Citation:

Miller, J., J. Chua, J. Heng, K. Houck, D. Zink, AND L. Loo. Nephrotoxicity Prediction and Structure-Activity Analyses of ToxCast Chemicals Based on High-Throughput Image-Based Phenotypic Profiling. 11th World Congress on Alternatives and Animal Use in the Life Sciences (WC11), Maastricht, N/A, NETHERLANDS, August 23 - September 03, 2021.

Impact/Purpose:

Abstract submitted to the 11th World Congress on Alternatives and Animal Use in the Life Sciences (WC11) held in August 2021. Application of a predictive model for nephrotoxicity using high-content, phenotypic imaging to a set of 328 ToxCast chemicals.

Description:

Kidneys are major targets for environmental chemicals. The identification of chemical families or moieties associated with nephrotoxicity may help flag potential chemical hazards. However, the lack of kidney-specific bioactivity data for many high-exposure chemicals have hampered the efforts to systematically identify toxicity-relevant substructures. We have previously developed highly predictive in vitro proximal tubule cell (PTC) toxicity models using high-throughput, imaging-based phenotypic profiling (HIPPTox) [1]. The models used the human kidney cell line HK-2 and primary human PTCs (HPTC), validated with 42 reference chemicals with known PTC- or non-PTC toxicities. Here we present preliminary results on the reproducibility of the PTC toxicity models and their applications to 328 chemicals from the US Environmental Protection Agency’s (EPA) ToxCast library selected for predicted high human exposure but lacking extensive safety data. We confirmed that our models are reproducible, and the 3 or 5 most predictive phenotypic features identified from HIPPTox can achieve 82% and 81% balanced accuracy, respectively, in predicting reference chemicals’ PTC toxicity. Then, we applied the models to the 328 chemicals and 180 of them were predicted to be PTC-toxic, 144 to be non-PTC-toxic, and 4 were inconclusive. Finally, we performed an enrichment analysis of ToxPrint substructures [2] in chemicals predicted to be PTC-toxic, identifying several chemical moeities associated with PTC toxicity. They include 5-member heterocycles present in conazoles, which were found to be nephrotoxic in previous in vivo animal studies. Our work demonstrates the reproducibility and feasibility of applying our in vitro PTC models based on phenotypic profiling to a larger and more diverse chemical set, which in turn allows us to identify chemical structures that may be associated with nephrotoxicity. We are conducting mechanistic studies on some of the chemical moieties associated with PTC toxicity; once validated, these may be used as structural alerts for nephrotoxicity, and be applied to the entire ToxCast chemical library. The work presented here does not represent the official position of the U.S. EPA.

Record Details: