U.S. flag

An official website of the United States government

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

National Academies of Sciences, Engineering, and Medicine; Division on Earth and Life Studies; Board on Life Sciences; Board on Chemical Sciences and Technology; Committee on Strategies for Identifying and Addressing Potential Biodefense Vulnerabilities Posed by Synthetic Biology. Biodefense in the Age of Synthetic Biology. Washington (DC): National Academies Press (US); 2018 Jun 19.

Cover of Biodefense in the Age of Synthetic Biology

Biodefense in the Age of Synthetic Biology.

Show details

6Assessment of Concerns Related to Bioweapons that Alter the Human Host

While we typically think about biodefense in terms of either pathogens (Chapter 4) or biochemicals (Chapter 5), technological advances are now making possible additional capabilities and means of attack that are more closely related to the human body itself. The study included consideration of how increased knowledge about the microbiome and immune system may enable new means of delivering an agent; the potential for incursions into the human host through means not typical of pathogens or toxin-based bioweapons, such as through genetic modification; and how genes themselves may potentially be used as weapons. While some of these potential activities overlap with the activities discussed in previous chapters, it is valuable to consider them from a host-centric angle to assess how advances in knowledge and biotechnology tools might further alter the landscape of vulnerabilities and weapons available for exploitation by malicious actors.


Human health is highly dependent upon the human microbiome—the microorganisms that live on and within us, especially those associated with the gut, oral cavity, nasopharyngeal space, and skin. These populations of microbes are likely far easier to manipulate than the human host itself, making the microbiome a potentially accessible vector for attack. The human microbiome is the focus of a great deal of academic and commercial research, and microbiome manipulation is an area that is rapidly developing, as also discussed in Chapter 5. Several possible ways the microbiome could be manipulated to cause harm were considered; these possibilities were analyzed, in aggregate, to determine the level of concern warranted.

Delivery of harmful cargo via the microbiome. As discussed in Chapter 5, the engineering of microorganisms to produce hazardous chemicals or biochemicals (including toxins) poses a medium to high level of concern and the potential for making chemicals or biochemicals in situ via the microbiome warrants a high level of concern. The microbiome could be used as a vector for other types of harmful cargoes, as well. For example, microbes could be modified to produce functional small RNAs (e.g., microRNAs [miRNAs]) that could be transferred to the host via the gut or skin microbiome1 to cause a variety of health impacts.2 Microbes also could potentially be engineered to horizontally transfer a genetic cargo to the native microbiome to, for example, cause a host's own well-established microbes to produce a harmful biochemical. In such a scenario the harmful agent would be manufactured by organisms in the established microbiome, so the engineered microbe would need to infiltrate and persist within the microbiome only long enough to transfer its cargo to a sufficient number of native microbes. Thus, this approach would circumvent the challenges associated with establishing engineered microbes in otherwise occupied niches. There are many known instances of natural horizontal transfer events that result in the production of toxins (Kaper et al., 2004; Strauch et al., 2008; Khalil et al., 2016). It may be possible to harm a population by enhancing the spread of vectors or phage (viruses targeting bacteria [Krishnamurthy et al., 2016]) carrying such genetic cargoes. Synthetic biology methods could advance such a capability, for example, through the engineering of toxin:antitoxin couples that would help ensure retention of plasmids. It is also conceivable that microbes could one day be engineered to horizontally transfer genes directly to human cells.

Use of the microbiome to increase the impact of an attack. The microbiome can also potentially be exploited to design a more effective bioweapon or increase the impact of an attack. Knowledge of the human microbiome could be used to modify pathogens or their delivery mechanisms to allow more efficient propagation within or between populations, for example, by taking advantage of the frequent exchange of bacteria between humans and animals. In particular, domestic animals could be used as carriers for engineered agents transmitted via the microbiome. For example, engineered dog or cat microbiomes could be established via adulterated feedstocks or via purposeful contamination of populations in animal shelters or pet stores and then subsequently transmitted to humans. Natural transfers resulting from animal-human contact, such as the transfer of the parasite Toxoplasma gondii from cats to humans and the transfer of Campylobacter from dogs to humans, illustrate the feasibility of this approach (Jochem, 2017). Similarly, research into the role of the microbiome in pathogenesis could provide a roadmap as to how to generate improved pathogens that are better supported by their microbial peers. Studies involving wide-ranging transposon- or CRISPR-based deletion libraries of pathogens (Barquist et al., 2013) have provided many insights into pathogenesis that might have dual-use implications, and such libraries could prove useful in identifying which genes productively or specifically interact with endogenous flora to better establish a pathogen.

In addition to using the microbiome to spread toxins and pathogens, manipulating the microbiome might also prove to be a useful adjunct for other biological threats. Recent research shows, for example, that eukaryotic viruses utilize bacteria to improve their chances of infection (Kuss et al., 2011). It is also conceivable that an actor could introduce an initial agent into a population in order to trigger widespread treatment with broad-spectrum antibiotics and then take advantage of the treated population's “clean slate” to introduce or expand an engineered organism via the (now disrupted) microbiome. An actor taking this two-step approach could even incorporate antibiotic or antiviral resistance elements into the initial attack.

Engineered dysbiosis. Our ever-increasing understanding of the human microbiome may lead to opportunities for engineered dysbiosis—that is, the purposeful perturbation of the normally healthy microbiome. This could be accomplished either by causing a known dysbiosis or engineering a new one, and in either case would likely involve introducing otherwise nonpathogenic microorganisms that then lead to diminutions in human health and performance. Since the microbiome likely plays a key role in human immunity (Kau et al., 2011), dysbioses could also potentially be used to cause longer-term debilitation of a population's ability to defend against disease. Gut, oral, nasal, and skin microbiomes could be targets for such an approach. The degradation of military readiness due to continued operations in harsh climes is an ongoing issue. This situation could be made much worse by targeted additions to or alterations of the skin microbiome that lead to heightened chafing, rashes, windburn, and itchiness. While these are seemingly minor concerns, over time they could degrade military capabilities to the point of impacting readiness.

The assessment of concerns related to modifying the human microbiome is summarized here and described in detail below.

Usability of the TechnologyUsability as a WeaponRequirements of ActorsPotential for Mitigation
Level of concern for modifying the human microbiomeMedium-lowMediumMediumMedium-high

Usability of the Technology (Medium-Low Concern)

Engineering the microbiome for any of the purposes described above would be difficult in the near term, leading to a medium-low level of concern with regard to this factor. Given the current level of understanding of the microbiome, the genetic modification(s) required to effect desired phenotypic changes are not yet certain. Achieving desired phenotypic results might require the introduction of particular bacterial species or strains and/or particular genetic modifications of these species or strains. In most cases, microbiome engineering is likely to be further complicated by the need to make multiple genetic introductions or edits involving multiple symbiotic microbiome species. Activities in this area may also be hampered by limited understanding of the genomic diversity and plasticity of microbial communities. Today's genomic databases are built around consensus sequences and do not adequately store or link genomic variations from a single sample. The surprisingly large differences in genomic plasticity observed when the U.S. Food and Drug Administration first applied whole-genome sequencing to trace an Escherichia coli outbreak underscore the inadequacy of this approach (Eppinger et al., 2011) and also suggest the difficulties inherent in engineering the microbiome.

There are similar barriers to understanding how to rationally manipulate the environment to encourage particular microbial compositions. For example, the vast differences in human diets worldwide create a plethora of different microbial environments that would be difficult to uniformly engineer. Even if insertion of a pathogenic microbe were possible, metabolism in culture is so different from metabolism in a host that if a given metabolic pathway was altered to achieve a particular phenotype, alternative or secondary pathways might be uniquely turned on in the context of a human host, thus potentially damping or thwarting the desired microbiome phenotypic engineering outcome. However, the microbiome is an extremely active area of research, and capabilities are advancing rapidly, particularly with regard to understanding how environmental perturbations affect species representations (Candela et al., 2012; Ghaisas et al., 2016) and with regard to the development of phages to target bacteria. It will be important to monitor new developments as the enormous interest in the impact of human commensals on human health continues to drive research and investment and will impact the current bottleneck of limited microbiome understanding.

Usability as a Weapon (Medium Concern)

There are many known routes for the introduction of bacteria into populations; the gut, mouth, nasal, or skin microbiomes could potentially be infiltrated through ingestion, dermal, or other exposure routes via a wide variety of avenues, from contaminated food or water to airborne sprays. For the warfighter, the uniformity of the food supply chain may make food of particular concern as a vector for attack; additionally, products such as probiotics and herbal supplements, routinely used by many warfighters (Hughes et al., 2010; Daigle et al., 2015) could be exploited. It also may be possible to engineer a bioweapon to target populations with a specific microbiome profile; any adversary that begins to better parse, store, and analyze the data that are increasingly being collected about human microbiomes will also be in a better position for probabilistic targeting of microbiome threats (see also Chapter 7, Targeting). However, the predictability of the results for manipulation of the microbiome will be low and, unlike conventional pathogens, the opportunities for dissemination via human-to-human transmission are reduced. On balance, the availability of routes to introduce bacteria tempered by the lack of predictability leads to an overall level of medium concern for this factor.

Requirements of Actors (Medium Concern)

The probiotics industry is well established and highly distributed; probiotics are being engineered and manufactured by people around the world with relatively low levels of scientific expertise at small-scale facilities using basic equipment. Once a successful microbiome engineering approach is established, subsequent production of bioweapons could likely be achieved with a relatively small organizational footprint. However, a high level of expertise would likely be needed to perform the engineering required. On balance, the expertise required to overcome the technical challenges in combination with the low organizational footprint leads to a medium level of concern for this factor.

Potential for Mitigation (Medium-High Concern)

The ability to recognize and respond effectively to an attack involving the microbiome would likely vary depending on the approach used. Given the still nascent understanding of the succession of microbial populations, the targeted manipulation of the human microbiome is, generally speaking, likely to be difficult to detect or attribute. The effects of an engineered threat, stealthily introduced, might be easily passed off as part of a normal change in microbial composition, particularly if the effects are slow acting or chronic phenotypes (e.g., mental health deficits, immune suppression, skin rashes). If an attack were detected, the individuality and plasticity of the human microbiome would likely make attribution difficult. Additionally, given the proliferation of facilities involved in manufacturing probiotics, it could be difficult to distinguish intentional production of harmful probiotics from natural issues arising from contamination or other breakdowns in normal production quality control. However, the gut and other microbiomes are robust and regularly reestablish microbial equilibria after perturbation, and existing antibiotics may well be an effective countermeasure against engineered microbes. As a result, treating attack victims could be relatively straightforward, and existing public health and outbreak response measures could be well positioned to contain an attack. While the introduction of antibiotic resistance genes might restrict the possibilities for treatment, this problem differs little from the traditional concerns over the spread of antibiotic resistance in populations and can potentially be overcome through the use of novel antimicrobials, especially in small cohorts. The overall level of concern for this factor is medium-high; the high level of concern that such an attack would be difficult to detect is reduced somewhat by the ability to treat if it were detected.


Human immunity is the bulwark for protection against infectious disease. Two basic systems respond to the vast array of threats in the natural environment. The first is the innate immune system, a collection of nonspecific protective mechanisms triggered by pathogen-associated molecular patterns, such as lipoteichoic acid from Gram-positive bacteria or unmethylated CpG sequences in viral DNA. The second is the adaptive immune system, which generates highly specific antibody and T-cell responses tailored to individual diseases and disease variants. Many natural pathogens manipulate the human immune system, both by suppressing the immune response (e.g., immunodeficiency viruses) and by upregulating certain responses (e.g., respiratory syncytial virus, which induces the immune system to favor a response involving Type 2 T helper cells [Th2] and subsequently increases the proclivity toward asthma [Lotz and Peebles, 2012]). These examples suggest that it may be feasible to develop a bioweapon capable of manipulating or “engineering” the immune response. Several potential forms for such a bioweapon were considered:

Engineering immunodeficiency. Manipulating a target population to have decreased immunity could increase the impact of a biological attack. This goal could be pursued either by manipulating a pathogen to simultaneously reduce immunity and cause disease (Jackson et al., 2001) or by separately introducing an immune-suppressing agent and a bioweapon into a target population. Agents used to cause immunodeficiency could be pathogens (e.g., the insidious spread of HIV [human immunodeficiency virus]) or chemicals (see NRC [1992] and IPCS [1996] for discussions of chemicals that contribute to immunotoxicity). It is also possible that a disease agent could be tailored to the immune state of a population, either by engineering the agent to avoid extant adaptive or innate immune barriers or by actually taking advantage of those barriers (for further discussion see Chapter 7, Health-Associated Data and Bioinformatics).

Engineering hyperreactivity. The flip side of engineering immune deficiencies would be to attempt to cause immune hyperreactivity. Both pathogens and chemicals have been demonstrated to create a cytokine storm, a dangerous state that results from a positive feedback loop in the immune response. It may be possible to engineer an agent to purposefully trigger such a cascade. For example, some have suggested that the introduction of anthrax lethal toxin into a more benign disease vector could trigger a cytokine storm (Muehlbauer et al., 2007; Brojatsch et al., 2014; however, see Guichard et al., 2012 for a differing point of view). Similarly, the fact that there are already widespread responses in the human population to a limited number of well-known allergens (ACAAI, 2017) may provide a means of engineering biological threats that would trigger life-threatening IgE-mediated immune responses. The development and testing of new immunotherapies could also provide a roadmap for potentially engineering threats; for example, actors could learn from clinical studies in which anti-CD28 antibodies caused life-threatening cytokine storms (Suntharalingam et al., 2006).

Engineering autoimmunity. Natural autoimmune diseases cause significant disability and death. It may be possible to engineer a disease that causes the body to turn on itself. Mouse models for the stimulation of autoimmunity now exist. For example, Experimental Autoimmune Encephalomyelitis, which mimics the symptoms of the human malady multiple sclerosis, has been induced in mice by immunization with antigens that cause an immune response (autoantigens; see Miller et al., 2007). Normally, such self-immunization is prevented by the mechanisms that ensure exclusion of antibodies and T-cells that are self-reactive, but some pathogens may present antigens that are similar enough to the body's own proteins that the original immune response spreads from the pathogen to the new human target. Research into checkpoint inhibitors, compounds designed to unleash the human immune system to eradicate tumors, could also potentially inform efforts to purposely engineer autoimmunity. By overstimulating the immune system, checkpoint inhibitors have been shown to lead to autoimmunity, often in the form of colitis (June et al., 2017). In addition, particular compounds have been shown to lead to an autoimmune disease of the liver (Tanaka et al., 2017, 2018). One potential route of attack could be to introduce such compounds via the microbiome.

The assessment of concerns related to immunomodulation is summarized here and described in detail below.

Usability of the TechnologyUsability as a WeaponRequirements of ActorsPotential for Mitigation
Level of concern for modifying the human immune systemMediumMedium-lowLowHigh

Usability of the Technology (Medium Concern)

It is difficult to predict precisely the impact of engineering on a system as complex as the immune system. We are only now beginning to more fully understand the mechanisms for how the immune system recognizes foreign antigens, and many immune mechanisms, such as how immune memory guides future responses, remain opaque. In addition, much of the research in this area is on animals, and the results do not necessarily map well to humans. Furthermore, while there has been an explosion of new research into the causes of autoimmunity, the onset of autoimmune disease remains idiosyncratic (Rosen and Casciola-Rosen, 2016), and it would likely be difficult to create immunomodulatory weapons capable of causing reliable effects in populations as genetically and immunologically diverse as the United States. In particular, while an immune deficiency virus pandemic has emerged naturally, engineering the spread of immune deficiency is currently difficult to imagine.

However, even undirected efforts in this area could be successful enough to warrant concern. In experiments in which mousepox was augmented with interleukin-4 (IL-4) (Jackson et al., 2001), earlier studies had already discerned that vaccinia virus altered with IL-4 increased virulence in mice (van den Broek et al., 2000), but it came as a surprise that the altered mousepox virus could also overcome vaccination against mousepox. The failed clinical trial of anti-CD28 antibodies, in which patients suffered life-threatening cytokine storms after receiving doses 500 times lower than those shown safe in mouse models (Suntharalingam et al., 2006), offers another example. Although modeling studies indicated that the doses used would nearly saturate the T-cell population of a human (suggesting the potential for overactivation), the dramatic outcomes highlight the potential for inadvertent immune hyperreactivity as well as the dual-use potential of immunomodulation research. The concept of engineering a cytokine storm, especially in susceptible subpopulations, may become a concern when coupled with increasing knowledge of the immune system. For example, the growing knowledge of superantigens that hyperstimulate immunity could further increase the feasibility of such activities.

Our understanding of human immunity also represents an increasing, but unknown, area of concern. For example, with the advent of next-generation sequencing, the range of both B-cell and T-cell responses to vaccines can now be described in molecular detail. Similarly, the effectors of the pattern recognition receptors of the innate immune system are being defined to the point that engineering responses, both therapeutic and otherwise, are possible (Brubaker et al., 2015; Macho and Zipfel, 2015). In addition, the continuing explosion of work in immunotherapy broadly could potentially create a roadmap for the development of immunomodulatory weapons. As understanding of this phenomenon improves and as the ability to engineer protein structures improves, the opportunities for creating synthetic simulacrum of antigens already known to be present in autoimmune diseases will increase. The opportunities to engineer autoimmunity are likely tempered by the diversity of potential autoantigens that can be exploited, although this could also be viewed as a means of disease targeting as more and more personalized health data become available (see Chapter 7, Health-Associated Data and Bioinformatics).

On balance, given the challenges and both near- and longer-term opportunities, there is a medium level of concern with regard to usability of the technology for the variety of ways in which immunomodulation might be employed as a bioweapon.

Usability as a Weapon (Medium-Low Concern)

The connections between factors capable of influencing immunity and the actual immune response of individuals remain poorly understood. Although it is possible to imagine generic degradations to, or overstimulation or mis-stimulation of, the human immune system, it will initially be very difficult to target such threats to particular individuals or populations, and thereby to have a clear and predictable path to an overall impact on a population's health or on military readiness and response. However, although immunomodulation might not necessarily be the most effective approach for an adversary seeking to effect large-scale and immediate death or debilitation, this approach could nonetheless undermine a nation's capabilities. The 1918 influenza pandemic, likely abetted by an interplay between viral infectivity and poor public health, was a major factor in military preparations for the first World War (Byerly, 2010); this historical example serves as a reminder that a general decrease in immunity would even today have strategic consequences for the military machine. Nonetheless, because there are few ways to model or manipulate the human immune system other than by carrying out large-scale experiments on humans themselves, the amenability of this particular threat to improvement via the Design-Build-Test cycle is minimal, and predictability of results is likely to remain a significant barrier in the near term. Therefore, there is a medium-low level of concern with regard to this factor with the engineering of delivery systems amenable to delivery of immunomodulatory factors an area to monitor.

Requirements of Actors (Low Concern)

The expertise required to modulate human immunity with any degree of surety is likely quite high. In particular, choosing appropriate animal models for testing immunomodulatory interventions remains an art with only a few capable practitioners (Taneja and David, 2001; Benson et al., 2018). Moreover, several of the approaches considered would require an actor to not only successfully develop and deploy the immunomodulatory weapon itself but to successfully plan and execute a multipronged attack in which the immunomodulatory weapon is combined with another biological attack (such as deploying a pathogen after an initial attack causing immunodeficiency) or specialized public health knowledge (such as vulnerabilities created by vaccination patterns, see Chapter 7, Health-Associated Data and Bioinformatics). Such approaches therefore increase the already advanced level of expertise required to effect an immunomodulatory attack, leading to an overall low level of concern for this factor. However, fast-advancing research in immunotherapies may reduce some of these barriers and expand the availability of the appropriate knowledge and skills in the coming years.

Potential for Mitigation (High Concern)

Modulation or evasion of the human immune system is already a hallmark of many pathogens, many of which are constantly developing novel means to avoid immune surveillance (e.g., seasonal adoption of new glycosylation sites by influenza) (Tate et al., 2014). There are also likely many unknown or undercharacterized pathogens that are currently biasing immune responsivity. These natural dynamics would make differentiating between natural and synthetic threats a considerable challenge. It may be particularly daunting to identify the hand of a designer versus the opportunism of nature in a given epitope in a pathogen variant that leads to autoimmunity. The lack of knowledge regarding the mechanisms for discriminating self versus non-self would also increase the challenges associated with recognizing an attack and deploying effective countermeasures. For these reasons, there is a relatively high level of concern with regard to this factor.

Whereas public health measures can potentially be useful in countering a threat involving immunomodulation, recognizing a problem and deploying the appropriate countermeasures would not necessarily be easy or quick; the slow response to the AIDS epidemic, albeit almost 40 years ago, is a potential cautionary tale in this regard. The current state of knowledge regarding immunity is such that it is likely far easier to craft an immunomodulatory weapon than an effective response to one. Even if good countermeasures could be crafted, their expense would likely be inordinate, especially for more general attacks on population immunity.


In addition to using synthetic genes to impact human physiology through pathogens or modifications to the microbiome, it may also be possible to insert engineered genes directly into the human genome via horizontal transfer, in other words, to use “genes as weapons.” Recent improvements in the ability to deliver genetic information via horizontal transfer, for example, through tools such as CRISPR/Cas9, potentially open the way for synthetic or cross-species transfer of genetic information into human hosts. In addition to protein-encoding genes, genes that encode RNA products such as short hairpin RNAs (shRNAs) or miRNAs could potentially be exploited as weapons in their own right. In combination with technologies for the modification of genes or their expression, deepening insights into systems biology could open new opportunities for causing diseases that are outside the rubric of the types of threats typically focused on in biodefense. Several ways in which synthetic biology approaches could be used to horizontally transfer genetic information to a human target to cause harm were considered:

  • Deletions or additions of genes. If researchers can create mouse models of particular disease states based on the deletion or addition of particular genes, it follows that if the genomes of human beings could be similarly modified, such modifications could potentially cause a wide variety of noninfectious diseases. In particular, decades of research on genes associated with oncogenesis—oncogenes—have yielded many examples of gene changes that lead to cancer, including via infection by viruses and bacteria (Robinson and Dunning Hotopp, 2014; Cui et al., 2015; Sieber et al., 2016). Oncogenes could potentially be horizontally transferred to human cells via unnatural means. In this vein, CRISPR/Cas9 has been used to create point mutations, deletions, and complex chromosomal rearrangements in germline and somatic cells to develop mouse models for cancer (Mou et al., 2015).
  • Epigenetic modifications. Just as programmed genetic modifications are possible, it may also prove possible to use horizontal transfer to alter the epigenetic state of an organism in a way that causes harm. Epigenetic modifications are clearly of immense importance in gene expression and are implicated in disease states and pathogenicity. For example, it is now proving possible to predict the course of oncogenesis based on the epigenetic state of a tumor (Jones and Baylin, 2007). Sequence-specific epigenetic modifications can be carried out by small RNAs in other species, such as plants, but are not extensive in humans (He et al., 2011). However, the sequence-specific binding capabilities of Cas9 and other CRISPR elements may allow fusion proteins to carry out sequence-specific epigenetic modifications (Brocken et al., 2017). There are also chemicals that yield relatively nonspecific epigenetic changes (Bennett and Licht, 2018).
  • Small RNAs. Small RNAs are another example of functional genetic information that could be horizontally transferred. Small RNAs, although not a genome modification per se, are important because they may prove capable of modifying gene expression and bringing about phenotypic change. The large number of small interfering RNA (siRNA), short hairpin RNA (shRNA), micro RNA (miRNA) (Zhang et al., 2007; Huang et al., 2008), and other small-RNA library studies in a variety of species and cells from different species, including human, provides a potential roadmap of what sequences may lead to what disease states or to modulation of defenses against disease. Similarly, there are already numerous viral and other vectors that can encode and express small RNAs. The fact that many viral pathogens already seem to encode small RNAs that aid in their pathogenicity further underlines this possibility. For example, the oncogenic gamma herpesviruses Epstein-Barr virus (EBV) and Kaposi's sarcoma–associated herpesvirus (KSHV) encode miRNAs that clearly act as mediators of immune suppression (Cullen, 2013). While most gene delivery mechanisms would likely be facilitated by CRISPR elements, direct delivery of small RNAs via liposomes or other vehicles has proven possible in many cell types (Barton and Medzhitov, 2002; Wang et al., 2010; Miele et al., 2012), and more recently the delivery of entire messenger RNAs (mRNAs) has proven useful for vaccination and cellular reprogramming (Steinle et al., 2017). Naked RNA is generally considered to be fragile due its susceptibility to ribonuclease in the cell, and its delivery is largely confined to laboratory settings, but there are approaches for stabilizing RNAs (e.g., using liposomes, nanoparticles, synthetic polymers, cyclodextrins, ribonucleoproteins, and viral capsids [“armored” RNAs]) in use for many applications. RNA can be expressed from genes delivered as simple expression vectors, as low-fitness-burden cargoes on viral pathogens, or via CRISPR element insertion. One reason that RNA delivery is potentially a viable biological threat is that even a small initial skew in gene expression (such as the changes in gene expression normally caused by miRNAs) could greatly alter the probability of an initial cellular alteration. Even small amounts of a targeted RNA would not modify the genome per se, but might allow or encourage cells to begin the process of self-transformation to tumors, as evidenced by the fact that a large number of pro-oncogenic miRNAs have already been discovered (O'Bryan et al., 2017). In addition to RNAs produced by viruses, bacteria produce numerous small regulatory RNAs; introduction of these into the endogenous microbiome could lead to dysbiosis. Larger mRNAs can also be delivered via liposomes and nanoparticles or by RNA replication strategies being developed for vaccine production (see Chapter 8, Rapid Development of Self-Amplifying mRNA Vaccines); these methods could potentially be used to express deleterious cargo such as toxins or oncogenes, similar to threats related to DNA vectors.
  • CRISPR/Cas9. CRISPR elements can be harnessed for site-specific cleavage of genes, followed by homologous recombination via double-strand break repair or other mechanisms. This technology has revolutionized genome engineering. The fact that DNA recognition can be programmed by simple modification of an RNA element makes precision targeting of genome change much easier than previous technologies such as zinc finger endonucleases and TAL effector nuclease (TALEN)–mediated sequence-specific recognition of DNA. Another advantage of CRISPR technology is its broad host range; CRISPR elements are able to recognize and bind to DNA sequences in species other than those in which they originally evolved. Thus, the fact that gene editing technologies such as CRISPR make possible genomic changes in animal models that directly impact health and pathogenesis further implies that it may be possible to manipulate either germline or somatic cells to make such changes in humans. Significantly, the sequence specificity of CRISPR elements might also make possible ethnospecific targeting of gene-based weapons depending on the distributions of alleles (see also Chapter 7, Health-Associated Data and Bioinformatics). In terms of delivery, CRISPR elements could potentially be loaded onto a pathogen or delivered via the microbiome to modify human genomes in a way that would pose harm to individuals or populations.
  • Human gene drives. Because of the ability of CRISPR elements to modify genomes, they can be repurposed as selfish genetic elements in their own right, wherein their introduction into a naïve genome leads to their site-specific establishment. In sexually reproducing organisms, an appropriately modified CRISPR element or other homing endonuclease gene, when used as a gene drive, can spread throughout a population. Gene drives are well known in nature, such as the Drosophila P element, which moves nonspecifically through naïve populations based on sexual (vertical) transfer. Gene drives have recently proven to be extremely useful for engineering mosquito populations for infertility (Hammond et al., 2016) and they have been proposed for the attenuation of fitness in other undesirable species, as well (for more detail, see National Academies of Sciences, Engineering, and Medicine, 2016). Concerns related to the use of gene drives in human populations were assessed separately from other potential approaches involving horizontal gene transfer because fundamental differences in the mechanisms involved in these different types of activity engender significantly different levels of concern. The assessment of concerns related to the use of human gene drives is summarized below.
Usability of the TechnologyUsability as a WeaponRequirements of ActorsPotential for Mitigation
Level of concern for modifying the human genome using human gene drivesLow

Assessment of Concerns About Gene Drives

For a gene drive to spread in a population, typically many cycles of reproduction are required so that genes can be vertically transferred from one generation to the next. Because humans have a relatively long generation span due to our age of reproductive maturity, a gene drive would take thousands of years to spread throughout a human population in this manner. In addition, some resistance mechanisms to gene drives are already becoming apparent as barriers to their use (Champer et al., 2017). In short, because of the fundamental and insurmountable constraint of human reproductive cycle length, the level of concern with regard to human gene drives is very low and other factors beyond usability of the technology were not analyzed.

The assessment of concerns related to modifications to the human genome through approaches other than through gene drives is summarized here and described in detail below.

Usability of the TechnologyUsability as a WeaponRequirements of ActorsPotential for Mitigation
Level of concern for modifying the human genomeMedium-lowLowMedium-lowHigh

Assessment of Concerns About Genome Modifications Other Than Gene Drives

Usability of the Technology (Medium-Low Concern)

Engineering genes to infiltrate an individual's genome and cause harm is likely to be a technically challenging endeavor, leading to a medium-low level of concern with regard to this factor. Approaches focused on transient horizontal transfer of genes or small RNAs (e.g., via modified viral vectors) could be used, along with systems biology insights, to engineer changes in genes or gene expression to cause noninfectious disease, such as cancer or neurological debilitation, or to degrade immunity. For example, the use of engineered pathogens to deliver small RNAs that cause healthy cells to initiate tumors may be feasible with current knowledge and technology. However, there would be significant challenges to determining the right targets or edits, packaging the genetic cargo into viral vectors, and delivering it to appropriate host cells.

CRISPR-based genome editing technologies are advancing rapidly and could be used to create genetic modifications propagated through engineered pathogenic vectors or horizontal transfer to human cells. However, it would likely be difficult to implement such genome modifications, in part because of the size of the protein-based machinery required for DNA recognition and cleavage, which would impose a hefty fitness cost on the (likely viral) pathogen unless it is linked with the viral life cycle in some way. In other words, viral pathogens have no need to cleave genomes, and this would likely limit the viability of viruses carrying genome-cleaving machinery. That said, new alternatives to the ubiquitous CRISPR/Cas9 system, such as the smaller Cpf1 (Zetsche et al., 2015), Staphylococcus aureus Cas (Ran et al., 2015), or newly discovered CasX and CasY (Burstein et al., 2017) could reduce this barrier.

If an actor sought to cause cancer in targeted individuals, it might only be necessary to modify a small number of cells to initiate oncogenesis and cause a self-sustaining and potentially metastatic cancer. Thus, the mechanisms for delivery could be relatively inefficient and might not require a replicating pathogen for initial distribution. A sufficient gene modification could be accomplished, for example, by introducing the ribonucleoproteins (RNPs) of CRISPR elements by themselves, rather than as genes, with an accompanying protein translocation domain to transit cellular membranes (Liu et al., 2015; Kouranova et al., 2016). This makes a CRISPR RNP potentially more akin to a toxin than to a traditional pathogenic biological threat. Similarly, DNA need not replicate to lead to expression in cells; there are many circular and linear plasmid vectors that can be transiently transfected into a host and thereby provide transient expression of even a large cargo (Nafissi and Slavcev, 2012). This route could be used to facilitate delivery of CRISPR/Cas9 and accompanying oncogenic guide RNAs to a host. In addition, a number of RNA-based mechanisms for gene delivery have come to the fore as a result of recent thrusts to create RNA-based vaccines (Kranz et al., 2016; Pardi et al., 2017). These methods lead to amplification of the originally introduced nucleic acid, but do not otherwise spread between individuals. Thus, they could be used to facilitate oncogenesis in a specifically targeted population.

Usability as a Weapon (Low Concern)

Even were it to become more technologically feasible to use genes to cause oncogenesis, neurodegenerative disease, immunological collapse, or other undesirable states, in the absence of a pathogen or greatly advanced unnatural horizontal transfer mechanism to promote the dispersal of a gene, the ability of an actor to deliver genes for these purposes is limited. Therefore, given this barrier, the concern level regarding usability as a weapon is relatively low. The mechanisms of dispersal (other than pathogens themselves) are likely to be low yield, the probability of inculcation of the disease state is likely to be low, and the onset of the disease state is likely not rapid. However, these limitations do not necessarily preclude an actor from pursuing such a weapon, especially since such a weapon could still significantly impact morale and readiness. In addition, many of these envisioned genetic weapons would become substantially more insidious if the skin rather than the bloodstream could be utilized as a route of entry, and improvements in dermal delivery could greatly change the landscape of threat. The use of siRNAs as a means of targeting tyrosine hydroxylase or tyrosinase and thereby treating hyperpigmentated scars (Xiu-Hua et al., 2010) is instructive as to how this route may be actionable; it will be important to monitor future developments in this area.

Requirements of Actors (Medium-Low Concern)

Almost all of the technologies that might be instrumental in the use of genes as weapons are still in their translational infancy, practiced primarily in research laboratories and not in the clinic. Therefore, the concern level with regard to requirements of actors is medium-low. Achieving the types of potential bioweapons envisioned would likely require advanced research knowledge and experience, not just technical ability. Even advanced companies that would be best suited for the development of dual-use technologies, such as siRNAs, have yet to fully develop delivery methods for desired biomedical applications. One possible exception is the development of bioweapons designed to cause cancer; possible approaches for such an attack can be inferred from knowledge of how chemicals in the environment have impacted cancer epidemiology and from laboratory data on how to induce cancers in animals. An additional caveat is that the rapid spread of technologies for genome engineering via CRISPR element toolsets could potentially decrease the barrier to entry for actors. For example, gene editing could be used to engineer a gene drive into an endemic insect or other pest population to assist delivery of a noxious or infectious agent. In this scenario, even a poorly functioning gene drive might not have to be successful for very long to achieve an effect.

Potential for Mitigation (High Concern)

Overall, the relative level of concern related to the potential for mitigation of gene-based weapons is high. Although some types of impacts would be readily recognized and attributed to a purposeful attack, it would be extremely difficult to trace some impacts—an epidemic of new cancers, for example—to a bioweapon. Such an attack may unfold very slowly, gradually skewing the health of a population. This would make mitigation very difficult, as presaged by experiences with identifying, tracing, and addressing cancer epicenters near toxic waste sites over the past several decades. The considerable challenge of mitigating an intentional cancer epidemic is a primary driver for the high level of concern relating to mitigation for this potential threat. However, once a threat is recognized, established mitigation methods such as quarantine and potential new ones such as therapeutic genome editing could be effective against some types of gene-based weapons.

Given that exome sequence data are being generated at an exponential rate, the introduction of CRISPR elements in humans or other higher organisms would likely be identified quickly and immediately recognized as cause for alarm. The presence of previously unknown oncogenes in viruses not normally known to harbor oncogenes would also be an immediate cause for alarm. However, the surreptitious spread of an oncogenic small-RNA sequence, especially if it is embedded within a protein-encoding gene, might be less noticeable and thus evade detection.


  • The alteration of humans through mechanisms that are different than conventional pathogens is an important potential concern area. The reduction or removal of key bottlenecks and barriers in the future could make some of the approaches discussed in this chapter more feasible.
  • As understanding of microbiomes increases, the possibility of misuse also increases, and it may become feasible to use synthetic biology to engineer the microbiome to transfer toxic genes, debilitate human immunity, improve pathogen entry or spread, or create dysbioses.
  • The threat posed by human immune modulation is limited by current knowledge, but knowledge is accumulating rapidly enough that it may well become more feasible to predictably modify the human immune system.
  • Strategies to modify the human genome or alter gene expression in undesirable ways include gene editing, delivery of RNA molecules, and use of chemicals with epigenetic effects, although significant technical and delivery barriers remain that constrain feasibility.

While the traditional biodefense paradigm places agents such as pathogens or chemicals at the center of considerations of threat and vulnerability, this chapter attempts to reshape that paradigm by considering how interplay with and potential modifications of the human host might change the threat landscape. As understanding of the human microbiome, human immunity, and the human genome increases, the possibility of misuse also increases. In addition, advances in the understanding of individual genetic variability and in the ability to exploit individual variation may make it more feasible to target host-modifying attacks to individuals or subpopulations (further discussed in Chapter 7, Health-Associated Data and Bioinformatics).

The current state of knowledge of the human microbiome is rapidly increasing, and it may be feasible to use synthetic biology to engineer the microbiome to transfer toxic genes, debilitate human immunity, improve pathogen entry or spread, or create dysbioses. However, with the exception of the in situ production of a hazardous compound (as detailed in Chapter 5, Making Biochemicals Via In Situ Synthesis), these potential threats are of lesser concern than more traditional pathogen- and chemical-centered attacks. Despite being an active area of research, the microbiome is still not fully understood, and creating a microbe that could colonize and persist within an established commensal community is a significant challenge. Furthermore, the judicious use of antibiotics could be an effective countermeasure to attacks propagated through the microbiome. Indeed, given the strong push to improve human health via microbiome research and engineering, there may be far more robust opportunities for microbiome-based countermeasures than threats.

The overall concern posed by human immune modulation is similar to the overall concern posed by microbiome engineering, and for similar reasons. On the one hand, current knowledge limitations likely preclude this potential vulnerability from being exploited in a significant way in the near future. On the other hand, knowledge is accumulating at such a rapid clip that it may well become more feasible to predictably modify the human immune system, and the expertise needed to do so is likely to become more widespread in the coming years. In addition, even unpredictable modifications can still cause harm. While it could have been predicted that IL-4 insertion into the mousepox genome would lead to the virus's ability to overcome vaccination (Müllbacher and Lobigs, 2001), it is still unknown whether the same type of modification in a human variant of a virus would have similar dire consequences. In contrast, the development of an anti-CD28 antibody was judged safe enough based on the rigorous review accorded clinical trials, yet proved to be life-threatening (Suntharalingam et al., 2006). Overall, the engineering of hyperimmunity and subsequent pathogenesis seems a greater threat than the engineering of reduced immunity or autoimmunity. The former is acute and fits more readily with individual pathogens and weaponization scenarios; the latter are chronic and with enough foresight can potentially be dealt with at a societal level via the usual public health measures for containing communicable diseases.

Building on that analysis, while the assessment focused on the human immune system, it is important to keep in mind that there are other potential systems that may also prove to be vulnerable to manipulation. For example, human neurobiology is immensely complex, and there are already a variety of genetic and chemical means to manipulate the overall mental health of individuals. That said, it is difficult to engineer such systems for a particular outcome with any surety. It will be important to continue to monitor advances related to understanding and modifying these complex systems in the coming years.

The concept of genes as weapons encompasses the development of synthetic genes that could change human physiology, either on their own or potentially delivered as an augment to a known pathogen. This concept also encompasses the possibility of delivering synthetic genes for small RNAs (or the synthetic small RNAs themselves) that could impact host physiology via interference mechanisms. Genes have a unique position in the biological threat pantheon, being somewhere between pieces of genomes, in which case they can be considered as just parts of pathogens, and being toxins, chemical compounds capable of harm without necessarily replicating. There are multiple difficulties that surround their delivery and a limited number of military scenarios in which an adversary would find it worthwhile to alter human physiology over time frames longer than a single battle or campaign. That said, some scenarios, such as the use of dermal transfection to create shRNAs or miRNAs that alter human physiology, or the use of gene drives to alter insect populations to deliver noxious compounds to humans, may present more attractive options from the perspective of an adversary.

In addition, threats related to horizontal gene transfer in synergy with the threats posed by pathogens may lead to new modes of attack. Just as clinical trials of immunotherapies are increasingly a roadmap for engineering cytokine storms, the increasing knowledge on gene deletions, gene additions, and small-RNA modifications of human cells may provide a roadmap for the induction of noninfectious disease states that could be abetted by pathogen engineering (and, conversely, that could abet the spread of the pathogens themselves, such as via immunodeficiency viruses).

Relevant developments to monitor for each of these capabilities are summarized in Table 6-1.

TABLE 6-1. Bottlenecks and Barriers That Currently Constrain the Capabilities Considered and Developments That Could Reduce These Constraints.


Bottlenecks and Barriers That Currently Constrain the Capabilities Considered and Developments That Could Reduce These Constraints.



The transfer of small RNAs has been demonstrated in other organisms (Zhang et al., 2012), and small RNAs and other nucleic acids derived directly from the diet have been found circulating in higher organisms (Yang et al., 2015).


In human skin, application of anti-tyrosinase siRNAs leads to temporary changes in skin pigmentation (Kim et al., 2012).

Copyright 2018 by the National Academy of Sciences. All rights reserved.
Bookshelf ID: NBK535870


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (2.9M)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...