ICCVAM Biennial Report 2022-2023

In vitro toxicity assays are being applied to transform toxicology from an observational to a predictive science, improve throughput, and reduce costs. The qualitative linkage between in vitro and in vivo toxicity endpoints can be strengthened via application of in vitro to in vivo extrapolation (IVIVE) of dosimetry, which relates an in vitro concentration associated with bioactivity to an equivalent external exposure level. In some contexts, applications of IVIVE have advanced past the exploratory research stage and are beginning to gain acceptance for risk assessment of chemicals. The ICCVAM IVIVE Workgroup requested information from ICCVAM member agencies regarding the extent and context of their use of IVIVE. Surveyed agencies were also asked about programmatic needs, data gaps, and agency-specific guidance documents or publications related to IVIVE, as well as for information about modeling tools or software they had used or may use for facilitating IVIVE analysis and decision-making. This information was compiled into a review summarizing the workgroup’s findings, current challenges, and future needs (Chang et al. 2022). The review also proposes operational definitions for IVIVE, presents literature examples for several common toxicity endpoints, and highlights implications of IVIVE use in decision-making processes. This well-received paper was recognized by the journal Toxics as an "Annual Recommended Review for 2022."

In vitro assay data can provide valuable information about the effects of substances on biological systems. However, to determine safe exposure levels in real-life situations, it is necessary to consider the translation of in vitro results to the in vivo context. In vitro to in vivo extrapolation (IVIVE) uses pharmacokinetic models to relate concentrations of substances eliciting responses in an in vitro assay to an in vivo equivalent administered dose (EAD). NICEATM applied IVIVE to a group of volatile organic compounds with abundant pharmacokinetic and in vivo toxicity data. For each chemical, EADs that would result in internal concentrations (e.g., plasma or target tissue) equivalent to in vitro activity concentrations were estimated. The in vitro activity concentrations were obtained from in vitro assays measuring different endpoints (e.g., cytochrome p450 activation, transcriptome analysis, and genotoxicity) from public resources. EAD estimates were compared to published in vivo point-of-departure (PODs) or minimal risk levels provided by the ATSDR covering multiple target organs toxicities via inhalation exposure. The impact of mechanistic relevance of in vitro assays, target organs chosen for analysis, and the concordance between in vitro and in vivo exposure regimens on IVIVE outcomes will be evaluated and discussed in a presentation (Chang et al.) at the 2024 Annual Meeting of the Society of Toxicology. For most chemicals tested, close agreements between EAD estimates and rat in vivo PODs were observed, but most EADs were at least 10-fold higher than minimum risk levels, suggesting that a “modifying factor” may need to be established to approximate minimum risk levels based on in vitro assay data. This study provides proof-of-concept case examples to illustrate the utility new approach methodologies (NAMs) in informing human risk following exposure to inhaled substances.

The modern battlefield is consistently evolving, and the USAF Force Health Protection program must account for new technologies and chemical threats. While much progress has been made in the development of computational methods for predicting public health risk, the warfighter is expected to experience exposures under conditions that are substantially different from those experienced by the general U.S. population. Thus, USAF Predictive Risk Team (PRT) is evaluating the utility of existing models for operational exposure scenarios and modifying existing models when necessary.

The Air Force population differs from the general U.S. population with respect to age and gender distribution. PRT constructed an approach to use published demographic data (Mullenger and Zehner, 2020) to customize the httk pipeline to adequately model Air Force population. A Monte Carlo simulation was performed to generate 48 anthropometric characteristics that closely resemble the USAF population. Upon completion, an analysis was compiled to compare previous oral-equivalent doses for the general population to those derived for the military. This pipeline is being used as a foundation to perform a high throughput risk analysis on military-relevant chemicals and will continue to be expanded as necessary as the population changes in the future. Additional areas for development in 2024 include consideration of occupational and acute dosing scenarios and physiological changes due to extreme environments.

Neurotoxicity is of particular concern for the military due to the potential for cognitive, behavioral, and physiological effects. With an ever-evolving operational landscape, de novo generation of in vivo data for all potential neurotoxicants is not feasible. USAF Predictive Risk Team (PRT) evaluated the utility of high-throughput new approach methodologies (NAMs) for rapid risk assessment of 220 potentially neurotoxic chemicals. Chemical-specific point-of-departure (PODs) were derived from novel in vitro neuronal assay data and ToxCast bioactivity concentrations. The AC50 (concentration with 50% maximum activity) was derived for each chemical/assay combination and PODs were calculated from the 5th percentile of the AC50 values. In vitro to in vivo extrapolation (IVIVE) was performed using the EPA httk model to estimate the equivalent administered dose (EAD) for each POD. Monte Carlo calculations from httk were modified to reflect USAF active-duty demographics. Neurotoxicity-associated endpoints produced higher PODs compared to all CompTox in vitro endpoints combined. In vitro-derived EADs were lower than in vivo-derived PODs for most of the chemicals in both the U.S. and USAF populations. In vitro-derived EADs were also lower than 45% of provisional reference doses calculated using one-year Military Exposure Guideline procedures. An uncertainty factor of 1000 results in conservative estimated exposure limits (vs. Military Exposure Guidelines) for almost all the chemicals. This case study supports the use of NAMs to derive conservative PODs for risk assessment. A manuscript is currently being prepared for publication in 2024.

Warfighters are subject to a variety of chemical exposures, and PBK models can be used to predict chemical metabolism and potential toxicity given a specific exposure concentration. However, most in silico techniques for estimation of metabolism are only suitable for low exposure concentrations. Saturable metabolism is a key component for extending current high-throughput PBK models beyond environmental exposures to occupationally relevant concentrations. To adequately support human health decision-making, USAF Predictive Risk Team (PRT) scientists identified a need for in silico approaches to estimate chemical metabolism parameters suitable for describing the rate of compound biotransformation over a wide range of exposure concentrations. Sweeney and Sterner (2022) describe the reconstruction of published, insufficiently validated QSAR and the application of these QSARs to jet fuel components. A subsequent publication (Sweeney 2022) used these QSAR estimates, along with other techniques, to demonstrate the value of smaller families of highly similar compounds for generating reliable QSAR-derived metabolism estimates and their application to internal dosimetry-based risk assessment for both low dose (chronic) and higher dose (acute) risk assessment activities. PRT scientists are coordinating a newly established DoD QSAR interest group and are exploring future collaborations that will expand the application of these tools. Additionally, the team is assessing the feasibility of adding saturable metabolism into the EPA httk model.

PFAS are organofluorine chemicals manufactured and used for decades in products and applications such as firefighting foams. PFAS are highly persistent and bioaccumulative in the environment, biota, and humans, and concern about their health effects has driven efforts to identify and evaluate possible substitutes. Scientists at the Defense Centers for Public Health used in silico and in vitro approaches to predict toxicity of six fluorine-free firefighting formulations were assessed for toxicity. Ingredient information provided by the manufacturers missed up to 70% of the constituents per formulation. QSARs were completed for some of the known ingredients, and in vitro assays (the Ames test for genotoxicity, the Microtox test for cytotoxicity, and the EpiDerm assay for skin irritation) were conducted to fill data gaps to help rank the toxicity of these formulations before progression to in vivo tests. Direct comparisons of these products through these QSARs and in vitro screens provided early data to prevent potential regrettable substitutions when selecting alternative PFAS-free firefighting foams.

The use of pesticides to control invasive species is a key component of integrated pest management plans. Traditionally, the development of new chemical controls required testing on numerous animals. In vitro cytotoxicity testing, which reduces animal use, is becoming more common to prioritize candidate toxicants and permits high-throughput testing. However, it remains unknown whether in vitro cytotoxicity values (i.e., effective concentrations or EC50 values) are representative of in vivo toxicity values (i.e., lethal concentration for 50% of organisms or LC50). DOI’s USGS has begun proof-of-concept studies on the use of fish cytotoxicity screening assays to prioritize new candidate pesticides to control nuisance fishes. Recent studies have quantified the cytotoxicity EC50 values of the fish toxicant Antimycin-A using a commercially available gill cell line (RTgill-W1) from rainbow trout (Oncorhynchus mykiss). This study utilized the procedure described in OECD Test Guideline 249 and the CellTiter-Glo 2.0 (Promega) cell viability assay for assessing RTgill-W1 cytotoxicity. In vitro to in vivo extrapolation (IVIVE) was applied to assay results and indicated that, for most valuable metrics, the ratio of in vitro to in vivo toxicity was approximately 1. These results demonstrate that toxicity as measured in rainbow trout gill cell lines is predictive of whole-animal toxicity and shows promise for the development of additional fish gill cell lines for screening of pesticide candidates.

EPA uses the in vivo acute rat oral LD50 to assign hazard classifications for acute oral toxicity for pesticides before they are approved for marketing. These classifications determine the precautionary statements placed on the pesticide label for acute human exposure. The in vivo acute rat oral LD50 is also used by EPA as a surrogate for acute oral toxicity to all mammalian wildlife. The rat LD50 is predicted by the Collaborative Acute Toxicity Modeling Suite (CATMoS), an in silico predictive tool for estimating acute oral toxicity based on molecular structure. EPA and NICEATM scientists and collaborators evaluated how well CATMoS predicted acute oral toxicity LD50 values of pesticides with available in vivo acute rat oral LD50 data (Bishop et al. 2024). The evaluation included 177 pesticides registered for use in the United States, mostly fungicides, herbicides, and insecticides. Most of the evaluated pesticides fell into the lower-toxicity Categories III and IV, and for these CATMoS predictions were found to correlate well with in vivo data, although it was felt that in some cases CATMoS estimates for toxicity might result in a more stringent label warning than animal tests would require. There were too few chemicals evaluated from the higher-toxicity Categories I and II to support a conclusion about CATMoS performance for more toxic substances. This analysis will help inform whether CATMoS can be used to estimate acute oral toxicity from pesticides to identify toxicity categories and assess risk to wildlife.

An AOP is a model that identifies the sequence of molecular and cellular events required to produce a toxic effect when an organism is exposed to a substance. As most AOPs are defined using a single or small number of species, they have a narrow taxonomic domain of applicability (tDOA). Defining the tDOA of an AOP is critical for use in regulatory decision-making for ecotoxicity, particularly when considering protection of untested species. Structural and functional conservation are two elements that can be considered when defining the tDOA. Publicly accessible bioinformatics approaches, such as the SeqAPASS tool, take advantage of existing and growing databases of protein sequence and structural information to provide lines of evidence toward structural conservation of key events and key event relationships of an AOP. It is anticipated that SeqAPASS results could readily be combined with data derived from empirical toxicity studies to provide evidence of both structural and functional conservation to define the tDOA for AOPs and elements of AOPs. Such data could be incorporated into resources such as the AOP-Wiki as lines of evidence toward biological plausibility for the tDOA. EPA scientists developed a case study describing the process of using bioinformatics to define the tDOA of an AOP using an AOP linking the activation of the nicotinic acetylcholine receptor to colony death/failure in the honeybee (Apis mellifera). Although the AOP was developed to gain a particular biological understanding relative to honeybee health, applicability to other Apis bees, as well as non-Apis bees, has yet to be defined. The EPA study demonstrates how bioinformatics can be utilized to rapidly take advantage of existing protein sequence and structural knowledge to enhance and inform the tDOA and elements of AOPs, focusing on providing evidence of structural conservation across species.

Chemical risk evaluations under TSCA are often conducted with limited test data, as TSCA does not have minimum data requirements. EPA is evaluating how this limitation can be addressed using Web-based Interspecies Correlation Estimation (Web-ICE), a model that can predict toxicity values for environmental species absent from a dataset and provide a more robust dataset to estimate toxicity thresholds. Web-ICE is an online tool that estimates acute toxicity values for aquatic and terrestrial species using surrogate species data and least-squared regressions. Each Web-ICE model represents the relationship of inherent sensitivity between a surrogate species and a predicted taxon (species, genus, or family). Web-ICE models use surrogate species sensitivity as an input to estimate the sensitivity of all available taxa. For the industrial chemicals examined so far under TSCA, Web-ICE predictions increased the number of aquatic species represented in the dataset and provided more species representation. Additionally, the use of Web-ICE predictions with empirical data to create species sensitivity distributions provided a data-driven way of accounting for uncertainties in aquatic hazard characterization. A poster describing the project (Koehrn et al.) was presented at the 2022 meeting of SETAC North America.

QSAR models may be used to assess the potential acute toxicity of chemicals in fish for pesticides or pesticide degradation products for which there are little or no toxicity data available. However, data used to develop QSARs often do not represent the broad range of pesticidal structures and modes-of-action that contribute to pesticide toxicity. EPA Office of Pesticide Programs and Office of Research and Development scientists compared the performance of three existing QSAR models with a random forest model newly developed for this purpose. The new random forest model predicted toxicity better than the existing models, likely because it was trained using only pesticide data and a targeted predictor set, and because its algorithm accounts for predictor interactions and non-linear relationships. The new random forest model is being incorporated into a graphical user interface tool with the goal of making it publicly available to web users.

EPA’s Endocrine Disruptor Screening Program (EDSP) is responsible for determining the potential for certain chemicals to cause adverse effects in humans and wildlife via endocrine pathways. One goal of the EDSP is to evaluate how broadly results can be concluded across taxa. Two EPA projects are using computational approaches to assess how potential endocrine disruptors might affect a diversity of mammalian and nonmammalian species.

EPA scientists and collaborators assessed the cross-species conservation of androgen receptor-modulated pathways by using computational analyses and systematic literature review approaches to conduct a comprehensive analysis of existing in silico, in vitro, and in vivo data (Vliet et al. 2023). Results indicate that androgen receptors are conserved across vertebrate species, which could thus be predicted to share similarly susceptibility to chemicals that interact with the human androgen receptor. This study demonstrated a framework for utilizing bioinformatics and existing data to build weight-of-evidence for cross-species extrapolation and provides a technical basis for extrapolating human androgen receptor-based data to prioritize hazard in nonmammalian vertebrate species.

Thyroid hormone system-disrupting compounds are considered potential threats for human and environmental health. Multiple AOPs for thyroid hormone system disruption are being developed in different taxa. Combining these AOPs results in a cross-species AOP network for thyroid hormone system disruption which may provide an evidence-based foundation for extrapolating these data across vertebrate species and bridging the gap between human and environmental health. A review by EPA scientists and collaborators (Haigis et al. 2023) aimed to advance the description of the taxonomic domain of applicability in the network to improve its utility for cross-species extrapolation. The study focused on molecular initiating events and adverse outcomes, evaluating both the taxa they are likely applicable to and where evidence for applicability to various taxa exists in the context of thyroid system disruption. The evaluation showed that all molecular initiating events in the AOP network are applicable to mammals. There was some evidence of structural conservation across vertebrate taxa, especially for fish and amphibians and to a lesser extent for birds. The results of this evaluation are summarized in a conceptual AOP network that helps prioritize specific parts of AOPs for a more detailed evaluation.

The Stemina devTOX quickPredict (devTOXqP) assay, an in vitro human induced pluripotent stem cell assay, assesses a chemical's potential to induce developmental toxicity. Assessments are based on a chemical’s developmental toxicity potential (dTP) concentration in the assay. A study conducted by EPA, FDA, and NIEHS scientists and collaborators applied in vitro to in vivo extrapolation (IVIVE) approaches to see if the devTOXqP assay could quantitatively predict in vivo developmental toxicity lowest-effect levels for the prototypical teratogen valproic acid and a group of comparable structures (Chang et al. 2022). Equivalent administered dose (EADs) that would lead to plasma concentrations equivalent to the in vitro dTP concentrations for valproic acid and its analogues were quantitatively similar to in vivo data from both rats and humans, where available, and the derived rank order of the chemicals was consistent with observed in vivo developmental toxicity. This study highlighted the importance of pharmacokinetic considerations when using in vitro assays and demonstrates the usefulness of the devTOXqP assay to quantitatively assess a chemical's developmental toxicity potency.

Seizures have been reported among e-cigarette users, in particular youth or young adults. FDA scientists used chemoinformatic computational models to compare chemicals documented to be present in e-cigarettes with known neuroactive compounds, with the goals of predicting blood-brain barrier penetration potential, central nervous system activity, and structural similarities (Stratford et al. 2024). The e-cigarette chemicals identified showed structural similarity to neuroactive compounds based on chemical fingerprint similarity analyses. Most chemicals studied were predicted to cross the blood-brain barrier and were also predicted to have central nervous system activity. This study showed that computer-based models can be useful to screen e-cigarette chemicals, allowing for prioritization for further possible in vitro and in vivo testing and potential early identification of central nervous system toxicity.

FDA is required to assess the environmental impact of its tobacco regulatory actions per the National Environmental Policy Act. Increases in use of waterpipe tobacco products raises concerns about environmental impacts from waterpipe waste disposal. To characterize the scope of this issue, FDA scientists compiled a comprehensive list of waterpipe wastewater chemical concentrations from literature (Termeh-Zonoozi et al. 2023). Chemicals were then selected for risk assessment by estimating persistence, bioaccumulation, and aquatic toxicity characteristics and hazardous concentration values. Of 38 chemicals with concentration data, 20 were found to be listed as harmful or potentially harmful constituents in tobacco smoke and tobacco products by FDA, and 15 are EPA hazardous waste substances. Six metals on both lists were selected for future risk assessments, as were three non-metals because of their persistence and/or toxicity. The presence of multiple hazardous compounds in waterpipe wastewater highlights the importance of awareness on the proper disposal of waterpipe wastewater in residential and retail settings. Future studies can build on the hazard characterization provided in this study through fate and transport modeling, exposure characterization and risk assessments of waterpipe wastewater chemicals.

FDA is required to assess the environmental impact of its tobacco regulatory actions per the National Environmental Policy Act. Increased use and sales of e-cigarettes raises concerns about the potential environmental impacts throughout their life cycle. However, few available research studies focus on the environmental impacts and ecotoxicity of e-cigarettes. Using a combination of available laboratory data and structure‒activity relationship models, FDA scientists compiled a list of e-liquid chemicals to be considered for future environmental impact and risk assessments (Venugopal et al. 2023). Characteristics considered included environmental persistence, bioaccumulation, and aquatic toxicity. Of the 421 unique e-liquid chemicals considered, 35 are considered hazardous constituents by EPA, 42 are considered by FDA to be harmful or potentially harmful constituents in tobacco products and smoke, and 20 were included on both lists. The study ultimately identified 81 chemicals that should be considered for future environmental impact and risk assessments, including tobacco-specific compounds, polycyclic aromatic hydrocarbons, flavors, metals, phthalates, and flame retardants. This study underscores the importance of awareness and education when handling or disposing of e-liquids/e-cigarettes and aim to inform strategies to prevent and reduce hazards from e-cigarettes.

Flavor chemicals contribute to the appeal and toxicity of e-cigarettes and other tobacco products, and the assortment of flavor chemicals available for use in tobacco products is extensive. FDA scientists used a chemistry-driven computational approach to evaluate flavor chemicals based on intrinsic hazardous structures and reactivity of chemicals (Goel et al. 2022). A library of 3,012 unique flavor chemicals was evaluated based on physicochemical properties, GHS health hazard classification, structural alerts linked to the chemical’s reactivity, instability, or toxicity, and substructures shared with known respiratory toxicants. Computational analysis of the constructed flavor library flagged 638 chemicals with GHS classified respiratory health hazards, 1,079 chemicals with at least one structural alert, and 2,297 chemicals with substructures of concern. From further analysis of a subset of 173 chemicals, four general structures with an increased potential for respiratory toxicity were identified. This study indicated that computational methods are efficient tools for hazard identification and understanding structure-toxicity relationship. With appropriate context of use and interpretation, in silico methods may provide scientific evidence to support toxicological evaluations of chemicals in or emitted from tobacco products.

Cardiovascular disease is the leading cause of death for people of most ethnicities in the United States. In a pioneering effort to minimize animal testing to evaluate chemicals for potential cardiotoxicity, NIEHS scientists applied new approach methodologies (NAMs) that blend in vitro, in chemico, and in silico methods to investigate potential cardiotoxicity of over 800 substances. These substances, characterized by widespread human exposure, included personal care product ingredients, flame retardants, herbicides, pesticides, pharmaceuticals, and industrial byproducts. A systems-based modeling workflow including PBPK models was used to transform data from molecular and cellular assays relevant to cardiovascular endpoints into human daily equivalent administered doses (EADs). The study compared these in vitro-derived EADs against both human exposure predictions and in vivo toxicological data to gauge human-relevant risks. It also integrated geospatial analyses to evaluate the compounded risks across diverse U.S. populations, spotlighting communities at disproportionate risk. Through this comprehensive approach, which merged high-throughput screening (HTS) assays, PBPK modeling, and exposure data, the project aimed to refine human health risk assessments for chemicals posing cardiovascular hazards. This endeavor marks a significant stride in transitioning from molecular insights to actionable public health interventions, striving to replace animal testing with human biology-based strategies for chemical safety evaluation. A paper describing this work will be submitted for publication in 2024.

Animal models are currently used to predict whether a chemical might cause human neurotoxicity, which can lead to adverse effects such as seizures. However, research suggests that animal models have limited reliability, particularly in predicting drug safety for the central nervous system. The prediction of whether a drug might cause seizures, specifically, is unreliable due to a significant failure rate of novel drugs in human clinical trials caused by unforeseen toxicity, revealing the inadequacy of these models. To address this issue, a collaboration between NIEHS and industry was initiated to identify potential biological targets associated with seizures. By combining targets from established AOPs and drug discovery databases, a seizure-specific AOP network was generated. Resources including NICEATM’s Integrated Chemical Environment and the European Molecular Biology Laboratory’s ChEMBL database were employed to identify new approach methodologies (NAMs) that could measure identified targets. The search identified both compounds likely to induce seizure and compounds that have tested negative for seizure effects. A proof-of-concept evaluation to assess in vitro mechanistic assay data availability across the seizure-relevant targets was conducted. Through this comprehensive approach, the current landscape of seizure risk-informative testing assay availability was assessed, laying the groundwork for future predictive modeling efforts with the ultimate goal of enhancing the ability to anticipate and mitigate seizure-related adverse effects. This project was presented at the 2023 annual meeting of the American College of Toxicology (Behl et al.), and a manuscript is in preparation to be submitted in 2024.