This Article Abstract Full Text (PDF) All Versions of this Article: 25/6/869    most recent Final Manuscript Author Manuscript Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Copyright Permission Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Herynk, M. H. Articles by Fuqua, S. A. W. Search for Related Content PubMed PubMed Citation Articles by Herynk, M. H. Articles by Fuqua, S. A. W.

Estrogen Receptor Mutations in Human Disease

Breast Center, Baylor College of Medicine, Houston, Texas 77030

Correspondence: Address all correspondence and requests for reprints to: Suzanne A. W. Fuqua, Ph.D., Breast Center, Baylor College of Medicine, One Baylor Plaza, Mailstop 600, Houston, Texas 77030. E-mail: suzannef{at}breastcenter.tmc.edu

	Abstract

Top Abstract I. Introduction II. ER Structure III. Roles of ER{alpha}... IV. ERs in Human... V. Splicing and Genetic... VI. Splicing and Genetic... VII. Discussion References

 
As early as the 1800s, the actions of estrogen have been implicated in the development and progression of breast cancer. The estrogen receptor (ER) was identified in the late 1950s and purified a few years later. However, it was not until the 1980s that the first ER was molecularly cloned, and in the mid 1990s, a second ER was cloned. These two related receptors are now called ER and ERß, respectively. Since their discovery, much research has focused on identifying alterations within the coding sequence of these receptors in clinical samples. As a result, a large number of naturally occurring splice variants of both ER and ERß have been identified in normal epithelium and diseased or cancerous tissues. In contrast, only a few point mutations have been identified in human patient samples from a variety of disease states, including breast cancer, endometrial cancer, and psychiatric diseases. To elucidate the mechanism of action for these variant isoforms or mutant receptors, experimental mutagenesis has been used to analyze the function of distinct amino acid residues in the ERs. This review will focus on ER and ERß alterations in breast cancer.

I. Introduction

II. ER Structure

III. Roles of ER and ERß

IV. ERs in Human Breast Cancer

V. Splicing and Genetic Alterations of ER

A. Alternative exons in the 5'UTR

B. mRNA splice variants

C. ER experimental splice variants

D. mRNA splice variants summary

E. Natural mutations of ER identified in human tissue samples

F. Natural mutations conclusions

G. Experimental mutations

H. ER experimental mutations summary

VI. Splicing and Genetic Alterations of ERß

A. Alternative exons in the 5'UTR

B. Natural mRNA splice variants

C. mRNA splice variants summary

D. Natural point mutations identified in human tissue samples

E. Experimental point mutations

F. ERß summary

VII. Discussion

	I. Introduction

Top Abstract I. Introduction II. ER Structure III. Roles of ER{alpha}... IV. ERs in Human... V. Splicing and Genetic... VI. Splicing and Genetic... VII. Discussion References

 
ABERRANT ESTROGENIC SIGNALING has long been associated with human disease such as schizophrenia, Parkinson’s [reviewed by Cyr et al. (1)], osteoporosis [reviewed by Gennari et al. (2) and Grumbach (3)], cancers of the breast [reviewed by Russo and Russo (4)], the colon [reviewed by Di Leo et al. (5)], and ovarian tissues (6). It has long been appreciated that estrogenic signaling plays a critical role in the development of breast cancer (7, 8, 9, 10). As early as the late 1800s, it was recognized that oophorectomy of premenopausal women with metastatic breast cancer caused tumor regression in approximately one third of these patients (9, 10). In the 1950s, Jensen and Jacobson (11, 12) used tritium-labeled, 17ß-estradiol to demonstrate that estradiol was specifically retained by estrogen target tissues. The specific tissue targeting of radiolabeled estradiol led them to hypothesize that a receptor must exist for this molecule. In the next decade, an estrogen receptor (ER) was identified by Toft and Gorski (13) and isolated from several mammalian species, including rat and human (14, 15, 16). However, it was not until the mid 1980s that the first ER, now called ER, was cloned by two groups of investigators (17, 18, 19, 20). In the mid 1990s, a second ER, called ERß, was identified in a library scan of rat (21) and subsequently cloned from several species including the mouse, human, and fish (21, 22, 23). At first, a human ERß with 477 amino acids was reported (23). A few months later, Enmark et al. (24) reported the identification of an ERß mRNA species with a size of 485 amino acids, and it was hypothesized to reflect full-length ERß. The following year, Ogawa et al. (25) reported the cloning of an additional ERß species consisting of 530 amino acids, which is now considered to represent full-length ERß. A few months later, Moore et al. (26) also identified the same 530-amino acid sequence as the full-length ERß, as well as various isoforms. As has been extensively shown for ER, ERß expression has also been associated with cancers of the breast (27, 28, 29, 30), colon (31, 32), and ovarian tissues (6, 33). Additionally, studies with ER and ERß knockout (KO) mice have revealed a role for ER signaling in bone formation, male and female sexual maturation, fertility, cardiovascular and angiogenesis effects, and behavior (34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57). These effects of ER and ERß KOs are completely reviewed by Couse and Korach (58) and will not be covered here. Other topics, such as the role of the two ERs in human disease (59, 60, 61), ER structure and function (62), ER interaction with other cellular signaling molecules (63, 64), ER coregulators (62, 65, 66), and the pharmacology of selective estrogen receptor modulators (SERMs) (67, 68), are also extensively reviewed elsewhere, and hence will only be modestly reviewed here for the purpose of background information. This review will instead focus on the genetic alterations that have been identified in the human ERs, including sequence mutations and RNA splice variants.

	II. ER Structure

Top Abstract I. Introduction II. ER Structure III. Roles of ER{alpha}... IV. ERs in Human... V. Splicing and Genetic... VI. Splicing and Genetic... VII. Discussion References

 
ER and ERß are separate genes, and do not represent splice variants. Accordingly, ER is found on chromosome 6q, whereas ERß is localized to chromosome 14q (24, 69, 70). To fully understand the consequences of specific ER mutations or variants, one must first be familiar with the functional domains of the two ERs. Figure 1 shows the six structural domains (termed domains A–F) of ER (71), and Fig. 2 presents the ERß structural domains as defined by Ogawa et al. (25). There is a predicted 96% homology in the DNA binding domain (C), and a 53% homology between the E/F domains, but the A, B, and hinge (D) domains are not well conserved between ER and ERß as reported by Ogawa et al. (25). In addition to their structural domains, the ERs contain defined functional domains. The transactivation domain termed activation function (AF)-1 is contained within the amino-terminal A and B domains and contains ligand-independent activation function (72, 73, 74, 75). In addition, the A/B region contains a coregulatory domain, which binds various ER coactivators and corepressors that modulate ER-mediated transcriptional activity. The C domain is composed of two zinc finger motifs and encodes the DNA binding domain that is responsible for binding to specific estrogen response elements (EREs) within the promoters of estrogen-responsive genes (72, 76, 77). The ER dimerization domain is discontinuous, is split between the C and E domains, and is required for the ERs to dimerize, allowing binding to the entire ERE site (72). The structural D domain contains the hinge region, part of the ligand-dependent, transactivation domain AF-2a and a portion of the ER nuclear localization signal (78, 79). The carboxy-terminal E and F regions contain the ligand binding domain and the ligand-dependent AF-2 transactivation domain (78). Finally, as previously mentioned, this carboxy-terminal region is also involved in receptor dimerization, the binding of coregulatory proteins, and the binding of chaperone proteins, such as heat shock proteins 70 and 90 (80, 81).

View larger version (15K): [in this window] [in a new window]   FIG. 1. Domains of ER. A, The mRNA sequence of ER. Alternative promoters are shown to the left of +1. The shaded box shows the ER coding region. Exons are numbered in the corresponding blocked region with the nucleotide number above. ATG start codon and the TAG stop codon are shown below. B, The protein domains are labeled A–F, nucleotide numbers corresponding to the start of each domain are above, with amino acid numbers below. Relative positions of some of the known functional domains are represented by solid bars below. BD, Binding domain.

View larger version (12K): [in this window] [in a new window]   FIG. 2. Domains of ERß. A, The mRNA sequence of ERß. Alternative promoters are shown to the left of +1. The shaded box shows the ERß coding region. Exons are numbered in the corresponding blocked region with the nucleotide number above. ATG start codon and the TAG stop codon are shown below. B, The protein domains are labeled A–F, nucleotide numbers corresponding to the start of each domain are above, with amino acid numbers below. Relative positions of some functional domains are represented by solid bars below.

The ligand-independent AF-1 domain can be activated by cAMP, dopamine, vanadate, and growth factors such as epidermal growth factor (EGF) and IGF (77, 101, 102, 103, 104, 105, 106). The AF-1 domain demonstrates ligand-independent activation that is closely related to its phosphorylation status (102, 105, 107). The A/B region also contains a coregulator binding domain, where coactivators such as p68 and steroid receptor coactivator (SRC)-1 and corepressors such as Ssn3 bind and modulate ER-mediated transcriptional activity (108, 109, 110, 111, 112). AF-2 and AF-2a demonstrate ligand-dependent activity (72, 73, 77, 78, 113). AF-2 also contains coregulator binding sites for the coactivators SRC-1, -2, -3 and CREB binding protein (CBP), as well as the corepressors thyroid hormone receptor interacting protein (TRIP1) and repressor of estrogen activity (REA), to name only a few (108, 110, 114, 115, 116, 117, 118, 119, 120). Mutations in these AF-1 and AF-2 functional domains can alter ER signaling by activating or inactivating the protein or by altering the binding of coregulator proteins and, indirectly, modulating ER signaling.

The crystal structure of ER, ERß, and several other nuclear receptors bound to agonists has demonstrated that ligand binding to the carboxy-terminal hydrophobic pocket induces helix 12 to position itself over the pocket (121, 122, 123, 124, 125). This repositioning stabilizes helix 12, allowing it to recruit transcriptional coactivators required for full agonist action (126, 127). When ER is bound to partial agonists or antagonists such as tamoxifen, Faslodex, or raloxifene, the "bulky" side chain of the compound prevents helix 12 from adopting the agonist bound position over the pocket, thus antagonizing coactivator binding to the ERs (124, 128, 129, 130, 131, 132). Compounds without bulky side chains, such as genestein or THC (5,11-cis-diethyl-5,6,11,12-tetrahydrochrysene-2,8-diol) inhibit full ER activation by stabilizing nonproductive conformations of the ligand-binding pocket (128, 129). The mechanism of tamoxifen-mediated repression has been well studied and has demonstrated that tamoxifen-bound ER can recruit corepressors such as nuclear receptor corepressor (NCoR) and silencing mediator of retinoid and thyroid receptor (SMRT) (133, 134, 135, 136, 137, 138). In contrast, ERß has been shown to recruit these corepressors in the presence of agonists, but not antagonists (139). Furthermore, reduced expression of NCoR and SMRT has been correlated with tamoxifen resistance in breast tumors (134, 135, 136, 138, 140), demonstrating the importance of putative helix 12 mutations. This review will focus on ER mutations that lead to alterations within the protein sequence, particularly those that alter AF-1 or AF-2 function. The many ER mRNA splice variants that have been identified, which predict variant forms of the ER proteins, will also be discussed.

	III. Roles of ER and ERß

Top Abstract I. Introduction II. ER Structure III. Roles of ER{alpha}... IV. ERs in Human... V. Splicing and Genetic... VI. Splicing and Genetic... VII. Discussion References

 
With the relatively recent discovery of ERß, the separate cellular roles for ER and ERß are just now being elucidated. ER and ERß have been demonstrated to form heterodimers, as well as homodimers, further complicating their individual and/or combined function within a cell. Although both receptors bind estrogen with similar affinities, ERß appears to have a stronger affinity for phytoestrogens (141, 142, 143). ERß also exhibits reduced transactivation in most cells when directly compared with ER, and this is likely due to the weaker AF-1 activity of ERß (144, 145, 146). In contrast, AF-2 activity is similar for both ER and ERß, depending on the cell-type (147, 148). These receptors display differential ligand-induced activity; therefore, it is predicted that they could also differentially regulate transactivation of heterologous promoters (149). Paech et al. (149) have shown that estrogens up-regulate ER activating protein (AP)-1 activity, whereas ERß AP-1 activity is reduced. In contrast, the antiestrogens tamoxifen, raloxifene, and ICI 164,384 all increase AP-1 activity of both ER and ERß. Furthermore, serotonin-1A has been shown to be specifically up-regulated through nuclear factor-B (NF-B) induced by ER signaling, but not ERß, demonstrating receptor gene specificity (150). When MDA-MB-231 cells were engineered to overexpress either ER or ERß, both ER and ERß decreased in vitro invasion (151); however, ER-overexpressing cells reduced proliferation in a ligand-dependent manner, whereas ERß was able to repress proliferation in a ligand-independent manner, thus demonstrating divergent roles for these receptors. Although ER and ERß share similar ligand specificities and some signaling actions, they appear to respond to ligands in a receptor-specific manner.

Work with ER KO mice has demonstrated that these receptors are not dependent on or under the control of each other, because mice deficient in one receptor are not lacking in the other (152, 153, 154, 155, 156). Mammary glands of ER KO mice have normal embryonic and fetal development, but these glands never develop beyond the newborn stage (41). In contrast, mammary glands of ERß KO mice develop normally with ductal structures that fill the fatpad but exhibit reduced side branching in nulliparous glands (37, 157), thus demonstrating that ER is the predominant receptor involved in mammary gland development. It should be noted that, in a number of studies, ER KO mice, although lacking the full-length ER, express shorter isoforms of ER mRNA and protein with some residual signaling activity (38, 158, 159). Examination of mammary glands from pregnant and lactating ERß KO mice has revealed that ERß expression is required for normal lobuloalveolar development (157). Additionally, analysis of tight junction proteins, gap junction proteins, smooth muscle actin, and Ki67 expression all suggested that ERß KO mice have less well-differentiated mammary glands compared with wild-type mice (157). Because ER KO mice do not develop normal adult mammary glands, it will be necessary to develop conditional KOs to more adequately model human breast cancer development and progression and to fully delineate hormone action during development.

Because ER KO mice have significantly impaired mammary gland development, it is perhaps not surprising that these mice would be resistant to 7,12-dimethyl benz[a]anthracene (DMBA)-induced mammary tumors (160). Interestingly, when mice lacking ER expression were crossed with transgenic mice overexpressing the mouse mammary tumor virus-Wnt-1 oncogene or the mouse mammary tumor virus-Her2/neu oncogene, mammary tumors did indeed develop (161, 162); however, tumor development was delayed when compared with mice expressing wild-type receptor (161, 162). Collectively, these data demonstrate that although ER can contribute to mammary tumorigenesis, it is not absolutely required.

	IV. ERs in Human Breast Cancer

Top Abstract I. Introduction II. ER Structure III. Roles of ER{alpha}... IV. ERs in Human... V. Splicing and Genetic... VI. Splicing and Genetic... VII. Discussion References

 
ER signaling is known to be necessary for the proper development and maturation of the mammary gland by stimulating DNA synthesis and promoting bud formation (163, 164, 165, 166). It is estimated that only 7–10% of the epithelial cells in the normal human breast express ER, and it has been shown that this expression fluctuates with the menstrual cycle (167, 168, 169, 170, 171, 172, 173). Although only a small percentage of the cells in the normal breast express ER, these are not the same cells as those that are proliferating (174, 175, 176). In contrast, ERß expression is relatively high in the normal breast, with 80–85% of the cells expressing ERß, which is again inversely correlated with cellular proliferation (59, 177). In contrast, ERß expression does not appear to change during the menstrual cycle (178, 179, 180). Although ER signaling is required for normal mammary gland development, it has been hypothesized that aberrant signaling could lead to abnormal cellular proliferation and survival, potentially participating in the development and progression of breast cancer. Similar to invasive breast cancer, low-grade ductal carcinoma in situ (DCIS) has been demonstrated to have 75% of the cells expressing high levels of ER, but high-grade DCIS has approximately 30% of the cells expressing low levels of ER (181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, 192, 193). Additionally, it has been shown that DCIS lesions have reduced ERß expression compared with normal epithelium, with high-grade DCIS showing the most significant reduction in ERß levels (177, 178). Although reduced, invasive breast carcinomas show high levels of ER and ERß with approximately two thirds of the tumors staining positive by immunohistochemistry (28, 180, 194, 195, 196). To date, few large studies have been performed analyzing ERß protein expression in normal breast, early lesions, and invasive cancers. Two recent studies, although having varying cutoffs, have demonstrated reduced intralesional ERß expression in DCIS when compared with normal epithelium (197, 198). Additionally, these studies demonstrated a further loss of ERß expression from DCIS to invasive cancer. One study demonstrated a 21% reduction in tumors expressing ERß (194), whereas a second study demonstrated a reduction in intralesional ERß expression and not a reduced number of invasive breast cancers expressing ERß (198). Interestingly, when primary tumors and matched lymph node metastases where compared, there was not a significant reduction in lymph node metastases staining positive for ERß expression (197). Although it has been much hypothesized that ERß has tumor-suppressor-like activity in the breast (199), ERß would have to be unique among tumor suppressors if expressed in over 75% of invasive lesions (178). Undeniably, much more work is required to understand this paradox. Collectively, these data indicate that although ERß levels may vary, ER expression levels rise during tumor progression.

The potential role of ERß in breast cancer progression is highly controversial. Many studies have suggested that ERß expression is a favorable prognostic indicator, whereas additional studies have suggested that ERß expression is associated with known factors of poor clinical outcome. Basically, two types of prognostic studies have been performed to date, those evaluating RNA levels, and those evaluating protein expression. It is interesting to note that many of the studies indicating that ERß is a poor prognostic indicator have evaluated only the RNA levels by quantitative or semiquantitative PCR techniques. Many of these RNA-based studies have correlated ERß expression with markers of a poor prognosis, such as EGF receptor expression and high tumor grade, and an inverse correlation between ERß expression and progesterone receptor (PR) status (28, 30, 200, 201). However, a few studies evaluating RNA have demonstrated that ERß expression is reduced in breast cancer compared with normal epithelium, and that it is inversely correlated with proliferation (evaluated by measuring Ki67 levels) (177, 202), thereby suggesting that ERß expression is a favorable prognostic indicator. However, PCR analysis of RNA levels from tumor samples will also measure ERß mRNA in the "normal" surrounding cells, stroma, and contaminating immune cells present in homogenized tissues. Additionally, many PCR primers may also amplify alternatively spliced RNA variants, thereby increasing the false-positive rate or perhaps skewing results toward higher expression levels. Thus, protein analyses would more precisely measure ERß expression levels in clinical samples.

Studies evaluating ERß protein expression appear to be much less contradictory. Direct protein analyses generally suggest that ERß protein expression is a favorable prognostic indicator, correlating with known biomarkers such as low histological grade, ER and PR expression, longer disease-free survival, and response to tamoxifen (180, 195, 196, 197, 203, 204, 205). Although these studies do not always agree on the specific associations with known prognostic indicators, they do generally agree that similar to ER, ERß expression is a favorable prognostic indicator. A few protein-based studies have suggested that ERß expression is associated with high proliferation (Ki67 expression) and high tumor grade (206, 207). However, these studies have varying cut-off points for being classified as ERß-positive, requiring 20–25% of the cells staining positive for ERß. Furthermore, these later studies examined small tumor subsets. Mann et al. (205), have demonstrated that tumors with as little as 10% of the cells expressing ERß have a more favorable response to tamoxifen, thus suggesting that only 10% positive cells may be a reasonable cut-off point for classification of ERß status in tumors, as has been adopted for ER and tamoxifen responses (208). However, these methods give an estimate of the percentage of positive cells, but they do not account for expression levels in individual cells. The Allred score, used in several studies, measures both the percentage positive and the relative intensity, thus providing a semiquantitative method of ER protein expression (209). It is clear that the role of ERß in breast carcinogenesis has not been fully elucidated; however, direct protein analysis strongly suggests that ERß is an indicator of a more favorable clinical outcome. A uniformly adopted classification of ERß expression will be required to reconcile these issues and may help to clarify the potential role of ERß in breast cancer progression.

It has also been suggested that it is not necessarily the individual level of ER or ERß that is clinically relevant, but the ratio of ER:ERß that may change and impact tumorigenesis. In support of this hypothesis, it has been shown that ER-positive breast cancer has a mean higher ER:ERß ratio when compared with the normal tissue; in contrast, estrogen-independent, ER-negative cancer exhibits a low ER:ERß ratio (201, 210). Further complicating this hypothesis, however, is the existence of ER and ERß splice variants that may contribute to measurements and result in an overestimation of mRNA or protein expression. It is evident that the interplay between ER and ERß could be complicated; data exist that both may have roles within normal breast epithelial cells and that up-regulation or down-regulation of one receptor could potentially upset a physiological balance. Thus, although a definitive role for ERß in breast carcinogenesis has not yet been demonstrated, it can be concluded that ER appears to play the dominant role in the breast.

	V. Splicing and Genetic Alterations of ER

Top Abstract I. Introduction II. ER Structure III. Roles of ER{alpha}... IV. ERs in Human... V. Splicing and Genetic... VI. Splicing and Genetic... VII. Discussion References

 
A. Alternative exons in the 5'UTR
The majority of efforts to understand ER action have focused on its functional domains and protein-protein interactions. Over the past several years, many investigators have separately identified and named as many as eight upstream untranslated ER exons (69, 211, 212, 213, 214, 215, 216). Because these exons have been identified and differently named, this review will use the nomenclature suggested by Flouriot et al. (216), as modified by Kos et al. (217) and shown in Fig. 1A. ER exon 1, as defined by Green et al. (18), contains an acceptor splice site at +163, permitting the splicing of several different exons encoding various 5' untranslated regions (UTRs). To date, at least seven different promoters have been described, allowing for a relative amount of tissue specificity [for a complete review, see Kos et al. (217)]. The most common promoter found expressed in tissues and cell lines is encoded in exon 1 and is termed promoter A. What is now called promoter C was first described in 1991 (214), and a longer version of promoter C was described a couple of years later (212, 214). In subsequent years, exons A–E were described. These various upstream exons have been shown to affect reporter gene expression levels (218). Numerous AUG start codons are found in these various 5'UTRs and are thought to inhibit the scanning ribosomes from reaching the start codon responsible for full-length ER translation, thus reducing ER protein expression (218). The tissue specificity of these various 5'UTR promoters has been examined (216, 219). It was found that the promoters within 2-kb pairs of the acceptor splice site, namely promoters A, B, and C, are predominately used in cell lines and tissues expressing relatively high levels of ER, whereas the more distal promoters, named E and F, are found in tissues where ER is less abundant, such as the liver and osteoblasts (219). Although these promoters demonstrate some tissue-specific expression, Brand et al. (211) have reported the identification of two new exons, called T1 and T2, that are expressed predominantly in the testis and the epididymis. Because these different promoters can indeed control tissue-specific expression, as well as ER levels, the inappropriate splicing of these promoters may affect the expression and ER signaling activities.

B. mRNA splice variants
ER splice variants have been detected in a number of different normal tissues, including the breast, endometrium, and pituitary tissues, as well as smooth muscle cells and peripheral blood mononuclear cells (220, 221, 222, 223, 224). Additionally, ER mRNA splice variants have been detected in various tumor types including breast cancer, endometrial carcinoma, prolactinoma, systemic lupus erythematosus, and meningiomas, to name a few (see Table 1 and references therein). In the vast majority of cases, wild-type ER is coexpressed along with variant ER mRNAs (225, 226, 227). Although many of these variants have been predominately detected in diseased tissues, a number of studies have been unable to demonstrate differences in the expression levels, or individual patterns of mRNA splice variants when comparing normal controls from unaffected patients to diseased tissues, suggesting that these variants may also play a role in normal physiological processes (220, 221). A large number of splice variants have been reported; therefore, only the most common RNA splice variants will be discussed herein. Unfortunately, the study of predicted proteins of these RNA variants is rare, but we will include a discussion of protein isoforms when appropriate.

View this table: [in this window] [in a new window]   TABLE 1. ER splice variants

1. ER exon 2 deletion (2).

The 2 variant has been detected in a number of tissues. It is interesting to note that in diseases such as breast cancer and systemic lupus erythematosus, where this variant has been identified in both the normal and diseased tissue, no associations were found between relative expression and the particular disease state (221, 231). Additionally, 2 has also been identified in normal endometrium, but not in hyperplastic endometrium or endometrial adenocarcinoma (232, 233). However, this ER variant was not found in the normal pituitary but has been identified in prolactinoma, a tumor arising from the pituitary (222). These data suggest that the exon 2 deleted ER splice variant may be regulated in a tissue-specific manner, but probably does not play a significant role in tumorigenesis.

2. ER exon 3 deletion (3).
The ER 3 splice variant results in an in-frame shift that is missing part of the DNA binding domain. This deletion results in a protein with dominant-negative activity that is able to suppress estrogen-induced transcriptional activity (228, 234). Because this variant is unable to bind DNA, its ability to act as a dominant-negative isoform most likely occurs through dimerization with wild-type ER (228). Stable transfection of ER 3 into the ER-positive MCF-7 breast cancer cell line resulted in an 80% reduction in invasion using a chick embryo chorioallantoic membrane assay (235). Treatment of these 3-expressing cells with estrogen also reduced soft-agar colony-forming ability to below basal levels (235). Because 3 is able to inhibit estrogen-induced transcriptional activity, its expression could potentially affect ER signaling.

The 3 variant has been detected in prolactinoma, endometrial hyperplasia, and breast cancer (222, 225, 233). Comparison of the ER 3:wild-type ER ratios demonstrated that normal breast tissues have a median ratio of 3.4, whereas breast cancers, due to 3 down-regulation, have a median ratio of 0.11 (235). Additionally, 3 has been detected in the majority of ER-positive, PR-negative breast tumors (225), suggesting a role in the discordant receptor phenotype. 3 was not detected in normal pituitary, normal endometrium, or endometrial carcinoma (222, 233). This variant was also found in 19 of 21 endometrial hyperplasias but was not detected in 29 endometrial carcinomas (233). The dominant negative activity of 3 may serve to reduce normal estrogenic signaling, thereby influencing tumor progression and growth. It is interesting to note the tissue-type manner in which this variant has been found. Although prolactinomas and endometrial hyperplasias expressed 3, endometrial carcinomas did not; and furthermore, normal breast tissues have significantly reduced levels, prompting the question does 3 have contrasting roles in alternatively promoting or protecting from tumorigenesis in a tissue-type dependent manner? In addition, ER 3 expression is reduced by over 30-fold in breast cancer when compared with normal breast epithelium; and because ER 3 has dominant negative activity [requiring significantly higher levels of 3 for dominant negative activity (236)], the loss of ER 3 may "relieve" normal physiological repression of ER signaling. To our knowledge, no antibodies specific for 3 exist to test this hypothesis.

3. ER exon 4 deletion (4).
Deletion of ER exon 4 results in an in-frame deletion that encodes a protein lacking a nuclear localization signal, the AF-2a activation domain, and part of the hormone binding domain. The resulting variant isoform is unable to bind hormone or DNA, and therefore has no basal or estrogen-induced transcriptional activity (237). Additionally, this alternately spliced ER isoform does not appear to interfere with wild-type ER activity (237). Although these data suggest that expression of this variant would be of no significant consequence to a cell, its RNA expression has been associated with biomarkers of a more favorable clinical outcome for breast cancer, such as low grade and high PR levels (225, 238). In addition, 4 RNA is more common in PR-positive breast cancer (225, 238). When endometrial hyperplasia and adenocarcinomas were examined for the presence ER 4, it was found in 17 of 21 hyperplasias but was absent in 29 adenocarcinoma samples (233). It is not understood how 4 expression could associate with markers of a more favorable clinical outcome for breast cancer; because it has no appreciable binding or transcriptional activity, larger studies will be required to demonstrate that 4 expression is indeed significantly associated with a more favorable clinical outcome for either breast or endometrial carcinoma?

4. ER exon 5 deletion (5).
We first identified the ER exon 5 deletion isoform that results in the introduction of a stop codon within the ligand binding domain (239). Although the resulting 40-kDa protein lacks most of the ligand binding domain, it retains AF-1 activity and DNA binding ability (239). The 5 retains its ligand-independent transactivation domain AF-1; therefore, it is not surprising that the encoded protein is constitutively active in a yeast transactivation assay or transfected into some breast cancer cell lines (239, 240, 241), although not all cell lines (242). Many studies have attempted to correlate expression of the 5 splice variant with tamoxifen resistance in clinical samples. One study by Madsen et al. (230) found similar levels of 5 mRNA in both tamoxifen-resistant and tamoxifen-sensitive MCF-7 cells. In support of this in vitro result, two separate studies reported by Daffada et al. (243) and Zhang et al. (225) examined 120 and 109 primary breast tumors, respectively, and did not find any correlation between 5 mRNA expression levels and the tamoxifen-resistant phenotype (225, 243). In contrast, several in vitro studies in our laboratory have shown that overexpression of ER 5 in MCF-7 cells confers relative tamoxifen resistance (240, 244). In these studies, we used exogenously expressed 5 in MCF-7 cells, whereas the studies by Madsen et al. looked for levels of this splice variant in acquired tamoxifen-resistant cells. These data suggest that although 5 is able to confer tamoxifen resistance in vitro, it is probably not a major mechanism of de novo or acquired tamoxifen resistance in invasive breast cancer.

ER 5 was first identified in a small number of ER-negative, but PR-positive breast cancers, and these tumors tend to have higher expression levels of 5 (239, 245) relative to wild-type ER. In addition to primary breast cancer, 5 has also been found in normal breast tissue, and is expressed at increased levels in breast cancer metastases (225, 231, 246, 247). Interestingly, this variant has not been found in breast hyperplasias but has been found at reduced levels in normal tissue adjacent to breast cancer (231, 240, 248). Regardless of these correlative findings, 5 levels have not associated with known clinical prognostic indicators such as ER or PR status, tumor size, or S-phase fraction (225). Thus, although expression of ER 5 mRNA is elevated in breast cancers compared with normal tissues, collectively, these data do not support a dominant role for ER 5 in breast tumorigenesis, and convincing evidence for protein expression in patient samples has not been presented.

In addition to breast cancer, other tumor types exhibit expression of 5; pituitary tumors, but not normal pituitary, express 5 (222). In addition, this is the only ER splice variant whose expression has been detected at a significantly increased level in endometrial carcinomas, compared with endometrial hyperplasias (233). Peripheral blood mononuclear cells in patients with systemic lupus erythematosus also express either wild-type ER or the 5 variant, but not both isoforms (249, 250). Thus, at present one can conclude that although 5 does not appear to play a significant role in breast tumorigenesis, there may be an as yet undefined role for 5 in other tumor types.

5. ER exon 6 deletion (6).
There have been few reports of an ER exon 6 deletion variant. Poola and Speirs (251), using RT-PCR assays, first found this variant in one of 35 normal breast samples, but in seven of 38 clinical breast cancer specimens. Using this same technique, 6 has been detected in several ER-positive breast cancer cell lines, including MCF-7 and T47D (252) (see Table 1 for a complete listing). The exon 6 region encodes a portion of the hormone binding and dimerization domains, and it has been demonstrated using experimental mutagenesis (see Table 3) that many important functional amino acids reside within this exon. Although this region is rarely deleted, it is commonly duplicated (Table 1), and three naturally occurring mutations lie within this exon (Table 2). These data are suggestive of an important role for exon 6 in ER signaling, and may help to explain why this particular exon is rarely deleted in vivo.

View this table: [in this window] [in a new window]   TABLE 3. Experimental mutations in ER

View this table: [in this window] [in a new window]   TABLE 2. ER point mutations found in patients

6. ER exon 7 deletion (7).

7. Multiple exon deletions.
In addition to single exon deletions, a number of multiple and partial exon deletions have been found in different normal and neoplastic tissues (Table 1). Although normal breast tissue expresses primarily single exon deletion variants, breast cancer tissues appear to express a number of multiexon splice variants, (251), but their clinical significance remains to be fully discovered. One study by Leygue et al. (238) demonstrated that deletions in exons 2–4 or exons 3–7 are associated with high-grade breast cancers and elevated ER expression. In addition, many of these multiple ER splice variants do not appear to affect the transactivation function of wild-type ER (257). Thus, although a large number of multiple exon splice variants have been identified to date, their role, if any, in breast cancer remains unclear.

8. ER insertion and exon duplications.
In contrast to the lack of single exon 6 deletion variants, a number of the ER variants with multiple exon duplications or multi base-pair insertions involve exon 6; 7.5% of breast cancers examined (n = 212) demonstrate a duplication of exon 6 (226). This duplication results in a truncation immediately after the duplicated exon, leading to a 50-kDa protein that lacks the AF-2 and dimerization domains (226). A duplication of both exons 6 and 7 results in an 80-kDa protein that has lost the ability to bind ligands, such as estrogen or tamoxifen (258, 259, 260, 261). This duplication was originally identified in an estrogen-independent MCF-7 subline and is the result of a genomic rearrangement, rather than alternative RNA splicing (258). A 69-nucleotide insertion between exons 5 and 6 has also been identified in three of 212 breast cancer tumor specimens analyzed (226). This 69-nucleotide insertion results from a point mutation in the intron creating a consensus splice donor site (262). Because a consensus splice acceptor site is located 5' to the 69-nucleotide sequence, this short intron sequence is recognized as an exon (262). Karnik et al. (263) examined tamoxifen-resistant breast cancers for mutations and identified a 42-bp insertion within exon 6. As with the other major ER splice variants, the individual importance of ER exon 6 remains unclear, because it is rarely deleted and is a common site for duplications, suggesting that exon 6 may play an important role in the general function of ER.

C. ER experimental splice variants
The human ER (HE) series of experimental splice variants are widely used both as experimental variants and as backbone plasmids for additional mutational analysis. Originally, 14 deletion and/or truncation mutants were made and called HE1–HE14, and they helped to identify the functional domains of ER (264). In subsequent years, many additional variants with single or multiple deletions and N-terminal and C-terminal truncations have been made; however, these are too numerous to mention, so they will not be discussed here (73, 75, 81, 265). Figure 3 shows the deleted portions of the ER protein in the original HE series mutants (264). HE5–HE9 mutants were unable to bind estradiol, indicating that the hormone binding domain was within amino acids 301–552 (264). Deletions that altered domain C (HE3, HE4, or HE11) or domain D (HE12) reduced nuclear association and ERE binding (81, 264). These original mutants as well as the continued series of deletions and truncations have been valuable tools in mapping the functional domains of ER.

View larger version (22K): [in this window] [in a new window]   FIG. 3. Experimental deletion mutants. The original HE series of experimental deletion mutants as published by Kumar et al. (264 ) is shown. The relative positions of the deleted regions are shown by a break in the horizontal solid black bars. The deleted amino acids are shown in the far right column, with the mutants name in the far left column. The vertical black lines correspond to the A–F domains of ER.

Compared with the number of mRNA splice variants, relatively few variant protein isoforms have been examined, in part, due to the lack of antibodies with the ability to detect specific deleted exons. Although PCR and sequencing can determine the ER exact deletion and/or insertion, antibodies reacting with a different size band only give the relative size and the epitope, but not the isoform identity. A number of groups have compared amino-terminal, carboxy-terminal, and exon-specific antibodies in immunohistochemical studies. These studies have demonstrated that the pattern of immunohistochemical staining is not always identical for antibodies reacting with different epitopes (266, 267). Typically, the amino-terminal signal was found to be stronger than the carboxy-terminal signal, suggesting the presence of truncated variants of ER in clinical samples (266, 268). Western blot analysis of tumor samples has demonstrated a number of different ER isoforms, corresponding to the predicted protein sizes of specific molecular weight forms of mRNA splice variants (245, 246, 269). For instance, Desai et al. (246) developed an ER 5-specific antibody and demonstrated a positive correlation between ER 5 expression and disease-free survival in breast cancer patients. Thus, whereas the number of detected alternatively spliced ER mRNAs is large, the number of variant isoforms that appear to be stably translated in vivo is infrequent.

In vitro analysis of ER variant isoforms has demonstrated a range of activities, including constitutive activation and dominant negative phenotypes. Theoretically, a small amount of a constitutively active isoform could significantly up-regulate total ER activity. In contrast, numerous studies have demonstrated that significantly higher amounts of dominant negative receptor are required to obtain significant inhibition of wild-type ER signaling. For instance, dose-dependent inhibition of ER transactivation has been demonstrated with a number of different dominant negative constructs, including ER 3 and ER 7 as well as ERß 5 (236, 254, 270). Erenburg et al. (235), using transient transfection assays, demonstrated that 10 times more ER 3 is required to obtain an 80% inhibition of estrogen-induced ER transactivation in HeLa cells. MCF-7 cells stably expressing ER 3 at varying levels showed a dose-dependent reduction in pS2 mRNA, with the greatest reduction in cells expressing higher levels of variant than wild-type protein. These in vitro studies support the concept that significantly higher amounts of dominant negative ER, compared with wild-type receptor, are required to inhibit ER function. Because it has been shown that relatively few tumors demonstrate variant expression levels higher than wild-type ER (225), thus is the expression of variant ER isoforms really playing a significant role in this small subset of tumors? Although 10 times more DNA was required to reduce transactivation in transient transfection assays, stable transfection and expression of similar or even slightly less ER 3 did significantly reduce anchorage-dependent growth, soft-agar colony-forming ability, and in vitro invasion of MCF-7 breast cancer cells (235). This later result could suggest that although low level dominant negative inhibition may not dramatically affect transactivation, it may significantly impair the ability of a cell to survive and grow, at least in vitro.

E. Natural mutations of ER identified in human tissue samples
Several groups have reported single mutations within the ER sequence, however many of these are silent mutations or polymorphisms that do not affect the protein sequence. Some of the identified mutations do result in an altered ER protein sequence and have been detected in a variety of tissues and diseases, including breast cancer and its metastases, endometrial cancer, and physiological disorders (Table 2). Only those ER mutations that have been identified in breast cancers will be discussed here.

Although two thirds of all breast cancers express ER, mutations within the ER sequence are relatively rare. One study found that mutations occur in only 1% of primary tumors (271). It has been suggested that the mutation frequency may be higher in metastatic breast tumors, and some of these have been correlated with tamoxifen resistance and estrogen independence (263, 272). For instance, recent studies have demonstrated increased mutation rates in metastatic breast lesions (272). We have identified a somatic mutation at nucleotide 908 of ER in 34% of hyperplastic breast tissues (273). This later study used newer, more sensitive detection techniques, which, due to the heterogeneity of tumors, may be required to accurately detect ER mutations to determine realistic mutation rates. Although the rates of mutations are low in the primary tumor and increased in metastases, many of the single mutations identified to date have been demonstrated to influence breast cancer cell behavior.

1. Mutations that do not affect ER functionality.
Some of the ER point mutations do not demonstrate any observable phenotype different from that of wild-type ER. For instance, ER S47T and K531E mutations identified from metastatic breast cancer specimens were analyzed for their ability to transactivate four different ERE reporter constructs in HeLa and MDA-MB-231 breast cancer cells (272). The transactivation profiles of these two mutants did not differ from the transactivation status of the wild-type ER. Roodi et al. (271) have also identified two single mutations from an individual patient with a breast cancer metastasis. These two mutations, ER N69K in the AF-1 transactivation domain and ER M396V in the ligand binding domain, were identified from a tumor that was clinically ER-negative (271). A missense mutation in the ER D domain was identified as a leucine to proline substitution at amino acid 296 (274). An ER E352V mutation was also identified in a breast cancer patient who responded to adjuvant tamoxifen therapy (263). Although these later mutations do not observably alter ER functionality, many mutations have been identified that do dramatically affect the regulation of wild-type ER and will be discussed next.

2. ER A86V.
One of the first naturally occurring missense mutations identified in breast cancer specimens was ER A86V (275, 276). This mutation was found to be present in a subset of patients originally identified with a silent mutation at nucleotide 261 (275). This combination of a silent mutation and a missense mutation has been found to be associated with several distinct phenotypes. First, women expressing this variant allele are significantly taller (277), and ER-positive breast cancer patients have an elevated risk of spontaneous abortions if this double mutation is present (278). Additionally, ER-positive breast tumors expressing this allele tend to have lower levels of ER protein when compared with tumors expressing wild-type ER (275). These data all demonstrate that a single mutation in ER can have profound biological effects. It will prove useful when these studies are expanded to larger patient populations.

Because ER-positive tumors expressing this variant allele demonstrated reduced ER protein levels, Schmutzler et al. (279) hypothesized that it may be one mechanism leading to ER-negative tumors. To test this hypothesis, they examined genomic DNA, but found an approximately equal distribution of this variant allele (279). In total, of the tissue samples from 483 women that were analyzed, 59 were found to carry this variant allele (279). Additionally, blood samples from healthy women expressed the variant allele 9.5% (279) of the time. Interestingly, all carriers were heterozygous, suggesting that homozygous expression of this variant allele may not be viable. However, through power calculations, these authors determined that at least 960 samples would have to be tested to determine whether indeed the A86V homozygous cell is not viable.

3. ER K303R.
Our group has identified a K303R somatic mutation at the border of the hinge and hormone binding domains of ER in 34% of hyperplastic breast lesions (273). Additionally, we have detected this same mutation in the majority of primary breast cancers (our unpublished results). In contrast to our findings in samples from the United States, Zhang et al. (280) were unable to find this mutation in a cohort of breast cancer samples from Japanese women. The reason for this ethnic difference is currently under investigation. Because this mutation is present only in tumor samples, and not in normal tissues, it may have an important role in tumor progression.

To examine a potential role for this mutation in tumor progression, MCF-7 cells with enforced expression of K303R ER were analyzed. Both the wild-type and the K303R ER mutant receptors have similar binding affinities for estrogen and tamoxifen (273). Although these receptors have similar binding affinities, cells expressing the mutant receptor responded to 10^–12 M estrogen as well as they responded to 10^–9 M estrogen in an in vitro growth assay. MCF-7 cells expressing the wild-type receptor had a doubling time of 2.2 d and 1.3 d when grown in 10^–12 and 10^–9 M estrogen, respectively (273). In contrast, cells expressing the mutant receptor had a doubling time of 1.3 d, regardless of whether the cells were grown in 10^–12 or 10^–9 M estrogen (273). Because the estrogen binding affinities were identical, the binding of the ER coactivator transcriptional intermediary factor 2 (TIF2) was analyzed. In vitro binding experiments revealed increased ER association with transcriptional intermediary factor 2 at low levels of hormone, suggesting one potential mechanism for the observed increased sensitivity to estrogen (273). We hypothesize that the K303R ER mutation reduces the concentration of hormone required for the formation of the coactivator:ER hydrophobic groove binding surface (131, 273). Lysine at amino acid 303 has also been identified as a major acetylation site within ER (281). ER acetylation was shown to modulate the response to ligand and may, in part, be a component of the mechanism leading to the increased estrogen hypersensitivity associated with this mutant receptor (281). These data demonstrate that the K303R ER mutation has acquired the ability to respond to much lower concentrations of estrogen, e.g., a gain of function mutation. Because the hypersensitive mutation has been found in a large percentage of primary tumor samples, we are currently examining its clinical role as a prognostic biomarker.

4. ER 437Stop
The 437Stop mutation results from the deletion of a nucleotide in codon 432 leading to a frameshift and the introduction of a stop codon at residue 437 (263). This mutation was first identified in a patient that had relapsed while on tamoxifen (263). Interestingly, this mutation was found only in the metastatic lesion and not in the primary tumor, suggesting a role in tamoxifen resistance and/or metastatic spread (263). Furthermore, Graham et al. (282) have identified an ER 417Stop mutation in the tamoxifen-resistant T47D_CO cell line. These two mutants suggest that loss of the ligand binding domain could be one mechanism, albeit infrequently, of acquired tamoxifen resistance. However, there have been no reports of these truncated proteins in vivo in tamoxifen-treated patients.

5. ER Y537N/Y537S.
Before identifying Y537 as a site of natural mutation in breast cancer, much work had already been done with mutations at this site, demonstrating that this tyrosine is an important phosphorylation site with potential roles in regulating ligand binding, homodimerization, and transactivation of ER. These experiments, as well as additional Y537 experimental mutations are discussed in Section V.G. dealing with experimental mutations. Because of this earlier work, it was very exciting when Kohler et al. (283) identified a Y537S mutation in endometrial cancer and we found the Y537N mutation in an ER-negative metastatic breast cancer patient (272). Because many publications have analyzed the Y537S mutation in conjunction with other Y537 mutants, the in vitro analysis of the Y537S mutation will be further discussed in Section V.G. dealing with experimental mutations. Transactivation analysis of the Y537N mutation was analyzed in both HeLa and MDA-MB-231 breast cancer cells utilizing four different ER-responsive promoter constructs: vitellogenin, pS2, cathepsin D, and lactoferrin (272). It was demonstrated that, in the absence of estrogen, the Y537N mutant exhibited 5- to 20-fold higher levels of transactivation (vs. wild-type ER). Because this tyrosine residue is an important ER phosphorylation site, a mutation at this site may allow ER to escape phosphorylation-mediated controls (272, 284, 285, 286) and provide a cell with a potential selective tumorigenic advantage. Because only a relatively few metastatic tumors have been examined for mutations at Y537, its frequency may be underappreciated.

6. ER silent mutations.
In addition to the numerous nonsense and missense mutations that have been found in ER, several silent mutations have also been identified. These mutations do not alter the protein sequence, so they would not be predicted to significantly affect ER function. In fact, the majority of polymorphisms identified do not have an observable phenotype associated with them, regardless of the tissue of origin (263, 271, 287). Although the ER codon 10 polymorphism has been found in numerous normal and diseased tissues, it is not commonly associated with the diseased state (221, 283, 288, 289). However, two other silent mutations, one in codon 87 and one in codon 325, have been reported to be associated with distinct phenotypes. The BstU1 restriction fragment length polymorphism (RFLP) is caused by a silent mutation in codon 87 (275). This mutation was subsequently found to be linked with the A86V missense mutation, and it is associated with spontaneous abortions and reduced ER levels in breast cancer patients with ER-positive tumors, as discussed above. Multiple studies have demonstrated a correlation between a silent mutation at codon 325 and breast cancer risk (271, 290, 291). Additionally, this mutation has been associated with low femoral neck bone mineral density (BMD), but not lumbar spine BMD (292). Unfortunately, all of the correlative studies have involved a small number of patients. Although these silent mutations do not directly affect the protein sequence, they may indirectly affect protein function through alterations in RNA turnover rates, such as RN