High-throughput transcriptomics analysis of equipotent and human relevant mixtures of BPA alternatives reveal additive effects in vitro

Abstract

While many jurisdictions have phased out use of bisphenol A (BPA), there is increasing exposure to mixtures of BPA alternatives. Like BPA, some alternatives perturb nuclear hormone receptors and are endocrine disruptors. We used high-throughput transcriptomics (HTTr) to evaluate the potency and modes of action of seven mixtures of BPA alternatives and their 12 individual components in breast cancer cells. Our aim was to explore whether alternatives present in mixtures act additively. MCF-7 cells were exposed to chemicals (0.001-50 µM) for 48 h and gene expression analysis was used to measure global and estrogen receptor alpha (ERα)-specific transcriptomic changes. Transcriptomic points of departure (tPODs) were derived using benchmark concentration (BMC) modelling. We first identified concentrations at which global transcriptional activity was robustly altered. Then, we applied a ERα transcriptomic biomarker to identify ERα agonists and predict ERα activation tPODs. We employed mixtures modeling to predict potency of BPA alternatives and test for additive effects in vitro. Ingenuity pathway analysis (IPA; Qiagen) was used to identify upstream regulators and canonical pathways from genes fitting BMCs. BPAF was the most potent individual chemical tested overall, followed by BPA and BPC. All seven mixtures had additive effects across all tPODs modeled. The ERα transcriptomic biomarker classified all mixtures as ERα activators along with several BPA alternatives. All mixtures and most individual components perturbed similar upstream regulators and pathways, suggesting common modes of action. These data support the value of HTTr in identifying additive effects and toxicological potency of mixtures in vitro.

Keywords: Endocrine disruptor; Mixtures; Potency; Regulatory toxicology; Transcriptomics.

Fig. 1

Chemical structure of bisphenol A (BPA) and alternative chemicals

Fig. 2

Gene accumulation plot; data were prefiltered using the Williams trend test (p < 0.05) and an absolute fold-change filter of ≥ 1.5, and post filtered with the following settings in BMDExpress v3: best BMD/BMDL < 20, best BMDU/BMDL < 40, and best fitPvalue >  = 0.1

Fig. 3

Comparison of transcriptomic points of departure (tPODs); tPODs were derived by prefiltering data using the Williams trend test (p < 0.05) and an absolute fold-change filter of ≥ 1.5 and postfiltered with the following settings in BMDExpress v3: best BMD/BMDL < 20, Best BMDU/BMDL < 40, and best fitPvalue >  = 0.1; tPODs (shown in µM) representing the 25th rank ordered gene benchmark concentration (BMC) and the 5th percentile gene BMC for the estrogen receptor alpha (ERα) biomarker gene set are shown; BMCL and BMCU are used for the lower and upper bounds, respectively; predictions based on additivity of individual components are shown as diamonds for the mixtures; most chemicals are shown in decreasing order of potency based on tPODs from the 25th gene BMC, followed by the mixtures

Fig. 4

Ingenuity pathway analysis; a list of genes fitting benchmark concentration models, the Williams trend test p value, and the maximum fold-change were imported into IPA; filters were set to Z score ≥ 2.0 and p value < 0.05; orange denotes predicted activation and blue predicted inactivation (dots signify Z score < 2); a top 20 upstream regulators; b top 20 canonical pathways