Lessons learned from evaluating defined chemical mixtures in a high throughput estrogen receptor assay system.

Abstract

In this paper, we provide a proof of concept evaluating the utility of the U.S. Tox21 high-throughput screening approach to assess the hazard of chemical mixtures using two estrogen receptor assays. A subset of chemicals identified in Phase I of the Tox21 program as active in the estrogen receptor (ER) agonist assay were used to design mixtures for testing in Phase II. Individual chemicals and mixtures were evaluated in two cell-based estrogen receptor alpha (ERα) activation assays: One incorporating a transfected ligand-binding domain in an ERα β-lactamase reporter cell line (ER-bla) and the full-length endogenous receptor in the MCF7 cell line with a luciferase reporter gene (ER-luc). Concentration-response data from individual chemicals were used to predict the joint effect based on mixtures modeling methods and were compared to observed mixtures data to assess model fit. The models tended to overpredict mixture responses in the ER-bla assay, while predictions were closer to observed responses in the ER-luc assay, indicating that a full-length endogenous estrogen receptor is a preferred model for high-throughput mixture analysis. Lessons learned from this research include the importance of analyzing the individual chemicals used for predictions and the mixtures in the same experimental paradigm to minimize variation, developing methods for imputing missing values from incomplete concentration-response curves, and establishing criteria to determine when inactive chemicals should be omitted from mixture predictions.