Glucocorticoid repression of the mouse gonadotropin-releasing hormone gene is mediated by promoter elements that are recognized by heteromeric complexes containing glucocorticoid receptor.

We have identified two regions of the mouse gonadotropin-releasing hormone (GnRH) promoter, one between −237 and −201 (distal element) and the other between −184 and −150 (proximal element), which are required for glucocorticoid repression in transiently transfected GT1-7 cells. These sequences show no similarity to known positive or negative glucocorticoid response elements (nGREs) and do not function when placed upstream of heterologous viral promoters. The glucocorticoid receptor (GR) does not bind directly to the distal or proximal promoter elements but may participate in glucocorticoid repression of GnRH gene transcription by virtue of its association within multiprotein complexes at these nGREs. Electrophoretic mobility shift assays with GT1-7 nuclear extract demonstrate the presence of GR-containing protein complexes on GnRH nGREs. One protein that co-occupies the distal nGRE in vitro along with GR is the POU domain transcription factor Oct-1. Thus, the tethering of GR to the GnRH distal nGRE, by virtue of a direct or indirect association with DNA-bound Oct-1, could play a role in hormone-dependent transcriptional repression of the GnRH gene. In contrast, Oct-1 does not appear to be a component of the GR-containing protein complex that is bound to the proximal nGRE.

We have identified two regions of the mouse gonadotropin-releasing hormone (GnRH) promoter, one between ؊237 and ؊201 (distal element) and the other between ؊184 and ؊150 (proximal element), which are required for glucocorticoid repression in transiently transfected GT1-7 cells. These sequences show no similarity to known positive or negative glucocorticoid response elements (nGREs) and do not function when placed upstream of heterologous viral promoters. The glucocorticoid receptor (GR) does not bind directly to the distal or proximal promoter elements but may participate in glucocorticoid repression of GnRH gene transcription by virtue of its association within multiprotein complexes at these nGREs. Electrophoretic mobility shift assays with GT1-7 nuclear extract demonstrate the presence of GR-containing protein complexes on GnRH nGREs. One protein that co-occupies the distal nGRE in vitro along with GR is the POU domain transcription factor Oct-1. Thus, the tethering of GR to the GnRH distal nGRE, by virtue of a direct or indirect association with DNA-bound Oct-1, could play a role in hormone-dependent transcriptional repression of the GnRH gene. In contrast, Oct-1 does not appear to be a component of the GR-containing protein complex that is bound to the proximal nGRE.
The hypothalamic decapeptide, gonadotropin-releasing hormone (GnRH), 1 is at the top of the endocrine hierarchy that controls reproductive function. GnRH is secreted from widely dispersed neurons along the rostral hypothalamus and basal forebrain into the pituitary portal vessels to control release of the gonadotropins. The sparsity and distribution of GnRH neurons have rendered difficult in vivo cellular and molecular studies aimed at elucidating the regulation of GnRH gene expression. However, such studies have become possible in vitro due to the development of immortalized, GnRH-secreting hypothalamic cell lines (GT1 cells) from a genetically engineered brain tumor in a transgenic mouse (1). These cell lines (GT1-1, GT1-3, and GT1-7) have been used to examine the regulation of GnRH secretion and gene expression by a variety of neurotransmitters (2)(3)(4)(5), steroid hormones (6,7), and second messengers involved in intracellular signaling (8 -11).
One potential application of the GT1 cell lines, of particular interest to our laboratories, concerns the putative role of adrenal steroid hormones in regulation of the hypothalamic-pituitary-gonadal axis. Both stress-related reproductive disorders and high cortisol levels in women have been associated with reductions in GnRH and luteinizing hormone (12)(13)(14). Although glucocorticoids can exert direct effects on the pituitary to suppress gonadotropin secretion (15), results obtained in castrated male rhesus monkeys implicate an action at the hypothalamic level to inhibit GnRH synthesis and/or release (16). Studies in immature female rats also support the notion that the effects of glucocorticoids on the hypothalamic-pituitary-gonadal axis may be exerted at a hypothalamic site (17). At the molecular level, responses to glucocorticoid hormones are mediated by the glucocorticoid receptor (GR) protein, a member of the steroid/thyroid hormone superfamily of nuclear receptors (18 -20). Recently Ahima and Harlan (21) demonstrated that a subset of GnRH neurons in the rat brain contain immunoreactive GR, raising the possibility that genomic effects of GR within these neurons could directly influence GnRH gene expression and secretion and thus the hypothalamic-pituitary-gonadal axis. Using GT1-3 and GT1-7 cells as model systems, we carried this analysis to the molecular level by demonstrating that these cells contain functional GR and that the potent glucocorticoid agonist, dexamethasone, inhibited the transcriptional activity of both the endogenous mouse GnRH gene and transfected reporter genes under the control of the rat GnRH promoter (7). In the present study, we have mapped two negative glucocorticoid responsive elements (nGREs) within the mouse GnRH promoter and identified a novel mechanism for GR-mediated repression, which involves the tethering of GR to nGREs via an association with other DNA-bound transcription factors.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-DNA fragments containing mouse GnRH 5Јflanking sequences extending to the first exon, Ϫ3446 to ϩ24 and Ϫ552 to ϩ24, were subcloned into the promoterless luciferase vector pXP2 between BamHI and XhoI sites. A series of 5Ј-deletion fragments, between approximately 500 and 100 base pairs (bp) in length, were generated by mung bean nuclease/exonuclease III digestion and linked to pXP2 between BamHI and XhoI sites. All constructs were confirmed by DNA sequencing. Maintenance of the appropriate transcription initiation sites in the 5Ј-deletion constructs was verified by reverse transcription and reverse transcription-polymerase chain reaction (data not shown).
Cell Culture and Transfections-GT1-7 cells were generously provided by Drs. Richard Weiner (University of California, San Francisco, CA) and Pamela Mellon (University of California, San Diego, CA). GT1-7 cells were grown in monolayer culture in Eagle's minimum essential medium (Formula 78 -5470EF purchased from Life Technologies, Inc.) containing glucose (4.5 g/liter), NaHCO 3 (2.2 g/liter), and Hepes (5.96 g/liter), and supplemented with 10% fetal bovine serum. For transfections, cells were plated in 6-cm Petri dishes at 1-3 ϫ 10 6 cells/dish and used at 50 -80% confluency (generally within 1 or 2 days). Approximately 2 h prior to transfection, the medium was withdrawn, and fresh medium (Eagle's minimum essential medium ϩ 10% dextrancoated charcoal-stripped fetal bovine serum) was added. Cultures were maintained in medium containing 10% dextran-coated charcoalstripped-stripped fetal bovine serum until harvesting. Transfections involved 32-54 dishes so that all experimental and control conditions could be compared in the same experiment. Cells were transfected by the calcium phosphate procedure using reagents from Life Technologies, Inc. as described previously (7,22). For each plasmid to be transfected, a sufficient quantity of calcium phosphate-DNA precipitate was prepared so that the same precipitate could be dispensed to all control and hormone-treated dishes in triplicate. In this way, luciferase activity in control and experimental cultures could be compared directly without the need to correct for transfection efficiency with an additional reporter gene. After 4 h, cells were washed and incubated with fresh medium (control medium or medium containing 10 Ϫ6 M dexamethasone (Steraloids, Wilton, NH) for 20 h unless otherwise indicated. Cell lysates were prepared in 25 mM Tris-phosphate, pH 7.8, 2 mM DTT, 2 mM 1,2-diaminocyclohexane-N,N,NЈ, NЈ-tetraacetic acid, 10% glycerol, 1% Triton X-100 (Promega Corp., Madison, WI) and analyzed for protein content (23). Luciferase activity was determined in aliquots containing equivalent total protein (usually 40 g) by the procedure of deWet et al. (24) using a Monolight 2010 luminometer (Analytical Luminescence Laboratory, San Diego, CA). For comparison of the basal transcriptional activity of the series of 5Ј-flanking deletion constructs, a cytomegalovirus-␤-galactosidase reporter plasmid was co-transfected with the GnRH-luciferase plasmids. ␤-galactosidase activity was measured using the CPRG (chlorophenol red-␤-D-galactopyranoside) substrate (25), and luciferase activities were normalized for differences in transfection efficiency based on ␤-galactosidase activities.
Statistical Analyses-Transfections of each DNA construct were performed in triplicate. For comparison of different experiments, luciferase activities from dexamethasone-treated cultures were converted to a percentage of those of control cultures. Differences in dexamethasone suppressibility of the various 5Ј-deleted GnRH promoter constructs were tested by one-way analysis of variance followed by t test using the BMDP statistical software package (University of California Press, Berkeley, CA).
Antibodies-The anti-GR antibody, BuGR2 (26), was a gift from Dr. Robert Harrison (University of Rochester Medical School, Rochester, NY). The anti-chicken PR antibody, PR22 (27) was a gift from Dr. David Toft (Mayo Clinic, Rochester, MN). The antibody to Oct-1 was purchased from Santa Cruz Biologicals (Santa Cruz, CA).
Nuclear Extracts-Nuclear extracts were prepared by a modification of the rapid nuclear extraction protocol (28). Cells were grown to 80% confluence in 10-cm dishes and fed with fresh media. The next day, the cells were treated with 10 Ϫ7 M dexamethasone for 6 h, washed once with phosphate-buffered saline, and harvested in phosphate-buffered saline. The resulting cell pellet was resuspended in 400 l of ice-cold Buffer A, which consists of 10 mM Hepes, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM DTT, 0.5 mM of the protease inhibitor phenylmethylsulfonyl fluoride, and 1 g/ml of the protease inhibitors leupeptin, pepstatin, and aprotinin. The cells were incubated on ice for 10 min and lysed by vigorous vortexing for 10 s, and a nuclear pellet obtained by centrifugation for 10 s at 4°C. The nuclear pellet was resuspended in 100 l of ice-cold Buffer C, which consists of 20 mM Hepes, pH 7.9, 25% glycerol, 0.42 M NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, 0.5 mM DTT, and 1 g/ml of the protease inhibitors leupeptin, aprotinin, and pepstatin. This mixture was incubated on ice for 20 min at 4°C. The extract was centrifuged for 2 min, and the supernatant aliquoted and stored at Ϫ80°C.
Electrophoretic Mobility Shift Assays (EMSAs)-For EMSAs (29) of GT1-7 nuclear extracts with radioactively labeled oligonucleotide probes, single-stranded oligonucleotides were 32 P-end-labeled with T4 polynucleotide kinase (30). The single-stranded oligonucleotides were hybridized to their complementary strands to generate double-stranded probes. 0.05 M of 32 P-end-labeled double-stranded oligonucleotide was incubated in 1 ϫ gel shift buffer (100 mM Tris-HCl, pH 8.0, 50% glycerol, 10 mM EDTA, 10 mM DTT) and 1 g/ml poly(dI-dC) in a 15-l reaction volume with 5 g of GT1-7 extract for 15 min at room temperature. To resolve protein-DNA complexes, the reaction mix was run on a 10% native polyacrylamide (75:1) gels (for probes L7 and mouse mammary tumor virus (MMTV) GRE) or 5% native polyacrylamide (39:1) gels (for probe L75) in 1 ϫ TBE. Gels were electrophoresed in 1 ϫ TBE at 180 V for approximately 30 min. For competition assays, unlabeled doublestranded competitor DNA was incubated first with GT1-7 nuclear extract for 10 min, and then the radiolabeled probe was added. For supershift assays with the monoclonal GR antibody, 1 l of ascites fluid containing the antibody was first incubated with the GT1-7 extract for 10 min at room temperature, after which the radiolabeled probe was added, and the incubation was continued for another 15 min. For supershift assays with the Oct-1 polyclonal antibody (Santa Cruz Biologicals), probe and extract were incubated for 15 min at room temperature, subsequent to which the antibody was added, and the reaction was incubated for an additional 30 min.
For EMSAs of the MMTV GRE with purified, DNA-binding domain of rat GR (XGR556), 0.05 M of 32 P-end-labeled probe was incubated with 25 ng of XGR556 in 1 ϫ binding buffer (40 mM Tris-HCl, pH 7.9, 2 mM EDTA, 20% glycerol, 0.2% Nonidet P-40, 2 mM DTT) in a 15-l reaction volume for 15 min at room temperature. One g of poly(dI-dC) was included in the reaction to minimize nonspecific binding. Protein-DNA complexes were resolved on a 10% polyacrylamide gel (75:1), and electrophoresis was carried out in 1 ϫ TBE buffer at 180 V for 45 min.
The MMTV GRE (31) consists of base pairs from Ϫ191 to Ϫ159 of the MMTV long terminal repeat (LTR) and has the following sequence: 5Ј-GTTTATGGTTACAAACTGTTCTTAAAACAAGGA-3Ј.
The oligonucleotide TAT3 (32), which consists of the GRE from the tyrosine aminotransferase (TAT) gene, from bp Ϫ2510 to Ϫ2490 has the following sequence: 5Ј-TCGACTGTACAGGATGTTCTAGCTAG-3Ј.
The octamer-binding sequence, Oct-1, was derived from the MMTV LTR and is located between bp Ϫ56 and Ϫ37 (33). It consists of the following sequence: 5Ј-AGCTTAGTCTTATGTAAATGCTTATGTAAAC-3Ј.
Only the sequence of the top strands of the MMTV GRE, TAT GRE, and the Oct-1-binding sequence are shown. The complementary bottom strands were also synthesized, and two strands were hybridized to generate double-stranded probes.

Glucocorticoid Repression of the Mouse GnRH Gene Is Mediated by Two
Promoter Elements-In order to localize sequence elements within the GnRH promoter that are responsible for GR-mediated transcriptional repression, dexamethasone effects were assessed on a series of plasmids containing 5Ј-flanking deletions of the GnRH promoter using the luciferase gene as a reporter. As shown in Table I, basal promoter activity was reduced 75-80% upon deletion of sequences between Ϫ3446 and Ϫ522 bp, suggesting removal of an enhancer element that has been demonstrated to reside within Ϫ1863 to Ϫ1571 bp of the rat promoter (34). Further deletion to Ϫ447 resulted in a drop of promoter activity to 6 -7% of the Ϫ3446 construct. Basal activity of the smallest promoter fragment tested (Ϫ108 bp) was reduced more than 99% compared with the full-length promoter. These results are generally in agreement with those of Kepa et al. (35) who carried out similar experiments using the rat promoter. Fig. 1 shows the effect of dexamethasone on promoter activity of this series of GnRH-luciferase fusion genes (results of 4 -29 experiments, each performed in triplicate). Dexamethasone treatment resulted in a 60 -70% repression of luciferase activity of the Ϫ3446, Ϫ522, Ϫ447, and Ϫ250 promoter constructs. Glucocorticoid sensitivity of the mouse GnRH promoter decreased slightly between Ϫ237 and Ϫ216 bp but was lost mainly with deletion of the sequences from Ϫ216 to Ϫ201 bp and from Ϫ184 to Ϫ150 bp. Basal activity of the Ϫ108 base pair construct was too low to reliably measure the effect of dexamethasone. Thus, the mouse GnRH promoter possesses two nGREs that contribute to glucocorticoid-mediated transcriptional repression. The proximal nGRE is located between Ϫ184 and Ϫ150 while the distal nGRE is located between Ϫ237 and Ϫ201. Both nGREs, either separately or together, exhibit minimal activity when fused to heterologous promoters derived from the herpes simplex virus thymidine kinase gene or Rous sarcoma virus LTR (data not shown).
Interestingly, an analogous region of the rat GnRH promoter (Ϫ126 to Ϫ22) appears to mediate repression by activators of protein kinase C such as 12-O-tetradecanoylphorbol-13-acetate (36). Since the mouse and rat GnRH promoters utilize different start sites, the numbering of 5Ј-flanking sequences varies. Thus, sequences from Ϫ237 to Ϫ150 of the mouse promoter correspond to Ϫ132 to Ϫ45 of the rat promoter and are 95% identical (36). The relationship between 12-O-tetradecanoylphorbol-13-acetate and dexamethasone repression is currently unknown but will be the subject of future studies.
GR Does Not Bind Directly to the mGnRH nGREs-The nGREs within the mouse GnRH promoter (Fig. 2) do not bear any homology to either positive GREs (18) or any nGREs identified in other genes (37)(38)(39)(40). Although nGREs do not share extensive sequence homology, some possess at least one copy (sometimes degenerate) of a half-site found in positively acting GREs (i.e. 5Ј-TGTTCT-3Ј or its palindromic complement 5Ј-AGAACA-3Ј). In some cases these half-sites are sufficient to direct the binding of GR (40,41). There is one 5/6 and one 4/6 match to this GRE half-site consensus within the proximal nGRE of the mouse GnRH promoter, and one 4/6 match within the distal nGRE (Fig. 2). Therefore, we examined whether glucocorticoid repression might be due to direct binding of GR to the GnRH nGREs. However, using both DNase I footprinting and EMSAs, we were unable to detect binding of the purified DNA-binding domain of GR (XGR556) to GnRH nGREs (data not shown). To confirm the lack of GR binding to GnRH nGREs, we also performed competition analysis to determine whether the GnRH nGREs could compete with a radiolabeled GRE derived from the MMTV LTR for binding to XGR556. Up to a 1000-fold excess of either unlabeled MMTV GRE (Fig. 3, lanes Dexamethasone had no effect on a control plasmid, human ␤-actin-luciferase, as indicated.

TABLE I
Basal activity of mouse GnRH 5Ј-flanking deletions GT1-7 cells were transiently transfected with the series of 5Ј-flanking deletions of the mouse GnRH gene inserted in a luciferase reporter vector. Cells were harvested 20 h later for protein and luciferase activity. Values represent the mean Ϯ S.E. of triplicate transfections in a single experiment. Luciferase activities were normalized for ␤-galactosidase activity obtained by co-transfecting cultures with cytomegalovirus-␤-galactosidase. Similar results were obtained in replicate experiments. *, the (Ϫ108) construct was not analyzed in this experiment; therefore, this value represents the mean of five other experiments.

3-5)
or the GnRH distal (Fig. 3, lanes 6 -8) and proximal nGREs (Fig. 3, lanes 9 -11) were used as competitors in this assay. As expected, unlabeled MMTV GRE was able to compete with the radiolabeled MMTV GRE probe for binding to XGR556, whereas the GnRH nGREs were not able to compete for binding of XGR556 to this probe even when present in 1000-fold molar excess. Thus, the GnRH nGREs do not bind the purified DNA-binding domain of GR with any appreciable affinity in vitro. GnRH nGREs Interact in Vitro with Factors Present in GT1-7 Nuclear Extract-While many examples of GR-mediated repression of transcription have emerged, no unifying mechanism is evident. In general, GR is capable of mediating transcriptional repression either via its direct interaction with DNA elements (37) or through protein-protein interactions that functionally inactivate specific transcription factors (42,43). Since GR did not appear to bind the GnRH nGREs in vitro, we used EMSAs to examine whether proteins in GT1-7 nuclear extracts bind to the mouse GnRH nGREs. Nuclear extracts were prepared from GT1-7 cells and incubated with radiolabeled L7 or L75 oligonucleotide probes representing sequences between Ϫ237 to Ϫ201 (distal nGRE) and Ϫ184 to Ϫ150 bp (proximal nGRE), respectively, of the mouse GnRH promoter. The formation of protein-DNA complexes was monitored by EMSAs. Eraly et al. (36) have shown by in vitro DNase I footprinting experiments that GT1-7 nuclear proteins form complexes with the promoter proximal region of the rat GnRH promoter, a region that is 95% homologous to mouse promoter proximal sequence.
Incubation of GT1-7 nuclear extract with the L7 probe resulted in the formation of several protein-DNA complexes (Fig.  4A, lane 2). All of these complexes represent proteins bound specifically to the L7 oligonucleotide, since their formation can be competed by excess unlabeled L7 oligonucleotide (Fig. 4A,  lane 3). When unlabeled oligonucleotides spanning different regions of the Ϫ237 to Ϫ201 sequence were tested for their ability to compete with the L7 probe for binding to GT1-7 nuclear extract, L71 and L72, which represent base pairs Ϫ237 to Ϫ218 and Ϫ222 to Ϫ210, respectively, did not compete for formation of any of the complexes. (Fig. 4A, lanes 4 and 5,  respectively). Only L73, which represents bp Ϫ218 to Ϫ201, competed effectively for formation of complex C1 and C6 (Fig.  4A, lane 6), suggesting that the proteins present in these complexes bind to the Ϫ218 to Ϫ201 bp region of the L7 probe.
Since complexes C2-C5 were competed only by L7 and not by L71, L72, or L73, these complexes must form only when the entire Ϫ237 to Ϫ201 bp sequence is present. Therefore, they either represent degradation products of protein complexes that bind tightly only when the entire Ϫ237 to Ϫ201 bp region is present, or their binding sequence is not completely included within the truncated L71, L72, or L73 oligonucleotides. It is not clear why some complexes such as C4 and C5 increase in intensity in the presence of oligonucleotides L71, L72, and L73. Perhaps factors present in crude nuclear extracts that destabilize these complexes are competed by various oligonucleotides. That only L73 is able to compete effectively for formation of complexes suggests that the Ϫ218 to Ϫ201 bp region binds GT1-7 nuclear proteins stably. This result is consistent with the promoter deletion analysis, which implicates this region as important for dexamethasone mediated repression of the mouse GnRH gene. None of the complexes was competed by the TAT3 oligonucleotide (Fig. 4A, lane 7), which contains the GRE sequence from the TAT promoter (32), ruling out the possibility that endogenous GR in GT1-7 extracts binds directly to the L7 probe.
The number of complexes formed between the L7 oligonucleotide and GT1-7 nuclear proteins varies with different batches of extract. For example, a nonspecific complex (NS) (Fig. 4B, lane 2) that is not effectively competed by excess unlabeled L7 or L73 oligonucleotide (Fig. 4B, lanes 3 and 4) is sometimes observed. In addition, complexes C2-C6 are not always resolved in the EMSAs (Fig. 4B, lanes 2 and 3). To determine if any new complexes are induced by dexamethasone treatment of cells, we performed the EMSA with GT1-7 extracts from both dexamethasone-treated and untreated cells. We were unable to detect novel dexamethasone-induced complexes (data not shown).
When the L75 oligonucleotide (Ϫ184 to Ϫ150 bp, proximal nGRE) was used as the labeled probe in EMSAs with GT1-7 nuclear extracts, a number of complexes were formed (Fig. 5, lane 2) whose formation could be competed for with unlabeled L75 oligonucleotides (Fig. 5, lane 3). These EMSAs with the L7

FIG. 2. The GnRH nGREs and oligonucleotides used in EMSAs.
The sequence of the GnRH distal nGRE (Ϫ237 to Ϫ201, oligonucleotide L7) and proximal nGRE (Ϫ184 to Ϫ150, oligonucleotide L75) are shown. These oligonucleotides were synthesized with BamHI and XhoI sites at their 5Ј and 3Ј termini, respectively. Also shown are the locations within the distal nGRE of the L71 (Ϫ237 to Ϫ218), L72 (Ϫ228 to Ϫ211), and L73 (Ϫ218 to Ϫ201) oligonucleotides. Potential matches to a GRE half-site in the L7 and L75 oligonucleotides (distal and proximal nGREs, respectively) are overlined. The asterisks in L7 and L75 oligonucleotide sequences represent the mismatches from the consensus GRE sequence half-sites.  6, 7, and 8), or L75 (proximal nGRE) (lanes 9, 10, and 11) oligonucleotides was incubated with GT1-7 nuclear extract for 10 min prior to the addition of MMTV GRE probe. and the L75 probes suggest that GT1-7 nuclear extracts contain proteins that are capable of binding specifically to the mouse GnRH nGREs.

GR Is Part of a Multiprotein Complex Formed on L7 and L75
Oligonucleotides-The inability of pure GR to bind to the GnRH nGREs as well as the inability of cold GREs to compete for binding of GT1-7 extracts to the L7 (see Fig. 4A) and L75 probes (data not shown) ruled out the possibility of direct GR interactions with GnRH nGREs. However, we wished to examine whether endogenous GR in GT1-7 cells, although unable to bind directly to the GnRH nGREs, is part of the protein-DNA complexes formed by GT1-7 cell nuclear extracts and the GnRH nGREs. In order to detect GR in these complexes, a GR antibody was added during the incubation of GT1-7 nuclear extracts with GnRH promoter oligonucleotides. GR-containing complexes should either be supershifted or removed upon incubation of extracts with antibody. Indeed, a specific complex, complex C1, formed upon incubation of GT1-7 nuclear extract, and the L7 oligonucleotide was supershifted when the anti-GR antibody BuGR2 was included in the incubations (Fig. 6A, lane  3), suggesting that GR is indeed included within a protein complex that binds to the Ϫ237 to Ϫ201 bp region of the mGnRH promoter. Since formation of complex C1 was also competed by the L73 oligonucleotide (see Fig. 4A), this suggests that the GR-containing protein complex recognizes the mouse GnRH promoter sequence between positions Ϫ218 and Ϫ201. The supershift of complex C1 is specific for the GR antibody, since a monoclonal antibody to chicken PR (PR22) did not alter the pattern of complexes observed on the L7 oligonucleotide (Fig. 6A, lane 4). The presence of GR in a protein complex that binds to the L7 oligonucleotide may explain why no differences in complex formation are seen between dexamethasone-treated and untreated extracts. Since GR is known to become activated in vitro by the extract purification protocol, GR may be activated and part of the nuclear multiprotein complex even in the absence of dexamethasone. In addition to alterations in complex C1, the inclusion of BuGR2 antibody in EMSAs of the L7 probe led to an increased intensity of complexes C3 and C4. While this may represent a stabilization of weakly bound GR, the functional significance of this is unclear, since only complex C1 and complex C6 formed on the L7 oligonucleotide were shown to be competed by the L73 oligonucleotide, which pos- sesses the sequence (Ϫ218 to Ϫ201) important for dexamethasone-mediated repression.
Similarly, to analyze whether protein-DNA complexes formed on the L75 oligonucleotide (Ϫ184 to Ϫ150) contain GR, we included the BuGR2 antibody when incubating the L75 probe with GT1-7 nuclear extract. One of the complexes was removed by BuGR2 (Fig. 6B, lane 3, arrowhead), suggesting that this complex likewise contains GR. Furthermore, none of the other complexes was altered by the presence of the anti-GR antibody. An antibody to chicken PR did not alter the mobility of the complex, which was removed by BuGR2 (Fig. 6B, lane 4). Thus, both of the promoter elements shown by 5Ј-flanking deletion analysis to be involved in dexamethasone-mediated repression of the mouse GnRH promoter are bound in vitro by specific multiprotein complexes that contain GR. We hypothesize that GR is present in this complex by virtue of its interaction with a DNA-bound factor, since the purified receptor itself does not bind to these sequences with any appreciable affinity in vitro.
Oct-1 and GR Are Present in the Same Multiprotein Complex on the Mouse GnRH Distal nGRE-Inspection of the region between Ϫ218 and Ϫ201 reveals an A ϩ T-rich sequence, which resembles a consensus binding site for Oct-1, a protein belonging to the POU domain family of transcription factors (44 -46). The A ϩ T-rich sequence between Ϫ218 and Ϫ201 contains a 6/8 match on the coding strand to the consensus 5Ј-ATG-CAAT-3Ј octamer-binding sequence (Fig. 7) and a 5/8 match on the noncoding strand. Since GR has been shown to interact with one member of this family, Oct-1, and repress its binding to an A ϩ T-rich octamer site in the histone H2B promoter (47), we examined whether Oct-1 was a component of protein-DNA complexes formed by GT1-7 nuclear extracts on the GnRH nGREs. Oct-1 and other POU domain family members have been detected in GT1-7 cells (48). As shown in Fig. 8A, an Oct-1-binding sequence from the MMTV promoter (31) effectively competed for the formation of complex C1 (Fig. 8A, lane  3) in EMSAs of the L7 probe with GT1-7 cell nuclear extract. Similarly, a consensus Oct-1-binding sequence generated by mutating the mouse GnRH promoter sequence between Ϫ218 and Ϫ201 (Fig. 7, L73M) was able to compete for binding of GT1-7 nuclear extract to the L7 probe (Fig. 8A, lane 4). This competition analysis suggests that Oct-1 may be part of a protein complex formed by GT1-7 nuclear extracts and the L7 probe.
Since different POU domain proteins can bind to the Oct-1 consensus sequence with varying affinities, the competition analysis does not on its own establish the exclusive presence of Oct-1, but it reveals the presence of some POU domain protein in this complex. We therefore performed EMSAs of GT1-7 nuclear extract with the L7 probe in the presence of an Oct-1specific antibody to establish whether Oct-1 is, in fact, part of the complexes formed on the L7 probe. The inclusion of the Oct-1 antibody in the binding assays resulted in the disappearance of complex C1 (Fig. 8B, lane 3), the same complex whose FIG. 6. Effects of GR antibody on the binding of GT1-7 nuclear proteins to the distal (oligonucleotide L7) and proximal (oligonucleotide L75) nGREs. A, 1 l of the anti-GR antibody, BuGR2, or the anti-chicken PR antibody, PR22, was incubated with 5 g of GT1-7 nuclear extract (lanes 3 and 4, respectively) for 5 min at room temperature prior to the addition of 0.05 pmol of 32 P-end-labeled L7 probe as described under "Experimental Procedures." The reaction was then incubated for an additional 15 min at room temperature, and protein-DNA complexes (C1-C6) resolved on a 10% polyacrylamide gel (75:1). B, 1 l of the anti-GR antibody, BuGR2, or the anti-chicken PR antibody, PR22, was incubated with 5 g of GT1-7 nuclear extract (lanes 3 and 4, respectively) for 5 min at room temperature prior to the addition of 0.05 pmol of 32 P-end-labeled L75 probe as described under "Experimental Procedures." The reaction was then incubated for an additional 15 min at room temperature, and protein-DNA complexes were resolved on a 5% polyacrylamide gel (37:1). The protein-DNA complex that is removed by the BuGR2 antibody is indicated by a closed arrowhead. 7. Sequence of the L7 and L73M oligonucleotides. The sequence of the L7 oligonucleotide (distal nGRE, Ϫ237 to Ϫ201) is shown. The region that is homologous to the Oct-1-binding site is overlined and underlined, and the mismatches to the perfect Oct-1 site are shown by asterisks. The L7 sequence exhibits a 5/8 match to a perfect Oct-1-binding sequence on the top strand and a 6/8 match on the bottom strand. The L73M oligonucleotide exhibits a perfect match to a consensus Oct-1-binding sequence (5Ј-ATGCAAT-3Ј) on the bottom strand (underlined) and was derived by mutating two nucleotides in the L73 sequence (Ϫ218 to Ϫ201).
formation was competed by the Oct-1 consensus sequence L73M (see Fig. 8A). Therefore, Oct-1 is indeed part of the protein-DNA complex formed by GT1-7 nuclear extract on the L7 probe. Complex C1, which contains Oct-1, was shown to be also the GR-containing complex which binds to the GnRH nGRE between Ϫ218 and Ϫ201 (see Fig. 4, A and B). Therefore, GR and Oct-1 are both part of a heteromeric protein complex that binds to mouse GnRH sequences between Ϫ218 and Ϫ201. Since formation of complex C1 can be competed by excess Oct-1-binding oligonucleotide but not by a GRE oligonucleotide, this suggests that Oct-1 but not GR may be binding directly to the GnRH nGRE.
The proximal nGRE contains potential binding sites for a number of different families of transcription factors including the helix-loop-helix, GATA, and POU domain families of transcription factors. Initial attempts to examine binding of GT1-7 nuclear proteins to various subfragments of the L75 oligonucleotide generated negative results, most likely due to the inherent weakness of nuclear protein binding to the L75 oligo-nucleotide. However, at least one protein that binds to the proximal nGRE must be a member of the POU domain family, since a specific complex formed on the L75 oligonucleotide can be competed with the consensus Oct-1 oligonucleotide L73M (data not shown). Since this complex is not supershifted by the Oct-1 antibody (data not shown), a POU domain protein other than Oct-1 must bind to the proximal nGRE. Although we have not yet determined whether GR and the POU domain protein are part of the same complex on the proximal nGRE, the lack of Oct-1 involvement distinguishes the mechanism of GR action at the proximal nGRE from that at the distal nGRE. DISCUSSION We have identified two regions of the mouse GnRH promoter, one between Ϫ237 and Ϫ201 (distal nGRE) and another between Ϫ184 and Ϫ150 (proximal nGRE), which mediate transcriptional repression by glucocorticoids. These sequences show no similarity to known positive or negative GREs (18 -20) and do not mediate glucocorticoid repression through direct interactions with the GR. Rather, GRs appear to participate in glucocorticoid repression of GnRH gene transcription by virtue of their association within a multiprotein complex at the nGREs. While the precise composition of the complexes formed in vitro on the GnRH nGREs have not been defined, the distal nGREs appears to be recognized by a POU domain transcription factor present within GT1-7 cells. Antibody supershift experiments identified Oct-1 as the POU domain family member bound to the distal nGRE.
While precedence has been established for the involvement of other transcription factors in GR-mediated repression (40,42,43,47,49,50), co-occupancy of GR with other transcription factors at nGREs is not typically observed. For example, the interaction of Oct-1 with GR precludes its binding in vitro to an octamer site derived from the histone H2B promoter and could be responsible for glucocorticoid-dependent repression of this promoter (47). This contrasts with results presented in this report, where Oct-1 and GR were shown to co-occupy an element in vitro (i.e. the distal nGRE) that functions in glucocorticoid repression. Since endogenous GR and Oct-1 levels were not manipulated in the GT1-7 cells from which nuclear extracts were prepared, our ability to detect a DNA-bound complex in vitro that possesses both GR and Oct-1 may be more reflective of events that occur in vivo.
Unlike the octamer site at the H2B promoter (47), the distal nGRE does not exhibit a perfect match to the consensus ATG-CAAT octamer-binding sequence. The sequence between Ϫ218 and Ϫ201 of the mouse GnRH promoter exhibits a 6/8 match to the consensus octamer sequence on the bottom strand and a 5/8 match on the top strand. The highly conserved adenosine residues at positions 1 and 7 (51), are present on both strands. The mismatches at positions 3 and 4 on both strands are known to be tolerated (51) and allow Oct-1 binding in vitro. Similarly, the mismatch at position 5 on the top strand is also tolerated. It remains to be determined whether these departures from the octamer consensus sequence within the GnRH distal nGRE impact the nature of GR interactions with factors bound at this site, which could include Oct-1. Sequences that flank the binding site for POU domain transcription factors have been shown to affect steroid hormone interactions with POU domain transcription factors, as exemplified by estrogen receptor interactions with Pit-1 in the prolactin promoter (52). Therefore, the octamer core, as well as the flanking sequence, at the distal nGRE may determine not only the affinity of Oct-1 binding to that site but also the nature of additional proteins that comprise a multiprotein complex.
Since Oct-1 appears to make direct DNA contact at the distal nGRE, it may function to nucleate the formation of multipro- FIG. 8. Effects of an Oct-1-binding oligonucleotide and an anti-Oct-1 antibody on GT1-7 nuclear protein binding to the distal nGRE (oligonucleotide L7). A, 5 g of GT1-7 nuclear extract was incubated with 0.05 pmol of 32 P-end-labeled L7 oligonucleotide in 1 ϫ binding buffer for 15 min at room temperature as described under "Experimental Procedures" (lanes 2-4). Protein-DNA complexes were resolved on a 10% polyacrylamide gel (75:1). For the competition analysis, 250-fold molar excess of either an Oct-1-binding sequence from the MMTV LTR (lane 3) or the L73M oligonucleotide (lane 4) were included in the binding assay. Complex C1, which is competed by both Oct-1binding sequences (lanes 3 and 4) is shown by an arrowhead. B, for supershift assays with the Oct-1 antibody, 0.05 pmol of 32 P-end-labeled L73 oligonucleotide was incubated with 5 g of GT1-7 nuclear extract for 15 min at room temperature. Then, 2 l of an Oct-1-specific antibody was added, and the reaction was incubated for an additional 30 min at room temperature. Complex C1, which is removed upon incubation with the antibody (lane 3), is shown by an arrowhead. tein complexes at this site, which could provide the GnRH gene with the potential to be regulated by various signal transduction pathways. For example, we speculate that by virtue of its association within a multicomponent complex (which includes Oct-1) at the distal nGRE, GR imparts a hormone-dependent decrease in GnRH promoter activity. Given the ubiquitous expression of Oct-1, and the restricted expression of GnRH to a limited number of neurons within the hypothalamus, GnRH neuron-specific co-activators may also comprise a subset of proteins that are recruited to the GnRH promoter by DNAbound Oct-1. In an analogous manner, the Oct-1 homeodomain serves to nucleate the formation of multiprotein complexes at herpes simplex virus immediate early promoters, which participate in promoter activation (53,54). This model may be invoked to explain why sequences required for glucocorticoid repression of mouse GnRH gene transcription coincide precisely with those sequences in the nearly identical region of the rat GnRH promoter that are required for transcriptional repression brought about by activation of protein kinase C by tumor promoters such as 12-O-tetradecanoylphorbol-13-acetate (36). Although 12-O-tetradecanoylphorbol-13-acetate repression of rat GnRH gene transcription does not appear to involve activation of AP-1 family members (36), perhaps novel factors are recruited to the nGRE site via their interaction with DNA-bound Oct-1 upon protein kinase C activation. Alternatively, such protein kinase C-responsive factors may already be present within a complex at the nGRE and be poised to respond to protein kinase C activation.
Our current results also suggest that the activity of GnRH nGREs is most efficient when linked to their native promoter. While this survey of promoter specificity of the GnRH nGRE is certainly not exhaustive, promoter specificity has also been observed for the proopiomelanocortin nGRE. Unlike the GnRH nGRE, the proopiomelanocortin nGRE directs the binding of GR, although in an unique trimer arrangement that has not been observed in any positive or negative GREs examined to date (37). Perhaps the GnRH and proopiomelanocortin nGREs share the feature of exposing a unique surface of GR that has the potential to negatively impact the activity of only a limited number of transcription factors. The resolution of the apparent promoter specificity of GnRH nGRE activity awaits further experimentation.
Some POU domain transcription factors such as those implicated in brain development and neurogenesis (55-60) are expressed in a tissue-specific manner, whereas Oct-1 is expressed in a number of different cell types. However, Oct-1 can contribute to tissue-specific expression by exhibiting a requirement for tissue-specific coactivators in the activation of promoters such as the immunoglobulin promoter in B cells (61,62). Similarly, Oct-1 can activate the herpes simplex virus early promoter by interacting with the viral protein VP16 (53,54). These studies highlight the POU domain protein's contribution to developmental regulation and tissue-specific expression and the potential role of co-activators in tissue-specific gene expression. Thus, although the activity of a GT1 cell-specific enhancer in the rat GnRH 5Ј-flanking region has been shown to be highly dependent upon the binding of Oct-1 (48), it is possible that the restricted expression of GnRH to the GnRH neurons requires tissue specific co-activators. As we have shown in this report, the function of POU domain proteins may also be integrated with hormonal signals that function to alter neuronal specific gene expression in response to various environmental cues.