As discussed in Chapter 1, the committee was asked to consider modern approaches for predicting acute, debilitating chemical toxicity and to suggest an overall conceptual approach that uses emerging science to evaluate acute hazards to deployed military personnel. This chapter first discusses current and future needs for toxicity evaluations of chemical-warfare agents, recognizing the increasing number and types of chemicals that are potentially available to adversaries. It then describes the conceptual framework and strategy developed by the committee for systematically applying modern approaches to the prediction of acute toxicity. The overall approach, which is illustrated in Figures 2-1 and 2-2, consists of three components (relevant terms are defined in Box 2-1):
- A conceptual framework that links chemical structure, physicochemical properties, biochemical properties, and biological activity to acute toxicity.
- A suite of databases, assays, models, and tools that are based on modern in vitro, nonmammalian in vivo, and in silico approaches applicable to predicting acute toxicity.
- A tiered prioritization strategy for using databases, assays, models, and tools to predict acute toxicity in a manner that balances the need for accuracy and timeliness.
Later chapters in this report provide details of the types of databases, assays, models, and tools that are available for evaluating acute toxicity, their integration, and next steps that are needed to begin implementing the committee’s framework and strategy.
Historically, most chemical-warfare agents have belonged to the following chemical classes: nerve agents (such as sarin and soman), blister or vesicant agents (such as phosgene oxime and sulfur mustards), blood agents (such as cyanide), and pulmonary agents (such as chlorine and phosgene) (DHHS 2014). Those agents have been well studied, and a detailed mechanistic understanding that is based on human data is available for some. For example, the organophosphorus (OP) nerve agents are potent inhibitors of acetylcholinesterase and result in acute cholinergic effects that occur minutes or hours after exposure. Knowledge of their mechanisms of toxicity can be useful in the development of therapeutic countermeasures (Sharma et al. 2015) and in the development of in vitro tests. For example, in vitro methods have been developed for the evaluation of cholinesterase inhibition by nerve-gas agents (Worek et al. 2007). Data on some early, sensitive responses to such agents can support development of acute exposure limits for the general public. For example, a number of studies indicate that pupil constriction (miosis) is the most sensitive acute response to human exposure to OP nerve agents (such as sarin), and such end points have been used as part of the basis of acute exposure limits (NRC 2005).
An assay is a laboratory system designed to measure a physical, chemical, or biological end point.
A model is a quantitative or qualitative representation of a hypothesis that attempts to explain how different observations are related to one another. In the context of this report, the hypothesis typically concerns how physical, chemical, or biological data (“inputs”) can be used to predict biological outcomes of a given exposure (“outputs”) in a whole animal or human qualitatively or quantitatively.
A tool is an application of a model or set of models, such as in a software package, designed to be routinely used in an applied setting as opposed to a research or development setting.
In vitro approaches include high-throughput screening, other in vitro assays, and more complex systems, such as organotypic cell cultures.
Nonmammalian in vivo approaches include fish, amphibian, nematode, and insect models.
In silico approaches include computational modeling, structure–activity relationship analysis, analysis of physicochemical characteristics, and read-across techniques (see Chapter 3).
In vivo testing approaches have been developed and applied to assess the toxicity of chemical-warfare agents. The vast majority of available toxicity information has come from traditional toxicity studies in which adverse biological responses were measured in laboratory animals that were exposed to high doses of a test agent. The acute-toxicity data are often used to provide estimates of the amount of an agent that would be required to kill 50% of a population of test animals, such as a lethal dose 50% (LD50) or a lethal concentration 50% (LC50). In addition, pharmacokinetic studies that were designed to identify species differences in chemical absorption, distribution, metabolism, and excretion (Tenberken et al. 2010; Benson et al. 2011a,b) and specialized pharmacokinetic models (such as ones that use a human or porcine skin flap) that were developed to evaluate absorption and toxicity of some chemical-warfare agents (Riviere et al. 1995; Monteiro-Riviere and Inman 1997; Vallet et al. 2008) have been undergoing incremental refinement since their inception. A limitation of the in vivo studies, however, is that they tend to be low-throughput, require consideration of species differences in response, and often provide little insight into a chemical’s mechanism of action.
Only a few chemicals have been formally classified as chemical-warfare agents. However, the list of chemicals that could potentially be used by an adversary against deployed US personnel is large and continues to grow as more chemicals enter the marketplace. Therefore, Department of Defense (DOD) efforts to evaluate potential chemical-warfare agents need to consider a wide array of chemicals beyond traditional chemical-warfare agents, including toxins of biological origin (such as trichothecenes, saxitoxin, and tetrodotoxin), industrial chemicals (such as ammonia), pesticides (such as sodium monofluoroacetate), and pharmaceutical agents (such as cocaine and amphetamine) (Holstege et al. 2007). The ability of an adversary to use those or other chemicals will depend on their or their precursors’ availability and weaponizability and on other factors that were deemed beyond the scope of the committee’s work but that might be important in deciding which agents to evaluate for acute toxicity.
To determine the best way to assess the growing list of registered chemical substances, the committee considered the adverse effects of highly toxic agents, including those of classical chemical-warfare agents, and identified the following organ systems to be of greatest importance for evaluating acute, debilitating hazards: cardiovascular, respiratory, hepatic, renal, skeletomuscular, immune, and nervous systems, including special senses (vision and hearing). Sufficient perturbation in those organ systems can lead to a progression in the severity of effects that can result in incapacitation or death of the whole organism.
Given ethical considerations, additional acute human-toxicity data are unlikely to be available except in cases of accidental release or deliberate attack for which exposure estimates are typically highly uncertain or unknown. And, available traditional toxicity-testing data provide little information about acute, debilitating toxicity. For example, information about chronic, reproductive, or developmental hazards—although important for chemical risk assessment in occupational or environmental settings—is of secondary concern in a military environment where acute, debilitating hazards are of immediate importance. As with other toxicity-testing programs, DOD recognizes that it would be prohibitively expensive and time-consuming to test all potential agents with traditional whole-animal toxicity-testing approaches even if such testing were limited to evaluations of acute toxicity. Moreover, traditional in vivo testing, particularly for acute toxicity, often does not provide information on the cellular or biological mechanisms of toxicity or in some cases even identify the target organ system.
Although some of the more modern, biological assay-based approaches have been used to elucidate mechanisms of action of many of the classical chemical-warfare agents described above, they have not been used to identify potential chemical-warfare agents. Nonetheless, the fact that some high-throughput screening data on chemical-warfare agents already exist suggests the feasibility of using such approaches to evaluate agents and provides important “reference” data with which results on other agents can be compared. The modern predictive approaches can also inform decisions as to whether additional mammalian in vivo testing of an agent is needed and might be able to provide information about the cellular and biological mechanistic events associated with acute toxicity and indicate whether additional testing should focus on a specific organ system or biological target.
A predictive-toxicology program to assess acute toxicity ideally will build on knowledge about the cellular targets and mechanisms of action that are related to acute human toxicity. Acute toxicity depends on fewer biological and chemical pathways than those envisioned by NRC (2007) for a general toxicity evaluation. It could be more straightforward, although still challenging, to predict the potential for acute toxicity than the potential for toxicity in the general public in a variety of organ systems, life stages, populations, and exposure timeframes. Specifically, clinical toxicologists have recognized several cellular or biological targets that are often associated with the acute lethal or debilitating effects of chemicals. Table 2-1 provides an overview of those cellular targets and relevant examples and lists some chemicals that affect the targets. It should be noted that there is not necessarily a one-to-one correspondence between mechanistic targets and organ-system targets because multiple mechanisms could affect a single organ system, a single mechanism could affect multiple organ systems, and debilitation or death could occur from multiorgan failure.
|Biological Process or Cellular Target||Example||Chemical or Biological Agent||Example Target Organ System||Examples of in vitro Assay Approachesc|
|Change in neurotransmitter function|
|Altered axonal transport||Disruption of microtubule function||Vinca alkaloids β, β'-iminodipropionitrile||Nervous||Tubulin polymerization assessed with flow cytometry (Morrison and Hergenrother 2012)|
|Altered impulse conduction by axonal membrane||Blocking of Na+ ion channel||Tetrodotoxin||Nervous||Cell-based assays of the membrane potential that use fluorescent dye (Hill et al. 2014)|
|Reduced precursor availability or neurotransmitter synthesis and storage||Inhibition of acetylcholine uptake into synaptic vesicle||Vesamicol
|Nervous||PC12 cell-based microelectrode assay (Cui et al. 2006; Chen et al. 2008)|
|Altered neurotransmitter release||Blocking of release of acetylcholine at neuromuscular junction||Botulinum toxin||Nervous||PC12 cell-based system for in vitro measurements of neurotransmitter release events (Yakushenko et al. 2013)|
|Presynaptic release of acetylcholine and other neurotransmitters||α-latrotoxin|
|Altered neurotransmitter binding at receptor sites||Neurotransmitter agonists||Opioids, benzodiazepines, nicotine, anatoxin-a, kainic acid||Nervous||Review of selected methods to assess receptor binding (Dunlop et al. 2007); use of stably transfected HEK cells expressing human D2, D3, or D4 dopamine receptors as a screening tool (Vangveravong et al. 2006; Xiao et al. 2014)|
|Neurotransmitter antagonists||Curare, α-bungarotoxin, 3-quinuclidinyl benzilate|
|Impaired neurotransmitter inactivation mechanisms||Acetylcholinesterase inhibition
Altered dopamine transporter
Altered serotonin reuptake
Altered dopamine reuptake
|Nerve gas agents
|Nervous||Zebrafish-based (Jin et al. 2013) and enzyme-based (Wille et al. 2010) assays for acetylcholinesterase inhibitors|
|Altered ion flow|
|Altered electrical conduction of heart or cardiomyocyte contractility||Sodium–potassium
|Digoxin||Cardiovascular||Assessment of altered cardiomyocyte contraction (Himmel 2013; Pointon et al. 2013, 2015; Sirenko et al. 2013; Scott et al. 2014) and electrophysiology (Lopez-Izquierdo et al. 2014); organotypic zebrafish heart model (Pieperhoff et al. 2014)|
|Altered ion pump (Na+, Ca++, K+) activity||Inhibit K+ channel function
Inhibit Na+ channel function
|Dendrodotoxin, 4-aminopyridine Tetrodotoxin, saxitoxin||Cardiovascular||Comparison of in vitro potency of saxitoxin in cultured neurons with in vivo results (Jellett et al. 1992; Vale et al. 2008).|
|Increased permeability of cellular membranes|
|Pore formation||Na+/H+ antiporter||Ionophores||Cardiovascular||Assessment of cell permeability and other end points in multiple strains of mouse embryonic fibroblasts (Suzuki et al. 2014)|
|Ion-channel interactions||Transient receptor potential cation channel, subfamily A, member 1 (TRPA1) activation||Sulfur mustard||Respiratory||Role of TRPA1 as a chemosensor (Büch et al. 2013; Stenger et al. in press)|
|Chemical reactivity||Acylation of proteins and lipids (pulmonary edema)||Phosgene||Respiratory||Human epithelial lung cells as a system to investigate pulmonary edema (Wijte et al. 2011)|
|Mitochondrial dysfunction||Multiple mechanisms||Various||Multiple||Various HTS of mitochondrial dysfunction (Jensen and Rekling 2010; Sakamuru et al. 2012; Vongs et al. 2011; Attene-Ramos et al. 2013, 2015; Sirenko et al. 2014b; Wills et al. 2013)|
|Reduced ATP production||Inhibition of oxidative phosphorylation||Fluoroacetate, cyanide, chlordecone, bromethalin||Nervous, cardiovascular, multiple||Monitoring of ATP production or cell concentrations (Steinhoff et al. 2015)|
|Activation of apoptotic pathways||Multiple||Cisplatin, doxorubicin||Multiple||Cell-imaging methods for cultured cardiomyocytes (Mioulane et al. 2012)|
|Altered oxygen transport|
|Competitive binding to hemoglobin||Carboxyhemoglobin production||Carbon monoxide||Multiple||In vitro assessment of carbon monoxide and cyanide binding to hemoglobin using human blood (Thoren et al. 2013)|
|Irritant or cytotoxic effects||Pulmonary edema||Phosgene, chlorine, methylisocyanate||Respiratory||Microfluidic system that mimics alveolar-capillary interface of human lung (Huh et al. 2012)|
|Oxidative stress or ROS formation|
|Lipid peroxidation||Hepatic injury||Acetaminophen, carbon tetrachloride||Hepatic||Lipid peroxidation cell-based and cell-free assays (Kelesidis et al. 2014)|
|ROS formation||Renal injury||Aminoglycosides||Renal||HTS assays to measure ROS formation (Adams et al. 2013; Prasad et al. 2013; Zielonka et al. 2014)|
|Altered prostaglandin synthesis||Vascular dysfunction||NSAIDs||Cardiovascular||HTS assay for prostaglandin E synthase activity (Andersson et al. 2012)|
|Biological Process or Cellular Target||Example||Chemical or Biological Agent||Example Target Organ System||Examples of in vitro Assay Approachesc|
|Damage to DNA and subcellular systems|
|Genetic damage||Multiple||Multiple||Multiple||HTS assays to measure genetic damage (Gutzkow et al. 2013; Li et al. 2013; Wasalathanthri et al. 2013; Watson et al. 2014; Bandi et al. 2014; Falk et al. 2014; van der Linden et al. 2014)|
|DNA or protein adduct formation||DNA alkylation||Aflatoxins, cisplatinin (kidney), sulfur mustard||Multiple||Medium-throughput methods for quantification of sulfur mustard adducts to proteins (Andacht et al. 2014; Pantazides et al. 2015)|
|Altered protein synthesis||Inactivation of ribosomes||Ricin||Multiple||Assay for the measurement of adenine released from ribosomes or small stem-loop RNAs by ricin toxin A-chain catalysis (Sturm and Schramm 2009)|
|Disruption of cytoskeleton||Actin or cytoskeleton disassembly||Phalloidin, microcystin||Multiple||Assessment of cytoskeleton integrity in a hepatocyte model (Sirenko et al. 2014a)|
|Immunogenic interactions with cell macromolecules||Alteration of mammalian immune system function||Endotoxin, anthrax exotoxins||Immune||Reviews of endotoxin and anthrax toxins (Thorn 2001; Liu et al. 2013).|
|Autoimmunity||Autoimmune hepatitis and necrotizing myositis||Statins||Multiple||Reviews of statins and myositis (Jones et al. 2014) and hepatotoxicity (deLemos et al. 2014)|
aThe lists of chemicals and biological targets shown here are not intended to be complete; rather, this table shows a variety of plausible biological targets and responses that need to be considered in evaluating chemicals for acute toxicity that could debilitate or kill deployed troops.
bBoldface: Listed in Chemical Weapons Convention or is a suspected chemical agent of concern.
cThere are relatively few applications of these methods to the prediction of acute toxicity; thus, the information is provided for illustrative purposes only to demonstrate the types of approaches used to date (see Chapter 4 for additional information).
Abbreviations: ATP, adenosine triphosphate; DNA, deoxyribonucleic acid; HTS, high-throughput screening; NSAID, nonsteroidal anti-inflammatory drug; PC, pheochromocytoma; RNA, ribonucleic acid; ROS, reactive oxygen species.
The relatively detailed knowledge of the multiple mechanisms by which chemicals can cause acute toxicity supports the basic premise of predictive toxicology that whole-animal toxicity can be predicted on the basis of information on lower levels of complexity down to the level of chemical structure. That premise forms the basis of the conceptual framework developed by the committee, illustrated in Figure 2-1. Specifically, it is hypothesized that chemical structure, physicochemical properties, biochemical properties, or biological activity in isolated cells and tissues or in nonmammalian organisms can predict acute mammalian toxicity. The predictions can arise through observations of empirical or statistical correlations or through knowledge of the relevant mechanistic pathways, either of which could potentially be coupled with toxicokinetic information.
Databases, Assays, Models, and Tools
Evaluating the potential for acute toxicity by using the conceptual framework of predictive toxicology requires a suite of databases, assays, models, and tools to cover the relevant physical, chemical, biological, and toxicological space. In general, “input” information on chemical structure, physicochemical properties, biochemical properties, and biological activity that is used to make predictions will be obtained from relevant databases or assays. Chemical-structure data might range from chemical-grouping data (for example, reaction chemistry domains, such as Michael acceptors) to quantitative descriptors of chemical structure (for example, topological descriptors and semiempirical quantum chemical descriptors). Physicochemical-property data include quantities measured in physical or chemical assays, such as boiling point, pH, pKa, and KOW.1 Biochemical measures are usually measures of specific molecular interactions (such as DNA binding and receptor activation) with biological molecules, such as nucleic acids, proteins (including enzymes and receptors), and lipids. Finally, biological activity might include both specific measures of function (such as acetylcholinesterase inhibition) and nonspecific measures of toxicity (such as cytotoxicity from in vitro assays and LC50 estimates obtained from assays that use Drosophila). Databases and assays for chemical structures and physicochemical properties are discussed in Chapter 3 and Appendix B, and assays for biochemical properties and biological activity in Chapter 4.
In this same context, the prediction “outputs” consist of estimates of end points related to acute toxicity. The end points might be related to particular mechanisms known to cause acute toxicity (see, for example, Table 2-1), end points related to specific organ system targets (noted above), or nonspecific end points, such as death and cytotoxicity. Data on those end points for chemicals of known toxicity (such as classical chemical-warfare agents) can serve as “training” and “test” data for building models or tools to predict the same end points for chemicals on which such data are lacking. Finally, the specific form of the outputs might be qualitative (such as active or inactive), semiquantitative (such as a ranking), or quantitative (such as a numerical estimate of dose). However, as discussed further below, quantitative estimates are likely to be of greatest use for military applications.
A variety of models and tools might be used to provide toxicity estimates. Models and tools might be qualitative (such as decision trees) or quantitative (such as statistical regression) and might include statistically based (or machine-learning–based) models, biologically based models, or a mixture of the two. No model or tool is universally applicable, so it is important that a model’s or tool’s domain of applicability is characterized in terms of the chemical space in which it is predictive and the relevant toxicological end points that are covered. Moreover, toxicokinetic models might need to be integrated into the predictions to address absorption, distribution, metabolism, and excretion relevant to acute toxicity. Finally, models and tools differ with respect to the uncertainty or confidence in their predicted outputs.
1Kow is the octanol-water partition coefficient.
As described further in Chapters 3-5, there are many available databases, assays, models, and tools that could be used to predict acute toxicity. Because they vary in their required level of effort, their relevance to acute toxicity, their domain of applicability, the extent to which they address toxicokinetics, and the uncertainty or confidence in their predictions, the committee developed an overall strategy for using them to evaluate acute toxicity. The committee’s strategy is described next.
Prioritization Strategy for Evaluating Acute Toxicity
Effective implementation of predictive models and tools depends on first identifying the ultimate (and acceptable) use of the predictive outputs. The committee’s task states that DOD needs to understand “the relative threat of the increasingly long list of registered chemical substances, particularly in terms of potential acute hazard.” The committee interprets that statement to mean that the goal of the predictive-toxicology approach is to prioritize substances in the sense of identifying those of greater and less concern for acute toxicity. Three key issues must be considered in developing a strategy for prioritization: the need for quantitative measures of potency and their uncertainties, the need to minimize false negatives, and the need to screen a large number of chemicals rapidly.
The first key issue is that prioritization with respect to toxicity inherently requires a quantitative measure of potency and a characterization of uncertainty. Ideally, potency should be defined in absolute units, such as an acute oral LD50 in milligrams per kilogram per day. Relative potency measures might be informative if they include reference chemicals that have known toxicity and, in that case, could be converted to absolute potency measures if toxicokinetic information is also available to make any necessary adjustments. Qualitative outputs, such as binary categorizations of “active” or “inactive,” might be useful as an additional output to target testing for specific end points but are not useful by themselves. Furthermore, in the absence of human data, there will always be inaccuracies in predicting human toxicity, so it is important to characterize the uncertainty
or confidence associated with any predicted potency value. Because a decision-maker might have defined tolerance for errors (such as for false negatives and false positives2), the degree of uncertainty or confidence in a prediction can influence the decision that is made about a particular substance. Therefore, an estimated confidence interval is essential to any prioritization strategy.
The topic of uncertainty leads to the second key issue: given that this task is meant to prevent death and debilitating injuries of US military personnel, it is expected that there will be a low tolerance for false negatives. A likely consequence of reducing the number of false negatives is that a higher percentage of chemicals will be retained for assessment with more accurate but more resource-intensive approaches. The overall time needed to complete the review for the whole chemical space would increase accordingly. However, the timeframe to complete an assessment of thousands of chemicals could be unacceptably long, and a chemical could be successfully weaponized in that timeframe before a decision has been made.
The timeframe raises the third key issue: the prioritization strategy needs to be able to screen chemicals in a manner that allows rapid identification of the ones that pose the greatest risk. A rapid-screening scenario could be acceptable if follow-up screening is conducted to ensure that all potential chemical threats are eventually identified. It is critical that such an approach incorporate a short timeline that progresses efficiently through a multitiered approach to allow timely reconsideration of chemicals that are not originally classified as posing the greatest risk. Lessons learned from the first round of screening could then be leveraged effectively in the reassessment and enable a more informed review and follow-up validation of the initial approach that can also be rapidly implemented. The risk of using this approach lies in a time lag that could result in weaponization of a chemical that was originally not deemed to pose a great threat.
The policy tradeoff of balancing a low tolerance for false negatives with a need to identify important hazards rapidly is beyond the scope of the committee’s charge. However, as a general approach, the committee found that the policy tradeoff could be managed through a tiered prioritization approach as illustrated in Figure 2-2. Specifically, the committee’s proposed prioritization strategy proceeds through a number of tiers that apply successively more predictive and resource-intensive approaches than the previous ones. At each tier, a chemical is placed into one of three general categories:
- High confidence of low toxicity. These chemicals would be deselected for further study and are considered to have a low relative acute toxicity. The requirement that the determination be made with high confidence addresses the low tolerance for false negatives.
- High confidence of high toxicity. These chemicals would be selected and considered to have a high relative acute toxicity. The requirement that the determination be made with high confidence focuses attention quickly on chemicals that might pose a high risk.
- Uncertain toxicity due to inadequate data. The remaining chemicals would be candidates for moving to the next tier of evaluation for acute toxicity.3 The uncertainties might stem from available predictions of high uncertainty or low confidence or from inadequate coverage of end points deemed important for evaluating acute toxicity. Depending on resource constraints, it might be reasonable to assess the chemicals by using additional factors unrelated to toxicity, such as weaponizability. Thus, some chemicals might be further deselected for further study because they pose a low threat owing to factors unrelated to toxicity (discussion of such factors is beyond the committee’s charge). If additional evaluation of toxicity is determined to be needed, the chemical would be moved to the next tier of hazard evaluation to reduce uncertainty concerning the potential for acute, debilitating toxicity. Uncertainty might also be reduced through additional research into and development of approaches to improve acute-toxicity prediction, that is, by
2In this context, a “false negative” occurs when a chemical is identified as having low toxicity when it actually has high toxicity, and “false positive” occurs when a chemical is identified as having high toxicity when it actually has low toxicity.
3Some chemicals in categories (a) and (b) might be carried to a higher tier for validation purposes.
decreasing the number of chemicals in category (c) and increasing the ability to discriminate between categories (a) and (b).
The categorization can be based on a single end point (possibly based on multiple approaches) or on multiple end points. The committee notes that an end point could be a clinical outcome or a molecular initiating event (see Figure 3-1). If science advances in such a way that adverse-outcome pathways of interest to DOD are known, the strategy shown in Figure 2-2 could rely on nontesting and biological assay-based approaches that evaluate molecular initiating events or measurable key events in the pathways.
The committee broadly grouped the available approaches to predicting acute toxicity into four tiers, beginning with an initial chemical characterization (Tier 0), proceeding to nontesting approaches (Tier 1), then to biological assay-based approaches (Tier 2), which includes nonmammalian animal species, and ultimately to traditional whole-animal toxicity testing (Tier 3). The tiers are described further in Box 2-2.
FIGURE 2-2 Prioritization strategy based on a tiered approach for using predictive-toxicology models and tools to evaluate agents for acute toxicity. The strategy can be applied to a single end point (such as lethality, neurotoxicity, and cytotoxicity) and to multiple end points.
A tiered approach to predicting toxicity consists of successively more predictive and resource-intensive approaches to evaluating toxicity (see Figure 2-2). DOD might deselect a chemical at any tier on the basis of factors unrelated to toxicity, such as availability or weaponizability.
Tier 0 would be an initial chemical characterization of toxicity and physicochemical properties based on existing data. In addition to characterizing acute toxicity, traditional toxicity data can be used to build and test predictive-toxicology models in Tiers 1 and 2, and physicochemical data might be important for understanding potential exposure routes, bioavailability, target-tissue distribution, and potential physical hazards or chemical reactivity associated with an agent. Chapter 3 and Appendix B discuss the availability, accessibility, and sources of acute-toxicity data and other data useful for initial chemical characterization.
Tier 1 uses models and tools that make predictions based on chemical structure and physicochemical properties. Such models and tools, discussed in Chapter 3, are termed nontesting approaches because they do not involve any additional toxicity testing and data generation. Such approaches include the use of structure–activity relationships, quantitative structure–activity relationships, and read-across. As discussed further in Chapter 3, the available approaches and tools differ in their potential applicability to prediction of acute toxicity, their chemical domain of applicability, and their predictive power and degree of uncertainty. There are a number of gaps in chemical space, biological space, and predictivity; for many chemicals or end points, predictions based on nontesting approaches will often be highly uncertain.
Tier 2 is the conduct of biological assays to generate data to reduce uncertainty in the toxicity evaluation. Biological assays in this tier include specific protein assays, cell-based phenotypic assays, organotypic models, and nonmammalian in vivo animal models. Toxicity predictions based on such data, discussed in Chapter 4, are termed biological assay-based. Ideally, this biological testing focuses on specific biological targets that are based on information from previous tiers. However, it is also likely to include nonspecific toxicity end points, such as cytotoxicity. As with nontesting approaches, available biological assay-based approaches and tools differ in their potential applicability to prediction of acute toxicity, their chemical domain of applicability, and their predictive power and degree of uncertainty. There are a number of gaps in chemical space, biological space, and predictivity; for many chemicals or end points (although one hopes fewer than in Tier 1), predictions based on biological assay-based approaches will be highly uncertain.
Tier 3 is the conduct of mammalian in vivo testing. These traditional approaches are not part of the committee’s task. However, the committee notes that as in Tier 2, ideally this toxicity testing will focus on specific biological targets that are based on information from previous tiers. There could also be specific gaps or limitations identified in earlier tiers that could be addressed with additional research or development of new models and tools.
A key step in each tier is integration and decision-making (described in Chapter 5). Even within a tier, such as nontesting approaches (Tier 1), there might be diverse outputs and predictions from different models or tools that need to be synthesized. For example, a simple integration approach could be in the form of a scorecard that counts “positive” and “negative” results from available nontesting approaches; a more sophisticated integration approach might aggregate different predictions. In addition, Tier 2 integration should consider the previous results of nontesting approaches with the newly generated biological assay data. Absorption, distribution, metabolism, and excretion (ADME) considerations can be integrated to provide relevant information, such as chemical bioavailability or distribution to target organs.
The committee envisions that a decision as to whether a chemical is categorized as having high toxicity, low toxicity, or inadequate data could be made for each end point that is relevant to acute toxicity (see examples in Figure 2-1). As discussed previously, such decisions would be based on quantitative toxicity estimates for each toxicity end point and associated levels of confidence or confidence intervals. Defining the specific “thresholds” for assigning a chemical to each category will require expert judgment on the part of DOD. However, reference chemicals with known high and low toxicities could help to inform those boundaries. Overall, chemicals would also be assigned to categories for multiple individual end points that reflect different types of acute toxicity although, as noted in Chapters 3 and 4, there are many gaps in coverage of end points related to acute toxicity at all tiers. Therefore, noting the gaps as part of the prioritization strategy provides guidance on how to target testing in later tiers. And, it is up to DOD to determine the extent of coverage of end points that is adequate for it to make sufficiently reliable decisions at each tier.
- Finding: There are multiple mechanisms by which chemicals can elicit acute, debilitating toxicity, and these mechanisms provide support for a predictive-toxicology conceptual framework that predicts system, tissue, or organism toxicity on the basis of chemical structure, physicochemical properties, biochemical properties, or biological activity in isolated cells, tissues, and lower organisms.
- Finding: Such a conceptual framework that includes databases, assays, models, and tools that are applicable to prediction of acute toxicity could be used to evaluate a large number of chemicals for acute-toxicity potential more rapidly than traditional, mammalian in vivo studies.
- Finding: In prioritizing chemicals in terms of their potential to cause acute toxicity, DOD will need to balance a relatively low tolerance for false negatives with a need to evaluate a large number of chemicals rapidly. Regardless of how DOD decides to balance those objectives, they can be managed through a tiered prioritization strategy that applies successively more predictive and resource-intensive approaches as needed.
- Recommendation: The committee recommends a prioritization strategy that broadly groups approaches to prediction of acute toxicity into four tiers, beginning with an initial chemical characterization (Tier 0), moving to nontesting approaches (Tier 1), then to biological assay-based approaches (Tier 2), and finally to traditional mammalian in vivo testing (Tier 3). Progression through the tiers will require intermediate integration steps that consider the diverse data within a tier and among tiers. The prioritization strategy can be applied to single or multiple end points.
- Recommendation: As part of the prioritization strategy, the committee recommends placing chemicals into one of three general categories at each tier: “high confidence of high toxicity,” “high confidence of low toxicity,” and “inadequate data to evaluate toxicity confidently.” Chemicals placed in the last category, “inadequate data,” are moved to the next tier for additional, more resource-intensive evaluation. Quantitative estimates of how potent the chemicals might be and of the confidence or uncertainty in each estimate will be needed to place chemicals into categories. DOD will need to use expert judgment to define specifically how chemicals are to be assigned to the different categories.
Adams, D.J., Z.V. Boskovic, J.R. Theriault, A.J. Wang, A.M. Stern, B.K. Wagner, A.F. Shamji, and S.L. Schreiber. 2013. Discovery of small-molecule enhancers of reactive oxygen species that are nontoxic or cause genotype-selective cell death. ACS Chem Biol. 8(5):923-929.
Andacht, T.M., B.G. Pantazides, B.S. Crow, A. Fidder, D. Noort, J.D. Thomas, T.A. Blake, and R.C. Johnson. 2014. An enhanced throughput method for quantification of sulfur mustard adducts to human serum albumin via isotope dilution tandem mass spectrometry. J. Anal. Toxicol. 38(1):8-15.
Andersson, S., M. Norman, R. Olsson, R. Smith, G. Liu, and J. Nord. 2012. High-precision, room temperature screening assay for inhibitors of microsomal prostaglandin E synthase-1. J. Biomol. Screen. 17(10):1372-1378.
Attene-Ramos, M.S., R. Huang, S. Sakamuru, K.L. Witt, G.C. Beeson, L. Shou, R.G. Schnellmann, C.C. Beeson, R.R. Tice, C.P. Austin, and M. Xia. 2013. Systematic study of mitochondrial toxicity of environmental chemicals using quantitative high throughput screening. Chem. Res. Toxicol. 26(9):1323-1332.
Attene-Ramos, M.S., R. Huang, S. Michael, K.L. Witt, A. Richard, R.R. Tice, A. Simeonov, C.P. Austin, and M. Xia. 2015. Profiling of the Tox21 chemical collection for mitochondrial function to identify compounds that acutely decrease mitochondrial membrane potential. Environ. Health Perspect. 123(1):49-56.
Bandi, S., P. Viswanathan, and S. Gupta. 2014. Evaluation of cytotoxicity and DNA damage response with analysis of intracellular ATM signaling pathways. Assay Drug Dev. Technol. 12(5):272-281.
Benson, J.M., B.M. Tibbetts, W.M. Weber, and G.R. Grotendorst. 2011a. Uptake, tissue distribution, and excretion of 14C-sulfur mustard vapor following inhalation in F344 rats and cutaneous exposure in hairless guinea pigs. J. Toxicol. Environ. Health A 74(13):875-885.
Benson, J.M., A.P. Gomez, M.L. Wolf, B.M. Tibbetts, and T.H. March. 2011b. The acute toxicity, tissue distribution, and histopathology of inhaled ricin in Sprague Dawley rats and BALB/c mice. Inhal. Toxicol. 23(5):247-256.
Büch, T.R., E.A. Schäfer, M.T. Demmel, I. Boekhoff, H. Thiermann, T. Gudermann, D. Steinritz, and A. Schmidt. 2013. Functional expression of the transient receptor potential channel TRPA1, a sensor for toxic lung inhalants, in pulmonary epithelial cells. Chem. Biol. Interact. 206(3):462-471.
Chen, Y., C. Guo, L. Lim, S. Cheong, Q. Zhang, K. Tang, and J. Reboud. 2008. Compact microelectrode array system: Tool for in situ monitoring of drug effects on neurotransmitter release from neural cells. Anal Chem. 80(4):1133-1140.
Cui, H.F., J.S. Ye, Y. Chen, S.C. Chong, and F.S. Sheu. 2006. Microelectrode array biochip: Tool for in vitro drug screening based on the detection of a drug effect on dopamine release from PC12 cells. Anal Chem. 78(18):6347-6355.
deLemos, A.S., D.M. Foureau, C. Jacobs, W. Ahrens, M.W. Russo, and H.L. Bonkovsky. 2014. Drug-induced liver injury with autoimmune features. Semin. Liver Dis. 34(2):194-204.
DHHS (US Department of Health & Human Services). 2014. Chemical Warfare. Classes of Chemical Agents. Specialized Information System [online]. Available: http://sis.nlm.nih.gov/enviro/chemicalwarfare.html#a1 [accessed March 13, 2015].
Dunlop, J., R. Roncarati, B. Jow, H. Bothmann, T. Lock, D. Kowal, M. Bowlby, and G.C. Terstappen. 2007. In vitro screening strategies for nicotinic receptor ligands. Biochem. Pharmacol. 74(8):1172-1181.
Falk, M., M. Hausmann, E. Lukášová, A. Biswas, G. Hildenbrand, M. Davídková, E. Krasavin, Z. Kleibl, I. Falková, L. Ježková, L. Štefančíková, J. Ševčík, M. Hofer, A. Bačíková, P. Matula, A. Boreyko, J. Vachelová, A. Michaelidesová, and S. Kozubek. 2014. Determining Omics spatiotemporal dimensions using exciting new nanoscopy techniques to assess complex cell responses to DNA damage: Part A--radiomics. Crit. Rev. Eukaryot. Gene Expr. 24(3):205-223.
Gutzkow, K.B., T.M. Langleite, S. Meier, A. Graupner, A.R. Collins, and G. Brunborg. 2013. High-throughput comet assay using 96 minigels. Mutagenesis 28(3):333-340.
Hill, A.J., N.A. Jones, I. Smith, C.L. Hill, C.M. Williams, G.J. Stephens, and B.J. Whalley. 2014. Voltage-gated sodium (NaV) channel blockade by plant cannabinoids does not confer anticonvulsant effects per se. Neurosci. Lett. 566:269-274.
Himmel, H.M. 2013. Drug-induced functional cardiotoxicity screening in stem cell-derived human and mouse cardiomyocytes: Effects of reference compounds. J. Pharmacol. Toxicol. Methods 68(1):97-111.
Holstege, C.P., L.K. Bechtel, T.H. Reilly, B.P. Wispelwey, and S.G. Dobmeier. 2007. Unusual but potential agents of terrorists. Emerg. Med. Clin. North Am. 25(2):549-566.
Huh, D., D.C. Leslie, B.D. Matthews, J.P. Fraser, S. Jurek, G.A. Hamilton, K.S. Thorneloe, M.A. McAlexander, and D.E. Ingber. 2012. A human disease model of drug toxicity-induced pulmonary edema in a lung-on-a-chip microdevice. Sci. Transl. Med. 4(159):159ra147.
Jellett, J.F., L.J. Marks, J.E. Stewart, M.L. Dorey, W. Watson-Wright, and J.F. Lawrence. 1992. Paralytic shellfish poison (saxitoxin family) bioassays: Automated endpoint determination and standardization
of the in vitro tissue culture bioassay, and comparison with the standard mouse bioassay. Toxicon. 30(10):1143-1156.
Jensen, K.H., and C. Rekling. 2010. Development of a no-wash assay for mitochondrial membrane potential using the styryl dye DASPEI. J. Biomol. Screen. 15(9):1071-1081.
Jin, S., K.S. Sarkar, Y.N. Jin, Y. Liu, D. Kokel, T.J. Van Ham, L.D. Roberts, R.E. Gerszten, C.A. Macrae, and R.T. Peterson. 2013. An in vivo zebrafish screen identifies organophosphate antidotes with diverse mechanisms of action. J. Biomol. Screen. 18(1):108-115.
Jones, J.D., H.L. Kirsch, R.L. Wortmann, and M.H. Pillinger. 2014. The causes of drug-induced muscle toxicity. Curr. Opin. Rheumatol. 26(6):697-703.
Kelesidis, T., C.K. Roberts, D. Huynh, O. Martínez-Maza, J.S. Currier, S.T. Reddy, and O.O. Yang. 2014. A high throughput biochemical fluorometric method for measuring lipid peroxidation in HDL PLoS One 9(11):e111716.
Li, Y., X. Feng, W. Du, Y. Li, and B.F. Liu. 2013. Ultrahigh-throughput approach for analyzing single-cell genomic damage with an agarose-based microfluidic comet array. Anal. Chem. 85(8):4066-4073.
Liu, S., Y. Zhang, M. Moayeri, J. Liu, D. Crown, R.J. Fattah, A.N. Wein, Z.X. Yu, T. Finkel, and S.H. Leppla. 2013. Key tissue targets responsible for anthrax-toxin-induced lethality. Nature 501(7465):63-68.
Lopez-Izquierdo, A., M. Warren, M. Riedel, S. Cho, S. Lai, R.L. Lux, K.W. Spitzer, I.J. Benjamin, M. Tristani-Firouzi, and C.J. Jou. 2014. A near-infrared fluorescent voltage-sensitive dye allows for moderate-throughput electrophysiological analyses of human induced pluripotent stem cell-derived cardiomyocytes. Am. J. Physiol. Heart Circ. Physiol. 307(9):H1370-H1377.
Mioulane, M., G. Foldes, N.N. Ali, M.D. Schneider, and S.E. Harding. 2012. Development of high content imaging methods for cell death detection in human pluripotent stem cell-derived cardiomyocytes. J. Cardiovasc. Transl. Res. 5(5):593-604.
Monteiro-Riviere, N.A., and A.O. Inman. 1997. Ultrastructural characterization of sulfur mustard-induced vesication in isolated perfused porcine skin. Microsc. Res. Tech. 37(3):229-241.
Morrison, K.C., and P.J. Hergenrother. 2012. Whole cell microtubule analysis by flow cytometry. Anal. Biochem. 420(1):26-32.
NRC (National Research Council). 2005. Impact of Revised Airborne Exposure Limits on Non-Stockpile Chemical Materiel Program Activities. Washington, DC: National Academies Press.
NRC (National Research Council). 2007. Toxicity Testing in the 21st Century: A Vision and a Strategy. Washington, DC: National Academies Press.
Pantazides, B.G., B.S. Crow, J.W. Garton, J.A. Quiñones-González, T.A. Blake, J.D. Thomas, and R.C. Johnson. 2015. Simplified method for quantifying sulfur mustard adducts to blood proteins by Ultrahigh Pressure Liquid Chromatography-Isotope Dilution Tandem Mass Spectrometry. Chem. Res. Toxicol. 28(2):256-261.
Pieperhoff, S., K.S. Wilson, J. Baily, K. de Mora, S. Maqsood, S. Vass, J. Taylor, J. Del-Pozo, C.A. MacRae, J.J. Mullins, and M.A. Denvir. 2014. Heart on a plate: Histological and functional assessment of isolated adult zebrafish hearts maintained in culture. PLoS One 9(5):e96771.
Pointon, A., N. Abi-Gerges, M.J. Cross, and J.E. Sidaway. 2013. Phenotypic profiling of structural cardiotoxins in vitro reveals dependency on multiple mechanisms of toxicity. Toxicol. Sci. 132(2):317-326.
Pointon, A., A.R. Harmer, I.L. Dale, N. Abi-Gerges, J. Bowes, C. Pollard, and H. Garside. 2015. Assessment of cardiomyocyte contraction in human-induced pluripotent stem cell-derived cardiomyocytes. Toxicol Sci. 144(2):227-237.
Prasad, R.Y., J.K. McGee, M.G. Killius, D.A. Suarez, C.F. Blackman, D.M. DeMarini, and S.O. Simmons. 2013. Investigating oxidative stress and inflammatory responses elicited by silver nanoparticles using high-throughput reporter genes in HepG2 cells: effect of size, surface coating, and intracellular uptake. Toxicol. In Vitro 27(6):2013-2021.
Riviere, J.E., J.D. Brooks, P.L. Williams, and N.A. Monteiro-Riviere. 1995. Toxicokinetics of topical sulfur mustard penetration, disposition, and vascular toxicity in isolated perfused porcine skin. Toxicol. Appl. Pharmacol. 135(1):25-34.
Sakamuru, S., X. Li, M.S. Attene-Ramos, R. Huang, J. Lu, L. Shou, M. Shen, R.R. Tice, C.P. Austin, and M. Xia. 2012. Application of a homogenous membrane potential assay to assess mitochondrial function. Physiol. Genomics 44(9):495-503.
Scott, C.W., X. Zhang, N. Abi-Gerges, S.D. Lamore, Y.A. Abassi, and M.F. Peters. 2014. An impedance-based cellular assay using human iPSC-derived cardiomyocytes to quantify modulators of cardiac contractility. Toxicol Sci. 142(2):331-338.
Sharma, R., B. Gupta, N. Singh, J.R. Acharya, K. Musilek, K. Kuca, and K.K. Ghosh. 2015. Development and structural modifications of cholinesterase reactivators against chemical warfare agents in last decade: A review. Mini Rev. Med. Chem. 15(1):58-72.
Sirenko, O., E.F. Cromwell, C. Crittenden, J.A. Wignall, F.A. Wright, and I. Rusyn. 2013. Assessment of beating parameters in human induced pluripotent stem cells enables quantitative in vitro screening for cardiotoxicity. Toxicol. Appl. Pharmacol. 273(3):500-507.
Sirenko, O., J. Hesley, I. Rusyn, and E.F. Cromwell. 2014a. High-content assays for hepatotoxicity using induced pluripotent stem cell-derived cells. Assay Drug Dev. Technol. 12(1):43-54.
Sirenko, O., J. Hesley, I. Rusyn, and E.F. Cromwell. 2014b. High-content high-throughput assays for characterizing the viability and morphology of human iPSC-derived neuronal cultures. Assay Drug Dev. Technol. 12(9-10):536-547.
Steinhoff, R.F., M. Ivarsson, T. Habicher, T.K. Villiger, J. Boertz, J. Krismer, S.R. Fagerer, M. Soos, M. Morbidelli, M. Pabst, and R. Zenobi. 2015. High-throughput nucleoside phosphate monitoring in mammalian cell fed-batch cultivation using quantitative matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. Biotechnol J. 10(1):190-198.
Stenger, B., F. Zehfuß, H. Mückter, A. Schmidt, F. Balszuweit, E. Schäfer, T. Büch, T. Gudermann, H. Thiermann, and D. Steinritz. In press. Activation of the chemosensing transient receptor potential channel A1 (TRPA1) by alkylating agents. Arch. Toxicol.
Sturm, M.B., and V.L. Schramm. 2009. Detecting ricin: Sensitive luminescent assay for ricin A-chain ribosome depurination kinetics. Anal. Chem. 81(8):2847-2853.
Suzuki, O.T., A. Frick, B.B. Parks, O.J. Trask Jr., N. Butz, B. Steffy, E. Chan, D.K. Scoville, E. Healy, C. Benton, P.E. McQuaid, R.S. Thomas, and T. Wiltshire. 2014. A cellular genetics approach identifies gene-drug interactions and pinpoints drug toxicity pathway nodes. Front. Genet. 5:272.
Tenberken, O., J. Mikler, I. Hill, K. Weatherby, H. Thiermann, F. Worek, and G. Reiter. 2010. Toxicokinetics of tabun enantiomers in anaesthetized swine after intravenous tabun administration. Toxicol. Lett. 198(2):177-181.
Thoren, T.M., K.S. Thompson, P.S. Cardona, A.K. Chaturvedi, and D.V. Canfield. 2013. In vitro absorption of atmospheric carbon monoxide and hydrogen cyanide in undisturbed pooled blood. J. Anal. Toxicol. 37(4):203-207.
Thorn, J. 2001. The inflammatory response in humans after inhalation of bacterial endotoxin: A review. Inflamm. Res. 50(5):254-261.
Vale, C., A. Alfonso, M.R. Vieytes, X.M. Romarís, F. Arévalo, A.M. Botana, and L.M. Botana. 2008. In vitro and in vivo evaluation of paralytic shellfish poisoning toxin potency and the influence of the pH of extraction. Anal. Chem. 80(5):1770-1776.
Vallet, V., C. Cruz, J. Licausi, A. Bazire, G. Lallement, and I. Boudry. 2008. Percutaneous penetration and distribution of VX using in vitro pig or human excised skin validation of demeton-S-methyl as adequate simulant for VX skin permeation investigations. Toxicology 246(1):73-82.
van der Linden, S.C., A.R. von Bergh, B.M. van Vught-Lussenburg, L.R. Jonker, M. Teunis, C.A. Krul, and B. van der Burg. 2014. Development of a panel of high-throughput reporter-gene assays to detect genotoxicity and oxidative stress. Mutat. Res. Genet. Toxicol. Environ. Mutagen. 760:23-32.
Vangveravong, S., E. McElveen, M. Taylor, J. Xu, Z. Tu, R.R. Luedtke, and R.H. Mach. 2006. Synthesis and characterization of selective dopamine D2 receptor antagonists. Bioorg. Med. Chem. 14(3):815-825.
Vongs, A., K.J. Solly, L. Kiss, D.J. Macneil, and C.I. Rosenblum. 2011. A miniaturized homogenous assay of mitochondrial membrane potential. Assay Drug Dev. Technol. 9(4):373-381.
Wasalathanthri, D.P., S. Malla, I. Bist, C.K. Tang, R.C. Faria, and J.F. Rusling. 2013. High-throughput metabolic genotoxicity screening with a fluidic microwell chip and electrochemiluminescence. Lab. Chip. 13(23):4554-4562.
Watson, C., J. Ge, J. Cohen, G. Pyrgiotakis, B.P. Engelward, and P. Demokritou. 2014. High-throughput screening platform for engineered nanoparticle-mediated genotoxicity using CometChip technology. ACS Nano. 8(3):2118-2133.
Wijte, D., M.J. Alblas, D. Noort, J.P. Langenberg, and H.P. van Helden. 2011. Toxic effects following phosgene exposure of human epithelial lung cells in vitro using a CULTEX® system. Toxicol. In Vitro 25(8):2080-2087.
Wille, T., H. Thiermann, and F. Worek. 2010. Development of a high-throughput screening for nerve agent detoxifying materials using a fully-automated robot-assisted biological assay. Toxicol. In Vitro 24(3):1026-1031.
Wills, L.P., G.C. Beeson, R.E. Trager, C.C. Lindsey, C.C. Beeson, Y.K. Peterson, and R.G. Schnellmann. 2013. High-throughput respirometric assay identifies predictive toxicophore of mitochondrial injury. Toxicol. Appl. Pharmacol. 272(2):490-502.
Worek, F., P. Eyer, N. Aurbek, L. Szinicz, and H. Thiermann. 2007. Recent advances in evaluation of oxime efficacy in nerve agent poisoning by in vitro analysis. Toxicol. Appl. Pharmacol. 219(2-3):226-234.
Xiao, J., R.B. Free, E. Barnaeva, J.L. Conroy, T. Doyle, B. Miller, M. Bryant-Genevier, M.K. Taylor, X. Hu, A.E. Dulcey, N. Southall, M. Ferrer, S. Titus, W. Zheng, D.R. Sibley, and J.J. Marugan. 2014. Discovery, optimization, and characterization of novel D2 dopamine receptor selective antagonists. J. Med. Chem. 57(8):3450-3463.
Yakushenko, A., E. Kätelhön, and B. Wolfrum. 2013. Parallel on-chip analysis of single vesicle neurotransmitter release. Anal. Chem. 85(11):5483-5490.
Zielonka, J., G. Cheng, M. Zielonka, T. Ganesh, A. Sun, J. Joseph, R. Michalski, W.J. O'Brien, J.D. Lambeth, and B. Kalyanaraman. 2014. High-throughput assays for superoxide and hydrogen peroxide: Design of a screening workflow to identify inhibitors of NADPH oxidases. J. Biol. Chem. 289(23):16176-16189.