Estrogen Signaling via Estrogen Receptor β*

Estrogens act by binding to and activating two estrogen receptors (ERs), ERα and ERβ. Transcriptional regulation by ERs is controlled by a complex array of factors such as ER-ligand binding, the DNA sequence bound by ERs, ER-interacting cofactors, and chromatin context. This minireview will provide an overview of the most recent advances in the identification of ERβ-regulated target gene networks and ERβ DNA-binding sites. We also highlight the recent work establishing new roles of ERβ signaling, including protective functions in the epithelial-mesenchymal transition and in atherosclerosis, as well as regulation of cell proliferation in the colon.


Estrogens act by binding to and activating two estrogen receptors (ERs), ER␣ and ER␤. Transcriptional regulation by
ERs is controlled by a complex array of factors such as ER-ligand binding, the DNA sequence bound by ERs, ER-interacting cofactors, and chromatin context. This minireview will provide an overview of the most recent advances in the identification of ER␤-regulated target gene networks and ER␤ DNAbinding sites. We also highlight the recent work establishing new roles of ER␤ signaling, including protective functions in the epithelial-mesenchymal transition and in atherosclerosis, as well as regulation of cell proliferation in the colon.
Estrogens, the main female sex steroids, control many cellular processes, including growth, differentiation, and function of the reproductive systems. In the non-pregnant female, estrone and estradiol (E 2 ) 2 are the two main forms of estrogens, whereas estriol is the main estrogen in pregnancy. Estrogens interact with two estrogen receptors (ERs), ER␣ and ER␤, and exert their effects through diverse signaling pathways that mediate genomic and nongenomic events, resulting in tissue-specific responses. Estrogen-regulated gene expression is controlled by a complex array of factors such as ERligand binding, the DNA sequence bound by ERs, ER-interacting cofactors, and chromatin context. The transcriptional responses to estrogen signaling depend on ligand identity and availability, the cellular concentration and localization of ERs, levels of various coregulator proteins and other signal transduction components, and the chromatin state (1). The discovery of a second ER, ER␤, in 1996 (2) prompted renewed efforts to investigate the mechanisms of action of estrogenic molecules. There is now compelling evidence that ER␤ is involved in various types of cancer (breast, ovarian, colorectal, prostate, and endometrial), in bone and brain physiology, and in the cardiovascular system and inflammation (3,4). Recently, global analysis of gene expression profiles and identification of protein-DNA interactions have begun to reveal the molecular architecture of ER␤ binding to DNA and the subsequent effects on gene regulatory networks. In this minireview, we will discuss the current knowledge of gene regulatory networks influenced through ER␤, as well as several novel discoveries pertaining to roles of ER␤ in epithelial-mesenchymal transition (EMT), atherosclerosis, and the colon.

ER␤ Gene and Its Protein Variant ER␤2
ER␤ is a member of the nuclear receptor superfamily and shares common structural characteristics with the other members of this family, including five distinguishable domains denoted A-F ( Fig. 1) (5). The human ER␤ gene (ESR2) is located on chromosome 14q23.2, spanning ϳ61.2 kb. The ER␤ protein is produced from eight exons. Additionally, there are two untranslated exons, 0N and 0K, in the 5Ј-region and an exon at the 3Ј-end that can be spliced to exon 7 to produce the alternative ER␤ isoform, ER␤2 (also called ER␤cx) (6). Thus, ER␤2 has a unique C terminus, where the amino acids corresponding to exon 8 are replaced with 26 unique amino acids. The full-length human ER␤ (also named ER␤1) protein includes 530 amino acids with an estimated molecular mass of 59.2 kDa, whereas ER␤2 encodes a protein of 495 amino acid residues with a predicted molecular mass of 55.5 kDa. ER␤2 has undetectable affinity for E 2 and other tested ligands. ER␤2 was suggested to be a dominant-negative inhibitor of ER␣ (6). Further mechanistic study revealed that ER␤2 induces proteasome-dependent degradation of ER␣, leading to suppression of ER␣ signaling (7). Although additional mRNA isoforms of ER␤ arising from differential splicing have been described, only ER␤2 has been identified at the protein level (8,9).

Mechanisms of ER Signaling
Upon ligand activation, ERs can regulate biological processes by divergent pathways (Fig. 2) (1). The so-called classical signaling occurs through direct binding of ER dimers to estrogen-responsive elements (EREs) in the regulatory regions of estrogenresponsive genes, followed by recruitment of coregulators to the transcription start site. The consensus ERE consists of a 5-bp palindrome with a 3-bp spacer: GGTCAnnnTGACC. However, many natural EREs deviate substantially from the consensus sequence (10). Estrogen also modulates gene expression by a second mechanism in which ERs interact with other transcription factors such as AP-1 (activating protein-1) and Sp-1 (stimulating protein-1) through a process referred to as transcription factor cross-talk. Furthermore, estrogen may elicit effects through nongenomic mechanisms, which occur much more rapidly. This action has been shown to involve the activation of downstream cascades such as PKA, PKC, and MAPK via membrane-localized ERs. Recently, an orphan G protein-coupled receptor (GPR30) in the cell membrane was reported to mediate nongenomic estrogen signaling. Subsequent studies by others demonstrated that the activities of GPR30 in response to estrogen were through its ability to induce expression of ER␣36, a novel variant of ER␣, and that in turn, ER␣36, acted as an extranuclear ER to mediate nongenomic estrogen signaling (11). It is still possible that addi-tional membrane receptors for estrogen are involved in mediating nongenomic estrogen action. The mechanistic details of activation through nongenomic pathways such as target genes remain to be characterized.
ERs can also be activated by extracellular signals in the absence of ligand. Growth factor signaling or stimulation of other signaling pathways leads to activation of kinases that can phosphorylate and thereby activate ERs or associated coregulators in the absence of ligand. As an example, the HER2 downstream signaling molecules ERK1 and ERK2 can phosphorylate ER, leading to ligand-independent receptor activation (12). The biological significance of this ER signaling remains unclear.

Genome-wide Profiling of ER␤ Gene Expression Programs
There have been a number of studies in the past few years aimed at comprehensively unraveling the complete estrogenregulated gene expression programs in cancer cells. These reports can be attributed to the introduction of microarrays for global gene expression profiling. DNA microarray technology allows quantitative monitoring of changes in the ex-pression of thousands of genes simultaneously and has been described in several configurations, including oligonucleotide arrays and microarrays of cDNAs spotted on glass slides. During the past few years, the development of high-throughput DNA sequencing (HTS) methods for global gene expression profiling, also known as "RNA-Seq," has challenged microarray technology because of its superior capability for detection of genes expressed at low levels, alternative splice variants, and novel transcripts (13). However, to our knowledge, no studies that explore HTS to assay genome-wide transcriptional regulation by estrogen have been reported.
Several reports have described global gene expression profiles in ER␣-expressing breast cancer cell lines in response to E 2 treatment (14). The available studies have reported different numbers of E 2 /ER␣-regulated genes in MCF-7 breast cancer cells, ranging from ϳ200 to ϳ1500. These discrepancies may be attributed to differences in the length of the E 2 treatment, application of different microarray platforms, and different analysis strategies (14). Two studies that aimed to identify E 2 /ER␣ direct targets by short-term E 2 treatment (3 h) in  MCF-7 cells identified similar numbers of E 2 target genes. In one of the studies, 122 genes were identified as stimulated by E 2 , and 95 genes were identified as inhibited by E 2 (15). In the other study, 134 genes were up-regulated, and 141 genes were down-regulated after E 2 treatment (16). However, a comparison of E 2 -regulated genes between the studies has not been reported. Overall, gene expression profiling and candidate gene analysis have identified several well known estrogenregulated genes in breast cancer cells such as TFF1, CCND1, IGFBP4, C3, ADORA1, GREB1, and MYC. Furthermore, gene expression profiling has identified categories of ER␣-regulated genes, including those that modulate the cell cycle, transcriptional regulation, morphogenesis, and apoptosis, compatible with a role of estrogen in inducing ER␣-expressing breast cancer cell proliferation and survival (17).
With regard to identification of genes regulated by ER␤, gene expression studies have been performed mainly in cancer cell lines stably expressing ER␤ either alone or together with ER␣. This is due to the lack of immortalized cell lines expressing high levels of endogenous ER␤. The literature is still contradictory regarding which cell lines express endogenous ER␤ mRNA and/or protein. For example, although MCF-7 cells are generally considered to be ER␤-negative, some studies reported that MCF-7 cells express endogenous ER␤ (18,19). Thus, the challenge now is to identify cell lines widely accepted to express endogenous ER␤ also in the absence of ER␣. Studies examining gene expression profiling in ER␣-positive breast cancer cell lines stably expressing ER␤ have provided insights into the interplay between ER␣ and ER␤ for gene regulation (20 -23). In these studies, several common features were observed: (i) ER␣ and ER␤ share some target genes, although each receptor also appears to have distinct sets of downstream target genes; (ii) coexpression of ER␤ with ER␣ significantly impacts the E 2 -induced transcriptional response by ER␣; and (iii) expression of ER␤ inhibits E 2 /ER␣induced cell proliferation. These studies have also provided new mechanistic insights into the suppressive effect of ER␤ on estrogen-stimulated cell proliferation. For example, Williams et al. (24) reported that, of the categories of genes down-regulated by ER␤, the "regulation of cell proliferation" category was the most overrepresented one. Chang et al. (20) showed that ER␤ regulated multiple components of TGF␤ signaling, consistent with the observations that TGF␤ is normally associated with the suppression of breast cancer cell proliferation. Additional experiments such as time course studies of ER␤-induced genes combined with examination of ER␤-DNA binding by ChIP will help to identify direct ER␤ target genes.
ER␤-specific effects on gene expression have been investigated in three different cell lines lacking expression of endogenous ER␣ and ER␤, namely U2OS (25), HEK293 (26), and Hs578T (23) cells. Of the 76 ER␤-regulated genes, only 17 genes were commonly regulated by both ER␣ and ER␤, suggesting that the transcriptional effects of E 2 via ER␣ or ER␤, respectively, are largely distinct in U2OS cells. After a 24-h E 2 treatment, 61 and 95 genes were identified as ER␤-regulated genes in HEK293 and Hs578T cells, respectively, as judged by a Ͼ2-fold induction in response to ER␤ activation. However, there were only three genes (PTGER4, ENPP2, and DKK1) commonly regulated in both HEK293 and Hs578T cells, suggesting that ER␤ evokes distinct gene responses in different types of target cells. However, some of these discrepancies may be attributed to different expression levels of ER␤ and differences in the array designs. Despite the differences, both studies reported inhibition of cell proliferation by ER␤ expression independently of ER␣, suggesting a similar function of ER␤ in different cell types. Further studies are needed, however, to clarify molecular mechanisms by which ER␤ elicits inhibitory effects on cell proliferation.

Global Identification of ER␤ DNA-binding Regions
ChIP is a powerful method for studying protein-DNA interactions in vivo. In recent years, the development of wholegenome analyses by combining the ChIP assay with highthroughput genomic technologies has enabled researchers to gain new insight into interactions between ERs and regulatory networks contributing to gene regulation. The currently available ChIP-based methods for examining ER binding include ChIP-chip and ChIP-DSL (DNA selection and ligation), based on hybridization, or ChIP-PET and ChIP-Seq, based on HTS. These global studies of ER-DNA binding have revealed that ER-binding sites can be located at a large distance from the proximal promoter region of genes. Based on these findings, an enhancer-promoter looping mechanism has been proposed for transcriptional regulation by ERs (27). Very recently, Fullwood et al. (28) used ChIA-PET (chromatin interaction analysis by paired-end tag sequencing) to demonstrate that most distal ER␣-binding sites are anchored at gene promoters through long-range chromatin interactions, suggesting that chromatin interactions constitute an important mechanism for gene regulation.
Although ER␣ DNA-binding regions have been extensively profiled, ER␤ DNA-binding regions have been less well characterized. To date, three studies examined ER␤-binding sites in ER␣-positive cells engineered to express ER␤ (29 -31). All of these studies used the MCF-7 cell line and the ChIP-chip platform. Charn et al. (29) examined the localization of ER␣ and ER␤ DNA-binding regions in MCF-7 cells engineered to express one or both ERs and in response to E 2 . They identified a higher number of sites bound by ER␣ (4405 sites) than by ER␤ (1897 sites), but the majority of the ER␤ sites (73%) also bound ER␣ when each of the two ER subtypes was present alone, consistent with a model in which ER␣ and ER␤ can recognize the same ERE motif. However, fewer sites (33%) were shared when both ERs were present, suggesting a competition between the ER subtypes in their selection of DNA-binding sites.
Recently, our group (31) described 1457 high-confidence ER␤-binding sites on a genome-wide scale in MCF-7 cells using the ChIP-chip approach. Interestingly, ϳ60% of the genomic regions bound by ER␤ contained both AP-1-like binding regions and ERE-like sites, suggesting a functional interaction between AP-1 and ER␤ signaling. Furthermore, we demonstrated co-occupancy of ER␤ and AP-1 on chromatin and that siRNA-mediated knockdown of c-Fos or c-Jun expression decreased ER␤ recruitment to chromatin. These results suggest that the AP-1 transcription factor collaborates with ER␤ in mediating estrogen responses in breast cancer cells.
ER␣ and ER␤ exhibit distinct as well as overlapping functions at the level of DNA binding. In a study by Liu et al. (30), a high degree of overlap between ER␣and ER␤-binding sites was found. However, the regions bound by ER␣ had distinct properties in terms of genome landscape, sequence features, and conservation compared with regions that were bound by ER␤. For example, ER␤-bound regions included GC-rich motifs, whereas ER␣-bound regions had an overrepresentation of TA-rich motifs, including forkhead-binding sites. Differences in the properties of bound regions might explain some of the differences in gene expression programs and physiological effects exerted by the two ER subtypes.
We compared the ER␤-binding sites described in our study (31) with those described by Charn et al. (29) and found only a limited overlap (170 sites) between the two data sets. The discrepancies can be attributed to many factors, including differences in (i) MCF-7 cell stocks, (ii) the ratio of ER␣ and ER␤ levels, (iii) antibodies, (iv) data analysis protocols, and (v) biological handling of cell lines. In addition, the arrays used by Charn et al. (29) to identify binding regions cover only selected regions of the genome, whereas we have used arrays covering the complete genome. Notably, in the case of the ER␣ binding data, considerable variation in the number of binding sites and a limited overlap between studies have also been reported (14).
Overall, several common features of ER␤-DNA binding are revealed by these studies: (i) ER␤ shares, to a large extent, common binding regions with ER␣; (ii) ER␤-binding sites are enriched for ERE-like sites and AP-1 sites; and (iii) ER␤-binding sites are preferentially located at long distances from the proximal promoter region. Importantly, the profiling of ER␤binding sites in vivo, such as in mouse and mammary tumor tissues that express ER␤, remains to be determined.

Novel Discoveries Pertaining to ER␤
Regulation of EMT-EMT is an essential process for normal development and is implicated in cancer progression to an invasive state (32). A recent article by Mak et al. (33) reported that ER␤ can regulate EMT in prostate cancer. Using the PC3 prostate cancer cell line that endogenously expresses both ER␣ and ER␤, they showed that induction of EMT by treatment with TGF␤ or exposure to hypoxia was paralleled by a reduction in ER␤, suggesting that loss of ER␤ promotes EMT in prostate cancer cells. Under these conditions, ER␣ levels were not affected, implying an ER␤-specific function. Loss of ER␤ led to increased VEGF-A production, which drove EMT by enhancing nuclear localization of Snail1. The investigators also demonstrated that decreased ER␤ expression was correlated with a higher Gleason grading for prostate tumors. It is suggested that ER␤ functions as a "gatekeeper" of the epithelial phenotype in the prostate gland. Intriguingly, a protective role of ER␤ against the induction of EMT in laryngeal carcinomas was also suggested by studies showing a positive correlation between the expression of ER␤ and the maintenance of the EMT marker E-cadherin in the plasma membrane of tumor cells (34). Moreover, a recent study by our group (35) revealed that EMT is involved in the progression of benign prostatic hyperplasia and is associated with high levels of ER␤2. We propose that ER␤2 can suppress expression of ER␤1, leading to EMT. Thus, ER␤ may impede EMT in different types of tumors.
Regulation of Cell Growth in Nontransformed Colonocytes-Both clinical and animal studies show that estrogen replacement therapy reduces the risk of colon tumor formation (36,37). ER␤ is the predominant ER in the colonic epithelium (38), suggesting that effects of estrogen in the colon are mediated by ER␤. Additionally, ER␤ expression has been shown to be inversely associated with colon tumor incidence (39). Using a nonmalignant cell line originally isolated from young adult mouse colonocytes, Weige et al. (40) showed that E 2 treatment reduced cell growth and induced apoptosis. Furthermore, an ER␤-mediated mechanism was shown to be required for colonic cell growth control in mice in that E 2treated ovariectomized wild-type mice exhibited significantly fewer aberrant crypt foci and increased apoptotic activity in colonic epithelia compared with ER␤ knock-out mice. This study indicated that E 2 treatment protects colonocytes from malignant transformation by increasing apoptotic activity through an ER␤-mediated mechanism, consistent with other findings indicating ER␤ to be protective against tumor formation (37).
Protective Role against Atherosclerosis-Atherosclerosis is a complex progressive disease characterized by alterations in endothelial function, smooth muscle cell proliferation, coagulation, and inflammation. Epidemiological and animal studies have suggested that E 2 protects against development of atherosclerosis (41). ER␣ and ER␤ are expressed in cells predominantly in vascular tissues such as endothelial cells (42), vascular smooth muscle cells (43), and macrophages (44). Using ER␣and ER␤-null mice, both ERs have been shown to be necessary and sufficient for estrogen-mediated protection against vascular injury (45). In apoE-deficient mice crossed with ER␣ Ϫ/Ϫ and ER␤ Ϫ/Ϫ mice, respectively, ER␣, but not ER␤, is a major mediator of the atheroprotective effects of E 2 (46). In contrast, using wild-type and ER␣-deficient mice consuming an atherogenic diet, Villablanca et al. (47) showed that the protection from development of early atherosclerotic lesions is dependent on estrogen but independent of ER␣, indicating an ER␤-mediated process. HSP27 (heat shock protein of 27 kDa) prevents many processes associated with the formation of atherosclerotic plaques. Rayner et al. (48) reported that treatment with diarylpropionitrile, an ER␤-selective agonist, induced an extracellular release of HSP27 in macrophages and an increase in serum HSP27 levels in a mouse model of atherosclerosis, suggesting a novel mechanism regulating atherosclerosis specifically through ER␤. Clearly, there is continued need to define the individual contribution of ER␣ and ER␤ in atherosclerosis, the stage of atherosclerosis when estrogen is atheroprotective, and the target genes involved in the atheroprotective effects of estrogen.

Summary and Perspectives
Recent genomic mapping of ER␤ DNA-binding chromatin regions and gene expression profiling have provided a global view of estrogen signaling via ER␤. ER␤ binds mainly outside of the proximal promoter regions, suggesting that long-range chromatin interactions may constitute an important mecha-nism for transcriptional regulation of ER␤ target genes. It is now evident that other DNA-binding transcription factors such as AP-1 collaborate with ER␤ in mediating its actions. Future studies need to correlate ER␤-binding regions with genes regulated by ER␤. To understand how these general patterns of ER␤ genomic localization direct physiological responses to ER␤ ligands and to what extent location of ER␤ determines transcriptional responses remains a great challenge. Importantly, further studies are required to elaborate on the role of ER␤ in vivo, including profiling of ER␤-binding sites in mouse and mammary tumor tissues. We have just begun to recognize newer aspects of ER␤ function, for example its roles in EMT and atherosclerosis. Further development of better and specific anti-ER␤ antibodies and ER␤-selective agonists will contribute to increasing our knowledge of the molecular basis of ER␤ signaling and is likely to uncover hitherto unknown physiological functions of ER␤. Better insights into the physiology and molecular biology of ER␤-mediated signaling may lead to identification of novel targets for effective therapeutic intervention against estrogen-related diseases.