Potential Benefits of Gain-of-Function Research
The benefits that have resulted from the billions of dollars invested in biomedical research over the past several decades are seldom disputed. Biomedical research has made enormous contributions to the understanding of disease and the development of cures through the creation of countless innovations for improving and protecting human health, including new animal models and more effective vaccines and drugs. However, as pointed out by Dr. Ronald Atlas, from the University of Louisville and one of the symposium planning committee members, the benefits of basic biomedical research for medical practice and public health may be long term and their value not immediately evident. The results of particular types of research cannot always be predicted, and benefits are often serendipitous. Because it is not possible to predict what breakthroughs may occur as a result of fundamental research, it is impossible to quantify the benefits of fundamental research for risk/benefit analyses. Long-term research benefits are achievable, but it is not possible to specify what these are when the research is initiated.
Research using Gain-of-Function (GoF) techniques is no different with respect to what it can achieve in the long term, at least according to many of the symposium participants. Atlas noted that, although there was no attempt to achieve a consensus, no disagreement was voiced to the repeated claims of various presenters that in the short term GoF research is helpful for adapting viruses to growth in culture and for developing essential animal models for emerging pathogens, such as Middle East Respiratory Syndrome coronavirus (MERS-CoV), and escape mutations
to understand drug resistance and viral evasion of the immune system. In the long-term it may also allow the generation of information that is not obtainable through other methods, but whether all the long-term benefits envisioned for GoF research will actually be realized is still unclear. Vaccine producers in particular disagree on whether GoF methods are essential for vaccine development, so the contributions of GoF research to vaccine development need careful evaluation. Increasing reliance on gene sequences to predict phenotypes may increase GoF research’s importance over time. As was clear from the presentations in Session 4 of the symposium, there is wide recognition that it is not yet possible to predict phenotype from genotype, but Dr. Philip Dormitzer, from Novartis Vaccines and a member of the symposium planning committee, noted that as more genotype-phenotype linkages are established, it may enable keeping certain viral characteristics out of vaccine strains.
Two symposium sessions were devoted to presentations on the potential benefits of GoF research, one focusing on the role of GoF in surveillance, detection, and prediction and the other on its role in treatment and response.
SURVEILLANCE, DETECTION, AND PREDICTION
The first presentation in Session 4 was given by Dr. Stacey Schultz-Cherry, St. Jude Children’s Research Hospital, who discussed the information garnered from GoF studies about what she believes are its public health implications. Her home institution is one of five National Institute of Allergy and Infectious Diseases (NIAID) Centers of Excellence for Influenza Research and Surveillance in the United States and focuses on the animal-human interface. St. Jude is also a World Health Organization (WHO) collaborating center for studies on the ecology of influenza and is part of a global influenza surveillance and response system that includes six WHO collaborating centers and 144 national influenza centers throughout the world. St. Jude collaborates with colleagues in the animal health sector and their main role is to decide on the influenza strains that are incorporated into the seasonal flu vaccines. They also decide whether vaccines or candidate vaccine viruses are needed for emerging zoonotic threats.
The national influenza centers conduct viral strain surveillance throughout the year, looking at the genetic information from human as well as emerging zoonotic viruses. Every February and September, representatives from the WHO centers and central regulatory laboratories as well as animal health experts go through the surveillance data to decide on which viruses to choose as vaccine strains. This information is given to the vaccine manufacturers and regulatory agencies, and 6-9 months later
the vaccines become available. She described many of the complexities of the process. She noted, in particular, that determining the function of amino acid changes in the viruses circulating in the field is one of the key tasks. As an example, she discussed an ongoing outbreak of H5 viruses in Cambodia. Through GoF research, it has been determined that the presence of certain genetic markers in the outbreak strain suggested that this particular virus could be more readily transmitted, at least in ferrets. This information has provided the persuasive factor to move forward with the development of a vaccine.
Schultz-Cherry noted that GoF research-derived information is also used for risk assessment. The U.S. Centers for Disease Control and Prevention has developed a risk assessment tool, the Influenza Risk Assessment Tool, to rank the risk associated with particular viruses. She stated that the result of using the Tool is not a prediction of the next pandemic, but rather an objective means of prioritizing viruses for future risk management. The Tool looks at the properties of a virus. What kind of receptors does it bind to? Is it more mammalian or avian? Does it transmit in animal models, or does it have molecular signatures that would suggest transmissibility? What is its genomic variation? She stated that all of this information, especially the molecular determinants of transmissibility, has been generated through GoF studies at some point, perhaps even as far back as the 1970s. She stated that the ability to prioritize is important because of limited resources; vaccines cannot be made for every new emerging virus.
Schultz-Cherry’s final points dealt with the limitations of these studies. Phenotype still cannot be predicted from genotype. We may know a lot from studies of particular amino acid changes in one strain of virus that may not apply to another strain. She noted that opponents of GoF research have said that this is a reason to not continue this work. She would argue, however, that inability to predict phenotype is precisely why GoF studies must continue so that eventually this inability can be overcome.
During the discussion following the presentations, Schultz-Cherry was asked what is the trajectory of the information being used for vaccine candidate selection? She explained that the risk assessment tool is continually updated to add new information about molecular determinants of virulence and transmissibility. She believes that the more information we have, the better we will be able to predict the risk of a pandemic and then use that prediction to prioritize vaccine strain selections and make the vaccines available.
Dr. Christophe Fraser of Imperial College, London next spoke about potential pandemics. He began by stating that he would scrutinize the benefits of GoF experiments using a narrow definition of GoF as dealing
with the transmissibility of the highest risk potential pandemic pathogens (PPPs). He is the Deputy Director of the Center for Outbreak Analysis and Modeling, which is also a WHO Collaborating Centre for Infectious Disease Modelling, located in London. He and his colleagues at the Centre have worked on the Severe Acute Respiratory Syndrome (SARS) outbreak, the initial response to the 2009 influenza pandemic, and have synthesized a variety of surveillance, neurological, and epidemiological information. In 2014, their work turned to both MERS, for which they were trying to quantify its transmissibility to humans, and Ebola as part of the WHO response team. He noted that, on a global scale, the interventions in the event of an outbreak are quite simple—well-organized classical public health tools. The key aspect is timeliness, and the classical tools are diagnostics, social distancing, and risk communication. Probably the area most lacking at the moment on a global scale is rapid diagnostics to allowed triaging of people, which has been made very clear with Ebola. Data systems, multidisciplinary validation, and sharing of data and samples are all required. There is also a huge role for basic science, but once an epidemic has started, the value of information from this limited realm of GoF work on transmissibility is unclear. The role of such work is clearly going to be in predicting pandemics. He stated, however, that H5N1, H7N9, MERS, and Ebola had all clearly been identified as threats prior to any GoF-PPP experiments, although this is less the case for the 2009 H1N1 outbreak and SARS. Nevertheless, the failure to predict outbreaks of the first four pathogens he listed was due to surveillance gaps, not a lack of understanding. Of the viruses that emerged in 2009, there were no closely related viruses found by surveillance in any swine populations for 12 years prior to the emergence of H1N1. MERS also emerged from a complete surveillance gap.
The next utility that has been claimed for GoF research-derived data is for predicting emergence. The data from the two experiments on H5N1 transmissibility were plugged into a model by Colin Russell and Derek Smith (Russell et al., 2012), who concluded that it is not possible to calculate the level of pandemic risk precisely because of uncertainties in some aspects of the biology. Fraser stated that he very much endorses that statement; it is not possible to calculate the level of risk from the mutational landscape. The aim in Russell et al. (2012) was to conduct basic science to understand the factors that increase or decrease risk, not to assess the actual risk. Russell’s work built on earlier work that attempted to predict pandemics. The earlier work from Jamie Lloyd Smith tried to establish a general rule, which is that infection begets transmission and transmission begets epidemics. Things that can cause transmission are much more likely to result in epidemics than things that are not already transmissible.
The WHO uses an empirical, rather than a theoretical, approach,
meaning that alarm bells should be based on human cases and clusters and the key is surveillance and sharing of data. However, as Fraser had previously noted, there are limitations, especially given that for many years there was reluctance to acknowledge clusters of infections because of the fear of escalating the WHO alert levels and the resulting consequences. In terms of surveillance and response, it is of course very useful to know what viruses are out there, but it is promptness that is critical. To contain an epidemic at its source, there is a window of days in which to intervene. Once the epidemic gets going, the scale of the problem will double every week. The most suitable response would be based on the timely reporting of cases.
Fraser believes that pre-pandemic vaccine strain selection is the crux of the argument. Timely development of vaccines could be transformative. Vaccine seed stocks can speed this up, but there are other rate-limiting steps, especially international agreements on the regulation and conduct of human trials. He also believes that the objectives should be to:
- prioritize strains with evidence of infection and transmission;
- cover antigen space, and monitor antigenic drift;
- plug gaps in surveillance;
- make more/faster seed stocks (Dormitzer et al., 2013)?
Fraser concluded with the following:
- The direct benefits for enhanced surveillance and model-based prediction of GoF experiments with PPP should not be overstated.
- The indirect benefits of basic science are likely huge, but the rationale for working with dangerous pathogens requires benefits that outweigh risks and opportunity costs.
- The benefits of GoF with PPP for pre-pandemic vaccine production should be probed in depth.
- The risks are real and present (Lipsitch and Inglesby, 2014).
A participant asked Fraser about what he would require to be confident about using data from GoF or other experiments in his modeling? He responded that the tools required for this lengthy, although worthwhile, journey must be available. The issue centers on the risk taken at the beginning of the journey. Earlier in the morning, Fineberg mentioned that, by their nature, pandemics provide many years to think about the tools but only infrequent and limited time to acutally test them. Weather forecasting has improved dramatically because weather forecasters can test their models daily and receive many complaints when they are wrong. The situation with pandemics is not like that.
Dr. Colin Russell of Cambridge University Infectious Diseases responded to the two previous presentations as the last speaker of Session 4. He noted that both of the previous speakers touched on the ability to predict risks for pandemic viruses and on the ability to produce vaccines in a timely manner, and to ensure that there are enough vaccines to go around and provide a chance to mitigate the early spread of disease. However, the more we learn about nature, the more we understand that there are a vast number of undescribed viruses out there, many known only through sequence data. He stated that genotype to phenotype prediction is one of the holy grails of influenza biology research. However, much more research is required to reach this goal. He referred to a National Institutes of Health workshop for which he was lead organizer in the fall of 2013 that brought together experts in virology, epidemiology, and other fields. It included participants from both sides of the GoF debate, and a key focus of the meeting was to rectify the limitations in the ability to make inferences about the phenotype of influenza viruses from genetic sequence data alone. A full report of this workshop was published in October (Russell et al., 2014). A key question in the discussions was whether the effects of mutations are dependent on the viruses in which they occur. A variety of studies suggest that the effects of particular mutations are strongly likely to depend on the genetic context in which they appear. First, in 2006 Jane Stevens, Ian Wilson, and others published a paper in the journal Science (Stevens et al., 2006) about GoF research, investigating the potential for a virus to switch receptor binding from avian-like to human-like. This work was among the first to demonstrate that single amino acid substitutions could cause such a switch. But the authors concluded that knowledge of genetic changes in circulating virus isolates by themselves obviously cannot be used to predict the impact of receptor binding specificity, let alone affect the results of future mutations (Stevens et al., 2006). It is worth bearing in mind, Russell stated, that there is a great degree of genetic diversity in the H5 virus. Other studies have found that the effects of mutations in other H5 viruses depend on the clade of H5 viruses in which the substitutions were produced. These residues alone cannot be used as reference points with respect to specificity in H5N1 strains, but when combined with other data, the presence or absence of these mutations can be informative. None of this should be in any way construed to undermine the value of the studies, but highlight the impressive need for further work. In short, Russell believes that, given the incomplete state of knowledge, there is a risk of overestimating what is known based on sequence data alone. Focusing too much attention on the presence or absence of particular mutations may cause other mutations or even other traits yet to be identified to be overlooked.
Gavin Huntley-Fenner asked the panel members what sort of public
health system would be needed to justify the status quo and whether the risks and benefits of GoF research are balanced from this public health perspective. Fraser answered that transmissible viruses makes GoF research a very special case. In terms of general basic science, we never have to justify that to the same degree, luckily, because otherwise we would find it difficult to move forward. Basic science is a much broader portfolio where the risks are very small. The real crux of the GoF issue is separating out that very small number of experiments. We need a much wider frame for all experiments, where occupational health risks are not an order of magnitude higher than public health risks.
Laurie Garrett of the Council on Foreign Relations commented to Schultz-Cherry that her statement that the risk assessment model would be adjusted differently if H5 was in Canada speaks to the core of the whole problem. Risk is about rich people, which is about 5 percent of the global population, if that. She stated that we have never once delivered vaccine to poor people around the world for any epidemic/pandemic situation in the history of the planet, have never delivered clinical tools, and have never delivered diagnostic tools. Garrett had just come out of quarantine for Ebola, and there is nothing that can possibly be called a rapid diagnostic available for Ebola. So when the Council on Foreign Relations reviewed the whole question of GoF use and issued its memorandum to the White House (available at www.CFR.org), it concluded that the most fundamental problem is that the International Health Regulations have never been fully implemented. Garrett stated that none of the wealthy nations has assisted poor nations to raise them to capacity and that “none of the benefits will ever be available to the majority of planet Earth and none of them are getting the toolkit to minimize or mitigate risk. We are having a very American conversation that excludes the rest of the planet.”
Schultz-Cherry responded that her remark about having H5 in Canada was designed to make people think about risk versus benefit and to reflect that doing more work can democratize the surveillance process. With more work, it could become cheap and easy to assess the threat of viruses. If this could be done, we could radically change the way we do surveillance worldwide and we would not have the same sort of geographic distributional issues that are of concern now.
Dr. Gregory Koblentz, George Mason University, asked Schultz-Cherry about the proven accuracy of the risk assessment tool used for selecting flu strains for yearly vaccines. She, in turn, called on Dr. Ruben Donis of the CDC to comment more about the risk assessment tool. Donis noted that the risk assessment tool is a product of the global community of scientists working on both human and animal health. It is a product of the realization of the gaps in surveillance that were noted in Fraser’s presentation. It was developed to ensure that we have a comprehensive
way of evaluating all the possible viruses that are circulating in animals that could reassort, recombine, and change the phenotype and eventually emerge as pandemic viruses. The tool attempts to develop a comprehensive review of all of the potential threats.
Via the web, Dr. Daniel Perez, University of Maryland, asked whether the potential of strains that resulted in past pandemics to affect humans would have been moderated if we had had the opportunity to sequence them. Fraser stated that understanding how a virus expands its host range from swine to humans requires a lot of information. The validation of the genotype to phenotype prediction tools really should address that question. Russell added that he did not think that having sequence information at the time of earlier pandemics would have forewarned of the emergence of those viruses, which again speaks to the incomplete nature of knowledge and the critical need for further work.
Another participant pointed out that there is probably a very large number of variables involved in understanding viral pathogenicity. Given the number of variables, is there much chance of doing anything useful? Russell and Fraser both agreed that this is a very complicated problem, which is why more experimental work is needed to help reduce the dimensionality. But what we currently know cannot help us very much in understanding what will occur in the next 5 years. However, science is an incremental process. The increases in understanding that have been achieved from the work that has been done so far have been helpful. In terms of translating directly into public health improvements, that is a pretty substantial leap to make. But saying we will not get there will not undermine science. Nevertheless, tools that can deal with perhaps thousands of genetic traits and phenotypes are needed. It is not about the mutations but rather about the function of the mutations. We could reach the state where we sufficiently understand the traits that a virus needs to adapt to humans and identify ways to test for those that are either independent of sequence or a metalevel of sequences.
Another participant made the point that had the 2009 pandemic strain been seen in animals instead of humans, it might have been falsely viewed as having low virulence and transmissibility and would have been discounted. Fraser agreed that the fact that our knowledge is incomplete right now creates a risk of discounting viruses that lack a certain number of substitutions when in fact we should be concerned about the risk.
Dr. Ron Fouchier, Erasmus MC, commented that he believes a lot is being asked of papers that were only published in 2012 and for which the follow-up work has been shut down twice for extended periods. This is work in which the phenotypes, not just the genotypes, are being studied. He agreed with Fraser that although he cannot yet predict phenotypes from genotypes, the assays produced by his work are being used to look
at phenotypes in surveillance, which means a better job is already being done. He made a plea for more basic science to follow up on his work, which is still in the early stage. Fraser responded that the basic science is not under question. The question is: Should we be starting with experiments that have orders of magnitude higher risk than other work in the area?
TREATMENT AND RESPONSE
Session 5, moderated by Baruch Fischhoff, consisted of a panel discussion with four speakers. Each panelist was given about 5 minutes and then the session was opened up for discussion.
The first speaker was Philip Dormitzer, who described how GoF research and the regulation around research affect the real-world case of trying to apply virology to a public health situation. For the purpose of his talk, Dormitzer described the chronology for the production and delivery of the 2009 H1N1 influenza pandemic vaccine, an “historical reminder,” for which the response was the “fastest ever, but still came after the disease peaked” (Borse et al., 2013). In fact, an estimate published in Emerging Infectious Diseases (EID) showed that for every week of acceleration of vaccine supply, an additional 300,000 to 430,000 U.S. cases could have been prevented. Dormitzer explained that Novartis, in collaboration with the J. Craig Venter Institute (JCVI) and Synthetic Genomics Vaccines (SGVI), are now working together to establish a process for rapid generation of synthetic influenza viruses that includes GoF studies based on sequence motif data to guide the genetic assembly of the vaccine. For instance, the Novartis research team routinely screens for phenotypic traits of interest and can specifically remove or mutate strains with either polybasic cleavage sites in the hemagglutinins (HA) (found in highly pathogenic avian influenza viruses [HPAIV]) or neuraminidase (NA) gene markers of resistance. For that specific example, Dormitzer explained that the process from the identification of the relevant HA and NA sequences for the new influenza strain to the genetic identity confirmation of the vaccine virus lasted about 1 week. However, the next phase leading to the first large-scale clinical trial took months because of various well-intentioned regulations and policies to protect the food supply in the United States. Notably, because Novartis could not obtain a U.S. Department of Agriculture permit, this phase involved international research collaboration with Germany before taking the vaccine back to the United States, which unintentionally slowed down the human vaccine development. Under U.S. government regulations on select agents, vaccine development against HPAIV is counter-productive because “you can’t really put an entire manufacturing facility under select agent conditions and
still have a factory that can produce seasonal vaccines in an economically competitive way” and in a timely manner. Also, as Dormitzer pointed out, he “couldn’t apply any of this [GoF research] technology.” Therefore, if adaptation of vaccine virus to increase yield or more modern synthetic biology were captured by GoF regulations, then additional unintended impediments to timely vaccine supply could be created.
Next, Ralph Baric presented his view on the impact of GoF restrictions to the emerging coronavirus vaccine and therapeutic research. Baric started his talk by reiterating that no vaccine has been approved for MERS-CoV or SARS-CoV in the midst of an ongoing MERS-CoV outbreak. Baric explained how new restrictions reduce public health preparedness to respond to future SARS-like CoV outbreaks. He explained that the original vaccine target for the SARS-CoV outbreak 2002-2004 strain was 99 percent identical between human and civet (Ge et al., 2013). However, metagenomic sequencing showed that bat SARS-like CoV (SL-CoV) with 65 percent to 95 percent sequence homology, can constitute a large pool of strains with pandemic potential against which countermeasures need to be developed. To evaluate whether the existing vaccine and drugs work on these strains, Baric’s team and others used two types of approaches. The first was based on the production of CoV pseudotypes coated with virus spike-like proteins that can potentially engage the human angiotensin converting enzyme II (ACE2), which is the SARS-CoV cellular receptor molecule. This method constitutes a safe and ethical research alternative approach. Similarly, chimeric recombinant viruses that encode spike-like proteins as part of the virus particle can also be used. While studies using pseudotypes and structure-based prediction confirmed the existence of a bat SL-CoV that can infect human cells, only studies using GoF chimeric virus identified an additional bat SL-CoV as a potential threat. Baric noted that both bat SL-CoV were less virulent in a mouse model. Importantly for public health implications, data further showed that existing vaccine and human monoclonal antibody therapy failed to protect against these two newly identified bat SL-CoVs, leading Baric to point out that “we are vulnerable” to SL-CoV bat strains that currently exist in nature. The second part of Baric’s talk described how robust animal models are essential for vaccine/drug design, safety testing, and performance outcomes. He explained that SARS-CoV replicates poorly in mice (Frieman et al., 2012) and although his team and Subbarao’s lab have developed mouse-adapted strains, the in vivo correlates of infection vary widely depending on the model used. For example, he described some collaborative work done on inbred and outbred mice demonstrating that in some cases the vaccine could have caused increased mortality in some individuals and emphasized the need for better animal models for SARS-CoV vaccine research. In the case of MERS-CoV, the epidemic is ongoing and no robust
animal model exists because routine GOF studies, including passage in small animal models, have failed. Baric called for an immediate lifting of the restrictions on MERS-CoV research on animal model development. This was echoed by other participants during the final discussion. For example, Peter Hale of the Foundation for Vaccine Research stated that he thought the inclusion of the coronaviruses in the “pause” was “muddying the waters” and that he did not detect any enthusiasm among SARS and MERS investigators to increase their transmissibility. This point was also made strongly during the discussion following the session.
The next speaker was Dr. Jerry Weir from the Food and Drug Administration’s Center for Biologics Evaluation and Research, whose team participates in the selection of strains for the yearly influenza vaccines and regulates viral vaccines to ensure that they are safe and efficacious for human use. Weir offered some comments about how the regulatory process views some of the experiments and techniques addressed by the symposium speakers. He stated that there are actually not very many, if any, regulatory issues associated with the type of virus manipulations that were under discussion (i.e., improved types of seed development, reverse genetics, manipulation of virus genomes to improve vaccine virus stability or performance). Manufacturers already licensed can submit a supplement to the license that is evaluated for using a fairly standard process. In lieu of giving examples of how GoF research can influence a process, Weir mentioned a few challenges that still remain in vaccine development for the influenza virus. In general, for the seasonal strain selection and the preparation of pandemic vaccine strains, the major challenge is the existence of very large gaps in our knowledge of how genotype sequences relate to phenotypic changes. Weir stated that strain prediction and selection remain a “guessing game … for which improvements are desperately needed.” In addition, for other factors such as transmissibility or virulence, a lot is not known and improvements are also needed there. To complicate the matter, the incorporation of four, instead of three influenza strains in the seasonal vaccine is a challenge every year for the different players in the global community that pick the vaccine strains as well as the manufacturers who need to deliver the vaccines in a timely manner. For them the yields of vaccine viruses need to be improved with the challenge of limiting factors such as poorly growing strains among the four chosen. In his view, Weir believes that, as broadly defined, “GoF studies have had an enormous influence on how we develop vaccines over the years … and can help improve the process with the challenges that we still face.”
The final speaker was Mark Denison, who explained his view of GoF studies in MERS-CoV and SARS-CoV countermeasure development and how oversight or regulation might be limiting. Denison reminded the
audience about the basic research and ongoing challenges that remain in the development of therapeutics to SARS and MERS-CoVs, emphasizing, like other speakers, the need for in vivo and in vitro models to identify common mechanisms and determinants of resistance. He then moved to a case study involving GoF research and asked the audience whether they would consider giving or taking “a live vaccine with a virus that has an engineered increased mutation rate,” for which only a few people raised their hands. The question was an introduction to a series of studies showing that CoVs, contrary to other viruses, express a proofreading exonuclease (ExoN) normally only found in bacteria and eukaryotes. When this ExoN was inactivated, the CoV mutation rate was increased by 20-fold. Normally, mutations allow tremendous variation in viral populations and presumably increase adaptation, fitness, virulence, and therefore public health risks. However, GoF studies demonstrated that SARS-CoV with the inactivated ExoN were less fit, attenuated in a mouse model of lethal SARS-CoV, could not compete with the wild-type virus, and could therefore be used as a target for therapeutics development. This work was also adapted to other RNA viruses with encouraging results. Denison used this case study to reflect on the implications of new regulations and guidelines if he wanted to create a mutated strain of a virus and test it in an animal model. In conclusion, Denison stated that he believes that because assumptions are usually wrong, GoF research that includes “passage for adaptation and resistance in in vitro and animal models are essential components of therapeutics development” and that to his knowledge no bioinformatics or predictive safer alternative approaches are effective to develop new countermeasures.
Following the panel member’s presentations, there was discussion with the audience. Fraser asked Dormitzer how he would propose to reconcile, practically, the need to conduct very dangerous research without casting the net too wide. Dormitzer responded that what is first needed is a very clear and limited definition of the sorts of research that require particular attention. As Relman discussed, experiments that combine increased transmissibility, virulence, and and lack available countermeasures are very concerning. But we have to make sure that the definitions are not too broad so that they do not capture a lot of other work. Second, there needs to be a distinction between the highly diverse work performed for basic research and the much more restricted, but more urgent, work needed for vaccine development. A classic example is H5N1 vaccine development. There have been at least 26 H5N1 strains that have been attenuated all in the same way. But for the 27th one, the often months-long routine must be goone through again. We need clearly established, well-defined pathways to get vaccines quickly and not encumber the process with regulations.
Relman also stated that he does not think that there is a major question about the value of MERS and SARS research, even that research that currently falls under the rubric of GoF. Restrictions do, in fact, hamper the quest to develop countermeasures, etc. What he thinks is a more interesting question is whether there is a very discreet and specific set of experiments with MERS and SARS that you might not want to see undertaken. For example, would it be appropriate to deliberately start with a highly virulent human isolate of MERS and then attempt to add to that much enhanced human-to-human transmissibility by the respiratory route? Baric responded that he did not know of anyone doing transmissibility studies with the human coronaviruses. Unlike flu, there are currently no small animal models suitable for MERS or SARS transmissibility assays. This is mostly due to receptor incompatibility between the human and any small animal models. Optimization assays to enhance virus transmissibility between ferrets, for instance, would probably decrease the ability of that modified virus to bind to the human ACE2 receptor. Relman reformulated the question to include the possibility of using transgenic ferrets with the human receptor, but Baric explained that the human receptor itself is not sufficient and that other proteins are essential for viral transmissibility and, therefore, the results in transgenic models would not be predictable.
Denison added that nobody would have as a goal or would support trying to increase virulence and transmissibility of MERS or SARS. That is why he recommends the use of a case-based approach that looks at how we really do science. Denison shared his approach when sending a proposal through study sessions or review process at a funding institution. For him, instead of trying to define “boundaries of absolute,” the real question should always be, “What is the best approach to answer that question?” Then, depending on the stage of the review process, the response should be iterative to be adequately addressed.
Inglesby asked Dormitzer whether the annual process of production of flu vaccine relies on research using highly transmissible and highly virulent strains. Dormitzer responded that this is not the case and that the goal is quite opposite—to take a strain found in nature and transform it into something that can be manufactured efficiently by increasing its growth rate in cell culture or eggs. Inglesby then asked whether virulence and transmissibility are traits that can be distinguished from increased growth capability. Dormitzer stated that there is precedent that shows that adapting viruses to grow better in cell culture does not, in general, increase their virulence or transmissibility, whereas passaging from animal to animal often does. He stated that we also need to distinguish between two things: the need for very rapid production of new antigenic variants, which should start on the day it is found that there is a new
variant causing disease, and the development of the vaccine backbone, which could be used in multiple variants and which you do not want to take forever. It is not the same issue when facing an emergency.
Dr. Simon Wain-Hobson, Institut Pasteur, echoed Denison’s presentation by citing work done on polio by John Holland 15 years ago that showed that when chemical mutagenesis is combined with a rapidly evolving RNA virus such as polio, the fitness of the virus goes down. Several members of the panel agreed. However, Denison raised the issue of perception in the current environment and under the current policy circumstances. Such proposals might not necessarily be vetted even though most of the time we can not know the answers until the experiments are conducted.
Another questioner from the webcast asked Dormitzer whether, in his opinion, GoF research is essential for future development of intervention strategies against various pathogens. Dormitzer responded that he thinks it depends on whether you are talking about the short- or long-term. He stated that GoF research is not going to help pick next year’s flu vaccine, but if one is making viruses for use in manufacturing and a certain genetic motif that correlated with high transmissibility is known, then one could make sure that the motif is not included in the vaccine strain. GoF research has utility for such purposes. The other thing is that vaccine manufacturers are increasingly figuring out how to take genetic data and use it to predict what they want to make. That would be a genuine utility if it could be done. The question is whether we can do that kind of science in a way that does not create more problems than it solves. There is potential for GoF research to improve vaccine production, but it is not today except for limited instances; it has long-term potential for this purpose as long as the work can be done without inordinate risk. Denison asked whether any “bad” GoF experiments were performed to discover the polybasic cleavage site associated with high virulence. Dormitzer listed what is believed to have led to this discovery, including studies on correlations between the presence of these sites and clinical observation of virulence in birds; discovery of plausible mechanisms looking at cleavage proteins expressed in different cells; and loss-of-function and GoF studies to make sure that the gene identified is the correct one. Denison’s point was that a series of experiment led to that conclusion.
It was clear, however, that there is a substantial disagreement over the value of GoF research for vaccine development. Lamb, in a later session (Session 8), stated that he thought we should modify the mantra that GoF research is useful for vaccine and antiviral drug development. He thinks that this point is overused and oversold. Hale also commented during the final discussion that he agreed with Lamb, we do need to modify the mantra that this research will help develop vaccines and antivirals.
He said that he and his Foundation fully endorse that sentiment. It is an argument that is made over and over again without evidence to substantiate it. He believes that in terms of development of better vaccines, GoF research has little or no benefit, and if there is any benefit, then it is tiny and way down the road. In the meantime, he said, it is not worth the risk and there are other priorities.
Dormitzer responded to the latter comment and acknowledged that the community of people who make vaccines is divided just as much of the symposium audience was divided. He stated that the basis for that division is informative. Flu vaccines today are still made by very, very old techniques. One looks at what is spreading, sees if it has changed, and then picks the strain. There is not a lot of basic science in this; rather it is 1960s science. In 2009 we were not able to get the H1N1 vaccine out until after the outbreak had peaked, and many people have commented that the current flu vaccines, although somewhat effective, are not good enough. A lot of people who work on vaccines think we need to do things better. One way to do things better is to take advantage of the available information, particularly sequence-based information, so we can do things faster and make vaccines better. Information from GoF research can contribute to identifying risks earlier so countermeasures can be taken earlier. Dormitzer said he does not think it is the case that GoF research is essential to the current vaccine system as it is generally practiced today, but it is not useless. It is clearly part of the trend to understand and predict what can be done better and to help respond quickly. That does not mean it is open season to do what you want and forget the risks. A balance is needed. But he was firm in his statement that the vaccine producers are not universally of the opinion that there is no use for GoF research.
Koblentz asked whether the coronavirus researchers had a sense and could comment on why MERS and SARS were included in the “pause” on GoF along with influenza. Denison believes that, despite the circumstances, the inclusion of SARS- and MERS-CoV in the “pause” demonstrates that this is not about one virus but more about the issue of how we address critical questions in science and what constitutes appropriate review and safety among the different research institutions. He believes that whatever the question asked, whether about replication or virulence and transmissibility, the science should be the same and should follow an iterative process that incorporates risks, milestones, and points to change along the way.
As a follow-up from Relman’s question on transmissibility in MERS and SARS animal models, Koblentz asked Baric to clarify which set of experiments he would use to study transmissibility. Baric explained that many variables are needed to make a model to enhance transmissibility, but if he had the perfect model to do these experiments he would not do
them. Later during the discussion, prompted by Inglesby, Baric added that because the CoV interaction barriers are species specific, the only real absolute model that could be used would be human, so he certainly would not do the experiment.
Fraser asked Denison to clarify what he meant when he said that no one would want to increase the pathogenesis or transmissibility of MERS and, therefore, that the regulation should not apply to MERS and SARS research, especially because this is what the debate is about. Denison explained that he thinks that increasing transmissibility of human coronaviruses is not a goal. He then described the importance of research on wild-type or genetically modified animal models or cell cultures to understand determinants of pathogenesis or virulence factors. No one has the goal to increase these characteristics, but researchers need to be able to study the virus or they would need to rely on epidemiology and surveillance, which are not adequate to answer the question. Denison also stated that to his knowledge there is no other approach to develop countermeasures and vaccines.
Richard Roberts, New England Biolabs, asked whether experiments on dangerous traits that exist in highly pathogenic and virulent strains could also be done on strains that have already been incapacitated in some way. Denison agreed that on a case-by-case basis, if it is possible, then a safer approach is always preferred, but that it depends on the genetic background of the strain of interest. As an example, Denison explained that sometimes a certain type of loss- or gain-of-function experiment is undertaken on BSL-2 strains that are 90 percent identical to more virulent strains, but that the small genetic background differences and therefore structure can greatly influence the outcome of the experiment. When one has strains that are not genetically identical or from the same clade, it may not be possible to make the right determination without doing the experiments.
A participant from the Department of State noted that although there may be ways to do the research in a safer manner, Denison had just argued that in a competitive environment the research question should be answered in the best and most direct way to get funding. The participant wondered whether, in this competitive context, a researcher would prefer the safest, but perhaps more indirect, option assuming it would get at the question. Denison commented that sometimes there are safer options such as when he used a mouse model for hepatitis virus to identify determinant proteins such as those for proofreading. This approach is of public health importance because it proves that certain mechanisms might be a useful target across multiple strains, including those we have not yet tested, such as the basic cleavage site. Returning to his earlier comment about the funding, Denison explained that professors not only try to
educate students to do the best science in the best way, but also ask them about finding alternatives that will eventually answer the question in a less direct way. Dormitzer added that although one may get NIH funding through a grant that incorporates safety considerations, institutional safety boards and questionnaires about dual use research of concern are procedures already in place to make sure it is not only the most direct way to a scientific answer taken, but also that safety is considered.
Each of the panel members was then given an opportunity for closing remarks. Dormitzer’s closing remarks were that he believes that there is long-term potential in GoF research. He believes that we must be very careful with any sort of restrictions or regulations to make sure we do not inadvertently capture a lot of work that is not only good for basic science, but also a core part of the public health response. He stated that as a practitioner of vaccine development, he has realized that there really are road blocks that were never intended by the people who drafted the restrictions.
Baric agreed with those comments and affirmed the importance of reverse genetics and GoF research in understanding viral pathogenesis as well as vaccine and therapeutic design. NIH should be very careful about delineating the boundaries of the restrictions to be placed on the research community because there could be dire consequences if these restrictions are too broad. Weir affirmed one of his earlier points: if we had great vaccines for all of these agents, we might be having a different discussion, but the fact is that we do not.
Denison closed by proposing an iterative process whereby scientists do a review along the way. For critical pathogens of high human consequence there should be a mechanism that allows for a case-based, iterative approach that identifies problems along the way. Investigators need to have their research supported and be allowed to integrate best practices when doing GoF research.