Editor's Highlight: Transcriptome Profiling Reveals Bisphenol A Alternatives Activate Estrogen Receptor Alpha in Human Breast Cancer Cells

Abstract

Plasticizers with estrogenic activity, such as bisphenol A (BPA), have potential adverse health effects in humans. Due to mounting evidence of these health effects, BPA is being phased out and replaced by other bisphenol variants in "BPA-free" products. We have compared estrogenic activity of BPA with 6 bisphenol analogues [bisphenol S (BPS); bisphenol F (BPF); bisphenol AP (BPAP); bisphenol AF (BPAF); bisphenol Z (BPZ); bisphenol B (BPB)] in 3 human breast cancer cell lines. Estrogenicity was assessed (10-11-10-4 M) by cell growth in an estrogen receptor (ER)-mediated cell proliferation assay, and by the induction of estrogen response element-mediated transcription in a luciferase assay. BPAF was the most potent bisphenol, followed by BPB > BPZ ∼ BPA > BPF ∼ BPAP > BPS. The addition of ICI 182,780 antagonized the activation of ERs. Data mining of ToxCast high-throughput screening assays confirm our results but also show divergence in the sensitivities of the assays. Gene expression profiles were determined in MCF-7 cells by microarray analysis. The comparison of transcriptome profile alterations resulting from BPA alternatives with an ERα gene expression biomarker further indicates that all BPA alternatives act as ERα agonists in MCF-7 cells. These results were confirmed by Illumina-based RNA sequencing. In conclusion, BPA alternatives are not necessarily less estrogenic than BPA in human breast cancer cells. BPAF, BPB, and BPZ were more estrogenic than BPA. These findings point to the importance of better understanding the risk of adverse effects from exposure to BPA alternatives, including hormone-dependent breast cancer.

Keywords: bisphenol; cell culture; endocrine; estrogens; toxicogenomics.

Figure 1

Molecular structures of BPA and bisphenol alternatives in comparison to the natural hormone 17β-estradiol.

Figure 2

BPA alternatives can effectively substitute for estradiol in promoting growth through ERs of human breast cancer cells. (A) Proliferative effect of BPA and bisphenol alternatives on mammary cells in an E-Screen bioassay. After 24 h of steroid hormone starvation cells were treated for 6 days with the test compounds. Numbers of cells was measured by an MTT colorimetric assay. Results are expressed as proliferative effect percentage relative to the proliferation of cells under hormone-free conditions. Data are the mean ±SE of 3 independent experiments, with each performed in triplicate. (B) BPA alternatives stimulate ERE-mediated transcription. After 24 h of steroid hormone starvation, T47D-Kluc cells were treated with the test compounds for 24 h. Cells were then lysed and subjected to a bioluminescence luciferase reporter gene assay. (C) Luciferase assays of T47D-Kluc cells treated with BPA alternatives in the absence (black) or presence (red) of ICI 182,780, an ER antagonist. Results show that ICI 182,780 represses ERE-mediated transcription induced by BPA alternatives.

Figure 3

Variation in the transcriptome profiles of MCF-7 cells exposed to BPA alternatives and 17β-estradiol. (A) PCA analysis of gene expression profiles showed a clear separation between the negative control and the BPA alternatives. The different colors indicate different treatments. (B) Expression of the 1000 genes presenting the most pronounced variations between the different samples was evaluated according to a one-way ANOVA. Hierarchical clustering revealed different patterns in gene expression between the BPA alternatives and 17β-estradiol-treated cells.

Figure 4

Alteration of MCF-7 transcriptome profiles caused by BPA alternatives bear the signature of ER activation. Transcriptome analysis of MCF-7 cells treated with BPA and alternative bisphenols reflect cell cycle changes (A), response to hormone (B), as well as an association to breast cancer (C). The P-values are determined by hypergeometric calculation and adjusted using a Benjamini and Hochberg approach. Ratios show the total number of network objects belonging to each term in comparison to those disturbed by the treatments. The total of network objects (ie, all DEG) recognized by Metacore are indicated in parentheses (D). Overrepresentation of ER binding motifs in the promoters of the differentially expressed genes. The analysis conducted with the transcription factor analysis tool of Metacore. A total of 1262 ER binding sites is found in the 42 909 protein-based objects in the Metacore background list. A, number of targets in the activated dataset regulated by ESR1; E, expected mean of hypergeometric distribution; P-value calculated using hypergeometric distribution. (E) A gene expression biomarker confirms that BPA and bisphenol alternatives are ER agonists. Lists of statistically significant genes from MCF-7 cells treated with BPA and bisphenol alternatives or the natural hormone 17β-estradiol were examined against an ER gene expression biomarker signature consisting of 46 genes. The heat map shows the expression of genes in the biomarker after exposure to the indicated compound. Fold-change values for the ER biomarker are the average across 7 agonist treatments. (F) Bar plots showing the significance of the correlation by their −log10 P-values. Classification of activation or suppression required P < .0001. The number of genes overlapping the ER biomarker is indicated at the top of each bar.

Figure 5

Comparison of RNA-Seq and microarray platforms in determining endocrine disrupting effects of BPA and bisphenol alternatives. RNA extracted from MCF-7 cells was subjected to a full transcriptome profiling using the Illumina RNA sequencing or the microarray technology under similar conditions. (A) Pearson correlation coefficients between the RNA-Seq and microarray data. The fold changes in gene function having an altered expression by the two methods are presented. (B) Venn diagrams showing the number of genes uniquely or commonly disturbed. (C) Gene ontology analysis of terms associated with MCF-7 hormone induced proliferation in the transcriptome profiles obtained by RNA-Seq or microarray analysis. Ratios are showing the total number of network objects belonging to each term in comparison to those disturbed by the treatments. The total of network objects (ie, all DEG) recognized by Metacore are indicated in parentheses. (D) Heat map of genes whose expression was statistically significantly altered examined against an ER gene expression biomarker. (E) Bar plots showing the significance of the correlation by their −log10 P-values.