Mapping ERβ Genomic Binding Sites Reveals Unique Genomic Features and Identifies EBF1 as an ERβ Interactor

Considerable effort by numerous laboratories has resulted in an improved understanding of estrogen and SERM action mediated by the two estrogen receptors, ERα and ERβ. However, many of the targets for ERβ in cell physiology remain elusive. Here, the C4-12/Flag.ERβ cell line which stably expressed Flag.ERβ is used to study ERβ genomic functions without ERα interference. Mapping ERβ binding sites in these cells reveals ERβ unique distribution and motif enrichment patterns. Accompanying our mapping results, nascent RNA profiling is performed on cells at the same treatment time. The combined results allow the identification of ERβ target genes. Gene ontology analysis reveals that ERβ targets are enriched in differentiation, development and apoptosis. Concurrently, E2 treatment suppresses proliferation in these cells. Within ERβ binding sites, while the most prevalent binding motif is the canonical ERE, motifs of known ER interactors are also enriched in ERβ binding sites. Moreover, among enriched binding motifs are those of GFI, REST and EBF1, which are unique to ERβ binding sites in these cells. Further characterization confirms the association between EBF1 and the estrogen receptors, which favors the N-terminal region of the receptor. Furthermore, EBF1 negatively regulates ERs at the protein level. In summary, by studying ERβ genomic functions in our cell model, we confirm the anti-proliferative role of ERβ and discover the novel cross talk of ERβ with EBF1 which has various implications in normal physiology.


Introduction
Estrogens regulate homeostasis, development and reproduction by exerting their effects on a number of tissues, including the mammary gland, the brain, the cardiovascular system, the liver, the musculoskeletal system, the intestines and the immune system. These effects are mediated by one or both of the two estrogen receptors, ERa and ERb. Given the diverse roles that estrogens exert on physiology, many studies have been devoted to understanding ER biology in these tissues [1,2]. The importance of ERa in the genesis, treatment and prevention of breast cancer prevention is also well recognized, which has resulted in improved management of ERa-positive breast cancers. While ERb is also expressed in many breast tumors, its role remains elusive and controversial [3,4]. In the prostate and colon, ERb is considered the predominant ER subtype, with a potential role as a tumor suppressor [5,6]. In the brain and the ovaries, distinct functions have been reported for both ERs [7][8][9].
As members of the nuclear receptor superfamily, ERa and ERb contain several canonical functional domains, including the activation function 1 (AF1) at the N-terminus, followed by the DNA-binding domain (DBD), the hinge region, and ligandbinding domain (LBD), which also contains the activation function 2 (AF2) at the C-terminal end. Despite being encoded by different genes, ERa and ERb share significant homology. The two DBD's are almost identical (97% homology), which allows both ERs to recognize a consensus estrogen responsive element (ERE) (GGTCAnnnTGACC) with equal efficacy. The two LBD's share approximately 60% of their amino acid sequences, resulting in overlapping as well as distinct ligand recognition. The N-terminal region has the least similarity between the two ERs, which explains much of their differential roles in regulating cell physiology [10][11][12]. This region has also been suggested to regulate the proteasome-mediated degradation of ERb [13].
Because the two ERs regulate cell physiology predominantly at the transcriptional level, understanding their interactions with the genome is crucial to elucidating ER biology. Thanks to advances in DNA sequencing technologies, the global interaction between ERa and the genome has been extensively characterized. Genome-wide mapping of ERa binding events in ERa-positive MCF-7 breast cancer cells has provided significant information regarding global distribution, motif enrichment patterns and target genes, which confirm many known ERa activities, including regulation of proliferation [14][15][16].
While the roles of ERa in transcriptional regulation in response to ligand binding are well studied, much remains to be learned regarding transcriptional regulation mediated by ERb [4,17]. ERb mechanistic studies have been limited by two major challenges.
The first of these is the lack of immortalized cell lines expressing significant amounts of endogenous ERb. Although ERb expression has been reported in several established cell lines, these observations remain controversial [18][19][20][21][22]. The second challenge is the general lack of validated, specific antibodies to detect ERb in cells, tissues and tissue/cell extracts [23]. As a consequence, no genome-wide analysis of endogenous ERb action has been reported yet, to the best of our knowledge.
To circumvent these issues, ERb genomic functions have been investigated in cell lines engineered to express this receptor exogenously [19,24,25]. In MCF7 breast cancer cells that overexpress recombinant ERb, shared binding sites have been observed for both ERs. Binding sites unique to ERb have also been detected [16,19,24]. Similarly, in U2OS osteosarcoma cells engineered to express ERa, ERb or both ERs, the two ER subtypes have overlapping sets of target genes as well as distinct target genes [25,26]. In addition, the interplay between these ERs at the genomic level has also been reported in these models [16,19,24,[26][27][28][29][30]. Although the genomic functions of ERb in a physiologically relevant context remain to be determined, these studies provide significant insights into ERb actions at the genomic level. However, ERb genomic functions in the absence of ERa still require further investigation.
In order to study ERb genomic functions in the absence of ERa interference, we used MCF-7/C4-12 cells, a derivative of MCF-7 cells that has lost ERa expression [31]. Using lentiviral infection, we generated C4-12/Flag.ERb cells that stably express Flagtagged ERb. To identify true ERb target genes, ERb genomic binding sites were mapped by ChIP-seq analysis while global nascent RNA generated at the time of binding was also captured with a GRO-seq assay [32,33], Using this approach, 3166 E2mediated ERb binding sites were identified and 342 were found differentially regulated by ERb. Furthermore, a novel cross talk between ERb and Early B-cell Factor 1 (EBF1) was also identified and characterized.
Generation of the C4-12/Flag.ERb Cell Line HEK293T cells were transfected with the Flag-ERb pCDH, VSV-G, and deltaR vectors on day 1 using Lipofectamine 2000 (Life Technologies). After overnight incubation, transfecting media was replaced with normal culturing media to induce the production of lentiviruses. On day 3, media with lentivirus were collected, filtered, and used to treat MCF7/C4-12 cells [33] in the presence of polybrene. On day 4, lentiviral media was removed, treated with bleach, and discarded; the infected cells were culture in normal culturing media until they needed to be further propagated. The DNA cassette being incorporated into the host genome carried a green fluorescent protein (GFP) that was used as a selective marker of positively infected cells. Infected cells with high GFP expression were sorted and further propagated. After two rounds of sorting, resulting cell line was referred to as the C4-12/Flag.ERb, in which the expressions of GFP and Flag-ERb were detectable up to 50 passages.

Generation and Characterization of ChIP-seq Libraries
The genomic DNA samples collected from ChIP assays (both inputs and immunoprecipitation products) were processed with Illumina ChIP-seq Sample Preparation Kit. Samples were then sequenced with Illumina Genomie Analyzer II and aligned to hg18 (High-throughput Genomics Analysis Core Facility, University of Chicago, Chicago IL). QuEST [34] was used as the peakcalling software, using default parameters recommended to analyze transcription factor ChIP-seq data. All ChIP-seq data is deposited in the Gene Expression Omnibus (GEO) database at National Center for Biotechnology Information (accession number GSE48161).
The global distribution of ChIP-seq peaks was analyzed using the CEAS package [35]. Distribution around transcription start sites (TSS) was analyzed using in-house algorithms. Enrichment of transcription factor binding motifs was analyzed with CLOVER [36] and the JASPAR public database.
Global Run-on followed by Sequencing (GRO-seq) GRO-seq was performed as previously described [32] with limited modifications. Briefly, after 1 hour of E2 or Vehicle treatment, nuclei from C4-12/Flag-ERb cells were extracted and processed with nuclear run-on assay. The nascent RNA products were ligated to adaptors prior to reverse transcription reaction. The resultant cDNA libraries were then sequenced using Illumina HiSeq2000 (Center for Genome Research and Biocomputing at Oregon State University, Corvallis OR). Sequencing reads were analyzed as described previously [32]. In order to identify E2regulated genes, we focused on RefSeq-annotated genes, counting reads in a fixed window between +1 kb and +13 kb relative to the transcription start site of each gene, so as to avoid possible complications introduced by paused polymerases [37], and to allow easy side-by-side comparison among samples. The normalized expression value of 1e-5 (normalized against the sample total read counts) was used as the threshold to select genes for further analyses. Comparing E2 treated samples versus vehicle treated samples, genes with FC.1.2 were considered upregulated; those with FC,0.8 were considered downregulated. All GRO-seq data is deposited in the Gene Expression Omnibus (GEO) database at National Center for Biotechnology Information (accession number GSE48161), and the scripts are available upon request.
Immunoprecipitation MCF7/C4-12 cells were transfected in 10 cm plates with Flag-ERb and V5-EBF1 plasmids using FuGENE. Transfected cells were then treated with E2 or Vehicle for 4 hours. Cells were lysed in CoIP-lysis buffer (10 mM Tris pH 7.4, 150 mM NaCl, 0.5% NP40, PICIII) followed by brief sonication. The cell lystate was incubated with 2 ug anti-Flag M2 antibody overnight at 4uC. The complex of interest was precipitated with 30 uL Dynabeads Protein G (Life Technologies) pre-blocked with BSA for 1 hour at 4uC. After the beads were washed, the protein complexes were eluted with SDS loading buffer and then subjected to Western Blot analysis. Three independent experiments was performed. The results of a representative experiment are shown.

Luciferase Reporter Assay
MCF7/C4-12 cells were transfected in 48-well plates with 3xERE-Luc, pRL-TK, ERa or ERb, and pcDNA or interactor plasmids using FuGENE. Transfected cells were then treated with E2 or Vehicle for 24 hours. Cells were lysed in Passive Lysis Buffer (Promega) supplemented with PICIII and 1 mM DTT. Lysate was then transferred to 96-well plate for the Dual Luciferase Reporter assays (Promega). Samples and treatments were quadruplicated in each experiment. Statistical analysis was performed using the unpaired Student's t test; p,0.05 was considered significant. Three independent experiments was performed. The results of a representative experiment are shown as mean 6 SD.

RT-qPCR
C4-12/Flag.ERb cells were transfected in 6-well plates with pcDNA or EBF1 plasmids using FuGENE (Promega), then treated with E2 or Vehicle for 2 hours. RNA was extracted using Trizol (Life Technologies). cDNA was generated using the Hi Capacity RNA-to-cDNA kit (Life Technologies). qPCR was performed using Fast SYBR Green Master Mix (Life Technologies). GAPDH was used as normalization control. Samples and treatments were triplicated in each experiment. Statistical analysis was performed using the unpaired Student's t test; p,0.05 was considered significant. Three independent experiments was performed. The results of a representative experiment are shown as mean 6 SD.

Cell Proliferation Assay
C4-12/Flag.ERb cells were transfected in 24-well plates with pcDNA or EBF1 plasmids using FuGENE (Promega). Cells were treated with E2 or Vehicle for 3 days. Cell confluency, used to quantify cell growth, was measured in an Incucyte FLR live content imaging system (Essen Bioscience). Samples and treatments were triplicated in each experiment. Statistical analysis was performed using the unpaired Student's t test; p,0.05 was considered significant. Three independent experiments was performed. The results of a representative experiment are shown as mean 6 SD.

C4-12/Flag.ERb as a Cell Model to Study ERb Genomic Functions
Due to the high homology between the DNA-binding domains of the two ERs, ERb genomic functions should be investigated in a system without ERa interference. For this reason, we used MCF-7.C4-12 cells, a derivative of the MCF-7 breast cancer cell line that no longer expresses any detectable estrogen receptors, as our cell model [31,33].
To overcome the lack of good ERb antibodies for ChIP experiments, a Flag-tagged version of ERb was used, which allowed the specific detection and immunoprecipitation of this nuclear receptor in subsequent experiments.
As shown in figure 1A, Flag.ERb is stably expressed in C4-12/ Flag.ERb cells at a low level compared to transiently transfected cells (because of low expression level, the ERb band in the input lane was faint). The fusion protein could be immunoprecipitated with anti-Flag M2 antibody, confirming the stable expression of ERb in these cells. An ERE-luciferase reporter assay was used to demonstrate that Flag.ERb is functional in these cells. As shown in figure 1B, luciferase activity is up regulated upon E2 treatment, indicating the expression of functional ERb.
In summary, we successfully generated a C4-12/Flag.ERb cell line in which Flag.ERb is stably expressed and functional. The  ERb binding sites were converted to 400-base long polynucleotides, centered at the binding summit. CLOVER motif analysis [36] was used to identify enrichment of transcription factor binding motifs in this library. JASPAR transcription factor binding motif database was used as the library of known motifs; CpG island, promoter regions and chromosome 20 were used for background analysis. (D) Enrichment of motifs known to associate with ER binding sites. (E) Enrichment of motifs unique to ERb binding sites in C4-12/Flag.ERb cells. doi:10.1371/journal.pone.0071355.g003 E2 induction. No binding sites were found in vehicle treated cells. Several of these binding sites were chosen at random for validation with ChIP-qPCR. As shown in figure 2, Flag.ERb is recruited to these genomic loci in an E2-dependent manner.
As shown in figure 3A, a small percentage of ERb binding sites overlap with previously reported ERa binding sites in MCF-7 cells [14,15]. Global distribution analysis showed that ERb exhibited a pattern similar to ERa in MCF-7 cells: 14% of ERb binding sites were found at the proximal promoter regions (defined as the regions within 1 kb of any transcription start site); most binding events occurred in intergenic regions or within introns ( Figure 3B). ERb binding sites also exhibited high density surrounding the TSS (figure 3C).
ERb binding sites were also analyzed for enriched transcription factor binding motifs using CLOVER [36]. The estrogen responsive element (ERE) and half-ERE were found to be the most enriched in our data. AP1, AP2, FOXA, FOXO, CREB and GATA binding motifs were also enriched at ERb binding sites ( figure 3D). Interestingly, we also detected GFI, REST and EBF1 binding motifs, which had not been previously reported to be associated with ERa genomic binding sites (figure 3E).

Profiling Nascent Transcripts of ERb Target Genes
After C4-12/Flag.ERb cells were treated with E2 for 1 hour (Vehicle treated cells were used as negative control), nascent RNA was collected and profiled by GRO-seq analysis. The list of genes differentially regulated by E2 induction was combined with genes associated with ERb (defined as those with at least 1 ERb binding site within 210 kb of TSS and +10 kb of TTS) to generate a list of In order to validate results obtained from our GRO-seq assay, mRNA was collected after 2 hours of E2 induction. A few randomly selected transcripts were analyzed with RT-qPCR. As shown in figure 4C-D, the majority of these transcripts were validated using this approach. Figure 4E shows ERb ChIP and GROseq signals at CDKN1A and KLF10, two representative ERb targets. These genes had ERb binding close to the promoter regions and were upregulated upon E2 treatment ( Figure 4C, 4E).
Gene ontology analyses of ERb target genes revealed specific enrichments in transcription regulation, metabolic processes, differentiation, development and apoptosis (figure 4A). Because the same gene ontology categories were found enriched in both up-and down-regulated groups, there is likely a potential molecular switch in response to E2 via ERb. The fact that differentiation, development and apoptosis categories were enriched in ERb target genes suggests that ERb is antiproliferative in this cell model. To test this hypothesis, a cell proliferation assay was performed on C4-12/Flag.ERb cells treated with E2 (or Vehicle as negative control). In agreement with the gene ontology analyses, E2 treatment significantly suppressed C4-12/Flag.ERb cell proliferation, further supporting the hypothesis that ERb can function as an anti-proliferative factor (figure 5).

Association between ERb and EBF1
Because the EBF1 binding motif was enriched at ERb binding sites, the interaction between EBF1 protein and a few ERb binding sites was investigated. When EBF1 was transiently expressed in C4-12/Flag.ERb cells, it was recruited to several ERb binding sites in an E2 dependent manner ( Figures 6A). Moreover, the recruitment of ERb at these sites was enhanced in the presence of EBF1 and E2 (figure 6B).
We then used co-immunoprecipitation to test for an interaction between these two proteins. As shown in figure 6C, V5-tagged EBF1 was detected in the ERb immunoprecipitate, indicating an interaction between the two proteins. However, when V5-EBF1 was immunoprecipitated with V5 antibody, no ERb was detected (data not shown). These results suggest a stoichiometry in which most of the ERb interacts with EBF1, while only a small portion of EBF1 interacts with ERb, indicating that ERb levels are limiting. Notably, cells used in these assays were treated with MG132 because ERb levels were very low ( figure 6C).
Because the presence of EBF1 correlated with low ERb protein levels, we wished to test the influence of EBF1 on ERb protein levels and function. Due to the low ERb expression in stably transfected C4-12/Flag.ERb cells, transiently over-expressed ERb was assayed in the presence and absence of EBF1. As shown in figure 7A, EBF1 significantly reduced ERb protein stability while ERb transcript levels remained unchanged ( Figure 7B), indicating that EBF1 regulates ERb stability at the protein level. ERb transcriptional activity was also suppressed, as measured by an ERE-luciferase reporter assay ( figure 7C). This suppressive effect was also observed for endogenous target genes. Thus, when C4-12/Flag.ERb cells were transfected with EBF1, ERb target gene expression in response to E2 was significantly reduced (figure 7D).
To investigate the role of this cross talk at the phenotypic level, we asked if EBF1 could influence the anti-proliferative effects of ERb. C4-12/Flag.ERb cells were transiently transfected with EBF1 and assayed for cell proliferation. As shown in figure 7E, EBF1 significantly suppressed the proliferation of these cells, in agreement with the previously suggested concept that EBF1 is a tumor suppressor [38]. Interestingly, the proliferation of cells transfected with EBF1 became ligand independent.
Altogether, these results confirm that EBF1 negatively affects ERb protein stability, which in turn down regulates ERb transcriptional and phenotypic effects.

EBF1 Differentially Regulates ERa and ERb
While EBF1-ERb interaction attenuates ERb protein stability, it is unclear if this interaction is ERb specific. We therefore tested the relationship between EBF1 and ERa. In cells transiently transfected with both EBF1 and ERa, as shown in figure 8A, EBF1 is found in the ERa immunoprecipitate, indicating an interaction between the two proteins. This interaction correlates with a decrease in ERa protein levels ( Figure 8B). EBF1 also suppresses ERa downstream activity, as measured by reporter assay (figure 8C). Although EBF1 expression affects both ERa and ERb stability, the suppressive effect was more dramatic on ERb, indicating a differential regulation of the two estrogen receptors by EBF1.
To further investigate this differential regulation of ERa and ERb by EBF1, we compared EBF1-mediated attenuation of EREluciferase activity of four different ER constructs: full-length ERa, full-length ERb, a chimeric construct with ERa N-terminal domain followed ERb C-terminal domain (ERa/b), and a chimeric construct with ERb N-terminal domain followed by ERa C-terminal domain (ERb/a) ( figure 8C). This approach was chosen, instead of investigating endogenous gene expression levels, so that all samples had the same elements except for those being tested; moreover, variations due to genomic interference were eliminated. Moreover, the identities of common genes being equally regulated by either ERalpha or ERbeta remains elusive and controversial. Therefore, the ERE-Luc reporter assay was an appropriate assay to investigate this cross talk. According to figure 8C, even though EBF1 significantly suppressed the activity of all four constructs, the magnitude of suppression was not the same. In agreement with the observation that EBF1 decreases ERb protein expression more dramatically than ERa, the transcriptional activity of ERb was also more reduced in the presence of EBF1 compared to ERa (figure 8C). Interestingly, constructs with ERb N-terminal domain (full-length ERb and ERb/a) were more affected by EBF1 than those with the ERa Nterminal domain. These results indicate that the N-terminal domains of the two receptors are responsible for the differential regulation of ERa and ERb by EBF1.

Discussion
Mapping ERa genomic binding sites in MCF7 cells [14] is considered a major milestone in ER research, with significant implications in understanding normal biology and pathophysiology. In particular, the ERa genomic landscape helped identify ERa direct target genes when used in combination with gene expression experiments. Moreover, these findings also revealed distinct ERa binding patterns, which were the basis for further investigations of ERa participation in chromatin looping as well as the identification of ERa binding partners [14,16,39]. ERa genomic binding signatures have also been associated with differential clinical outcomes in breast cancer [40].
Our understanding of ERb function is still very limited compared to that of ERa. Even though the presence of ERb in diverse tissues and cancers has been demonstrated [12,41], the expression of this receptor in cultured cell lines remains controversial. This is largely due to the lack of reliable antibodies to detect ERb protein [4,23]. These limitations have hindered investigations of ERb function.
To circumvent these challenges and to study ERb genomic function, we generated a cell line (C4-12/Flag.ERb cells) that maintains stable Flag-ERb expression. Because these cells were derived from the MCF7/C4-12 cells, which no longer express detectable ERs, ERb functions could be studied without ERa interference. The Flag epitope allowed us to detect and immunoprecipitate ERb with high efficiency and specificity. Figure 6. The association between ERb and EBF1. C4-12/Flag.ERb cells were transfected withV5-tagged EBF1 and incubated for 1 hr with 10 nM E2. Cells were then treated with 1% formaldehyde and processed by ChIP assay using (A) anti-V5 antibody and (B) anti-Flag M2 antibody. (C) C4-12/Flag.ERb cells were transfected with V5-tagged EBF1, treated with 10 nM E2, lysed and subjected to co-immunoprecipitation using anti-Flag M2 antibody to immunoprecipitate complexes containing Flag.ERb. Anti-Flag M2 antibody and anti-V5 antibody were used on Western blots to detect V5.EBF1 and Flag.ERb, respectively. doi:10.1371/journal.pone.0071355.g006 Using this approach, we mapped ERb to 3166 genomic sites in cells treated with 10 nM E2 for 1 hour. The 45-minute-to-1-hour window of E2 treatment time has been reported to be the first peak of ERa-chromatin binding in MCF7 cells and thus represents ER early transcriptional activities in response to hormone activation [14][15][16]19,24]. At this time point, most ERb binding sites were found in intronic or intergenic regions. Motif analyses of these regions revealed enrichments for many transcription factors that have been shown to interact with ERa. Moreover, motifs that have not previously been associated with ERa binding sites are also enriched, suggesting distinct genomic characteristics of ERb binding sites.
Interestingly, no ERb-chromatin binding events were detected in C4-12/Flag.ERb cells treated with Vehicle (Ethanol). In a study using U2OS cells that express ERb via a tet-off system (U2OS/ERb cells), apo ERb mapped to many genomic loci [25].  The difference between these findings is likely due to the different methods of expressing ERb. In our stable cell model, ERb expression is very low and at equilibrium, resulting in no detectable ERb-chromatin binding in the absence of ligand. The induced, transient expression of ERb in U2OS/ERb cells might yield constitutively active ERb if the amount of expressed protein exceeds the capacity of endogenous chaperones to keep ERb inactive in the absence of ligand. In addition, the difference between their results and ours could also reflect inherent cellspecific differences between the cell models.
The functionality of ERb binding sites was addressed by using a GRO-seq assay to profile nascent RNA generated at the time of ERb-chromatin binding [32]. Conventionally, in order to study the genomic functions of ERs in response to E2, gene expression profiling is performed a few hours after mapping the binding events, allowing mRNA processing to reach completion [14,19,24,25,32]. Although this approach has been applied to the study of ER target genes, the association between ER binding events and transcriptional regulation is not direct because samples are harvested and measured at different times. Here, we combined ChIP-seq and GRO-seq to map ERb-chromatin interaction sites and to profile actively transcribed genes at the time of binding events, respectively, to assess the functionality of ERb binding sites as well as to identify true ERb target genes.
According to our ChIP-seq results, most of the ERb binding sites identified in our study do not overlap with ERa sites in MCF7 cells (less than 20%) [14,15,19]. This observation indicates that ERb has a distinct set of target genes, in agreement with previously published studies [16,19,24]. According to our GROseq results, in C4-12/Flag.ERb cells, ERb target genes are enriched in differentiation, development and apoptosis pathways. Our results are consistent with the observation that ERb elicits an antiproliferative effect on MCF-7 C4-12/Flag.ERb cells in response to E2 stimulation.
It is important to notice that the MCF7 cells and the MCF7/ C4-12 cells are not the same due to the lack of ERa expression in the latter. Because ERa is a major transcription factor that regulates a wide array of cellular processes, this difference may lead to differences in cellular physiology [33]. This could contribute to the different genomic landscapes of ERa in MCF7 and ERb in C4-12/Flag. ERb cells. However, the MCF7/C4-12 cells were derived from the MCF7 cells, and thus the most related to the MCF7 line. On the other hand, the lack of ERa expression would allow the investigation of ERb genomic function without the interference from the former receptor. The more desirable comparison would be between ERb landscape in C4-12/Flag. ERb cells and ERa landscape in C4-12 cells stably expressing ERa at similar level. However, the investigation of ERa genomic function in such cell line would be beyond the scope of this study. Further studies are required to address this issue.
The global distribution of ERb sites in C4-12/Flag.ERb cells is similar to that of ERa in MCF7 cells, with several ERb-unique features. There are more ERb binding events in the proximal promoter (.13% versus ,7%) and distal intergenic regions (.42% versus 23%) than for ERa in MCF-7 cells [15]. Furthermore, ERb binding sites exhibit high density proximal to the TSS of target genes, which is not seen for the ERa genomic landscape in MCF-7 cells [14,24]. Our results are consistent with other studies that have mapped ERb binding sites in MCF-7 cells expressing ERb [24], which suggests that this behavior is an ERbunique feature. However, it is not yet known whether this distribution pattern reflects endogenous ERb behavior.
Motif analyses further showed similarities as well as distinct characteristics between ERb binding sites and those of ERa. Similar to ERa, most ERb binding sites are enriched in ERE, AP1, AP2, FOXA1, CREB, and GATA motifs, indicating the similarity of ERa and ERb functional patterns at the genomic level. Similar results have been reported in other ERb mapping studies [16,24]. In addition, several transcription factor binding motifs that are not associated with ERa are enriched at ERb binding sites, including binding motifs for GFI1, REST, and EBF1. These differences between ERa and ERb genomic landscapes suggest that ERb target genes likely have different promoter composition and/or structures compared to ERa target genes.
It was interesting to find the EBF1 binding motif enriched in ERb binding sites. Early B-cell Factor 1 is a crucial transcription factor that drives the maturation of B-cell development. Even though estrogens have been suggested to influence the immune system, an association between EBF1 and estrogen signaling has not been reported. Our finding is the first to suggest a cross talk between EBF1 and ERb. In addition, the ERE motif was also found in EBF1 binding sites (unpublished data). This association was further validated when EBF1 was co-immunoprecipitated with ERb. However, ERb was not detected in EBF1 immunoprecipitated samples, indicating that the stoichiometry of this association was not one to one. While most of the expressed ERb associated with EBF1, only a small fraction of EBF1 was involved, indicating that ERb is limiting.
EBF1 over-expression correlated with down regulation of ERb protein levels. Further experiments revealed that EBF1 negatively regulates both ERs, although the effect was not of the same magnitude for ERa. These observations suggest that EBF1 might be involved in hormone resistant breast cancer. ERa and ERb have both shared and distinct roles in breast cancer biology, some of which might be antagonistic. One could imagine that EBF-1 functions to differentially modulate the balance between ERa and ERb activities in breast cancers that express both ER subtypes, resulting in diverse transcriptional and phenotypic consequences.
Additionally, among four ER constructs (ERa, ERb, ERa/b, and ERb/a), those carrying the N-terminal domain of ERb were more sensitive to EBF1, suggesting that the association between EBF1 and both ERs involved the N-terminal region of ER. These results are consistent with the suggestion that the N-terminal region is involved in ERb degradation [13].
Because EBF1 over-expression correlated with the down regulation of ERb, when EBF1 was transiently expressed in C4-12/Flag.ERb cells, these cells became insensitive to hormone treatment. EBF1 exogenous expression also suppressed C4-12/ Flag.ERb cell proliferation. This effect is likely to be independent of ER because EBF proteins have been proposed to have antiproliferative or tumor-suppressive effects [38,42], Therefore EBF1 over expression may suppress C4-12/Flag.ERb cell growth independent of its influence on ERb (figure 9).
Even though our findings are based on engineered cell lines, opposing roles for EBF1 and ER have been reported in normal physiological processes, such as adipogenesis. While ERs suppress adipogenesis, EBF1 promotes the differentiation of adipocytes [7,38,[43][44][45][46][47][48][49][50][51][52][53]. Our observed cross talk between EBF1 and estrogen signaling suggests a connection between the two differentiating mechanisms: EBF1 suppresses the expression and activity of ER, which consequently facilitates adipogenesis by up regulating adipogenic genes as well as by releasing PPARc from the inhibition exerted by estrogen signaling [51,52]. Further investigations will be needed to validate this hypothesis.
Another example of the opposing roles EBF1 and ER is found in the differentiation of B cells. EBF1 is a required factor that drives of B cell differentiation to completion. Estrogens, however, inhibit differentiation [7,[54][55][56][57][58][59]. Our results support a possible mechanistic explanation for the opposing roles of EBF1 and ERs in lymphopoiesis. In this model, while EBF1 directly promotes B lymphocyte differentiation, it also mediates the degradation of the two estrogen receptors. By effectively suppressing these potent inhibitors of B cell production, this behavior might represent yet another mode of EBF1 action to promote B cell maturation. Further investigations will be needed to validate this hypothesis.
In summary, with the mapping ERb genomic binding sites in C4-12/Flag.ERb cells, we have identified features of ERb genomic functions that are distinct from those of ERa and from other reported actions of ERb. In our cell model, ERb is recruited to and regulates a unique set of genes, some of which suppress cell proliferation. Our analyses also reveal cross talk between ERb and EBF1. We demonstrate that EBF1 is a negative regulator of both ERa and ERb. This antagonistic relationship between EBF1 and ERs could explain their opposing roles in different physiological processes. Future experiments will be required to characterize this association in physiologically relevant contexts.