ICCVAM Biennial Report 2024-2025

Gene expression profiling with alternative toxicological cell systems is a valuable tool for exploring mechanisms of chemical interactions in biological systems. However, interpretation of these data for decision-making is challenging. EPA and NIEHS scientists are addressing this gap by building interpretable bridges among environmental chemicals and specific molecular targets associated with biological response such as nuclear receptors, enzymes, and transporters. Over 300 reference chemicals have been identified and procured to build robust transcriptomic datasets in various cell systems, including the U-2-OS human osteosarcoma cell line and differentiated cultures of HepaRG and 3D microtissues of primary human hepatocytes, using the TempO-Seq™ technology for next-generation sequencing of cell lysates. Experiments covered a wide range of exposures to capture the initial stages of biological responses to environmental chemicals, using transcriptomic patterns to define similarity and help infer mechanisms for data-poor chemicals. This efficient strategy has the potential to greatly increase the mechanistic understanding of adaptive responses to environmental chemicals at the gene expression level and inform next steps in toxicology research investigations. The research teams have completed data generation for all three cell systems, and manuscripts for each cell system are in preparation. Furthermore, these reference chemicals were shared with colleagues at the University of Leiden to add renal proximal tubule epithelial cells to the list of cell types. A manuscript by the researchers involved in this project has been prepared for submission in 2026.

High-throughput screening (HTS) programs such as Tox21 gather data on chemical bioactivity from a variety of targets and activities to prioritize assays for further testing or better understand mechanisms of toxicity. However, interpreting such data is complicated by nonspecific chemical reactivity. To help address this problem, specific HTS assays have been developed to measure autofluorescence, and data from these assays have been used to train machine learning models to predict reactivity.

The fluorescence-based MSTI ((E)-2-(4-mercaptostyryl)-1,3,3-trimethyl-3H-indol-1-ium) HTS assay identifies thiol-reactive compounds. EPA and NCATS scientists used this assay that identifies thiol-reactive compounds to screen 7,872 unique chemicals in the Tox21 10K chemical library (Patlewicz et al. 2023). Active chemicals were compared to profiling outcomes using structural alerts encoding electrophilic information. A random forest classification model based on chemical fingerprints was then developed to predict assay outcomes. The classification model shows promise as a tool to screen untested chemicals for their potential electrophilic reactivity based solely on chemical structure. MSTI assay data were subsequently evaluated for their utility in predicting compound promiscuity and cytotoxicity by comparing the activity of the Tox21 10K compounds in the MSTI assay to their activities across the entire Tox21 assay panel (Kruger et al. 2024). The predictive performance showed a sensitivity of 55% and specificity of 97%; chemical features such as halides and carboxamides resulted in false negatives and false positives, respectively. While this high-throughput assay allows researchers to identify whether a compound will be cytotoxic and/or promiscuous in an efficient manner, this study also reveals potential challenges in its use.

High-throughput screening (HTS) assays that have been used in the Tox21 testing program to date generally lack characteristics of metabolic activity found in living systems, which can potentially increase or decrease the toxicity evaluation of chemicals. As a result, HTS results may not accurately reflect in vivo activity. Scientists at EPA, NCATS, and NIEHS used several approaches to address this problem: adding human or rat liver microsomes into the existing assays (Ooka et al. 2022), transfecting cells with mRNAs encoding human metabolic enzymes, or using metabolically capable human HepaRG cells. The retrofitted assays were then used to screen the Tox21 10K chemical library to identify chemicals that are either bioactivated or detoxified by metabolic activity. Chemicals that exhibit metabolism-mediated shifts in bioactivity were further screened to identify the responsible metabolic enzymes, including the primary cytochrome P450 (CYP) drug-metabolizing enzymes CYP1A2, 3A4, 2C9, 2C19, and 2D6. The screening data were analyzed by grouping chemicals based on structural similarity, which enabled the identification of specific and nonspecific CYP inhibition patterns (Sakamuru et al. 2025). This information ultimately supports the development of predictive models for CYP inhibitory activities.

Retrofitted assays are also being used to explore how metabolism might affect interaction of chemicals with the androgen receptor pathway. Rat liver microsomes—small, membrane-bound vesicles formed from fragments of the endoplasmic reticulum—were incorporated into androgen pathway assays for screening the Millipore–Sigma Library of Pharmacologically Active Compounds (Lynch et al. 2025). Known androgenically active compounds, as well as at least one novel chemical, were identified using this newly optimized and physiologically relevant model. The 10K compound collection was screened during 2024 and 2025 to identify novel compounds requiring metabolism for androgen receptor antagonist activity; follow-up studies are ongoing in 2026.

Over the past decade, there has been a worldwide trend toward earlier breast development in girls. The rapid pace of this trend suggests the involvement of environmental factors. Scientists at NCATS and NIEHS examined the potential effects of endocrine-disrupting chemicals on important components of the biochemical pathway responsible for pubertal timing (Yang et al. 2024). In these studies, chemicals in the Tox21 10K collection were tested in human cell-based assays that measure activation or inhibition of the gonadotropin-releasing hormone receptor and kisspeptin receptor. These receptors are expressed in the hypothalamus and are important players in the control of pubertal timing. The 58 chemicals that exhibited activity included chemicals known to be active in these pathways and chemicals for which this activity was novel; these were characterized further in other cell-based assays and molecular docking studies. Further studies are necessary to investigate whether exposure to these chemicals might play a role in the increasing prevalence of early puberty in girls.

The developing child is vulnerable to genetic, pharmacological, or chemical disruption of the retinoid biochemical pathway, especially during early growth and differentiation of embryonic tissues. Susceptibility of this pathway to chemical disruption is an important regulatory concern for developmental and reproductive hazard identification. In this study, which involves all four Tox21 participating offices, Tox21 data are being mined and modeled to identify potential retinoid pathway disruptors. Results from a dozen Tox21 molecular targets mapping to retinoid pathway targets identified over 100 structurally diverse chemicals with relevant bioactivity. Computational tools and approaches are now being built to integrate these data with embryological knowledge and construct data-driven models for developmental hazard prediction. Earlier publications from the project:

• Identified candidate reference chemicals for use in in vitro assays relevant to the pathway (Baker et al. 2022).
• Described a computational model for fetal skeletal defects potentially linked to disruption of retinoic acid signaling (Pierro et al. 2022).
• Applied in vitro to in vivo extrapolation to characterize potency of chemicals interacting with the pathway (Chang et al. 2022).
• Developed and validated a high-throughput assay measuring activity of the metabolic enzyme CYP26A1 (Sakamuru et al. 2024); this assay was then used to first screen a library of pharmacologically active chemicals and then the Tox21 10K compound collection. Chemicals with inhibitory activity were selected for follow-up assays planned for 2026.

More recently, a high-throughput cytokine detection platform using human induced pluripotent stem cell-derived microglia was developed and validated to identify potential neurotoxicants (Yang et al. 2025) by testing nine environmental chemicals with evidence of cytokine-inducing potential in other in vitro assays.

To date, Tox21 high-throughput screening (HTS) assays have focused primarily on selected nuclear receptor and stress response pathways. This relatively limited focus suggests that activity in other toxicity pathways has not been adequately assessed; it is likely that some unexplored pathways relate to unanticipated adverse drug effects. Therefore, expanding the coverage of biological responses by adding assays that probe underrepresented pathways in the current Tox21 assay portfolio may improve the predictivity of Tox21 data. Scientists at FDA, NCATS, and NIEHS are systematically identifying these underrepresented pathways in a data-driven approach and nominating assays for development and Tox21 chemical screening. The data generated will be used to build models for human toxicity prediction, focusing on common adverse drug effects such as drug-induced liver injury and cardiotoxicity. The initial survey identified targets and pathways of interest, including CYP450 metabolic enzymes and G protein-coupled receptors. In 2022 and 2023, HTS assays for the proposed new targets and pathways were screened against the Tox21 10K chemical library. Predictive models for hepatotoxicity and cardiotoxicity were built using chemical structure and Tox21 in vitro assay data (Ye et al. 2022). The impact of the newly generated assay data on improving model performance was evaluated, and predictions from the models were validated in relevant assays. The expanded Tox21 assay panel was found to significantly enhance the capacity of assay data for in vivo toxicity prediction, underscoring the importance of a diverse and comprehensive in vitro assay portfolio in advancing safety assessment (Xu et al. 2025).