Integrative genomics of gene and metabolic regulation by estrogen receptors α and β, and their coregulators

Abstract

The closely related transcription factors (TFs), estrogen receptors ERα and ERβ, regulate divergent gene expression programs and proliferative outcomes in breast cancer. Utilizing breast cancer cells with ERα, ERβ, or both receptors as a model system to define the basis for differing response specification by related TFs, we show that these TFs and their key coregulators, SRC3 and RIP140, generate overlapping as well as unique chromatin-binding and transcription-regulating modules. Cistrome and transcriptome analyses and the use of clustering algorithms delineated 11 clusters representing different chromatin-bound receptor and coregulator assemblies that could be functionally associated through enrichment analysis with distinct patterns of gene regulation and preferential coregulator usage, RIP140 with ERβ and SRC3 with ERα. The receptors modified each other's transcriptional effect, and ERβ countered the proliferative drive of ERα through several novel mechanisms associated with specific binding-site clusters. Our findings delineate distinct TF-coregulator assemblies that function as control nodes, specifying precise patterns of gene regulation, proliferation, and metabolism, as exemplified by two of the most important nuclear hormone receptors in human breast cancer.

Figure 1

Genome-wide analysis of ERα, ERβ, SRC3 and RIP140 chromatin binding by ChIP-seq. (A) Experimental design for ChIP-seq, gene microarray, and clustering analyses. Veh, vehicle; E2, 10 nM 17β-estradiol. (B) Comparison of ERα- and ERβ-binding sites in the different cell backgrounds. (C) Comparison of RIP140-binding sites in the different cell backgrounds. (D) Comparison of SRC3-binding sites in the different cell backgrounds. (E) Comparison of the number of ER and coregulator (SRC3, RIP140)-binding sites in ERα cells or ERβ cells. (F) Overlap of RIP140- or SRC3-binding sites with ERs in ERα/ERβ cells.

Figure 2

Categorization of binding sites into clusters based on ERα, ERβ, SRC3, and RIP140 cistromes. (A) Clustering of the binding sites. seqMINER software was used for clustering based on colocalization of different factors in different cell backgrounds within a 300-bp window in both directions. (B) Conservation of binding sites among vertebrates: cistrome conservation tool was used to compare conservation of binding sites from different clusters among vertebrates. (C) Genomic location of binding sites: web-based CEAS tool was used for identifying genomic location of binding sites from each cluster.

Figure 3

Patterns of E2-regulated gene expression in ERα, ERα/ERβ, and ERβ cells. (A) Comparison of number of 17β-estradiol (E2) regulated genes after 4 h of ligand treatment in different cell backgrounds. Genespring Venn Diagram tool was used to compare overlap between genes regulated ⩾1.8-fold with FDR of 0.01 in each background. (B) Comparison of number of E2-regulated genes after 24 h of ligand treatment in the different cell backgrounds. (C) Scatter plots for gene regulations in the different cell backgrounds: scatter plots of E2-regulated genes in each cell background were generated using the Genespring scatterplot tool. The top three panels show how genes that are regulated by E2 in ERα cells are also being regulated in ERα/ERβ cells and in ERβ cells. The middle three panels show how genes that are regulated by E2 in ERα/ERβ cells are also being regulated in ERα cells and in ERβ cells. The bottom three panels show how genes that are regulated by E2 in ERβ cells are also being regulated in ERα cells and in ERα/ERβ cells. E2-upregulated genes are shown in red and E2-downregulated genes are shown in blue.

Figure 4

Association of ER-mediated gene expression with receptor and coregulator cistromes. (A) Comparison of number of genes with at least one binding site within 20 kb and E2-upregulated or -downregulated genes. (B) Association of gene regulations in different cell backgrounds with binding site clusters: enrichment of binding-site clusters (C1–C11) associated with upregulated and downregulated genes in each cell background after 4 or 24 h of estradiol treatment is shown. Level of enrichment is designated by the color scale. Cluster 3.0 was used to cluster the data. Data were visualized using Java Treeview. (C) Clustering of genes that are regulated by E2 treatment and genes that have binding sites from different clusters: BED files for E2-regulated genes were generated using Genespring. The seqMINER enrichment based method was used to calculate the density array for the E2-regulated gene BED files and BED files obtained for each cluster from Figure 2A. Density array was clustered using Cluster 3.0 and data was visualized using Java Treeview. (D) The pS2 gene and associated binding sites for ERα, ERβ, RIP140, and SRC3 in the three cell backgrounds. (E) The pS2 gene regulation by E2 (10⁻¹³ to 10⁻⁷ M) in the different cell backgrounds. (F) The pS2 gene ChIP for ERα, ERβ, and coregulators in cells treated with control vehicle (0.01% ethanol) or 10 nM E2. (G) The PgR gene and associated binding sites in cells treated with control vehicle (0.01% ethanol) or 10 nM E2. (H) The PgR gene regulation by E2 in the different cell backgrounds. (I) The PgR gene ChIP for ERα, ERβ, and coregulators in cells treated with control vehicle (0.01% ethanol) or 10 nM E2. (J) The OTUB2 gene and associated binding sites in cells treated with control vehicle (0.01% ethanol) or 10 nM E2. (K) The OTUB2 gene regulation by E2 in the different cell backgrounds. (L) The OTUB2 gene ChIP for ERα, ERβ, and coregulators in cells treated with control vehicle (0.01% ethanol) or 10 nM E2.

Figure 5

Pathway analysis of the E2-regulated genes associated with the ERα and ERβ-binding site clusters. (A) Clustering of gene ontologies that are differentially enriched in ERα cells versus ERβ cells: ClueGO plugin of Cytoscape was used to identify differentially regulated GO terms between ERα and ERβ cells. (B) Expanded view of M-phase-associated clusters. (C) Expanded view of additional M-phase-associated clusters.

Figure 6

ERα supports strong E2-mediated cell proliferation; ERβ reduces the proliferative stimulation of ERα in ERα/ERβ cells, and ERβ by itself supports very little if any hormone-enhanced proliferation. (A) Cell proliferation: ERα, ERα/ERβ, and ERβ cells in six-well plates were treated with Veh (0.01% EtOH) or 10 nM E2 (day 0). Treatment was repeated at day 2. Cell number was monitored by MTS assay. Treatment groups were monitored in triplicate in two separate experiments. (B) FACS analysis: cell cycle stages were analyzed by flow cytometry using BD-FACS Canto. Cells were fixed in 70% ethanol, stained for 30 min with 20 μg/ml PI in Triton-X in the presence of DNAse-free RNAse A, and PI staining was measured. (C) Gene expression of four M-phase-related genes in cells treated with control vehicle or 10 nM E2 for 24 h. mRNA level in vehicle-treated ERα cells was set at 1.

Figure 7

Impact of ERβ on ERK1/2 activation, RIP140 gene expression, and adipogenesis, and an ERβ and RIP140 gene signature and its association with overall survival and recurrence status of breast cancer patients. (A) Modulation of MAPK pathway activation by ERβ. MCF-7ERα cells were infected with the indicated amounts of ERβ-adenovirus (multiplicity of infection, MOI) and were treated with 10 nM E2 for 15 min. ERK1/2 phosphorylation was monitored using pMAPK antibody (9101; Cell Signaling). Total ERK2 (as loading control), and ERα and ERβ were also monitored. C represents control cells infected with 40 MOI of AdGal and no AdERβ. (B) Analysis of RIP140 mRNA levels in ERα, ERα/ERβ, and ERβ cells treated with control vehicle or 10 nM E2 for 24 h. (C) Regulation of genes involved in adipogenesis in the three cell types treated with vehicle or E2 for 24 h. (D) Total triglyceride levels in cells were measured. Assays were performed twice in triplicate. (E) Patient survival analysis using our 20-gene ERβ and RIP140 signature and van de Vijver et al (2002), Kao et al (2011), Finak et al (2008) and Esserman et al (2012) breast cancer data sets. All the data sets were obtained from the Oncomine database. (F) Hierarchical clustering of our signature genes using expression values obtained from the van de Vijver et al (2002), data set. (G) Comparison of average expression levels of the signature genes based on survival and recurrence data of patients.

Figure 8

Model depicting modulation of gene expression, metabolism, and proliferative properties of breast cancer cells by ERα and ERβ, based on findings in this report. The three cell contexts are illustrated, ERα at top, ERα/ERβ in middle (as both homo and heterodimers), and ERβ at bottom. Sites are shown with both SRC3 and RIP140 (right), and RIP140 and other factors, but not SRC3 (left). Representative associations between binding site clusters and cellular regulations are illustrated. These include antiproliferative actions of ERβ through utilization of novel mechanisms, including modulation of transcription by ERα, changes in MAPK signaling, and alteration in adipogenesis in breast cancer cells.