Related Developments That May Impact the Ability to Effect an Attack Using a Synthetic Biology–Enabled Weapon
Synthetic biology is a sophisticated, programmable platform that could in theory enable the development of a wide range of biological and chemical weapons. However, for a capability to warrant concern in the context of this study, it must not only be possible to create an agent in the laboratory but also to use the agent to effect an attack. For many of the potential malicious applications of synthetic biology that were considered, the level of concern raised by technological capabilities is tempered by constraints related to the need to produce the agent in volumes needed to achieve the desired scope of casualty, keep it stable until use, and deliver it to the population in a manner that yields the desired harm. Despite the impressive capabilities afforded by synthetic biology and other modern biotechnologies, these requirements, many of which are the same barriers to weaponization that have constrained the development of bioweapons in the past, are in many cases an important limiting factor in the context of synthetic biology–enabled weapons.
However, these challenges may well be overcome in the future, either by advances in synthetic biology or by developments in other fields. This chapter explores some developments that may become more important in this respect in the coming years. While a comprehensive analysis of technologies being pursued outside of synthetic biology was not conducted as part of this study, these examples are offered to highlight a few areas that will be important to monitor, because they could converge with synthetic biology advancements and ultimately reduce or eliminate barriers to the use of synthetic biology–enabled weapons.
Within the factor usability as a weapon, the report’s framework for assessing the potential for the weaponization of agents produced using synthetic biology identifies questions around production, fidelity, stability, delivery, testing, and targeting. Aspects of these attributes as they relate to specific potential applications of synthetic biology are discussed in Chapters 4–6; broader challenges and considerations related to them are described briefly in the following sections. In general, the challenges posed by each attribute largely depend on the potential nature and scope of an intended attack, which could range, for example, from a targeted assassination of one individual to mass casualty across a population. Although a variety of potential circumstances were considered in the assessments presented in this report, it was generally assumed that an actor would seek to develop the bioweapon covertly and minimize the likelihood of attribution once the agent is deployed. However, the possibility of assigning attribu-
tion for a biological attack is not necessarily a deterrent for terror groups, who may choose to affirm their own responsibility or power and who may not fear discovery and subsequent retribution.
Challenges associated with agent production largely depend on the quantity desired. Large-scale production of a bioweapon is extremely challenging because many agents lose infectivity or other features during scale-up. Although synthetic biology technologies may enable improved cell culture methods, innovations in fermentation, and improved ways to mass produce particular chemical and biological components, the large-scale production of bioweapons is still likely to require significant financial and intellectual resources. On the other hand, mass production may not be needed to perpetrate smaller, more narrowly focused attacks or attacks that can be spread by a replicating pathogen.
Fidelity and Testing
Although it is possible to design and build biological constructs or systems without testing, significant synthetic biology achievements are typically rooted in repeated Design-Build-Test cycles, with testing being a crucial step in the process. Testing in computer simulations, cell cultures, or animal models is a labor- and time-intensive process, and learning from the testing process to make design improvements for the next Design-Build-Test iteration can require a great deal of expertise and experience. Success in computer simulations, cell cultures, and animal models does not necessarily guarantee success in humans, because of differences in evolutionary pressures. Fidelity is also not guaranteed, and it can take repeated process improvements to develop a system that will reliably produce the same results every time, especially at scale. Some synthetic biology approaches, such as directed evolution, integrate testing together with other steps in the process, potentially offering a more streamlined option to circumvent resource-intensive testing steps. It is also conceivable that malicious actors would forego some of the rigorous testing that other researchers would perform, since the standard of success—creating an agent capable of doing “enough” harm—is markedly different from the standards involved in publishing results in a scientific journal. Malicious actors may also be able and willing to test in human subjects, unhindered by the moral considerations and ethical frameworks that guide other research efforts. Despite these caveats, however, developing a synthetic biology–enabled bioweapon would likely still require significant testing to achieve a product that is reliable and effective enough for the actor’s purposes.
A critical consideration in the development of a bioweapon is the capability to deliver it to the intended target population. At smaller scales, delivering a bioweapon can be as simple as contaminating food or water, sticking victims with a needle, or even smearing the agent on victims’ skin (CBC, 2017). Larger-scale attacks typically involve some form of aerosol dispersal, such as via a spray or an explosion, which may require that the agent not only be prepared at the optimal particle size for inhalation but also be able to withstand freeze drying, suspension in aerosol preparations, packaging processes, long-term storage, and adverse environmental conditions such as ultraviolet sunlight or extreme temperatures (Frerichs et al., 2004). Such requirements may impose significant barriers to bioweapon development, even with available biotechnologies. While synthetic biology could potentially be used to increase a pathogen’s environmental stability, infectivity, transmissibility, or tolerance for weapons delivery systems, maintaining potency or viability throughout the production, storage, and delivery process is still likely to present a significant challenge, particularly for large-scale attacks.
The agent’s ability to be transmitted from one individual to another is an important consideration in terms of both production scale and delivery. A communicable agent could theoretically be deployed in small amounts at multiple locations and allowed to spread on its own. Some actors may even find volunteers willing to spread infection by becoming infected themselves, akin to suicide bombers.
Attacks may target individual people; groups of people who share a common geography, occupation, ethnicity, or other attribute; or entire populations. Historically, targeting of bioweapons has been based largely on geographic location of the intended victims. Biotechnology advances may offer new opportunities for a malicious actor to influence the overall impact of an attack or the specific individuals affected, such that an agent could be deployed over a broad geographic area but only sicken targeted individuals. For example, actors may consider designing a bioweapon to target particular subpopulations based on their genes or prior exposure to vaccines, or even seek to suppress the immune system of victims to “prime” a population for a subsequent attack. These capabilities, which were feared decades ago but never reached any plausible capability, may be made increasingly feasible by the widespread availability of health and genomic data. While some fundamental barriers still likely limit the success and reliability of such an effort—for example, the United States’ genetic diversity may make the U.S. population resistant to targeting based on ethnicity—it is nonetheless crucial to continue to monitor developments that could facilitate targeting of particular populations.
The challenges associated with effecting an attack using a synthetic biology–enabled weapon may be overcome by emergent (new) or convergent capabilities. In the context of technology, convergence occurs when different technologies, often from different fields, create synergies that significantly advance capabilities when they are combined (Roco, 2008). In other contexts, convergence has been described as the formation of a framework to solve scientific and societal challenges that exist at the interfaces of multiple fields (NRC, 2014). In either conceptualization, the merging of diverse areas of expertise can stimulate innovation, from basic science discovery to translational application, which can advance beneficial and malicious goals alike. Convergence can happen through gradual advances over time or occur quite suddenly, taking everyone by surprise. This study considered how developments in multiple fields may converge with biotechnological developments to enable new breakthroughs in the Design-Build-Test cycle or act as “force multipliers” in advancing synthetic biology capabilities. Convergence, of course, can go both ways; as synthetic biology incorporates technologies from other fields, so too will other fields incorporate approaches from synthetic biology, potentially leading to more interdisciplinary collaboration and further breakthroughs. While synergies among technologies are included in the framework within usability of the technology, it is useful to consider how emergent and convergent technologies may allow breakthroughs specifically in aspects relevant to weaponization, since these factors are thought to be in many cases a significant limitation.
To that end, several examples were identified to explore technologies being pursued in fields and toward ends that are not directly related to synthetic biology, yet may converge with biotechnology in ways that help overcome some of the challenges related to creating weapons with synthetic biology. These include gene therapy, nanotechnology, automation, additive manufacturing, genomic data, and health informatics. The potential impacts of these technologies are discussed below and summarized in Table 7-1.
Gene therapy has been in development for use in therapeutics for several decades (Moss, 2014), and it can take a number of forms. In an approach known as ex vivo gene therapy, tissues are genetically altered in the cell culture and then transplanted into the body (Hacein-Bey-Abina et al., 2002). Although ex vivo gene therapy is not likely a viable approach for delivering bioweapons, the ability to transduce cells and tissues ex vivo could inform vector improvement and design and provide proof of principle for novel means of delivering substances, thereby providing an in vitro test capability for small-scale bioweapon design and development.
Another approach, known as in vivo gene therapy, might have other implications for bioweapons development. Using this approach, a component (usually a viral vector) is introduced into the body, potentially to a specific target tissue, where it delivers genetic material that creates the desired therapeutic function (Naldini et al., 1996; Kay et al., 2001). Viral vectors are typically chosen as the delivery vehicles because of their naturally evolved ability to
TABLE 7-1 Summary of How Selected Examples of Convergent Technologies May Affect Challenges of Effecting an Attack Using a Synthetic Biology–Enabled Weapona
aShading indicates which attribute each example aligns with most closely.
target specific cells of the human body; their disease-causing genes are removed and replaced with the engineered genetic components. As gene therapy viral vectors continue to be optimized for therapeutic use, their capability to act as delivery vehicles for bioweapons, such as toxin-producing pathways (as discussed in Chapter 5, Making Biochemicals Via In Situ Synthesis) will advance apace.
Gene therapy vectors being researched include adenovirus, adeno-associated viruses, alphaviruses, herpesviruses, retrovirus/lentiviruses, and vaccinia virus (see Table 7-2); gene therapies using retroviruses, adeno-associated virus, and adenoviruses have already advanced to human clinical trials (Edelstein et al., 2007) and in some cases to clinical approval (FDA, 2017a,b; Spark Therapeutics, 2017). The ability of these vectors to transfer genes into cells and the permanence of the edits they make differ from vector to vector. The size of the viral genome is also important, because the size of the engineered gene that can be transferred is limited to what the virus can successfully carry. While problems such as host immune responses, off-target effects, and decay of continued expression have been barriers to successful gene therapy (Verma and Somia, 1997; Mingozzi and High, 2013), work to address these barriers is being conducted and these challenges might not be of concern to an actor seeking to use the approach to deliver a bioweapon as long as the intended victims experience the intended illness or lethality. As gene therapy vectors continue to be made more efficient and coaxed to carry larger transgenes, gene therapy research could pave the way toward circumventing some of the barriers related to delivery of bioweapons.
Most gene therapies today are delivered via injections to target tissues, a route ill-suited to stealthy or widespread delivery of a weaponized gene therapy vector (though perhaps a viable strategy for targeted assassination). The development of inhalable gene therapy is advancing rapidly, however, particularly for treatments of respiratory diseases such as chronic obstructive pulmonary disease and cystic fibrosis (Zarogoulidis et al., 2013). Advances such as these may provide more expanded capability in the future as the aerosol therapy market continues to drive
TABLE 7-2 Characteristics of Viral Vectors Used in Gene Therapies
|Characteristic||Adenovirus||Adeno-Associated Virus||Alphavirus||Herpesvirus||Retrovirus/Lentivirus||Vaccinia (Poxvirus)|
|Genome||dsDNA||ssDNA||ssRNA (+)||dsDNA||ssRNA (+)||dsDNA|
|Host genome integration||No||No||No||No||Yes||No|
|Transgene expression||Transient||Potential for long lasting||Transient||Potential for long lasting||Long lasting||Transient|
|Maximum size of transgene(s)||7.5kb||4.5kb||7.5kb||30kb||8kb||25kb|
innovation for therapeutics. Efforts toward aerosolized delivery of vaccines are also advancing rapidly; this research may contribute to innovations in routes of delivery for gene therapies (Low et al., 2015). As these technologies progress and new therapeutics come to market, facilities manufacturing aerosolized therapeutics are likely to proliferate, raising the possibility not only that such approaches may be misused for the creation of bioweapons but also that apparently aboveboard manufacturing facilities could mask subversive programs to develop bioweapons delivery systems.
Although the viral vectors used in gene therapies are heavily engineered to remove the genes that cause disease and these viruses are used under exacting conditions that guard against spread, viruses have a history of evolving around constraints, and it remains possible that a single-use gene therapy vector could become “lytic,” leading to the spread of a disease. This is of limited concern for work involving many of the viruses in Table 7-2, which have often been heavily engineered to not propagate in the host. However, there has been a rise in the use of viruses, especially measles and vaccinia, for so-called oncolytic therapies in which the virus replicates in a cancer cell and spreads to surrounding cells (Haddad, 2017). Future studies that chart the evolution of oncolytic viruses in human hosts could potentially become roadmaps for the design and construction of effective bioweapons, if only because they bring into high relief the characteristics of the virus that have the greatest impact on tropism, spread, and pathology.
Nanotechnology is driving innovations in the delivery of gene therapies and other therapeutics. Actors with access to nanotechnology tools could adapt these platforms for malicious use, with implications for delivery of pathogens or toxins as well as targeting attacks. Smaller vehicles in general have much better pharmacokinetic and pharmacodynamic properties, making them more effective in penetrating tissues and cells. Nanoparticles used in drug formulations include imprinted polymers, dendrimers, vesicles, nanospheres, nanocapsules, micelles, carbon nanotubes, liposomes, and nanoemulsions (IAP, 2015), and additional nanocarriers are also being researched, including DNA- and viral-based systems.
Engineered nanotechnology could be used to assist in the weaponization of an agent in numerous ways (Kosal, 2009). For example, nanotechnology could be used to create microcapsules or nanocapsules that encase the agent and improve stability or delivery (Koroleva et al., 2016); to make delivery particles more environmentally stable; to create storage devices for biological products; to create specialized nanoparticles that respond to ultraviolet light (Jalani et al., 2016), are activated remotely, or are engineered to evade the immune system (Zolnik et al., 2010; Rodriguez et al., 2013); to confer the ability to penetrate skin or invade into tiny bronchioles in the lung, cross the blood-brain barrier (Saraiva et al., 2016), or target other specific tissues; or to provide advanced aerosolization capability. An example of one nanoparticle formulation and its use as a delivery platform is discussed in Box 7-1.
Automation is growing rapidly in nearly every field. In biology, the growth of automation is evident in the integration of technologies such as microfluidics, mass spectrometry, bioinformatics, and machine learning into laboratory processes. Automation tools allow researchers to screen ever-larger collections of genetic sequences or physical samples for a wide variety of properties; it is now possible to produce and screen hundreds of thousands of clones and variants in a matter of weeks. Malicious actors could take advantage of these capabilities to, for example, streamline testing of agents, increase fidelity, and fine-tune targeting, potentially while evading mechanisms to detect or screen for malicious activity. Although sequence annotation is becoming more precise, many algorithms must still use unvalidated and unverified data (Poptsova and Gogarten, 2010). This creates “noise” in the system that could inform the design of bioagents or allow malicious actors to undermine legitimate research by, for example, deliberately submitting incorrect genomic data to public databases to mask one’s own work or to sabotage the detection efforts of others.
Standard laboratory robotics is now within the reach of virtually any laboratory. By enabling massively scaled-up experimentation and testing, these tools can significantly shorten the time frame of the Design-Build-Test cycle
overall and potentially improve the likelihood of producing the desired biological functionality. Microfluidic tools, which provide the capability to handle small volumes, control laminar fluid flows, and measure perturbations and timescales within biological systems, are becoming particularly common and are used in a wide variety of research arenas, including drug development and the development of sensors for detecting biomarkers, biohazards, or pollutants (Dittrich and Manz, 2006; Berkeley Lights, 2017). In synthetic biology, microfluidics tools are being adopted to make the testing of biological products or systems fast, inexpensive, and robust. By facilitating testing of many agents at small scale and potentially low cost, these tools could provide malicious actors the capability to develop bioweapons by systematically incorporating multiple genetic variations to synthesize and screen multiple variants (a combinatorial approach) rather than a precise, knowledge-based approach. In addition, the automation of protein design, enabled by mass spectrometry, potentially allows hundreds of thousands of variants to be tested, assessed, and used for refining the design of protein properties via machine learning algorithms (Huang et al., 2016). The combined use of automated design with microfluidics can potentially enable an actor to rapidly develop and test multiple versions of a potential agent at small scale, at low cost, and with relatively limited prior knowledge of how to engineer the desired phenotypic result. For desired results such as lethality, combinatorial design and screening could also provide enough confidence in the behavior of an agent that the actor may not need to pursue larger-scale testing, as well as provide a way to achieve proof of principle for facilitating fidelity during production scale-up. Finally, microfluidics in particular can also create synergies with other areas such as nanotechnology by facilitating the creation of homogeneous nanoparticles for agent delivery.
Additive manufacturing technologies, also known as 3D printing, have emerged to create advanced materials with superior performance, lower environmental impacts, or new functionalities. A variety of materials with
complex biological architectures have been successfully emulated in synthetic systems, such as spider silk and leather (Qin et al., 2015). Although the vast majority of commonly available 3D printing technologies have been unable to sustain living cells, this capability is rapidly advancing (Richards et al., 2013). Examples include the development of 3D printers to generate replacement organs or pharmaceutical testing tissues such as livers and hearts (Robbins et al., 2013); the use of a modified inkjet printer to print layers of Escherichia coli (Lehner et al., 2017); the printing of viable natto bacteria into clothing (Yao et al., 2015); and the proposed use of 3D printing to generate oncolytic viruses (Swenson, 2015).
It is conceivable that one could produce, with biological 3D printing, engineered microbes, viruses, toxins, or other biological products. This capability could also be used to create biological material that could be used as a platform to test bioagents at relatively low cost, or to explore techniques for ensuring bioagent fidelity. Such activities could likely be pursued surreptitiously, because the creation of a small amount of a highly infectious bioagent using a 3D printer would be hard to detect.
Currently, 3D printers tailored specifically for biologicals are still rather expensive and require high expertise; they are not available to the public in libraries and other common spaces as plastics-based 3D printers are. However, as the technology continues to advance, costs may decrease and these devices may become more widely available.
Health-Associated Data and Bioinformatics
In the era of genomics, it has become increasingly feasible to design medical therapeutics tailored to the genetic makeup of an individual or a population. This approach, known as “precision medicine,” relies on the ability to amass large amounts of human genomic data. Sequence data alone are not sufficient, however; it is also necessary to understand genotype–phenotype functional relationships, which often entails tracing epigenetic modifications, metabolism, and changes in protein expression in response to environmental or other factors. The data necessary for such insights can be extracted from blood tests, urinalysis, and a range of other data points stored in individual health records.
Approaches that attempt to link human genomic data with other health metadata are becoming the preferred models for the pharmaceutical industry, making this an extremely active area of research. Not only does this facilitate the pursuit of many more “precise” drug targets, but genomic data, in the context of health metadata, can also allow for reverse engineering approaches for the synthesis of novel small molecules with therapeutic potential (Kim et al., 2016).
None of these approaches is possible without sophisticated bioinformatics and machine learning capabilities that link, correlate, and analyze the data. Such sophisticated techniques also are highly dependent upon having enough correctly annotated data to be able to determine the biomarkers needed to identify specific human conditions of interest. This is likely to present a barrier, particularly for rare or complex multivariant conditions; the existence of more than 5 million known human genetic polymorphisms (Hall, 2011; but GHR  estimates as high as 10 million) hints at the difficulties of trying to determine causative disease factors even with thousands of well-curated patient samples.
While the tailoring of diseases (or spread of diseases) to subpopulations or individuals would not be an exact science, a relatively sophisticated adversary could seek to exploit genomic and health data. The use of genomic data, health metadata, and tailored bioinformatics will continue to advance in the realm of pharmaceutical research, and these advances could enable enhanced targeting capabilities for the development of bioweapons. The vast amount of healthcare data that are now available electronically and the multiple documented incursions into those data, including by foreign powers (Krebs on Security, 2013; Ponemon Institute, 2013; Filkins, 2014; Perakslis, 2014), raises the possibility that an adversary could bypass cybersecurity barriers, identify unique vulnerabilities for specific subpopulations, and then develop bioweapons tailored to target those vulnerabilities. For example, this approach could be used to develop ethnospecific bioweapons. Retroviruses integrate into the genome upon infection, and the integration mechanisms of these viruses could theoretically be altered to greatly favor one genotype over another. Similarly, the existence of population-specific differences in the sequences and structures of receptor proteins suggests that computational modeling, high-throughput screening, or directed evolution could be used to more finely direct an agent to target a specific subpopulation. While such targeting might be more
readily accomplished with known genetic subtypes (such as ethnic subgroups), it may also be possible to target geographic regions or nation-states semiselectively based on allelic distributions in human populations. It may even be possible to drive targeting to an even finer level, raising the specter of “personalized terrorism.”
An increasing knowledge of the human immune system and the ranges of individual responses to diseases also may open opportunities for probabilistic targeting of subpopulations. The ethnic prevalence of preexisting pathogens or the national prevalence of immunotypes (due to vaccination strategies in different countries) could, for example, be exploited in the design of bioweapons targeted to individuals with certain disease or vaccine exposures. General engineering of lowered immunity (discussed in Chapter 6, Modifying the Human Immune System) could lead to additional local endogenous viral reactivation. Similarly, given the somewhat regional nature of even highly cross-reactive allergens, knowledge of a subpopulation available from (stolen) health records might provide clues for probabilistic targeting of anaphylactic shock.
More insidiously, it is possible that some diseases could be engineered not only to target but to actively take advantage of known immune prevalences, in particular those related to vaccination. An extremely sophisticated adversary, knowing in advance the likely fitness landscape of a given pathogen, could release an engineered pathogen that is “designed to evolve” in particular ways upon encountering the most likely human immune response. For example, if an immunodominant epitope is known, and if previous modeling or experimentation had indicated the range of likely sequence substitutions in response to the antibodies already present due to vaccination, and if some of these sequence substitutions lead to increased engagement with a cell surface receptor, then the sequence of the pathogen could be poised in advance to evolve greater lethality or transmissibility. The advantage of this approach, from a malicious actor’s perspective, is that a milder form of a disease could spread broadly and then “self-activate” as a result of “designed evolution” to become a pandemic. As noted in Chapter 4, however, designing such a “new” pathogen is currently far from feasible.
The probabilistic targeting of a disease to unique subpopulations could be used to drive particular military outcomes. Although chicken pox vaccination reduces the importance of this particular example, if a large fraction of a given military cadre is known to have been exposed to a virus such as varicella zoster virus (which causes chicken pox) and is thus at risk to develop a subsequent disease such as shingles, attempting to reactivate and augment this disease might be a viable attack vector. Indeed, the use of probabilistic targeting might prove to be especially important for driving military outcomes in an age where public health measures in the military are virtually universal and can be readily distributed. Probabilistic targeting, combined with targeting via geographic distribution and timed introduction, might be amenable to a larger-scale attack on a region by a more ubiquitous pathogen that could be readily detected and shut down through conventional public health countermeasures.
While factors such as scale-up, stability, fidelity, and delivery are likely to continue to pose barriers to the weaponization of biological agents, a number of technological developments could create synergies with synthetic biology capabilities that allow malicious actors to overcome these barriers. In this chapter, five examples of convergent technologies at various stages of development (see Table 7-3) are presented that may help reduce barriers in various aspects of weaponization (see Table 7-1). It will be important to monitor future developments in these and other areas to identify and assess vulnerabilities that could facilitate bioweapons development. Such developments might result in significant raising of the level of concern related to the synthetic biology–enabled capabilities examined in this study (see Figure 9-1 and Table 9-1).
TABLE 7-3 Summary of Relative Maturity of Selected Convergent Technologiesa
|Technology||In Development||In Use by Developers of the Technology||In Use by Synthetic Biology Community||In Use by Molecular Biology Community||In Use by Amateur Biologists|
aFor each column, darker shading indicates the technology is in routine use for that community, lighter shading indicates emerging use, and white background indicates little or no use. Adoption flows from left to right in most cases.
This page intentionally left blank.