Chicken Ovalbumin Upstream Promoter-Transcription Factor Interacts with Estrogen Receptor, Binds to Estrogen Response Elements and Half-Sites, and Inhibits Estrogen-induced Gene Expression*

Chicken ovalbumin upstream promoter-transcription factor (COUP-TF) was identified as a low abundance protein in bovine uterus that co-purified with estrogen receptor (ER) in a ligand-independent manner and was separated from the ER by its lower retention on estrogen response element (ERE)-Sepharose. In gel mobility shift assays, COUP-TF bound as an apparent dimer to ERE and ERE half-sites. COUP-TF bound to an ERE half-site with high affinity,K d = 1.24 nm. In contrast, ER did not bind a single ERE half-site. None of the class II nuclear receptors analyzed, i.e. retinoic acid receptor, retinoid X receptor, thyroid receptor, peroxisome proliferator-activated receptor, or vitamin D receptor, were constituents of the COUP-TF·DNA binding complex detected in gel mobility shift assays. Direct interaction of COUP-TF with ER was indicated by GST “pull-down” and co-immunoprecipitation assays. The nature of the ER ligand influenced COUP-TF-ERE half-site binding. When ER was liganded by the antiestrogen 4-hydroxytamoxifen (4-OHT), COUP-TF-half-site interaction decreased. Conversely, COUP-TF transcribed and translated in vitroenhanced the ERE binding of purified estradiol (E2)-liganded ER but not 4-OHT-liganded ER. Co-transfection of ER-expressing MCF-7 human breast cancer cells with an expression vector for COUP-TFI resulted in a dose-dependent inhibition of E2-induced expression of a luciferase reporter gene under the control of three tandem copies of EREc38. The ability of COUP-TF to bind specifically to EREs and half-sites, to interact with ER, and to inhibit E2-induced gene expression suggests COUP-TF regulates ER action by both direct DNA binding competition and through protein-protein interactions.

The estrogen receptor (ER) 1 is a transactivating enhancer protein that is a member of the ligand-activated steroid/nuclear receptor gene superfamily of proteins that share two conserved regions, the DNA-binding (C) and the ligand binding (E) domains (1). The physiological sequelae of estrogen action involve a passive diffusion of ligands, e.g. estradiol (E 2 ), into cells and the binding of E 2 with high affinity and specificity to ER in the nucleus of target cells. Ligand binding initiates a series of steps forming an "activated," homodimeric E 2 ⅐ER complex that binds with high affinity to specific DNA sequences, estrogen response elements (ERE). Analysis of the 5Ј-regulatory regions of numerous estrogen-responsive genes revealed a 13-bp palindromic ERE consensus sequence: 5Ј-GGTCAnnnTGACC-3Ј (EREc, where n ϭ any nucleotide in the center spacer region). EREc is the minimal ERE that conferred estrogen responsiveness to reporter genes analyzed by transfection assay (2,3). However, the crystal structure of the ER DNA binding domain when bound to DNA showed that ER contacts both the 5Ј-C and A in the more extended palindrome 5Ј-CAGGTCAnnn-TGAC-CTG-3Ј (4), indicating that a longer inverted repeat (IR) may stabilize ER binding.
Most estrogen-responsive genes identified to date contain one or more imperfect EREs or multiple copies of the ERE half-site rather than EREc (5). The latter genes are also regulated by class II nuclear receptors, e.g. thyroid hormone receptor (TR), retinoic acid receptor (RAR), retinoid X receptor (RXR), and orphan receptors, e.g. estrogen-related receptor (ERR), that bind to direct and inverted repeats (DR and IR) of the ERE half-site 5Ј-AGGTCA-3Ј (6,7). Recent studies demonstrated that ER also binds various spaced DR and IR of the ERE half-site motif, albeit with significantly lower affinity when compared with ER binding to EREc (5,8).
Once bound to an ERE, the precise mechanism of transcriptional activation, or repression, by the ER is unknown. ERmediated effects on transcription are thought to involve interaction between the DNA-bound ER and transcription factors, coactivator proteins, e.g. ERAP160; RIP140, SPT6; SRC-1; TIF1, TIF2, or components of the TATA binding complex including TFIIB and TATA-box binding protein (reviewed in Ref. 9). Two distinct ER regions are involved in these interactions as follows: an N-terminal, ligand-independent activation function 1 (AF-1), and a C-terminal, ligand-dependent AF-2 (10,11). The total and relative activity of each AF varies with the promoter and cell type, indicating that the function of each AF is mediated by interaction with cell-specific proteins (12)(13)(14).
In previous work, we quantitated the effect of ERE sequences, spacing, and the role of sequences flanking the ERE on the affinity and stoichiometry of liganded ER-ERE interaction in vitro (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25). In sum, our results indicate a critical role for sequences flanking the ERE and the relative helix orientation of the EREs on ER binding affinity. We also showed that the ER ligand modulates ERE binding parameters in vitro and that the purity of the ER preparation influences the affinity of ERE binding. Highly purified ER binds to EREs with significantly lower affinity than partially purified ER, i.e. for E 2 -ER K d ϭ 1.74 versus 0.24 nM (25). The latter result implies that proteins present in a partially purified preparation of bovine uterine ER facilitate high affinity binding of ER to EREs. Here we identified COUP-TF as a constituent of the partially purified bovine uterine ER⅐ERE binding complex. COUP-TF is an orphan nuclear receptor that binds to DR or IR of the ERE half-site as a homodimer or as a heterodimer with RXR (26). We have examined the ability of COUP-TF to bind full and half-site EREs in vitro and to modulate the expression of an estrogen-responsive reporter gene in transiently transfected ER expression MCF-7 human breast cancer cells.

EXPERIMENTAL PROCEDURES
Preparation of Plasmids Containing EREs-The sequences of select synthetic single-stranded oligonucleotides are given in Table I. EREc38 is a 38-bp ERE consensus sequence (15). 1/2EREc38 and 1/2ERE3Јc38 contain half-site EREs. Double-stranded oligomers were ligated into the SmaI restriction site of the vector pGEM-7Zf(ϩ) (Promega) as described (18). The ␤RARE (retinoic acid response element) is a synthetic version of the mouse RAR type ␤ gene (27)(28)(29).
Single or multiple, head-to-tail, tandem copies of EREc38 were removed from pGEM-7Zf(ϩ) (18) by double digestion with KpnI and SacI and cloned directly into the upstream multiple cloning site in pGL3-Promoter vector (Promega). This places the ERE 83 nucleotides upstream of the SV40 promotor. The constructs containing EREc38 are called pGL3-1EREc38, 2EREc38, and 3EREc38 with the number indicating the number of tandem copies of EREc38.
Preparation of Recombinant COUP-TFI-The pRSV-COUP-TFI plasmid, encoding recombinant human COUP-TFI, was a gift of Dr. Sophia Y. Tsai of Baylor University (33,34). COUP-TFI was transcribed and translated in vitro using the TNT rabbit reticulocyte lysate system from Promega (Madison, WI) according to the manufacturer's instructions. As a positive control, a plasmid encoding luciferase was transcribed and translated in parallel with COUP-TFI. The relative amount of the translated COUP-TFI protein was determined by [ 35 S]methionine (1175 Ci/mmol from NEN Life Science Products) incorporation into proteins analyzed by 10% SDS-PAGE followed by treatment of the fixed gel with En 3 Hance (NEN Life Science Products) according to the manufacturer's instructions for fluorography. The dried gel was exposed to Kodak X-Omat film (Eastman Kodak Co., Rochester, NY.) for 8 -24 h.
Electrophoretic Mobility Shift Assay-Plasmid DNA was extracted and purified from transformed JM109 Escherichia coli by the Mega Prep procedure from Qiagen (Chatsworth, CA). ERE-containing plasmid DNA (pGEM7Zf(ϩ), as described above (18), was digested with EcoRI and BamHI, and the ERE-containing oligomers were purified by polyacrylamide gel electrophoresis and electroelution. The exact nucleotide sequence of the EcoRI-BamHI oligomer containing EREc38, inserted in the inverse orientation from that in Table I, is 5Ј-AATTC-GGTACCCCTGAATGTGTTATTTTAGCTC-AGGTCACTCTGACCT-GGGGT-TCGAAATCGATAAGCTT-3Ј and for 1/2EREc38: 5Ј-AATTC-GGTACCCCTGAATGTGTTATTTTAGCTCGAAATGCT-CTGACCTGGGGTTCGAAATCGATAAGCTTG-3Ј where two imperfect half-sites are in bold, the perfect half-site is underlined, and the plasmid DNA sequence is written in italics. ERE-containing oligomers were fill-in labeled with [ 32 P]dATP (800 Ci/mmol from NEN Life Science Products) using Klenow large fragment DNA polymerase I (New England Biolabs, Beverly, MA). Unincorporated nucleotides were removed by centrifugation through either a G-50 Sephadex spin column in TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0) or a Centri-Spin 20 column (Princeton Separations, Adelphia, NJ). The sizes of the ERE oligomers used were 77 and 115 bp for one or two tandem (head-to-tail) copies of EREc38, respectively; 77 bp for either 1/2EREc38 or 1/ 2ERE3Јc38; 106 bp for 1/2EREc; and 43 bp for the AT-rich region. Labeled oligomer (25,000 cpm) was incubated for 2 h at 4°C with partially or ERE-Sepharose affinity purified The amount of ER⅐ERE complex formed, and that of free ERE, was determined by excision of the corresponding regions from the dried gels into scintillation vials containing 3 ml of EcoScint A (National Diagnostics, Atlanta, GA), and the radioactivity was counted. The fraction of total 32 P-ERE in the ER⅐ERE complex was calculated as follows: F(t) ϭ (cpm in the ER⅐ERE complex)/(total cpm in the lane), where the (total cpm in the lane) ϭ (cpm in free ERE) ϩ (cpm in the ER⅐ERE complex) (35).
Antibodies-H222 monoclonal antibody (MAb) to ER was a gift of Abbott. H222 was diluted 1:10 in TE and 1.0 l was added to selected samples in each experiment to confirm the identity of ER protein in the retarded ER⅐ERE complexes. Monoclonal (mouse) anti-ER antibodies AER304, AER314, AER308, AER315, AER303, AER310, AER311, AER317, and AER320 (36) were generous gifts from Neomarkers (Lab Vision Corp., Fremont, CA). Polyclonal anti-ER antibody ER715 (37) was a gift of the National Hormone and Pituitary Program of the NIDDKD, National Institutes of Health.
The following antibodies were gifts of Dr. Abdulmaged M. Traish of Boston University: ER-specific polyclonal antibodies AT2A, AT3A, and AT3B recognize the DNA-binding domain of the human ER and show no reactivity to glucocorticoid receptor, progesterone receptor, or androgen receptor (38). MAbs 33 and 213, and MAbs NMT-1 and NMT-2 react specifically with ER (39 -41). MAbs to RAR␣ and RAR␥ were prepared by standard hybridoma technology (42). Polyclonal RAR and RXR antibody RAR␥IIIB were raised against a region just N-terminal to the first zinc finger and recognized all forms of RAR and RXR, but not any of the steroid receptors. 2 Polyclonal antibody R1AB was raised against a polypeptide encompassing amino acids 63-77 (TQSSSSEEIVPSPPS) to RAR␣. Polyclonal antibody ␥NTB (also called rNTAB) was raised against an oligopeptide corresponding to the amino acid sequence encompassing residues between 55 and 68 (STPSPATIETQSSS) of RAR␣. R1AB and rNTAB recognized RAR␣, but not ER, progesterone receptor, or glucocorticoid receptor (48).
Polyclonal antiserum to RXR was a gift from Dr. Pierre Chambon of the Université Louis Pasteur in Strasbourg, France. MAb to peroxisome proliferator-activated receptor was a gift from Dr. Michel Dauça of the Université de Nancy I in France. MAb to TR␣ and TR␤ were purchased from Affinity BioReagents, Golden, CO. MAb 9A7␥ to vitamin D receptor was a gift of Dr. Mark R. Haussler of the University of Arizona. Polyclonal antisera to COUP-TF were kindly provided by Drs. Sophia Y. Tsai of Baylor University (43) and Janet E. Mertz of the University of Wisconsin. Dr. Mertz also provided the polyclonal antiserum to hERR␣1. Polyclonal antiserum raised against a GST-hERR-1 fusion protein was a gift of Dr. Christine T. Teng of NIEHS (44). Purified polyclonal antibody to c-Jun (Ab-2) was a gift of Oncogene Science (Cambridge, MA).
Immunoprecipitation-Immunoprecipitation reactions were accomplished by incubation of partially purified ER with the designated antibody for 1 h at room temperature (RT). Aliquots (10 l) of a 50% slurry of Protein G-Sepharose or Protein A-Sepharose (Pharmacia Biotech Inc.) were added, depending on the source of primary antibody, and the incubation was continued for an additional hour at RT. The Sepharose resin was pelleted by sedimentation in a microcentrifuge and the supernatant used in gel mobility shift assay. When 3 H-liganded ER was used, the bound 3 H-ligand was extracted with scintillation fluid and counted.
SDS-Polyacrylamide Gel Electrophoresis, Fluorography, and Western Blotting-Proteins present at various stages of ER purification were analyzed on either 8 or 10% SDS-polyacrylamide gels. Protein concentrations were determined by the method of Bradford (45). Protein molecular weight standards (Mark12 and MultiMark from Novex, San Diego, CA., or Kaleidoscope standard from Bio-Rad) were electrophoresed in parallel with experimental samples in 25 mM Tris, 192 mM glycine, 0.1% SDS at 135 V for 1.5 h, and the gels were fixed in 10% acetic acid, 50% methanol for several hours. Silver staining was performed using the Bio-Rad Silver Stain Plus kit according to the manufacturer's instructions.
For Western blotting, two identical 8 or 10% SDS-PAGE mini-gels were electrophoresed as above. One gel was silver-stained and proteins in the other SDS-PAGE gel were electroblotted onto nitrocellulose (Pharmacia, Dublin, CA) or polyvinylidene difluoride (NEN Life Science Products) membranes. The transfer was monitored by transfer of the prestained protein markers and staining with 0.5% Ponceau S (Sigma) solution. Following the transfer, the membranes were incubated in a 5% Carnation nonfat dry milk Tris-buffered saline (TBS), pH 7.6, containing 0.1% Tween (Sigma) for 3 h at RT to saturate the nonspecific binding sites. In general, the membrane was incubated with a 1:1000 dilution of primary antibody in the 5% milk/TBS-Tween for 1-2 h at RT and washed three times with a large volume of TBS-Tween for a total of 30 min. The membrane was then incubated with a 1:3000 -1:10,000 dilution of secondary antibody in 5% milk/TBS-Tween. After rinsing, the interacting proteins were detected by chemiluminescence (Amersham Corp.) for horseradish peroxidase-conjugated antibodies and Western-Star (TROPIX, Bedford, MA) for alkaline phosphatase-conjugated antibodies on Reflection (DuPont) or BIOMAX ML (Kodak) film for 10 s to 30 min prior to processing.
GST Protein-Protein Interaction Assays-Plasmids directing the expression of glutathione S-transferase (GST) ER, COUP-TFI, and ERR␣1 were kindly provided by Dr. Janet E. Mertz of the University of Wisconsin. GST fusion proteins and GST expressed from pGEX-2TK were purified from E. coli BL-21 cells according to protocols supplied by Pharmacia. The concentrations of the glutathione-Sepharose-purified fusion proteins were determined by DC assay (Bio-Rad), and [ 3 H]E 2 binding to GST-ER was measured by hydroxyapatite assay (32). Proteins were monitored by separation on 10% polyacrylamide SDS-PAGE followed by Western blot with an anti-GST antibody (Pharmacia). For GST "pull-down" assays, identical amounts of purified GST fusion proteins were preincubated with 30 l of a 75% slurry of glutathione (GSH)-Sepharose in NENT buffer (20 mM Tris-HCl, pH 8.0; 100 mM NaCl, 1 mM EDTA, 6 mM MgCl 2 , 0.5% Nonidet P-40; 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride, 8% glycerol), washed twice with 1 ml of NENT and twice with 1 ml of Transcription Wash buffer (20 mM Tris-HCl, pH 8.0; 60 mM NaCl, 1 mM EDTA, 6 mM MgCl 2 , 0.05% Nonidet P-40; 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride, 8% glycerol). ER was incubated with the resulting GST fusion protein-bound GSH-Sepharose resin for 2 h at 4°C on a rotator. The resin was sedimented by centrifugation and the supernatant removed, and the pellet was rinsed three times with 1 ml of NENT. The 3 H-ligand counts were determined in the supernatant and rinses. The bound proteins were eluted from the resin by microwaving the samples in SDS loading buffer and resolved by SDS-PAGE (46). Replicate samples were run on duplicate gels. Proteins were transferred to polyvinylidene difluoride membranes, and the proteins remaining on the gel were examined by silver staining.
Southwestern Blotting-Proteins on the nitrocellulose membrane were denatured by washing the membrane in 6 M guanidine HCl in 1 ϫ binding buffer (250 mM HEPES, pH 7.9; 30 mM MgCl 2 , 40 mM KCl, 1 mM DTT) for two 5-min washes at 4°C (47). Renaturation of the proteins was achieved by two sequential washes in each of the following concentrations of guanidine HCl in 1 ϫ binding buffer: 3 M, 1.5 M, 0.75 M, 0.325 M, and 0.187 M. The membrane was rinsed twice for 30 min in 1 ϫ binding buffer alone, the same buffer containing 5% nonfat milk, and the same buffer containing 0.25% nonfat milk. The membrane was incubated overnight (14 h at 4°C) in 1 ϫ binding buffer containing 0.25% nonfat milk, 10 g/ml sonicated denatured salmon sperm DNA, and 10 6 cpm of 32 P-1/2EREc38. The membrane was rinsed in 1 ϫ binding buffer containing 0.25% nonfat milk and exposed to X-Omat AR (Kodak) film for 5-30 min prior to processing. Photoaffinity UV Cross-linking-UV cross-linking was performed in solution using heparin-agarose-purified E 2 -ER or TAz-ER and bromouridine-substituted 32 P-EREc38 or 32 P-1/2EREc38. Liganded ER was preincubated with 25,000 cpm (approximately 10 fmol) 32 P-EREc38 or 32 P-1/2EREc38 in the presence of 1.5 g of poly[d(I⅐C)] in a total reaction volume of 20 l for 1 h on ice. Tubes were opened and incubation continued on ice under a UV transilluminator (Spectroline model X-15G, Westbury, NY) at a distance of 6 cm for 0 -60 min. DNA polymerase I (Klenow fragment, 68 kDa) and T4 ligase (70 kDa) were cross-linked to 32 P-EREc38 to estimate the effect of the cross-linked probe on the migration of a protein of known size through SDS-PAGE (48). The reaction was terminated by addition 2.5 l of Laemmli buffer. The samples were boiled for 3 min, loaded onto a 10% SDS-polyacrylamide gel, and electrophoresed as described above. Each gel included pre-stained protein molecular weight standards, Kaleidoscope (Bio-Rad) and/or SeeBlue Standard (Novex). The gel was dried and exposed to X-Omat AR film at Ϫ70°C.
Cell Transfection-MCF-7 human breast cancer cells (2.5 ϫ 10 5 ) were plated in each well of a 12-well Corning plate in Iscove's modified Dulbecco's medium (all cell culture reagents were from Life Technologies, Inc.) without phenol red, supplemented with 10% stripped fetal bovine serum, and 1% penicillin-streptomycin. After 24 h, at 50% confluencey, the cells were transfected using liposome-mediated transfection (LipofectAMINE) (49). Cells were co-transfected with 0.5 g of pCMV-␤-gal (CLONTECH) and 0.6 g of pGL3-luc reporter vector per well using a DNA:liposome ratio of 1 g/10 nmol. The DNA was preincubated with LipofectAMINE in Opti-MEM I without insulin, estradiol, or other growth factors for 45 min. Four hours after transfection, 1 nM 17␤-estradiol (Sigma), 100 nM 4-OHT, or an equal volume of ethanol was added to the wells in duplicate. The cells were maintained in Iscove's modified Dulbecco's medium containing 1% stripped fetal bovine serum. The cells were lysed 24 h after transfection in 150 l of 1 ϫ reporter lysis buffer (Promega), and the cleared extract was assayed for luciferase and ␤-galactosidase activities.
␤-Galactosidase Assay-Fifty l of cell extract was assayed, in duplicate, in a total volume of 300 l containing 1 ϫ reporter lysis buffer and 2 ϫ assay buffer according to the manufacturer's protocol (Promega). Samples were incubated at 37°C for 30 min, and the reaction was stopped by addition of 500 l of 1 M sodium carbonate, and absorbance readings were measured at 420 nm. Activity was calculated from a standard curve generated from a series of dilutions of ␤-galactosidase (Promega).
Luciferase Assay-Luciferase activity was determined in 20 l of cell extract with the assay performed according to the manufacturer's protocol (Promega). Enzyme activity was measured using a luminometer (Lumat Lb 9501-0, Wallac) for 10 s per sample. Luciferase activity was expressed as relative light units and was normalized using ␤-galactosidase activity.

ER Binds EREc38 but Not an ERE Half-site in Vitro-
We examined the ability of partially purified E 2 -liganded ER (E 2 -ER) to bind to EREc38, a consensus ERE derived from three highly estrogen-responsive genes (15), versus a single half-site ERE, 1/2EREc38, which lacks the 3Ј-ERE half-site but was otherwise identical to EREc38 (Fig. 1, sequences in Table I).
Using gel mobility shift assay, we detected an ER⅐EREc38 binding complex, of which 85% was supershifted by the ERspecific antibody H222 (Fig. 1). The specificity of the ER⅐EREc38 complex was also demonstrated with displacement by excess unlabeled EREc38 in a concentration-dependent manner (data not shown).
Approximately 15% of the ER⅐ERE binding complex detected was not ER. This was indicated by the inability of H222 to shift or inhibit the appearance of this complex (Fig. 1). Similarly, none of the ER-specific antibodies tested altered the amount or appearance of this complex (42). Because the antisera used recognize epitopes spanning the entire ER protein, this result rules out the possibility that the non-supershifted complex was formed by a proteolytic ER product or a naturally occurring truncated ER variant. In contrast, ER did not bind to 1/2EREc38 (Fig. 1) or 1/ 2ERE3Јc38. Antibodies to ER did not supershift or inhibit the half-site complex, indicating it is not ER (Fig. 1, Ref. 42, and data not shown). However, a complex of similar mobility, but lower intensity, was bound to the ERE half-site. Denaturation of the ER preparation (by boiling) or treatment with 0.5 g/ml trypsin (10 min at RT) destroyed all DNA binding, indicating that a protein(s) is responsible for 1/2EREc38 binding. Because the activity responsible for this binding co-purified with ER by heparin-agarose, phenyl-Sepharose, Yellow 86, Green 19, Affi-Gel Blue, and Blue 4 affinity chromatography (data not shown), we called it estrogen receptor associated factor (ERAF). Since ERAF bound to an oligomer containing a single 5Ј-AGGTCA-3Ј sequence and ER did not bind this half-site, measuring ERAF⅐1/2EREc38 binding provided a sensitive and reproducible assay for its characterization. The relative amount of ERAF activity associated with ER did not change between bovine uterine ER preparations, with ER ligand, i.e. E 2 , 4-OHT, or TAz, or in the absence of added ligand.
Cross-linking of ERAF to 1/2EREc38 -Photoaffinity UV cross-linking was used to evaluate the molecular weight of ERAF. Heparin-agarose-purified E 2 -ER was incubated with 5-bromo-deoxyuridine-substituted 32 P-EREc38 or 32 P-1/ 2EREc38 and irradiated by UV light, and the protein-DNA adducts were resolved by SDS-PAGE. No cross-linking occurred without exposure to UV light. The amount of protein cross-linked was time-dependent. A band of approximately 146 kDa was cross-linked to 32 P-EREc38 at 20 min ( Fig. 2A). Additional bands of 234, 78, 50, and 47 kDa appeared at 30 min. UV cross-linking of highly purified TAz-ER, obtained by two sequential rounds of ERE-Sepharose purification (31), to EREc38 showed a band of 69 -77 kDa (data not shown). The specificity of the complex was demonstrated by the complete abrogation of its appearance by the addition of 25-fold molar excess unlabeled EREc38 (Fig. 2A, lane 4). We suggest that the 146-and 69 -78-kDa bands correspond to E 2 -ER dimer and monomer, respectively. The slightly larger size of the ER is consistent with reports showing that UV cross-linking of comparably sized double-stranded oligomers contributes ϳ10 kDa to the apparent size of the protein (48,50). We did not detect bands of 55 or 39 kDa that were cross-linked to the Xenopus vitellogenin A2 ERE by rat uterine E 2 -ER (51). Addition of H222 to E 2 -ER and 32 P-EREc38 increased the intensity of a distinct 146-kDa band (data not shown), corresponding in size to an ER homodimer or an ER monomer plus one light and one heavy chain of the antibody.
The size of proteins cross-linked to 32 P-1/2EREc38 was timedependent with a diffuse 78-kDa band appearing by 15 min and bands of 47 and 38 kDa at 45 min (Fig. 2B). Bands of similar sizes were obtained in 34 UV cross-linking experiments performed. Whether the 78-kDa band is identical in composition to that cross-linked to EREc38, i.e. ER monomer, is not known. Unlike EREc38, a 146-kDa complex was not cross-linked to 32 P-1/2EREc38 (Fig. 2B). H222 had no effect on the appearance of the bands cross-linked to 32 P-1/2EREc38, confirming that ER is not part of the complex. Addition of 25-fold molar excess

TABLE I
Sequences of the EREs and ␤RARE used in gel mobility shift assays These are sequences of the EREs cloned into the SmaI site of pGEM-7Zf(ϩ) and subsequently used in gel mobility shift assays as described under "Experimental Procedures." EREc38 is a 38-bp ERE consensus sequence (15). The underlined nucleotides correspond to the minimal core consensus ERE. Nucleotides in italics indicate differences from the EREc38 consensus sequence. The nucleotide in bold indicates an alteration the ERE inverted repeat (IR). ␤RARE is a synthetic version of the mouse RAR type ␤ gene (27)(28)(29)

5Ј-CCAGGTCAGAGCATTTCGAG-3Ј
AT-rich region unlabeled cognate response element abrogated the appearance of proteins cross-linked to 1/2EREc38 (data not shown), demonstrating the specificity of the complexes formed. As a control, cross-linking of DNA polymerase I Klenow fragment to EREc38 generated bands of 234, 121, and 76 kDa. Because of the multiple protein complexes cross-linked to 1/2EREc38, we postulate that ERAF consists of more than a single protein of less than 70 kDa that forms homo-and/or hetero-dimers on 1/2EREc38.

Effects of Antibodies to Nuclear Receptors on ER-EREc38 or ERAF⅐1/2EREc38
Binding-Class II nuclear receptors, e.g. RAR, TR, RXR, peroxisome proliferator-activated receptor, and vitamin D receptor, bind to the ERE half-site 5Ј-AGGTCA-3Ј with half-site spacing influencing binding specificity (52). We tested whether ERAF is a known class II nuclear receptor by examining whether selected antibodies altered ERAF⅐1/ 2EREc38 binding. Incubation of nuclear receptors with their cognate response elements and receptor-specific antibodies can result in four possible outcomes as follows: enhanced receptor-DNA binding, inhibition of receptor-DNA binding, a supershift of the receptor-DNA complex, or no effect, indicating that the epitope recognized by the antibody is not accessible under the assay conditions. None of the receptor antibodies tested bound either EREc38 or 1/2EREc38 by themselves. Antisera to RAR affected neither ER-EREc38 nor ERAF⅐1/2EREc38 binding (Fig. 3). Incubation with H222 plus RAR antisera likewise did not inhibit ERAF⅐1/2EREc38 binding (Fig. 3).
Although MAb to RAR␣ or RAR␥ and antiserum to RXR did not affect ER-EREc38 binding, these antibodies reduced ERAF⅐1/2EREc38 binding by Ͼ50% (data not shown). This effect was discrete since MAb to TR␣, TR␤, or vitamin D receptor did not affect ERAF⅐1/2EREc38 binding. MAb to peroxisome proliferator-activated receptor increased ERAF⅐1/ 2EREc38 binding by 2-fold. The reason for this result is unknown. Although Jun interacts directly with ER (9, 53), addition of a MAb to Jun did not affect ERAF⅐1/2EREc38 binding, implying that ERAF is not the AP-1 complex.

Recombinant Human COUP-TFI Binds 1/2EREc38 as an Apparent Homodimer and Migrates at the Same Position as
Bovine COUP-TF-To determine whether bovine COUP-TF had DNA binding properties similar to those of human (h) COUP-TF, a plasmid encoding recombinant hCOUP-TFI was transcribed and translated in vitro. As anticipated, [ 35 S]methionine incorporation revealed a prominent band of 47 kDa for hCOUP-TF (Fig. 5A). The hCOUP-TFI bound to both EREc38 and 1/2EREc38 in a dose-dependent manner and migrated at the same position as the bovine COUP-TF⅐1/2EREc38 complex (Fig. 5B). The specificity of COUP-TFI-DNA binding was tested by displacement, using a 10-fold excess of unlabeled cognate oligomer. These results demonstrate conclusively that COUP-TFI binds EREc38 and 1/2EREc38. The similarity in migration of the COUP-TFI⅐EREc38 or -1/2EREc38 complexes with that of homodimeric ER-EREc38, indicates that the minimal form of COUP-TFI bound to DNA is a dimer.
Determining the Specificity of COUP-TF⅐1/2EREc38 Binding-Specificity of the bovine COUP-TF⅐1/2EREc38 complex was demonstrated by competition with EREc38, a sequence variant ERE (Fig. 6, lanes 13-16), and EREc, but not by the AT-rich portion of EREc38 (Fig. 6, lanes 17-20) nor by the region of pGEM-Zf(ϩ) into which EREc38 is cloned, although it contains an imperfect half-site (see "Experimental Procedures"). No component of the partially purified ER preparation bound the AT-rich portion of the 1/2EREc38 construct (Fig. 6,  lanes 9 -12). Together, these results indicate that COUP-TF interacts specifically with the 5Ј-AGGTCA-3Ј in 1/2EREc38.
Relative Affinity of COUP-TF⅐1/2EREc38 Versus ER-EREc38 Binding-As one measure of the relative affinity of ER-EREc38 versus COUP-TF⅐1/2EREc38 binding, the stability of the complexes was measured at different KCl concentrations. We reported that maximum specific E 2 -ER-EREc38 binding occurred between 100 and 150 mM KCl (16). E 2 -ER-EREc38 binding was stable up to 500 mM KCl in gel mobility shift assays. Higher KCl concentrations drastically inhibited E 2 -ER-EREc38 binding. In contrast, COUP-TF⅐1/2EREc38 binding was considerably less stable with binding reduced to 59% at 200 mM KCl. This observation, with those shown in Figs. 1-6, indicates that the interaction between ER and EREc38 is of higher affinity than that of COUP-TF⅐1/2EREc38.
By titrating the concentration of 32 P-1/2EREc38 or 32 P-1/ 2ERE3Јc38 with a fixed concentration of partially purified E 2 -ER containing bovine COUP-TF, saturation binding experiments were performed (data not shown). Linear regression analysis of the data from Scatchard plots yielded a K d ϭ 1.24 Ϯ 0.20 nM for COUP-TF⅐1/2EREc38 and a K d ϭ 0.94 Ϯ 0.18 nM for COUP-TF⅐1/2ERE3Јc38 binding (mean Ϯ S.D. from four separate experiments performed at two different concentrations of partially purified E 2 -ER). Thus, COUP-TF binds to these halfsites with comparable high affinity and with values similar to  1 and 2) or 32 P-1/2EREc38 (lanes 3-10) was added to the reaction. In both panels, the arrow indicates the ERAF⅐1/2EREc38 complex. Free [ 32 P]DNA is indicated at the bottom of the gel. In both gels, 25,000 cpm of [ 32 P]DNA, final 10 fmol DNA/lane, were added to each reaction for a final total reaction volume of 60 l, and the incubation was continued for an additional hour at 4°C. Incubation and gel mobility shift assay conditions were identical to those described in Fig. 1. those reported for the binding of in vitro translated COUP-TFI to various retinoid Z receptor (RZR/ROR) elements (55).
Interaction of COUP-TF with ER in Vitro-COUP-TF⅐1/ 2EREc38 binding activity was inhibited by approximately 45% by immunoprecipitation of the partially purified bovine ER preparation with ER MAb H222. In contrast, all of the ER-EREc38 binding activity was removed by immunoprecipitation with H222. These results suggest that COUP-TF interacts directly with ER or with a tightly ER-associated protein. Further evidence of a direct COUP-TF-ER interaction was shown by incubating partially purified bovine COUP-TF with ERE-Sepharose-purified E 2 -ER, which does not contain COUP-TF (Fig. 3) or class II nuclear receptors (25,31,42), and demonstrating that COUP-TF⅐1/2EREc38 binding is enhanced by purified E 2 -ER in a dose-dependent manner (Fig. 7). In contrast, purified 4-OHT-ER did not enhance COUP-TF⅐1/2EREc38 binding. Purified ER did not alter the mobility of the COUP-TF⅐1/2EREc38 complex, indicating that ER did not stably bind COUP-TF⅐DNA complex.
Direct COUP-TF-ER and ERR␣1-ER Interaction in Vitro-To address whether COUP-TF interacts directly with ER, GST pull-down assays were performed. Fusion proteins bound to GSH-Sepharose were incubated with [ 3 H]TAz-ER or [ 3 H]E 2 -ER, and specifically retained proteins were eluted by denaturation, resolved by SDS-PAGE, and evaluated by Western blotting using antibodies to ER and GST. ER was retained by the GSH affinity resin in the presence of GST-COUP-TF, GST-hERR␣1, and GST-hER, but not by resin with GST or by GSH-Sepharose alone (Fig. 8). To our knowledge, this is the first demonstration of a direct protein-protein interaction between ER and COUP-TF and confirms the recent report of a direct ER-ERR␣1 interaction (54).
COUP-TF Inhibits E 2 -mediated Transcriptional Activation in Vitro-To address the biological significance of ER-COUP-TF interaction, we co-transfected ER-positive MCF-7 human breast cancer cells with an expression vector for COUP-TFI and examined E 2 -induced expression of a luciferase reporter gene under the control of three tandem copies of EREc38 (Fig. 9). E 2 induced a 24-fold induction of luciferase activity from three tandem copies of EREc38. Addition of a 100-fold excess of 4-OHT inhibited the E 2 -mediated induction by Ͼ90%. Co-expression of COUP-TFI inhibited E 2 -stimulated luciferase activity in a dose-dependent manner. Co-transfection with identical concentrations of the pCMV5, a control vector, did not inhibit E 2 -induced luciferase activity, indicating the specificity of the COUP-TF effect (data not shown). Increased expression of COUP-TF in cells co-transfected with the COUP-TF expression vector was confirmed by Western blotting of a slot blot of whole cell extracts from the treated MCF-7 cells (data not shown). DISCUSSION An interplay between various nuclear receptors competing for binding to the same target sites in gene promoters modulates hormone-responsive gene expression in mammalian cells. The specificity of transcriptional activation is conferred by the cellular levels of cognate ligands, receptors, co-activator proteins, and the chromatin structure of the target gene. We previously showed that partially purified ER binds EREs with higher affinity and stability than highly purified ER (25). In this report, we provide data implicating the COUP-TF orphan receptor as a constituent of heparin-agarose-purified ER. We show that COUP-TF, like ERR-1 (44,54), co-purifies with ER on a number of separation matrices, but COUP-TF is separated from the ER on ERE-Sepharose because it binds with lower affinity. Using GST fusion proteins in pull-down assays, we provide the first demonstration of a direct protein-protein interaction between ER and COUP-TF and confirm the recent report of direct ER-ERR␣1 interaction (54). COUP-TF, unlike ER, binds not only to the fully palindromic ERE, i.e. EREc38, but also to a single ERE half-site. This is the first report documenting COUP-TF binding as a dimer to an ERE half-site. The specificity of COUP-TF⅐1/ 2EREc38 binding was demonstrated by antibody supershift and by competition with unlabeled oligonucleotides. Bovine COUP-TF interacts specifically with a single 5Ј-AGGTCA-3Ј half-site. In contrast to our results, COUP-TFI translated in vitro was unable to bind to an ERE half-site (33,56,57). However, our results corroborate a recent report that COUP-TF homodimers bind nuclear receptor RZR/ROR response elements consisting of a single 5Ј-extended half-site (55). We note that the RZR/ROR response element half-sites in the latter report were flanked by an imperfect half-site, forming an imperfect ERE (55). The influence of this additional half-site on COUP-TF binding is unknown.
We believe that a COUP-TF dimer binds 1/2EREc38 based on the mobility of the COUP-TF⅐1/2EREc38 and ER⅐EREc38 complexes (Fig. 1), indicating a size for the COUP-TF⅐1/2ERE complex that is slightly smaller than that of the 130-kDa ER  Table I). Assay incubation and gel mobility shift assay conditions were identical to those described in Fig. 1. These results indicate the specificity of COUP-TF binding to the ERE half-site, 5Ј-AGGTCA-3Ј.

FIG. 7. Addition of purified E 2 -but not 4-OHT-ER increases COUP-TF⅐1/2EREc38 binding.
A fixed amount (20 l) of partially purified bovine COUP-TF was incubated with ERE affinity purified E 2 -ER or 4-OHT-ER (amounts are indicated) and 32 P-1/2EREc38 (10 fmol/reaction). COUP-TF⅐1/2EREc38 binding was determined by gel mobility shift assay as described in Fig. 1. The amount of shifted complex was plotted as F(t) as described under "Experimental Procedures." Proteins were transferred to polyvinylidene difluoride membranes, probed with MAb AER311 to ER, and the interacting proteins visualized by chemiluminescence. Sizes of protein markers are indicated in kDa. The experiment has been repeated three times with comparable results.
FIG. 9. E 2 -induced luciferase reporter gene activity from EREc38 is inhibited by co-expression of COUP-TFI. MCF-7 cells were co-transfected with pGL3-pro-3(EREc38) (0.6 g), pCMV-␤gal (0.5 g), and the indicated amounts of pRSV-COUP-TFI. Four hours after plating, the cells were treated with ethanol (EtOH), 1 nM E 2 , 1 nM E 2 plus 100 nM 4-OHT, or 100 nM 4-OHT as indicated. The cells were harvested 24 h after treatment, and the cell extracts were assayed for luciferase and ␤-galactosidase activities. The fold induction of luciferase activity was normalized for ␤-galactosidase and is expressed as the ratio of relative light units between treatment groups and the ethanol control. Data are the mean Ϯ S.E. from five different experiments.
homodimer. Since COUP-TFs range from 43 to 53 kDa (43,56,58), we speculate that the COUP-TF⅐1/2EREc38 complex contains a COUP-TF homo-or heterodimer or possibly a homo-or heterotrimer. COUP-TF is a dimer in solution and binds as a dimer to divergent response elements, e.g. DR0-DR12 and IR, indicating that COUP-TF assumes different conformations to accommodate structural and spatial changes in the DNA (59). The specificity of nuclear receptor response element binding is conferred by the interaction of each receptor monomer with individual half-sites of different spacing and orientation (60). However, a third level of DNA recognition was recently suggested by the cooperative binding of certain nuclear receptors, e.g. RXRs, as higher order oligomers to response elements containing highly reiterated half-sites (61). Recognition of reiterated elements was suggested to extend to other nuclear hormone receptors (59).
Additional evidence that the COUP-TF binds as a dimer comes from two Southwestern experiments. These showed that neither ER nor COUP-TF monomers are capable of binding EREc38 or 1/2EREc38. In contrast, class II nuclear receptors that bind EREs as monomers, e.g. H-2RIIBP (62), are readily detected in Southwestern blots. Our results also demonstrate that COUP-TF differs from orphan nuclear receptors, e.g. ROR, NGFI-B, FTz-F1, ERR␣1, and Rev-ErbA, that bind as monomers to 5ЈAT-rich-extended half-sites (54,63).
A possible interpretation of our observations is that COUP-TF binds 1/2EREc38 because there is a second half-site in the construct. Inspection of the nucleotide sequences flanking 1/2EREc38 in the EcoRI-BamHI fragment from pGEM-7Zf(ϩ) revealed two imperfect half-sites, 5Ј-GGTAC-3Ј and 5Ј-AGCTCG-3Ј (italicized letters indicate changes from the consensus), located 32 and 8 bp, respectively, 5Ј to the complete half-site. Importantly, COUP-TF did not bind an oligomer containing the imperfect 5Ј-GGTAC-3Ј (data not shown), nor did this oligomer compete for COUP-TF⅐1/2EREc38 binding (Fig.  6). Indeed, the intensity of COUP-TF bound to an oligomer containing only a consensus half-site plus the AT-rich region was equal to that of the COUP-TF⅐1/2EREc38 complex. Finally, although the K d values for COUP-TF-DNA interaction were not reported, COUP-TF bound with progressively lower affinity to a series of constructs as the distance between the half-sites increased (33). In contrast, bovine COUP-TF bound 1/2EREc38 with a K d ϭ 1.24 nM. These data support our conclusion that COUP-TF binds with high affinity and specificity to a single ERE half-site and preclude the possibility that a "cryptic" halfsite(s) is responsible for binding.
Whether bovine COUP-TF bound to 1/2EREc38 is a homodimer or consists of a heterodimer of COUP-TF and another protein(s) is unclear. Our results raise the latter possibility since UV cross-linking of the protein-1/2EREc38 complex generated a diffuse band centering at 78 kDa in addition to a 48-kDa band that we believe is bovine COUP-TF. Although COUP-TF heterodimerizes with RAR, TR, and RXR (59), those receptors were not detected in the COUP-TF⅐1/2EREc38 complex. COUP-TF also interacts directly with co-repressors N-CoR and SMRT (64). However, neither co-repressor protein appeared to be a constituent of the bovine COUP-TF⅐1/ 2EREc38 complex since proteins corresponding to their sizes, i.e. N-CoR ϭ 270 kDa (65) and SMRT ϭ 168 kDa (66), would be expected to slow the migration of the complex more than that detected in gel shift. Finally, bands of neither 270 nor 168 kDa were cross-linked to 1/2EREc38.
We showed that ER interacts directly with COUP-TFI and ERR␣1 in solution. The latter observation agrees with a recent report on ERR␣1-ER interaction (54). We emphasize that neither our results (18, 20 -22) nor those reported previously (67) indicate that COUP-TF heterodimerizes with ER bound to EREs nor does ER stably participate in COUP-TF⅐ERE halfsite binding. In an interesting parallel to our findings, two recent reports showed interaction of COUP-TF with orphan receptors nur77 (68) and HNF-4 (69) only in solution but not when bound to DNA.
COUP-TF is highly conserved in evolution, and "knock-out" mutations of COUP-TFI and COUP-TFII are lethal, indicating that COUP-TF performs essential functions in vivo (26,70). To examine COUP-TF interaction with ER from another estrogenresponsive tissue, we partially purified ER from MCF-7 cells (data not shown). In results virtually identical to those described for bovine ER, ER in the MCF-7 extract bound specifically to EREc38 but not to 1/2EREc38. The MCF-7 extract showed ERAF⅐1/2EREc38 binding that appeared identical to the bovine COUP-TF⅐1/2EREc38. In contrast, a similar preparation of recombinant human ER expressed in yeast (71) showed no 1/2EREc38 binding but did bind specifically to EREc38. 3 We conclude that yeast do not express a homologous protein capable of half-site binding.
The potential biological role of ER-COUP-TF interaction was suggested by the ability of COUP-TF to inhibit E 2 -induced expression of a luciferase reporter gene under the control of three tandem copies of EREc38 in MCF-7 cells. We postulate that in estrogen target tissues such as the uterus, ER is present together with other nuclear and orphan receptors, co-activators, and co-repressors that interact with DNA and perhaps, with each other to regulate gene transcription.