Predicting the effects of per- and polyfluoroalkyl substance mixtures on peroxisome proliferator-activated receptor...

Human exposure to per- and polyfluoroalkyl substances (PFAS) is ubiquitous, with mixtures of PFAS detected in drinking water, food, household dust, and other exposure sources. Animal toxicity studies and human epidemiology indicate that PFAS may act through shared mechanisms including activation of peroxisome proliferator activated receptor α (PPARα). However, the effect of PFAS mixtures on human relevant molecular initiating events remains an important data gap in the PFAS literature. Here, we tested the ability of modeling approaches to predict the effect of diverse PPARα ligands on receptor activity using Cos7 cells transiently transfected with a full length human PPARα (hPPARα) expression construct and a peroxisome proliferator response element-driven luciferase reporter. Cells were treated for 24 h with two full hPPARα agonists (pemafibrate and GW7647), a full and a partial hPPARα agonist (pemafibrate and mono(2-ethylhexyl) phthalate), or a full hPPARα agonist and a competitive antagonist (pemafibrate and GW6471). Receptor activity was modeled with three additive approaches: effect summation, relative potency factors (RPF), and generalized concentration addition (GCA). While RPF and GCA accurately predicted activity for mixtures of full hPPARα agonists, only GCA predicted activity for full and partial hPPARα agonists and a full agonist and antagonist. We then generated concentration response curves for seven PFAS, which were well-fit with three-parameter Hill functions. The four perfluorinated carboxylic acids (PFCA) tended to act as full hPPARα agonists while the three perfluorinated sulfonic acids (PFSA) tended to act as partial agonists that varied in efficacy between 28-67 % of the full agonist, positive control level. GCA and RPF performed equally well at predicting the effects of mixtures with three PFCAs, but only GCA predicted experimental activity with mixtures of PFSAs and a mixture of PFCAs and PFSAs at ratios found in the general population. We conclude that of the three approaches, GCA most accurately models the effect of PFAS mixtures on hPPARα activity in vitro. Understanding the differences in efficacy with which PFAS activate hPPARα is essential for accurately predicting the effects of PFAS mixtures. As PFAS can activate multiple nuclear receptors, future analyses should examine mixtures effects in intact cells where multiple molecular initiating events contribute to proximate effects and functional changes.