Genome-Wide Mapping of Estrogen Receptor-Β–Binding Regions Reveals Extensive Cross-Talk with Transcription Factor Activator Protein-1

Published OnlineFirst May 25, 2010; DOI: 10.1158/0008-5472.CAN-09-4407

Tumor and Stem Cell Biology Cancer Research Genome-Wide Mapping of Estrogen Receptor-β–Binding Regions Reveals Extensive Cross-Talk with Transcription Factor Activator Protein-1

Chunyan Zhao1, Hui Gao1, Yawen Liu2, Zoi Papoutsi1, Sadaf Jaffrey1, Jan-Åke Gustafsson1,3, and Karin Dahlman-Wright1

Abstract Estrogen signaling can occur through a nonclassical pathway involving the interaction of estrogen receptors (ER) with other transcription factors such as activator protein-1 (AP-1) and SP-1. However, there is little mechanistic understanding about this pathway, with conflicting results from in vitro investigations. In this study, we applied the ChIP-on-chip approach to identify ERβ-binding sites on a genome-wide scale, identifying 1,457 high-confidence binding sites in ERβ-overexpressing MCF7 breast cancer cells. Genes containing ERβ-binding sites can be regulated by E2. Notably, ∼60% of the genomic regions bound by ERβ contained AP-1–like binding regions and estrogen response element–like sites, suggesting a functional association between AP-1 and ERβ signaling. Chromatin immunoprecipitation (ChIP) analysis confirmed the association of AP-1, which is composed of the oncogenic transcription factors c-Fos and c-Jun, to ERβ-bound DNA regions. Using a re-ChIP assay, we showed co-occupancy of ERβ and AP-1 on chromatin. Short interfering RNA–mediated knockdown of c-Fos or c-Jun expression decreased ERβ recruitment to chromatin, consistent with the role of AP-1 in mediating estrogen signaling in breast cancer cells. Additionally, ERα and ERβ recruitment to AP-1/ERβ target regions exhibited gene-dependent differences in response to antiestrogens. Together, our results broaden insights into ERβ DNA-binding at the genomic level by revealing crosstalk with the AP-1 transcription factor. Cancer Res; 70(12); 5174–83. ©2010 AACR.

Introduction tor complexes including Fos/Jun [activator protein-1 (AP-1)– responsive elements] or SP-1 (4, 5). However, there is little Estrogens bind to and activate two estrogen receptors mechanistic understanding about this pathway, with (ERα and ERβ), and exert their effects through a complex conflicting results from in vitro investigations (6, 7). A small array of signaling pathways that mediate genomic and non- number of genes containing an AP-1 site in their promoters genomic events (1, 2). ERs regulate gene expression through have been shown to be regulated by ERα, such as collagenase distinct DNA response elements. The classical mechanism of (8), human insulin-like growth factor I (9), and the human estrogen signaling is through an estrogen response element choline acetyltransferase gene (10). (ERE). In this process, ER dimerizes and interacts with EREs The AP-1 transcription factor is a key component of many associated with target genes, followed by the recruitment of a signal transduction pathways. AP-1 was first known as a 12-O- variety of coregulators to alter chromatin structure and facil- tetradecanoylphorbyl-13-acetate (TPA)–inducible transcrip- itate recruitment of the RNA polymerase II (PolII) transcription factor because the TPA response element was identified tional machinery (2, 3). Estrogen signaling could also occur as a binding site for AP-1 in many cellular and viral genes (11). through a nonclassical pathway in which liganded ERs are The AP-1 transcription factor is a dimeric complex consisting tethered to DNA via association with other transcription fac- of homodimers of Jun family members or heterodimers of Jun and Fos family members. AP-1 dimers regulate the expression of AP-1 target genes by binding to consensus DNA-regulatory elements (12). For example, c-Jun homodimers and c-Jun/c-

Authors' Affiliations: 1Department of Biosciences and Nutrition, Novum, Fos heterodimers bind to the consensus TPA response ele- Karolinska Institutet, Huddinge, Sweden; 2Department of Epidemiology ment sequence TGAC/GTCA (13). and Biostatistics, School of Public Health, Jilin University, Changchun, In reporter assays, in the context of an isolated AP-1 ele- China; and 3Center for Nuclear Receptors and Cell Signaling, Department α β of Biology and Biochemistry, University of Houston, Houston, Texas ment, ER and ER were shown to signal in opposite ways when complexed with estrogen: with ERα, estrogen-activated Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/). transcription, whereas with ERβ, estrogen-inhibited tran- Corresponding Author: Chunyan Zhao, Department of Biosciences and scription (7). However, it remains unclear whether AP-1 is Nutrition, Novum, Karolinska Institutet, S-141 57 Huddinge, Sweden. more generally recruited to promoter regions of ER target Phone: 46-8608-9273; Fax: 46-8774-5538; E-mail: [email protected]. genes, particularly in the context of intact chromatin and doi: 10.1158/0008-5472.CAN-09-4407 what is the potential molecular interplay between these sig- ©2010 American Association for Cancer Research. naling pathways at the level of chromatin.

5174 Cancer Res; 70(12) June 15, 2010

Association of AP-1 and ERβ

Recently, the combined technique of chromatin immuno- the University of California Santa Cruz genome browser. precipitation (ChIP) with genomic DNA microarrays (ChIP- Clover (21) was used to screen the sequences against a on-chip; ref. 14) or with DNA sequencing (ChIP-PET and precompiled library of motifs, JASPAR, to find the statisti- ChIP-Seq; refs. 15, 16) has been used for the identification cally overrepresented motifs as described in ref. (22). The of ERα-binding sites at the whole genome level. In this study, similarity of enriched motifs was analyzed using the web using ChIP-on-chip, we mapped ERβ-binding DNA regions tool STAMP at default settings (23). on a global scale and explored mechanisms of cross-talk between ERβ and AP-1. Quantitative real-time PCR Total RNA was extracted using the RNeasy kit (Qiagen). Materials and Methods Real-time PCR was performed as previously described (18).

Generation of a stable MCF7 tet-off ERβ clone RNA interference The generation of a stable MCF7 tet-off ERβ clone has pre- Short interfering RNA (siRNA) targeting c-Fos (si_c-Fos) or viously been described by Papoutsi and colleagues (17). c-Jun (si_c-Jun) and nonspecific siRNA (si_Control) were pur- chased from Santa Cruz Biotechnology. siRNA transfections Western blot analysis were carried out using a final concentration of 50 nmol/L oligo Western blotting was carried out as previously described using the Interferin transfection reagent (Polyplus) according (18) using the following antibodies: anti-Flag (M5; Sigma), to the recommendations of the manufacturer. anti–c-Fos (H-125; Santa Cruz Biotechnology), and anti–c- Jun (sc-45; Santa Cruz Biotechnology). Luciferase assay ERβ-binding regions identified from ChIP-on-chip for an ChIP/re-immunoprecipitation association with the EREG,FOXA1,MYC,VAV3,CAPN2, ChIP analysis was performed as previously described (18). and LTBP1 genes were amplified from genomic DNA using The antibodies used were as follows: anti-ERβ rabbit poly- PCR amplification and cloned into the pGL3 promoter clonal antibody LBD (19), anti-ERα (HC-20; Santa Cruz Bio- luciferase vector (Promega) at MluIandXhoIsites.The technology), anti–c-Fos (H-125; Santa Cruz Biotechnology), primer pairs used to amplify these regions are shown in anti–c-Jun (sc-45; Santa Cruz Biotechnology), anti-Flag (M5; Supplementary Table S1. Hormone-depleted HeLa cells were Sigma), and normal rabbit IgG (Santa Cruz Biotechnology). transfected using LipofectAMINE 2000 (Invitrogen Life Re-ChIP was carried out as described (17). The immunopre- Technologies). Luciferase activity was measured using the cipitated DNA was amplified by real-time PCR using Plati- Dual-Luciferase Reporter Assay (Promega). To control for num SYBR green quantitative PCR supermix uracil DNA transfection efficiency, luciferase activity was normalized to glycosylase (Invitrogen). Renilla luciferase activity.

Probe labeling and microarray hybridization Fluorescent-activated cell sorting ChIP DNA samples (three E2-treated ERβ ChIP samples and MCF7 tet-off ERβ cells were seeded in phenol red–free three input DNA samples) were amplified and labeled as de- DMEM supplemented with 5% charcoal dextran–stripped scribed (19). Six micrograms of labeled products were hybrid- fetal bovine serum for 24 hours and were transfected with ized to Affymetrix human tiling 2.0R arrays (Affymetrix). control siRNA or c-Fos siRNA for 48 hours; the cells were then stimulated with 10 nmol/L of E2 or vehicle for 72 hours. Affymetrix data analysis Cells were collected and stained with propidium iodide. To obtain high-confidence ERβ-binding sites, the scanned Staining was measured using flow cytometry by the fluores- output files were analyzed by two independent methods: cence intensity (FL-1, 530 nm) of 10,000 cells. Tiling Analysis Software version 1.1 (Affymetrix) using a threshold of P ≤ 0.0005, and Model-based Analysis of Results Tiling arrays (MAT; ref. 20) using a MAT score cutoff of 120. A total of 1,710 and 1,878 ERβ DNA-binding regions Global mapping of ERβ-binding regions in human were identified using either the Tiling Analysis Software breast cancer cells or the MAT analysis method. The predicted sites had a high A stable cell line, MCF7 ERβ, that expresses a tetracycline- degree of concordance, and a total of 1,457 ERβ-binding re- regulated version of ERβ, was used as a model system gions were identified by both analysis strategies. The same for mapping ERβ-binding regions on a genome-wide data set was also analyzed only by MAT using a MAT score scale. Figure 1A shows that high levels of ERβ protein cutoff of 50 to obtain a more extensive set of ERβ-binding were expressed in the absence of tetracycline, but not in 4,766 sites. the presence of tetracycline. We have previously reported that the protein ratio of ERα/ERβ is 0.7:1 in this cell Identification of common motifs in line (17). ERβ-binding regions There was an 84-fold enrichment of ERβ-binding (rela- The genomic DNA corresponding to every ChIP-enriched tive to control IgG antibody) to the known ER-binding site region identified by whole tiling array was retrieved from in the pS2/TFF1 gene promoter (Fig. 1B). No enrichment of

www.aacrjournals.org Cancer Res; 70(12) June 15, 2010 5175

Zhao et al.

Figure 1. Confirmation of ERβ expression in MCF7 tet-off ERβ cells and its binding to the pS2/TFF1 gene promoter (A and B). A, cells were cultured in the presence or absence of tetracycline for 16 h. Expression of Flag-tagged ERβ was determined by Western blot analysis using the anti-Flag antibody. B, cells were cultured in the presence or absence of tetracycline for 16 h. Cells were then treated with 10 nmol/L of E2 for 45 min and subjected to ChIP assays. Columns, mean fold enrichment of ERβ relative to IgG; bars, SD (n = 3). C, location of ERβ-binding sites relative to the nearest genes in the University of California Santa Cruz Known Gene database shows a large majority of sites distal to the genes (>1 kb) or within intragenic regions.

ERβ-binding was observed to a nonbinding control region of analyzed using the web tool STAMP (23). All the significantly the pS2 gene (intron 2 of the pS2 gene). No enrichment for enriched motifs could be subgrouped into three big clusters ERβ-binding was observed in the presence of tetracycline, (ERE-like sites, Forkhead sites, and AP-1 sites). Interestingly, when ERβ is not expressed, showing that there is no cross- only 5% of ERβ-interacting regions include only ERE or ERE reactivity of the anti-ERβ antibody with ERα. half-sites, whereas ∼60% of ERβ-interacting regions contain We performed ChIP-on-chip analyses using the Affymetrix AP-1–like binding sites together with ERE-like sites (Fig. 2B). Human Tiling 2.0 microarrays representing the entire nonre- Among them, 45% contain also forkhead-like binding sites. petitive human genome sequence tiled at 35 bp resolution. Our analysis suggests interactions of the AP-1 family and Using MAT analysis with a less stringent cutoff, we identified the Forkhead family of transcription factors with ERβ at 4,766 ERβ-binding sites. Additionally, 1,457 high-confidence the level of chromatin binding. In previous studies (24, 25), ERβ DNA-binding regions (Supplementary Table S2) were FOXA1 has been shown to define a domain of the estrogen derived by only including DNA-binding regions that were response. identified by two independent analysis strategies, TAS and MAT with stringent cutoffs. These regions were used for sub- Comparisons of the ERβ cistrome with previously sequent motif analysis. determined ER cistromes ChIP-quantitative PCR analysis of samples from inde- We compared our ERβ cistromes, 4,766 sites or 1,457 sites pendent biological triplicates validated the binding of derived using less or more strict inclusion criteria, re- ERβ to randomly selected sites (27 of 27; Supplementary spectively, with the published ERα cistrome of 5,782 binding Fig. S1). Figure 1C shows that the largest fraction (58%) sites (14) and the recently published ERβ cistrome of 1,744 of binding regions map to enhancers, defined as 1 to 300 kb sites (26), all derived from MCF7 cells (Fig. 2C). Fifty-one per- from either end of the genes. Forty percent of the binding cent (2,439 sites) of our less stringent set of ERβ sites and 76% regions were within genes. Only 3% of ERβ-binding re- (1,112 sites) of our stringent set of ERβ sites, respectively, gions were within 1 kb from either end of the genes. were also identified as ERα-binding sites by Carroll and col- These results are in good agreement with published data on leagues (14). More recently, Charn and colleagues (26), using a ERα (15, 19). custom-designed tiling array that contains approximate- ly 77,000 genomic regions, have identified 1,744 ERβ-binding DNA motifs in ERβ-binding regions suggest the sites. As shown in Fig. 2C, only 327 (19%) and 170 (10%) of involvement of AP-1 in ERβ-mediated gene their identified ERβ-binding sites were present in our less transcription stringent and more stringent ERβ data sets, respectively. Fifteen motifs were identified as significantly enriched We further compared the predominant transcription sites in the identified ERβ-binding regions (Fig. 2A). Some po- factor motifs enriched in our 1,457 ERβ-binding sites with sition weight matrices for different motifs closely resemble those enriched in the 1,744 ERβ-binding sites identified by each other. The similarity between enriched motifs was Charn and colleagues (ref. 26; Supplementary Table S3).

5176 Cancer Res; 70(12) June 15, 2010 Cancer Research

Association of AP-1 and ERβ

Significant enrichment of ERE-like sites and AP-1 sites was specifically upregulated (FOXA1) or downregulated (NEDD9) found in both data sets. However, Forkhead sites were highly by ERβ. enriched only in our data set. AP-1 is recruited to ERβ-binding regions Genes containing ERβ-binding sites could be Having shown by motif analysis that the AP-1 motif is regulated by E2 significantly enriched in identified ERβ-binding regions, we To address how ERβ-binding relates to gene expression, examined whether AP-1 was recruited to ERβ-binding re- we quantified expression levels for 10 genes that were gions containing the consensus AP-1 sites. We randomly se- found to have ERβ-binding sites in their proximal promoters lected 25 ERβ-binding regions that contained a consensus and 10 genes containing ERβ-binding sites in enhancer AP-1–binding site G/ATGAC/GTCA. Three binding regions regions. We observed that for 9 of 10 genes having ERβ- without a predicted AP-1–binding sequence were selected binding sites in promoters, the expression of ERβ clearly as negative controls. We confirmed ERβ recruitment to modulated gene expression levels (Fig. 3, top). Three of all the selected sites (Supplementary Fig. S2). ChIP followed 10 genes having ERβ-binding sites in enhancers were by real-time PCR revealed that 18 of the regions selected regulated by ERβ expression (Fig. 3, bottom). Consistent on the basis that they included an AP-1 consensus site re- with previous reports (27), we observed that ERβ could cruited c-Fos and c-Jun, whereas regions that did not include modulate ERα-mediated transcriptional activity positively this site did not recruit c-Fos and c-Jun (Fig. 4A; see Supple- and negatively. For instance, ERβ expression inhibited mentary Table S4 for the complete list of these binding ERα-mediated E2 induction of genes such as pS2, PKIB, regions). Using Pearson's correlation coefficient analysis, a PDZK1, GREB1, EREG, MYC, and NBPF4. The ERβ-mediated positive correlation (r =0.9,P < 0.05) was observed for inhibition of ERα-mediated E2 induction of the pS2 gene the recruitment of c-Fos and c-Jun to these regions. Six is in line with previously published data (28). We also ob- regions adjacent to the EREG, FOXA1, MYC, VAV3, CAPN2, served that ERβ reversed ERα-mediated stimulation by E2 and LTBP1 genes displayed the highest recruitment of (ADORA1 and IL20) or had no influence on ERα-mediated c-Fos and c-Jun, and these regions were selected for the sub- stimulation by E2 (PLAC1). Interestingly, some genes are sequent studies.

Figure 2. Identification of enriched motifs within the ERβ-binding regions (A and B). A, the corresponding distance tree of enriched motifs in the ERβ-binding regions identified. B, the distribution and occurrence of ERE-like sites, AP-1–like sites, and Forkhead-like sites within the identified ERβ-binding sites. C, Venn diagram of the overlap between the 4,766 sites or 1,457 sites derived using less or more strict inclusion criteria in this study and the 5,782 ERα sites discovered by Carroll and colleagues (14) or the 1,744 ERβ sites discovered by Charn and colleagues (26), respectively. ChIP-on-chip data of 5,782 ERα sites by Carroll and colleagues (14) was downloaded (44).

www.aacrjournals.org Cancer Res; 70(12) June 15, 2010 5177

Zhao et al.

AP-1 facilitates ERβ association with chromatin ChIP assay was performed using an anti-ERβ antibody. We used RNAi-mediated inhibition of c-Fos and c-Jun to Then, before reversal of protein-DNA cross-linking, the investigate the role of AP-1 for ERβ recruitment. The siRNAs anti-ERβ ChIP DNA was subjected to re-ChIP using an used effectively decreased both the mRNA and protein levels anti–c-Fos antibody. As shown in Fig. 5A, this assay confirms of c-Fos and c-Jun in MCF7 cells (Fig. 4B and C). Importantly, the co-occupancy of c-Fos and ERβ on the selected regions, siRNA directed against c-Fos or c-Jun did not affect ERβ whereas no co-occupancy of c-Fos and ERβ at control protein levels (Fig. 4C). Figure 4D shows that knockdown of regions was observed. c-Fos or c-Jun resulted in a significant decrease in ERβ recruitment to the investigated regions, EREG, FOXA1, MYC, ERβ/AP-1–binding regions function as VAV3, CAPN2, and LTBP1, whereas the recruitment of ERβ transcriptional enhancers to control regions (PDZK1, MAT1A, and VANGL1) was not Reporter assays were used to assay the transcriptional ac- affected. The reduction of recruitment of ERβ to these regions tivity of DNA regions including ERβ- and AP-1–binding sites. upon knockdown of c-Fos was also confirmed using a differ- As shown in Fig. 5B, four out of the six regions tested showed ent antibody (anti-Flag antibody; Supplementary Fig. S3). an increase in luciferase activity upon E2 treatment. To con- These data indicate that AP-1 plays an important role in en- firm that these regions harbor a functional AP-1 site, cells hancing ERβ association with chromatin. were treated with TPA, which is a potent activator of AP-1. All six binding regions showed greatly increased activity up- Co-occupancy of AP-1 and ERβ at AP-1/ERβ on TPA treatment. Moreover, TPA-stimulated AP-1 activity target DNA regions was further increased by the addition of E2, suggesting that Corecruitment of AP-1 and ERβ to AP-1/ERβ target ERβ and AP-1 were working synergistically to transactivate DNA regions was studied using a re-ChIP assay. First, the the reporter construct.

ERβ expression and c-Fos silencing inhibit E2-stimulated cell proliferation MCF7 tet-off ERβ cells expressing only ERα or both ERα and ERβ were transiently transfected with control siRNA or c-Fos siRNA. In line with published data (29), we found that ERβ expression significantly decreased E2-induced cell proliferation when compared with cells expressing only ERα (compare si_control/ERα+ERβ versus si_control/ERα only; Fig. 5C). Also, we observed that silencing of c-Fos in cells expressing only ERα resulted in a decrease in E2-induced cell proliferation compared with cells transfected with siRNA control (compare si_c-Fos/ERα only versus si_control/ERα only). These results are in agreement with previous studies regarding the role of c-Fos in MCF7 cell proliferation (30). In cells expressing both ERα and ERβ, we observed that silencing of c-Fos completely abolished E2-induced cell proliferation compared with cells transfected with siRNA control (compare si_c-Fos/ERα+ERβ versus si_control/ERα+ERβ), which might be due to a combination of the inhibitory effects of both ERβ and si_c-Fos on cell proliferation.

c-Fos silencing affects ERβ-mediated transcription We investigated the effect of silencing of c-Fos on estrogen- stimulated expression of two genes, FOXA1 and PKIB, which were found to be regulated by ERβ (shown in Fig. 3) and to recruit both ERβ and AP-1 to their promoters (shown in Fig. 4A and Supplementary Fig. S2, respectively). The reduction of c-Fos levels significantly attenuated the estrogen- responsive induction of these two genes (Fig. 5D), demon- Figure 3. Estrogen-dependent changes in the expression of 20 selected β genes in MCF7 tet-off ERβ cells expressing only ERα or both ERα and strating that the ER -mediated changes in gene regulation ERβ. Cells were grown in phenol red–free DMEM supplemented with are sensitive to changes in the levels of c-Fos. 5% charcoal dextran–stripped fetal bovine serum for at least 3 d and were treated with 10 nmol/L of E2 or vehicle for 24 h. Columns, mean Differential recruitment of ERα and ERβ to AP-1/ERβ fold change compared with vehicle-treated cells; bars, SD (n = 2). *, P < 0.05 relative to vehicle-treated controls determined by t test. target DNA regions in response to ligands Top, 10 genes containing ERβ-binding sites in their proximal promoters. To understand the effect of different ligands on transcrip- Bottom, 10 genes containing ERβ-binding sites in enhancers. tional activities of ERα and ERβ via AP-1–directed pathways,

5178 Cancer Res; 70(12) June 15, 2010 Cancer Research

Association of AP-1 and ERβ

Figure 4. Recruitment of c-Fos and c-Jun to ERβ-binding regions and effects of specific targeted knockdown of c-Fos or c-Jun on ERβ recruitment. A, MCF7 tet-off ERβ cells were cultured in the absence of tetracycline for 16 h followed by treatment with 10 nmol/L of E2 for 45 min. ChIP DNA was analyzed by real-time PCR for the binding regions close to 18 genes that include AP-1 sites, as well as for ERβ-binding regions close to three genes that did not include a predicted AP-1 element as negative controls. Columns, mean fold enrichment of c-Fos or c-Jun relative to IgG; bars, SD (n = 2). *, P < 0.05 compared with IgG controls. B, C, and D, MCF7 tet-off ERβ cells were transfected with siRNA to c-Fos (si_c-Fos), c-Jun (si_c-Jun), or nonspecific siRNA (si_control) for 72 h. B, siRNA-mediated knockdown of c-Fos or c-Jun significantly reduces c-Fos and c-Jun mRNA levels. Columns, mean; bars, SD (n = 2). *, P < 0.05 relative to si_control. C, c-Fos or c-Jun depletion significantly reduces the protein levels of c-Fos or c-Jun, but does not affect ERβ protein levels, as determined by Western blot analysis using the anti–c-Fos, anti–c-Jun, and anti-Flag antibody, respectively. D, ChIP assays using the anti-ERβ antibody were performed after E2 treatment of si_c-Fos, si_c-Jun, or si_control transfected cells. ChIP DNA was analyzed by real-time PCR using primers spanning six ERβ-binding regions containing AP-1 elements, as well as three ERβ-binding regions without a predicted AP-1 element as negative controls. Columns, mean fold enrichment of ERβ relative to IgG; bars, SD (n = 2). *, P < 0.05 relative to si_control.

we examined the recruitment of ERα and ERβ to identified ments decreased the recruitment of ERβ, but not ERα, at the AP-1/ERβ target DNA regions in response to ligands. Figure 6 MYC and FOXA1 binding sites. Only ERα association with shows the recruitment of ERα or ERβ to AP-1/ERβ target the VAV3 and EREG sites was markedly increased after regions for six genes, MYC, FOXA1, VAV3, EREG, CAPN2, tamoxifen treatment. Both tamoxifen and ICI increased the and LTBP1,usingChIPinMCF7ERβ cells after exposure recruitment of ERα at the CAPN2 and LTBP1 regions, but to vehicle, E2, ICI, or tamoxifen. Following E2 treatment, had no effect on ERβ recruitment. the recruitment of both ERα and ERβ to the six regions was greatly increased, but of note, we also observed significant ligand-independent recruitment of both ERs to these Discussion regions. The patterns of ERα or ERβ recruitment to these gene regions were thus similar in response to E2. However, A number of recent studies determined global binding re- ERα and ERβ recruitment exhibited gene-dependent differ- gions for ERα (14–16, 19). These studies have shown that on- ences in response to antiestrogens. ICI and tamoxifen treat- ly a minority of ERα-binding sites are located proximally to

www.aacrjournals.org Cancer Res; 70(12) June 15, 2010 5179

Zhao et al.

the gene promoters. This finding is in line with our data that riched in nuclear receptor binding regions (14, 15, 33, 35), only 3% of ERβ DNA-binding regions are within 1 kb from suggesting that nuclear receptors share some common either end of genes. A similar pattern for global DNA-binding themes in their patterns of interactions with other tran- has also been shown for androgen receptor, glucocorticoid scription factors. The importance of Forkhead transcription receptor, and vitamin D receptor binding to DNA (31–33). factors in mediating ERα binding to intact chromatin has Based on these findings, an enhancer-promoter looping been experimentally confirmed in breast cancer cell lines mechanism has been proposed for transcriptional regulation (24, 25). by nuclear receptors (34). Another possibility is that the dis- We compared the ERβ cistrome described in this study tal nuclear receptor binding sites regulate novel RNA tran- with that described by Charn and colleagues (26) and found scripts. In this study, we identified 1,457 ERβ DNA-binding only a limited overlap of ERβ-binding sites between the two regions by two independent analysis strategies with stringent data sets. It is possible that the differences between the two cutoffs. This gives a similar density of ERβ DNA-binding re- studies were due to the different MCF7 cell stocks used. gions as we have previously reported (19). When only MAT Several studies have shown that different MCF7 cell stocks with a less stringent cutoff was applied to the same data, we maintained in different laboratories are greatly diversified identified 4,766 ERβ-binding regions. A similar number of in their estrogen-induced proliferation and cytogenetic ERα-binding regions were also predicted by MAT analysis characteristics (36, 37). Differences in the ratio of ERα at a genome-wide scale (14). and ERβ levels might also have contributed to the observed Motif analysis identified several significantly enriched differences as Charn and colleagues (26) showed that the motifs including ERE-like sites, Forkhead sites, and AP-1 ratio between ERα and ERβ levels could affect the ERβ- sites. This finding is similar to previously published data binding site profile. In addition, the arrays used by Charn in which sites for the transcription factors AP-1, SP-1, Fork- and colleagues (26) to identify binding regions cover only head, Oct1, CREB, C/EBPα, and Myc were found to be en- selected regions of the genome, whereas we have used

Figure 5. A, co-occupancy of ERβ and c-Fos at ERβ-binding regions. MCF7 tet-off ERβ cells were cultured in the absence of tetracycline for 16 h and then treated with 10 nmol/L of E2 for 45 min. Columns, mean fold enrichment of ERβ/c-Fos relative to ERβ/IgG; bars, SD (n = 2). *, P < 0.05 compared with ERβ/IgG. B, ERβ-bound DNA regions containing AP-1 sites function as transcriptional enhancers. Hormone-depleted HeLa cells were cotransfected with the reporter plasmid, ERβ expression plasmid, and pRL-TK control plasmid, which contains a Renilla luciferase gene. An ERE reporter plasmid was used as a positive control, and the empty vector (Vector) was used as a negative control. Columns, mean; bars, SD (n = 3). *, P < 0.05 compared with vehicle; †, P < 0.05 compared with E2 or TPA alone. C, effects of ERβ expression and c-Fos silencing on E2-stimulated cell proliferation. Columns, mean; bars, SD (n = 2). *, P < 0.05 relative to si_Ctrl/ERα only/E2; †, P < 0.05 relative to si_Ctrl/ERα+ERβ/E2. D, effects of silencing c-Fos on FOXA1 and PKIB gene expression. Seventy-two hours after siRNA transfection, cells were treated with 10 nmol/L of E2 or vehicle for 24 h. The data are normalized to si_control/vehicle-treated conditions (mean ± SD, n = 3). *, P < 0.05 relative to si_Ctrl E2.

5180 Cancer Res; 70(12) June 15, 2010 Cancer Research

Association of AP-1 and ERβ

Figure 6. Differential recruitment of ERα and ERβ to AP-1/ERβ target DNA regions in response to ligands. MCF7 tet-off ERβ cells were cultured in the absence of tetracycline for 16 h and then treated with ligands (vehicle, 10 nmol/L E2, 1 μmol/L ICI, and 1 μmol/L tamoxifen) for 45 min. Columns, mean fold enrichment of ERα or ERβ relative to IgG; bars, SD (n = 2). *, P < 0.05 for ERα or ERβ binding versus vehicle.

arrays covering the complete genome. The use of different recruitment of ERα and ERβ. We propose that the different antibodies and possible differences in biological handling of ligand-induced recruitments of ERα and ERβ might enable cell lines likely account for some of the observed differences. ERα and ERβ to regulate unique sets of downstream target Additional studies are necessary to elucidate the discrepancy genes, as judged from a number of microarray experiments in ERβ cistromes between these two studies. (27, 38, 39). Based on in vitro investigations, a nonclassical model of We observed significant recruitment of both ERα and ERβ regulation by ER in which ER interacts with other transcrip- to chromatin in the absence of E2 (Fig. 6). Consistent with tion factors such as AP-1 and SP-1 has been proposed. How- these results, Carroll and colleagues (14) showed that when ever, no studies have examined the recruitment of ERα or 26 randomly selected ERα-binding sites were examined, ERα ERβ to AP-1 regions in intact chromatin. In this study, we was recruited to most of these sites in the absence of estro- confirmed the recruitment of c-Fos and c-Jun to 18 ERβ- gen. Our results showed that the ligand-independent recruit- binding regions (out of 25 tested sites) that contain a consen- ment of ER to chromatin was not limited to AP-1 containing sus AP-1 site. It is possible that other AP-1 family members ERβ-binding sites (Supplementary Fig. S4). In the case of the besides c-Fos and c-Jun were associated with those regions binding site in the pS2 gene, several studies have observed that did not recruit c-Fos and c-Jun. We also confirmed the that ERα and ERβ bind to this site in the absence of ligand co-occupancy of AP-1 and ERβ at AP-1/ERβ target DNA re- (19, 40–42). Binding of ER in a ligand-independent manner gions. Our results show the association of ERβ and AP-1 in could be due to the remaining estrogen in charcoal-stripped intact chromatin. fetal bovine serum. Alternatively, ER could be activated by ER subtype and gene-dependent patterns of recruit- extracellular signals such as epidermal growth factor in the ment were observed in response to antiestrogens. One ex- absence of ligands (43). planation for this observation is that the local chromatin In conclusion, our study provides novel and important landscape and histone modifications might be regulated insight into the regulation of ERβ target gene networks differently in response to different ligands, and thus, the and serves as a resource for the further elucidation of

www.aacrjournals.org Cancer Res; 70(12) June 15, 2010 5181

Zhao et al.

ERβ-regulated transcription. We provide evidence for the Acknowledgments global role of AP-1 in ERβ signaling as 60% of ERβ-interacting regions contain AP-1–like binding sites together with ERE- We are grateful to the Bioinformatic and Expression Analysis core facility at the Karolinska Institute (http://www.bea.ki.se/) for performing the Affymetrix assays. like sites and for 70% (18 out of 25 tested sites) of these, β we provide evidence for the involvement of AP-1 in ER Grant Support DNA-binding. Swedish Cancer Society (Cancerfonden) and from KaroBio AB. The costs of publication of this article were defrayed in part by the payment Disclosure of Potential Conflicts of Interest of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Jan-Åke Gustafsson is a shareholder of KaroBio AB and is a consultant for KaroBio AB and Bionovo. The other authors disclosed no potential conflicts Received 12/03/2009; revised 04/08/2010; accepted 04/08/2010; published of interest. OnlineFirst 05/25/2010.

References 1. Hall JM, Couse JF, Korach KS. The multifaceted mechanisms of ERα-andERβ-binding DNA regions. Proc Natl Acad Sci U S A estradiol and estrogen receptor signaling. J Biol Chem 2001;276: 2008;105:2604–9. 36869–72. 20. Johnson WE, Li W, Meyer CA, et al. Model-based analysis of tiling- 2. Nilsson S, Gustafsson JA. Estrogen receptor action. Crit Rev arrays for ChIP-chip. Proc Natl Acad Sci U S A 2006;103:12457–62. Eukaryot Gene Expr 2002;12:237–57. 21. Frith MC, Fu Y, Yu L, Chen JF, Hansen U, Weng Z. Detection of 3. Klinge CM. Estrogen receptor interaction with estrogen response functional DNA motifs via statistical over-representation. Nucleic elements. Nucleic Acids Res 2001;29:2905–19. Acids Res 2004;32:1372–81. 4. Kushner PJ, Agard DA, Greene GL, et al. Estrogen receptor 22. Gao H, Falt S, Sandelin A, Gustafsson JA, Dahlman-Wright K. pathways to AP-1. J Steroid Biochem Mol Biol 2000;74:311–7. Genome-wide identification of estrogen receptor α-binding sites in 5. SavilleB,WormkeM,WangF,etal.Ligand-,cell-,andestrogen mouse liver. Mol Endocrinol 2008;22:10–22. receptor subtype (α/β)-dependent activation at GC-rich (Sp1) 23. Mahony S, Benos PV. STAMP: a web tool for exploring DNA-binding promoter elements. J Biol Chem 2000;275:5379–87. motif similarities. Nucleic Acids Res 2007;35:253–8. 6. Cheung E, Acevedo ML, Cole PA, Kraus WL. Altered pharmacology 24. Carroll JS, Liu XS, Brodsky AS, et al. Chromosome-wide mapping and distinct coactivator usage for estrogen receptor-dependent of estrogen receptor binding reveals long-range regulation requiring transcription through activating protein-1. Proc Natl Acad Sci U S A the forkhead protein FoxA1. Cell 2005;122:33–43. 2005;102:559–64. 25. Laganiere J, Deblois G, Lefebvre C, Bataille AR, Robert F, Giguere V. 7. Paech K, Webb P, Kuiper GG, et al. Differential ligand activation of From the cover: location analysis of estrogen receptor α target estrogen receptors ERα and ERβ at AP1 sites. Science 1997;277: promoters reveals that FOXA1 defines a domain of the estrogen 1508–10. response. Proc Natl Acad Sci U S A 2005;102:11651–6. 8. Webb P, Nguyen P, Valentine C, et al. The estrogen receptor 26. Charn TH, Liu ET, Chang EC, Lee YK, Katzenellenbogen JA, enhances AP-1 activity by two distinct mechanisms with different Katzenellenbogen BS. Genome-wide dynamics of chromatin binding requirements for receptor transactivation functions. Mol Endocrinol of estrogen receptors α and β: mutual restriction and competitive 1999;13:1672–85. site selection. Mol Endocrinol 2010;24:47–59. 9. Umayahara Y, Kawamori R, Watada H, et al. Estrogen regulation of 27. Chang EC, Frasor J, Komm B, Katzenellenbogen BS. Impact of the insulin-like growth factor I gene transcription involves an AP-1 estrogen receptor β on gene networks regulated by estrogen recep- enhancer. J Biol Chem 1994;269:16433–42. tor α in breast cancer cells. Endocrinology 2006;147:4831–42. 10. Schmitt M, Bausero P, Simoni P, et al. Positive and negative effects 28. Williams C, Edvardsson K, Lewandowski SA, Strom A, Gustafsson of nuclear receptors on transcription activation by AP-1 of the human JA.Agenome-widestudyoftherepressiveeffectsofestrogen choline acetyltransferase proximal promoter. J Neurosci Res 1995; receptor β on estrogen receptor α signaling in breast cancer cells. 40:152–64. Oncogene 2008;27:1019–32. 11. Angel P, Baumann I, Stein B, Delius H, Rahmsdorf HJ, Herrlich P. 29. Chang EC, Charn TH, Park SH, et al. Estrogen receptors α and β 12-O-tetradecanoyl-phorbol-13-acetate induction of the human as determinants of gene expression: influence of ligand, dose, and collagenase gene is mediated by an inducible enhancer element chromatin binding. Mol Endocrinol 2008;22:1032–43. located in the 5′-flanking region. Mol Cell Biol 1987;7:2256–66. 30. Lu C, Shen Q, DuPre E, Kim H, Hilsenbeck S, Brown PH. cFos 12. Shaulian E, Karin M. AP-1 as a regulator of cell life and death. is critical for MCF-7 breast cancer cell growth. Oncogene 2005; Nat Cell Biol 2002;4:E131–6. 24:6516–24. 13. van Dam H, Castellazzi M. Distinct roles of Jun: Fos and Jun: 31. Pike JW, Meyer MB, Watanuki M, et al. Perspectives on mechanisms ATF dimers in oncogenesis. Oncogene 2001;20:2453–64. of gene regulation by 1,25-dihydroxyvitamin D3 and its receptor. 14. Carroll JS, Meyer CA, Song J, et al. Genome-wide analysis of J Steroid Biochem Mol Biol 2007;103:389–95. estrogen receptor binding sites. Nat Genet 2006;38:1289–97. 32. So AY, Chaivorapol C, Bolton EC, Li H, Yamamoto KR. Determinants 15. Lin CY, Vega VB, Thomsen JS, et al. Whole-genome cartography of cell- and gene-specific transcriptional regulation by the glucocor- of estrogen receptor α binding sites. PLoS Genet 2007;3:e87. ticoid receptor. PLoS Genet 2007;3:e94. 16. Welboren WJ, van Driel MA, Janssen-Megens EM, et al. ChIP-Seq 33. Wang Q, Li W, Liu XS, et al. A hierarchical network of transcription of ERα and RNA polymerase II defines genes differentially respond- factors governs androgen receptor-dependent prostate cancer ing to ligands. EMBO J 2009;28:1418–28. growth. Mol Cell 2007;27:380–92. 17. Papoutsi Z, Zhao C, Putnik M, Gustafsson JA, Dahlman-Wright K. 34. Kininis M, Kraus WL. A global view of transcriptional regulation Binding of estrogen receptor α/β heterodimers to chromatin in by nuclear receptors: gene expression, factor localization, and MCF-7 cells. J Mol Endocrinol 2009;43:65–72. DNA sequence analysis. Nucl Recept Signal 2008;6:e005. 18. Zhao C, Matthews J, Tujague M, et al. Estrogen receptor β2 35. Cheng AS, Jin VX, Fan M, et al. Combinatorial analysis of trans- negatively regulates the transactivation of estrogen receptor α in cription factor partners reveals recruitment of c-MYC to estrogen human breast cancer cells. Cancer Res 2007;67:3955–62. receptor-α responsive promoters. Mol Cell 2006;21:393–404. 19. Liu Y, Gao H, Marstrand TT, et al. The genome landscape of 36. Jones C, Payne J, Wells D, Delhanty JD, Lakhani SR, Kortenkamp A.

5182 Cancer Res; 70(12) June 15, 2010 Cancer Research

Association of AP-1 and ERβ

Comparative genomic hybridization reveals extensive variation 40. Metivier R, Gallais R, Tiffoche C, et al. Cyclical DNA methylation of among different MCF-7 cell stocks. Cancer Genet Cytogenet 2000; a transcriptionally active promoter. Nature 2008;452:45–50. 117:153–8. 41. Pan YF, Wansa KD, Liu MH, et al. Regulation of estrogen receptor- 37. Nugoli M, Chuchana P, Vendrell J, et al. Genetic variability in MCF-7 mediated long range transcription via evolutionarily conserved distal sublines: evidence of rapid genomic and RNA expression profile response elements. J Biol Chem 2008;283:32977–88. modifications. BMC Cancer 2003;3:13. 42. Reid G, Hubner MR, Metivier R, et al. Cyclic, proteasome-mediated 38. LinCY,StromA,LiKongS,etal.Inhibitoryeffectsofestrogen turnover of unliganded and liganded ERα on responsive promoters receptor β on specific hormone-responsive gene expression and is an integral feature of estrogen signaling. Mol Cell 2003;11: association with disease outcome in primary breast cancer. Breast 695–707. Cancer Res 2007;9:R25. 43. Curtis SW, Washburn T, Sewall C, et al. Physiological coupling 39. Secreto FJ, Monroe DG, Dutta S, Ingle JN, Spelsberg TC. Estrogen of growth factor and steroid receptor signaling pathways: estrogen receptor α/β isoforms, but not betacx, modulate unique patterns receptor knockout mice lack estrogen-like response to epidermal of gene expression and cell proliferation in Hs578T cells. J Cell growth factor. Proc Natl Acad Sci U S A 1996;93:12626–30. Biochem 2007;101:1125–47. 44. Available from: http://research.dfci.harvard.edu/brownlab/datasets/.

www.aacrjournals.org Cancer Res; 70(12) June 15, 2010 5183

Genome-Wide Mapping of Estrogen Receptor- −β Binding Regions Reveals Extensive Cross-Talk with Transcription Factor Activator Protein-1

Chunyan Zhao, Hui Gao, Yawen Liu, et al.

Cancer Res 2010;70:5174-5183. Published OnlineFirst May 25, 2010.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-09-4407

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2010/05/25/0008-5472.CAN-09-4407.DC1

Cited articles This article cites 43 articles, 13 of which you can access for free at: http://cancerres.aacrjournals.org/content/70/12/5174.full#ref-list-1

Citing articles This article has been cited by 19 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/70/12/5174.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/70/12/5174. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.