I. INTRODUCTION

Top Previous Next References

More than 30 years ago Jensen and Jacobsen (156) came to the conclusion, based on the specific binding of estradiol-17 (E2) in the uterus, that the biological effects of estrogen had to be mediated by a receptor protein. For 24 years this protein was extensively studied in several laboratories (57, 120), and in 1986, two groups reported the cloning of this estrogen receptor (ER) (121, 122). Until 1995, it was assumed that there was only one ER and that it was responsible for mediating all of the physiological and pharmacological effects of natural and synthetic estrogens and antiestrogens. However, in 1995, a second ER, ER, was cloned from a rat prostate cDNA library (182). The former ER is now called ER. Since then, several groups have cloned ER from various species (93, 242, 352, 355, 358) and have identified several ER isoforms (67, 208, 221, 240, 257, 271). The discovery of ER has forced a reevaluation of the biology of estrogen and, because of the abundance of ER in the male urogenital tract, has refocused attention on the role of estrogen in males.

	II. ESTROGEN RECEPTOR TYPES  AND  STRUCTURE AND FUNCTIONAL DOMAINS

Top Previous Next References

ER and ER belong to the steroid/thyroid hormone superfamily of nuclear receptors, members of which share a common structural architecture (99, 118, 123, 166, 218, 363). They are composed of three independent but interacting functional domains: the NH₂-terminal or A/B domain, the C or DNA-binding domain, and the D/E/F or ligand-binding domain (Fig. 1). Binding of a ligand to ER triggers conformational changes in the receptor and this leads, via a number of events, to changes in the rate of transcription of estrogen-regulated genes. These events, and the order in which they occur in the overall process, are not completely understood, but they include receptor dimerization, receptor-DNA interaction, recruitment of and interaction with coactivators and other transcription factors, and formation of a preinitiation complex (27, 28, 166, 176, 177, 228, 291, 322).

View larger version (7K): [in this window] [in a new window]   Fig. 1. Diagramatic representation of the domain structure of nuclear receptors. The A/B domain at the NH2 terminus contains the AF-1 site where other transcription factors interact. The C/D domain contains the two-zinc finger structure that binds to DNA, and the C/F domain contains the ligand binding pocket as well as the AF-2 domain that directly contacts coactivator peptides.

The N-terminal domain of nuclear receptors encodes a ligand-independent activation function (AF1) (30, 177, 224, 356), a region of the receptor involved in protein-protein interactions (226, 259, 384), and transcriptional activation of target-gene expression. Comparison of the AF1 domains of the two estrogen receptors has revealed that, in ER, this domain is very active in stimulation of reporter-gene expression from a variety of estrogen response element (ERE)-reporter constructs, in different cell lines (78), but the activity of the AF1 domain of ER under the same conditions is negligible. Another striking difference between the two receptors is their distinctive responses to the synthetic antiestrogens tamoxifen, raloxifene, and ICI-164,384. On an ERE-based reporter gene, these ligands are partial E2 agonists with ER but are pure E2 antagonists with ER (23, 223, 227). Dissimilarity in the NH₂-terminal regions of ER and ER is one possible explanation for the difference between the two receptors in their response to various ligands. In ER, two distinct parts of AF1 are required for the agonism of E2 and the partial agonism of tamoxifen, respectively (223). In ER, this dual function of AF1 is missing (227). The importance of ERAF1 in transcriptional activity therefore remains to be clarified.

The DNA binding domain (DBD) contains a two zinc finger structure, which plays an important role in receptor dimerization and in binding of receptors to specific DNA sequences (27, 95, 128a, 128b, 315, 367). The DBDs of ER and ER are highly homologous (93). In particular, the P box, a sequence which is critical for target-DNA recognition and specificity, is identical in the two receptors (371). Thus ER and ER can be expected to bind to various EREs with similar specificity and affinity.

The COOH-terminal, E/F-, or ligand-binding domain (LBD) mediates ligand binding, receptor dimerization, nuclear translocation, and transactivation of target gene expression (99, 118, 363). Amino acid residues that line the surface of the ligand binding cavity, or that interact directly with bound ligands, span the LBD from helix 3 to helix 12 (44, 274, 356). The LBD also harbors activation function 2 (AF2), which is a complex region whose structure and function are governed by the binding of ligands (83, 84, 102, 133, 216, 321). Crystallographic studies with the LBDs of ER and ER revealed that the AF2 interaction surface is composed of amino acids in helix 3, 4, 5, and 12 and that the position of helix 12 is altered by binding of ligands. When the ER LBD is complexed with the agonists, E2 or diethylstilbestrol (DES), helix 12 is positioned over the ligand-binding pocket and forms the surface for recruitment and interaction of coactivators (44, 321, 398). In contrast, in the ER- and ER-LBD complexes with raloxifene (44, 274) or the ER-LBD 4-OH-tamoxifen complex (321), helix 12 is displaced from its agonist position over the ligand-binding cavity and instead occupies the hydrophobic groove formed by helix 3, 4, and 5. In this position, helix 12 foils the coactivator interaction surface (Fig. 2). It is evident that different ligands induce different receptor conformations (223, 264) and that the positioning of helix 12 is the key event that permits discrimination between estrogen agonists (E2 and DES) and antagonists (raloxifene and 4-OH-tamoxifen).

View larger version (62K): [in this window] [in a new window]   Fig. 2. Crystal structure of the ligand binding domain of type  estrogen receptor (ER) in the presence of an agonist (genistein) and an antagonist (raloxifene). The picture depicts the two distinct positions of helix 12 which are adopted with the two ligands. (Photograph courtesy of KaroBio and Rod Hubbard, University of York.)

The LBDs of ER and ER share a high degree of homology in their primary amino acid sequence and are also very similar in their tertiary architecture. It is, therefore, not surprising that the majority of compounds tested so far bind to ER and ER with similar affinities (183) or have similar potencies in activation of ERE-mediated reporter gene expression (23). The phytoestrogen genistein is one naturally occurring ligand that has an ~30-fold higher affinity for ER. When genistein is bound to ER, helix 12 does not adopt an agonist conformation but has, instead, a position more similar to that seen with an antagonist. This result is unexpected because the molecular shape and volume of genistein and E2 are very similar (274) and because genistein is a partial (60-70% of E2) agonist with ER on an ERE-driven reporter gene (23). There is no explanation at present for the conformation of helix 12 in the ER-genistein structure. It may well be that there are subtle conformational differences between the two ER subtypes that are not discerned by comparing their primary amino acid sequences or crystallographic tertiary structures. Agonists induce conformations of ER that promote and stabilize ER-coactivator interaction (357) while ER-coactivator interaction reciprocally stabilizes ER-ligand interaction, markedly slowing the rate of dissociation of bound agonist from both ER and ER in vitro (115). Clearly, further structural studies with ER-ligand complexes are required to fully understand the subtle interrelationship between ligand binding and helix 12 orientation.

	III. ESTROGEN RECEPTOR SPLICE VARIANTS

Top Previous Next References

Several splice variants of ER have been described. Some have extended NH₂ termini, and others have truncations and/or insertions at the COOH terminus and in the LBD. A human ER cDNA encoding a protein of 530 amino acids was identified in 1998 (256). It was longer than the original rat ER clone, and this difference was due to an NH₂-terminal extension, composed of 45 amino acids. Later, a rodent ER isoform that was 64 amino acid residues longer than the original rat ER clone (ER-485) was reported (208). In addition to extensions of the NH₂ terminus, three groups have reported cloning of ER-503, an isoform with an in-frame insertion of 18 amino acids in the LBD (67, 128, 271). This isoform is a splice variant, and the insert is in the junction between exons 5 and 6 (93). In contrast to the NH₂-terminally extended ER isoforms, the affinity of ER-503 for E2 and other known ER ligands is lower than for ER and for ER-485. In one report, ER-503 acted as a dominant negative regulator by suppressing E2-dependent ER- and ER-mediated activation of gene transcription, but did not bind E2 (221). In other reports, ER-503 did exhibit E2-dependent transcriptional activation of a reporter gene, but the concentration of E2 needed was 100- to 1,000-fold higher than that for ER (271, 128). All ER isoforms, 503, 485, and 530, bind to consensus ERE and heterodimerize with each other and with ER (128, 271).

Transcripts encoding additional ER isoforms with variations at the extreme COOH terminus have been found in human testis cDNA libraries (257, 240). ERcx (257) is identical to ER-530 in exons 1-7, but exon 8 is completely different. The last 61 COOH-terminal amino acids (exon 8), encoding part of helix 11 and helix 12, have been replaced by 26 unique amino acid residues. Due to the exchange of the last exon, ERcx lacks amino acid residues important for ligand binding and those that constitute the core of the AF2 domain. It is, therefore, not surprising that ERcx does not bind E2 and has no capacity to activate transcription of an E2-sensitive reporter gene (257). Surprisingly, in view of its intact DBD, ERcx does not bind to a consensus ERE. Furthermore, ERcx shows preferential heterodimerization with ER rather than with ER, inhibiting ER DNA binding. Functionally, the heterodimerization of ERcx with ER has a dominant negative effect on ligand-dependent ER reporter-gene transactivation (257).

Moore et al. (240) have described five ER isoforms (ER 1-5). ER 1 corresponds to the previously described ER-530 (256), and the ER 2 variant is most likely identical to ERcx (257). However, ER 3-5 are novel splice variants with exchanges of the last exon of ER-530 for previously unknown exons. As with ERcx, neither of the novel COOH-terminal splice variants, ER 3-5, can be expected to bind E2 or activate transcription from an ERE-driven reporter, as they all lack amino acids important for ligand binding as well as the core of AF2. In contrast to what was reported for ERcx (257), two of the COOH-terminal splice variants, ER 2 and 3, do bind to a consensus ERE.

Various alternatively spliced forms of ER have also been described (108-110, 243, 351, 406). To date there is insufficient information as to whether or not all isoforms and splice variants of ER are expressed as proteins or whether they have any major biological and physiological role. Another source of variability in receptor function, and perhaps also dysfunction, is ER and - gene polymorphisms. ER polymorphisms have been linked to increased litter size in pigs (300, 325), breast cancer susceptibility (4, 313, 411), bone mineral density and osteoporosis (169, 236, 260), hypertension (201), spontaneous abortion (203, 311), and body height (202).

	IV. SPECIFIC ESTROGEN RECEPTOR TYPES:  AND  LIGANDS

Top Previous Next References

The discovery of ER has revitalized the search for improved tissue-selective estrogen receptor modulators (SERM). Such ligands could provide the benefits of estrogens and avoid unwanted side effects of E2. In the clinical setting, these pharmaceuticals would be used for prevention or treatment of menopausal symptoms, osteoporosis, cardiovascular disease, and breast cancer in women, or other estrogen-related indications affecting both men and women (72, 74, 124, 245, 287). Several large pharmaceutical companies are engaged in the development of ER- and ER-selective SERMs, but none has yet come to market. Katzenellenbogen and co-workers (230, 346) have synthesized ER subtype-specific ligands. The most ER-selective ligand had a 120-fold higher agonist potency for ER than for ER. Another selective ligand showed full ER agonism but pure ER antagonism (346). The ER-selective antagonist was further investigated by the synthesis of a series of analogs with substituents of various sizes in both cis- and trans-configurations (230). All analogs were agonists on ER, but only those with small substituents were ER agonists. As substituent size increased, the agonism on ER disappeared, and with larger substituents the ligands were pure ER-selective antagonists. The gradual ER-selective antagonism by this series of analogs was influenced by both size and shape of the substituent. It was concluded by the authors that less steric perturbation is required to induce an antagonist conformation in ER than in ER This could be due in part to the observation (274) that the volume of the binding cavity of ER is smaller than that of ER.

Are phytoestrogens natural SERMs? Phytoestrogens are nonsteroidal polyphenolic compounds present in several edible plants. On the basis of their chemical structure, phytoestrogens may be divided into four subclasses: isoflavonoids, flavonoids, coumestans, and mammalian lignans. The major dietary source of isoflavonoids (e.g., genistein, daidzein, formononetin, biochanin A) is soybean. Flavonoids (e.g., chrysin, apigenin, naringenin, kaempferol, quercetin) are more widely distributed in the plant kingdom and are present in several edible plants. Coumestans (e.g., coumestrol) are present in plants not commonly used in human diets, such as alfalfa sprouts. Mammalian lignans (e.g., enterolactone and enterodiol) are not present in human diets as such, but are ingested as precursors (plant lignans), which are converted to mammalian lignans by gut microflora. Plant lignans are present in fiber-rich foods, such as flaxseeds and unrefined grain products. Isoflavones and coumestrol interact with ER in vitro, although the activities of individual compounds with similar chemical structures vary remarkably (23, 183). Some flavonoids show modest estrogenic activity, while others are completely inactive (183). Mammalian lignans are very weak estrogens, and concentrations of 1 mM or more are required to show any ER- or ER-mediated activity (303, 308).

It has been suggested that dietary genistein could play a role in preventing the development of hormone-dependent diseases and conditions, such as breast and prostate cancer, cardiovascular disease, menopausal symptoms, and osteoporosis (reviewed in Ref. 341). This suggestion comes from epidemiological studies, showing an inverse correlation between the intake or serum concentrations of genistein and the risk of estrogen-related diseases and conditions. At present, there is no direct evidence for the beneficial action of genistein in humans. It has also been claimed that genistein would be devoid of the adverse effects typically found with E2, but this has not been shown convincingly either.

There are very little data from in vivo studies to demonstrate the tissue and/or ER selectivity of genistein that has been demonstrated in vitro (23). Most studies have been done with high doses of genistein, likely to exert both ER- and ER-mediated activities. Furthermore, careful dose-response studies have not been done. In most in vivo studies, genistein acts as an estrogen and induces effects similar to E2 in female and male experimental animals (308, 340). At low doses, genistein and other isoflavones have been shown to act selectively in the cardiovascular system. In female macaques, isoflavones have favorable effects on the cardiovascular system without affecting the reproductive tract (7). In female rats, genistein was as potent a vasculoprotectant as E2, but unlike E2 showed no uterotrophic activity (216a). At present, it is not known to what extent the vascular responses are mediated by ERs. Both ER subtypes are expressed in vascular tissues, but in addition, there is sufficient evidence to indicate the presence of non-receptor-mediated actions of E2 in the same tissues. Further studies are thus needed to demonstrate that genistein acts in a receptor- or tissue-selective manner in vivo.

	V. ESTROGEN RECEPTOR-DNA INTERACTIONS

Top Previous Next References

For several years it was thought that the only mechanism through which estrogens affected transcription of E2-sensitive genes was by direct binding of activated ER to EREs. These estrogen response elements were first observed in the 5'-flanking region of the Xenopus vitellogenin A2 gene (168). Today we know that ER and ER can also modulate the expression of genes without directly binding to DNA (Fig. 3). One example is the interaction between ER and the c-rel subunit of the NFB complex. This interaction prevents NFB from binding to and stimulating expression from the interleukin-6 (IL-6) promoter (112). In this way, E2 inhibits expression of the cytokine IL-6 (112, 292). Another example of indirect action on DNA is the physical interaction of ER with the Sp1 transcription factor (25, 277, 289). ER enhancement of Sp1 DNA binding is hormone independent (277), and both ER and ER can activate transcription of the retinoic acid receptor 1 (RAR-1) gene, presumably by the formation of an ER-Sp1 complex on GC-rich Sp1 sites in the RAR1 promoter (345, 409). Interestingly, in one study, ER activated RAR-1 promoter-reporter constructs in the presence of the estrogen antagonists 4-OH-tamoxifen, raloxifene, and ICI-164,384. E2 did not activate expression of the reporter but blocked the effect of the antagonists (409).

View larger version (26K): [in this window] [in a new window]   Fig. 3. Model representing the various modes through which estrogen receptors can modulate transcription of genes. In the first panel is depicted the classical interaction of the activated receptor with estrogen response elements (EREs) on DNA. In the other three panels are representations of the indirect effects of estrogen receptors on transcription interactions. This occurs through protein-protein interactions with the Sp1, AP1, and NFB proteins.

Both ER and ER can interact with the fos/jun transcription factor complex on AP1 sites to stimulate gene expression, however, with opposite effects in the presence of E2 (264, 383, 385). In the presence of ER, typical agonists such as E2 and DES as well as the antiestrogen tamoxifen function as agonists in the AP1 pathway. Raloxifene is only a partial agonist. In contrast, in the presence of ER, tamoxifen and raloxifene behave as fully competent agonists in the AP1 pathway, while E2 acts as an antagonist, inhibiting the activity of both tamoxifen and raloxifene (264). Analysis of the mechanism of ER-stimulated expression from an AP1 site showed that agonist-bound ER requires intact AF1 and AF2 functions and that it enhances AP1 activity via interactions with the p160 family of coactivators. In this "agonist" pathway the DBD of ER is essential. The "antiestrogen" pathway, triggered by antiestrogen-liganded ER, enhances AP1 activity in an AF1- and AF2-independent fashion but also requires an intact DBD. The DBD sequesters corepressors from the AP1 transcription complex, releasing it from suppression and subsequently resulting in AP1-dependent transcription (385).

The electrophilic/antioxidant response element (EpRE/ARE) is yet another site on DNA where ER and ER behave differently. In MCF-7 cells, antiestrogens but not E2 activate transcription of the quinone reductase gene and increase NAD(P)H:quinone oxidoreductase enzyme activity via an EpRE/ARE (239). Furthermore, E2 inhibits the agonistic effect of the antiestrogens on EpRE/ARE reporter constructs, and in this capacity, ER is more efficacious than ER (238). These findings suggest that antiestrogens are antioxidants and inducers of phase 2 detoxification enzymes, protecting cells from damage by oxygen radicals and other toxic by-products of metabolic oxidation. Furthermore, it would seem that these protective effects of antiestrogens are mediated by ER.

Another potentially important pathway of estrogen action is constituted by the very rapid, so-called, nongenomic effects of certain ligands (232). There is evidence that some of the rapid effects of nuclear receptor ligands require the presence of the receptor protein (237, 266, 293). In endothelial cells, E2-mediated membrane effects lead to sequential activation of ras, raf, mitogen-activated protein kinase kinase (MEK), and subsequently activation of mitogen-activated protein kinase (MAPK) (71). It has been suggested that this leads to activation of endothelial nitric oxide synthase (eNOS) and release of nitric oxide (NO). In neurons, membrane effects of estrogen lead to stimulation of src, ras, MEK, and MAPK, resulting in neuroprotection, and in osteoblasts the membrane effects of estrogen may be involved in control of apoptosis, cell proliferation, and differentiation (71).

	VI. PHOSPHORYLATION OF ESTROGEN RECEPTORS

Top Previous Next References

During the past 15 years, the dogma of the strict requirement of hormones for activation of steroid receptors has been challenged. Evidence for ligand-independent activation of these receptors through alternative signaling pathways has emerged (reviewed in Refs. 55, 91, 386, 387). In most instances, the mechanisms underlying this activation involve phosphorylation of the receptors (318, 386). The phosphorylation sites on native ER from calf, mouse, and rat uterus, and from human MCF-7 cells, have been extensively studied by analysis of the phosphoamino acid content of the receptor and/or its binding to anti-phosphotyrosine antibodies (21, 22, 233) There are still several controversial issues such as whether both tyrosine and serine are phosphorylated and which phosphorylations are ligand dependent. Some investigators report phosphorylation on tyrosine (3, 22, 231, 233) while others detect serine as the only phosphorylated amino acid (13, 51, 85, 191). A single tyrosine phosphorylation site in human ER at position 537 has been identified in vitro (12) and in vivo (11, 12, 14). Whether or not this phosphorylation is E2 dependent is still a point of contention (14).

Phosphorylation of ER on serine residues has been studied with recombinant human or mouse ER expressed in COS-1 cells in the presence of E2 (190, 200). The vast majority of phosphoserine residues were mapped to the NH₂-terminal A/B domain. Outside of this domain, sites have been identified at positions 236 in the DBD, 294 in the hinge region, and 305 in the LBD (numbering here refers to the position in the human sequence). Within the A/B domain, Ser-104, -106, -118, -154, and -167 were characterized as phosphorylation sites in human ER (hER) in the presence of E2 (160, 200). In the mouse ER (mER), three phosphoserine residues were clearly mapped in the A/B domain at positions 122, 156, and 158 (190). The precise location of the fourth phosphorylation site remains unsettled. It could correspond either to Ser-171 or to Ser-522. There is a high degree of homology between hER and mER, and all identified phosphoserines but one are conserved in these receptors (Fig. 3). Moreover, not only are the serine residues preserved at these positions, but their environments are well conserved too. The unique exception is position 156, which is Ser in mER and proline in hER. There is still disagreement about the importance of phosphorylation of Ser-104, -106, and -167 in hER (3, 13, 190, 200).

Even though ER is phosphorylated in the absence of E2 in different cell lines (12, 200), enhanced phosphorylation of the receptor occurs in response to physiological doses of E2 (3, 16, 85, 200). E2 treatment of MCF-7 cells has also been reported to result in the dephosphorylation of a single unspecified site in ER (11). The enzymes responsible for E2-dependent phosphorylation of ER seem to be diverse. They include an E2-dependent tyrosine kinase, which has been purified from calf uterus (52), and a casein kinase II in MCF-7 cells, which phosphorylates Ser-167 (11). In addition, serines other than Ser-167 are E2-stimulated phosphorylation sites (3, 158, 159, 200) that do not have casein kinase II binding sites in their environment (157a), and this suggests other novel and (so far) unspecified kinases.

In the absence of E2, other signaling pathways can modulate ER through phosphorylation. These include 1) regulators of the general cellular phosphorylation state, such as protein kinase A (PKA) (16, 46, 153) or protein kinase C (PKC) (152, 267, 189, 200); 2) extracellular signals such as peptide growth factors, cytokines, or neurotransmitters (56); and 3) cell cycle regulators. Peptide growth factors represent a large class of ER activators. Epidermal growth factor (EGF) can mimic the effects of E2 in the mouse reproductive tract, and pretreatment of mice with the pure anti-estrogen ICI-164,384 greatly diminishes the uterine response to EGF (151). In line with this cross-talk between ER and EGF, mice in which ER has been inactivaed (ERKO mice) lack uterine E2-like response to EGF even though the EGF signaling pathway is not disrupted (82). Other growth factors which activate ER signaling include insulin (249, 267, 268), insulin-like growth factor I (IGF-I) (16, 152, 215, 249), and transforming growth factor (TGF)- (152).

In all of these cross-couplings between growth factors and ER, the mediators are the guanine nucleotide binding protein p21^ras and the MAPK (46). p21^ras functions as an intermediate between the membrane-associated growth factor receptor/tyrosine kinase and MAPK phosphorylation cascades. The target site of these kinases on ER has been mapped to the AF-1 domain. In most studies, this target has been further localized to Ser-118 in hER, which corresponds to the consensus phosphorylation site for MAPK (46, 165). However, it appears that there are alternative pathways of cross-coupling between growth factors and ER. For example, in the SK-N-BE neuronal cell line, the target for the insulin activation of ER has been mapped to the AF-2 region (267). This observation suggests that different mediators of the growth factor signal, downstream of p21^ras, might be involved. In line with this hypothesis, EGF activation of ER in endothelial cells of the pulmonary vein is independent of MAPK and does not involve the same phosphorylation site in ER (164). Furthermore, the 90-kDa ribosomal S6 kinase, pp90^rsk1, a serine/threonine protein kinase which is phosphorylated by MAPK upon EGF stimulation, phosphorylates Ser-167 in hER (158).

Other extracellular signals that can modulate ER activity are heregulin (273), interleukin-2 (IL-2), and dopamine (279, 334). Dopamine is, so far, the only neurotransmitter identified as an ER activator. Dopamine activation of ER is distinct from activation by EGF, since a mutant hER, which is no longer activated by dopamine, can still be activated by EGF (15, 114). The mechanisms involved in ER activation by IL-2 or by dopamine remain elusive, and it has not yet been shown whether phosphorylation of the receptor is involved. Phosphorylation is clearly involved in heregulin activation of ER. In breast cancer cells, activation, by heregulin, of the heregulin receptor, HER-2 (a tyrosine kinase), leads to direct and rapid phosphorylation of ER on tyrosine residues, followed by transcription of the progesterone receptor gene. Heregulin promotes hormone-independent growth in human breast cancer cells, and overexpression of HER-2 results in the development of E2-independent growth, which is insensitive to both E2 and tamoxifen.

Cyclins are positive regulatory subunits of cyclin-dependent kinases (CDKs), a class of serine/threonine kinases that are major determinants in cell cycle progression. The activities of many proteins, including transcription factors, are modulated by CDKs. Because CDKs control cell division, dysregulation of their regulatory partners, the cyclins, has been implicated in the initiation and promotion of hyperplasia and oncogenesis (380, 389). Two different cyclins, cyclin A and D1, have been identified as ER activators (247, 280, 361, 412). Individual overexpression of these proteins elicits an increase of ER activity in the absence of E2. Distinct mechanisms seem to mediate the activation of ER by these two cyclins. Cyclin D1 activation of ER does not involve a phosphorylation, since it does not require participation of a CDK (361, 412, 247), but cyclin A does activate ER by a phosphorylation in the AF-1 domain (296, 361).

All of the steps in transcriptional activation of ER-dependent genes, i.e., ligand binding, ER dimerization, DNA binding, and the interaction with cofactors, appear to be influenced by phosphorylation of ER. Ligand binding of ER is regulated by phosphorylation of Tyr-537 (10, 231, 233). The dimerization function of ER appears to be enhanced by phosphorylation at Ser-236 in the DBD (62) and its DNA binding capacity by phosphorylation at Ser-167 (11, 51, 85). MAPK-mediated phosphorylation at Ser-118 in ER potentiates the interaction with the coactivator p68 (92).

In contrast to the numerous studies related to ER phosphorylation, only one group has so far investigated ER phosphorylation (357). This study demonstrated phosphorylation of recombinant ER when it was expressed in human embryonic kidney cells (293 T). The target sites for EGF-induced phosphorylation of ER are two serine residues located at positions 106 and 124 in consensus sequences for MAPK. Despite a poor overall homology between ER and ER in the A/B domain, Ser-106 is part of a motif shared with ER and other steroid receptors and corresponds to Ser-118 in human ER. Phosphorylation at Ser-106 and -124 enhanced the interaction with the coactivator SRC-1 (357).

	VII. TRANSCRIPTIONAL COFACTORS: COACTIVATORS, NEGATIVE COREGULATORS, AND COREPRESSORS

Top Previous Next References

Current models of eukaryotic gene regulation suggest the existence of distinct gene activity states i.e., repressed/silenced, basal/ ground state, and activated states (342, 343). Estrogen signaling is usually associated with gene activation, i.e., the switch between basal and activated state. As cellular environments change, ERs can associate with distinct subsets of cofactors depending on binding affinities and relative abundance of these factors (66, 125). Coactivators turn on target gene transcription, while negative coregulators and corepressors inhibit gene activation and possibly also turn off activated target genes. These proteins exist in multiple complexes, possess multiple enzymatic activities, and (in a simplified view) bridge ERs, to either chromatin components such as histones, to components of the basal transcription machinery, or to both. During the past 5 years, protein-protein interaction screenings and biochemical approaches have led to identification of a large number of cofactors which may act at different functional levels (for current review, see Refs. 107, 142, 229, 396, 397, 400). The majority of these cofactors bind to the LBD, which (as described above) plays a central role not only in binding of ligands but also in transforming the ligand signal. So far, only very few receptor-specific cofactors have been identified, and the various nuclear receptors appear to utilize similar cofactors.

Although all ER ligands bind exclusively to the LBD, binding of an agonist triggers AF-2 activity, but binding of antagonists does not. As described above, structural analyses of ER complexed with E2 or raloxifene/tamoxifen show that each class of ligand induces a different LBD conformation (44, 274, 321). These structural rearrangements are critical for ligand-induced transcriptional regulation, since it is the exposed surface of the receptor that recruits coactivators and other coregulatory proteins. Some characteristic AF-2 cofactors with relevance for agonist-bound ERs are illustrated in Figure 4. They can be categorized in several subgroups.

View larger version (24K): [in this window] [in a new window]   Fig. 4. Structural organization of cofactors with relevance for agonist-bound ERs. Highlighted are the nuclear receptor (NR) interaction domains with conserved LxxLL motifs, transcription activation domains (AD) and their potential target factors (CBP, ADA2, TBP), as well as potential DNA-binding domains.

This well-characterized coactivator family consists of three related members (reviewed in Ref. 228): SRC-1, the first identified nuclear receptor coactivator and founding member of the family (also p160-1, N-CoA1); SRC-2 (also TIF-2, GRIP1, N-CoA2); and SRC-3 (also P/CIP, ACTR, AIB1, RAC3, TRAM1). Critical for the function of these coactivators is the central nuclear receptor-interaction domain consisting of three equally spaced conserved LXXLL motifs, also called NR-box or LCD/LXD (reviewed in Ref. 400). These motifs represent the primary docking sites to the AF-2 domain and exist in many different cofactors (199). With regard to ERs, it has been found that, in the case of SRC-1 and -2, the second motif (NR-box 2) has the highest affinity for the agonist-bound ER (i.e., the AF-2 domain), while in SRC-3, the first motif serves as primary docking site (66, 344). Recent ER costructures with a single NR-box 2 peptide have visualized the interactions of the NR-box with the AF-2 domain (322). It is hoped that, in the future, costructural analysis of the entire SRC interaction domain including all three NR-boxes with dimeric ERs will be possible. An additional COOH-terminal region in SRC-1 has been implicated in binding to the NH₂-terminal ER AF-1 domain (140, 214, 384). This offers an interesting explanation for the collaboration of receptor NH₂ and COOH termini in activation, which is exhibited by ER but not ER.

All SRC coactivators contain two separate transcription activation domains, AD-1 and AD-2. AD-1 is involved in recruitment of CBP/p300 coactivators and acetyltransferases and AD-2 in recruitment of a second protein-modifying enzyme called coactivator-associated arginine methyltransferase (CARM1) (61). The COOH termini of SRC-1 and SRC-3 exhibit intrinsic lysine acetyltransferase activity. Together with evidence for the existence of histone acetyltransferase (HAT) complexes consisting of SRC, CBP/p300, and/or PCAF/GCN5 (400), these features strongly indicate that SRC coactivators function primarily by recruiting chromatin (i.e., histone)-modifying enzymatic activities to ligand-activated nuclear receptors and, thereby, to hormone-regulated target genes. Additionally, acetylation of lysine residues adjacent to the core of NR-box1 in SRC-3 abolishes the interactions with the AF-2 domain of ER and may represent one important mechanism for attenuation and feedback regulation in estrogen-regulated gene expression (64). A role for SRC coactivators in ER-mediated gene expression and, possibly in carcinogenesis, is suggested by the discovery of frequent AIB1 gene amplification in ER-positive breast and ovarian cancer cells (8) as well as by the phenotype of SRC-1 knock-out mice (399).

CBP/p300 are viewed as general coactivators and cointegrators involved in multiple signaling pathways. They are ubiquitously expressed, possess acetyltransferase activity, and contain docking sites for ERs, via NH₂-terminal NR-boxes. Several studies have highlighted their involvement in ER activation (127, 176), but it is yet unclear whether they bind directly to ERs under physiological conditions, i.e., in the presence of other competing coactivators. Instead, affinity considerations (407) and the existence of multiple interactions between CBP/p300 and other acetyltransferases argue for a scenario in which SRC coactivators are required for the recruitment of CBP/p300 to receptors. In support of a critical role of CBP/p300 in histone acetylation, studies utilizing receptors other than ERs have shown that CBP HAT activity was required for efficient target gene activation in vivo (175). Possibly, the requirement for multiple ER-associated HATs may also reflect the partially different substrate specificity (histones and nonhistone targets) of these enzymes. The role of CBP/p300 on ER functions has not yet been studied.

A large coactivator complex, referred to as TRAP/SMCC/DRIP/ARC complex, may connect ERs directly to the basal transcription machinery (155, 107). The complex was first identified biochemically from HeLa cells using the thyroid-hormone receptor (155 and references therein). The receptor-binding subunit of the entire complex, referred to as TRAP220 (TRIP2/PBP/DRIP205/RB18A), has been isolated independently in several laboratories using two-hybrid screenings. It is not known whether the two complexes act sequentially or independently on nuclear receptors (107, 155). It has been suggested, however, that acetylation of the AF-2 NR-box binding motif could be one possible mechanism for dissociation of the SRC coactivators from ERs, allowing other cofactor complexes such as TRAP/DRIP to bind (64). ER is less efficient than ER in recruiting the TRAP/DRIP complex in vitro (228), and this may indicate differences in the physical interaction of the ER subtypes with coactivator complexes.

Additional unique coactivators exist that are structurally distinct from CBP/p300, p160/SRC, or TRAP/DRIP proteins. However, functional connections to ERs have not yet been made for these cofactors. Examples are 1) NSD1, a bifunctional coactivator and corepressor containing a SET domain, a motif found in several chromatin-modifying proteins (146); 2) TIF1, which exhibits or associates with protein kinase activity and probably connects receptors with other chromatin components such as the heterochromatin protein HP1 and the SNF-2 component of the mammalian SWI/SNF-complex (106, 197, 294); 3) the PPAR coactivator PGC-1, which interacts with ERs probably via its NR-box motif and which exists in a complex with SRC-1 (254, 285, 286); and 4) a novel coactivator called ASC-2/RAP250 (48, 198). This coactivator contains one functional NR-box motif, is widely expressed in many tissues (48), and appears to be overamplified or overexpressed in certain cancers (198). It is currently not known whether ASC-2/RAP250 functions separately or in conjunction with other coactivator complexes.

Studies on several additional ER cofactors illustrate that not all AF-2 binding proteins act simply as coactivators. Two AF-2 interacting proteins, RIP140 and SHP (short heterodimerization partner) (see Fig. 5), exhibit negative coregulatory functions, because they can antagonize SRC-1 coactivators in vivo and compete for AF-2 binding in vitro (53, 161, 359, 360). Therefore, these cofactors may belong to a separate category of coregulators that are distinct from conventional corepressors. SHP is a remarkable and unusual orphan receptor that has a recognizable LBD but lacks a DBD (161, 162, 316). Both SHP and its closest relative DAX-1 interact with several nuclear receptors and inhibit their transcriptional activity (79, 316, 317). SHP binds directly to the AF-2 domain of ERs via two separate NR-box motifs and thus functions in the same way as AF-2 cofactors (161, 162). SHP mRNA is widely expressed in rat tissues including certain estrogen target tissues, and subcellular localization studies demonstrate that SHP is a nuclear protein. Under some circumstances ER may associate with corepressors utilizing SHP or DAX-1 as bridging proteins. This may recruit corepressors/deacetylases to ER target genes and thus antagonize acetyltransferase/coactivator complexes.

View larger version (24K): [in this window] [in a new window]   Fig. 5. Model illustrating the interplay between coactivators and coregulators/corepressors in estrogen signaling. The model may in part apply for antagonist signaling, yet the individual coregulator components such as coactivators for tamoxifen-bound ERs remain to be identified.

The related "conventional" corepressors N-CoR (nuclear receptor corepressor) and SMRT (silencing mediator of retinoid and thyroid receptors), associate with ER in the presence of antagonistic ligands (reviewed in Ref. 228), and it has been suggested that they may play a role in regulating ER activity in tumors treated with antiestrogens (335). There is evidence that histone deacetylase activity recruited by corepressor complexes such as N-CoR/SAP30/SIN3/HDAC2 is required for the transcriptional repression of tamoxifen-bound ER (188, 194, 335). Lavinsky et al. (194) have observed that acquisition of tamoxifen resistance in MCF-7 cells growing in athymic nude mice (a model of human breast cancer) is accompanied by a decrease in corepressor levels. They suggest that loss of corepressors may be one mechanism of tamoxifen resistance.

Because direct binding of antagonist-bound ER to corepressors has not been demonstrated, indirect recruitment mechanisms are also possible. In view of the insights gained from the structure of antagonist-bound ERs (44, 274, 321) and from corepressor interactions with retinoic acid, thyroid, and ecdysone receptors (405, 145, 362, 405), it is likely that ER antagonists, via the realignment of helix 12 and possibly the F-domain (251), expose as yet unidentified corepressor binding epitopes.

In a yeast two-hybrid screen with a dominant negative ER as bait, a novel corepressor was recently identified (238). This factor also repressed the agonist activity of both ER subtypes (but not that of other receptors) by competitive reversal of SRC-1 enhancement. The cofactor was named REA for "repressor of estrogen receptor activity." REA may, therefore, represent a bivalent cofactor with corepressive functions on both estrogen- and antiestrogen-bound ERs. Another candidate corepressor is the RSP5/RPF1 protein, which specifically represses the tamoxifen-mediated partial agonist activity of ER (255).

Although the involvement of corepressors in antagonist signaling may in part explain their antiestrogenic activity in breast cancer treatment, it is not easy to envisage how corepressors alone account for the agonistic activites of antiestrogens. The existence of novel coactivators that bind to ERs only in the presence of antagonists has been postulated, and a few candidates have been identified. One such coactivator, called L7-SPA, was identified using RU487-bound progesterone receptor as bait in yeast two-hybrid screenings, and subsequently demonstrated also to bind and enhance the agonistic properties the tamoxifen-bound ER (349). The hinge domain has been mapped as interaction site on PR, and it is likely that L7-SPA recognizes features that are conserved between antagonist-bound steroid receptors. In this respect, L7-SPA shares interaction features with corepressors, and this lends support to the idea that part of its function is competitive antagonism of corepressor function. As mentioned below, identification of synthetic peptides, which interact specifically with the antagonist-bound ER, suggest that there could be multiple sites on ER where coactivators in cells could interact only when the receptor is bound to an antagonist.

One major unanswered question is whether or not ligand-bound ER and ER contact different coactivators/cofactors. Because of the homology in their AF-2 domains, it was anticipated that the two ER subtypes would be similar in coactivator recruitment, but certain differences have been reported. For example, affinity of the ER interaction with SRC-3 is much higher than that observed for the ER (344), and in contrast to ER, ER apparently interacts in a ligand-independent manner with SRC-1 and SHP in vitro (78, 317, 357). Furthermore, as mentioned above, the TRAP/DRIP complex apparently binds only weakly to ER.

Such differences may be associated with distinct binding specificities of the NR-box interaction motifs (84, 216, 225). Clearly, preferential binding of certain coactivators to one of the ERs should have consequences for E2 signaling. In addition, because there are differences in the ligand-binding specificities of the two ERs, and because ligands can alter receptor conformations, ligand-binding specificity is likely to affect cofactor binding affinities and specificity. Indeed, such differences in receptor conformation have been found by an affinity-selection approach. This method was used to select peptides that specifically bound to ER subtypes in the presence of agonists or antagonists (255, 265). These experiments clearly showed that different agonists and antagonists induce different ER conformations, which in turn result in the recruitment of different ER-binding peptides.

Although no cause and effect relationship between cofactors and carcinogenesis has been established, gene amplification and overexpression of the ER coactivators SRC-3, AIB3/ASC-2/RAP250, and TRAP220/PBP have been found in breast and ovarian cancer (8, 198, 408). Additionally, changes in the expression levels of conventional corepressors, which associate with antagonist-bound ERs, may contribute to the phenomenon of tamoxifen resistance in breast cancer treatment. There could, for example, be imbalances or changes in the ratio between corepressors and coactivators (194). Interestingly, ER and SRC-1 are coexpressed in stromal cells but not in epithelial cells in the mammary gland (324). The response of the stroma to E2 differs from that of the epithelium, perhaps indicating that association between ERs and different coactivators may influence the responsiveness to E2. Lately, it has been suggested that transcriptional cross-talk between ERs and other major players in breast cancer such as BRCA1 or cyclin D1 occurs at the cofactor level via ligand-independent recruitment of acetyltransferases (100, 229). Although it is currently not known whether cofactor levels are related to carcinogenesis, it seems likely that changes in the levels or activity of cofactors could have profound effects on target gene expression. Such changes may shift the balance from differentiation to proliferation and contribute to the development of neoplastic diseases.

	VIII. ESTROGEN RECEPTOR-ARYL HYDROCARBON RECEPTOR INTERACTIONS

Top Previous Next References

More than two decades ago, it was first reported that ligands for the aryl hydrocarbon receptor (AhR) can counteract the effects of estrogens in intact animals (173). Like ER, AhR is a ligand-inducible transcription factor, but in contrast to ER, an endogenous ligand has yet to be identified. AhR is a member of the bHLH/PAS class of transcription factors, members of which are involved in development, oxygen homeostasis, and circadian rhythm (for review, see Ref. 80). AhR is the only ligand-activated bHLH/PAS protein identified, and because the majority of ligands are classified as environmental contaminants, AhR functions as a toxin sensor. Ligands for AhR include environmental polycyclic aromatic hydrocarbons (PAHs), 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) (275), related polychlorinated dibenzo-p-dioxins (PCDDs) and dibenzofurans (PCDF) (for review, see Ref. 276), and dietary indole carbinols present in cruciferous vegetables (35, 41, 65, 373). After ligand binding, AhR is found in the cell nucleus as a heterodimer with the AhR nuclear translocator Arnt (139). The AhR-Arnt heterodimer binds to specific genomic enhancer sequences termed dioxin- or xenobiotic-responsive elements (DREs or XREs), and this interaction leads to transcriptional activation. Those genes that are transcriptionally activated as a result of AhR interactions with 5'-upstream DREs have been described as members of the Ah gene battery and include phase I drug-metabolizing enzymes such as cytochromes P-450IA1 (CYP1A1), CYP1A2, and CYP1B1 and phase II enzymes including glutathione-S-transferase Ya subunit, UDP glutathione transferase, aldehyde dehydrogenase, and NAD(P)H quinone oxidoreductase (NQOR) (for review, see Ref. 126). In addition to the Ah battery, a number of E₂-regulated genes (see below) are regulated by AhR at either the transcriptional or posttranscriptional level.

In vivo, the most distinct antiestrogenic effects of TCDD in rodents are decreased uterine weight (113, 304), reduction in the incidence of mammary and uterine cancer (41, 172, 173), and inhibition of E₂ induction of PR, EGF receptor, and c-fos protooncogene in immature or ovariectomized rodent uterus (18, 19, 297). In humans, accidental exposure to TCDD, as occurred in 1979 in Seveso, has resulted in a decrease in mammary and endometrial cancer (31, 32). In vitro, several genes have been identified as targets for cross-talk between ER and AhR. Among E₂-regulated genes that are affected by TCDD treatment are cathepsin D (180), pS2 (402), prolactin receptor (211), c-fos (87, 88), hsp27 (278), TGF- and TGF- (111), as well as PR (129). E₂-induced cell proliferation and postconfluency production of foci are also decreased by TCDD (34, 116, 390).

TCDD is not a typical ER antagonist, because it does not compete with E2 for binding to ER (130). Instead, ligands for AhR might mediate their antagonistic effects on ER signaling by several separate pathways. These include the following.

1) The first pathway is by increasing the rate of metabolism of E2. TCDD increases the degradation of E2 by inducing CYP1A1 and CYP1B1, enzymes involved in the metabolic inactivation of estrogens (131, 304, 337, 338, 397a).

2) The second pathway is by decreasing levels of ERs. In vitro experiments have shown that TCDD treatment of MCF-7 human breast cancer cells and mouse hepatoma (Hepa 1c1c7) cells resulted in a dose-dependent decrease in levels of ER. There was no decrease in ER in Hepa cells expressing a mutated AhR (130, 403). In vivo, TCDD suppressed ER mRNA by 60% in an AhR-responsive mouse strain (C57BL/J6), but not in DBA/2J mice, a nonresponsive strain (353). Both the ER (390) and ER (unpublished observations) genes contain DREs in their 5'-flanking regions. In addition to transcriptional regulation of ER, downregulation of ER by TCDD may be achieved by changes in receptor synthesis, recycling, and/or degradation, as is the case in mammary carcinoma cell lines where TCDD treatment decreases the level of ER protein by a proteasome-dependent pathway (304).

3) The third pathway is by suppressing transcription of E₂-induced genes. As discussed above, ER can interact with DNA at ERE, Sp1, and AP-1 sites. A consensus DRE sequence has been identified that confers the antiestrogenic effect of liganded AhR complex on several genes. This inhibitory DRE (iDRE) has the core sequence GCGTG, which binds the AhR complex. In the cathepsin D gene (379), an iDRE lies between the ERE-half site and the Sp1 site in the ERE/SpI motif, disrupting the formation of the ER/Sp1-DNA complex (347). In the fos gene, liganded ER does not bind directly to DNA, but to DNA-bound Sp1. In this case an iDRE juxtaposes the Sp1 site, suggesting competition between Sp1 and AhR for DNA binding (87, 88). In the pS2 gene an iDRE overlaps the AP-1 site in the ERE/AP-1 motif (54, 402), and in the hsp27 gene an iDRE is located 100 bp downstream of the ERE/Sp1 motif close to the transcription start site, probably blocking access of the basal transcription machinery to the promoter (278).

4) The fourth pathway is by competing for shared cofactors. AhR interacts with various factors in the basal transcription complex (250, 301, 397a), including TATA-binding protein and transcription factor IIF (TFIIF) and TFIIB (347), whereas Arnt makes contact with TFIIF and CBP (170). Coactivators and corepressors might be involved not only in regulation of AhR, but also in cross-talk mechanisms between AhR and ER. Thus RIP140 (185, 210a), ERAP140, SMRT (250), and SP-1 protein (171) are involved in both AhR- and ER-mediated transcription and represent sites where the two receptors might compete.

	IX. ESTROGEN RECEPTOR-RELATED RECEPTORS: INTERPLAY WITH ESTROGEN RECEPTORS

Top Previous Next References

The estrogen receptor related receptors  and  (ERR, ERR) were first identified based on their amino acid identity to ER (118). The ERRs constitute a subgroup of orphan receptors that are evolutionarily and functionally most closely related to the steroid receptor subfamily of nuclear receptors, in particular the ERs (192, 193). These receptors display similarity to the ERs particularly in the DBD (>65% amino acid identity), and to a lesser extent in the LBD (~35% identity), but they do not bind estrogen or estrogen-like compounds. To date, three different members of this family have been identified, ERR and ERR (initially named ERR-1 and -2, Ref. 118) and recently, an ERR of human origin, probably ortologous to an ERR-3 (63, 98, 141).

Expression pattern analysis has revealed a structurally and temporally restricted expression of ERR during mouse embryo development. Essentially expression is confined to trophoblast progenitor cells in the developing chorion between day 6.5 and 7.5 days postcoitum (dpc). These structures eventually form the placenta (272), so it is not surprising that mice, homozygous for a targeted disruption of the ERR gene, die during early embryogenesis due to placental failure (213). In adult tissues, ERR expression levels appear to be low, and it is found only in a few organs (118, 272). ERR, in contrast, displays a more ubiquitous expression in adult tissues (118, 323, 333). During mouse embryogenesis, expression of ERR can be detected at day 8.5 in extra embryonic tissues, apparently sequential in time to expression of ERR, and also in the primitive heart. ERR expression is further detected from 10.5 dpc in the heart of the embryo and later at 13.5 dpc in skeletal muscle. At later stages, expression of ERR is detected in a number of tissues including adipose tissue and central nervous system