Open science involves the use of practices across the research life cycle that facilitate the transparency, reproducibility, and availability of scientific products and output. Prominent open science practices include registration of study protocols and preanalysis plans; materials, data, and code sharing; and publication of summary findings in open access outlets (1). To achieve openness as the default approach, initiatives are trying to use a systems approach to engage stakeholders — namely, scientific journals, funding agencies, and professional societies (2, 3). Proponents hope to realign the research enterprise with the values of transparency and reproducibility (4).

Institutional review boards (IRBs) are overlooked, yet critical, stakeholders for proponents to engage in open science initiatives (5). For instance, IRBs can require modifications to and even disapprove data sharing plans based on information in the application, protocol, and informed consent forms investigators submit for IRB review (6). Conversely, IRB workflows can increase if a greater number of submissions require higher levels of IRB review because of data sharing plans that risk disclosure of identifiable private information about living individuals. Practical tutorials on ethical open science primarily focus on educating study investigators about their responsibilities to protect human research subjects. IRBs, in turn, need to be informed about open science practices and the impact of their growth on IRB responsibilities to protect human research subjects. In other words, IRB members and professionals themselves should be part of the movement toward open science.

Although IRBs are ethics bodies concerned with moral obligations (7), they are best understood through a regulatory lens: that is, IRBs are principally compliance offices responsible for ensuring that an institution’s federally funded human-subject research is compliant with specific federal regulations. Institutions can assign additional responsibilities to their IRBs; yet even in these instances, the federal regulations strongly influence (if not determine) IRB operations for human-subject research at an institution (6). Furthermore, the open science movement is gaining traction alongside historic regulatory changes that have privacy, confidentiality, and reduction of IRB overreach and administrative burdens as core goals (8). Consequently, IRBs likely will be resistant to open science reform efforts — including those based on ethically sound arguments — that risk breaches of privacy and confidentiality, IRB overreach, or unnecessary burdens on IRB administration.

Consequently, IRBs likely will be resistant to open science reform efforts — including those based on ethically sound arguments — that risk breaches of privacy and confidentiality, IRB overreach, or unnecessary burdens on IRB administration.

First and foremost, IRB functions and operations that aim to enable open scientific research must avoid noncompliance with the Common Rule (45 CFR 46), the US federal regulations for the protection of human subjects in research. The regulations most relevant to open science are concerned with minimizing the risk of harm resulting from breaches in the privacy of subjects and confidentiality of data. These regulations are also a top priority for IRBs, and the recent revisions of the Common Rule explicitly note that advances in data matching and reidentification exacerbate risks of breaches of privacy and confidentiality (8). The result is a tension between the undeniable role that IRBs have in mitigating the risk of these breaches and the mounting expectations for investigators to apply practices (particularly data, code, and materials sharing) that increase the probability of this same risk.

In light of the above, investigators who plan to use open science practices would benefit from IRB procedures, guidance, templates, and expertise that can clarify how to practice open science while remaining compliant with the Common Rule. For example, study proposal forms and submission systems could include a specific field for each open science practice so that IRBs could systematically collect this information from investigators in their review of research. To assist investigators who are developing proposals, IRBs should create written guidance explaining how open science practices can escalate the level of IRB review and potentially prevent IRB approval. At a minimum, this guidance should address definitions of identifiable private information and biospecimens (§46.102(e)(5) and (6)); analytic technologies or techniques that generate identifiable private information and biospecimen (§46.102(e)(7)(ii)); exemption categories with criteria on the identifiability of subjects (§46.104(d)(2)–(4)); and adequate procedures for protecting the privacy of subjects and maintaining confidentiality of data (§46.111(a)(7)).

In addition, IRBs should provide templates of informed and broad consent forms with approved language on data, code, and materials sharing — especially statements about the confidentiality of records and future research use (§46.116). To fill in the gaps, IRBs can invite experts in open science practices, data matching, and reidentification as members (§46.107(a)) or consultants who assist in study review on a case-by-case basis for issues requiring expertise not available on the IRB (§46.107(e)).

Second, IRB enablement of open scientific research should be consistent with the ethical principles detailed in the Belmont Report. Published in 1979 by the National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research, the Belmont Report identifies the ethical principles and guidelines for the protection of human subjects of research. These principles serve as a foundation for — and provide an analytical framework to guide — the resolution of ethical issues in human subject research that are not explicitly covered by the specific regulations in the Common Rule. Several regulations discussed above entail “respect for persons” (informed consent offering subjects the opportunity to choose what will happen to their data) and “beneficence” (assessing risks that may occur from privacy and confidentiality breaches) that might constrain the use of open science practices.

However, when findings are not made public or other products of research not shared, the subsequent loss of generalizable knowledge from research also entails a violation of these principles (7), especially to those who chose to participate in research because they want to contribute to the furthering of knowledge. Consequently, the ethical principles of the Belmont Report also suggest that IRBs actually have a responsibility to support open science practices.

Lastly, research organizations that directly influence IRBs could help refine and clarify the role of IRBs in the movement toward open science. In the United States, the Office for Human Research Protections (OHRP) is responsible for ensuring that human subject protection regulations are appropriately and effectively applied to the changing needs of the research community, such as those caused by the movement toward a research enterprise “open by design” (2). Consequently, OHRP could create guidance that helps IRBs establish policies and procedures that enable transparent human subject research in compliance with regulations. The Association for the Accreditation of Human Research Protection Programs (AAHRPP) and Public Responsibility in Medicine and Research (PRIM&R) could then incorporate this guidance into their accreditation programs for institutions and certification programs for IRB professionals. In particular, these groups could adjudicate provocative proposals, such as requiring investigators to register studies prior to IRB approval and checking that investigators have shared results before closing continuing review (7).

In summary, initiatives focused on integrating open science into the research enterprise should engage IRBs in these endeavors. In addition to ensuring compliance with Common Rule regulations, IRBs can assist open science proponents in navigating other federal regulations for protecting human subjects in specific contexts (e.g., US Food and Drug Administration and Health Insurance Portability and Accountability Act requirements), regulations at other levels of governance (international, state, local, and tribal), and supplemental mandates for IRB operations made by their host institutions. Including IRBs in the movement toward open science will not only facilitate the “contribution of open science to producing better science” (4) but also maintain continued public trust in the research enterprise by protecting its most important stakeholders: the members of the public who participate in research.

Opinion: Why institutional review boards should have a role in the open science movement was originally published in Proceedings of the National Academy of Sciences on Medium, where people are continuing the conversation by highlighting and responding to this story.

In 2008, Satoshi Nakamoto introduced the world to Bitcoin, a volatile digital currency that’s untethered to any specific institution or country (1). (Satoshi Nakamoto was quickly revealed to be a pseudonym; the true identity of the inventor or inventors remains unknown.) In the years that followed, the value of Bitcoin has soared, plummeted, and soared again while giving rise to a raft of new cryptocurrencies of varying stability and legitimacy. Bitcoin and the others have their boosters and their detractors (see Opinion: Valuation, liquidity price, and stability of cryptocurrencies, https://www.pnas.org/content/115/6/1131).

But Bitcoin wasn’t just revolutionary as a virtual currency innovation. It also introduced a novel way for investors and others to obtain, monitor, and trade that currency. Users record and secure financial transactions along with a timestamp using blockchain, a ledger system designed by Nakamoto that’s unlike any traditional accounting system used before. A blockchain isn’t stored on a central computer or controlled by a single boss; instead, it’s distributed, which means every user on a blockchain network has a copy of the entire ledger. Information is also protected by a consensus algorithm — a computer program that enables the network to validate transactions.

Blockchain will forever be linked to cryptocurrencies, but many researchers have begun to explore untapped applications for blockchain’s secure ledger approach. Its built-in mechanisms of trust and attribution make it appealing as a way to organize networks where people want to share information — a potentially big asset for tracking information in all sorts of science-related systems. Applications include forging new ways of managing distributed electrical grids, tracking regulated food and drugs, monitoring devices in the ever-growing “Internet of Things” (IoT), and perhaps even quickly and accurately documenting how new scientific findings emerge.

But blockchain itself is also susceptible to hype. Some proposed applications may not make sense. “Really, there are only two or three things blockchain gives you that other systems don’t,” says computer scientist David Mazières, codirector of the Center for Blockchain Research at Stanford University in Palo Alto, CA. Those include a way for two people to have a transaction, especially if they don’t trust each other, and a tamper-proof record of the history of transactions. If a system doesn’t need those components, he says, there’s likely a better way to manage the information.

“One of the things I ask people is, how does your project take advantage of specific features of blockchain?” Mazières adds. “If you have a centrally trusted party, you don’t need a blockchain.” Although excited by recent advances in blockchain technologies, he says many projects don’t require blockchain’s unique attributes. “There are a lot of misguided efforts, as well.”

Even suitable projects may need to tailor their approach; Bitcoin’s blockchain version is an untenable model, skeptics emphasize. Its consensus algorithm is too cumbersome and energy intensive, it’s relatively slow to add transactions, and it requires lots of data storage because every user must keep a copy of the entire ledger. Publicly available blockchains also present privacy problems. Hence, in certain settings blockchain may need to be tweaked to fit the security, energy, and user demands of networks beyond cryptocurrencies.

Most importantly, says electrical engineer Krishna Ratakonda at IBM, people need to have an incentive to participate. A blockchain that collects data from customers without offering anything in return likely won’t attract many boosters. “These are the situations,” says Ratakonda, “where things fall apart.”

Keeping Track

A blockchain consists of a “block,” data written as an alphanumeric code. This can only be added if the network approves and validates it. In the case of Bitcoin, this occurs via a “Proof of Work” algorithm. That means any blocks added to the chain must be encoded with the solution to a computational problem that’s hard to crack but easy to verify (such as finding the prime factors of a big number). The entire network can validate the solution, verify the addition, and add the block to the chain with a timestamp. Bitcoin’s power is its size: Any user can observe this solution, and information is protected by consensus.

Bitcoin users have an incentive to keep the network running because they have invested significant energy and resources into their participation. There’s also a financial incentive. Users who solve these problems receive units of cryptocurrency. The Proof of Work requirement, meanwhile, discourages hacking because breaking into the blockchain entails investing time and energy to decode a single block but without any incentive or support from the network.

“Proof of Work was a good starting point to keep everyone honest,” says Ratakonda. It was a concept that everyone could trust, even if they didn’t trust each other, he notes. “That was really a necessity.”

But Proof of Work does require serious computing power, leading computer scientists to seek alternative means of certifying trustworthiness. For example, the Brooklyn Microgrid, which lets consumers generate electricity through solar panels and other means and sell it to their neighbors, approves users individually and validates transactions instead of requiring Proof of Work.

Killer Apps

Already, both private and public entities are pursuing blockchain systems. Walmart has announced plans to work with IBM to develop a blockchain-based system that can track the distribution of lettuce to easily track the source of Salmonella contamination or other outbreaks (2). Some insurance companies are testing blockchain approaches to verify coverage of claims, cut down on fraud, and reduce day-to-day costs (3). Amazon offers a service for consumers to create and maintain blockchain networks using Hyperledger, a blockchain framework. And the US Food & Drug Administration has launched a pilot plan that uses blockchain to track prescribed drugs (4).

The uses are likely to multiply. Many online devices, part of a growing IoT, are notoriously insecure against hacks. That’s in part, because hackers can find and exploit software weaknesses before the users can detect them or manufacturers can fix them. Blockchain may offer a fix. If its ledger is tamper-proof, it could reveal every transaction among smart devices, including unauthorized access by hackers. A smart lock could reveal who entered a house, and when, for example, and detect unauthorized entries. In 2017, a consortium of tech giants including Cisco and Bosch announced plans to explore new projects using blockchain to improve IoT security. “Blockchain is a near-perfect auditing instrument,” says software engineer Konstantinos Christidis, who contributes to Hyperledger Fabric, a blockchain framework hosted by the Linux Foundation and used by IBM.

But although blockchain may detect intruders, it won’t be able to stop them. And devices may be hacked in such a way that they’re not flagged by a blockchain. Ratakonda points out that IoT networks already have robust ways to share data, and blockchain wouldn’t necessarily offer any benefit in such cases. “If the existing ones are sufficient, why would you need a blockchain?”

Securing Virtual Communities

“Smart” electric grids already allow people with solar panels or wind turbines to sell energy to a power company. But some experts say blockchain-based systems could enable those people to sell energy directly to their neighbors without the costs associated with a large, central power company, as in the case of the Brooklyn Microgrid. Such a peer-to-peer approach could turn energy grids into local systems.

Starting in 2016, researcher Esther Mengelkamp at the Karlsruhe Institute of Technology in Germany began looking for ways to “create an energy community that is completely self-sufficient in terms of electricity, heat, and other kinds of energy,” she says. Surveys showed that people were interested but that communities lacked the means to establish this kind system. So Mengelkamp started investigating blockchain-based set-ups, among the best systems with “smart” contracts, which can be used to automatically execute transactions when the buyers and sellers reach a certain price or demand (5).

Features such as smart contracts could also help address a pressing issue in biomedicine: monetizing personal data in a way that’s fair to sellers and buyers alike while protecting privacy and ensuring transparency with regard to how that data used. “Everybody is talking about the value of data,” says physicist and mathematician Alex Zhavoronkov at Hong Kong-based Insilico Medicine, a company that focuses on new drug discovery by combining artificial intelligence with blockchain. “But people are struggling to properly evaluate it. How much is your data worth?”

It could be worth a lot. Zhavoronkov sees blockchain as a powerful organizing force for health care and biomedical data. “I think it will eventually lead to a new data economics,” he says.

Techniques such as high-throughput genetic sequencing are generating large datasets, presenting quandaries about how to collect, use, analyze, and control personal data. Zhavoronkov thinks private blockchains — in which patients have control over who can see and use blocks of information — offer an appealing organizing principle (6). Unlike public blockchains, these “permissioned” networks would only be visible to authorized users (although the networks would not be impervious to hacks).

Zhavoronkov has been working with a company called Longenesis in Hong Kong that develops private blockchain tools for hospitals and pharmaceutical companies to use with patients. “We treat data items as assets,” says the company’s CEO, Garri Zmudze. Their idea, says Zmudze, is to find a way for hospitals and other institutions to monetize patient data but do it in a way that’s transparent to — and controlled by — patients who choose to participate. Guaranteeing the security and privacy of such a network, he says, will be a pivotal issue.

Users can check the blockchain ledger see who uses their data and when and what happened to it. Every doctor’s visit, prescription, treatment plan, and exchange of information would be recorded on the ledger. If people consent to participate in a clinical trial, their consent — and the terms of that consent — would be recorded. Using smart contracts, they could automate that process.

Zmudze goes further, suggesting that the features of blockchain can reduce barriers to medical advances. If a pharmaceutical company in Norway wants to run an international study on men of a certain age and with a certain health condition, it can use blockchain to create a transaction with the information request, instead of having to recruit patients through collaborating institutions. And if patients agree to join, their consent is logged on the ledger.

Perhaps the biggest obstacle to this approach is security because blockchain ledgers reveal all transactions. Zmudze thinks that challenge is surmountable: A ledger would include transactions about the data but not the data itself. That way, that data could be deleted — without having to delete the entire blockchain. “The patient has the mechanism not only to control the data but also the mechanism to see the history,” says Zmudze.

Who Did What, and When

At the University of Chicago’s Knowledge Lab, sociologist James Evans has been mulling over ways that blockchain could revolutionize the scientific process itself.

A traditional published scientific article bundles a lot of information into a tidy, discrete package that often fails to accurately represent the work that went into it. “It’s a very chunky system,” Evans says. In the past four decades, the amount of effort that goes into producing a scientific paper has increased dramatically. A single biomedical article published now contains, on average, twice as many experiments as an article published in the late 1970s (7). “We’re packing more and more stuff into these papers, and in some sense we’re giving less detail,” says Evans.

That detail could be important. Researchers typically don’t share negative results, reports on failed experimental approaches, or outlines of experimental methods that worked but didn’t support the researcher’s hypothesis. “An awareness of experimental dead ends could help others avoid repeating the same mistakes,” says Evans. And novel methods that didn’t work for one experiment might be useful for another.

A blockchain network might help, Evans says. Blockchain could break down individual contributions to a scientific discovery. Instead of only looking to a published paper for its findings and experimental results, researchers could share new methods and techniques — and the potentially informative dead ends they hit along the way.

“Until a few institutions buy into this concept, it’s not going to kick off by itself. It has to get enough momentum that it becomes a meaningful thing.”

— James Evans

Mazières sees some advantages. Blockchain’s tamper-proof timestamp could help researchers establish who did what and when and even help them stay on task. “It might be useful to basically have [researchers] commit to what they’re doing or planned to do,” he says.

Challenges Ahead

Blockchain is far from perfect. In addition to energy and computing demands, Ratakonda cites transparency as a big challenge. Some blockchain platforms, such as Hyperledger and Quorum, which was developed by JP Morgan Chase, are transparent to anyone. “These I would recommend,” says Ratakonda. He worries about the proliferation of private blockchains in which not all users have access to the entire ledger.

But Ratakonda says privacy is “probably the key challenge” to implementing blockchain networks. A hacker who virtually masquerades as a patient, for example, might still be able to gain access to electronic health records — or decipher identities on a public network or get a person’s private key and control his or her data. Blockchains encrypted with today’s best cryptographic algorithms — and never updated — may still be vulnerable to future hackers.

And there are sometimes better, easier alternatives, as Mazières and other researchers point out. Smart contracts on their own, for example, don’t justify using a blockchain. “A smart contract is just another word for a program,” he says, and it’s possible to build in these programs without using a blockchain. In many cases, blockchain may not be the appropriate technological solution to the problem. Case in point: The researchers behind the Landau Microgrid Project, which is still in use, ultimately decided not to use blockchain technology — opting instead for a different centralized system that still allows for peer-to-peer energy trading.

Even when blockchain offers potential advantages, institutional buy-in will be a big challenge. A prime example is a blockchain-based system that improves crediting and funding allocations among researchers. “Until a few institutions buy into this concept, it’s not going to kick off by itself,” says Evans. “It has to get enough momentum that it becomes a meaningful thing.”

Published under the PNAS license.

Core Concept: Blockchain offers applications well beyond Bitcoin but faces its own limitations was originally published in Proceedings of the National Academy of Sciences on Medium, where people are continuing the conversation by highlighting and responding to this story.

More than a decade ago, amphibian microbial ecologist Reid Harris watched a mother salamander as she marched in a figure-eight pattern through her clutch of soft, jellylike eggs. He knew that her strange walk, rubbing up against her brood, transferred beneficial bacteria from her skin onto the eggs to fight off fungal infection. Then something clicked. Harris wondered if she might also be showing him the solution to a scourge threatening hundreds of other amphibian species around the world.

Then, as now, amphibians were dying in droves. At least one-third of species face extinction, with at least 90 confirmed gone and another 500 in decline (1, 2). The major culprit is the skin disease chytridiomycosis, caused by the waterborne fungus Batrachochytrium dendrobatidis, or Bd for short. Under a microscope, fungal bodies anchor themselves in the translucence of frog skin and often group together to form a collection of small spheres. Chytridiomycosis kills by disrupting ion and fluid transport, eventually stopping the host’s heart (3).

A particularly deadly Bd lineage arose in the Korean peninsula sometime in the last century, the source of a pandemic that fanned out across the world through the global amphibian trade, infecting about 700 species in six global regions over the past five decades (4). Naïve immune systems (mostly those of frogs) couldn’t fight off the highly contagious and aggressive pathogen, and chytridiomycosis spread like a leveling wave, becoming the most devastating wildlife epidemic ever recorded.

The stakes were already high when Harris, now emeritus at James Madison University in Harrisonburg, VA, had his epiphany. He suspected that the salamander mother’s protective antifungal bacteria could represent a line of defense against Bd marching right under his nose. There might be many bacteria, for that matter, already living on the backs of frogs, salamanders, newts, and toads able to fight off the fungus, he thought. If those bacterial herds could be managed, like a wildlife park, perhaps amphibian microbiomes could be manipulated to better defend against Bd.

Harris and colleagues set about testing a variety of amphibian skin bacteria to see which inhibited Bd growth (5, 6). In a landmark study published in 2009, the researchers used an antifungal bacteria isolated from salamander skin, Janthinobacterium lividum, known as J. liv for short, to protect mountain yellow-legged frogs from severe chytridiomycosis in lab experiments.

The finding kicked off an ongoing effort by Harris, who is director of disease mitigation for the Amphibian Survival Alliance in London, and others to better understand the relationship between amphibian microbiomes and Bd. Studies in the last 10 years show that Bd can change host bacterial communities, hinting at a possible link between the microbiome and disease. In parallel, researchers have also been searching far and wide for antifungal microbes such as J. liv that can be marshaled to protect wild animals under attack by Bd. They’ve found thousands of candidates that fight fungi in the lab, but few have been field tested. With some luck, this early work could lead to real solutions to a scourge that has victimized amphibians for years. “The idea,” Harris says, “is to augment anti-Bd bacteria on amphibian skin.”

Small Leaps

Around 1990, when researchers first realized that amphibians were dying globally, they didn’t even know what was killing the animals (7). By late in the decade, histological exams of sick and dead frogs linked abnormalities in the skin to a fungus in the phylum Chytridiomycota (8). Still, many had a hard time believing a fungus alone could wreak such havoc. Bd was the first chytrid fungus known to parasitize vertebrates, says Joyce Longcore, a mycologist at the University of Maine in Orono, who in 1997 was the first person to culture Bd in the lab (9). “I knew it was a big deal,” she says, “because it meant this thing we knew was killing amphibians was now in culture, and we could [do] research with it.”

Researchers soon found they could clear Bd infections by bathing frogs in antifungal drugs (10). But deploying such treatments in the wild was unrealistic, in part, because the fungus persists in the environment, meaning treated animals would be reinfected. This need for lasting solutions for wildlife was one reason Harris looked toward bacteria. “Ideally,” he says “the anti-Bd bacteria persists and is transmitted to other amphibians in the area, either through direct contact or indirectly through environmental transmission.”

That was the dream when Harris set about his 2009 experiment on mountain yellow-legged frogs, denizens of streams in the Sierra Nevada hard-hit by Bd. Harris tested the efficacy of a J. liv bath on six captive frogs, reared in the lab from wild eggs, before exposing the animals to Bd. After 139 days, all the treated frogs survived whereas most of a control group that didn’t get the probiotic bath had died (11).

This early success with J. liv was so exciting that biologist Matthew Becker, then a graduate student collaborating with Harris, wondered if it would work for another, far more troubled species. Hordes of small, sunny-yellow Panamanian golden frogs once gathered along forest streams in the Cordillera Mountains that snake down Panama’s narrow spine. They are a national icon — immortalized in marketplace tchotchkes of golden frogs kissing, dancing, and getting married — and a symbol of hope. But golden frogs haven’t been seen in the wild since 2009. They likely succumbed to Bd, leaving captive populations as all that’s left of the species.

Knowing there was no hope at all for reintroducing the species to the wild without protection from Bd, Becker repeated Harris’ experiment in 2011 with 20 Panamanian goldens from the Houston Zoo. But soaking in a probiotic bath of J. liv before exposure to Bd didn’t save the little yellow frogs. The bacteria multiplied on the frogs’ skin over the first two weeks; then bacterial numbers started declining, and the Bd infection grew more severe, finally sickening and killing all the animals (12).

Such differing outcomes in two frog species, in two nearly identical experiments, crushed any expectation that J. liv might be a silver bullet against Bd. The research community took a collective step back, says Becker, now a biologist at Liberty University in Lynchburg, VA. Biologists began to ask more basic questions about amphibian microbiomes and their relationship with the disease.

Clues on the Battlefield

More than a thousand articles have come out since 2011 exploring microbiome-related questions. Researchers now have a better sense of the variety of bacterial communities on different amphibian species, which may also help explain their differing vulnerability to Bd and responses to probiotic treatment (13). Scientists have also learned that Bd infection is associated with changes in the composition of a host’s bacterial community, although it hasn’t been clear if Bd causes those changes or merely capitalizes on them (14).

In a 2014 study led by microbial ecologist Andrea Jani and ecological modeler Cheryl Briggs, both at the University of California, Santa Barbara, field surveys showed that the relative abundances of certain bacterial species changed with the severity of Bd infections. To test whether Bd was actively changing the microbiome, Jani and Briggs experimentally infected 42 frogs with the fungus and then compared their skin microbiomes with uninfected controls over time. They found that the pathogen did, in fact, change the microbiome’s composition. That doesn’t necessarily imply that the microbiome protects against Bd or that microbiome changes make Bd infections worse, Jani says. Some lab research suggests the microbiome can protect frogs, Jani adds, “but we don’t know that it ever does in nature.”

New research, published in the past year, is starting to provide answers by examining multiple wild populations. Up in the Pyrenees Mountains, lacing the border between France and Spain, the chubby little midwife toad fell into steep decline over the last decade. But some pockets of toad populations are now bouncing back. Microbial ecologist Kieran Bates, a postdoctoral researcher at the University of Oxford and a visiting researcher at Imperial College London, wondered what set rebounding populations apart.

So Bates hiked to mountain lakes to collect and compare skin bacteria from toads in both struggling and recovering populations, looking to see if they had caught the same genotype of Bd. Perhaps, Bates thought, the rebounding populations had a milder, less virulent form of the fungus. But genome sequencing revealed that all the toads had the same type of Bd. What varied was the makeup of their commensal bacteria (15).

Bates saw consistent differences in the number and relative abundance of bacterial species between rebounding populations and declining ones. The finding “implies the microbiome is important in driving disease dynamics,” Bates says. One potential explanation, which has yet to be tested, is that antifungal microbes are more abundant in recovering groups.

To get a better handle on what else might explain differences in amphibian microbiomes, Becker, and more than 40 other authors launched a major survey, the results of which were published this year (16). The researchers looked at what factors shape the assemblage of skin bacteria on each frog, toad, and salamander species across a broad range of environments around the world. One major takeaway is that the number of bacterial species inhabiting amphibian skin is inversely related to local temperatures, with greater bacterial diversity in temperate zones, and less in the tropics.

“Whether it’s warm or cold, and whether it’s warm or cold all the time or changing throughout the year, are the two pieces shaping what’s on their skin,” says coauthor Molly Bletz, a postdoctoral researcher in microbial and disease ecology at the University of Massachusetts in Boston. It’s possible that different microbes colonize the skin seasonally in temperate zones, she says, whereas a few species take over year-round in the tropics.

Understanding what drives bacterial diversity could help researchers develop tailor-made probiotics for different species, Bletz and Becker say, ever mindful that the whole point is to fight chytridiomycosis.

Fungus Fighters

To that end, researchers are also hunting for bacterial species that carry their own antifungal defenses in the hope that they could be recruited to give the animals a better fighting chance against Bd. Having tested thousands of candidate bacteria, researchers have found that many do fight the disease, at least in the lab.

Doug Woodhams regularly wades into lily pad–covered ponds in search of antifungal bacteria living on wild amphibians. He hikes with a backpack loaded with Petri dishes, cotton swabs, and ice packs to collect microbes off their skin. Back in his disease ecology lab at the University of Massachusetts, Boston, Woodhams cultures the samples — searching for microbes that make chemical weapons like the striking violet pigment violacein that J. liv uses to inhibit Bd (17). The ideal, researchers say, is to find bacteria already present on an amphibian’s own skin that could be amplified on the animal, sprayed around its environment, or both, to become a lasting defensive force.

Woodhams estimates that he and other researchers have collected and tested about 7,400 candidate bacteria in the last decade and found almost 2,000 that have anti-Bd properties in culture. He keeps a digital database tracking them all. But of the thousands of candidates identified so far, only about

“We didn’t think it was going to be simple to begin with. But it’s a lot more complex than what we first thought out, as with most things.”

— Matthew Becker

a dozen have been tested on living captive frogs and even fewer on wild populations. Cost and red tape, such as the permitting needed to do a test in the wild, are the main bottlenecks, he says.

One of the earliest field tests was cut short by weather but did suggest the bacterial approach holds promise. In 2010, biologist Vance Vredenburg of San Francisco State University in CA, a coauthor of Harris’ 2009 study, and his team caught and tagged 100 wild yellow-legged frogs in the high-elevation lakes of Dusy Basin in the Sierras, where Bd was a newcomer at the time. The researchers soaked 80 frogs in a J. liv bath, leaving the other 20 as untreated controls, then released them all. After year one, untreated frogs had more severe Bd infections. After two years, the researchers couldn’t find any untreated frogs, suggesting they had succumbed. But 39% of the treated group had survived, which was “pretty good proof of concept,” Harris says. The pond froze the following winter, killing all the remaining frogs, so the team couldn’t find out how long the protection would have lasted. Still, Harris believes that the two-year field trial demonstrated the feasibility of a probiotic approach, at least for this species.

Undaunted, Woodhams hopes to begin a new field trial in western Massachusetts in the next year or two, testing antifungal bacteria on eastern red-spotted newts, an East Coast salamander species. In this case, it’s not Bd that’s threatening, but rather another chytrid fungus called Bsal (Batrachochytrium salamandrivorans). First turning up in the Netherlands in 2008, it’s devastated local fire salamander populations there (18, 19). Bsal causes chytridiomycosis predominantly in newts and salamanders. Although it hasn’t yet shown up in North America, models predict that if it does, native amphibians are in grave danger (20). Woodhams wants to be ready when it arrives.

He’s now swabbing these newts, looking for one (or several) potential probiotics that can inhibit both Bd and Bsal in the lab. In the field trials, he’ll catch and tag newts at several ponds, give a subset the antifungal bacteria, and spray the same probiotic around the lip of ponds. He’ll continue monitoring the Bd load on the newts and other amphibians at the ponds, recapturing test subjects every few months. Although the animals don’t get particularly sick from Bd, seeing how a probiotic fights that fungus, which is already a part of the wild environment, could indicate its potential to fight the much-worse Bsal, Woodhams says.

Keeping Perspective

When J. liv protected yellow-legged frogs but not Panamanian goldens, it confirmed that a single probiotic would not suffice to battle Bd or similar fungal epidemics. The host’s own microbiome makeup will likely play a key role in determining which populations are most vulnerable to begin with and which weapons will defend best in each case.

It’s now clearer, for example, why the same probiotic didn’t work for yellow-legged frogs and golden frogs. J. liv naturally occurs in the Sierras, but in unpublished sequencing data, Becker found J. liv at extremely low levels on golden frogs, suggesting it is not a natural part of their microbiome. In fact, he says, the frogs may have a mechanism to fight off the bacteria, which prevents it from protecting the animals against Bd.

That’s one reason why researchers are confident that the safest candidate probiotics are probably those already growing on local amphibian skin. That way, the microbes are also not being introduced to novel habitats. Even so, the main concern when releasing bacteria into the wild is that they will have unexpected consequences for other pond-dwellers, such as fish, water bugs, or dragonfly larvae.

If and when researchers do find a probiotic that excels in lab and field trials, the next step is testing it in the wild. That could mean catching and soaking amphibians, as Vredenburg has done and as Woodhams plans to do, as well as inoculating the environment by spraying around pond edges, a method that Woodhams intends to carry out. Most researchers agree that antifungal microbes alone are unlikely to solve the larger chytrid crisis but believe the probiotics do have great potential as one tool to fight Bd and now Bsal.

Even if probiotics don’t provide a cure-all, the research may also point to less direct forms of treatment. In an article published last spring (21), Woodhams applied four antifungal bacteria to yellow-legged frogs and then exposed them to Bd. Although none of the bacteria protected the frogs, their presence did appear to dial up the frogs’ own immune defenses, inducing greater secretion of antimicrobial peptides from their skin. Perhaps, Woodhams says, there’ll be a way to induce frog skin to upregulate its own defenses using bacteria or maybe even friendly fungi to get the frog immune system to defend itself from pathogenic fungi.

Despite such signs of progress, probing and dissecting the microbiome and searching for ways to fight the Bd scourge with bacteria remain daunting challenges with remedies far from guaranteed. “We didn’t think it was going to be simple to begin with,” Becker says. “But it’s a lot more complex than what we first thought out, as with most things.”

Published under the PNAS license.

News Feature: Fighting a fungal scourge was originally published in Proceedings of the National Academy of Sciences on Medium, where people are continuing the conversation by highlighting and responding to this story.

Adam Mann

Late one night in October 2015, Andrew Wetzel was fretting. For 15 days, his cosmological models had been swirling virtual dark matter around cybernetic gas and dust and slowly generating a synthetic galaxy approximately the size of our own Milky Way, and Wetzel was about to receive the results. “I finally got the plot up to compare our simulation with the Milky Way,” recalls Wetzel, an assistant professor of physics at the University of California, Davis. “I went to bed very happy that night.”

That’s because, for the first time, a simulation had accurately reproduced a realistic population of dwarf galaxies near the Milky Way. These small companions orbit larger galaxies like planets around a star, and for a decade and a half, they had presented astronomers with a big problem. Simulations suggested that thousands of companion galaxies should surround the Milky Way, but telescopes had seen only a handful. Researchers seemed to be in the dark about some important aspect of the cosmos.

With the latest simulations from Wetzel and his team, the mystery of the missing satellites might appear to be explained (1). But more questions are emerging, with newly discovered galaxies more diverse than anyone predicted, and simulations, as yet, unable to recreate their richness. There may even turn out to be too many dwarfs for simulations to contend with. Future supercomputer models and next-generation telescopes will address these conundrums, potentially illuminating a critical stage in the growth of the universe.

Satellite Swarm

In 1999, computers had finally grown powerful enough to model the birth and evolution of an entire universe, including dark energy, the mysterious phenomenon accelerating the expansion of space-time, and dark matter, a still-unknown substance outweighing luminous material by a factor of five. Starting at the Big Bang, these simulations use the laws of gravity to show matter interacting with itself and coalescing into a web-like network of galaxies, exactly like the one astronomers see. Researchers can also use the computer models to zoom in on a particular galaxy and observe its growth in detail. By doing so for a virtual Milky Way, two separate teams noticed a curious incongruity — matter around an object like our galaxy should clump into hundreds or even thousands of satellites (2, 3). At the time, scientists knew of only 11 dwarf galaxies within a million light-years of the Milky Way.

Those early simulations tracked only dark matter because computers did not yet have the processing power to include the complex behavior of ordinary matter. And it seemed possible that many more dwarf galaxies were hiding just below the threshold of detection. Still, one troubling explanation for the size of the discrepancy was that physicists were missing some crucial knowledge about dark matter.

As supercomputing power improved, researchers could finally simulate the evolution of ordinary gas and dust alongside that of dark matter. In models created by Wetzel and his team, the gravity of our galaxy’s disk pulls in and shreds many dwarf galaxies. Supernovas drive out their gas and dust, lowering their mass and making them easier for the Milky Way to catch and destroy. At the end of the simulations, only 13 virtual dwarf galaxies with 100,000 solar masses or more remain, closely recreating the true satellite population for objects that size.

Other teams have performed similar modeling, some yielding results that agree with observational data and some that still show tension; the details generally depend on the specific physics of ordinary matter included. Yet, with the numbers from computer simulations and those from telescope observation finally showing some agreement, many astronomers consider this a done deal. “The missing satellites problem is not something that keeps me up at night anymore,” says astrophysicist Alex Drlica-Wagner at the Fermi National Accelerator Laboratory near Batavia, IL.

Ghost Galaxies

But new surprises have sprung up since. In 2005, the Sloan Digital Sky Survey (SDSS) — which uses the Apache Point Observatory in Sunspot, NM, to repeatedly image the entire night sky in supreme detail — uncovered the first ultra-faint dwarf galaxy (4). These wispy objects are less than 100,000 solar masses and hold far fewer stars than those found in the previously known satellites. SDSS and other sky surveys have since increased the number of known Milky Way companions to more than 50. “They were like ghosts all around us,” says astrophysicist Elena D’Onghia of the University of Wisconsin–Madison.

“Rather than a missing satellites problem, it’s like a found satellites problem.”

— James Bullock

Even state-of-the-art simulations still can’t get below the 100,000-solar mass limit, but theory says that the Milky Way should have preferentially destroyed such lightweight objects with its tidal gravitational forces. This has cosmologist James Bullock of the University of California, Irvine wondering if there is an opposite problem brewing. “Rather than a missing satellites problem, it’s like a found satellites problem,” he says.

Another reason to suspect this possibility is a bizarre beast named Antlia 2. Uncovered last November (5), this is one of the largest satellites ever seen around the Milky Way, yet its stars are so thinly spread out that they almost blend in with the background. Theorists suspect it was once a smaller dwarf that got puffed up by internal supernova explosions and then strung out further by tidal interactions with the Milky Way. Bullock and others believe that if one such hidden giant exists, there could be more out there — an entire population of diffuse ultra-faint companion galaxies that were not predicted.

Any new discrepancies could help illuminate the nature of dark matter. The usual assumption is that dark matter particles barely interact with one another, but if they interact more strongly through new fundamental forces, that may affect how they clump together and how many dwarf galaxies they seed.

Both observations and simulations will need to improve before any lingering satellite problems are entirely resolved. Many theorists are now working toward forming ultra-faints in their next-generation models. Wetzel’s latest models of the Milky Way galaxy can produce objects 10 times smaller than previously possible, which should be able to resolve ultra-faint galaxies for the first time, he says. Astrophysicist Coral Wheeler of the California Institute of Technology in Pasadena and others are working from a different direction, bringing all of a supercomputer’s power to bear on simulating a single dwarf galaxy. This could give astronomers a better idea of how they form, as well as their expected population.

Even if theory and observation match in the Milky Way’s near neighborhood, that is only part of the story. Many more ultra-faint satellite galaxies are likely to exist too far away for today’s telescopes to detect. For now, astronomers use statistical modeling to infer how many more they might uncover with better instruments. A 2008 study estimated 300 to 600 within 1.3 million light-years of our galaxy (6). That will be tested by the Large Synoptic Survey Telescope (LSST), an observatory on the mountain Cerro Pachón in Chile, which should begin making scientific observations early next decade. With its huge 8.4-meter mirror, LSST will be able to see much fainter objects than its predecessors, potentially capturing hundreds of new dwarf galaxies and testing the latest simulations.

First Light

LSST will also help identify the smallest dwarfs, which could illuminate another mystery of the universe — the end of the cosmic dark ages. Shortly after the Big Bang, dark matter is thought to have begun collapsing into clumps. Ordinary gas and dust would have been pulled in to form protogalaxies. Next came the epoch known as reionization, when the first stars turned on. Their light ionized and heated hydrogen throughout the universe, increasing its pressure. The strong gravity of massive protogalaxies could overcome this and retain the material needed to make new stars, but low-mass protogalaxies couldn’t hold on to the energetic, high-pressure hydrogen, essentially snuffing out star formation inside them.

The ultra-faints bear witness to this period in cosmic history. The theory is that they formed before reionization really got going and then shut down star formation when the gas got too hot. “These are the fossils of reionization,” says Wheeler. The stars in these dwarf galaxies are all old, which fits this view. Discovering the minimum mass of dwarf galaxies that survive the process would help researchers pin down the timing of reionization as well as potentially what drove it — was it mostly heftier galaxies or their lighter siblings?

The missing satellites problem is now 20 years old, but it is still helping reorient astronomers’ thinking. Some of the smallest cosmic objects have a great deal left to teach about the larger universe. “This is how science works,” says Wetzel. “You claim victory but you’re always looking for where there are tensions. It’s exciting and pushing us to where to go next.”

Published under the PNAS license.

Inner Workings: Dwarf galaxies pose new questions about dark matter and the early universe that… was originally published in Proceedings of the National Academy of Sciences on Medium, where people are continuing the conversation by highlighting and responding to this story.

Just over 20 years ago, a Lyme disease vaccine called LYMErix was approved for sale in the United States. Researchers designed the vaccine to prevent the transmission of the tick-borne pathogen Borellia burgdorferi, which spurs a bacterial infection that can cause fever, headaches, and joint pain if left untreated.

LYMErix was on the market for just four years. Concerns over adverse reactions and a lukewarm reception from public health agencies led the vaccine’s manufacturer, SmithKline Beecham, to shelve the product in 2002. Since then, the need for a vaccine has grown. An estimated 300,000 people are diagnosed with Lyme disease in the United States annually, and reported cases of the disease have tripled since the 1990s. In some counties in the northeast United States, disease incidence has increased more than 300% over a 20-year period. “The people who live along the northeast corridor among the Mississippi River valley have suffered greatly because there is no Lyme vaccine,” says Gregory Poland, a vaccine researcher at the Mayo Clinic in Rochester, MN.

Now, a new round of Lyme disease vaccines is in development. European biotech company Valneva is in phase 2 clinical trials for a vaccine against six strains of Borrelia, which causes the disease in Europe and the United States. And researchers are working to develop a vaccine against Lyme disease based on a vaccine for dogs that was released in 2016. The pressure is on this time around to make the “perfect vaccine,” says immunologist Richard Marconi at the Virginia Commonwealth University in Richmond, one of the researchers working on the dog vaccine–inspired approach.

No doubt, concerns about the original vaccine are something drug developers and public health officials will have to grapple with as new vaccines strive for success. But although fears related to the original Lyme disease vaccine haven’t gone away, the science of vaccine development has changed significantly. Researchers are counting on that new science to put some of those fears to rest.

Still, despite ample promise — Valneva’s vaccine has obtained fast-track status from the US Food and Drug Administration (FDA) — there’s no way to know for sure what sort of reception these new vaccines will receive from the public. “There will be one more shot at bringing a human Lyme vaccine to market,” says Marconi, “and if that vaccine fails, the market will essentially disappear.”

First Try

In the late 1990s, the first Lyme disease vaccine came about during a “perfect storm” of adverse circumstances, says Poland, who’s also worked as a consultant for Valneva on two occasions. “Around that time, we begin to see just generally, a growing amount of anti-vaccine sentiment,” he says.

But anti-vaxxers aside, LYMErix was not a smashing success (1). The vaccine was only 80% effective at preventing the disease, and public health officials had no data on the effectiveness or safety of the vaccine for those under age 15 years, a cohort more at risk of developing Lyme disease because of outdoor play. Plus, it took three doses within two tick seasons to be effective, and intermittent boosters were, by all indications, going to be necessary, says Poland, because of the way the vaccine targets Borrelia.

The vaccine worked by inducing antibodies to outer surface protein A (OspA), which is expressed on the surface of the different strains of Borrelia. But Borrelia only exposes OspA when still inside the tick. When a tick takes in a vaccinated host’s blood, the antibodies in that blood can identify OspA and kill the pathogen.

However, if enough Borrelia makes it into a host, the bacteria essentially stops expressing those crucial OspA proteins. Hence, for an OspA vaccine to work, the person vaccinated has to produce enough antibodies to block transmission of the bacteria into the host — and that means many booster shots could be necessary (2).

Another strike against the vaccine: the public health agency responsible for recommending its use issued an ambiguous recommendation, saying that people “can use” the vaccine, rather than “must use,” says Poland.

But the biggest roadblock stemmed from concerns that the vaccine caused an autoimmune reaction in some vaccine recipients. The vaccine targeted a part of an antigen called the epitope, the section of the protein an antibody attaches to. Some researchers posited that this particular epitope resembled one found on human cells, which would then be attacked by antibodies, causing an arthritis-like reaction in the body (3). Others found a certain genotype could be associated with greater risk of both chronic Lyme disease symptoms and more risk of autoimmune reactions to the vaccine.

Although further review by the FDA and researchers found no connection between the vaccine and such adverse effects, the panel wanted to increase the number of enrollees for the phase 4 (aftermarket) safety trials. However, those results never came to light: with class-action lawsuits emerging from patients and dwindling market interest, LYMErix was removed from the market in 2002. And with market prospects dim, manufacturer Pasteur Mérieux Connaught dropped a separate, second vaccine, called ImuLyme, that had been in the works. Other companies, including Gaithersburg, MD-based MedImmune, have also tried and failed to get a vaccine off the ground.

Today, two main research strategies are targeting the bacteria that causes Lyme disease: one that continues to focus on OspA and another targeting a different outer surface protein. Each takes a different approach in hopes of avoiding the mistakes of the past.

New Candidates

Valneva is testing an OspA vaccine that is effective against six different subspecies (4) of Borrelia found in the United States and Europe, an improvement from the previous vaccine, which only worked on one serotype, Borrelia burgdorferi. Safety trials have also recruited people older than age two years, suggesting this vaccine could potentially be used in children. Although it targets the same outer surface proteins as the original Lyme disease vaccine, Valneva’s candidate does not make use of the epitope that was associated with purported autoimmune reactions, notes Valneva Chief Executive Officer Thomas Lingelbach. Instead, the researchers combined portions of different epitopes to target a variety of Borrelia strains. Lingelbach makes a point of noting that LYMErix was proven to be safe. Nevertheless, Valneva researchers decided to eliminate its epitope from their antigen to “not give any room to scientific speculation.”

But as with LYMErix, targeting OspA antigens means producing enough antibodies to stop pathogen transmission. Once inside the host, Borrelia switches outer surface proteins and would, therefore, slip by antibodies seeking OspA. Hence, researchers are testing the highest safe dose in clinical trials as they try to reduce the need for yearly booster shots. The goal is to deliver primary immunization over three doses and require a booster shot a year later that remains effective for another three years.

Patient groups may take more convincing. Ten percent to 20% of those diagnosed with Lyme disease deal with recurring symptoms, called chronic Lyme disease. And that constellation of chronic neurological symptoms is similar to what many patient advocates believe are the side effects of the previous Lyme disease vaccine. Pat Smith, president of the Lyme Disease Association, says any new vaccine manufacturer will first need to clarify what went wrong with the previous vaccine. “We need to see what happened, what really happened with LYMErix, and it needs to be done in a transparent fashion,” she says.

To address these concerns and make a more effective vaccine, Marconi and his colleagues have targeted an entirely different Osp — OspC. This obviates any need for the much-maligned OspA epitope while delivering immunity with fewer booster shots.

Unlike OspA, OspC is expressed once the bacteria enter the host bloodstream, meaning antibodies will continue to attack the bacteria in the human host, says Marconi. This could mean fewer booster shots; fewer circulating antibodies would be required to strike the bacteria before it transfers out of the tick. However, previous attempts at making an OspC vaccine proved challenging; the protein is highly variable, meaning it might protect against only one particular strain of Borrelia, says Marconi.

Marconi and his colleagues studied the variable epitopes of OspC that trigger a productive antibody response. Then, they combined those variations of epitopes into one single protein. They call it a chimeritope. This became a core component of a canine vaccine, called Vanguard crLyme, that the FDA approved in 2016. In that vaccine, both OspC and OspA antibodies work in concert to kill bacteria both inside and outside the tick (5).

In developing a vaccine for humans, Marconi wants to identify new antigens that will play the same role as OspA. Anticipating better acceptance of such a vaccine because of its track record in pets, Marconi says his group hopes to place its vaccine into phase 1 trials in the next few years.

Ideal Target

The ideal strategy, however, wouldn’t target bacteria or viruses but the tick itself. Such a vaccine would target proteins in the tick’s saliva or midgut to impair tick feeding before the parasite can transmit Lyme disease or any pathogen. In one study, researchers were able to lower transmission of Borrelia by vaccinating mice against a protein found in tick saliva and the gut, called tick histamine release factor (6).

A European academic consortium, known as the Anti-tick Vaccines to Prevent Tick-borne Diseases in Europe (ANTIDotE), is wrapping up a five-year project to identify tick proteins that could serve as vaccine targets to impair tick feeding or pathogen transmission or both, notes Joppe Hovius, the project coordinator. “We looked at proteins that could be involved in the transmission of multiple tick-borne pathogens,” he says (7).

To narrow the list of potential antigens, researchers used animal models to gauge how developing antibodies to these proteins would hinder tick feeding. Consortium researchers narrowed their candidate proteins to a “few promising results” that impede Lyme disease transmission and tick-borne encephalitis, Hovius says. The results of these experiments are not yet published; Hovius and his team hope to replicate their work as they seek more funding.

“There will be people who will be convinced by data. Then there will be others who will never be convinced.”

— Thomas Lingelbach

The big challenge so far is the genetic diversity among tick populations. One tick saliva protein may not be applicable to others. But if the researchers can find a widely conserved tick protein, they could potentially target the tick species in both Europe and the United States with one vaccine.

All of these advances look promising, and a better vaccine, whatever the approach, may be within reach. But the challenges of widespread acceptance still cast a pall on the field. Patient groups may have helped thwart LYMErix with public criticism, but, Smith says, “the failure of that vaccine occurred because it was so problematic, and that issue was not really addressed.” Clarifying the improvements inherent in new strategies will be crucial, she says.

It remains unclear how Valneva will address past failures if, in fact, the company brings a vaccine to market. With the vaccine still in safety trials, Lingelbach says it’s too early to consider what a marketing plan might look like. He does expect more acceptance of a Lyme disease vaccine in Europe because they’ve already been using a vaccine for tick-borne encephalitis virus for more than 40 years. By the time the company seeks a license for its vaccine, some 16,000 to 18,000 people will have been tested with their candidate, he adds.

But Lingelbach acknowledges the uphill battle. “There will be people who will be convinced by data,” he says. “Then there will be others who will never be convinced.”

Published under the PNAS license.

Inner Workings: Lyme disease vaccines face familiar challenges, both societal and scientific was originally published in Proceedings of the National Academy of Sciences on Medium, where people are continuing the conversation by highlighting and responding to this story.

Adam Mann

On September 3, 2018, an unpowered experimental sailplane made history by flying into the stratosphere. After leaving from El Calafate, a town near the Southern Patagonian Ice Field in Argentina, glider pilots Jim Payne and Tim Gardner surfed on enormous airborne waves emanating from the Andes Mountains. They achieved a world record height of 23,203 meters — higher than a Lockheed U-2 spy plane’s cruising altitude. No other unpowered aircraft has ever achieved such elevation. It was a feat made possible by not just human ingenuity but also the incredible strength of a phenomenon known as atmospheric gravity waves.

Unlike the similarly named (and perhaps more famous) cosmic gravitational waves, which are ripples in the fabric of space-time, atmospheric gravity waves are a wholly terrestrial occurrence. They emerge when parcels of air are forced upward, for instance, by a tall mountain range, moving from a dense atmospheric layer to a thinner one. The heavier blobs of air then succumb to the force of gravity and fall back down, resulting in a periodic oscillation that can carry energy and momentum over vast distances. They are the smallest atmospheric waves that researchers study, generally between a few hundred meters and a few hundred kilometers in size. But their cumulative effects can be dramatic.

The actions of atmospheric gravity waves bring energy from the troposphere — the lowest atmospheric layer, which ends at an altitude of about 10 kilometers — out past the edge of space, 500 or more kilometers above Earth’s surface. They play important roles in daily weather as well as long-term climate fluctuations, decelerating powerful jet streams and affecting the circulating polar vortexes that appear over our planet’s poles. They influence atmospheric turbulence, temperature, and chemistry, yet limited computing power continues to prevent their direct inclusion in most atmospheric simulations.

Although researchers believe they have a fairly good handle on the basics of atmospheric gravity waves, a more detailed and realistic look at the waves’ behavior would provide increased accuracy for simulations predicting both local weather and potential adverse effects from climate change. In recent years, new efforts have sprung up to provide better observations and modeling of these important players in our planet’s dynamic atmosphere.

“They seem small,” says atmospheric researcher Joan Alexander of NorthWest Research Associates in Boulder, CO. “But they’re affecting forecasting and predictions on many timescales, and if we don’t include them we get big biases in our models.”

A Rock in a Pond

Researchers have known about atmospheric gravity waves since the 19th century, but the waves remained obscure until 1960. Around then, atmospheric researcher Colin Hines of the Canadian Defense Research Board began wondering why the trails left by meteors, as they whooshed through the part of the ionosphere that sits at 80 to 110 kilometers above the surface, often contained irregular wave-like distortions. His calculations showed how gravity waves from below were crashing like ocean waves on the shore of the upper atmosphere, inducing the strange meteor trail patterns. Hines developed the first widely accepted theory on how gravity waves worked and how they influenced other events in the atmosphere (1).

Atmospheric gravity waves can be generated in a variety of ways, such as when air flows up and over mountains or in the updraft from a powerful thunderstorm. The air parcels may generate ripples that travel horizontally over the planet in a process similar to ocean waves rising and falling at the surface of the sea. More often, the undulations climb vertically, rising into different layers of the Earth’s atmosphere, which becomes exponentially thinner as altitude increases. Conservation of energy forces the waves to grow in amplitude in the lower-density air, allowing them to travel to distant elevations where they impart energy and momentum.

“It’s like dropping a rock in a pond,” says climate researcher Julio Bacmeister of the National Center for Atmospheric Research in Boulder. “When a parcel hits the stratosphere, it generates oscillations that propagate in all directions.”

Atmospheric gravity waves were first put into global weather forecasting models in the 1980s. Before that, numerical simulations that included the mesosphere — an atmospheric region between 50 and 85 kilometers — had produced temperature patterns that didn’t match our planet’s actual fluctuations. The mesosphere is particularly tricky to model because it gets unexpectedly colder in the summer and warmer in the winter, and at the time nobody could figure out why. Researchers realized that atmospheric gravity waves from near the surface were crashing against airstreams in the mesosphere, changing their normal flows. The effect forced winds in the mesosphere to pour from the Earth’s summer hemisphere toward the winter one, creating the anomalous temperature inversion (2).

But these early models were too coarse-grained to resolve the small-scale oscillations of relatively tiny atmospheric gravity waves. Instead, theorists relied on what is known as a parameterization — essentially a workaround in which researchers use their knowledge of the gravity waves’ actions to tell the simulation how to react to their influences, even while the researchers were unable to include the waves themselves (3, 4). Parameterizations have proven successful in reproducing most of our planet’s real-world activities and are still used in most cases for modeling the effects of atmospheric gravity waves. But Bacmeister says that the basic methods haven’t changed much since the first schemes were put together more than 30 years ago, and many in the field are now looking to improve their simulations.

Butterfly Effect

It’s relatively easy to parameterize the effects of stationary topographic features such as mountains. But winds can sometimes travel around mountains rather than over them, so a simplistic treatment might induce too much or too little change. The results of transient phenomena such as storm fronts and clouds are even harder to parameterize accurately because such events arise sporadically from complex causes. Gravity waves can also generate turbulence or smaller secondary waves, but state of the art computing power still isn’t sufficient to resolve such subtleties.

“It’s kind of like the butterfly effect,” says atmospheric researcher David Fritts, founder of the Colorado division of the aerospace company GATS, referring to the idea that tiny actions, such as a butterfly flapping its wings, can snowball into major consequences. “If you don’t allow for the butterflies, it’s hard to get the effect.”

At the International Space Science Institute in Bern, Switzerland, Alexander has recently brought together a team of experts in data assimilation and global weather and climate forecasting models to figure out how to best advance atmospheric gravity wave simulations. Although the small waves don’t affect the overall prediction that the Earth will warm in the coming decades, their actions can be crucial in determining whether some particular part of the globe will experience a greater potential for droughts, flooding, or other detrimental outcomes. The group hopes that getting a better handle on atmospheric gravity waves’ sources, increasing the resolution of their models, and producing improvements in parameterization schemes will help provide crucial information for planners seeking to minimize weather-related damage from routine events such as rain and temperature extremes, as well as from major incidents such as hurricanes, tsunamis, and tornadoes, which have been estimated to impact 30% of the global economy (5).

For the past decade, Earth-observing satellite missions have assisted with such missions, producing a golden era in climatologic and atmospheric data. Mostly invisible to the human eye, atmospheric gravity waves have historically been difficult to observe. Weather balloons launched twice a day from hundreds of locations around the planet provide information about atmospheric temperature and pressure and wind speeds and directions for researchers but do not provide full coverage over the oceans or throughout the entire vertical height of the atmosphere. From their perch in space, instruments such as NASA’s Atmospheric Infrared Sounder (AIRS) and the European Space Agency’s Atmospheric Dynamics Mission Aeolus (ADM-Aeolus) can watch for changes in winds, temperature, and humidity around the entire Earth, allowing researchers to build complex predictive weather models.

So far, there have been no satellite missions devoted specifically to atmospheric gravity waves, and researchers often rely on incidental data gathered by instruments on satellites designed for other purposes. But earlier this year, NASA green-lit the Atmospheric Waves Experiment (AWE), an observatory that will monitor how gravity waves high in the ionosphere interfere with radio and Global Positioning System (GPS) communications. The AWE is expected to fly to the International Space Station in August 2022.

Ground-based detectors such as radar and Light Detection and Ranging (LIDAR) instruments that can spot atmospheric shifts are also being deployed to gather gravity wave data. Combined with satellite observations, information from these campaigns is helping answer remaining questions in the field. For instance, global circulation models often fail to correctly replicate the behavior of the Antarctic polar vortex, an organized swirl of air over our planet’s South Pole. The simulations predict much stronger and colder winds than meteorologists have observed; researchers suspect that this is partly because of gravity waves originating from 60 degrees latitude south — a particularly windy region of our planet that does not get much observational coverage.

At the British Antarctic Survey, atmospheric physicist Tracy Moffat-Griffin is helping lead the Drake Passage Southern Ocean Wave Experiment (DRAGON-WEX), which will use radar and satellites to measure atmospheric gravity waves over the Southern Ocean and the Drake Passage at the tip of South America to investigate how they might be contributing to this discrepancy. Moffat-Griffin has also helped organize the Antarctic Gravity Wave Instrument Network (ANGWIN), an international effort to provide more comprehensive coverage of gravity waves over the South Pole. In their undulations, atmospheric gravity waves can create cold spots where clouds form. The surface of polar stratospheric clouds often induces ozone-destroying chemical reactions, contributing to the Earth’s ozone hole, so monitoring the waves is an important part of predicting its size and behavior.

“As in all things in meteorology, you have models and observations. You improve the models with observation, and there’s an interplay between the two.”

— Marvin Geller

Atmospheric gravity waves aren’t only an Earthly phenomenon. Observations from Japan’s Akatsuki mission currently orbiting Venus have revealed a 10,000-kilometer-wide bow-shaped structure in the planet’s upper atmosphere, which periodically appears and remains stationary for days at a time despite the turbulent Venusian clouds (6). Researchers believe the structure is caused by updrafts coming from mountains on the surface. Other gravity wave effects have been spotted from probes orbiting Mars, Jupiter, and Saturn (7⇓–9). Studying such events can help inform researchers’ general understanding of gravity waves, although thus far researchers have relatively little data with which to scrutinize these off-world cases, notes Moffat-Griffin.

Together with improved simulations, better monitoring will help atmospheric researchers gain a much greater understanding of gravity waves on Earth and beyond. “As in all things in meteorology, you have models and observations,” says atmospheric researcher Marvin Geller of Stony Brook University in New York. “You improve the models with observation, and there’s an interplay between the two.”

Published under the PNAS license.

Core Concept: To improve weather and climate models, researchers are chasing atmospheric gravity… was originally published in Proceedings of the National Academy of Sciences on Medium, where people are continuing the conversation by highlighting and responding to this story.

The distant rock has offered clues about planet formation and the state of the early solar system.

Within the cloud of icy rocks at the edge of the solar system lie objects that have remained virtually untouched since their formation more than four billion years ago. Last January, NASA’s New Horizons spacecraft made the first flyby of one such primitive sample, an object known as 2014 MU69 and nicknamed “Ultima Thule” (although that label has proved controversial*). After New Horizons’ successful flyby of Pluto in 2015, researchers were keen to study a primordial body that was within the craft’s reach. With MU69, that dream became a reality. The tiny object, one of only three possible destinations discovered after the mission launched, turned out to be an incredible target. “I really think we hit the jackpot,” says New Horizons principal investigator Alan Stern, of the Southwest Research Institute (SwRI) in Boulder, CO.

At first glance, MU69 looked much as researchers had imagined a pristine Kuiper Belt object (KBO) would appear, with a dark surface, rich in water ice and organic material, and relatively unscarred by craters. But when they looked closer, it offered plenty of surprises. From its shape to its spin to its composition, the distant rock is providing planetary researchers with a wealth of information about the conditions in the vicinity of the sun 4.5 billion years ago, and it’s even helping solve a decades-old puzzle about how the planets formed.

A Squashed Snowman

Researchers revealed their first results at the annual Lunar and Planetary Sciences conference in the Woodlands, TX, in March.† MU69 stretches roughly 35 kilometers (20 miles) from tip to tip and spins on its axis every 15.9 hours. Unlike most solar system inhabitants, which spin with their equator aimed at the sun, MU69 lies on its side, pole pointing sunward. The two lobes are roughly the same color, a reddish hue similar to other KBOs observed from Earth, which likely comes from materials known as tholins, formed when radiation from the sun modifies organic molecules.

Although the first image of MU69 suggested a pair of spheres, later observations revealed a big surprise. Thule is very thin, and Ultima even flatter: only 7 km thick and 22 km across. “If it’s not the flattest body in the solar system, it’s up there,” says team member William McKinnon, of Washington University in St. Louis, MO.

Most solar system objects this size are roughly spheroidal, with only a handful of exceptions. About a year before the flyby, Jeff Moore, New Horizons co-investigator at NASA Ames Research Center in Moffett Field, CA, had discussed the possibility that MU69 could be like some of the exceptions, showing images of the flattened moons of Saturn in a talk to fellow researchers. Forming in the disk of Saturn’s rings, rather than a more spherical cloud of particles, some of these moons have the appearance of a squashed walnut. “People scoffed,” Moore says. “They said, ‘whatever we’re seeing isn’t going to look like that.’”

But then New Horizons revealed the flattened hamburger lobes of Ultima and Thule. The similarity to Saturn’s moons suggests that they, too, were created in a swarm of particles spinning fast enough to form a disk. What’s more, the original axes of rotation for Ultima and Thule are almost parallel, only a few degrees apart.‡ The similar orientation suggests that the pair formed from a single cloud of material before joining together.

The final clue came from how the two lobes have retained their rounded shapes. If Ultima and Thule had slammed together at collision speeds typical of the Kuiper Belt, a few hundred meters per second, that would have been enough to deform or shatter them. Instead, they seem to have drifted gently into one another. Stern compares them to docking ships that came together without the violence common in the rest of the solar system.

This gentle docking maneuver makes sense if the pair started as part of the same cloud of material. Then they would have had similar speeds and wound up orbiting one another. As the pair orbited, other nearby material could have been ejected, reducing the system’s angular momentum and pulling the pair together; or something else entirely could have caused the two to become one. “That’s the bit I don’t think people have worked out yet,” says planetary formation researcher Harold Levison (SwRI), who is not on the New Horizons team but was present at the flyby.

A slow collision would mean that MU69’s interior probably remains loosely packed, a fluffy aggregate of ice and rock similar to the interior of comets (1). But had the lobes instead collided at high speed, their interiors would have been compacted.

Meanwhile, small lumps on the surface may be remnants from the birth cloud, the last pieces to accrete onto the KBO. Co-investigator Will Grundy, of Lowell Observatory in Flagstaff, AZ, says that the lumps appear to be homogeneous across the pair, again suggesting that they originated in the same cloud of debris.

Missing Craters

The early life of MU69 was a flurry of excitement. The two lobes formed and merged, and residual heat from the formation may have powered a bit of surface activity. But after half a billion years, most of the commotion had subsided. MU69 is too far out for the sun’s heat to change its surface. Apart from the occasional collision with a small piece of debris, the celestial snowman has remained essentially the same for the last four billion years. Although that may sound like a boring existence, it’s exactly why researchers wanted to visit the Kuiper Belt. Visiting MU69 is almost like traveling back in time to the birth of the solar system — with a few caveats.

When New Horizons flew by Pluto, researchers were stunned to see how few intermediate-sized craters§ covered the surface of the dwarf planet and its large moon, Charon. MU69 shows the same paucity, with only a handful of midsized excavations. “We see some very nice-looking craters on MU69, just not very many of them,” says co-investigator Kelsi Singer (SwRI).

The dearth reveals a lot about MU69’s virtually unchanged environment over the last 4.5 billion years. Objects that would generate such craters would be tens to hundreds of meters across. These could form directly from gas and dust or be chipped off a larger body in a collision, but the shortage of craters suggests that neither of these processes prevailed in the Kuiper Belt. “It looks like it just didn’t initially make a lot of small objects and just kept not making them,” Singer says. New Horizons’ resolution made it difficult to definitively identify craters, but there aren’t many potential midsized scars that could have come from medium impacts. The largest potential crater scars Thule, stretching about halfway across the lobe’s diameter. Nicknamed Maryland, the crater has raised questions on the timing of its formation.

Although the colliding object would have been about 10 times smaller than the crater it excavated, the collision would still have been significant. One way to avoid breaking Ultima and Thule apart would be to have Maryland smash into the smaller lobe before the pair came together, but it’s hard to determine which came first. Based on her previous work with impact craters, Singer doesn’t think that the Maryland smackdown would have been that destructive. Moore agrees, pointing out that the survival of MU69 would depend on its interior. The porous interiors of comets act like Styrofoam, absorbing the impact energy of a collision, he says. If MU69 is similarly porous on the inside, it could have absorbed the blow from the impact without flying apart. “The impact that formed Maryland might have had essentially no effects on the two lobes of the binary,” Moore says.

A comet-like interior could also explain some geological differences between the two lobes. Although Ultima is relatively smooth, Thule has more features on its surface. Moore thinks many of those features may have formed when the two lobes merged. He says that the orientation of grooves, regional uplifts, and surface depressions on Thule make them more likely to be tied to its gentle bump with Ultima than the hit from Maryland. Because it’s smaller, Thule would have been harder hit by the impact than the larger Ultima. “Thule basically might have been split at the seams,” Moore says, allowing methane, nitrogen, and other volatile material to escape its interior. Movement of that material could have driven the processes that shaped Thule’s landscape.

Planet Building

Fully understanding the images of MU69 will take some time, but the first view is already helping shed some light on how planets formed. As rocky pieces move through the gas and dust around a newborn star, they should collide with one another to form larger and larger bodies. It is thought that they would have built up to objects like MU69, which then came together to form larger objects known as planetesimals, which in turn built the planets, from Mercury to Pluto (2). But there is a missing chapter to this story.

Simulations can turn bits of dust into so-called pebbles up to a meter in size and transform larger, kilometer-sized objects into planetary cores, but they had a problem in the middle, going from small pebbles to rocks kilometers across (3). The computer models suggested collisions would blow apart growing planetary embryos in this size range. Known as the meter-sized barrier, the problem has plagued planetary formation for decades.

In recent years, new theories have emerged to help with the problem. One idea, known as the streaming instability (4), suggests that gas in the disk creates drag, forcing these small objects to concentrate into clumps. Gravity then pulls the clumps together to create kilometer-sized objects in a matter of tens of thousands to hundreds of thousands of years, depending on the solar distance no midsized material is necessary. A popular extension of this theory, known as pebble accretion, suggests that smaller objects continue to fall onto the kilometer-sized objects to rapidly build the cores of giant planets.

Both the pancake shape of MU69 and its dearth of midsized craters seem to validate these theories. If large objects form via the streaming instability, there will be few projectiles in the range of tens of meters to create those craters. And according to David Nesvorny, who models early solar system formation at SwRI, the flat lobes of MU69 are also telling. The shapes are too flat to have come from the initial spin of a cloud of material. Such a large cloud would have had too much angular momentum to hold together and, instead, split into two objects. A loose collection of material in the outer solar system could have gathered together, as proposed by the streaming instability. Gravity would have pulled most of it together, creating objects flatter than a sphere but not quite pancake-like, then pebble accretion would have piled most of the remaining pieces on top, creating the flattened lobes visible today. Eventually, Ultima and Thule would have drawn closer together, gently docking with one another.

The researchers haven’t yet simulated the process for MU69 specifically, although they hope to. If correct it could solve the problem of how to build a planet. “That meter-sized barrier that we all stressed out about, we just jumped right over it,” says astronomer Kevin Walsh of SwRI. “We all suspected something like that but these observations from the Kuiper Belt really seem to clinch it.”

Previous models have always worked with spherical bodies rather than pancakes, in part because of limited computer power, according to McKinnon. “[Simulations] might have tracked angular momentum, but they certainly didn’t try to build shapes,” he says. The new results from MU69 have likely changed that. “Now I’m sure people will [attempt to simulate final shapes] because we have the capabilities to do this sort of stuff.”

MU69 may live at the edge of the solar system, but the processes that formed it probably worked throughout, Nesvorny says. Earth and other planets likely formed from similarly flattened objects. And the process is not limited to our planetary neighborhood. “Elsewhere in exoplanetary systems, it might be quite common to form pancake and contact binaries,” he says.

New Horizons will be sending back data on MU69 until the end of 2020, and plenty of further study will be needed, but it already has provided a wealth of insight about both the Kuiper Belt and the entire solar system. “We had no right to expect that we would get so much from this one small body,” Stern says, “but we have.”

News Feature: “Celestial snowman” starts to reveal its secrets was originally published in Proceedings of the National Academy of Sciences on Medium, where people are continuing the conversation by highlighting and responding to this story.

Benjamin S. Halpern, Richard S. Cottrell, Julia L. Blanchard, Lex Bouwman, Halley E. Froehlich, Jessica A. Gephart, Nis Sand Jacobsen, Caitlin D. Kuempel, Peter B. McIntyre, Marc Metian, Daniel D. Moran, Kirsty L. Nash, Johannes Többen, and David R. Williams

Improving global food systems is essential to addressing climate change, mitigating biodiversity loss, and meeting both sustainability and human development goals. International assessments from the Intergovernmental Panel on Climate Change and Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services and business and technology innovations such as lab-grown and plant-based meat, as well as many consumer diet trends, can all be traced to studies that identify undesirable impacts of certain food systems. Yet the evidence underpinning many widely touted recommendations about what to grow and eat is remarkably sparse and generally biased.

We know that not all food is created equal in terms of environmental impact (1⇓–3). However, most past research has focused on only a few key foods (e.g., beef and staple crops) and only a few environmental stressors (particularly greenhouse gas emissions). In addition, these studies tend to be confined to a few countries, and many nations suffer from poor knowledge transfer between the scientific community and the public. Even the most wide-ranging assessments made to date (2, 4, 5) contain significant gaps (Fig. 1). These biases arise, first, because substantial portions of the global food system are inaccurately or insufficiently reported or effectively “hidden” from standard statistics (e.g., inland and small-scale fisheries, bushmeat, backyard farming) and, second, because of an incomplete knowledge of the suite of environmental impacts and how these propagate through the many linkages among food systems.

The web of food impacts and interactions are challenging to map, let alone fully assess. Yet characterizing these linkages is essential to understand the true cumulative impact of food production (2, 3). Without doing so, society risks unknowingly exceeding regional, or even global, environmental boundaries or missing opportunities to steer food consumption and policy toward more sustainable foods and practices. Furthermore, gaps in assessments are unevenly distributed. For example, in developing regions with rapidly growing human populations, unassessed foods underpin the nutrition of millions and provide essential sources of protein for more than three billion people (3, 6, 7). It is hard to imagine developing science-based food policy for regions where so many of the common foods have never been assessed in terms of environmental impacts.

Two major gaps in our understanding — substantial holes in food assessment studies and the nearly complete absence of linkages among foods in these assessments — limit our ability to link environmental impacts to food security. We show how this lack of understanding undermines decision makers’ capacity to develop policies ensuring the planet can sustainably meet future human food demands. We recommend tackling these gaps by improving the environmental impact assessment of food production and supporting the development of effective, integrative food policy.

Underassessed Foods and Impacts

Only by putting all foods on the “same table” can we comprehensively evaluate their relative environment impacts, and in turn, develop effective and efficient policies that ensure greater production with lower environmental cost across all foods. “Underassessment” is an acute problem for two reasons. Hidden, missing, and under-reported (hereafter collectively called underassessed) foods represent substantial amounts of food production in many countries around the world. Synthesizing data from open-access databases and published literature, we show that underassessed foods represent more than half of animal production in 76 countries (see Ext. Data 2; Fig. S1) and more than 25% of total food in 40 countries (Fig. 2). For those countries lacking data on their major food groups, it is impossible to generate informed plans for sustainability and food security. In addition, even prominent well-studied foods commonly suffer from very narrow examinations — often only greenhouse gas emissions implications and, perhaps, land and freshwater uses are taken into account (Fig. 3). The omission of marine stressors, such as acidification and plastic pollution, diminishes the value of existing marine food assessments and could expose many regions of the world to tragic environmental breakdowns.

We hypothesize that the environmental impacts of underassessed foods are likely more diverse than conventionally reported commercial agriculture and livestock systems. This variation stems, in part, from the great diversity of species that are farmed and harvested within these underassessed sectors and from the greater range of production and capture methods. Conventionally reported food consists of relatively few livestock species (cows, chickens, pigs, sheep, and goats) and just ten crops (8). In contrast, the wild harvest of aquatic and terrestrial species includes thousands of species, habitats, and capture technologies. These differences will affect the resilience of targeted stocks — and thus the sustainability of the harvest — but also lead to more types of environmental impact and greater variability within each impact (2). For example, although greenhouse gas emissions from marine fisheries are generally lower than estimates from other animal protein production, emissions can vary widely depending on motorization, species, gear type, and location (9).

The additional diversity that these underassessed foods provide — both within and between food types — offers greater opportunities for policymakers to mitigate the environmental impacts of food production. Policy could strategically support the kinds of production most likely to minimize the trade-offs between environmental and other objectives. For example, bushmeat hunting and inland fisheries have potentially serious impacts on native biodiversity and the societal benefits arising from biodiversity. However, these foods are often a vital protein source for marginalized people and have extremely low impacts on water pollution or greenhouse gas emissions compared with other foods. Only by incorporating data on these foods and associated environmental impacts into food system accounting will consumers be able to make fully informed choices and policymakers to identify and evaluate trade-offs within and across food systems. This information is requisite to managing for lower overall food system impacts.

Critical Linkages

Interdependencies among food production sectors are commonplace: manure is used to fertilize crops, land converted to crops is no longer available as pasture, and fish farms displace local wild fisheries. Some linkages are well studied and reveal the complexity of interactions. For example, studies of the ubiquitous linkage between fed animals and their feed have highlighted less-known results that roughly 4% of crops are fed to farmed seafood (8) and 27% of wild-caught fish is used as feed for farmed fish, pigs, and poultry (4).

But few studies have comprehensively accounted for the suite of linkages among systems, despite their significance for the cumulative impact of food production, and more indirect linkages abound. Agricultural pollution limits the location and yield of inland and coastal aquaculture (10), disease and genetic escapes from aquaculture pens can influence the health of wild fish stocks and vice versa (11), and landscape homogenization in the search for more efficient production can disrupt natural pollination and potentially decrease yields (12). Linkages can also generate win-win opportunities. For example, feeding seaweed to cows dramatically reduces methane emissions from cattle while also potentially reducing the need to convert new land to crop cultivation (13). Even more poorly known are the potential linkages among different types of underassessed foods — as when unregulated overfishing (often by foreign fleets) spurs bushmeat hunting (14) or when shrimp farming destroys mangroves critical as a nursery habitat for harvested wild stocks.

Connections among food sectors are likely becoming more pervasive as commodity markets become fully global and production expands into new areas, creating complex cascades of interdependencies. To avoid unintended negative impacts, as well as to identify and capitalize on potential win-win opportunities, requires clear, quantitative assessments of both positive and negative linkages among food systems that can, for example, help farmers make informed choices regarding raw material inputs and waste management. Integration of impacts and linkages into comprehensive assessments will also allow policymakers to evaluate new production systems in the context of wider regional and global food systems and sustainability objectives.

Only if researchers provide assessments and methodologies to support an integrative understanding will decision makers be able to anticipate how and where the different dimensions of environmental impacts will change under a given action. Such an approach is vital if we are to achieve increased production and harvest while decreasing the environmental impact of food. It is no longer safe to rely on natural resilience to ensure sustainable food production, given that many buffers to the planet’s ecological limits are already close to exhausted (15), and in some regions, local limits have already been surpassed.

Smarter Food Policy

The global food system is too complex, diverse, and contingent on environmental and socioeconomic context to allow for simple or singular policy recommendations. However, because food systems are linked globally through international supply chains and distributed environmental impacts (e.g., greenhouse gases), achieving sustainable food production for 10 billion people will require global-level coordination, through mechanisms such as intergovernmental agreements and trade deals. Such collaboration will need to favor more environmentally efficient forms of production while meeting food security and other sustainable development goals, decisions that require comprehensive and balanced knowledge on the impacts of different food production systems and linkages among them.

Addressing these shortcomings requires three linked approaches: First, funding organizations and researchers should prioritize the collation, collection, and synthesis of spatially explicit primary data on underassessed foods and underassessed impacts for well-studied foods. New studies should be evaluated for how much they augment existing data, and we suggest prioritizing initial data collection and collation and synthesis of understudied foods and environmental impacts over ever-more refined studies of well-documented systems. For example, many tropical and arid crops (e.g., sorghum, coffee, and sugar cane) and aquatic species (e.g., inland fisheries and seaweed mariculture) are underassessed and are critical in developing nations.

The diversity of food systems, the range of objectives and stakeholders, and the linkages among different components make forward-looking, proactive, and comprehensive food policy a major challenge at all levels from individuals to international organizations.

Importantly, including many underassessed foods in comprehensive assessments is immediately tractable because necessary data often already exist (e.g., country-level production and global-gridded databases of fisheries). Rapid advances in satellite remote sensing and artificial intelligence are already being used for monitoring biodiversity, as well as fishing, farming, and hunting activities. Where data gaps exist, they should be prioritized for expanded monitoring but could be temporarily filled by estimation or simulation. For example, remote sensing and geographic information system data can be integrated into environmental suitability models to predict bushmeat hunting patterns in Central Africa (16). Although some of these approaches are still in their infancy, they are improving as new data and technologies rapidly come online.

Second, the linkages among different foods and aspects of the food system need to be better defined and quantified. International organizations such as the Food and Agriculture Organization and other United Nations agencies are well placed to coordinate and lead such efforts, with the collation and provision of existing datasets being a priority. The quality of these datasets, however, critically depends on the statistical capacities of the reporting countries. In short, these agencies need additional financial support if they are to better track and report linkages among foods. Existing economic and environmental methods have great potential to improve our understanding and documentation of these linkages — for example, the use of hydrological models to predict where eutrophication will impact downstream food production. But again, it is important to restate that even the most sophisticated models today would omit large parts of the global food system because the data underpinning them are underreported.

Finally, we issue an urgent call for national and international policymakers to increase dialogue and the sharing of data and even personnel among departments tasked with different aspects of food systems. We must reverse trends toward decreased funding for gathering national and subnational food statistics. Coordination and adequate staffing are vital for enabling comparable and comprehensive assessments of food’s environmental impacts, understanding trade-offs and synergies among different objectives, and setting sustainable dietary recommendations for consumers. For example, if all Americans followed the US Department of Agriculture–recommended diet, US greenhouse gas emissions from food would actually increase by 12% (17). Although shifts toward alternative diet scenarios have been estimated to reduce land-use and greenhouse gases, the full suite of impacts (Ext. Data S3) and the linkages associated with feed, including fertilizer and other indirect and localized impacts, have not yet been accounted for in these assessments.

The diversity of food systems, the range of objectives and stakeholders, and the linkages among different components make forward-looking, proactive, and comprehensive food policy a major challenge at all levels from individuals to international organizations. To make effective food policy now and to plan for future scenarios and shifting linkages among food sectors we need a comprehensive baseline understanding that includes all food systems, their direct and indirect connections, and how these are changing. Armed with this information, a wide variety of positive outcomes becomes possible through improving efficiency within particular food systems, favoring positive over negative linkages, encouraging the consumption of low-impact foods, or any combination of these strategies. For example, governments could favor subsidies for low-impact or positive-linkage foods, food companies could develop (and market) foods they know scientifically are more environmentally efficient, and consumers could make better-informed decisions at supermarkets and restaurants. Some of these opportunities are already occurring because of increased awareness of the health and environmental impacts of food, but many are currently missed because of major gaps in assessments of the impacts and linkages in food systems.

We ask policymakers to work toward responsible food production that respects the planet’s limits. And we call on researchers to put all foods on the same table in even-handed environmental impact assessments.

Opinion: Putting all foods on the same table: Achieving sustainable food systems requires full… was originally published in Proceedings of the National Academy of Sciences on Medium, where people are continuing the conversation by highlighting and responding to this story.

Neuroscientists believe that immediately after a mammal is born, brain cells called microglia spring into action, pruning away connections between neurons. This may be a way for the brain to refine its neural networks. Until recently, neurobiologists had assumed the microglia worked by phagocytosis, reaching out a cup-shaped tendril to swallow the neuron-to-neuron bridges whole. But in 2017 when Laetitia Weinhard, then a graduate student at EMBL Rome in Italy, observed pruning via a combination of light-sheet and electron microscopy, she saw something others had missed.

Weinhard, now a postdoc at New York University, discovered that the microglia were instead nibbling off bits of neural connections in a process called trogocytosis, as she reported in 2018 (1). Her group is one of many that have recently found trogocytosis happening in unexpected places and for unsuspected purposes. “I think trogocytosis is really underappreciated,” says Katy Ralston, a microbiologist at the University of California, Davis. “It’s exploding…where it’s happening, and also what it’s being used for.”

Phagocytosis, when one cell engulfs something large such as another cell, is well understood among biologists. So is endocytosis, whereby one cell takes up something from outside into a membranous bubble. There’s also pinocytosis, the cellular ingestion of liquid. But trogocytosis — a word derived from the Greek for “gnaw” or “nibble” — entails one cell nipping bits off another.

The process has gone by different names, such as partial phagocytosis or cell cannibalism, but the basics are the same. A nibbling cell must somehow sever a piece from another cell. That bite then transfers membranes, and sometimes cytoplasmic contents, to the nibbler. This cellular gnawing was first observed decades ago among amoebas (2, 3) and then between cells in the immune system (4).

Using advanced microscopy to catch trogocytosis in action, researchers are seeing it in a diverse set of organisms and processes. Some nibbling cells prune away unwanted cellular bits during development. Others gnaw cells to death. Certain bacteria were recently spotted taking advantage of this interaction between two host cells to slip from an infected cell into an uninfected one. Trogocytosis might even be a drug target for diseases such as allergies and amoebic dysentery.

Information Exchange

Trogocytosis has received the most attention over the years among researchers who study immune cells, which seem to nibble other cells to sample what’s inside. An immune cell may even take a bit of gnawed-off membrane as its own, allowing it to display new proteins on its own surface to activate immune responses. When a cell advertises proteins it has acquired this way, it’s called cross-dressing.

In 2017, immunologist Kensuke Miyake and colleagues from the Tokyo Medical and Dental University in Japan reported an example of cross-dressing that solved a puzzle surrounding a type of immune cells called basophils (5). These cells are involved in activating an immune response to parasite attacks, as well as some allergic reactions, such as asthma.

Researchers had found that basophils use proteins called MHC class II (MHC-II) to present identifying bits of pathogen material, called antigens, to T cells (6⇓–8). This activates the T cells to mount a defense. But other researchers said this couldn’t be because basophils lacked the ability to efficiently process antigens or even make much of the MHC-II proteins (9). It was a “big controversy,” says Miyake.

Supporting the latter theory, Miyake and colleagues found little evidence that basophils could express genes to make their own MHC-II proteins. Yet, they also observed MHC-II protein on basophil surfaces. When they cocultured basophils with dendritic cells, a type well known for its antigen-presenting prowess, the team found the answer. Basophils were nibbling the dendritic cells, specifically taking pieces of membrane with MHC-II proteins to display as their own. If this basophil cross-dressing promotes allergy, then blocking it could be a route to treatment, suggests Miyake.

Nibbled to Death

In the case of certain cancers, it might be better to encourage trogocytosis. Recently, researchers have observed immune cells called neutrophils doing more extensive, even deadly, nibbling of cancer cells — revealing that the mechanism can involve more than just a taste of another cell (10).

Timo van den Berg, an immunologist at the Amsterdam University Medical Center and Sanquin Research in the Netherlands, was interested in how neutrophils kill cancer cells that have been labeled with antibody medicines. His team found that the neutrophils had to touch the cancer cells to do them in. But the neutrophils weren’t using one of their usual techniques, the release of granules full of toxins. Nor were they producing damaging reactive oxygen species.

Neutrophils cannot phagocytose something as large as a cancer cell, but, as van den Berg’s group saw under the microscope, neutrophils can nibble at a cancer cell until it disintegrates. “It was like piranhas attacking prey,” he says.

Van den Berg’s group coined a term for this death by a thousand nibbles: trogoptosis, based on the same Greek root as the word for another kind of cell death, apoptosis. He predicts neutrophils may kill a variety of targets in this way: “I think we are just seeing the tip of the iceberg.” Neutrophils have also been spotted nibbling large parasites to death (11).

Ralston, meanwhile, has observed as infectious amoebas kill host cells by nibbling. The parasite Entamoeba histolytica, which causes dysentery, destroys cells by trogocytosis (12). In a recent study in mBio, she shows that the amoeba also cross-dresses, displaying proteins from immune cells it has nibbled to disguise itself and evade immune attack (13).

Trogo-Transit

Like E. histolytica, other pathogens have been caught taking advantage of host-cell trogocytosis. Tom Kawula, a microbiologist at Washington State University in Pullman, and colleagues study a bacterium called Francisella tularensis, which causes tularemia infections in the skin, lungs, eyes, or digestive system. The team had noticed that F. tularensis traveled quickly between immune cells called macrophages. The researchers could treat a dish full of macrophages with, say, 500 bacteria, and within a day, more than 500 macrophages would be infected. There hadn’t been enough time for the bacteria to take hold, reproduce, and be released to infect other cells. “It wasn’t adding up,” says Kawula.

Graduate student Shaun Steele decided to take video of the process using bacteria labeled with green fluorescent protein. He saw infected and uninfected macrophages interacting and bacteria transferring between them before the cells separated again. When Steele labeled the cytosol of infected cells with red dye and then sorted them by color with flow cytometry, he saw that as uninfected cells picked up some red cytosol from another cell, the green bacteria came along at the same time (14).

“It was like piranhas attacking prey.”

— Timo van den Berg

Kawula thinks that Franscicella is not doing anything specific to instigate trogocytosis but rather hitching a ride in a natural host process that’s stimulated by infection. The team found that another bacterium, Salmonella enterica serovar Typhimurium, can also travel by trogocytosis. Being able to transit from cell to cell this way is an advantage for the pathogens, Kawula says, because they’re never exposed to immune attack in the extracellular space.

Gnawing During Development

Trogocytosis appears to be a basic tool that cells can use when phagocytosis may be too blunt an instrument. As in the case of microglia pruning neural connections, cells within a single organism use trogocytosis during development to nip and tuck other cells into the right shape.

Cornelius Gross, the neurobiologist who led the Rome team (1), suspects that in the brain, microglia use whole-cell phagocytosis early in development to eliminate entire neurons. Then, after birth, they switch to trogocytosis for a finer tuning of neural connections.

At the NYU School of Medicine in New York, developmental biologist Jeremy Nance and colleagues found trogocytosis similarly used for pruning in a different system: the primordial germ cells (PGCs) of the nematode Caenorhabditis elegans. Each larval worm possesses exactly two PGCs, which will go on to create all of its sperm and, in the case of hermaphrodites, eggs.

Researchers have long known that the PGCs, which are nestled up against the embryo’s intestines, produce large lobes that disappear by the time the worm matures (15). In a recent experiment, Nance’s then–graduate student Yusuff Abdu, now working at the Rockefeller University in New York, labeled the PGC membranes red to see what those lobes were doing. What he saw was unexpected: bits of red membrane turned up in the intestinal cells. Using fast light-sheet microscopy, he caught the intestinal cells in action, nibbling the lobes (16).

The researchers also used worms deficient in different genes to probe the mechanisms of trogocytosis: does it work like phagocytosis or endocytosis? The team found that nibbling uses elements of both those mechanisms, relying on not only the actin cytoskeleton like phagocytosis but also endocytic proteins to sever the neck of the lobe being consumed.

Nance suspects the purpose of trogocytosis in this context is protective. The PGC lobes are chock-full of mitochondria, which produce not just energy but also DNA-damaging free radicals. That’s dangerous in the germline. By nibbling off the lobes, the intestinal cells may be safeguarding the genomes of future generations.

Although there are only a handful of clear examples of trogocytosis, some researchers suspect the phenomenon is widespread. New microscopes are helping researchers spot the process, which happens quickly. And an uptick in papers describing trogocytosis should help others recognize it in their own systems, suggests Ralston. “It’s evolving, or emerging, as a big theme.”

Published under the PNAS license.

Core Concept: Cells nibble one another via the under-appreciated process of trogocytosis was originally published in Proceedings of the National Academy of Sciences on Medium, where people are continuing the conversation by highlighting and responding to this story.

In early 2017, researchers managed to slip a flexible sliver of polymer next to a pig’s heart. The device — placed between the heart and the fibrous wall that encases it, called the pericardium — squished and expanded with each contraction. It also converted the physical strain of its movement into electrical energy stashed into a capacitor. When hooked up to a commercial pacemaker, the device produced a steady pulse of 130 beats per minute — effectively using the heart’s own mechanical motions to power an implanted pacemaker.

The study, led by materials scientist Zhong Lin Wang of the Georgia Institute of Technology in Atlanta, demonstrated how self-powered electronics could lead to a new generation of smaller pacemakers and other implanted devices that improve safety and performance, last longer, and obviate the need for invasive battery-replacement surgeries (1). Wang and several other researchers envision a world in which batteries are a thing of the past: deep-brain implants powered by electric impulses, cochlear implants fueled by inner ear vibrations, bone implants that stimulate tissue repair, and shoes and clothes that turn every bodily movement into a source of power.

Crystal Clear

When solid materials, such as some ceramics, crystals, proteins, or bone, are pulled or stretched, charged particles in their crystal structure are dislodged, creating a flow of ions — and thus, an electric current known as piezoelectricity. This ability to convert mechanical energy to electricity has been known for centuries; one of its most famous applications was in developing piezoelectric sonar transducers to detect submarines during World War I. Piezoelectric materials are now widely used in cigarette lighters, quartz watches, and more.

But while studying the potential of these materials for self-powered biomedical devices, Wang’s team ran into a problem. Some of the devices they’d engineered delivered a power output several-fold higher than others — as it turned out, because of a tiny air bubble squished between polymer layers within the devices. That air resulted in an accumulation of electric charge as the layers rubbed against each other, a phenomenon known as the triboelectric effect. The discovery led Wang and his colleagues to shift their focus to triboelectric power, eventually leading to the recently published study.

Both triboelectric and piezoelectric systems rely on converting mechanical energy that’s abundant — and goes wasted in the human body — into electrical energy to power pacemakers and other devices. Such implants currently rely on batteries, which only last a few years before needing replacement. A pacemaker, for example, is replaced every 7 to 12 years via an invasive, complex surgery.

Self-powered devices also promise to be smaller, so they can be used in tight spaces, such as for implants within the brain. And unlike batteries, these devices can be made with flexible, biocompatible materials.

Researchers working in this niche field see the potential for electric power in every bodily movement: heartbeats, the breathing movements of the diaphragm, or the pounding of footsteps on pavement. Both theoretical and experimental studies have found that the energy from these motions can easily power implanted devices. For example, theoretical studies have found that simply breathing can generate 0.83 watts of power, whereas a cardiac pacemaker only needs 50 microwatts to work for 7 years (2, 3).

Inner Driving Force

In 2010, then-doctoral student Amin Karami was researching ways to harvest energy from the mechanical motions of airplane wings when his father passed away suddenly after a heart attack. He began to wonder: could he adapt his research to improve cardiac pacemakers? The work relied on a ceramic-based piezoelectric material. “But I hadn’t seen anything really fundamental about human energy harvesting,” recalls Karami, who’s now an assistant professor of mechanical engineering at the State University of New York in Buffalo. “I started thinking about where we see sustained motion in the human body both in healthy and sick people and that led me to the heartbeat itself.”

Although piezoelectric materials have a long history in other fields, adapting the technology for biomedical uses was not easy, Karami says. For starters, the heartbeat is wildly variable depending on a person’s activities. Second, technical details, such as optimizing power requirements for different working conditions of commercial pacemakers, are proprietary industrial knowledge — Karami and Wang say that to do their experiments, they had to do take a commercial pacemaker and dissect it in the lab to figure out the requirements. Any device that fits into the body would also need to be extremely reliable. “With energy harvesting for non-biomedical uses, it’s mostly cost that matters, not size,” Karami explains. “But this is a totally different world — if a pacemaker doesn’t work the person could die.” Karami, his doctoral advisor Daniel Inman, and their colleagues designed a slim, ceramic-based device that could, in principle, replace a pacemaker’s battery (4).

Like Karami, Canan Dagdeviren, assistant professor of media arts and sciences at the Massachusetts Institute of Technology (MIT) in Cambridge, also began by adapting technologies from other fields. While she was a graduate student, Dagdeviren found that many of her labmates worked on stretchable devices to generate piezoelectricity but not on harvesting biological energy. In 2014, Dagdeviren and her colleagues demonstrated how a tattoo-like, polymer-based device could be used to generate piezoelectricity and power a pacemaker placed in a mouse (5) (see Science and Culture: Wearable tech meets tattoo art in a bid to revolutionize both, https://www.pnas.org/content/115/14/3504). “Our paper was one of the first to show that such a device works even when the chest is closed, and that the average power generation is even higher than a modern pacemaker needs,” she says.

Ceramic-based materials work best to generate power, but they’re also brittle, Dagdeviren says. Polymer-based implants that are flexible and stretchable are sturdier and, thus, better suited to biomedical uses.

In recent studies, Dagdeviren and others have expanded their work to a range of applications beyond pacemakers, such as pH sensors, cochlear implants, sensors that can be placed in the brain to monitor seizures, and bone graft materials that can accelerate tissue repair. Most are powered by piezoelectric materials rather than other modes of energy harvesting.

For example, biomedical engineer Thanh Nguyen and his team at the University of Connecticut in Storrs used a biodegradable polymer to design a sensor that detects the force of breathing movements when placed on the diaphragm in mice (6). Typically, a device inserted via a catheter is used to track these forces as a way to monitor breathing under anesthesia or in patients who have respiratory disorders. The piezoelectric device was less invasive than the catheter-based sensor, delivered comparable data, and — unlike traditional sensors, which must be surgically removed — simply dissolves on its own several days after the procedure, Nguyen says.

Others have also tried to tap into alternative sources of energy in the body by using thermoelectric or chemo-electric devices. Electrical engineer Patrick Mercier at the University of California, San Diego, studies devices that generate electricity from body fluids — converting chemical to electrical energy — as a way to create wearable sensors to monitor the body’s many chemical activities. Such sensors could power themselves and offer a way to unobtrusively monitor blood glucose or other metabolites. In a recent study, Mercier and his colleagues created a wearable biofuel cell that generated electricity from the lactate present in human sweat (7). “If done correctly, such cells can extract quite a bit of power from these fuels,” Mercier says. “The potential power output is much larger than many other forms of energy harvesting.”

But because such systems are based on enzymes catalyzing electrochemical reactions, they have a short lifespan of just days. Further studies are needed to improve their longevity and shelf stability, Mercier says. “How can we make a sensor well-calibrated, so it can sit on the shelf for 6 months, you put it on and you know the readout is correct?”

Chemo-electric devices could function well as wearables and have the potential to work in parts of the body where relative movement is limited. Because of their short lifespan, they’re more likely to be useful in wearables than to power long-term implants such as pacemakers.

Slow Pace

Although these self-powered devices are inching into clinical trials, researchers point out that there’s currently no standardized way to evaluate these devices or report data on the design, efficiency, or lifespan of such implants. That makes it difficult to replicate results or performance.

Wang adds that both piezoelectric and triboelectric devices face three major hurdles: devising the right coating to ensure the device is biocompatible, finding ways to implant the device in a noninvasive manner, and ensuring that the implant is durable.

Durability and the lifespan of the device are particularly important considerations for the new generation of leadless pacemakers, Karami notes. Older pacemakers were box-like structures placed near the collarbone with wires leading to the heart. But more modern leadless pacemakers –which are about the size of a grain of rice — are placed directly into heart tissue. Their batteries die after about 7 years, at which point the device is simply silenced and a new one inserted. Because they are covered in scar tissue by that point, there is no way to extricate the old device. This could lead to infections and isn’t feasible in younger patients, who would need multiple such implants over the course of a lifetime. “You’re practically turning the heart into a little graveyard of all these pacemakers,” Karami says. “It’s one of the challenges currently and it’s far more significant than the previous surgery challenge.”

“They’re marvelous engineering but not always practical.“

— Amin Karami

The ideal solution, he adds, would be one that simply swaps out the battery in a pacemaker with a self-powered device that ensures the pacemaker doesn’t need replacement. But most energy harvesters are still intrusive. “They’re marvelous engineering but not always practical.” Karami says. “The real challenge is how to simply incorporate the energy harvester into the battery package of the pacemaker.”

Nonetheless, these researchers all expect energy-harvesting devices — as wearable sensors rather than implants — will be on the market within the next five years or so because that modality faces fewer regulatory hurdles than implants. These self-powered electronics would constantly monitor a wearer’s gait, pulse, blood pressure, or movement. They could benefit athletes, older individuals at risk of falls, or patients with chronic conditions that need constant monitoring.

Mercier says that devices of the future might combine different modalities — thermal, mechanical, or chemical — for the most efficient power generation. Because each kind of human energy varies among bodies and over time, “energy harvesting in humans is going to be stochastic by its very nature,” he says. “There’s no silver bullet.”

Published under the PNAS license.

Inner Workings: Self-powered biomedical devices tap into the body’s movements was originally published in Proceedings of the National Academy of Sciences on Medium, where people are continuing the conversation by highlighting and responding to this story.