A Framework for Tomorrow’s Pathogen Research

Final Report


Ravindra Gupta

Ameenah Gurib-Fakim

Shahid Jameel

David Relman


Jesse Bloom

Filippa Lentzos

February 2024

Ethical obligations to make research with pandemic risk more safe, secure, and responsible

The benefits of pathogen research are significant because it has dramatically improved the well-being of humans, other animals, and the environment—and has the potential to continue to do so in the future. However, the potential harms of research with potential pandemic pathogens are no less significant as these could affect entire populations.

This reality underscores the need for actionable and sustainable strategies and mechanisms to effectively maximize the potential benefits of research with potential pandemic pathogens and eliminate or mitigate the foreseeable potential harms, while attending to issues of equity and proportionality.

Ethical obligation to ensure high-probability benefits for public health.

Research with pandemic risks should have high-probability benefits for public health.

We acknowledge that the public health benefits of research with known and potential pandemic pathogens are difficult to assess as these would typically unfold over long time scales, with much uncertainty and with uneven distribution around the world. Moreover, these benefits depend upon disputed technical details about the research in question. As a result, claimed public health benefits are often vague or underspecified, complicating a comparison to risks.

In some instances, scientists may overstate the potential benefits of research. For example, supporters of the H5N1 influenza A virus enhancement research have claimed benefits for surveillance and vaccine design, notwithstanding that research had previously identified many of the mutations those studies determined to be important for transmission by safer methods without increasing the transmissibility of the actual H5N1 influenza A virus (Lipsitch 2016). Consequently, although such research certainly has increased human knowledge, it is fair to ask if the public health benefits of that knowledge are commensurate with the risk of a pandemic that could be caused by the accidental (or otherwise) release of a mammalian-transmissible H5N1 influenza virus.

Underpinning public health benefits assessments is the idea that the expected benefits of human pathogen research are greater if pathogens (and the specific variants under study) are circulating in humans and domestic animals. Conversely, the expected benefits are smaller if pathogens are solely found in wildlife or are extinct or enhanced variants of current pathogens that are not currently circulating. In turn, this is considered more beneficial than research on pathogens that are not expected to naturally evolve (for example, highly chimeric pathogens or pathogens modified with genes from other organisms). Specific frameworks have been proposed to make these types of assessments (Casagrande and Greene 2022). Taking the example of re-constructing the 1918 pandemic H1N1 influenza A virus, a qualitative assessment might rate the public health benefits of that research as lower than the public health benefits of research on today’s H5N1 avian influenza A strains because the 1918 strain is not currently circulating in humans or domestic animals and is unlikely to re-appear unless the re-constructed virus itself is released. On the other hand, understanding the characteristics that made historical pandemic viruses virulent or transmissible may provide insights that enable us to predict whether pathogens currently circulating in animals are potential human pandemic risks. Where feasible, these questions should be addressed using surrogate systems or by taking advantage of loss-of-function experiments on current human viruses (Johnson 2021; Yen 2011).

There are many types of research designed to mitigate the risk of pandemics, with enhancing the properties of potential pandemic agents being just one of these. Other approaches include studies of virus components using noninfectious or safe constructs (see Box 1) as well as non-virologic approaches such as countermeasure development, which is often not dependent on any knowledge of the specific phenotypes studied in enhanced potential pandemic pathogen research (e.g., transmissibility), or improvements in health care. However, it is important to note that in many cases countermeasures may need to be tested with actual live pathogens.

Moreover, a pandemic would result in a risk to lives even if the probability of an accident is low, and a risk to the lives of others cannot be justified purely by the promise of increased scientific knowledge. This highlights the need to review studies for risk to bystanders (non-participants) resulting from enhanced potential pandemic pathogen research (Eyal et al. 2019).

Once reviewers complete a systematic public health benefit assessment for proposed research with known and potential pathogens, they will still need to make judgements on how to evaluate the risks of a study, including any residual risk.

Ethical obligation to minimize risk of harm by using less-risky alternatives, where appropriate.

Biological risk assessments tend to be limited in scope to biosafety issues and a risk to laboratory personnel and their communities. For research with pandemic risk, from which the potential for harm extends to the entire human population, a more elaborate risk assessment is required.

As WHO sets out, researchers and their institutions have an obligation to use less-risky research when this would be equally beneficial (WHO 2022). When the potential benefits from a study that involves an enhanced potential pandemic pathogen could be achieved by other less-risky means, the lower-risk research design should be the choice. This analysis is consistent with the types of decisions routinely made by individual researchers and funding agencies, when they aim to choose an investment of time and money that optimizes the likely gain from the study rather than abstractly weighing doing versus not doing a study (Lipsitch 2018; Lipsitch and Inglesby 2014). Incentives, both positive and negative, need to be devised and promulgated, mindful of local sensitivities, to encourage the weighing of risk, along with time and money, in deciding on research design.

A significant percentage of the risk from virological research arises from the possibility of accidents or misuse of the pathogens themselves instead of the information gleaned from the research. Consequently, incorporating a virus into laboratory research increases exposure risk, even under the strictest biorisk management operating procedures. Whether the risks of the research are tolerable is a function of the value of the outcomes. The answer will be dependent on several factors, such as the type of virus under study (its infectivity, transmissibility, and virulence as well as the availability of countermeasures), the type of experiment (in silico, in vitro or in vivo), the amount and concentration of the handled virus, the biosafety level of the laboratory where the research is conducted, the level of training and standard operating procedures of laboratory staff, and the implemented oversight (e.g., institutional biosafety committees, animal-care-and-use committees, and dual-use committees). To reduce risk, it is prudent to evaluate whether incorporating potential pandemic viruses into the research is necessary: (1) scientifically (i.e., whether researchers could replace viruses with so-called “surrogate systems” to yield the same types and qualities of answers to questions) and (2) pragmatically (i.e., whether the virology community, including funding agencies, promotion committees, and publishers, will take research using surrogates seriously).

In the past, promising results obtained with surrogate systems (see Box 1) have not always withstood validation with the viruses they are intended to represent. These failures have led to concerns on the part of virologists, scientific journals, and funding agencies that virology research without using actual replication-competent viruses is subpar or incomplete. Researchers have quickly learned that successful publication of their results is often dependent on including replication-competent virus research because reviewers and editors almost inevitably request it, even if it is technically not necessary and does not contribute to the utility of the study. Consequently, studies with replication-competent viruses are often held in high regard, even if the use of a surrogate system would yield equally robust and scientifically useful results. The broader scientific community (in particular, journals, their staff, and reviewers) determines whether a study is deemed a success, which has a direct impact on the professional fate of researchers. The preference for research using replication-competent viruses thus drives a reliance on risky virus research. One path forward to overall less-risky research is for the scientific community to commit to using surrogate systems when feasible.

Examining whether surrogate systems yield equally robust and scientifically useful results has several advantages. It appropriately considers whether the extra knowledge gained from doing riskier research justifies extra risk. It appropriately weighs the fact that riskier research is necessarily more expensive than safer alternatives due to the extra biosafety and biosecurity controls needed. Such an analysis might well suggest that investing funds in safer alternatives could result in more generalizable knowledge and avoid the small sample sizes often used in expensive research on high-risk viruses (Linster et al. 2014; Herfst et al. 2018). At the same time, it is important to recognize that often there will not be a surrogate system that can effectively answer the question at hand.

Considering the best alternative to research with pandemic risks obviates arguments about the risk of “not doing” a set of studies, or the opportunity costs of a road not travelled. While it is true that the benefits of basic biomedical research may be long term and its value not immediately evident, some argue that not doing a study foregoes the unknowable potential benefits of that study. However, the same is true of the alternative study. Factoring in unknowable potential benefits is therefore an argument for doing science in general, rather than a point in favor of specifically doing risky research or foregoing all risky research.

Surrogate systems

Over recent decades, researchers have developed numerous surrogate systems for conducting research in the absence of replicating target viruses. These systems exist to simplify the complex virus-host system and allow focus on different aspects of it. They reduce risks for laboratory workers and publics and avoid the need to perform research in high- or maximum-containment laboratories. They can overcome the need for unavailable resources (e.g., target viruses and access to containment laboratories). In some cases, surrogates enable studying aspects of a virus lifecycle that can be dissected better with these systems than with infectious viruses. On the other hand, some findings obtained with surrogate systems may not always reflect the biological reality of the fully replicating virus. The potential pros and cons of surrogate systems can be seen in several examples:

Pseudotypes (Cui and Huang 2023; Radoshitzky et al. 2018; Steeds et al. 2020)   

Pseudotypes are created by expressing a viral entry protein on the surface of a virion that packages a reporter gene but lacks the full complement of viral genes needed to undergo multicycle growth and are thus unlikely to cause disease. Pseudotypes can be modified to incorporate the surface proteins of the viruses they are intended to represent (e.g., Nipah virus) instead of their own (e.g., HIV-1) surface proteins. Since the surface proteins of many viruses determine which cells and organs they infect and are the main targets of antibodies, pseudotypes can be used to study and develop medical countermeasures for the earliest events in viral infection. Importantly, these studies can be done at lower biosafety levels with minimal risk. However, for various technical reasons, pseudotype systems only work for some target viruses with surface proteins amenable for pseudotyping and, because the “geometry” of pseudotypes is not identical to particles of target viruses, results sometimes need to be confirmed with the actual target viruses. Thus, pseudotype research cannot always completely replace research with the bona fide pathogen but can instead reduce it. It is a suitable method for addressing many questions of biological and public-health relevance.

VLPs/trVLPs/biologically contained particles (Hoenen et al. 2011; Halfmann et al. 2008; Wenigenrath et al. 2010) 

Virus-like particles (VLPs) are replication-incompetent particles produced by the co-expression of certain viral structural proteins. They can be used in a similar manner as pseudotypes and have the advantage of correct particle “geometry.” Virus-like particles can be turned into transcriptionally active virus-like particles (trVLPs), i.e., target virus particles that contain truncated target virus genomes and behave like replicative entities in cells continuously producing virus components in trans (that is, provided by the researcher in various ways rather than from the truncated genomes). In the extreme, transcriptionally active virius-like particles can be turned into “biologically contained viruses.” This is achieved by introducing virus genomes that lack one or a few viral genes into cells that produce the missing genes. Transcriptionally-active-virus-like particles and biologically contained particles “behave” like true viruses but cannot replicate and cause disease in organisms because they lack critical components. Few such systems exist due to often formidable technical challenges and safety concerns of developing them (e.g., possible recombination and thereby creation of fully infectious viruses). Thus, research with these systems may be less risky than research with target viruses but is not considered risk-free.

Minigenomes (Hannemann 2020; Hoenen et al. 2011) 

Minigenomes are target viral genomes typically devoid of most genes. They contain viral genomic regions required for replication and/or transcription. Researchers manipulate cells so they provide the minimal replication/transcription proteins of a virus in trans, thereby resulting in minigenome replication. If a reporter gene is incorporated, the process will result in its transcription and translation, as well. These systems can be used to identify candidate medical countermeasures targeting, for instance, proteins that viruses use to replicate (viral polymerases) or the functions of an infected host cell (host factors) on which a virus depends for that process. However, the challenge of developing minigenomes increases with the complexity of the target virus and is crucially dependent on the knowledge of replication and transcription signals of the viral genome, which often are not known (and not easily determined) for target viruses. In addition, the minimal genome replication complexes represented by minigenomes do not capture all aspects of the viral life cycle.

Recombinant viruses (Gross et al. 2018; Fathi, Dahlke, and Addo 2019)

Another approach to decrease, but not abolish, risk is to create recombinant viruses that express target virus proteins instead of their own proteins. A classic example is vesicular stomatitis Indiana virus (VSIV) manipulated to express the Ebola virus glycoprotein instead of its own glycoprotein. VSIV infects insects, cattle, horses, and pigs and, rarely, leads to mild influenza-like disease in humans, whereas Ebola virus often causes fatal illness in humans. Researchers can use recombinant versions of VSIV expressing the surface proteins of a target virus to study cell entry processes similar to pseudotypes but in a fully replicative background. Recombinant VSIV expressing Ebola virus glycoprotein (“rVSV-ZEBOV-GP”) is currently considered sufficiently safe to be used as a vaccine against Ebola virus disease and is approved for this purpose by the European Union and the US Food and Drug Administration. However, such recombinant viruses are not necessarily attenuated, and their cell and host tropism depend on the incorporated protein. Therefore, these viruses may pose unknown risks. In addition, special regulatory approval may be necessary because their creation could be considered research of concern due to the incorporation of components of a potentially pandemic pathogen into a less dangerous background, leading to the possibility that a previously benign virus becomes a risky one.

Alternatively, researchers could render target viruses less risky by, for instance, by serially passaging the target virus in cell cultures and/or animals to select a weakened (attenuated) strain. However, creating attenuated viruses necessarily starts with replication of the target virus (and hence is “risky” to a degree). Also, any manipulation or selection includes inherent (even if minimal) “gain-of-function” risks (or at least the perception thereof), as it cannot be strictly assumed that any mutations resulting from serial passage will only result in attenuated viruses (although this is usually the case if the passage is done in common cell lines). In addition, there are concerns about the potential for reversion of attenuated viruses to wild-type viruses. The development of attenuated viruses is in some cases possible by rational approaches if reverse genetics systems are available that enable deleting or modifying known viral pathogenicity factors, or by recoding strategies that are known to confer attenuation (e.g., codon-pair deoptimization (Cai et al. 2020)). This would reduce the risk since it omits starting with replication of the authentic target virus.

Gene synthesis and protein expression

If the genomic sequence of a target virus is at least partially known, its genes can still be synthesized individually from known sequence fragments and their encoded proteins can be expressed in tissue culture. This approach enables the study of individual virus components and, to a degree, identification of candidate medical countermeasures that bind them or their cellular or viral interaction partners. However, many viral proteins do not fold or function correctly in the absence of other proteins that often have not been identified. Many viral proteins may not show the same (sub-)cellular localization compared to their localization within infected cells, and this approach often does not identify medical countermeasures that are active enough during target virus infections, during which exponential replication of viral protein components overcomes the fixed concentration of the countermeasure. Moreover, viral proteins may act in complex ways, or require the presence of other viral proteins to reveal their true function. Artifacts of over-expression systems are also a confounder.

In silico analyses (Versini et al. 2024; Gutnik et al. 2023; Ismi, Pulungan, and Afiahayati 2022)

In silico (computer simulation) methods, including those based on AI, have made great strides forward in predicting structures of target virus components (e.g., AlphaFold 2 and RoseTTAFold). However, while these methods often work well for predicting structures of isolated (e.g., secreted) virus proteins, they often yield suboptimal results for proteins that require interacting partners for correct folding. Thus, these methods become less useful with increasing virus complexity. Moreover, these systems may not yet accurately predict the folding of proteins with disordered regions, or domains that do not fit well with the known structures within a database. Furthermore, the function of a protein is often dependent on a complex cellular environment that cannot currently be modelled using in silico methods.

While virus surrogate systems are an important component of virological research, none of them is universally applicable to all viruses or experimental research questions. Importantly, due to caveats and limitations of each surrogate system, the results obtained with these systems often need to be confirmed using actual virus research and some of these systems may themselves pose novel risks or concerns. It should be noted that for many viruses, particularly those that have not been intensively studied before, researchers have not established surrogate systems yet or cannot establish them for functional/biological reasons. Therefore, there are often circumstances in which there is no alternative to studies involving authentic viruses.

Ethical obligation to correct inequities in benefit-sharing and research burdens

The benefits of scientific research often accrue differently across the range of stakeholders. For researchers and their institutions, publication, grant funding, professional advancement, and prestige are powerful incentives and benefits of successful research. For journals and funders, publishing and supporting successful research can represent high impact and strong returns on investments.

Public health benefits or benefits to particular communities, if they do transpire, are often delayed or only become apparent in the longer term. In the present setting, inequities in access and purchasing power have often led to earlier and larger public health benefits from the fruits of research for wealthier countries and for wealthier residents within those countries. This often contrasts with the risks. Pandemics, by definition, have widespread geographic impact and have often caused considerably greater harm to those already at economic and social disadvantage, both within countries and across the globe (Murray et al. 2006). Parties involved in caring for patients and managing the downstream consequences of potential outbreaks resulting from accidental, inadvertent, or intentional releases—public health authorities, clinicians, and other front-line workers—also do not benefit directly from research with known and potential pandemic pathogens and are not usually consulted as part of harm-benefit assessments. There is, therefore, often a mismatch between those who bear the risk (e.g., communities directly or economically affected by a biosafety incident) and those who might benefit from the products of the research.

Researchers and their institutions play an important role in harm–benefit assessments. They likely have the earliest and clearest insight into whether their particular research raises biosafety and biosecurity risks. When it does, researchers and institutions are usually best positioned to propose mitigation strategies or to find alternative lower-risk paths for pursuing their research. However, with research with known and potential pandemic pathogens, for which the stakes are higher and the inequities in the harm–benefit distribution across stakeholders greater, researchers and their institutions should not be the only ones conducting harm–benefit assessments; a broader range of stakeholder groups should be involved in consultation.

Ethical obligation to respect prohibitions on research when there is not a proportionate harm–benefit ratio

The most concerning research is that which could result in (1) uncontained community spread of a novel pathogen or variant of a pathogen among humans, other animals, plants, or the environment and cause harm, or (2) uncontained community spread of a novel pathogen that was already transmissible and capable of epidemic or pandemic spread but has been made more harmful. This could be the result of accidental, inadvertent, or intentional release of the known or potential pandemic pathogen. An independent and transparent review of risks and potential benefits of this kind of research should occur at national/federal levels, and, given that the risks are global, may also warrant review at the international level (Steinbruner et al. 2007). This level of review should not only precede the work but occur at regular intervals as new data are collected and new experiments are proposed.

Research with pandemic risks should proceed only when the research community and relevant oversight bodies can (1) demonstrate that the research would be conducted safely, securely, and responsibly; (2) demonstrate that no alternative and safer research could reach the same public health ends; and (3) provide adequate assurances of substantial benefits expected in the near term with a plausible plan for equitable global distribution of these benefits.

Next section ⇢