Ligand-, Cell-, and Estrogen Receptor Subtype ( a / b )-dependent Activation at GC-rich (Sp1) Promoter Elements*

17 b -Estradiol (E2) induces expression of several genes via estrogen receptor (ER)-Sp1 protein interactions with GC-rich promoter elements in which Sp1 but not ER binds DNA. This study reports the ligand- and cell context-dependent ER a /Sp1 and ER b /Sp1 action using an E2-responsive construct (pSp1) containing a GC-rich promoter. Both ER a and ER b proteins physically inter- act with Sp1 (coimmunoprecipitation) and preferentially bind to the C-terminal region of this protein in pull-down assays. E2- and antiestrogen-dependent transcriptional activation of ER a /Sp1 was observed in MCF-7, MDA-MB-231, and LnCaP cells, but not in HeLa cells. E2 did not affect or significantly decrease ER b /Sp1 action, and antiestrogens had minimal effects in the same 4 cell lines. Exchange of activation function-1 (AF-1) domains of ER subtypes gave chimeric ER a / b (AF- 1 a /AF-2 b ) and ER b / a (AF-1 b /AF-2 a ) proteins that resem-bled wild-type ER ( a or b ) in terms of physical associa-tion with Sp1 protein. Transcriptional activation studies with chimeric ER b / a and ER a / b showed that only ER a / b can activate transcription from an Sp1 element, not ER b / a . This indicates that the a ) interact with the transcription factor Sp1. In vitro transcribed and translated hER proteins were incubated with recombinant Sp1 protein and analyzed as described under “Experimental Procedures.” 35 S-Labeled ER a , ER b , hER a / b , and hER b / a proteins coincubated with unlabeled Sp1 protein are immunoprecipitated with Sp1 antibodies ( lanes 3, 6, 9, and 12, respectively). Over several experiments, levels of immunoprecipitated wild-type and chimeric ER were similar.

The estrogen receptor (ER) 1 is a member of the nuclear receptor superfamily of transcription factors that exhibit several common structural/functional domains (1,2). Estrogens, the endogenous ER ligands, influence gene expression and physiological responses in multiple target tissues (3,4), and mechanisms associated with selective estrogen and antiestrogen action have been extensively investigated. 17␤-Estradiol (E2)-induced transactivation in specific cell types is modulated by multiple factors, including expression of the ER, coactivator and corepressor proteins, other nuclear factors, and accessibility of responsive elements in target gene promoters (5)(6)(7)(8)(9). The recent discovery of a second form of the ER (i.e. ER ␤ ) extends the potential tissue-specific selectivity for induction of estrogenic and antiestrogenic responses (10,11).
ER ␣ and ER ␤ exhibit both differential and overlapping expression in various tissues and cell lines (10 -21). For example, in rats, ER ␤ mRNA transcripts were highly expressed in the prostate and ovary; moderate expression was observed in testis, uterus, lung, and bladder; and low ER ␤ expression was observed in the spinal cord, various brain sections, pituitary, epididymis, and thymus (10). In many of these same tissues, both ER ␤ and ER ␣ are co-expressed; however, it was evident that in tissues such as the epididymis, uterus, kidney, and adrenal, which express high levels of ER ␣ mRNA, low to nondetectable ER ␤ mRNA was detected. The functional tissuespecific differences in ER subtypes have not yet been delineated; however, estrogenic responses observed in murine neuronal cells, vascular tissue, and the male reproductive tract of ER ␣ knockout mice suggest that in these tissues, ER ␤ is active in addition to ER ␣ (22)(23)(24).
There is considerable overlap in the ligand binding and functional properties of ER ␣ and ER ␤ (12,(25)(26)(27)(28)(29)(30). Both proteins bind a diverse spectrum of ligands with similar (but not identical) relative binding affinities; ER ␣ and ER ␤ form homo-and heterodimers that bind estrogen-responsive elements (EREs) in gel mobility shift assays. ER ␣ and ER ␤ activate transcription from ERE-and AP1-containing constructs; however, the effects are ligand-, cell type-, and promoter-dependent (11, 25, 28 -34). A recent study reported that mutation of a conserved tyrosine to asparagine in the ligand binding domains of ER ␣ and ER ␤ conferred hormone-independent activation properties to both ER subtypes (31).
Recent studies in our laboratory have demonstrated physical and functional interactions between ER ␣ and the nuclear transcription factor Sp1 (35). E2 induced reporter gene activity in breast cancer cells transfected with pSp1, a construct containing a consensus GC-rich oligonucleotide insert linked to a bacterial CAT reporter gene (35). Sp1 protein plays an important role in regulation of mammalian and viral genes, and results of ongoing research have shown that E2 responsiveness of c-fos, cathepsin D, retinoic acid receptor 1␣, adenosine deaminase, E2F1, bcl-2, and insulin-like growth factor-binding protein 4 gene expression in breast cancer cells is linked to specific GC-rich promoter sequences that bind ER ␣ /Sp1 complex in which only the Sp1 protein binds DNA (35)(36)(37)(38)(39)(40)(41)(42). Therefore, we have investigated the physical and functional interactions of Sp1 and ER ␤ as a model for ER subtype-specific transactivation properties. The results demonstrate that like ER ␣ , ER ␤ physi-cally interacts with Sp1 protein; however, transactivation through GC-rich sites is cell type-, ligand-, and promoter-dependent. EXPERIMENTAL  , and T4-polynucleotide kinase were purchased from Roche Molecular Biochemicals. DME/F12 without phenol red, phosphate-buffered saline, acetyl coenzyme A, E2, 4Ј-hydroxytamoxifen, 100ϫ antibiotic/antimycotic solution were purchased from Sigma. Human ER ␣ (hER ␣ ) expression plasmid was originally supplied by Dr. Ming-jer Tsai (Baylor College of Medicine), and recloned into pcDNA3 (35); the pERE construct containing an ERE insert linked to a CAT reporter gene was also provided by Dr. Ming-jer Tsai. Sp1 protein was obtained from PanVera (Madison, WI) and plasmid preparation kits were purchased from Qiagen (Santa Clarita, CA); 40% polyacrylamide was obtained from National Diagnostics (Atlanta, GA). ICI 182,780 was provided by Dr. Alan Wakeling (Zeneca Pharmaceuticals, Macclesfield, United Kingdom). Constructs expressing GST-Sp1 (wildtype) and GST-Sp1 (variants) were kindly provided by Drs. H. Rotheneder and E. Wintersberger (43) (University of Vienna, Vienna, Austria), and a consensus GC-rich oligonucleotide was prepared by the Gene Technologies Laboratory, Texas A&M University, as described previously (35). All other chemicals and biochemicals were the highest quality available from commercial sources.
Cell Lines and Transient Transfection Assays-All cell lines were obtained from the American Type Culture Collection (Manassas, VA). MCF-7, MDA-MB-231, and HeLa cells were grown and maintained in DME/F12 (Sigma cell culture) supplemented with 5% fetal bovine serum (FBS). LnCaP cells were grown and maintained in RPMI 1640 medium (Sigma cell culture) supplemented with 10% FBS. MCF-7, MDA-MB-231, and HeLa cells were seeded onto 100 ϫ 20-mm plates in phenol-free DME/F12 media supplemented with 10% charcoal-stripped FBS. After either 24 (MCF-7, MDA-MB-231, and HeLa) or 48 h (Ln-CaP), cells were transfected by the calcium phosphate method with either 10 g of pSp1 or 10 g of pERE construct and 5 g of the appropriate hER expression plasmid. After 18 h, cells were rinsed once with phosphate-buffered saline and dosed with either Me 2 SO, E2 (10 nM), 4-OH tamoxifen (1 M), or ICI 182,780 (1 M) in either DME/F12 ϩ 2.5% charcoal-stripped FBS (MCF-7, MDA-MB-231, and HeLa) or DME/F12 ϩ 10% charcoal-stripped FBS (LnCaP) for 48 h. Cells were harvested by scraping the plates in 1 ml of phosphate-buffered saline. Cell lysates were prepared in 135 l of 0.25 M Tris-HCl, pH 7.5, by three freeze-thaw-sonication cycles. Protein concentrations from the lysates were determined by the method of Bradford using bovine serum albumin as a standard. CAT activity was determined using 0.2 mCi d-threo-[dichloroacetyl 1-14 C]chloramphenicol and 4 mM acetyl-CoA as substrates. After TLC, acetylated products were visualized and quantitated on either a Packard Instant Imager or a Betascope 603 blot analyzer. Treatment with various estrogens/antiestrogens were repeated at least three times, and results are expressed as means Ϯ S.E. and compared with the Me 2 SO control group (set at 100%) for each set of experiments.
Cloning and Oligonucleotides-Plasmid pSp1 was created by cloning a consensus Sp1 binding site into pBLTATA-CAT as described previously (35). Plasmid pERE 3 was created by cloning three consensus EREs (GGTCAnnnTGACC) separated by 10 bp into pBLTATA-CAT, 30 bp upstream of the TATA box. Plasmid pERE contains an estrogenresponsive element from the Xenopus vitellogenin A2 gene promoter (Ϫ332 to Ϫ318) (44). Plasmid pcDNA3hER ␣ was created by digesting pcDNA3 (Invitrogen) with EcoRI and inserting the hER ␣ at this site as described in Ref. 35. pcDNA3.1hER ␤ was created by first inserting a SacII site in between the NheI and PmeI sites in the polylinker of pcDNA3.1 (Invitrogen). Multiple forms of hER ␤ have been identified (11,21,45), and the form used in this study was identified by Enmark et al. (21). hER ␤ (21) was released from pCMV5hER ␤ by digesting with SacII and HindIII and subsequently ligated into the SacII and HindIII sites of pcDNA3.1. In order to make hER ␣/␤ , site directed mutagenesis (Transformer TM site-directed mutagenesis kit, CLONTECH; selection primer, 5Ј-AGA CCC AAG CTT GTT ATC GAG CTC GGA TCC ACT AG-3Ј; mutagenic primer, 5Ј-GTG TGC AAT GAC TAT GCT TCA GGA TAT CAT TAT GGA GTC TGG TCC TGT G-3Ј) was performed on pcDNA3hER ␣ , creating a silent mutation EcoRV site at nucleotide positions 874 and 877 (numbering based on a 2092-bp cDNA sequence). pcDNA3hER ␣ was then digested with BamHI and EcoRV, releasing a 650-bp fragment that contains the coding sequence for the AF-1 domain of hER ␣ along with a few coding sequences for the DNA binding domain. To prepare pcDNA3.1hER ␤ for the AF-1 domain of hER ␣ , this plasmid was cut with BamHI and NotI, blunt-ended, and religated to remove the BamHI site located downstream of the coding sequence for hER ␤ . Next, an oligo containing a BamHI site was inserted in the multiple cloning site region between NheI and SacII upstream of the coding region for hER ␤ . This plasmid was digested with BamHI and EcoRV to remove the AF-1 domain of hER ␤ . The 650-bp BamHI-EcoRV AF-1 domain of hER ␣ containing fragment from pcDNA3hER ␣ was cloned into BamHI-EcoRV digested pcDNA3.1hER ␤ , creating the plasmid pcDNA3.1hER ␣/␤ . This plasmid contains the coding sequence for the chimeric protein, hER ␣/␤ , which consists of the AF-1 domain of hER ␣ fused to the DNA binding and AF-2 domains of hER ␤ . To create the plasmid pcDNA3hER ␤/␣ , the Eco-RV-BamHI 350-bp fragment from pcDNA3.1hER ␤ (which contains coding sequence for AF-1 domain of hER ␤ ) was cloned into the EcoRV-BamHI cut vector, pcDNA3hER ␣ (this plasmid has the AF-1 domain of hER ␣ removed). The resulting plasmid when translated encodes the chimeric protein, hER ␤/␣ . Comparable chimeric proteins derived from a slightly higher molecular weight form of ER ␤ (11) have recently been described (33,34).
Creation of AF-1 Deletion Constructs-hER␣ deletion constructs HE344 (⌬3-50), HE343 (⌬3-61), HE304 (⌬3-79), HE302 (⌬3-101), and HE303 (⌬3-117) were kindly provided by Prof. Pierre Chambon (Strasbourg, France). These same deletions were created in hER ␣/␤ by overlap PCR mutagenesis. Briefly, one set of primers that span from BamHI in pcDNA3.1hER ␣/␤ to the first amino acid to be deleted (in all cases, amino acid (aa) 3) are PCR-amplified. The primer at aa-3 has an overlapping region (about 15-20 bp) that begins at the next amino acid in the deletion construct (i.e. aa 51 in ⌬3-50 hER ␣/␤ ) In a separate, simultaneous reaction, another set of primers that span from the last amino acid deleted (this primer also contains an overlap region of about 15-20 bp that begins at aa 3 and spans to upstream sequences in the vector) to the EcoRV site in pcDNA3.1hER ␣/␤ are PCR-amplified. The PCR products from both reactions were purified using Wizard PCR Prep DNA purification system (Promega, Madison, WI). Next, the two products are added together, and the overlapping region will anneal. This reaction is then PCR-amplified with the BamHI and EcoRV primers. After amplification, the resulting PCR product contains the desired mutation and is subsequently cloned into BamHI-EcoRV digested pcDNA3.1hER ␣/␤ . Construct 79 -117 hER ␣ was created by PCR amplification. The 5Ј primer (beginning at aa 79) has a BamHI overhang, and the 3Ј primer (beginning at aa 117) has an EcoRV overhang. After PCR, the 79 -117 hER ␣ DNA fragment was digested with BamHI and EcoRV. pcDNA3AhER ␣ was digested with BamHI and EcoRV to remove the wild-type AF-1 domain, and the BamHI/EcoRV 79 -117 hER ␣ DNA fragment was subsequently cloned into this vector. The clones were then sequenced to confirm that the proper deletions were introduced as well as to determine whether any unexpected mutations were introduced by Taq polymerase.
In Vitro Transcription and Translation-hER ␣ , hER ␤ , hER ␣/␤ , and hER ␤/␣ proteins were synthesized in vitro using the TNT T7 quickcoupled transcription/translation system (Promega) in the presence of 1 mM methionine (gel shift experiments). Using [ 35 S]methionine (Amersham Pharmacia Biotech) as substrate, it was shown that expression of the ER subtypes was comparable, and the corresponding 35 S-labeled proteins were utilized for GST pull-down assays as described previously (35).
Coimmunoprecipitation of Sp1/Various hER Protein Complexes-To determine whether the hER proteins (␣, ␤, ␣/␤, and ␤/␣) interact with Sp1 protein, monoclonal antibodies against Sp1 were used. Briefly, hER ␣ , hER ␤ , hER ␣/␤ , or hER ␤/␣ was in vitro transcribed and translated in the presence of [ 35 S]methionine using the T7 quick-coupled transcription/translation kit (Promega). Ten l of hER protein (␣, ␤, ␣/␤, or ␤/␣) was incubated with 10 ng of pure Sp1 protein at 4°C for 2 h. Sp1 monoclonal antibodies prebound to protein G-agarose beads (Santa Cruz Biotechnology) were added to the reaction and incubated at 4°C for an additional 6 h (overnight). Sp1/hER-protein G complexes were spun down and washed according to the manufacturer's recommendations. Proteins were subsequently separated by SDS-polyacrylamide gel electrophoresis (5% gel) and visualized by autoradiography.
GST Pull-down Assay-DH5␣ bacteria cells transformed with either plasmids for the empty GST vector, the GST-Sp1 (wild-type), or the GST-Sp1 truncation mutants were induced with 0.05 mM isopropyl-1thio-␤-D-galactopyranoside to express the appropriate proteins. Ninety min after induction, cells were collected and resuspended in sonication buffer (150 mM KCl, 40 mM HEPES (pH 7.5), 0.5 mM EDTA, 5.0 mM MgCl 2 , 1.0 mM dithiothreitol, 0.05% Nonidet P-40) supplemented with 1.0 mM phenylmethylsulfonyl fluoride and 10 g/l aprotinin. Cells were lysed by sonication, and crude bacterial extracts containing the GST fusion proteins were incubated with glutathione-Sepharose beads. The glutathione-Sepharose beads bound with GST fusion proteins were incubated with in vitro translated [ 35  Electrophoretic Mobility Shift Assays-Electrophoretic mobility shift assays were used to determine interactions of in vitro translated hER ␣ , hER ␤ , or hER ␣/␤ with Sp1. Additionally, gel shift analysis was used to characterize hER ␣/␤ binding to an ERE. One l of in vitro translated protein or 1 l of unprogrammed lysate, 1 mg of poly[d(I-C)], and 2.5 ng of pure Sp1 protein (Promega) (if required) were incubated at 25°C in 1ϫ binding buffer (6% glycerol, 1 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM dithiothreitol, 50 mM NaCl, 10 mM Tris-HCl (pH 8.0) 0.1 mg/ml of bovine serum albumin, and 0.125 mg/ml poly[d(I-C)]) for 15 min. Either radiolabeled consensus Sp1 oligonucleotide, wild-type ERE, or mutant ERE (50,000 cpm) was added to the reaction and incubated at 25°C for 30 min. Samples were loaded onto a 4% polyacrylamide gel electrophoresis and run at 110 V in 0.09 M Tris, 0.09 M borate, 2 mM EDTA (pH 8.3). Protein-DNA binding was visualized by autoradiography and quantitated by densitometry using the Molecular Dynamics Zero-D software package (Molecular Dynamics, Sunnyvale, CA) and a Sharp JX-330 scanner (Mahwah, NJ).
Statistical Analysis-Statistical significance was determined by analysis of variance and Scheffe's test, and the levels of probability are noted. Results are expressed as means Ϯ S.E. for at least three separate (replicate) experiments for each treatment group.
Ligand-and Cell Type-dependent Activation by ER ␣ /Sp1 and ER ␤ /Sp1-After determining that both ER ␣ and ER ␤ physically associate with Sp1, we then investigated the comparative activities of both proteins in transfection studies. Celltype and ligand-dependent ER/Sp1-mediated transactivation were investigated in breast, cervical, endometrial, and prostate cancer cells cotransfected with ER ␤ or ER ␣ expression plasmids plus pSp1 construct, which contains a consensus Sp1 oligonucleotide insert linked to a TATA promoter and a CAT reporter gene (35). Parallel studies showed that E2 activated gene expression in the same cell lines transiently co-transfected with ER ␣ or ER ␤ expression plasmids and an E2-responsive pERE construct (data not shown). These results were comparable to those previously reported (25)(26)(27)(28)(29)(32)(33)(34)(35). In ER ␣ -positive MCF-7 human breast cancer cells cotransfected with ER ␣ plus pSp1, E2 and the antiestrogens 4Ј-hydroxytamoxifen and ICI 182,780 induced reporter gene activity (Fig. 2). As a control, MCF-7 cells were cotransfected with the pcDNA3 vector and pSp1, and no ligand-induced effects were observed (data not shown); this is due, in part, to overexpression of the plasmid in transfected cells and limited endogenous expression of the ER. The requirement for cotransfecting ER in ER-positive MCF-7 has previously been reported for other E2-responsive constructs containing promoter inserts from the cathepsin D, fos, myc, pS2 progesterone receptor, and heat shock protein 27 genes (35-42, 46 -48).
In cells cotransfected with ER ␤ , only the antiestrogens 4Јhydroxytamoxifen and ICI 182,780 were weak ER ␤ agonists in MCF-7 cells, whereas E2 was inactive (Fig. 2). Ligand-dependent ER ␣ /Sp1 activation in ER-negative MDA-MB-231 human breast cancer cells showed that E2, 4Ј-hydroxytamoxifen, and ICI 182,780 were agonists, whereas ER ␤ /Sp1 reporter gene activation was not observed for estrogens or antiestrogens. The GST-Sp1 wild-type (wt) and deletion constructs of GST-Sp1 were incubated with [ 35 S]hER ␤ and analyzed as described previously. hER ␤ interacts with wild-type GST-Sp1 but not control GST (lanes 6 and 5, respectively). Also, hER ␤ interacts with the C-terminal portion of Sp1 (lanes 8 and 9, respectively) but not the N-terminal portion of Sp1 (lane 7). Neither GST nor GST-wt Sp1 interact with any proteins in the unprogrammed lysate (lanes 2 and 3, respectively). Molecular weights are given on the outside of the lane; UPL, unprogrammed lysate. B, hER ␤ enhances Sp1 binding to a 32 P-labeled consensus Sp1 oligonucleotide. As a control, two different amounts of hER␣ were incubated with Sp1 protein and addition of hER ␣ enhanced formation of the Sp1-[ 32 P]Sp1 complex as compared with UPL (compare lanes 2 and 3 with lane 1). Also, equivalent amounts of hER ␤ were incubated with Sp1 protein and a consensus [ 32 P]Sp1 oligonucleotide, and this also enhanced Sp1 binding to the [ 32 P]Sp1 as compared with UPL (compare lanes 4 and 5 with lane 1). complexity of ER ␤ /Sp1 versus ER ␣ /Sp1 ligand-dependent reporter gene activation is illustrated by results obtained in HeLa cells (cervical) and androgen-responsive LnCaP (prostate). In HeLa cells, E2, 4Ј-hydroxytamoxifen and ICI 182,780 did not activate ER ␣ /Sp1 or ER ␤ /Sp1; moreover, all three ligands significantly decreased ER ␤ /Sp1-dependent CAT activity compared with controls (Fig. 2). In contrast, the overall pattern of ligand-dependent induction via ER ␤ /Sp1 and ER ␣ /Sp1 in LnCaP cells was similar to that observed in MCF-7 cells (Fig.  2); E2, 4Ј-hydroxytamoxifen, and ICI 182,780 were ER ␣ agonists, whereas only 4Ј-hydroxytamoxifen (minimal) and ICI 182,780, but not E2, were ER ␤ /Sp1 agonists. In summary, estrogens and some antiestrogens activate ER ␣ /Sp1 in MCF-7, LnCaP, and MDA-MB-231, whereas these same ligands induce minimal to nondetectable ER ␤ /Sp1-mediated responses. In contrast, neither estrogens nor antiestrogens activated ER ␣ /Sp1 or ER ␤ /Sp1 in HeLa cells.
Repression of ER ␣ /Sp1-mediated Transactivation by ER ␤ -With the exception of results obtained in HeLa cells, estrogens activated ER ␣ /Sp1-mediated gene expression, whereas ER ␤ / Sp1 action was minimally induced or significantly decreased in all four cell lines (Fig. 2). Data summarized in Fig. 3A show that ER ␣ /Sp1-mediated transcriptional activation by E2 in MDA-MB-231 cells was significantly repressed after cotransfection with ER ␤ expression, whereas the failure of E2 to induce reporter gene activity in cells transfected with ER ␤ was overcome by adding increasing amounts of ER ␣ expression plasmid (Fig. 3A). Thus, relative amounts of ER subtype determined the activity of E2, suggesting that hormone-mediated transactivation of genes through ER/Sp1 interaction with GCrich sites may be highly susceptible to relative expression of ER ␣ or ER ␤ in various tissues/cell-types.
Expression of Chimeric ER ␣/␤ and ER ␤/␣ Proteins and Their Interactions with Sp1-ER ␣ /Sp1-mediated transactivation requires both activation function-1 (AF-1) and AF-2 but not the DNA binding domain of the ER (35). The ligand binding properties of ER ␣ and ER ␤ are similar for the ligands used in this study (12), and there is 58 and 96% amino acid sequence identity between domains E and C, which contain AF-2 (and ligand binding activity), and DNA binding function, respectively (27). In contrast, there is Ͻ20% sequence identity between the N-terminal regions of ER ␣ and ER ␤ , which contain AF-1 (21). Therefore, it is possible that differences in transactivation between ER ␣ /Sp1 and ER ␤ /Sp1 at GC-rich sites may be related to structural and functional differences between AF-1 domains of ER ␣ and ER ␤ . We therefore constructed two chimeric proteins for further study; namely ER ␤/␣ , containing the AF-1 domain of ER ␤ and the ligand and DNA binding domain (AF-2) of ER ␣ , and ER ␣/␤ , containing the AF2 and DNA binding domains of ER ␤ and the AF-1 domain of ER ␣ . Two recent papers utilized a similar approach in constructing ER ␤/␣ and ER ␣/␤ chimeras (33,34). Results summarized in Fig. 4A illustrate the molecular structure of ER ␣ , ER ␤ , and chimeric proteins and the electrophoretic separation of the 35 S-labeled proteins obtained after in vitro expression (Fig. 4B). Results summarized in Fig. 4C show that ER ␣ , ER ␤ , ER ␣/␤ , and ER ␤/␣ specifically bind to [ 32 P]ERE in gel mobility shift assays ( lanes  2-4, 6, and 8). Comparative mobilities of the retarded band are consistent with formation of homodimeric ER-DNA complexes for wild-type and chimeric proteins. Moreover, Sp1 antibodies immunoprecipitated all four [ 35 S]ER proteins after coincubation with unlabeled Sp1 protein (Fig. 4D), and these results were similar to previous reports showing physical interactions between ER ␣ and Sp1 proteins (35). GST-Sp1 pull-down assays using [ 35 S]ER ␣/␤ and [ 35 S]ER ␤/␣ also gave results similar to those observed for wild-type ER ␣ and ER ␤ proteins ( Fig. 1) with both chimeric proteins preferentially binding the C-terminal domain of Sp1 (data not shown). Thus, both chimeric ER ␣/␤ and ER ␤/␣ physically interact with Sp1 protein, and therefore, differences in their functional (transactivation) properties cannot be related to their failure to bind Sp1 protein.
Role of AF-1 and AF-2 Domains of ER ␣ and ER ␤ in ER/Sp1mediated Transcriptional Activation-The functional activity of wild-type and chimeric ER proteins was recently reported in HepG2 cells and transiently transfected with C3-luc, an E2responsive construct containing an insert from the complement C3 gene promoter (34). In HepG2 cells, E2 induced luciferase activity and, in this study, E2 also induced luciferase activity in HeLa cells transiently transfected with pC3-luc and ER ␣ , ER ␤ , ER ␣/␤ , or ER ␤/␣ (data not shown), thus confirming E2 responsiveness of these ER plasmids with pC3-luc. Differences in E2-dependent ER ␤ /Sp1 and ER ␣ /Sp1 transactivation through GC-rich sites were primarily seen in MCF-7, MDA-MB-231, and LnCaP cells, in which induced reporter gene activity was only observed in cells transiently transfected with ER ␣ but not ER ␤ (Fig. 2). Therefore, these cell lines were utilized as models to determine the role of AF-1 and AF-2 domains of ER ␣ and ER ␤ in determining ER/Sp1-mediated responses. The results summarized in Fig. 5 show that E2-dependent transactivation via Sp1 interactions at a GC-rich site was observed for both ER ␣ and chimeric ER ␣/␤ , but not ER ␤ or ER ␤/␣ , in all three cell lines examined. The comparable activities of ER ␣ and ER ␣/␤ observed in these cells suggest that the N-terminal AF-1 domain in ER ␣ is primarily responsible for E2-mediated transactivation at a GC-rich site, and this is observed using the AF-2 domain from either ER ␣ or ER ␤ . ER ␤ /Sp1 and ER ␤/␣ /Sp1 did not respond to hormone, suggesting that the AF-1 domain of ER ␤ is primarily responsible for failure to observe ER ␤ /Sp1 transactivation.
Region of AF-1 Domain of ER ␣ Responsible for Sp1 Transactivation-In order to further define the region in the AF-1  (aa 191-595). B, in vitro transcription and translation was performed on all four hER subtypes (␣, ␤, ␣/␤, and ␤/␣) to confirm the predicted size and relative expression of each of the proteins. The observed molecular mass of hER ␣ (67 kDa), hER ␤ (52 kDa), hER ␣/␤ (65 kDa), and hER ␤/␣ (60 kDa) after SDS-polyacrylamide gel electrophoresis was consistent with calculated values for these proteins. C, in vitro gel shift assay confirming that hER ␣/␤ and hER ␤/␣ bind to a palindromic [ 32 P]ERE. In vitro transcribed and translated hER ␣/␤ and hER ␤/␣ were incubated with [ 32 P]ERE and analyzed on a 5% acrylamide gel as described under "Experimental Procedures." Both hER ␣/␤ (lane 4) and hER ␤/␣ (lane 8) form a specific retarded [ 32 P]ERE band, as do controls for hER ␣ and hER ␤ . Reticulocyte lysate alone did not form a retarded band (data not shown). D, coimmunoprecipitation assay showing that all four hER proteins (␣, ␤, ␣/␤, and ␤/␣) interact with the transcription factor Sp1. In vitro transcribed and translated hER proteins were incubated with recombinant Sp1 protein and analyzed as described under "Experimental Procedures." 35 S-Labeled ER ␣ , ER ␤ , hER ␣/␤ , and hER ␤/␣ proteins coincubated with unlabeled Sp1 protein are immunoprecipitated with Sp1 antibodies (lanes 3, 6, 9, and 12, respectively). Over several experiments, levels of immunoprecipitated wild-type and chimeric ER were similar. domain responsible for induction on an Sp1 element, several deletion constructs were made in both ER ␣ and ER ␣/␤ . These constructs were cotransfected into MDA-MB-231 cells with pSp1 or pERE 3 (contains three consensus EREs), and the results are shown in Fig. 6. For ER ␣ activation on an Sp1, removal of amino acids 3-50 did not affect transcriptional activation, and deletions up to amino acid 79 retained about 50% of wild-type induction. Deletion to amino acid 101 resulted in only a 20% retention of the activity observed for wild-type ER ␣ (Fig.  6A). The importance of aa 79 -117 for ER ␣ /Sp1 action was confirmed in transactivation assays using the chimeric protein containing this amino acid sequence fused to the DBD-AF2 (domain C-F) region of ER ␣ (Fig. 6A). A relatively high response (53%) was observed using pSp1, whereas only a minimal response (3%) was observed with pERE 3 . For the corresponding Sp1-dependent ER ␣/␤ activation, removal of amino acids 3-50 retained about 70% of wild-type ER ␣/␤ -induced activity, and this response remained until deletion of aa 3-117. Thus, similar, but not identical, results were obtained for AF-1 deletion constructs of ER ␣ and ER ␣/␤ . Loss of hormone-dependent ER/Sp1 action with these deletion constructs was not associated with loss of Sp1 binding because AF-1 domain-deleted variants of both ER subtypes coimmunoprecipitate with Sp1 protein (data not shown).
Our results for ER ␣ deletion constructs on a pERE 3 indicate transcriptional activation similar to wild-type ER ␣ until all 180 amino acids are removed (Fig. 6A). These data overlapped with previous studies that indicated considerable cell context-dependent differences in transactivation using these same deletion constructs (49). In contrast, AF-1 deletion constructs of ER ␣/␤ gave a different pattern of induction than the ER ␣ -AF-1 deletion mutants, and these differences may be associated with the ER ␤ C-terminal region of the chimeric ER ␣/␤ protein. DISCUSSION Transcriptional regulation of genes by members of the nuclear receptor superfamily is highly complex and involves direct or indirect interactions with multiple nuclear proteins, including TATA binding-associated factors, coactivators, corepressors, modulators of histone acetylation, and integrating factors, such as p300 (1)(2)(3)(4)(5)(6)(7)(8). The complexity is increased for ligand-activated members of this superfamily, such as the ER, because transactivation is also dependent on ER subtype (ER ␣ or ER ␤ ) and ligand structure, which can affect receptor conformation and subsequent interaction with other nuclear factors (50,51). Interactions of ER ␣ and ER ␤ homo-and heterodimers with EREs have been extensively investigated and used for developing models of ER-mediated transactivation (11, 25, 28 -33). However, other reports and more recent studies by Kushner and co-workers (30,52,56) and others (53-55) have demonstrated ligand-dependent activation of ER in which the receptor does not directly bind promoter DNA but transduces the signal through ER-AP1 (protein-protein) interactions. Studies in this laboratory have identified protein-protein interactions of ER ␣ /Sp1 with specific GC-rich promoter elements that are required for transcriptional activation of several genes by E2, and these include cathepsin D, adenosine deaminase, c-fos, retinoic acid receptor ␣1, insulin-like growth factor-binding protein 4, bcl-2, and E2F1 (35)(36)(37)(38)(39)(40)(41)(42).
Results for ER ␣ /Sp1 were primarily obtained in E2-responsive MCF-7 breast cancer cells; therefore, the initial objectives of this study were focused on investigating ligand and cell context-specific differences between ER ␣ /Sp1 versus ER ␤ /Sp1 action using a construct containing a consensus GC-rich promoter insert. Physical interactions of ER ␣ and ER ␤ with Sp1 protein and GC-rich oligonucleotides were indistinguishable; both hormone receptor subtypes physically interact with Sp1 protein (coimmunoprecipitation and pull-down assays), preferentially bind the C-terminal DNA binding domain of Sp1, and enhance Sp1-DNA interactions with a consensus GC-rich oligonucleotide in gel mobility shift assays. The failure to observe ER ␣ or ER ␤ /Sp1-DNA ternary complexes in the latter assay is not uncommon for interacting nuclear proteins, and we have also shown that ER ␣ does not supershift an AP1-DNA complex (data not shown). These data are similar to the results of other studies showing that cyclin D1, sterol regulatory element-binding protein, and human T-cell leukemia virus type-1 Tax enhanced binding of ER, Sp1, and bZIP to their cognate sequences without forming ternary complexes (57)(58)(59). Interestingly, the physical interactions between both ER subtypes and Sp1 are not sufficient to induce transcription. Although ER ␤ physically interacts with Sp1, this interaction is not functional, whereas for ER ␣ , there exists both a physical and functional interaction with Sp1 (35) (Figs. 1 and 2). This result suggests that another event(s) must occur in order to facilitate E2-mediated induction on an Sp1 element via the ER.
During the course of these studies, both chimeric ER ␣/␤ and ER ␤/␣ proteins were described by other groups (33,34). Their studies indicated that E2 induced reporter gene activity in Chinese hamster ovary and HEC-1 cells transiently transfected with an ERE-dependent construct and wild-type or chimeric (ER ␣/␤ and ER ␤/␣ ) ER expression plasmids (33). Both wild-type and chimeric ER expression plasmids were also tested in the four cell lines used in this study, and E2 induced reporter gene activity (data not shown) as described previously (33). Like wild-type ER ␣ and ER ␤ , both chimeric proteins physically interacted with Sp1 protein and preferentially bound the C-terminal region of Sp1 (data not shown); however, results of functional studies in MDA-MB-231, LnCaP and MCF-7 cells demonstrated that only ER ␣/␤ /Sp1 mediated transcriptional activation by E2, not ER ␤/␣ /Sp1. Similar results were recently reported in HepG2 cells using a GC-rich construct from the retinoic acid receptor ␣1 gene promoter cotransfected with wild-type and ER chimeras (34), and these data are consistent with our previous study showing that E2 responsiveness of this region of the promoter is associated with ER ␣ /Sp1 interactions with three GC-rich motifs (37). From these data, we conclude that the differences in ER ␣ /Sp1 versus ER ␤ /Sp1 action are associated with their respective AF1 domains and that only AF1 of ER ␣ mediates functional interaction of the ER/Sp1 complex with other nuclear proteins and downstream basal transcription factors.
AF1 deletion mutants of ER ␣ and ER ␣/␤ were used to further define the region within this domain required for ER ␣ /Sp1 action in MDA-MB-231 cells, and the results are compared with a control pERE 3 construct containing three tandem EREs.
Stepwise deletion of ER ␣ resulted in decreased activation through the ERE as previously reported in other cell lines (49), whereas AF-1 domain deletions gave different patterns of transactivation compared with ER ␣ for an ERE-dependent promoter, and this may be related to differential interactions between the AF-1 and AF-2 domains of the chimeric protein.
The pattern of Sp1-dependent activation by deletion mutants of ER ␣ and ER ␣/␤ was also slightly different. Both ER ␣ and ER ␣/␤ are still functional when the first 79 amino acids are deleted. For ER ␣ when the first 101 amino acids are deleted, transcriptional induction by E2 is lost; however, for ER ␣/␤ , transcriptional activation was not lost until the first 117 amino acid were deleted. This slightly different pattern for ER ␣ and ER ␣/␤ may be attributed to different conformational changes of the The region deleted is denoted by a line, and the deleted amino acids are indicated. CAT activity induced by E2 (relative to Me 2 SO) is set at 100%, and fold induction is indicated in parentheses. The transcriptional activity for each of the deletion constructs is represented as a percentage of wild-type ER ␣ induction, and significant (p Ͻ 0.05) induction by E2 is noted with an asterisk (*). DBD, DNA binding domain; LBD, ligand binding domain. Each value is an average (Ϯ10%) of a triplicate experiment. B, transcriptional activity of the ER ␣/␤ deletion series on pSp1 and pERE 3 . The amino acids of the AF-1 domain (A/B) of ER ␣ are represented in gray, and the remaining portion of ER ␤ is represented in black. The region deleted is denoted by a line, and the deleted amino acids are indicated. The transcriptional activity for each of the deletion constructs is represented as a percentage of wild-type ER ␣/␤ induction as described above and a significant induction by E2 is noted with an asterisk (*). Each value is an average (Ϯ10%) of a triplicate experiment.
AF-1 domain of ER ␣ , which are dependent on the remaining ER context (i.e. ER ␣ or ER ␤ ). The importance of aa 79 -117 in ER ␣ /Sp1 action was confirmed with the ER ␣ chimeric protein (79 -117 ER ␣ ), in which the AF-1 domain was replaced only with aa 79 -117 from this same domain. The 79 -117 ER ␣ protein was active in cells transfected with pSp1 but not in cells transfected with pERE (Fig. 6A). This result indicates that the amino acids 79 -117 of ER ␣ are specific for ER ␣ /Sp1-mediated transactivation, suggesting that this region may be important for interactions with other proteins (e.g. coactivators) specific to the ER ␣ /Sp1 mechanism. Fig. 7 illustrates a possible model to explain differences between ER ␣ /Sp1 and ER ␤ /Sp1. We have shown that both ER ␣ and ER ␤ physically associate with Sp1 in the absence of ligand, but upon binding of ligand, only ER ␣ will transactivate from an Sp1 element. We hypothesize that ER ␣ undergoes a ligandinduced conformational change that allows the AF-1 domain of ER ␣ or Sp1 protein to associate with positive regulators (possibly an AF-1 specific coactivator) that result in transactivation. In contrast, upon binding E2, the AF-1 domain of ER ␤ or possibly the Sp1 protein is unable to associate with positive regulators, and induction is not observed. An alternate hypothesis might also involve corepressors that are released from ER ␣ (but not ER ␤ ) upon binding of ligand.
Endoh et al. (60) recently demonstrated that p68 RNA helicase acted as an AF1-dependent transcriptional coactivator of ER ␣ , but not ER ␤ , in COS-1 cells transiently transfected with an ERE-dependent construct. Moreover, they also showed that the central region (aa 56 -127) of the ER ␣ -A/B domain is required for both physical and functional interactions with p68 and that this region overlaps with the region in ER ␣ (aa 79 -117) required for ER ␣ /Sp1-mediated transactivation (Fig. 6). However, results of preliminary studies in breast cancer cells showed that p68 and representative classes of p160 coactivators did not enhance ER ␣ /Sp1 action, 2 and current studies on identification of other AF-1-dependent coactivators are in progress. Results from our studies add to the increasingly complex array of factors that influence estrogen-induced responses in mammalian cells but are consistent with the multiple temporal-and tissue-specific effects regulated by this hormone. FIG. 7. Hypothetical model to explain ER ␣ /Sp1 and ER ␤ /Sp1 differences. In the step 1, Sp1 associates with both ER ␣ and ER ␤ and bind DNA at GCrich elements. Upon binding of ligand (step 2), the AF-1 domain (A/B) of hER ␣ undergoes a conformational change (denoted by transformation in shape from a rectangle to a triangle) required for interactions of ER ␣ , Sp1, or both with coactivators or general transcription factors (GTFs) and for induction through the GCrich element. In contrast, ligand binding to ER ␤ does not result in formation of a transcriptionally active complex.