskov, A. & Slotboom, D. J. Structural ensemble of a glutamate transporter homologue in lipid nanodisc environment. Nat. Commun. 11, 998 (2020). 162. Hansen, S. B. Lipid agonism: the PIP2 paradigm of ligand-gated ion channels. Biochim. Biophys. Acta 1851, 620–628 (2015). 163. Heinz, V. et al. Osmotic stress response in BetP: how lipids and K+ team up to overcome downregulation. Preprint at bioRxiv https://doi.org/10.1101/2022.06.02.493408 (2022). 164. Perez, C., Khafizov, K., Forrest, L. R., Kramer, R. & Ziegler, C. The role of trimerization in the osmoregulated betaine transporter BetP. EMBO Rep. 12, 804–810 (2011). 165. Leray, X. et al. Tonic inhibition of the chloride/proton antiporter ClC-7 by PI(3,5)P2 is crucial for lysosomal pH maintenance. eLife 11, e74136 (2022). 166. Pedersen, S. F. & Counillon, L. The SLC9A–C mammalian Na+/H+ exchanger family: molecules, mechanisms, and physiology. Physiol. Rev. 99, 2015–2113 (2019). 167. Tang, H. et al. The solute carrier SPNS2 recruits PI(4,5)P2 to synergistically regulate transport of sphingosine-1-phosphate. Mol. Cell 83, 2739–2752.e2735 (2023). 168. Zhang, L. et al. Cholesterol stimulates the cellular uptake of L-carnitine by the carnitine/ organic cation transporter novel 2 (OCTN2). J. Biol. Chem. 296, 100204 (2021). 169. Raunser, S. et al. Heterologously expressed GLT-1 associates in approximately 200-nm protein–lipid islands. Biophys. J. 91, 3718–3726 (2006). 170. Butchbach, M. E., Tian, G., Guo, H. & Lin, C. L. Association of excitatory amino acid transporters, especially EAAT2, with cholesterol-rich lipid raft microdomains: importance for excitatory amino acid transporter localization and function. J. Biol. Chem. 279, 34388–34396 (2004). 171. Yang, D., Zhao, Z., Tajkhorshid, E. & Gouaux, E. Structures and membrane interactions of native serotonin transporter in complexes with psychostimulants. Proc. Natl Acad. Sci. USA 120, e2304602120 (2023). 172. Laursen, L. et al. Cholesterol binding to a conserved site modulates the conformation, pharmacology, and transport kinetics of the human serotonin transporter. J. Biol. Chem. 293, 3510–3523 (2018). 173. Hong, W. C. & Amara, S. G. Membrane cholesterol modulates the outward facing conformation of the dopamine transporter and alters cocaine binding. J. Biol. Chem. 285, 32616–32626 (2010). 174. Martens, C. et al. Direct protein-lipid interactions shape the conformational landscape of secondary transporters. Nat. Commun. 9, 4151 (2018). In this study, hydrogen–deuterium exchange mass spectrometry and molecular dynamics simulations show how lipid compositions can influence conformational preferences and dynamics in MFS transporters. 175. Andersson, M. et al. Proton-coupled dynamics in lactose permease. Structure 20, 1893–1904 (2012).

974 | Nature | Vol 626 | 29 February 2024

176. Bogdanov, M. & Dowhan, W. Phosphatidylethanolamine is required for in vivo function of the membrane-associated lactose permease of Escherichia coli. J. Biol. Chem. 270, 732–739 (1995). 177. Hariharan, P. et al. Structural and functional characterization of protein-lipid interactions of the Salmonella typhimurium melibiose transporter MelB. BMC Biol. 16, 85 (2018). 178. Suades, A. et al. Establishing mammalian GLUT kinetics and lipid composition influences in a reconstituted-liposome system. Nat. Commun. 14, 4070 (2023). 179. van ‘t Klooster, J. S. et al. Membrane lipid requirements of the lysine transporter Lyp1 from Saccharomyces cerevisiae. J. Mol. Biol. 432, 4023–4031 (2020). 180. Hresko, R. C., Kraft, T. E., Quigley, A., Carpenter, E. P. & Hruz, P. W. Mammalian glucose transporter activity is dependent upon anionic and conical phospholipids. J. Biol. Chem. 291, 17271–17282 (2016). 181. Reddy K. D. et al. Uncoupled substrate binding underlies the evolutionary switch between Na+ and H+-coupled prokaryotic aspartate transporters. Preprint at bioRxiv https://doi.org/ 10.1101/2023.12.03.569786 (2023). 182. Goudsmits, J. M. H., Slotboom, D. J. & van Oijen, A. M. Single-molecule visualization of conformational changes and substrate transport in the vitamin B(12) ABC importer BtuCD–F. Nat. Commun. 8, 1652 (2017). 183. Fitzgerald, G. A. et al. Quantifying secondary transport at single-molecule resolution. Nature 575, 528–534 (2019). 184. Ciftci, D. et al. Single-molecule transport kinetics of a glutamate transporter homolog shows static disorder. Sci. Adv. 6, eaaz1949 (2020). 185. Ciftci, D. et al. Linking function to global and local dynamics in an elevator-type transporter. Proc. Natl Acad. Sci. USA 118, e2025520118 (2021). 186. Fang, X. Z. et al. NRT1.1 dual-affinity nitrate transport/signalling and its roles in plant abiotic stress resistance. Front. Plant Sci. 12, 715694 (2021). 187. Tao, X., Zhao, C. & MacKinnon, R. Membrane protein isolation and structure determination in cell-derived membrane vesicles. Proc. Natl Acad. Sci. USA 120, e2302325120 (2023). 188. Windler, F. et al. The solute carrier SLC9C1 is a Na+/H+-exchanger gated by an S4-type voltage-sensor and cyclic-nucleotide binding. Nat. Commun. 9, 2809 (2018). 189. Jones, S. A. et al. Structural basis of purine nucleotide inhibition of human uncoupling protein 1. Sci. Adv. 9, eadh4251 (2023). 190. Kang, Y. & Chen, L. Structural basis for the binding of DNP and purine nucleotides onto UCP1. Nature 620, 226–231 (2023). Acknowledgements D.D. is a Wallenberg Scholar funded by the Knut and Alice Wallenberg Foundation and further acknowledges support from the European Research Council (ERC) Consolidator Grant EXCHANGE (ERC-CoG-820187), The Swedish Research Council (202104709) and the Göran Gustafsson Foundation. O.B. is a Howard Hughes Medical Institute (H.H.M.I) Investigator and further acknowledges support from the National Institute of Neurological Disorders and Stroke grant R37NS085318. Author contributions Both authors contributed to the writing of this manuscript. Competing interests The authors declare no competing interests. Additional information Correspondence and requests for materials should be addressed to David Drew or Olga Boudker. Peer review information Nature thanks Bert Poolman and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Reprints and permissions information is available at http://www.nature.com/reprints. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law. © Springer Nature Limited 2024

Article

Most of the photons that reionized the Universe came from dwarf galaxies https://doi.org/10.1038/s41586-024-07043-6 Received: 16 August 2023 Accepted: 8 January 2024 Published online: 28 February 2024 Check for updates

Hakim Atek1 ✉, Ivo Labbé2, Lukas J. Furtak3, Iryna Chemerynska1, Seiji Fujimoto4, David J. Setton5, Tim B. Miller6, Pascal Oesch7,8, Rachel Bezanson5, Sedona H. Price5, Pratika Dayal9, Adi Zitrin2, Vasily Kokorev9, John R. Weaver10, Gabriel Brammer8, Pieter van Dokkum11, Christina C. Williams12,13, Sam E. Cutler10, Robert Feldmann14, Yoshinobu Fudamoto15,16, Jenny E. Greene17, Joel Leja18,19,20, Michael V. Maseda21, Adam Muzzin22, Richard Pan23, Casey Papovich24,25, Erica J. Nelson26, Themiya Nanayakkara2, Daniel P. Stark13, Mauro Stefanon27, Katherine A. Suess28,29, Bingjie Wang18,19,20 & Katherine E. Whitaker8,10

The identification of sources driving cosmic reionization, a major phase transition from neutral hydrogen to ionized plasma around 600–800 Myr after the Big Bang1–3, has been a matter of debate4. Some models suggest that high ionizing emissivity and escape fractions (fesc) from quasars support their role in driving cosmic reionization5,6. Others propose that the high fesc values from bright galaxies generate sufficient ionizing radiation to drive this process7. Finally, a few studies suggest that the number density of faint galaxies, when combined with a stellar-mass-dependent model of ionizing efficiency and fesc, can effectively dominate cosmic reionization8,9. However, so far, comprehensive spectroscopic studies of low-mass galaxies have not been done because of their extreme faintness. Here we report an analysis of eight ultra-faint galaxies (in a very small field) during the epoch of reionization with absolute magnitudes between MUV ≈ −17 mag and −15 mag (down to 0.005L⋆ (refs. 10,11)). We find that faint galaxies during the first thousand million years of the Universe produce ionizing photons with log[ξion (Hz erg−1)] = 25.80 ± 0.14, a factor of 4 higher than commonly assumed values12. If this field is representative of the large-scale distribution of faint galaxies, the rate of ionizing photons exceeds that needed for reionization, even for escape fractions of the order of 5%.

We combine ultra-deep James Webb Space Telescope ( JWST) imaging data with ancillary Hubble Space Telescope (HST) imaging data of the gravitational lensing cluster Abell 2744 (A2744) to photometrically select extremely faint galaxy candidates in the epoch of reionization. A crucial component of our study is the use of strong gravitational lensing to amplify the intrinsically faint flux of distant sources. An accurate estimate of the magnification factor is required to retrieve the intrinsic luminosity of sources. This step relies on a good knowledge of the total mass distribution in the galaxy cluster. Here we use the most recent lensing model (v.1.1) published for the UNCOVER (Ultradeep

NIRSpec and NIRCam Observations before the Epoch of Reionization) survey. The magnification factors for our galaxy sample range from µ ≈ 2 to µ ≈ 27. The values are reported in Table 1, together with 1σ uncertainties. The second part of the UNCOVER programme consists of ultra-deep follow-up spectroscopy with the NIRSpec instrument. We used the Multi-Shutter Assembly to obtain multi-object spectroscopy in seven pointings, totalling an exposure time ranging from 2.7 h to 17.4 h. Figure 1 shows the position of these sources in the A2744 field with the associated regions of high magnification and the configuration of the NIRSpec slits. Simultaneous spectral fits to the continuum and

Institut d’Astrophysique de Paris, CNRS, Sorbonne Université, Paris, France. 2Centre for Astrophysics and Supercomputing, Swinburne University of Technology, Melbourne, Victoria, Australia. Physics Department, Ben-Gurion University of the Negev, Be’er Sheva, Israel. 4Department of Astronomy, The University of Texas at Austin, Austin, TX, USA. 5Department of Physics and

1

3

Astronomy and PITT PACC, University of Pittsburgh, Pittsburgh, PA, USA. 6Center for Interdisciplinary Exploration and Research in Astrophysics (CIERA) and Department of Physics and Astronomy, Northwestern University, Evanston, IL, USA. 7Department of Astronomy, University of Geneva, Versoix, Switzerland. 8Cosmic Dawn Center (DAWN), Niels Bohr Institute, University of Copenhagen, Copenhagen, Denmark. 9Kapteyn Astronomical Institute, University of Groningen, Groningen, The Netherlands. 10Department of Astronomy, University of Massachusetts, Amherst, MA, USA. 11Department of Astronomy, Yale University, New Haven, CT, USA. 12NSF’s National Optical-Infrared Astronomy Research Laboratory, Tucson, AZ, USA. 13Steward Observatory, University of Arizona, Tucson, AZ, USA. 14Institute for Computational Science, University of Zurich, Zurich, Switzerland. 15Waseda Research Institute for Science and Engineering, Faculty of Science and Engineering, Waseda University, Tokyo, Japan. 16National Astronomical Observatory of Japan, Tokyo, Japan. 17Department of Astrophysical Sciences, Princeton University, Princeton, NJ, USA. 18Department of Astronomy and Astrophysics, The Pennsylvania State University, University Park, PA, USA. 19Institute for Computational and Data Sciences, The Pennsylvania State University, University Park, PA, USA. 20Institute for Gravitation and the Cosmos, The Pennsylvania State University, University Park, PA, USA. 21Department of Astronomy, University of Wisconsin, Madison, WI, USA. 22Department of Physics and Astronomy, York University, Toronto, Ontario, Canada. 23Department of Physics and Astronomy, Tufts University, Medford, MA, USA. Department of Physics and Astronomy, Texas A&M University, College Station, TX, USA. 25George P. and Cynthia Woods Mitchell Institute for Fundamental Physics and Astronomy, Texas A&M University, College Station, TX, USA. 26Department for Astrophysical and Planetary Science, University of Colorado, Boulder, CO, USA. 27Departament d’Astronomia i Astrofìsica, Universitat de València, Valencia, Spain. 28Department of Astronomy and Astrophysics, University of California, Santa Cruz, CA, USA. 29Kavli Institute for Particle Astrophysics and Cosmology and Department of Physics, Stanford University, Stanford, CA, USA. ✉e-mail: [email protected] 24

Nature | Vol 626 | 29 February 2024 | 975

Article Table 1 | Summary of the sample properties Source

RA (J2000)

Dec (J2000)

Exptime (h)

µ

zphot

zspec

MUV(AB)

18924

3.581044

−30.389561

17.4

26.6 ± 7.1

7.90.2 0.3

7.70

−15.47 ± 0.08

16155

3.582953

−30.395232

17.4

11.1 ± 3.8

6.70.1 0.1

6.87

−16.29 ± 0.08

23920

3.572830

−30.380026

3.7

3.3 ± 0.1

6.40.4 4.9

6.00

−16.18 ± 0.10

12899

3.582353

−30.402732

10.2

13.9 ± 0.9

6.60.2 0.2

6.88

−15.34 ± 0.11

8613

3.600602

−30.410271

2.7

9.3 ± 0.6

6.50.1 0.1

6.38

−16.97 ± 0.04

23619

3.607272

−30.380578

7.5

1.8 ± 0.2

6.70.2 0.3

6.72

−16.55 ± 0.16

38335

3.541383

−30.357435

2.7

2.3 ± 0.2

6.42.2 1.8

6.23

−16.89 ± 0.13

27335

3.625081

−30.375261

7.5

1.4 ± 0.1

6.90.5 0.1

6.76

−17.17 ± 0.08

The exposure time (Exptime) corresponds to the total of all NIRSpec observations for each source. The magnification factors (µ) are computed at the spectroscopic redshift of the source using the most recent UNCOVER lensing model. We also added systematic uncertainties derived from a comparison with an independent lensing model34. The photometric redshift (zphot) is measured with the Eazy software. The spectroscopic redshift (zspec) is measured from the best msaexp fit. The typical best-fit error is σ z spec = 0.01. Absolute magnitude MUV(AB) is measured in the rest-frame UV using the observed magnitude derived from the UNCOVER photometric catalogue corrected for magnification. RA, right ascension; Dec, declination.

976 | Nature | Vol 626 | 29 February 2024

result is the high value of log[ξion (Hz erg−1)] = 25.80 ± 0.14 observed in faint (MUV > −16.5) galaxies, compared with the canonical value of log[ξion (Hz erg−1)] = 25.2 (ref. 12) commonly assumed in reionization

18924

8613

16155

23619

12899

23920

38335

27335

−30º 21′ 38335

−30º 22′ 27335

Dec (J2000)

23920

23619

−30º 23′

18924 16155 12899

−30º 24′ 8613

s

s

14 h 0

h

14

m

m

in

in

6

12

s 0

0

h

14

m

in

18

s 24 in m 14 h 0

h

14

m

in

30

s

−30º 25′

0

the emission lines provide estimates of the spectroscopic redshifts of these sources, which lie between z ≈ 6 and z ≈ 7.7 (Table 1). The spectral extraction and fitting procedure are discussed in the Methods. Because of the gravitational magnification, we measure extremely faint line fluxes down to f = 5 × 10−21 erg s−1 cm−2. We also derive intrinsic absolute magnitudes as faint as MUV  ≲ −15.34 mag, which is nearly two magnitudes fainter than the faintest galaxies discovered in JWST spectroscopic surveys13–15 at the epoch of reionization so far (Fig. 2). Assu⋆ ming a characteristic magnitude of M UV = − 21.15 mag at z = 7, these ⋆ galaxies are as faint as 0.005L . In light of the steep faint-end slope of the galaxy UV luminosity function at z > 6 (refs. 10,16), these galaxies probably provide the bulk of the UV radiation at the epoch of reionization3,9. To infer the stellar populations of these systems, we perform joint spectrophotometric spectral energy distribution (SED) fits using the Bagpipes (Bayesian Analysis of Galaxies for Physical Inference and Parameter Estimation) software package. Accounting for the magnification, we derive extremely low stellar masses between +0.07 log(M⋆ /M⊙) = 5.88+0.13 −0.08 and log(M⋆ /M⊙) = 7.12−0.08 . Our results also show that these galaxies harbour very young stellar populations, with stellar ages mostly around a few million years (Table 2). This picture is also supported by their blue UV continuum slopes, derived from our SED fitting, in the range of β = [−2.07, −2.53]. These values are generally indicative of a young massive stellar population and low dust attenuation. The ability of galaxies to reionize the Universe depends on their production of ionizing photon density per unit of time and the fraction of this radiation that escapes to ionize the intergalactic neutral gas. This quantity can be summarized by the following relation: ̇ = fesc ξion ρUV, where ρUV is the non-ionizing UV luminosity density nion at 1,500 Å, ξion is the ionizing photon production efficiency that represents the number of ionizing photons (Lyman continuum photons, LyC) per unit UV luminosity density and fesc is the fraction of this LyC radiation that escapes the galaxy to ionize the intergalactic medium3. It is now well-established that faint galaxies (MUV > −18) are the dominant source of UV radiation during the reionization period10,17,18, although recent JWST observations show several faint active galactic nuclei (AGNs) at 3  10) and cyan (µ > 100) contours, which are derived from the latest lensing model35. The position of each source is marked with a red circle. Two of the sources (12899, 16155) are predicted to be multiply imaged by the lens model, but only the marked image of each system was targeted spectroscopically. On the top of the image, we show an RGB (red, green and blue) image of each source and the positions of each NIRSpec slitlet on top of the target. RA, right ascension; Dec, declination.

–21

26.00 fesc > 5% 25.75

–20

fesc > 14%

log[ξion (Hz erg–1)]

25.50 –19

MUV

–18

25.25

fesc > 20%

Canonical values

25.00 24.75

–17

24.50

–16

JWST/GLASS (ref. 15) Literature results JWST/JADES

–15

This work 6

7

8

9

Redshift Fig. 2 | Spectroscopic observations of the faintest galaxies during the epoch of reionization. Various literature results from ground-based, HST and JWST observations are shown with orange squares14. The blue stars represent the spectroscopic sample of the JWST/GLASS (Grism Lens-Amplified Survey from Space) survey presented in ref. 15. The green circles are derived from the latest data release of the deep spectroscopic observations of the JWST/JADES (JWST Advanced Deep Extragalactic Survey) programme13. The horizontal grey line denotes the limit of the deepest JWST spectroscopic programmes.

models, or previous studies at this epoch. For example, the measured efficiency in a population of Lyα emitters, which are thought to have larger ionization radiation than the average galaxy population, is around log[ξion (Hz erg−1)] = 25.4 (ref. 21). Our measured value is consistent with the maximum values predicted by the BPASS (Binary Population and Spectral Synthesis)22 stellar population models for a dust-free galaxy with a constant star formation and a stellar age of less than 3 Myr and a 0.1 Z⊙ metallicity. This large ionizing efficiency in faint galaxies implies that modest values of fesc are sufficient for galaxies to reionize the Universe by z = 6. Until now, most models of reionization needed to

Table 2 | Summary of the physical properties of the sample derived from SED fitting with Bagpipes Source log (M★/M☼)

SFRHα (M☼ yr)

SFRUV (M☼ yr)

β

t50 (Myr)

12 + log(O/H)

− 2.39+−0.12 0.10

18924

0.33 ± 0.02 0.01+−0.14 5.88+−0.13 0.08 0.07

2.23+−0.68 0.85

6.95 ± 0.15

16155

6.61+−0.07 0.06

+0.07 0.92 ± 0.04 0.04+−0.08 0.06 − 2.09−0.08

3.96+−0.92 0.66

7.01 ± 0.19

23920

6.30+−0.03 0.03

1.32 ± 0.04

1.12+−0.32 0.11

6.84 ± 0.06

12899

+0.09 0.49 ± 0.02 0.04+−0.12 6.54+−0.14 0.19 0.15 − 2.51−0.07

8613

7.12+−0.07 0.08

0.78 ± 0.07 0.16+−0.08 − 2.53+−0.04 0.07 0.03

6.97 ± 0.18 25.73+−6.47 6.33

23619

6.57+−0.10 0.06

38335

6.83+−0.25 0.20

27335

0.73 ± 0.10 6.73+−0.15 0.08

0.02+−0.03 0.03

− 2.45+−0.03 0.03

6.70 ± 0.15 28.66+−15.51 11.98

0.85 ± 0.07

0.04+−0.11 0.05

− 2.51+−0.13 0.07

1.08+−0.22 0.07

7.19 ± 0.20

1.00 ± 0.16

0.07+−0.34 0.15

− 2.07+−0.29 0.24

6.45+−4.39 2.29

7.46 ± 0.32

1.56+−1.33 0.52

6.99 ± 0.18

+0.22 0.05+−0.17 0.07 − 2.35−0.11

For each source, the median posterior and associated uncertainties from the best-fit models are given for the stellar mass (log(M★/M☼)), the star-formation rate (SFR), the UV continuum slope β and the half-mass age (t50). The oxygen abundance computed from strong optical lines is also reported. The SFR derived from the Hα emission is also reported.

24.25 0

2

Ref. 80 Ref. 82 Ref. 81 Ref. 85

Ref. 83 Ref. 84 This work MUV > –16.5 mag This work MUV < –16.5 mag

4

6

8

10

Redshift Fig. 3 | The ionizing photon production efficiency of faint galaxies during the epoch of reionization. Our ξion measurements are marked with an orange star (light and dark shades for galaxies brighter and fainter than MUV = −16.5, respectively). The grey-shaded horizontal line represents the canonical values assumed when assessing the contribution of galaxies to reionization. Various literature results are also shown and listed in the Methods. Assuming a fiducial UV luminosity density10,17, we plot the minimum fesc required to maintain reionization at each given value of ξion (horizontal lines). The error bars represent 1σ uncertainties.

assume large values of fesc, typically around 20%, to accommodate the relatively low ionizing photon emissivity observed in high-z galaxies. Some models required lower fesc values with combinations of specific galaxy properties and a small contribution from AGNs8. Direct measurements of fesc z = 0−4 analogues, reported typical or sample-averaged values below 10% (ref. 23), albeit with a large dispersion and higher values have been observed in individual objects. As we can see in Fig. 3, a volume-averaged escape fraction as low as fesc = 5% is sufficient for faint galaxies to maintain reionization. To fully understand cosmic reionization by star-forming galaxies, we compute the spectroscopic UV luminosity function based on the present sample. We put spectroscopic constraints on the prevalence of ultra-faint galaxies. Although with a small sample size, our measurements provide confirmation of the steep faint-end slope of the UV luminosity function at z ~ 7, in agreement with the photometric UV luminosity function derived from Hubble Frontier Fields observations24. By integrating this UV luminosity function down to a faint-end limit of MUV = −15 mag, we determine a UV luminosity density of log[ρUV (erg s−1 Mpc−3)] = 26.22. Now combining these two quantities, we can obtain the total ionizing emissivity of galaxies for different values of fesc, accounting for the contribution of the ultra-faint population. The result is shown in Fig. 4. Galaxies produce enough ionizing photons to maintain reionization at z ≈ 7 (ref. 25), assuming on average as little as 5% of this radiation escapes from the galaxies to heat the intergalactic medium. We can go a step further by indirectly estimating fesc using the UV continuum slope we measured for these galaxies. Specifically, we follow an approach pioneered by recent studies of nearby galaxies that calibrated indirect indicators of fesc of LyC emission. In particular, a strong correlation is observed between the 1,550 observed β obs slope and fesc (ref. 26). Adopting the UV-slope β (Table 2) as a proxy for fesc, we infer escape fractions within fesc = [0.045, 0.16], well in the range of assumed values in Fig. 4. Another indirect indicator of fesc that has been explored in recent studies is the star-formation surface density ΣSFR (ref. 27). For these compact sources, we measure log[ΣSFR/M☼(yr kpc2)] ≈ 0.2−2. These values are commonly observed in Nature | Vol 626 | 29 February 2024 | 977

Article 52.0

3.

Required ionizing emissivity This work: ξion ×

4.

UV

fesc = 15%

5. 6.

51.0

fesc = 5% Galaxies can a reionize the Universe

50.5

7. 8.

50.0

49.5

49.0

–22

–20

–18 –16 MUV (mag)

UNCOVER (this work)

9.

JADES + NGDEEP limit

log[n˙ ion (Mpc–3 s–1)]

51.5

10. 11. 12.

13.

–14

–12

Fig. 4 | The total ionizing emissivity of galaxies at z ~ 7. The total ionizing photon production rate density, derived from the prevalence and the ionizing efficiency of galaxies, as a function of the faint integration limit. The blue region delimits the two cases, in which fesc = 5% and fesc = 15%. The grey-shaded region is the threshold required to maintain the Universe ionized at z = 7. The grey vertical line marks the magnitude limit of the deepest JWST spectroscopic surveys to date. The orange vertical line shows the limit probed by this work. At this luminosity, galaxies produce enough radiation to reionize the Universe.

LyC leakers28,29 and also predictive of high fesc according to published best-fit relations of fesc–ΣSFR (ref. 7). We note that measurements of ξion can be markedly affected by dust attenuation. This concern also applies to our estimate of stochastic star formation through the SFR(Hα)/SFR(UV) ratio. However, as indicated by the blue UV continuum slopes that we observe, we expect the dust content to be small in these galaxies. This assumption is also supported by the low Balmer decrement Hα/Hβ, for which we measure an average value of 3.3 ± 0.5. Therefore, dust attenuation should not markedly affect these quantities. We note that we used indirect indicators, which come with a significant scatter, to estimate fesc, because direct measurements of LyC at the epoch of reionization are impossible. The stochastic nature of star formation in these low-mass galaxies also makes the fesc highly variable, because it mainly relies on stellar and supernovae feedback clearing the ISM for LyC escape30. However, on average, hydrodynamical simulations predict higher fesc in lower-mass galaxies31,32. Again, based on the ionizing photon production we estimated, modest values of fesc around 5% are sufficient. We also note that our conclusions are based on observations obtained in one field, and are therefore not immune to field-to-field variations or environmental effects. For instance, the ionizing properties of faint galaxies can be affected differently by reionization radiation if they reside in over-dense regions33. More observations in an independent field should provide further insights in that regard.

14. 15. 16. 17.

18.

19.

20.

21. 22. 23.

24.

25. 26. 27. 28. 29. 30.

31. 32. 33.

Online content

34.

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-024-07043-6.

35.

1. 2.

Dayal, P. & Ferrara, A. Early galaxy formation and its large-scale effects. Phys. Rep. 780–782, 1–64 (2018). Mason, C. A., Naidu, R. P., Tacchella, S. & Leja, J. Model-independent constraints on the hydrogen-ionizing emissivity at z > 6. Mon. Not. R. Astron. Soc. 489, 2669–2676 (2019).

978 | Nature | Vol 626 | 29 February 2024

Robertson, B. E. et al. Identification and properties of intense star-forming galaxies at redshifts z > 10. Nat. Astron. 7, 611–621 (2023). Robertson, B. E. Galaxy formation and reionization: key unknowns and expected breakthroughs by the James Webb Space Telescope. Annu. Rev. Astron. Astrophys. 60, 121–158 (2022). Madau, P. & Haardt, F. Cosmic reionization after Planck: could quasars do it all? Astrophys. J. Lett. 813, L8 (2015). Mitra, S., Choudhury, T. R. & Ferrara, A. Cosmic reionization after Planck II: contribution from quasars. Mon. Not. R. Astron. Soc. 473, 1416–1425 (2018). Naidu, R. P. et al. Rapid reionization by the oligarchs: the case for massive, UV-bright, star-forming galaxies with high escape fractions. Astrophys. J. 892, 109 (2020). Finkelstein, S. L. et al. Conditions for reionizing the Universe with a low galaxy ionizing photon escape fraction. Astrophys. J. 879, 36 (2019). Dayal, P. et al. Reionization with galaxies and active galactic nuclei. Mon. Not. R. Astron. Soc. 495, 3065–3078 (2020). Finkelstein, S. L. et al. The evolution of the galaxy rest-frame ultraviolet luminosity function over the first two billion years. Astrophys. J. 810, 71 (2015). Bouwens, R. J. et al. UV luminosity functions at redshifts z ∼ 4 to z ∼ 10: 10,000 galaxies from HST legacy fields. Astrophys. J. 803, 34 (2015). Robertson, B. E., Ellis, R. S., Furlanetto, S. R. & Dunlop, J. S. Cosmic reionization and early star-forming galaxies: a joint analysis of new constraints from Planck and the Hubble Space Telescope. Astrophys. J. Lett. 802, L19 (2015). Bunker, A. J. et al. JADES NIRSpec initial data release for the Hubble Ultra Deep Field: redshifts and line fluxes of distant galaxies from the deepest JWST Cycle 1 NIRSpec Multi-Object spectroscopy. Preprint at https://doi.org/10.48550/arXiv.2306.02467 (2023). Roberts-Borsani, G. et al. The nature of an ultra-faint galaxy in the cosmic dark ages seen with JWST. Nature 618, 480–483 (2023). Mascia, S. et al. Closing in on the sources of cosmic reionization: first results from the GLASS-JWST program. Astron. Astrophys. 672, A155 (2023). Ishigaki, M. et al. Full-data results of Hubble Frontier Fields: UV luminosity functions at z ∼ 6–10 and a consistent picture of cosmic reionization. Astrophys. J. 854, 73 (2018). Atek, H. et al. Are ultra-faint galaxies at z = 6–8 responsible for cosmic reionization? Combined constraints from the Hubble Frontier Fields clusters and parallels. Astrophys. J. 814, 69 (2015). Bouwens, R. J., Oesch, P. A., Illingworth, G. D., Ellis, R. S. & Stefanon, M. The z ∼ 6 luminosity function fainter than −15 mag from the Hubble Frontier Fields: the impact of magnification uncertainties. Astrophys. J. 843, 129 (2017). Matthee, J. et al. Little Red Dots: an abundant population of faint AGN at z ~ 5 revealed by the EIGER and FRESCO JWST surveys. Preprint at https://doi.org/10.48550/ arXiv.2306.05448 (2023). Fujimoto, S. et al. CEERS spectroscopic confirmation of NIRCam-selected z ≳ 8 galaxy candidates with JWST/NIRSpec: initial characterization of their properties. Astrophys. J. Lett. 949, L25 (2023). Simmonds, C. et al. The ionizing photon production efficiency at z ∼ 6 for Lyman-alpha emitters using JEMS and MUSE. Mon. Not. R. Astron. Soc. 523, 5468–5486 (2023). Stanway, E. R. & Eldridge, J. J. Re-evaluating old stellar populations. Mon. Not. R. Astron. Soc. 479, 75–93 (2018). Pahl, A. J., Shapley, A., Steidel, C. C., Chen, Y. & Reddy, N. A. An uncontaminated measurement of the escaping Lyman continuum at z ∼ 3. Mon. Not. R. Astron. Soc. 505, 2447–2467 (2021). Atek, H., Richard, J., Kneib, J.-P. & Schaerer, D. The extreme faint end of the UV luminosity function at z ∼ 6 through gravitational telescopes: a comprehensive assessment of strong lensing uncertainties. Mon. Not. R. Astron. Soc. 479, 5184–5195 (2018). Gnedin, N. Y. & Madau, P. Modeling cosmic reionization. Living Rev. Comput. Astrophys. 8, 3 (2022). Chisholm, J. et al. The far-ultraviolet continuum slope as a Lyman Continuum escape estimator at high redshift. Mon. Not. R. Astron. Soc. 517, 5104–5120 (2022). Naidu, R. P. et al. Two remarkably luminous galaxy candidates at z ≈ 10−12 revealed by JWST. Astrophys. J. Lett. 940, L14 (2022). Naidu, R. P. et al. The HDUV Survey: six Lyman continuum emitter candidates at z ~ 2 revealed by HST UV Imaging. Astrophys. J. 847, 12 (2017). Vanzella, E. et al. Direct Lyman continuum and Ly α escape observed at redshift 4. Mon. Not. R. Astron. Soc. 476, L15–L19 (2018). Trebitsch, M., Blaizot, J., Rosdahl, J., Devriendt, J. & Slyz, A. Fluctuating feedback-regulated escape fraction of ionizing radiation in low-mass, high-redshift galaxies. Mon. Not. R. Astron. Soc. 470, 224–239 (2017). Ma, X. et al. No missing photons for reionization: moderate ionizing photon escape fractions from the FIRE-2 simulations. Mon. Not. R. Astron. Soc. 498, 2001–2017 (2020). Yeh, J. Y.-C. et al. The THESAN project: ionizing escape fractions of reionization-era galaxies. Mon. Not. R. Astron. Soc. 520, 2757–2780 (2023). Hutter, A., Dayal, P., Legrand, L., Gottlöber, S. & Yepes, G. Astraeus – III. The environment and physical properties of reionization sources. Mon. Not. R. Astron. Soc. 506, 215–228 (2021). Bergamini, P. et al. New high-precision strong lensing modeling of Abell 2744. Preparing for JWST observations. Astron. Astrophys. 670, A60 (2023). Furtak, L. J. et al. UNCOVERing the extended strong lensing structures of Abell 2744 with the deepest JWST imaging. Mon. Not. R. Astron. Soc. 523, 4568–4582 (2023).

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law. © The Author(s), under exclusive licence to Springer Nature Limited 2024

Methods Throughout the paper, we use absolute magnitudes36 and a standard cosmology with H0 = 70 km s−1 Mpc−1, ΩΛ = 0.7 and Ωm = 0.3.

Observations and sample selection The UNCOVER dataset consists of both imaging and spectroscopic observations of the lensing cluster A2744. The imaging observations and data reduction are described in detail in the survey and catalogue papers37,38. Here we briefly summarize the imaging and photometric products used in the present paper. HST imaging consists of seven broadband filters (F435W, F606W, F814W, F105W, F125W, F140W and F160W). The NIRCam39 images include short-wavelength broadband filters (F115W, F150W and F200W), long-wavelength broadbands (F277W, F356W and F444W) and one medium-band filter (F410M). Data were processed and drizzled into 0.04 arcsec pixel−1 mosaics using the Grism redshift and line analysis software for space-based spectroscopy40 (Grizli; v.1.6.0.dev99). In terms of ancillary data, the Hubble Frontier Fields (HFF) programme41 has obtained deep optical and NIR observations of the core area of A2744 with the Advanced Camera for Surveys (435W, F606W and F814W), and Wide-Field Camera Three (F105W, F125W, F140W and F160W). A wider area around the cluster has also been covered by the BUFFALO (Beyond Ultra-deep Frontier Fields and Legacy Observations) programme42 in almost identical broadband filters (without F435W and F140W). All HST observations were drizzled to the same pixel scale and the same orientation as the NIRCam mosaics. The second part of the UNCOVER programme consists of ultra-deep follow-up spectroscopy with the NIRSpec instrument43. Data were obtained between 31 July 2023 and 2 August 2023. Observations use the Prism mode and the Multi-Shutter Assembly44 of NIRSpec to observe more than 650 targets. To optimize background subtraction, each target was observed with a three-slitlet nodding strategy. Observations were split into seven pointings, with important overlap at the centre, providing total on-target exposure times ranging from about 2.7–17.4 h. The spectral resolution is wavelength dependent and varies between R ~ 30–300 over the full wavelength range λ ~ 0.6–5.3 µm. Data were reduced using the JWST/NIRSpec analysis software msaexp v.0.6.10. The processing is based on level 2 MAST products, using the CRDS context file jwst_1100.pmap. The software performs basic reduction steps, including flat-field, bias, 1/f noise and snowball correction, wavelength and photometric calibrations of individual exposure frames45. The extraction of one-dimensional (1D) spectra from individual exposures is operated on an inverse-weighted stack of two-dimensional spectrum in the dispersion direction, following an optimal extraction procedure46. Then the software fits a Gaussian profile along the cross-dispersion direction to define the 1D extraction aperture. Finally, we compute the final deep 1D spectrum by performing inverse-variance stacking of the individual spectra. To account for slit loss effects, we apply a wavelength-(broadband-)dependent correction factor to re-scale the 1D spectrum to the observed NIRCam aperture photometry. We show an example of the imaging and spectroscopic data in Extended Data Fig. 1. A clear Lyman break at rest-frame wavelength λrest = 1,216 Å is observed, together with multiple strong emission lines, including Hα + [NII], [OIII]λλ4960, 5008, Hβ, Hγ and [OII]λ3727. The selection of our sample combines several criteria to constrain the photometric redshifts of the sources. First, we applied a colour–colour selection, based on a flux dropout in the HST optical filters caused by rest-frame Lyman-α absorption by residual intergalactic hydrogen gas. This selection consolidates most of the sources identified in the HFF data24,47 at 6