Estrogen-related Receptor (cid:1) 1 Actively Antagonizes Estrogen Receptor-regulated Transcription in MCF-7 Mammary Cells*

The estrogen-related receptor (cid:1) (ERR (cid:1) ) is an orphan member of the nuclear receptor superfamily. We show that the major isoform of the human ERR (cid:1) gene, ERR (cid:1) 1, can sequence-specifically bind a consensus palindromic estrogen response element (ERE) and directly compete with estrogen receptor (cid:1) (ER (cid:1) ) for binding. ERR (cid:1) 1 activates or represses ERE-regulated transcription in a cell type-dependent manner, repressing in ER-positive MCF-7 cells while activating in ER-negative HeLa cells. Thus, ERR (cid:1) 1 can function both as a modulator of estrogen responsiveness and as an estrogen-independent activator. Repression likely occurs in the absence of exogenous ligand since charcoal treatment of the serum had no effect on silencing activity. Mutational analysis re-vealed that repression is not simply the result of competition between ER (cid:1) and ERR (cid:1) 1 for binding to the DNA. Rather, it also requires the presence of sequences within the carboxyl-terminal E/F domain of ERR (cid:1) 1. Thus, ERR (cid:1) 1 can function as either an active repressor or a constitutive activator of ERE-dependent transcription. We hypothesize that ERR (cid:1) 1 can play a critical role in the etiology of some breast cancers, thereby providing a novel therapeutic target in their treatment.

The estrogen-related receptor ␣ (ERR␣) is an orphan member of the nuclear receptor superfamily. We show that the major isoform of the human ERR␣ gene, ERR␣1, can sequence-specifically bind a consensus palindromic estrogen response element (ERE) and directly compete with estrogen receptor ␣ (ER␣) for binding. ERR␣1 activates or represses ERE-regulated transcription in a cell type-dependent manner, repressing in ER-positive MCF-7 cells while activating in ER-negative HeLa cells. Thus, ERR␣1 can function both as a modulator of estrogen responsiveness and as an estrogen-independent activator. Repression likely occurs in the absence of exogenous ligand since charcoal treatment of the serum had no effect on silencing activity. Mutational analysis revealed that repression is not simply the result of competition between ER␣ and ERR␣1 for binding to the DNA. Rather, it also requires the presence of sequences within the carboxyl-terminal E/F domain of ERR␣1. Thus, ERR␣1 can function as either an active repressor or a constitutive activator of ERE-dependent transcription. We hypothesize that ERR␣1 can play a critical role in the etiology of some breast cancers, thereby providing a novel therapeutic target in their treatment.
The nuclear receptor (NR) 1 superfamily is comprised of hundreds of transcription factors that regulate a vast array of genes and physiological responses (1)(2)(3)(4)(5)(6)(7)(8)(9). Most nuclear receptors share a similar structural organization (Fig. 1A). The amino-terminal A/B domain can function as a hormone-independent activator of transcription. The highly conserved C domain contains the DNA binding domain (DBD) that confers sequence-specific DNA binding activity. A hinge region, called the D domain, bridges the C domain with the carboxyl-terminal E/F domain that includes the receptor-specific ligand binding domain (LBD) of the protein. The binding of appropriate ligands results in conformation changes leading to alterations in the transcriptional properties of the receptor, including the exposure of a transcriptional activation region within the carboxyl end. Although many nuclear receptor superfamily members bind known ligands (e.g., steroids, retinoids, thyroid hor-mones), some, termed orphan receptors, share significant sequence similarity in their LBDs with their ligand binding family members but lack as-yet known naturally occurring ligands (7)(8)(9)(10).
Among the first orphan receptors identified were the estrogen-related receptors ERR␣ and ERR␤ (officially named NR3B1 and NR3B2, respectively) (10). They were cloned by low stringency screening of cDNA libraries with probes corresponding to the DBD of estrogen receptor ␣ (ER␣) (10). Subsequently, ERR␣1 was identified as the major isoform present in HeLa cells ( Fig. 1A) (11,12). The DBD of human ERR␣1 shares 70% amino acid similarity with the DBD of human ER␣; the LBD shares 35% amino acid identity. A third member of the ERR family, ERR␥ (NR3B3), has also been identified (13)(14)(15). These three ERRs are closely related by sequence similarity but encoded by different genes.
Despite sequence similarity with ERs in the LBD, the ERRs do not bind 17␤-estradiol (11,15,16), and the identification of naturally occurring ligands for ERR family members has remained elusive. Vanacker et al. (17) reported that a serum component removable by treatment with charcoal regulates ERR␣-dependent transcription. However, remaining unclear is whether this factor(s) acts directly by binding to ERR␣ or indirectly through a signal transduction pathway. Yang and Chen (18) found that the pesticides toxaphene and chlordane decrease the activity of ERR␣, whereas others reported that ERRs can constitutively interact with co-activators independent of any ligand (19 -23). Interestingly, the synthetic estrogen diethylstilbestrol has been shown to antagonize the activation function of ERR family members by disrupting ERR interactions with coactivators (22, 23); 4-hydroxytamoxifen acts likewise, but only with ERR␥, not ERR␣ (24). Thus, ligands appear to affect the activities of ERRs, but via non-classical mechanisms.
The ERRs also differ somewhat from the ERs in their binding site specificities. They recognize estrogen response elements (EREs) (11,(25)(26)(27)(28)(29); however, ERR␣1 binds with even higher affinity to the consensus steroidogenic factor-1 response element-extended half-site sequence 5Ј-TCAAGGTCA-3Ј (11,27,26). Interestingly, the sequence 5Ј-TAAAGGTCA-3Ј is also recognized by ERR␣ but not by steroidogenic factor-1 (17). Therefore, some genes likely contain estrogen-related receptor response elements regulated only by ERR family members. Thus, ERR family members likely signal via cross-talk with other nuclear receptors through common binding sites as well as ERR-specific genes via ERR-binding sites.
The ERR family members have been shown to function in numerous cell types as transcriptional activators of promoters containing EREs, steroidogenic factor-1 response elements, and ER response elements (16 -31). Nevertheless, we found that ERR␣1 repressed rather than activated transcription in ER-negative CV-1 cells when it binds sites within the late promoter of SV40 (11,32). Thus, ERRs likely modulate gene expression via several mechanisms.
To better understand the multiple activities of ERR␣1, we investigated here the effects of ERR␣1 on expression from an ERE-regulated promoter in ER-positive versus ER-negative cells. We show that ERR␣1 can function as either a repressor or activator of ERE-mediated transcription in a cell type-dependent manner. The mechanism of repression involves interactions with cellular corepressor(s) as well as binding to the ERE. We propose ERR␣1 likely plays roles in the etiology of some breast cancers and the progression from an ER-dependent to ERindependent state.

MATERIALS AND METHODS
Plasmids-All plasmid DNAs were constructed by standard recombinant DNA techniques. Plasmid p3xERE-TK-luc, a gift from V. C. Jordan, contains three tandem copies of the palindromic ERE sequence 5Ј-TAAGCTTAGGTCACAGTGACCTAAGCTTA-3Ј, placed upstream of a minimal herpes simplex thymidine kinase (TK) promoter (nucleotides -109 to ϩ52 relative to the transcriptional start site of the TK promoter), directing expression of the luciferase coding sequence (33). The ERE-negative control plasmid, pTK-luc, was generated from p3xERE-TK-luc by cleavage at the two HindIII sites directly surrounding the three EREs and ligation.
Plasmid pcDNA3.1-hERR␣1 encodes wild-type human ERR␣1 expressed from the cytomegalovirus promoter. It was constructed by reverse transcription-PCR amplification of hERR␣1 mRNA isolated from normal human mammary gland RNA (CLONTECH, Palo Alto, CA) followed by PCR-based cloning of the coding region into pcDNA3.1/V5-His (Invitrogen). To ensure efficient initiation and termination of translation of the ERR␣ open reading frame, the cloning was performed using the primers 5Ј-gacttcGCCACCATGAGCAGCCAGGT-GGTGGTGCATTGA-3Ј (lowercase letters indicate an EcoRI site, underlined letters indicate the translation initiation codon, and bold letters indicate bases altered to optimize translation initiation while maintaining coding of the wild-type ERR␣1 protein) and 5Ј-ggatccT-CAGTCCATCATGGCCTCGAGCAt-3Ј (lowercase letters indicate a BamHI site, and underlined letters indicate the translation termination codon). DNA sequence analysis confirmed that the protein encoded by this plasmid corresponds to the wild-type ERR␣1 referenced in Gen-Bank TM entry NM-004451.
Plasmid pcDNA3.1-hERR␣1 P-box contains three amino acid substitution mutations (E97G/A98S/A101V) within the predicted P-box of the ERR␣1 DNA binding domain that abrogate the ability of the protein to bind DNA. It was constructed by PCR amplification of the open reading frame of ERR␣1 in two directly abutting fragments corresponding to the amino and carboxyl termini of the protein. The ERR␣1 P-box aminoterminal fragment was amplified using as primers the wild-type translation initiation codon-containing primer and 5Ј-phosphate-GGACCC-ACAGGATGCCACACCATAGTGGTA-3Ј. The ERR␣1 P-box carboxylterminal fragment was amplified using as primers the wild-type translation termination codon-containing primer and 5Ј-phosphate-TGCAAAGTCTTCTTCAAGAGGACCATCCA-3Ј. The resulting PCR products were digested with EcoRI and BamHI, respectively, ligated together, and re-amplified using the wild-type initiation and termination codon-containing primers to produce the full-length ERR␣1 P-box mutant.
Cells-The ER-positive, human mammary carcinoma MCF-7 cell line was cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS), 6 ng of insulin/ml, 3 g of glutamine/ml, and 100 units of penicillin and streptomycin/ml. The ER-negative, human cervical HeLa cell line and the monkey kidney COS-M6 cell line were cultured in Dulbecco's modified Eagle's medium supplemented with 10% FBS and 100 units of penicillin and streptomycin/ml. When cells were cultured in estrogen-free medium, referred to here as stripped medium, dextran-coated charcoal-treated FBS (34) replaced whole FBS and phenol red-free RPMI 1640 replaced RPMI 1640.
Transient Transfections and Luciferase Assays-To assess the role ERR␣1 plays in regulating transcription of an ERE-containing promoter, MCF-7 or HeLa cells grown in 12-well tissue culture plates were co-transfected in parallel with 0.5 g of pTK-luc versus p3xERE-TK-luc along with the indicated amounts of the empty cloning vector pcDNA3.1, the ERR␣1 expression plasmid pcDNA3.1-hERR␣1, or mutant variants thereof. Transfections were performed with the aid of the TransIT LT1 transfection reagent (PanVera, Madison, WI) as previously described (35). To examine the effects of ER ligands, cells were maintained in stripped medium for 48 h before transfection and the addition of 17␤-estradiol (E 2 ) (Sigma) or the pure anti-estrogen ICI-182,780 (Astra Zeneca, London, UK) dissolved in ethanol and diluted in medium to obtain the indicated concentrations. Cells were harvested 48 h post-transfection, and lysates were assayed for luciferase activity normalized to protein concentration as previously described (36).

FIG. 1. Schematic representations of ER and ERR family members.
A, comparison of the sequence similarity between human ER␣ and human ERR␣1. The letters A-F indicate the domains typically found in NRs. DBD and LBD denote the DNA binding and ligand binding domains, respectively. The numbers refer to the amino acid residues from the amino terminus. The percentages indicate the amino acid sequence similarity between the corresponding domains of the two proteins, with ER␣ sequences set at 100%. B, structures of the variants of ERR␣1 studied here. Amino acid substitution mutations are indicated at the sites of the numbered residues. formed essentially as described by Reese et al. (37) using whole-cell extracts prepared as described previously (35). Briefly, whole-cell extracts obtained from five 10-mm dishes of COS-M6 cells that had been transfected 48 h previously with 3 g/dish of the desired expression plasmid served as the source of NRs. Transfections were performed with the aid of the TransIT LT1 transfection reagent as previously described (35). The radiolabeled double-stranded synthetic oligonucleotide 5Ј-TAAGCTTAGGTCACAGTGACCTAAGCTTA-3Ј served as the ERE probe. One to five l of extract (10 -100 g of protein) was preincubated on ice for 20 min in a 16-l reaction mixture containing 20 mM HEPES (pH 7.4), 1 mM dithiothreitol, 100 mM NaCl, 10% glycerol (v/v), 3 g of BSA, and 4 g of poly(dI-dC). Radiolabeled probe (ϳ1.0 ng) was added, and the mixture was incubated for 15 min at room temperature. The samples were loaded directly onto a 5% non-denaturing polyacrylamide gel with 0.5ϫ Tris-buffered EDTA as running buffer and electrophoresed at 200 V for 2 h at 4°C. Immunoshift assays were performed by the addition of the indicated antiserum at the preincubation step. The ER␣-specific antiserum was the monoclonal antibody H222 (kindly provided by Dr. Geoffrey Greene). The hERR␣1-specific antiserum was the previously described polyclonal one raised in rabbits against glutathione S-transferase GST-hERR␣1   (11).
Western Blots-To determine whether the NRs were efficiently and correctly expressed, 5-10 g of whole-cell extract containing the overexpressed NR were resolved by SDS, 12% PAGE. The proteins in the gel were electroblotted onto a nitrocellulose membrane. The membranes were probed with a rabbit polyclonal antiserum raised against GST-hERR␣1 (11) followed by anti-rabbit IgG peroxidase (1:1000 dilution). The retained antibodies were detected by enhanced chemiluminescence.

Competition between ERR␣1 and ER␣ for Binding an ERE-
Johnston et al. (11) and others (12) showed previously that ERR␣1 can bind to some naturally occurring EREs. To test whether ERR␣1 recognizes the palindromic ERE sequence, 5Ј-TAAGCTTAGGTCACAGTGACCTAAGCTTA-3Ј, EMSAs were performed using whole-cell extracts obtained from COS-M6 cells that contained overexpressed ER␣ or ERR␣1 as protein source and a radiolabeled, double-stranded synthetic oligonucleotide that contained the palindromic ERE sequence as probe. As expected, both ER␣ and ERR␣1 bound to this synthetic ERE ( Fig. 2A). ER␣ generated a protein-DNA complex ( Fig To determine whether the binding of ERR␣1 and ER␣ to this ERE is mutually exclusive, cooperative, or competitive, EMSAs were performed with a constant amount of ER␣ plus various amounts of ERR␣1 mixed together in the same binding reaction. The addition of increasing amounts of ERR␣1 yielded an increase in the amount of ERR␣1-DNA complex along with a corresponding decrease in the amount of ER␣-DNA complex (Fig. 2B, lanes 3-6). Similar results were obtained when this competition experiment was performed with extracts of COS-M6 cells that had been co-transfected with various molar ratios of the ER␣and ERR␣-expressing plasmids (data not shown). Thus, the binding of ERR␣1 and ER␣ to this palindromic ERE is mutually exclusive, with ERR␣1 effectively competing with ER␣ for binding when present in sufficient amounts.
ERR␣1 Represses Transcription in MCF-7 Cells-Because ERR␣1 can interfere with the binding of ER␣ to this ERE, might it affect estrogen-responsive transcription? To answer this question, we co-transfected ER-positive mammary MCF-7 cells in parallel with p3xERE-TK-luc, a reporter plasmid con-  (11); lane 6, ERR␣1-containing extract plus ER␣-specific antibody H222; lane 7, ERR␣1-containing extract plus ERR␣1-specific polyclonal antiserum. The arrows indicate the specific DNA-protein complexes and free probe DNA. The asterisk denotes antibody-supershifted complexes. B, ER␣ and ERR␣1 compete for binding to the ERE. The radiolabeled ERE probe was incubated with a constant amount of whole-cell extract containing ER␣ by itself (lane 2) or with increasing amounts of whole-cell extract containing ERR␣1 (lanes 3-6). Lanes 7 and 8 contained 2.5 g of ER␣-containing extract plus 2.5 g of ERR␣1-containing extract preincubated with ER␣-specific antibody or hERR␣1-specific polyclonal antiserum, respectively. taining three tandem copies of this ERE (Fig. 3A) or the EREnegative control plasmid, pTK-luc, together with 0.12 or 0.25 g of the ERR␣1 expression plasmid, pcDNA3.1-hERR␣1, or the empty parental expression plasmid, pcDNA3.1. After incubation for 48 h in medium containing whole FBS, the cells were harvested and assayed for luciferase activity. The presence of the EREs conferred an ϳ150-fold increase in transcriptional activity above the level observed in the cells transfected with the ERE-negative reporter plasmid (Fig. 3B). This ERE-dependent activity was extremely sensitive to treatment with the anti-estrogen ICI-182,780, indicating the activation is dependent upon ER and the presence of estrogens in the FBS (Fig. 3B). Overexpression of ERR␣1 repressed this ERE-dependent transcriptional activity ϳ3-4-fold ( Fig. 3B; see also Fig. 7B) while exhibiting little if any effect on expression of the control pTKluc reporter plasmid (Fig. 3, B and C). Thus, ERR␣1 inhibits the estrogen-responsive activation of transcription from this ERE-containing promoter in this ER-positive MCF-7 cell line.
ERR␣1 Modulates Estrogen-responsiveness in MCF-7 Cells-To directly examine the effect of ERR␣1 on the response to estrogen, MCF-7 cells were cultured in medium containing estrogen-free charcoal-stripped FBS and cotransfected as described above but in the absence or presence of E 2 . In the absence of exogenous E 2, p3xERE-TK-luc was expressed at an ϳ50-fold higher level than pTK-luc, activity that was largely ablated by treatment with the anti-estrogen ICI-182,780 (Fig.  4A). Once again, overexpression of ERR␣1 repressed ERE-dependent transcription ϳ3-fold while having little if any effect on transcription of the control reporter (Fig. 4A).
The addition of 1 ϫ 10 Ϫ10 M E 2 , a physiological concentration, to the medium dramatically induced transcriptional activity of the ERE-containing reporter plasmid ϳ10-fold above the level observed in the absence of exogenous E 2 (Fig. 4, B versus A). Consistent with this induction being mediated by ER, it was completely eliminated by incubation of the cells with the antiestrogen ICI-182,780 (Fig. 4B). Strikingly, overexpression of ERR␣1 inhibited this E 2-mediated activation of transcription from the ERE-containing promoter 3-5-fold, while, again, having little if any affect on expression of the ERE-negative control, pTK-luc (Fig. 4B).
Lastly, when the co-transfected cells were incubated with 1 ϫ 10 Ϫ8 M E 2, a non-physiological concentration, ERE-dependent transcription was stimulated an additional 3-fold over the level of activity observed with 10 Ϫ10 M E 2 to a level ϳ30-fold above the activity observed in the absence of exogenous E 2 (Fig.  4C). Again, this activity was extremely sensitive to treatment with the anti-estrogen ICI-182,780 (Fig. 4C), consistent with high concentrations of E 2 generating high levels of liganded, active ER. However, overexpression of ERR␣1 no longer inhibited this E 2 -mediated activation of transcription from p3x ERE-TK-luc (Fig. 4C). Thus, we conclude that ERR␣1 can modulate estrogen responsiveness when its concentration relative to E 2occupied ER␣ is sufficient to allow effective competition for binding to the ERE. ERR␣1 Represses Transcription by an Active Mechanism-What is the mechanism by which ERR␣1 inhibits ER-mediated transcriptional activation? One possibility is that it simply competes with ER␣ for mutually exclusive binding to EREs, thereby blocking binding of the transcriptional activator. Alternatively, ERR␣1 may contain a regulatory domain(s) as well as a DNA binding domain that plays an active role in modulating transcription from ERE-containing promoters.
A second variant examined was ERR␣1 76 -423 . This aminoterminal-deleted variant of ERR␣1 lacks the A/B domain but retains the entire ligand and DNA binding domains (Fig. 1B). Immunoblots indicated that ERR␣1 76 -423 appeared to accumulate in transfected cells to somewhat lower levels than fulllength ERR␣1 (Fig. 5A, lanes 5 versus 2, with lane 5 containing 3-fold more whole-cell extract). It is unclear whether the ERR␣1 76 -423 variant protein lacks some of the epitopes recognized by the polyclonal ERR␣1-specific antiserum used here, thus resulting in it being detected at lower efficiency, or that it actually accumulated to lower levels because of differences in rates of synthesis or stability. Regardless, sufficient quantities of protein accumulated for studies of DNA binding and transcriptional activity. As expected, hERR␣1 76 -423 was found both to bind to the palindromic ERE (Fig. 5B, lane 7) and to repress transcription of p3xERE-TK-luc approximately 2-3-fold (Fig.  7B). Thus, although both ERR␣1 76 -423 and ERR␣1 1-173 bind to the ERE, only ERR␣1 76 -423 represses ERE-dependent transcription. Therefore, a domain(s) of ERR␣1 mapping within the carboxyl-terminal region of the protein in addition to its DNA binding domain is required for repression. These findings support an active model of transcriptional repression in which ERR␣1 represses transcription by recruiting cellular corepressor(s) to the promoter.
As is true for most NRs, ERR␣1 contains a coactivator binding motif or NR box. The ERR␣1 NR box, located between amino acids 413 and 418, is comprised of the sequence LX-LXXL. This sequence differs slightly from the consensus NR box motif, LXLXXL (38 -42). To examine the effect of inactivation of this coactivator binding motif on transcriptional activity, we constructed pcDNA3.1-hERR␣1 413A/418A . The ERR␣ encoded by this plasmid contains alanine substitution mutations in place of the leucine residues at amino acids 413 and 418. Immunoblots indicated that ERR␣1 413A/418A efficiently accumulated in transfected cells (Fig. 5A, lane 6). EMSAs showed that it specifically bound the palindromic ERE (Fig. 5B, lanes 9  and 10). Interestingly, ERR␣1 413A/418A repressed ERE-depend-ent transcription more efficiently than did full-length hERR␣1, i.e. approximately 5-8-fold versus 3-4-fold, respectively (Fig.  7B). Somewhat surprisingly, ERR␣1 413A/418A up-regulated ERE-independent transcription ϳ2-fold (Fig. 7A), likely the result of sequestration by overexpressed ERR␣1 413A/418A of corepressors utilized by this control promoter . Thus, the repressive effect of ERR␣1 413A/418A on ERE-dependent transcription was 10 -16-fold if normalized to the control. Quite likely, ERR␣1 contains both corepressor and coactivator binding domains, with these domains acting in concert to determine the overall effect of ERR␣1 on transcription. Thus, inactivation of the coactivator binding NR box motif potentiates repression by ERR␣1.
Last, we constructed pcDNA3.1-hERR␣1 P-box . This plasmid encodes a variant of ERR␣1 containing three amino acid substitution mutations within the DNA binding domain (Fig. 1B). ERR␣1 P-box accumulated to normal levels in transfected cells (Fig. 5A, lanes 3 versus 2). As expected, it was incapable of binding to the palindromic ERE (Fig. 5B, lane 4). ERR␣1 P-box also failed to interfere with the binding of either ER␣ (Fig. 6A,  lanes 7-9) or wild-type ERR␣1 (Fig. 6B, lanes 6 and 7) to the palindromic ERE. Most interestingly, overexpression of ERR␣1 P-box led to a 2-3-fold induction of ERE-dependent transcription (Fig. 7) rather than repression or no effect. Induction could not have been a consequence of sequestration of endogenous ERR␣1 away from the ERE via protein-protein interactions since ERR␣1 P-box did not interfere with binding of wildtype ERR␣1 to the ERE (Fig. 6B). Rather, ERR␣1 P-box likely sequestered cellular corepressors away from DNA-bound endogenous ERR␣1, thereby relieving repression. These findings provide further support for the hypothesis that ERR␣1 probably functions as a repressor of E 2 -stimulated, ERE-dependent transcription via an active mechanism.
To determine whether the transcriptional repression observed here was dependent upon the cell line, we repeated the cotransfection experiments as described above except using ER-negative HeLa cells in place of ER-positive MCF-7 cells (Fig. 8). Contrary to the results obtained in MCF-7 cells (Figs. 3B and 7B), overexpression of ERR␣1 in HeLa cells resulted in a 2.5fold activation of transcription from the p3xERE-TK-luc reporter plasmid (Fig. 8A). A similar level of ERE-dependent activation was also observed in CV-1 and COS-M6 cells, other ER-negative cell lines (data not shown). Thus, ERR␣1 is a constitutive, estrogen-independent activator of transcription in these ER-negative cell lines. We conclude that ERR␣1 can function as either a repressor or activator of ERE-dependent transcription in a cell type-specific manner.
Also noteworthy is the fact that the transcriptional activity of the ERE-containing p3xERE-TK-luc plasmid was already ϳ100-fold higher than that of its matched ERE-negative control plasmid, pTK-luc, even in the absence of overexpressed ERR␣1 (Fig. 8, A versus B). Unlike in MCF-7 cells (Fig. 3B), in HeLa cells this ERE-dependent activity was completely insensitive to the anti-estrogen ICI-182780 (Fig. 8A) and, therefore, not mediated by ERs. Because HeLa cells contain high endogenous levels of ERR␣1 (Ref. 32, data not shown), we conclude that endogenous ERR␣1, not ER␣, likely mediated this high ERE-dependent transcriptional activity in these cells. Furthermore, the only modest induction observed in HeLa cells with overexpressed ERR␣1 was likely due to the already abundant presence of endogenous ERR␣1. Thus, we conclude that ERR␣1 FIG. 6. Competition between ERR␣1 variants and ER␣ or ERR␣1 for binding to the palindromic ERE. A, ERR␣1 1-173 , but not ERR␣1 P-box , competes with ER␣ for binding to the ERE. The radiolabeled ERE probe was incubated with a constant amount of whole-cell extract expressing ER␣ alone (lane 2) or together with increasing amounts of whole-cell extract containing overexpressed ERR␣1 1-173 (lanes 4 -6) or ERR␣1 P-box (lanes 7-9). Lane 1, probe alone; lane 3, extract was preincubated with the ER␣-specific antibody, H222. The arrows indicate the specific and nonspecific (NS) DNA complexes and free probe. B, ERR␣1 1-173 , but not ERR␣1 P-box , competes with wild-type ERR␣1 for binding to the ERE. Experiments were performed as described in panel A above but with a constant amount of whole-cell extract containing overexpressed wild-type ERR␣1 in place of ER␣. is a strong constitutive activator of ERE-dependent transcription in HeLa cells. DISCUSSION We examined here the transcriptional properties of ERR␣1 when it acts via binding an ERE. We showed that ERR␣1 directly competes with ER␣ for binding to a consensus palindromic ERE (Fig. 2) and down-modulates the transcriptional response to estrogen in an ERE-dependent manner in MCF-7 cells (Figs. 3 and 4). Using variants of ERR␣1, we further showed that repression is not simply the result of ERR␣1 interfering with the binding of ER␣ to DNA; rather, it occurs via an active mechanism (Figs. 5-7). Interestingly, ERR␣1 functions as an activator rather than a repressor of this same promoter via its EREs in ER-negative HeLa cells (Fig. 8). Thus, ERR␣1 operates as an active repressor or activator of ERE-dependent transcription based upon other properties of the cell.
Down-modulation of Estrogen Response by ERR␣-What is the mechanism by which overexpression of ERR␣1 in ER-positive MCF-7 cells leads to antagonism of the response of an ERE-containing promoter to estrogens? We showed that ERR␣1 competes with ER␣ for binding to the consensus palindromic ERE (Fig. 2). We hypothesize that estrogen responsiveness is governed by the percentage of EREs occupied by ER␣, with ERE occupancy determined by the relative concentrations of E 2 -activated ER␣ and ERR␣1 in the cell. MCF-7 cells contain high endogenous levels of ER␣ (33) that exist in a ligandactivated complex when E 2 is present. In this case, most EREs are bound by ligand-activated ER␣, and expression of the reporter gene is high (Fig. 4C). On the other hand, when ERR␣1 is overexpressed and there is little ligand-activated ER␣ present, most EREs are bound by ERR␣1, and expression of the reporter gene is low (Fig. 4A). In this way, expression of the ERE-containing promoter is regulated by cross-talk between these two nuclear receptors. Thus, the level of expression of an ERE-dependent gene depends in part upon the relative amounts of ERR␣1 and ligand-activated ER␣ in the cell.
Mechanism of Repression by ERR␣1-Previously, Burbach et al. (43) showed that COUP-TF1 represses estrogen-dependent stimulation of the oxytocin gene by simply competing with ER␣ for binding to an ERE. However, based upon analysis of variants of ERR␣1, we conclude here that repression by ERR␣1 involves, instead, an active silencing mechanism. First, ERR␣1 1-173 retains its DNA binding activity (Ref. 21;Figs. 5 and 6), yet failed to repress transcription (Fig. 7). Thus, simply blocking the binding of ER␣ is not sufficient for ERR␣1 to repress ERE-mediated transcription. Second, ERR␣1 76 -423 , a variant lacking the amino-terminal domain but retaining both the DNA binding and carboxyl-terminal domains repressed transcription as well as full-length ERR␣1 (Fig. 7). Therefore, in addition to the DNA binding domain, a region within the carboxyl terminus is required for ERR␣1 to repress E 2 -stimulated, ERE-dependent transcription. Third, ERR␣1 413A/418A , a variant containing mutations only within the LXLXXL coactivator binding NR box motif, repressed E 2 -stimulated, ERE-dependent transcription more efficiently than did wild-type ERR␣1 (Fig. 7). We interpret this latter result to indicate that FIG. 7. ERR␣1 represses transcription by an active silencing mechanism. MCF-7 cells were co-transfected with 0.5 g of (A) pTK-luc or (B) p3xERE-TK-luc and 0.12 g or 0.25 g of the empty vector pcDNA3.1, pcDNA3. 1-hERR␣1, pcDNA3.1-hERR␣1 1-173 , pcDNA3.1-hERR␣1 76 -423 , pcDNA3. 1-hERR␣1 413A/418A , or pcDNA3.1-hERR␣1 P-box. After incubation for 48 h in medium containing whole FBS, the cells were harvested, and luciferase activity was determined with normalization to the protein concentration of each extract. The data are presented in panel A relative to the activity observed with pTK-luc plus 0.12 g of pcDNA3.1; they are presented in panel B relative to the activity observed with p3xERE-TK-luc plus 0.12 g pcDNA3.1. All data shown represent means plus the S.E. from three separate experiments, each performed in triplicate.

FIG. 8. ERR␣1 activates rather than represses ERE-dependent transcription in HeLa cells.
Experimental details are identical to the ones described in Fig. 3B, except that ER-negative HeLa cells were used in place of ER-positive MCF-7 cells. The data in panels A and B are presented in the same format as in Fig. 3, panels B and C, respectively. ablation of the NR box disrupts the balance of ERR␣1-bound co-regulators, thereby allowing any putative corepressor bound to ERR␣1 to act more effectively. Last, ERR␣1 P-box , a variant whose DNA binding activity was abrogated but coregulator binding domains were left intact, specifically up-regulated rather than antagonized ERE-dependent transcription (Fig. 7). This latter finding is likely a consequence of repression domains present within ERR␣1 P-box competing with endogenous wild-type ERR␣1 for binding cellular corepressors, thereby preventing endogenous ERE-bound ERR␣1 from antagonizing transcription. Furthermore, ERR␣1 can function as an active repressor even in the absence of ER. For example, we have found that ERR␣1 represses SV40 late gene expression in ER-negative CV-1 cells both from the natural ERR response elements overlapping the transcription initiation site of the SV40 major late promoter (11) and when this ERR response element is relocated to 50 bp upstream of the transcription initiation site (data not shown). Taken together with previous findings of others (20), these results provide evidence that ERR␣1 contains both repression and activation domains. We have also shown elsewhere (44) that silencing mediator for retinoid and thyroid hormone receptors (SMRT) is one of the corepressors that can bind ERR␣1, binding within the hinge region of ERR␣1. Additional experiments will be needed to identify the corepressors of ERR␣1 and to definitively map their sites of binding.
Contrary to our findings with ERR␣1 1-173 , Zhang and Teng (21) reported that the amino-terminal region of ERR␣1 contains repressor activity. However, they assayed the effects of Gal4DBD-ERR␣1 chimeras on expression of a Gal4 reporter rather than non-chimeric variants of ERR␣1 binding via the ERR␣1 DNA binding domain to an ERE. Whether these differences in experimental design can account for the seemingly contradictory conclusion is not yet clear.
ERR␣1 has also been shown to bind ER␣ directly (11). Thus, alternative, non-mutually exclusive hypotheses to explain the ability of ERR␣1 to down-modulate ERE-dependent transcription include (i) ERR␣1 forming true heterodimers with ER␣ that can bind EREs and (ii) ERR␣1 interacting with ER␣ in ways that abrogate the ability of ER␣ to bind EREs. However, we failed to observe ER␣-ERR␣ heterodimeric complexes in either the experiments presented here (Fig. 2) or EMSAs performed using whole-cell extracts obtained from COS-M6 cells co-transfected with the ER␣ and ERR␣ expression plasmids (data not shown). Moreover, the presence of ERR␣1 P-box failed to interfere with the binding of ER␣ to DNA (Fig. 6A). Taken collectively, these data indicate that ERR␣1 likely functions as a repressor independently of any ability to bind ER␣.
Activation of Transcription by ERR␣1-Confirming prior reports (17-31), we have also observed that ERR␣1 can activate transcription from an ERE-regulated promoter (Fig. 8). We show here for the first time that whether ERR␣1 functions as a repressor or activator of a specific promoter can depend upon the cell type (Figs. 3 versus 8). What factors determine the activity of ERR␣1? Several possibilities exist. First, the activities of ERR␣1 might be ligand-dependent. Previous reports appear to be contradictory as to the existence of an exogenous activating ligand. One indicated that transcriptional activation by ERR␣ depends upon a component present in serum (17). Others claimed that ERRs are not activated by naturally occurring ligands (19,20). In the experiments reported here, the same serum was present in the medium in which the HeLa and MCF-7 cells were cultured; nevertheless, ERR␣1 exhibited markedly different activities in these cell types (Figs. 3 versus  8). Thus, if an activating ligand of ERR␣1 exists in FBS, it probably does not exclusively determine the activity of hERR␣1. Furthermore, we found that charcoal-dextran treatment of the serum did not affect the silencing activity of ERR␣1 (Fig. 4A), supporting the notion that ERR␣1 functions as a repressor independently of an exogenous ligand. One alternative possibility is that various cell types may or may not endogenously synthesize the putative ligand of ERR␣1, thereby determining the transcriptional properties of ERR␣1 in those cells. Second, the differences in transcriptional activity observed here might be a reflection of differences in the coregulators present in these cell types. Third, by analogy with ER␣ (45)(46)(47)(48), the phosphorylation state of ERR␣1 may affect its functional activities. Indeed, Sladek et al. (26) showed that murine ERR␣1 can be phosphorylated in vivo. Likewise, we have found that human ERR␣1 can be phosphorylated in vitro by MAP kinase. 2 Model for ERR␣1 Modulation of Estrogen Responsiveness-Based upon the data presented here, we postulate that ERR␣1 plays key roles in the regulation of estrogen-responsive genes by efficiently binding EREs (Ref. 11; data not shown), leading either to modulation of the response to estrogens or functional substitution for ER as a constitutive activator of ERE-dependent transcription. Furthermore, the cellular concentrations of ER␣ and ERR␣1, together with the differential transcriptional properties of ERR␣1, determine the transcriptional response of an ERE-regulated promoter. For example, when the concentrations of both ER␣ and ERR␣1 are low or the level of the repressor form of ERR␣1 is high, an ERE-dependent gene is expressed at intermediate or low levels (Fig. 9, rows 1 and 2, respectively). Low and high concentrations of the repressor form of ERR␣1 relative to high amounts of active ER complex yield intermediate or high ERE-dependent gene expression (Fig. 9, rows 3 and 4, respectively). Last, in the absence of active ER, the activator form of ERR␣1 can constitutively activate ERE-dependent transcription (Fig. 9, rows 5 and 6).
Both estrogens acting through ERs and kinase signaling pathways contribute to the initiation and progression of some breast cancers. Because ERR␣1 plays multiple roles in regulation of ERE-dependent transcription (Fig. 9), we hypothesis that the functionality of ERR␣1, possibly modulated by kinase signaling events, leads to the development or progression of some breast cancers. We propose that the silencing activity of ERR␣1 tightly regulates estrogen responsiveness in normal breast cells (Fig. 9, rows 1 and 2). Some cancerous cells attain very high levels of ER␣, thereby maximizing the mitogenic affects of estrogen (Fig. 9, rows 3 and 4). In addition, some breast cancers present as ER-negative (49 -51) or develop re-2 E. Ariazi, unpublished data.  9. Model for ERR␣1 modulation of estrogen responsiveness. See "Discussion" for details. The subscript rep denotes ERR␣1 that is functioning as a repressor; the subscript act denotes ERR␣1 that is functioning as an activator. Plus (ϩ) and minus (Ϫ) symbols indicate relative levels of ERE-dependent transcription. sistance to hormonal treatment (52). Under either of these circumstances, ERR␣1 may functionally substitute for ER if it is in an active form, thereby constitutively activating EREregulated transcription (Fig. 9, rows 5 and 6). Thus, the conversion of ERR␣1 from a repressor to an activator by a mechanism(s) yet to be determined may be a critical step in the progression to a hormone-independent phenotype.