Limits...
Global analysis of estrogen receptor beta binding to breast cancer cell genome reveals an extensive interplay with estrogen receptor alpha for target gene regulation.

Grober OM, Mutarelli M, Giurato G, Ravo M, Cicatiello L, De Filippo MR, Ferraro L, Nassa G, Papa MF, Paris O, Tarallo R, Luo S, Schroth GP, Benes V, Weisz A - BMC Genomics (2011)

Bottom Line: Expression of full-length ERβ in hormone-responsive, ERα-positive MCF-7 cells resulted in a marked reduction in cell proliferation in response to estrogen and marked effects on the cell transcriptome.Of 921 genes differentially regulated by estrogen in ERβ+ vs ERβ- cells, 424 showed one or more ERβ site within 10 kb.ERβ binding in close proximity of several miRNA genes and in the mitochondrial genome, suggests the possible involvement of this receptor in small non-coding RNA biogenesis and mitochondrial genome functions.

View Article: PubMed Central - HTML - PubMed

Affiliation: Department of General Pathology, Second University of Naples, vico L, De Crecchio 7, 80138 Napoli, Italy.

ABSTRACT

Background: Estrogen receptors alpha (ERα) and beta (ERβ) are transcription factors (TFs) that mediate estrogen signaling and define the hormone-responsive phenotype of breast cancer (BC). The two receptors can be found co-expressed and play specific, often opposite, roles, with ERβ being able to modulate the effects of ERα on gene transcription and cell proliferation. ERβ is frequently lost in BC, where its presence generally correlates with a better prognosis of the disease. The identification of the genomic targets of ERβ in hormone-responsive BC cells is thus a critical step to elucidate the roles of this receptor in estrogen signaling and tumor cell biology.

Results: Expression of full-length ERβ in hormone-responsive, ERα-positive MCF-7 cells resulted in a marked reduction in cell proliferation in response to estrogen and marked effects on the cell transcriptome. By ChIP-Seq we identified 9702 ERβ and 6024 ERα binding sites in estrogen-stimulated cells, comprising sites occupied by either ERβ, ERα or both ER subtypes. A search for TF binding matrices revealed that the majority of the binding sites identified comprise one or more Estrogen Response Element and the remaining show binding matrixes for other TFs known to mediate ER interaction with chromatin by tethering, including AP2, E2F and SP1. Of 921 genes differentially regulated by estrogen in ERβ+ vs ERβ- cells, 424 showed one or more ERβ site within 10 kb. These putative primary ERβ target genes control cell proliferation, death, differentiation, motility and adhesion, signal transduction and transcription, key cellular processes that might explain the biological and clinical phenotype of tumors expressing this ER subtype. ERβ binding in close proximity of several miRNA genes and in the mitochondrial genome, suggests the possible involvement of this receptor in small non-coding RNA biogenesis and mitochondrial genome functions.

Conclusions: Results indicate that the vast majority of the genomic targets of ERβ can bind also ERα, suggesting that the overall action of ERβ on the genome of hormone-responsive BC cells depends mainly on the relative concentration of both ERs in the cell.

Show MeSH

Related in: MedlinePlus

Sequence analysis of ERα, ERβ and ERα+ERβ binding sites. (A) Venn diagram showing a summary of ERα and ERβ binding sites identified in TAP-ERβ cells by ChIP-Seq. (B) Classification of ERα and β binding sites based on the presence of a perfectly or imperfectly palindromic Estrogen Response Element (ERE, green), an ERE hemipalindrome (hERE, blue) or no ERE (none, red). (C) ERE motif matrices identified in each of the three ER binding regions indicated (left), classification of the binding sites belonging to each region according to the presence of ERE (center) and grid summarizing the results of TFBS matrix enrichment (overrepresentation) analyses performed on the binding sites groups indicated (right). Z-Score cut-off was 3.0 and only TFBSs showing an over-representation score ≥4.0 in at least one of the ERE- (none) binding site groups. Light grey cells indicate a Z-Score <3.0 while dark grey cells indicate absence of the matrix.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3025958&req=5

Figure 3: Sequence analysis of ERα, ERβ and ERα+ERβ binding sites. (A) Venn diagram showing a summary of ERα and ERβ binding sites identified in TAP-ERβ cells by ChIP-Seq. (B) Classification of ERα and β binding sites based on the presence of a perfectly or imperfectly palindromic Estrogen Response Element (ERE, green), an ERE hemipalindrome (hERE, blue) or no ERE (none, red). (C) ERE motif matrices identified in each of the three ER binding regions indicated (left), classification of the binding sites belonging to each region according to the presence of ERE (center) and grid summarizing the results of TFBS matrix enrichment (overrepresentation) analyses performed on the binding sites groups indicated (right). Z-Score cut-off was 3.0 and only TFBSs showing an over-representation score ≥4.0 in at least one of the ERE- (none) binding site groups. Light grey cells indicate a Z-Score <3.0 while dark grey cells indicate absence of the matrix.

Mentions: The widespread effects of ERβ on MCF-7 cell transcriptome are likely to result from multiple effects of this receptor in the cells, including direct regulation of primary response genes via genomic or non genomic mechanisms, and indirect gene regulation events mediated by the products of primary genes. The primary ERβ target genes are most likely to comprise also master regulators of complex cellular responses to the receptor, mediating its effects on the biological and clinical phenotype of BC cells. To identify such primary genomic targets and investigate the mechanisms that allow their regulation by ERβ, a global analysis of in vivo binding of this receptor to the genome was carried out in TAP-ERβ cells by chromatin immunoprecipitation coupled to massively parallel sequencing (ChIP-Seq) [38], that allows detailed mapping of in vivo TF binding to the genome. In parallel, we studied ERα binding to the genome under the same conditions, to allow comparative analyses between the two ER subtypes. Replicate chromatin samples were prepared from both Ct-ERβ and Nt-ERβ cells before and after stimulation for 45 minutes with 10-8M E2 and DNA-bound proteins were immunoprecipitated either with antibodies against the N- and C-terminus of ERα, or with IgGs binding with high affinity the TAP moiety of tagged ERβ (see Methods). Preliminary testing on several known ERβ binding sites, including the promoter-near region of pS2/TFF1 gene [26], confirmed that the method selected to immunoprecipitate chromatin-bound Ct-ERβ and Nt-ERβ was efficient and specific (data not shown). The resulting DNAs were used to generate ChIP-Seq libraries for ERα and ERβ, respectively, that were then sequenced with the Illumina Genome Analyzer. The sequence tags obtained were then aligned to the human genome sequence and peaks enriched in the libraries generated after E2-treatment were identified using MACS (Model-based Analysis of ChIP-Seq) [39]. This led to the identification of 9702 binding sites for ERβ and 6024 sites for ERα, of which 4506 were shared by both receptors (Figure 3A), with an average False Discovery Rate (FDR) of 3%. The full list of these binding sites is available, with relevant information, in Additional Files 1 and 2. Interestingly, about half (4862) ERβ binding sites identified map within transcription units, mainly (3942 sites) in intronic regions. This distribution is maintained also in 424 ERβ-regulated transcription units (see below), where 966 ERβ binding sites located in the gene or within 10 kbps from it are distributed as follows: 154 in promoter regions, 51 in exons, 471 within one or more introns and the remaining either upstream of promoters (156) or downstream of the gene (134). In both cases the ERβ binding sites within genes did not show any preference with respect to exon or intron position nor for know intragenic regulatory elements (splice sites, polyadenylation sites, etc). It should be mentioned that the number of ERβ binding sites identified is significantly higher that those mapped in MCF-7 cells by ChIP-on-chips [30,31], possibly for technical differences due to ERβ expression levels in the different MCF-7 cell-derived clones used, in immunoprecipitation efficiency and/or in DNA analysis. Since only Zhao et al. [31] performed an unbiased search for ERβ binding with whole-genome chips, we could confront our results only with those reported in that study. This showed that 86% of high confidence ERβ sites described in that study appear also in our dataset. The binding sites identified here were then subjected to sequence analysis, searching first for the presence of EREs (Estrogen Receptor Elements), the characteristic ER binding signature (Figure 3B). This analysis revealed that in all three cases (i.e. ERβ, ERα and ERβ+ERα) a high percentage of sites displayed one or more imperfectly palindromic ERE (ERE+), with a slightly higher positivity in ERα sites (58.89 vs 53.51%). As ERs have been shown to bind both in vitro and in vivo to PuGGTCA hemi-palindromes (hEREs), we searched the sequence of the remaining (ERE-) sites for perfect matches to this sequence. Results showed that almost half of them indeed contained one or more hEREs. The percentages of sites not carrying a known ER-binding element (ERE- and hERE-) were similar for both receptors (ERα: 22.34%, ERβ: 28.38%). We observed that ERα and ERβ binding sites were often found in close proximity to each other, a confounding factor when attempting to discern and analyze separately ER subtype-specific sites and target genes. This could be due to the limits of the ChIP-Seq technology or of the algorithm used for peak selection. To overcome this problem, and allow the identification of potential ER subtype-specific sites, we used a cartographic approach to group nearby binding sites that might be the result, at least in part, of shortfalls of the mapping methods applied. Each binding peak was thus elongated in both directions by 1000 bp and the overlapping ones obtained were merged into 8536 ERβ and 5371 ERα 'extended' binding regions. These regions were intersected to define ERα only, ERβ only or ERα+ERβ binding regions. In this way we could identify 1271 ERα-only and 4541 ERβ-only binding sites, comprised in these regions, none of which showed nearby binding of the other receptor. These were named: ER subtype 'prevalent' sites. The binding peak sequences included in each of the three regions obtained (ERα only, ERβ only or ERα+ERβ) were then re-analyzed for the presence of ERE or hERE elements. In this way we could observe that sites within the ERα+ERβ regions showed now a much higher percentage of ERE+ sequences (62.90%), respect to those present in the ERα-only or ERβ-only regions (45.63% and 44.62%, respectively, Figure 3C). Since all three types of sites showed almost identical proportions of hERE+, this result suggests that perfectly or imperfectly palindromic EREs are preferential binding sequences for ERα-ERβ heterodimers, while ERα and ERβ homodimers appear to be more flexible in DNA recognition. ERE+ sequences were then analyzed in more detail with MEME [40], to investigate if the three classes of sites identified showed any difference in the relative base composition of their respective ERE signatures. For each list of sequences, the most significant position-specific probability matrix generated by MEME was compared to the matrices present in the JASPAR transcription factor binding profile database [41], using the STAMP tool-kit for DNA motif comparison [42]. As shown in Figure 3C (left panels), this analysis revealed that the ERE matrices derived from the three types of binding regions identified (ERα selective, ERβ selective and ERα+ERβ) are identical and, as a consequence, that ERβ does not appear to display ERE variant selectivity.


Global analysis of estrogen receptor beta binding to breast cancer cell genome reveals an extensive interplay with estrogen receptor alpha for target gene regulation.

Grober OM, Mutarelli M, Giurato G, Ravo M, Cicatiello L, De Filippo MR, Ferraro L, Nassa G, Papa MF, Paris O, Tarallo R, Luo S, Schroth GP, Benes V, Weisz A - BMC Genomics (2011)

Sequence analysis of ERα, ERβ and ERα+ERβ binding sites. (A) Venn diagram showing a summary of ERα and ERβ binding sites identified in TAP-ERβ cells by ChIP-Seq. (B) Classification of ERα and β binding sites based on the presence of a perfectly or imperfectly palindromic Estrogen Response Element (ERE, green), an ERE hemipalindrome (hERE, blue) or no ERE (none, red). (C) ERE motif matrices identified in each of the three ER binding regions indicated (left), classification of the binding sites belonging to each region according to the presence of ERE (center) and grid summarizing the results of TFBS matrix enrichment (overrepresentation) analyses performed on the binding sites groups indicated (right). Z-Score cut-off was 3.0 and only TFBSs showing an over-representation score ≥4.0 in at least one of the ERE- (none) binding site groups. Light grey cells indicate a Z-Score <3.0 while dark grey cells indicate absence of the matrix.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3025958&req=5

Figure 3: Sequence analysis of ERα, ERβ and ERα+ERβ binding sites. (A) Venn diagram showing a summary of ERα and ERβ binding sites identified in TAP-ERβ cells by ChIP-Seq. (B) Classification of ERα and β binding sites based on the presence of a perfectly or imperfectly palindromic Estrogen Response Element (ERE, green), an ERE hemipalindrome (hERE, blue) or no ERE (none, red). (C) ERE motif matrices identified in each of the three ER binding regions indicated (left), classification of the binding sites belonging to each region according to the presence of ERE (center) and grid summarizing the results of TFBS matrix enrichment (overrepresentation) analyses performed on the binding sites groups indicated (right). Z-Score cut-off was 3.0 and only TFBSs showing an over-representation score ≥4.0 in at least one of the ERE- (none) binding site groups. Light grey cells indicate a Z-Score <3.0 while dark grey cells indicate absence of the matrix.
Mentions: The widespread effects of ERβ on MCF-7 cell transcriptome are likely to result from multiple effects of this receptor in the cells, including direct regulation of primary response genes via genomic or non genomic mechanisms, and indirect gene regulation events mediated by the products of primary genes. The primary ERβ target genes are most likely to comprise also master regulators of complex cellular responses to the receptor, mediating its effects on the biological and clinical phenotype of BC cells. To identify such primary genomic targets and investigate the mechanisms that allow their regulation by ERβ, a global analysis of in vivo binding of this receptor to the genome was carried out in TAP-ERβ cells by chromatin immunoprecipitation coupled to massively parallel sequencing (ChIP-Seq) [38], that allows detailed mapping of in vivo TF binding to the genome. In parallel, we studied ERα binding to the genome under the same conditions, to allow comparative analyses between the two ER subtypes. Replicate chromatin samples were prepared from both Ct-ERβ and Nt-ERβ cells before and after stimulation for 45 minutes with 10-8M E2 and DNA-bound proteins were immunoprecipitated either with antibodies against the N- and C-terminus of ERα, or with IgGs binding with high affinity the TAP moiety of tagged ERβ (see Methods). Preliminary testing on several known ERβ binding sites, including the promoter-near region of pS2/TFF1 gene [26], confirmed that the method selected to immunoprecipitate chromatin-bound Ct-ERβ and Nt-ERβ was efficient and specific (data not shown). The resulting DNAs were used to generate ChIP-Seq libraries for ERα and ERβ, respectively, that were then sequenced with the Illumina Genome Analyzer. The sequence tags obtained were then aligned to the human genome sequence and peaks enriched in the libraries generated after E2-treatment were identified using MACS (Model-based Analysis of ChIP-Seq) [39]. This led to the identification of 9702 binding sites for ERβ and 6024 sites for ERα, of which 4506 were shared by both receptors (Figure 3A), with an average False Discovery Rate (FDR) of 3%. The full list of these binding sites is available, with relevant information, in Additional Files 1 and 2. Interestingly, about half (4862) ERβ binding sites identified map within transcription units, mainly (3942 sites) in intronic regions. This distribution is maintained also in 424 ERβ-regulated transcription units (see below), where 966 ERβ binding sites located in the gene or within 10 kbps from it are distributed as follows: 154 in promoter regions, 51 in exons, 471 within one or more introns and the remaining either upstream of promoters (156) or downstream of the gene (134). In both cases the ERβ binding sites within genes did not show any preference with respect to exon or intron position nor for know intragenic regulatory elements (splice sites, polyadenylation sites, etc). It should be mentioned that the number of ERβ binding sites identified is significantly higher that those mapped in MCF-7 cells by ChIP-on-chips [30,31], possibly for technical differences due to ERβ expression levels in the different MCF-7 cell-derived clones used, in immunoprecipitation efficiency and/or in DNA analysis. Since only Zhao et al. [31] performed an unbiased search for ERβ binding with whole-genome chips, we could confront our results only with those reported in that study. This showed that 86% of high confidence ERβ sites described in that study appear also in our dataset. The binding sites identified here were then subjected to sequence analysis, searching first for the presence of EREs (Estrogen Receptor Elements), the characteristic ER binding signature (Figure 3B). This analysis revealed that in all three cases (i.e. ERβ, ERα and ERβ+ERα) a high percentage of sites displayed one or more imperfectly palindromic ERE (ERE+), with a slightly higher positivity in ERα sites (58.89 vs 53.51%). As ERs have been shown to bind both in vitro and in vivo to PuGGTCA hemi-palindromes (hEREs), we searched the sequence of the remaining (ERE-) sites for perfect matches to this sequence. Results showed that almost half of them indeed contained one or more hEREs. The percentages of sites not carrying a known ER-binding element (ERE- and hERE-) were similar for both receptors (ERα: 22.34%, ERβ: 28.38%). We observed that ERα and ERβ binding sites were often found in close proximity to each other, a confounding factor when attempting to discern and analyze separately ER subtype-specific sites and target genes. This could be due to the limits of the ChIP-Seq technology or of the algorithm used for peak selection. To overcome this problem, and allow the identification of potential ER subtype-specific sites, we used a cartographic approach to group nearby binding sites that might be the result, at least in part, of shortfalls of the mapping methods applied. Each binding peak was thus elongated in both directions by 1000 bp and the overlapping ones obtained were merged into 8536 ERβ and 5371 ERα 'extended' binding regions. These regions were intersected to define ERα only, ERβ only or ERα+ERβ binding regions. In this way we could identify 1271 ERα-only and 4541 ERβ-only binding sites, comprised in these regions, none of which showed nearby binding of the other receptor. These were named: ER subtype 'prevalent' sites. The binding peak sequences included in each of the three regions obtained (ERα only, ERβ only or ERα+ERβ) were then re-analyzed for the presence of ERE or hERE elements. In this way we could observe that sites within the ERα+ERβ regions showed now a much higher percentage of ERE+ sequences (62.90%), respect to those present in the ERα-only or ERβ-only regions (45.63% and 44.62%, respectively, Figure 3C). Since all three types of sites showed almost identical proportions of hERE+, this result suggests that perfectly or imperfectly palindromic EREs are preferential binding sequences for ERα-ERβ heterodimers, while ERα and ERβ homodimers appear to be more flexible in DNA recognition. ERE+ sequences were then analyzed in more detail with MEME [40], to investigate if the three classes of sites identified showed any difference in the relative base composition of their respective ERE signatures. For each list of sequences, the most significant position-specific probability matrix generated by MEME was compared to the matrices present in the JASPAR transcription factor binding profile database [41], using the STAMP tool-kit for DNA motif comparison [42]. As shown in Figure 3C (left panels), this analysis revealed that the ERE matrices derived from the three types of binding regions identified (ERα selective, ERβ selective and ERα+ERβ) are identical and, as a consequence, that ERβ does not appear to display ERE variant selectivity.

Bottom Line: Expression of full-length ERβ in hormone-responsive, ERα-positive MCF-7 cells resulted in a marked reduction in cell proliferation in response to estrogen and marked effects on the cell transcriptome.Of 921 genes differentially regulated by estrogen in ERβ+ vs ERβ- cells, 424 showed one or more ERβ site within 10 kb.ERβ binding in close proximity of several miRNA genes and in the mitochondrial genome, suggests the possible involvement of this receptor in small non-coding RNA biogenesis and mitochondrial genome functions.

View Article: PubMed Central - HTML - PubMed

Affiliation: Department of General Pathology, Second University of Naples, vico L, De Crecchio 7, 80138 Napoli, Italy.

ABSTRACT

Background: Estrogen receptors alpha (ERα) and beta (ERβ) are transcription factors (TFs) that mediate estrogen signaling and define the hormone-responsive phenotype of breast cancer (BC). The two receptors can be found co-expressed and play specific, often opposite, roles, with ERβ being able to modulate the effects of ERα on gene transcription and cell proliferation. ERβ is frequently lost in BC, where its presence generally correlates with a better prognosis of the disease. The identification of the genomic targets of ERβ in hormone-responsive BC cells is thus a critical step to elucidate the roles of this receptor in estrogen signaling and tumor cell biology.

Results: Expression of full-length ERβ in hormone-responsive, ERα-positive MCF-7 cells resulted in a marked reduction in cell proliferation in response to estrogen and marked effects on the cell transcriptome. By ChIP-Seq we identified 9702 ERβ and 6024 ERα binding sites in estrogen-stimulated cells, comprising sites occupied by either ERβ, ERα or both ER subtypes. A search for TF binding matrices revealed that the majority of the binding sites identified comprise one or more Estrogen Response Element and the remaining show binding matrixes for other TFs known to mediate ER interaction with chromatin by tethering, including AP2, E2F and SP1. Of 921 genes differentially regulated by estrogen in ERβ+ vs ERβ- cells, 424 showed one or more ERβ site within 10 kb. These putative primary ERβ target genes control cell proliferation, death, differentiation, motility and adhesion, signal transduction and transcription, key cellular processes that might explain the biological and clinical phenotype of tumors expressing this ER subtype. ERβ binding in close proximity of several miRNA genes and in the mitochondrial genome, suggests the possible involvement of this receptor in small non-coding RNA biogenesis and mitochondrial genome functions.

Conclusions: Results indicate that the vast majority of the genomic targets of ERβ can bind also ERα, suggesting that the overall action of ERβ on the genome of hormone-responsive BC cells depends mainly on the relative concentration of both ERs in the cell.

Show MeSH
Related in: MedlinePlus