Truncated Estrogen Receptor Product-1 Suppresses Estrogen Receptor Transactivation by Dimerization with Estrogen Receptors α and β*

The estrogen receptor (ER) is a ligand-activated transcription factor that acts as a homodimer. Truncated estrogen receptor product-1 (TERP-1) is a pituitary-specific, estrogen-induced, isoform of rat ERα that is transcribed from a unique start site and contains only the C-terminal region of the full-length receptor. TERP-1 does not affect transcription directly but suppresses ligand-activated ERα and ERβ activity. Because TERP-1 contains a dimerization domain and part of the coactivator binding pocket, we hypothesized that it modulates ER function by direct interactions with full-length ER or the steroid receptor coactivator, SRC-1. Localization studies demonstrate that TERP-1 is present in the cytoplasm and nucleus of transfected cells and colocalizes with nuclear ER. Protein binding studies show that TERP-1 forms heterodimers with both ERα and ERβ and inhibits ERα binding to its cognate DNA response element. TERP-1 also binds SRC-1, and increasing levels of SRC-1 decrease the TERP-1-ERα interactions, in agreement with the rescue of TERP-1-suppressed ERα transcriptional activity by SRC-1. Mutational analysis of TERP-1 and ERα in the activation helix and the AF-2 dimerization helix indicates that TERP-1 acts predominantly through dimerization with ERα. Therefore, TERP-1 suppression of ER transcription occurs primarily by formation of inactive heterodimers and secondarily by competition for coactivators.


The estrogen receptor (ER) is a ligand-activated transcription factor that acts as a homodimer. Truncated estrogen receptor product-1 (TERP-1) is a pituitary-specific, estrogen-induced, isoform of rat ER␣ that is transcribed from a unique start site and contains only the C-terminal region of the full-length receptor. TERP-1 does not affect transcription directly but suppresses ligand-activated ER␣ and ER␤ activity. Because TERP-1 contains a dimerization domain and part of the coactivator binding pocket, we hypothesized that it modulates ER function by direct interactions with full-length ER or the steroid receptor coactivator, SRC-1. Localization studies demonstrate that TERP-1 is present in the cytoplasm and nucleus of transfected cells and colocalizes with nuclear ER. Protein binding studies show that TERP-1 forms heterodimers with both ER␣ and ER␤ and inhibits ER␣ binding to its cognate DNA response element. TERP-1 also binds SRC-1, and increasing levels of SRC-1 decrease the TERP-1-ER␣ interactions, in agreement with the rescue of TERP-1-suppressed ER␣ transcriptional activity by SRC-1. Mutational analysis of TERP-1 and ER␣ in the activation helix and the AF-2 dimerization helix indicates that TERP-1 acts predominantly through dimerization with ER␣. Therefore, TERP-1 suppression of ER transcription occurs primarily by formation of inactive heterodimers and secondarily by competition for coactivators.
Estrogen receptors (ERs) 1 are ligand-activated transcription factors that bind to cognate estrogen response elements (EREs) on DNA to influence target gene activity in a variety of responsive tissues including breast, uterus, liver, and pituitary (1). ER isoforms, ER␣ and ER␤, belong to the nuclear receptor superfamily that includes both steroid and nonsteroid nuclear receptors such as thyroid hormone receptors (TR), retinoic acid receptors (RAR), and retinoid X receptors (RXR), as well as an increasing number of orphan receptors with no known ligand (2,3). Nuclear receptor family members share distinct structural and functional domains. The N terminus contains an activation function (AF-1), which is ligand-independent. A DNA-binding domain (DBD) consisting of two zinc fingers is located in the central region of the receptor, and a second activating function (AF-2) is located within the ligand-binding domain (LBD) at the C terminus of the protein. The AF-2 function requires ligand binding for transcriptional activity, and the contribution of AF-1 and AF-2 to receptor activity is both cell type-and promoter-dependent. Nuclear localization signals are located in both the DBD and LBD, and dimerization domains exist in the DBD and in helices 10/11 of the LBD (1)(2)(3)(4).
Nuclear receptor activity can be modulated by several mechanisms including ligand binding, formation of heterodimers with other receptors, and binding of and competition for coregulatory proteins. ER and other steroid receptors are thought to act primarily as homodimers, although ER␣-ER␤ dimers can be observed on DNA and in solution (5)(6)(7). Because the isoforms are expressed in a tissue-and cell-specific manner (8,9) and have differential affinities for synthetic and environmental ligands (10), dimer formation can have important biological consequences. For example, ER isoforms ␣ and ␤ have markedly different responses to estrogen antagonists, and the ratio of the two isoforms could determine cellular responses to these drugs (10 -12).
Cellular specificity of receptor responses may occur in part by selective expression of coregulatory proteins that interact with the receptors in the C-terminal regions and thus influence interaction of the ligand-receptor complex with the transcriptional machinery (13,14). Several coactivator and corepressor proteins that bind to distinct receptor regions have been isolated. Corepressor proteins, such as nuclear receptor corepressor (NCoR; Ref. 15) and silencing mediator of retinoic acid and thyroid hormone receptors (SMRT), bind to the NCoR (15,16) box in the receptor hinge region between the DBD and LBD. Corepressors are recruited to unliganded receptors, such as the TR and RAR, and act as potent transcriptional repressors in the absence of ligand (15,16). NCoR and SMRT do not appear to have a significant role in ligand-activated ER activity but can contribute to the regulation of transcription in antagonistbound ER complexes (17). Coactivator proteins, including steroid receptor coactivator-1 (SRC-1; Ref. 18), the related glucocorticoid receptor activating protein-1 (GRIP-1; Ref. 19), and * This work was supported by National Institutes of Health Grant R01 HD25719/DK57082 (to M. A. S.) and funds from the Lalor Foundation (to D. A. S.). This work was also supported by NICHD, National Institutes of Health Grant U54 HD28934 through a cooperative agreement as part of the Specialized Cooperative Centers Program in Reproductive Research. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
In addition to coactivators and corepressors, orphan receptors and other coregulatory proteins have been found to influence nuclear receptor activity both directly and indirectly. For example, short heterodimer partner (SHP), an orphan receptor isolated from human and rat liver cDNA libraries, can suppress transcriptional activation by RAR and ER (25)(26)(27). SHP suppression of RAR-RXR heterodimer activity occurs through formation of SHP-RAR heterodimers and inhibition of receptor binding to DNA (25). In contrast, SHP inhibition of ER activity may involve heterodimerization and competition for coactivators without influencing DNA binding (26,27). The nuclear protein, receptor interacting protein-140 (28,29), stimulates ER activity directly by binding to the receptor; however, it inhibits activation of the nuclear receptor peroxisome proliferator-activated receptor (PPAR) by SRC-1 and lowers overall PPAR activation by competing with SRC-1 for binding to PPAR (30).
Our laboratory cloned and characterized a novel isoform of ER␣ from female rat pituitary, truncated estrogen receptor product-1 (TERP-1; 31). TERP-1 is transcribed from a unique start site and contains exons 5-8 of ER␣. The resultant protein contains the majority of the LBD but lacks the DBD and AF-1 regions and cannot bind estradiol (E) effectively. TERP-1 mRNA and protein expression is tissue-selective and is stimulated dramatically by estrogen and throughout the estrous cycle (32). This regulation is distinct from that of full-length ER␣ and during proestrus and diestrus TERP-1 is expressed at mRNA and protein levels equal to or in excess to that of fulllength ER␣ (31)(32)(33). Previous work demonstrated that TERP-1 modulates the activity of full-length ER isoforms in a tissueand promoter-specific manner at ratios that are observed physiologically. Specifically, low TERP-1:ER ratios stimulate ER activity, whereas TERP-1:ER ratios equal to or greater than 1:1 suppress ER stimulation of gene transcription (7). In these investigations we have addressed the suppressive actions of TERP-1 and show that TERP-1 can bind directly to both ER␣ and ER␤ and can compete with ER for binding to the coactivator SRC-1. Our data demonstrate that TERP-1 suppresses ER activity through both direct and indirect mechanisms and that direct binding to ER is the predominant pathway.
Cell Culture and Transfections-The ER negative monkey kidney Cos-1 cell line was maintained in Dulbecco's minimal essential medium with 10% newborn calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin. Cells used for transcription assays and immunofluorescence (on coverslips) were plated in 30-mm wells at a density of 200,000/ well. For transcription experiments (performed 3-6 times for each study), phenol red-free Dulbecco's minimal essential medium with 5% charcoal-stripped newborn calf serum was used. 24 h after plating, cells were transfected by the calcium phosphate method (7). Cells were transfected with 1 g of the luciferase reporter plasmid and various expression constructs (shown in figure legends), and total DNA was standardized with pcDNA vector. 16 h later, cells were washed with phosphate-buffered saline (PBS; pH 7.4), incubated for 24 h, and then collected to assess luciferase activity or fixed for immunofluorescence. Where indicated, cells were treated with 10 nM E during the 24-h period post transfection. Luciferase activity was assessed on a Turner-20E luminometer using a Promega luciferase assay kit, and samples were normalized by assessing lysate protein with Bio-Rad protein dye. Data are presented as arbitrary light units and normalized for protein levels.
Immunofluorescence-Cos-1 cells were transfected with CMV vectors expressing TERP-FLAG (4 g) and ER␣ (1 g) and were incubated in the absence or presence of 10 nM E. Cells were fixed with 4% paraformaldehyde in PBS (10 min), permeabilized with 0.2% Triton X-100 (2 min), and blocked with 5% nonfat dry milk (1 h; Carnation) in PBS. TERP-FLAG was detected using the monoclonal antibody anti-FLAG M2 (25 g/ml; VWR) and an ER␣ antibody, ER C1355 (1:7500). ER C1355 was developed by our laboratory and is directed against the last 14 amino acids of rat ER␣ (32). ER␣ was detected using ER C1355 and ER 715 (1:500; gift of Dr. Jack Gorski), which is directed against the hinge region (amino acids 270 -284) of rat ER␣ (36). All antibody incubations were performed in PBS containing 2% bovine serum albumin (Sigma). 1-h incubations at room temperature were performed with anti-FLAG, and overnight incubations at 4°C were performed with ER 715. Anti-mouse IgG-conjugated Texas Red (1:500; Calbiochem) was used to detect TERP-FLAG (anti-FLAG), and anti-rabbit IgG-conjugated FITC (1:400; Calbiochem) was used to detect ER␣ (ER 715). Anti-rabbit IgG-conjugated Texas Red or FITC was used to detect TERP-FLAG or ER␣ (ER C1355) in individual experiments. For single and double label immunofluorescence, a Nikon Microphot-SA microscope was used. The Texas Red and FITC filters were from Chroma Technologies. Cellular sectioning was performed with an IX70 microscope (Olympus America). To acquire the image, Isee software from Inovision Corporation was used, and Deltavision from Applied Precision, Inc. was used for processing the data.
GST Pull-downs-BL21 bacterial cells were transformed with constructs expressing GST, GST-TERP, GST-TERP-E547K, GST-TERP-L509R, or GST-ER␣-LBD. Luria Broth (100 ml) containing 50 g/ml ampicillin was inoculated with 1 ml of bacteria and incubated in an orbital shaker at 37°C. Bacteria were grown to A 600 ϭ 0.5, induced with 0.1 mM isopropyl ␤-thiogalactopyranoside and shaken overnight at room temperature. The bacterial pellet was resuspended in 5 ml of buffer containing 50 mM Tris, pH 7.5, 0.5 mM EDTA, 300 mM NaCl, 10 mg/ml lysozyme, and 1 mM dithiothreitol. 100 l of 10% Nonidet P-40 was added, and after 10 min, the lysate was frozen at Ϫ70°C in an ethanol bath. Lysate was thawed at room temperature and then incubated for 1 h in 5 ml of buffer containing 1.5 M NaCl, 12 mM MgCl 2 , DNase I (5 g), 10 g/ml leupeptin, 1 g/ml pepstatin A, and 0.1 mM phenylmethylsulfonylfluoride. Lysates were passed through a 20-gauge needle and centrifuged for 30 min at 8000 ϫ g. Soluble lysate was conjugated with glutathione beads (Sigma) overnight at 4°C. Beads were washed with PBS, and protein concentrations were assessed after electrophoresis on 12% polyacrylamide denaturing gels by Coomassie Blue stain and immunoblotting (GST-TERP, GST-TERP-E547K, GST-TERP-L509R, and GST-ER␣-LBD) with ER C1355. For pull-down experiments, approximately 1 g of GST alone or 1 g of GST fusion protein was used in each sample incubation. Bovine serum albumin (20 g/ml) was added to each incubation containing [ 35 S]methionine-labeled (0.04 mCi/50-l reaction) in vitro translated proteins (T N T Rabbit Reticulocyte Transcription/Translation Kit; Promega) or approximately 100 g of transfected Cos-1 whole cell extracts. Cos-1 cells were transfected with 1.6 g of ER␣ in 60-mm dishes with 10% newborn calf serum for whole cell extract collection (37). Where indicated, 10 nM estradiol (E), 1 M 4-hydroxyamoxifen (OHT), 10 M ICI 182,780 (ICI), or ethanol vehicle was added to the incubations. Total volume was adjusted to 150 l with GST wash buffer (10 mM MgCl 2 , 150 mM KCl, 20 mM HEPES, 10% glycerol, and 0.12% Nonidet P-40). Beads and proteins were incubated for 1.5 h at 4°C and then centrifuged and washed four times in GST wash buffer. Beads were resuspended in 10 l of SDS loading buffer and boiled for 5 min. Proteins were electrophoresed on SDS-containing 12% acrylamide gels at 150 V, along with standard protein molecular weight markers (Benchmark, Life Technologies, Inc.). Gels containing [ 35 S]methionine-labeled proteins were dried and exposed to film overnight at room temperature. Whole cell extract proteins were transferred to nitrocellulose, and Western blotting with ER C1355 was performed as described previously (7). Where noted, band intensity was measured by densitometric scanning of the resultant autoradiograph as previously reported (32).
Electrophoretic Mobility Shift Assays-ER binding to DNA was determined by electrophoretic mobility shift assays as described previously (35). Briefly, double-stranded DNA containing one copy of the vitellogenin A2 ERE (5Ј-TAACTGTCCAAAGTCAGGTCACAGTGAC-CTGATCAAAGTTAATGTG-3Ј) was end-labeled with [␣-32 P]dCTP and the Klenow fragment of DNA polymerase II. In vitro translated proteins (made in the presence of 10 nM E) were incubated for 20 min at room temperature with labeled ERE (ϳ100,000 cpm) in buffer containing 10 mM Tris, pH 7.6, 1 mM EGTA, 1 mM dithiothreitol, 0.1 M KCl, 2 g of poly(dI⅐dC)⅐poly(dI⅐dC), and 50 ng of salmon sperm DNA. For cotranslation studies, 500 ng of ER␣ was translated with either 500 ng or 2 g of TERP-1, and the amount of ER␣ protein was normalized by immunoblots. ER␣ protein was supershifted by the addition of 1 l of a 1:10 dilution of ER C1355 antibody to the protein-DNA incubations. Incubation time and conditions were identical for reactions incubated with or without antibody. The protein-DNA complexes were subjected to electrophoresis on 5% polyacrylamide, nondenaturing gels (25 mM Tris, 190 mM glycine, 1 mM EDTA, pH 7.9) at 150 V at 4°C. Gels were dried and exposed to film overnight at Ϫ70°C.

TERP-1 Can Colocalize with ER␣ in the Nucleus-
The mechanism by which TERP-1 suppresses ER transactivation could be direct or indirect and will be governed by its intracellular location. TERP-1 contains one of the two nuclear localization signals present in ER␣ and could reside in either the cytoplasm or nucleus. Therefore, we used double label immunofluorescence to localize TERP-1 in transfected Cos-1 cells in the presence or absence of ER␣ and ligand. Because TERP-1 contains the identical amino acid sequence for most of the LBD of ER␣ (amino acids 393-600), it was necessary to differentiate the two proteins by an epitope tag (Fig. 1). We constructed the TERP-FLAG expression vector in which the 8-amino acid FLAG sequence is fused to the C terminus of TERP-1. TERP-FLAG and TERP-1 have similar effects on ER␣ transactivation as measured by transient transfection assays, and immunofluorescence of TERP-1 or TERP-FLAG using the C-terminalspecific C1355 antibody demonstrated similar localization patterns (not shown). In a series of parallel experiments, the same pattern for TERP-FLAG was observed with either the C1355 or FLAG antibodies, and exclusively nuclear staining for ER␣ was observed with the C1355 or ER715 antibodies with or without E treatment (data not shown). Fig. 2 depicts cells cotransfected with ER␣ and TERP-FLAG and treated with E. Transfection conditions were identical to those in which TERP-1 suppresses ER␣ activity (7). In these cells, TERP-FLAG (red color) appeared in both the cytoplasm and nucleus (Fig. 2, A, B, C, and E), and ER␣ (green color) was found only in the nucleus (Fig. 2D). Panels A and B show the same contrasfected cell, with colocalization of TERP-FLAG and ER␣ in the nucleus as indicated by the yellow color in Fig. 2B. The subcellular location of TERP-FLAG was not altered by E treatment or the presence of the full-length ER␣, and TERP-FLAG did not alter ER␣ localization in the nucleus. A second cotransfected cell (Fig. 2, C-E) shows individual fluorescence of red TERP-FLAG in both cytoplasm and nucleus (Fig. 2C) and green nuclear ER␣ (Fig. 2D). Identical subcellular relationships were observed for the two proteins in five independent experiments. Nuclear localization of TERP-FLAG in this cell was further verified by cell sectioning and computer imaging. The mosaic (Fig. 2E) shows every third cellular section (0.4-m sections) and demonstrates consistent staining for TERP-FLAG throughout the cell and distinct dark nucleoli. Thus, TERP-1 does not inhibit ER activation by sequestering ER protein in the cytoplasm, and the colocalization of both TERP-FLAG and ER␣ in the nucleus suggests that direct interactions between the two proteins could occur.
GST-TERP Binds ER␣ and ER␤-Because TERP-1 protein can colocalize with ER␣ and can modify the transcriptional activity of both ER␣ and ER␤ (7), we tested the ability of TERP-1 to form dimers with both full-length receptors. A GST-TERP fusion protein was constructed and used in pull-down experiments with transfected Cos-1 whole cell extracts and in vitro translated [ 35 S]methionine-labeled ER proteins. GST-TERP specifically bound ER regardless of the protein source. For example, both ER␣ from transfected Cos-1 cells and cellfree translated ER␣ strongly bound GST-TERP but did not effectively interact with GST alone (Fig. 3). TERP-1 cannot bind E, but its biological effects require E and an activated ER and are inhibited by the antiestrogens OHT and ICI (7). Therefore, we examined the interactions of TERP-1 and both ER isoforms in the presence and absence of E, OHT, and ICI (Fig.  3). An equal amount of purified GST alone or GST fusion protein was incubated with equivalent amounts of in vitro translated [ 35 S]methionine-labeled ER protein. Formation of TERP-1-ER␣ heterodimers was not ligand-dependent but was ligand-enhanced. This was seen most dramatically in the pres- ence of ICI, where binding was increased approximately 8.5fold as measured by densitometry, and this trend was seen in each of three experiments. In contrast, TERP-1 binding to ER␤ was completely ligand-independent. These data show that TERP-1 interaction with either ER isoform does not require ligand, in agreement with the ability of ER isoforms to form dimers in the absence of ligand, and in contrast to the liganddependent interactions of receptors and coregulatory proteins (18,38). TERP-1 predominantly formed heterodimeric complexes, although weak homodimers were seen (not shown). The ability of TERP-1 to form heterodimers with the full-length receptors suggests that TERP-1 could suppress ER activity by direct binding to the receptor through the dimerization domains in helix 10/11 and the formation of inactive complexes.
TERP-1 Inhibition of ER␣ Activity through the AF-1 and AF-2 Domains-TERP-1 modulation of AF-1 and AF-2 transactivation was tested in cotransfection studies and compared with effects on the full-length ER␣ (Fig. 4). All three constructs (proteins shown in Fig. 1) stimulated promoter activity of the luciferase reporter gene. Both ER␣-and ER␣-AF-2-stimulated activities of the ERE-containing construct were increased by E, and both activities were inhibited by cotransfected TERP-1 by an average of 70 -80%. Interestingly, high concentrations of TERP-1 also inhibited AF-1 activity in the same experiments by approximately 40%, suggesting that some less effective transcriptional complex was formed. Both AF-1 and AF-2 have less activity than the full-length ER␣ in the same experiments, indicating that interaction between the two activating functions of the receptor is required for full biological activity. The data indicate that TERP suppresses both AF-1 and AF-2 activity. The inhibition of ER␣ activity was similar to suppression of ER␣-AF2, in agreement with previous data showing that the suppressive effects of TERP-1 require a ligand-bound ER and that the antiestrogen OHT, which suppresses AF-2 but maintains AF-1 activity (14), completely abolishes the effects of TERP-1 (7).
TERP-1 Binds to the Coactivator SRC-1 and Competes for Binding with ER␣-Cotransfection of increasing amounts of SRC-1 can partially overcome the inhibitory effects of TERP-1 in Cos-1 cells (Fig. 5A). Previous work demonstrated that this rescue is dose-dependent (7). Because TERP-1 includes regions of ER␣ that are important for coactivator binding, particularly helix 12 (amino acids 544 -552 in rat ER␣), we examined the ability of TERP-1 to interact directly with SRC-1 in pull-down assays with in vitro translated [ 35 S]methionine-labeled proteins (Fig. 5B). In addition to binding ER␣, GST-TERP also bound SRC-1. The interaction between TERP-1 and SRC-1 was specific, because the corepressor proteins SMRT and NCoR did not bind to GST-TERP (not shown). Suppression of ER␣ activ-ity could occur by competition of TERP-1 and ER␣ for available coactivators. To test this possibility, we performed competition experiments with GST-TERP in the presence of a constant amount of ER␣ and increasing SRC-1 (Fig. 5B). Increasing amounts of SRC-1 displaced ER␣ binding to TERP-1 up to 64%. To assess whether this competition could be biologically relevant, we compared the ability of the ER␣-LBD and TERP-1 to individually bind SRC-1 in the same experiment. As shown in Fig. 5C, both molecules can bind SRC-1. ER␣-LBD binding to SRC-1 was greatly enhanced by the presence of E as shown previously (18), whereas SRC-1 binding to TERP-1 was not. TERP-1 bound SRC-1 slightly better than unliganded ER␣-LBD but less well than liganded receptor. Thus, TERP-1 could compete with ER for binding to coactivators but only at times when TERP-1 protein levels are high, which can occur physiologically (32). This would limit the ability of coactivators to bind activated ER␣, thus limiting the ability of the ligand-activated receptor to stimulate transcription.

Point Mutations in TERP-1 Demonstrate That Suppression of ER␣ and ER␤ Occurs Predominantly through Dimerization
with the ER-Because TERP-1 influences the AF-2 function in the C-terminal region of ER␣, we examined two potential mechanisms of the actions of TERP-1. First, because TERP-1 contains the dimerization domain in helices 10/11 (amino acids 479 -522), it could interact with the ER directly. The second potential mechanism involves the interactions of coactivators, like SRC-1 with ER or TERP-1. TERP-1 contains the coactivator binding interface and could sequester coactivator proteins away from activated ERs. Therefore, two single point mutations were made in TERP-1 that would individually disrupt the dimerization domain (TERP-1-L509R; Ref. 34) and the helix 12 coactivator binding pocket (TERP-1-E547K; Ref. 22). The same individual mutations were made in the full-length ER␣ to ensure that the receptors were functionally compromised in luciferase reporter gene assays. Wild type ER␣ activity was effectively stimulated by E; however, both ER␣-E547K and ER␣-L509R had compromised E-induced activity (Fig. 6A). GST pull-down experiments showed that ER␣ interactions with GST-TERP-L509R (Fig. 6B) are greatly reduced relative to wild type GST-TERP, whereas those with GST-TERP-E547K are not. TERP-1 mutated in helix 12 inhibited ER␣ activity as well as wild type TERP-1 and partially inhibited ER␤ activity (Fig. 6C). However, the dimerization mutation in TERP-1 severely compromised its ability to inhibit both ER␣ and ER␤ activity (Fig. 6C) and even slightly increased ER␤ activity. The failure of the L509R mutant to inhibit ER was not due to protein instability, because immunoblots of transfected cells demonstrated levels of TERP-1-L509R equal to that of wild type TERP-1 (not shown). Transfection of a TERP-1 construct containing both mutations (TERP-1-RK) had no suppressive effect on ER actions, indicating that these two pathways likely explain all suppressive effects of TERP-1. Overall these data suggest that TERP-1 and ER␣ interact directly through dimerization domains, and this is the primary pathway of TERP-1 inhibition. In comparison, interaction of TERP-1 with coactivator proteins is a secondary mechanism for TERP-1 actions.
TERP-1 Reduces the Ability of ER␣ to Bind to an ERE-One possible mechanism by which direct dimerization of TERP-1 to ER␣ could suppress ER␣ activity is by inhibition of ER␣ binding to its defined DNA element (ERE). To test this hypothesis, electrophoretic mobility shift assays were performed with a labeled ERE and in vitro translated proteins. The amount of ER␣ in each lane was standardized by immunoblotting with ER C1355, and unprogrammed reticulocyte lysate was added where necessary to maintain a constant total lysate volume. As expected, TERP-1 does not bind to the ERE, because it does not contain a DBD. In contrast, ER␣ forms specific high affinity complexes as shown by the shaded arrowhead in Fig. 7, and this complex is shifted upon addition of C1355 antibody as indicated by the open arrowhead in Fig. 7. No smaller DNAprotein complexes potentially representing ER␣/TERP-1 dimers on DNA were observed. ER␣ cotranslated with TERP-1 bound DNA less efficiently than ER␣ alone. Thus, TERP-1 interacts with ER␣ through dimerization domains to form ER dimers with reduced DNA binding ability.

DISCUSSION
The stimulation of transcription by activated estrogen receptor has both general and cell-specific aspects that contribute to biological responses (8 -12). This has particular importance therapeutically with the use of selective estrogen receptor modulators that have both tissue-and cell-specific effects to treat osteoporosis, memory loss, and heart disease and as treatments or possible preventives for steroid-dependent cancers of the breast and uterus (39). Among the major factors contributing to tissue-selective responses are the complement of receptor isoforms present in a given cell type and the number and types of proteins capable of binding to the receptors and modulating their function. Several groups have isolated activating proteins termed coactivators and inhibitory proteins called corepressors that bind directly to specific sites on nuclear receptors and alter their transcriptional capability (13,14). Others have defined molecules termed orphan receptors that have some structural similarity to the nuclear receptors and modify their activity by direct interactions with the nuclear receptor ligand-binding domain (25-27, 40, 41). We have identified a novel form of ER␣, TERP-1, which acts as a tissue-specific, highly regulated suppressor of ER action (7,31). Here we show that it exerts its activity primarily by binding to both full-length receptor isoforms, ER␣ and ER␤, and secondarily by acting as an intracellular buffer to prevent binding of coactivators to the full-length ER forms.

TERP-1 Suppresses ER Activity Primarily by Direct Protein-Protein
Interactions-Transient transfection studies with constructs representing either the N-terminal AF-1 or ligand-binding region AF-2 domains of ER␣ indicate that TERP-1 suppression of ER␣ transactivation is equal to suppression of the C-terminal AF-2 function, although suppression of AF-1 also is observed. Direct physical and functional interactions between the N-and C-terminal regions of the estrogen, progesterone, and androgen receptors and interactions between the N-terminal regions of the steroid receptors and SRC-1 have been reported (42)(43)(44)(45)(46), perhaps explaining some inhibition of AF-1 activity by TERP-1 in transfection assays. TERP-1 does not inhibit ER␣ action by sequestering the full-length receptor in the cytoplasm (Fig. 2) or by inhibiting E binding by ER␣, as we previously demonstrated (7). Instead, TERP-1 acts primarily by binding to ERs through the dimerization helix and suppressing the transcriptional activity of ligand-activated receptor.
Although a direct interaction between TERP-1 and ER␣ was not observed in stringent immunoprecipitation studies, this

FIG. 6. A mutation in the dimerization domain diminishes TERP-1 suppressive effects on ER and inhibits TERP-1 interactions with ER␣. A,
Cos-1 cells were transfected with 1 g of each ER expression vector, treated, and assayed for luciferase activity as in Fig. 4. Data shown are the means Ϯ S.E. from three experiments containing duplicate wells. *, p ϭ Ͻ0.05 versus ligand-activated ER␣ wild type control. B, 1 g of GST, GST-TERP, GST-TERP-L509R, or GST-TERP-E547K was used in each incubation containing 6 l of labeled in vitro translated ER␣ with 10 nM E. The input lane (I) contains 1 l of ER␣ protein, with the migration position indicated by the arrowhead. Proteins were detected by autoradiography, and values beneath the lanes indicate relative densitometric volumes of ER␣ bands. C, Cos-1 cells were transfected with 1 g of ER␣ or ER␤ and 4 g of each TERP-1 expression vector and were treated and assayed for luciferase reporter gene activity as in Fig. 4. Data presented are the means Ϯ S.D., are the average of two (ER␤) or three (ER␣) experiments with duplicate wells, and are representative of three (ER␤) to six (ER␣) independent experiments. *, p Ͻ 0.05 versus ligand-activated ER␣ or ER␤ control in both panels. approach may not always detect important physiological interactions. It is now clear that the dimerization function is critical for TERP-1 suppression of ER activity because the L509R mutation, which has previously been shown to inhibit ER␣ homodimerization and compromise its function ( Fig. 6 and Ref. 34), eliminates the suppressive effects of TERP-1 on ER transactivation. This mutation also dramatically decreases the interactions between TERP-1 and ER␣ (Fig. 6). The TERP-1-ER␣ heterodimer complex must therefore be less effective at stimulating transcription from an ERE, potentially by decreasing binding to DNA as demonstrated by gel shift analysis. Alternatively, a heterodimeric complex bound to DNA could be less effective at associating with coactivator and integrator proteins and the transcription initiation complex. However, there is no evidence for smaller complexes in DNA binding assays representing heterodimer binding to DNA, making this possibility less likely. ER␣ and ER␤ can also form heterodimers through this region in helix 10/11, and the heterodimers bind to DNA and influence gene transcription (5)(6)(7). The dimerization mechanism is also likely important for TERP-1 effects on ER␤, as demonstrated by the transfection studies in Fig. 6. Because TERP-1-ER interactions, particularly with ER␤, occur in the absence of ligand, TERP-1 could suppress activity of ER-regulated genes modulated through ligand-independent pathways. However, levels of TERP-1 are tightly regulated by E, and significant mRNA levels are observed only after E treatment and at times in the reproductive cycle when E is high (32). Thus, TERP-1 will serve as an E-induced regulator of ER function in vivo.
A secondary mechanism for TERP-1 suppression of ER activity is competition for binding to coactivator proteins. Nuclear receptors associate with several different activator proteins, including SRC-1 and GRIP-1, through the C-terminal portion of the receptor, although the precise areas of interac-tion within ER␣ vary slightly with each coactivator (29). Cotransfection of SRC-1 partially overcomes the inhibitory effects of TERP-1 on ER␣ transactivation ( Fig. 5A and Ref. 7). TERP-1 contains a portion of the coactivator binding interface, helix 12 of the LBD, but lacks the most N-terminal helices of the LBD that are important for coactivator binding, helices 3 and 5. Mutations in each of these helices inhibit binding of SRC-1 and GRIP-1 to TR␤ and GRIP-1 to human ER␣ (22). Our data indicate that TERP-1 interacts with SRC-1, as well as with ER, and could thus compete with ER for SRC-1 binding. In agreement with this possibility, Jeyakumar et al. (47) demonstrated that a small peptide consisting of amino acids 437-456 of TR␤, equivalent to helix 12, inhibited SRC-1 binding to the entire receptor. Other investigators have also found that RIP140, a coregulatory protein binding at a somewhat different region on PPAR than SRC-1, antagonizes receptor activation by SRC-1 by effectively competing for binding to the receptor (30). The ER␣-LBD binds SRC-1 more effectively than does TERP-1 in GST-pull-down experiments. This suggests that competition of TERP-1 and ER for SRC-1 is likely to occur only when SRC-1 levels are low or when TERP-1 levels are high, as has been noted during proestrus (32).
A related mechanism by which the TERP-1 helix 12 region could inhibit ER action is by direct occupation of the coactivator pocket, as has been observed for the homologous helix 12 of the ER␣ when bound to the selective estrogen receptor modulators tamoxifen or raloxifene (21,48). The helix 12 mutant, TERP-1-E547K, is as effective at suppressing ER␣ transactivation as the wild type TERP-1 in these studies but only partially inhibits ER␤ transactivation, demonstrating that suppression through helix 12 of TERP-1 is more important for ER␤ than ER␣. This mechanism could be important in some cellular contexts, when ER␤ is high (9, 10), or when coactivator proteins themselves are modulated. For example, SRC-1 levels are altered in clonal pituitary cells after treatment with E or thyroid hormone, and this could alter responsiveness of nuclear receptors in those cells (49). The ability of SRC-1 to rescue TERP-1 actions demonstrates a functional role for this pathway.
TERP-1 Is Functionally Similar to Nuclear Receptor Coregulatory Proteins SHP and DAX-1-Nuclear receptor function can also be modulated by the formation of specific heterodimers with orphan receptors that have structural similarities with the nuclear receptors but do not bind known ligands. Association of RARs, TRs, and vitamin D receptors with the previously characterized RXRs results in heterodimers with increased binding affinity to DNA cognate response elements and enhanced transcriptional responses to retinoic acid, thyroid hormones, and vitamin D (50). RXRs themselves bind to DNA and associate with nuclear receptors via dimerization domains similar to those found in the nuclear receptor LBDs. The ER isoforms do not interact with RXR family members, and ER function is not affected by RXRs.
More recently, a new class of suppressive orphan receptor molecules has been defined. These proteins, including SHP and DAX-1, contain putative LBDs but lack DNA-binding regions. DAX-1 combines preferentially with the orphan receptor steroidogenic factor-1 with which it is coexpressed in the adrenals, gonads, hypothalamus, and pituitary gland (40,41,51,52), whereas SHP is expressed in a variety of tissues and suppresses the activities of RAR, TR, and ER (25)(26)(27). SHP and DAX-1 are proposed to exert their effects via several mechanisms. These include dimerization with target nuclear receptors to form nonproductive transcriptional complexes, inhibition of heterodimer binding to DNA, and competition for coactivator proteins.
Like these orphan receptors, TERP-1 does not bind DNA or ligand and has no transcriptional effects on its own. All three proteins exert their suppressive effects by binding directly to nuclear receptor partners, but the physical nature of the interaction is different. Although DAX-1 and SHP both contain putative dimerization helices, they do not associate with target receptors through those C-terminal regions. DAX-1 association with steroidogenic factor-1 occurs through an N-terminal region that does not contain a dimerization helix (51). Similarly, interaction of SHP with target receptors does not occur through the SHP C-terminal putative dimerization helix 10/11 but instead requires the central interaction domain between amino acids 92 and 148 (25)(26)(27)53). This region does not contain functional NR box motifs (LXXLL) required for other cofactor interactions with nuclear receptors but may require a similar region in activation helix 12 within the AF-2 region of the nuclear receptors (27). Mutation of this region in ER␣ does not disrupt the interaction with TERP-1 (Fig. 6). SHP interactions with nuclear receptors are generally ligand-dependent, with the exception of its association with ER␤. In comparison, TERP-1 binds to both ERs in the absence of ligand, as has been noted for dimerization of the ER isoform homodimers and heterodimers (38), and occurs through the dimerization helix 10/11 regions of both partners. The protein-protein interactions occur even in the presence of antiestrogens, reinforcing the importance of the dimerization helix in this process and in biological function.
Formation of heterodimers between nuclear receptors and orphan receptors suppresses transcription through at least three defined mechanisms, including direct transcriptional repression, inhibition of DNA binding, and changes in coactivator or corepressor binding to nuclear receptors. Both DAX-1 and SHP have separate, transferable transcription repressor functions separated structurally from receptor interaction domains, as measured in the yeast two-hybrid system. TERP-1 is unlikely to have a separate, transferable suppressive function, and there is no evidence for suppressive activity with other receptor partners such as TR (7). Neither DAX-1 actions on steroidogenic factor-1 nor SHP inhibition of ER occur by interference with DNA binding (26,27,40,51,52). However, SHP suppression of TR, RAR, and RXR activity is postulated to occur through the demonstrated direct inhibition of receptor heterodimer binding to DNA (25), as we have demonstrated for TERP-1 and ER.
SHP, DAX-1, and TERP-1 have all been proposed to interfere with the AF-2 function of nuclear receptors, partly by competition for binding of coactivators. In support of this second mechanism, SHP inhibition of ER␣ and ER␤ activity can be rescued by cotransfection with high levels of the coactivator transcriptional intermediary factor-2 (27). TERP-1 also suppresses AF-2 activity of ER by competition for coactivator binding, and these effects can be rescued by addition of SRC-1 (7). DAX-1 may also recruit the binding of corepressors to steroidogenic factor-1 (52), but this activity has not been tested for SHP inhibition of TR and RAR. The previously described corepressors NCoR and SMRT appear to have a minor role in ER activation by natural ligand, and because TERP-1 does not bind N-COR or SMRT, 2 this pathway is unlikely to play a major role in TERP-1 suppression. Recently, however, an ER-selective coregulator (repressor of estrogen receptor activity, or REA) has been identified (54). REA interacts with the ligandbinding domain of ER in a ligand-independent fashion and suppresses ER activity. REA or similar molecules could interact with TERP-1, and titration of such repressors by TERP-1 could conceivably explain the stimulatory effects of TERP-1 noted at low TERP-1:ER ratios (7). The relative contribution of the dimerization and coactivator binding pathways to SHP and DAX-1 activities have not been directly evaluated. Our mutation studies indicate that the dimerization pathway is the most important one for TERP-1 suppression of ER activity.
TERP-1 Is a Novel Tissue-specific Suppressor of ER Action--Two novel features of TERP-1 distinguish it from SHP and DAX-1. TERP-1 protein represents a portion of a functional receptor, ER␣, and has strict tissue-specific expression of the mRNA and protein. Other mRNA variants of both ER␣ and ER␤ have been described, and some of these have altered biological functions when tested in transfection assays (55)(56)(57). However, these variants all occur through exon splicing of ER mRNA, comprise only a small percentage of total ER mRNA, and do not appear to be regulated in response to physiological state. Not all of these variants have been verified at the protein level, and they affect the biological activity of full-length ER only at ratios far in excess to those demonstrated in cells. In contrast, TERP-1 is transcribed from a transcriptional start site distinct from that of full-length ER␣ mRNA (31). TERP-1 mRNA and protein are expressed only in the pituitary, and the physiological levels observed can be equal to or greater than that for full-length ER␣ (32). Detectable TERP-1 mRNA has been found only in pituitary cells that express ER. Many of these cells express ER␣ and ER␤, and TERP-1 can suppress the activity of both of these ER forms. It is not known whether expression of full-length ER is a requirement for TERP-1 expression. TERP-1 mRNA levels are dramatically stimulated by E and increase 50-fold on the day of proestrus to become the predominant ER isoform (32). This increase in TERP-1 coincides with the pituitary hormone surge and ovulation and correlates with the physiological switch from positive to negative E feedback in the hypothalamic-pituitary axis at this time (58). SHP, which can also suppress ER function, is expressed in many estrogen-responsive tissues but, significantly, not in the pituitary (27). This regulatory function in the pituitary can be provided by TERP-1, which acts a novel, highly regulated, tissue-specific suppressor of ER action.