Heterodimerization of Mineralocorticoid and Glucocorticoid Receptors at a Novel Negative Response Element of the 5-HT1A Receptor Gene*

Negative regulation of neuronal serotonin (5-HT1A) receptor levels by glucocorticoids in vivo may contribute to depression. Both types I (mineralocorticoid) and II (glucocorticoid) receptors (MR and GR, respectively) participate in corticosteroid-induced transcriptional repression of the 5-HT1A gene; however, the precise mechanism is unclear. A direct repeat 6-base pair glucocorticoid response element (GRE) half-site 5′-TGTCCT separated by 6 nucleotides was conserved in human, mouse, and rat 5-HT1A receptor promoters. In SN-48 neuronal cells that express MR, GR, and 5-HT1A receptors, deletion or inactivation of the nGRE (negative GRE) eliminated negative regulation of the rat 5-HT1A or heterologous promoters by corticosteroids, whereas its inclusion conferred corticosteroid-induced inhibition to a heterologous promoter. Bacterially expressed recombinant MR and GR preferentially bound to the nGRE as a heterodimer, as identified in nuclear extracts of MR/GR-transfected COS-7 cells, and with higher affinity than MR or GR homodimers. In SN48 and COS-7 cells, concentration-dependent coactivation of MR and GR was required for maximal inhibitory action by corticosteroids and was abrogated in the L501P-GR mutant lacking DNA binding activity. Corticosteroid-mediated transcriptional inhibition was greater for MR/GR in combination than for MR or GR alone. These data represent the first identification of an nMRE/GRE and indicate that heterodimerization of MR and GR mediates direct corticosteroid-induced transrepression of the 5-HT1A receptor promoter.

of the 5-HT1A receptor gene results in increased anxiety (3)(4)(5). Limbic brain regions involved in mood control, such as the hippocampus and septum, are richly innervated by serotonergic neurons (6). Glucocorticoids exert strong negative regulation on the serotonergic system (7)(8)(9), and the limbic system is particularly sensitive to glucocorticoids because both high and low affinity glucocorticoid receptor types I and II (i.e. MR and GR) are present (10 -12). The 5-HT1A receptor plays an important dual role in the serotonergic system, because it is expressed both postsynaptically in the limbic system and presynaptically as the primary somatodendritic autoreceptor of serotonergic raphe neurons (13,14). The level of expression of somatodendritic 5-HT1A receptors determines serotonergic tone, in part, by inhibiting raphe firing activity and is thus implicated in mood disorders, such as depression and generalized anxiety, that involve the serotonergic system (11,15,16). Major depression has also been associated with disruption of the hypothalamic-pituitary-adrenal axis; as many as 50% of depressed patients fail to suppress corticotrophin and cortisol levels upon dexamethasone challenge. Interactions between glucocorticoids and the serotonin system may be critical in the origin or maintenance of the depressed state.
The 5-HT1A receptor is strongly expressed in the hippocampus, septum, and other limbic areas (17,18) and functions as a key regulator of the limbic system. Transcription of the 5-HT1A receptor gene is negatively regulated by corticosteroids especially in the limbic system. In the rat, adrenalectomy is followed by a rapid increase (within hours) in de novo 5-HT1A RNA synthesis (as measured by nuclear run-on assay (19)), total 5-HT1A mRNA, and binding sites. These changes are completely suppressed by low concentrations of corticosterone that preferentially activate MR (20 -23). The extent of increase in 5-HT1A receptor gene transcription upon adrenalectomy was most robust in the hippocampus and septum, with smaller changes in the raphe nuclei. Using MR-and GR-selective ligands (in rat) and gene knockout approaches (in mouse) (24 -26), MR (primarily) and GR have been implicated in negative regulation of 5-HT1A gene expression, but the precise mechanism remains unclear.
Positive gene regulation by glucocorticoids results in an 8 -12-fold induction of transcription and is mediated by binding of GR/MR homo-or heterodimers to a canonical GRE composed of an inverted repeat of GRE half-sites (TGTTCT) separated by 3 nucleotides (27,28). However, negative regulation of gene transcription by glucocorticoids results in 40 -70% inhibition of transcription and involves diverse mechanisms (29). Direct steroid-induced repression of target gene transcription occurs at an nGRE (e.g. proopiomelanocortin (30)). Indirect mechanisms are mediated via interactions of glucocorticoid receptors with a variety of other transcription factors such as Pbx and * This research was supported by the Canadian Institute of Health Research (CINR). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  Oct-1 (31,32), XTF (33), or c-Jun and c-Fos (34). The mechanism of glucocorticoid action on the 5-HT1A gene was examined using deletion constructs of the promoter to localize the site of gene repression by corticosteroids in SN-48 cells, a septal cell line that expresses 5-HT1A receptors. Because interactions between MR and GR appear to play an important role in glucocorticoid-mediated regulation of 5-HT1A receptor transcription in vivo, gel shift and cotransfection experiments were designed to assess functional interactions at the 5-HT1A promoter. Our findings indicate that a novel nGRE confers direct negative regulation by corticosteroids to the rat 5-HT1A receptor or heterologous promoters and suggest that MR and GR form a novel head-to-tail heterodimeric complex with the nGRE to mediate glucocorticoid-induced gene repression.
Cell Culture and Transfections-SN-48 and COS-7 cells were maintained in Dulbecco's modified Eagle's medium (DMEM, Life Technologies, Inc.) supplemented with 10% fetal calf serum at 37°C in 5% CO 2 . SN-48 and COS-7 cells were grown to 50 -60% confluence, and the medium in each plate were replaced 12 h before transfection. Calcium phosphate coprecipitation was used for both cell lines (35). Cells were transfected with the following amounts of normal or mutant constructs: 5-HT1A gene deletion constructs (10 g); 3xnGRE/SV40 (1 g); 1xn-GRE/SV40 (3 g); 1.2 pM p6RMR (5.4 g); 1.2 pM p6RGR (5 g); 0.6 pM MR/GR; or as indicated. Cells were incubated with DNA precipitates for about 10 h and then were supplemented with charcoal-stripped, steroid-free fetal calf serum for 36 h. SN-48 cells were differentiated by a reduction of fetal calf serum (charcoal-treated serum) to 1% v/v and the addition of 1 M retinoic acid (39). Cells were resuspended in reporter lysis buffer (Promega) for detection of luciferase or ␤-galactosidase activity as described (40) or collected for extraction of nuclear protein.
Recombinant Protein Expression-pGEX-2T-X568 encoding GST-GR fusion protein containing amino acids 407-568 of the rat GR (DNA binding domain) was a gift from Dr. Robert Haché (41). pGEX-2TKS-X922 encoding GST-MR fusion protein containing amino acids 105-922 of rat MR (GST-MR) was created by cloning the NotI-HindIII fragment of p6RMR in frame into the NotI-HindIII sites of pGEX-2TKS, modified from pGEX-2TK (Amersham Pharmacia Biotech) by Dr. M. Ghahremani (University of Ottawa). GST-GR-or GST-MR-transformed Eschrichia coli cells were grown at 37°C overnight followed by induction with 1 mM isopropy-1-thio-␤-D-galactopyranoside for 4 -6 h at 23°C. Induced bacteria were resuspended in lysis buffer (25 mM Hepes, pH 7.9, 100 mM KCl, 20% glycerol, 2 mM EDTA, 0.2 mM phenymethylsulfonyl fluoride) supplemented with 100 g/ml lysozyme and incubated for 10 min at 4°C. Cells and DNA were sheared by passage through a series of 18-, 20-, and 25-gauge needles followed by sonication in the presence of 0.1% Nonidet P-40 at 4°C. Lysates were centrifuged for 10 min at 10,000 ϫ g and the supernatants incubated with 3 ml of glutathione-Sepharose 4B beads (Amersham Pharmacia Biotech) for 90 min. The beads were subsequently washed four times with 10 bed volumes of lysis buffer and three times with 20 bed volumes of 60% lysis buffer containing 0.1% Nonidet P-40, eluted in glutathione elution buffer (Amersham Pharmacia Biotech), and stored at Ϫ80°C. For transfections, GST-MR was subcloned into the SmaI site of pcDNA3 vector, and Flag-GR was created by cloning the polymerase chain reaction product of GR in frame into the EcoRI site of pcDNA3-Flag vector downstream from the Flag epitope.
DNase I Footprint Assay-The 1186-luc construct was digested with MluI, end-labeled using Klenow DNA polymerase with [␣-32 P]dCTP (3000 Ci/mmol) (40), and digested with HindIII to generate a 1186-bp fragment, which was purified on an agarose gel. The 60-l reaction contained: 20,000 cpm of labeled probe, 3 g of poly(dI-dC) (Roche Molecular Biochemicals), nuclear extract or recombinant GST, GR, or MR in binding buffer (20 mM Hepes, 0.2 mM EDTA, 0.2 mM EGTA, 100 mM KCl, 5 mM MgCl 2, 5% glycerol, and 2 mM dithiothreitol, pH 7.9) (40). Following incubation samples were digested with DNase I (Amersham Pharmacia Biotech) for 2 min at room temperature with the addition of 15 g of yeast tRNA as carrier. After phenol/chloroform extraction, samples were resuspended in formamide dye solution and electrophoresed through 8% polyacrylamide/urea gel. The total nuclear protein for all extracts was quantitated according to the Bradford method with bovine albumin as standard.
Electrophoretic Mobility Shift Assay (EMSA)-Complementary 39-bp oligonucleotides (Ϫ1180 to Ϫ1141 bp) containing the dexamethasoneresponsive element were end-labeled with [␥-32 P]dATP and used in EMSA (40). Nuclear extracts or recombinant GR and MR were preincubated at room temperature for 20 min with or without competitor DNA or antibodies in a 20-l reaction containing EMSA buffer (20 mM Hepes, 0.2 mM EDTA, 0.2 mM EGTA, 100 mM KCl, 5% glycerol, and 2 mM dithiothreitol, pH 7.9), and 2 g of poly(dI-dC). 32 P end-labeled probe (50,000 cpm) was added and incubated for an additional 20 min at room temperature. The reaction was separated on a 5% polyacrylamide gel at 4°C, which was dried and exposed to film overnight at Ϫ80°C with an intensifying screen.
Western Blot Analysis-Nuclear protein was extracted from SN48 cells as described above. Western blot was performed as described previously (42). Proteins were electrophoresed on sodium lauryl sulfate containing 10% polyacrylamide gel. The MR or GR antibody (1/1000) was rabbit polyclonal antibody (Santa Cruz Biotechnology, Inc.). The secondary antibody (1/3000) and chemiluminescence procedures were conducted according to the manufacturer (Amersham Pharmacia Biotech).
Statistical Analysis-The statistical significance was evaluated using a Student's t test for two-group comparisons or by one-way analysis of variance with the Tukey test for multiple comparisons among more than two groups.

Identification of nGRE-like Sequence in the Rat 5-HT1A
Receptor Promoter-We have previously identified a major TATAdriven transcriptional initiation site at Ϫ967 bp from the translational initiation ATG of the rat 5-HT1A receptor gene, which is regulated by proximal GC-rich enhancer elements and by upstream dual repressor elements that confer neuron-specific expression (35,40). To identify DNA elements that confer corticosteroid-mediated transcriptional repression, a series of six 5Ј-deletion fragments fused to a luciferase reporter gene was transfected in septal SN-48 cells differentiated to a neuronal phenotype, and luciferase activity in the absence or presence of 1 M dexamethasone was measured (Fig. 1). Dexamethasone induced a 40% reduction in luciferase activity for the 1519-or 1186-luciferase constructs, but no significant effect was observed for any of the shorter luciferase constructs (Fig. 1A), including 995-, 800-, and 618-luciferase (not shown). Within the dexamethasone-responsive region between positions Ϫ1186 and Ϫ1145 bp, a GRE-like sequence (Ϫ1169 to Ϫ1148 bp) TGTCCT-nnnnnn-TGTCCT was identified, and the 6-bp spacing between repeated 6-bp GRE half-sites (TGTCCT) was conserved in human and mouse 5-HT1A promoters. Mutations were incorporated to disrupt the GRE half-sites (1519m-and 1186m-luciferase, see "Materials and Methods"), resulting in a complete loss of corticosteroid transrepression for both constructs (Fig. 1B). Similarly, in rat raphe RN46A cells (35), another 5-HT1A-expressing neuronal cell line, 1 M dexamethasone induced a 32% reduction in transcriptional activity of the 1186-luciferase; and this was absent in the mutant 1186m construct (data not shown). Thus, the nGRE was critical for dexamethasone-mediated negative regulation of 5-HT1A receptor gene transcription in neuronal cells.
MR and GR Associate with the 5-HT1A nGRE-To determine whether nuclear proteins interact with the putative nGRE, the DNase I protection assay was conducted using SN-48 nuclear extracts or purified recombinant glucocorticoid receptors (Fig.  2). DNase I digestion using nuclear extracts revealed a pair of strongly protected segments between Ϫ1173 and Ϫ1145 bp and corresponding to the two GRE-like half-sites (Fig. 2B, lanes 2 and 3). The interaction between GR and MR at the nGRE was examined directly using recombinant GR and MR (Fig. 2, A and B, lanes 4 -9). Purified GR or MR or GR/MR heterodimerization protected the same nGRE sequence observed using SN-48 nuclear extracts (Fig. 2B, lanes 2 and 3), but GST protein did not protected this region ( Fig. 2A, lanes 2 and 3), suggesting that the DNA-protein complex observed involves a direct interaction of GR or MR at two half-sites on the nGRE sequence.
To examine further the interaction of GR or MR with the nGRE, an EMSA was conducted using double-stranded 39-bp oligonucleotides incorporating the nGRE sequence incubated with nuclear protein extracts from SN-48 cells (Fig. 3) or with recombinant GR and MR proteins (Fig. 4). A specific protein complex from SN-48 nuclei associated with the nGRE and was competed by excess unlabeled nGRE oligonucleotide but not by the unrelated E2F sequence (Fig. 3A, lane 4) or by the mutated nGRE sequence (data not shown). The inclusion of either anti-GR or anti-MR antibody displaced this complex (Fig. 3A, lanes 5 and 6). As observed by others (32,43), the interaction of antibody with either rat GR or MR prevented high affinity binding to labeled nMRE/GRE (39-oligo), displacing the DNAprotein complex rather than inducing a supershift of the complex. By contrast, inclusion of either nonspecific antibody or preimmune serum did not displace the complex (Fig. 3A, lanes  7 and 8). Using GR antibody against MR or MR antibody against GR in Fig. 4, A (lane 6) and B (lane 6), we did not find a cross-reaction of MR and GR antibodies to each other, arguing against nonspecific actions of the antisera. To determine directly whether MR and GR are expressed in SN48 cells, Western blots were conducted using nuclear extract from SN48 cells (Fig. 3B). In parallel samples single MR-and GR-reactive species of 120-and 95-kDa, respectively, were detected with comparable levels of expression as estimated by densitometry. In Fig. 3C, the density of the protein-DNA complex was increased in a concentration-dependent manner by prior treatment of SN-48 cells with dexamethasone, suggesting glucocorticoid-mediated translocation of GR/MR to the nucleus.
MR and GR Form a Heterodimeric Complex with the 5-HT1A nGRE-To examine whether GR or MR binds directly to the nGRE sequence, EMSA was performed using purified recombinant GR or MR proteins (Fig. 4). In the presence of GR or MR, single protein-DNA complexes were detected, in agreement with evidence from DNase I protection analysis (Fig. 2). These complexes were not present in the absence of protein and were displaced by unlabeled nGRE (Fig. 4, A and B) or by the addition of the appropriate antibody to GR (Fig. 4A, lane 5) or MR (Fig. 4B, lane 5). In contrast, noncorresponding NF1 antibody or preimmune serum did not compete (Fig. 4, A and B,  lanes 6 and 7). Thus GR or MR alone can bind directly to the nGRE sequence, which is displaced by anti-GR or anti-MR, respectively. The interaction between GR and MR at the nGRE was examined directly by EMSA using recombinant GR and MR (Fig. 4C). Incubation of labeled nGRE with GR, MR, or MR/GR in an equal plasmid molar ratio resulted in distinct complexes with readily distinguishable electrophoretic mobilities, presumably representing GR and MR receptor homodimers (predicted molecular mass ϭ 98 and 196 kDa, respectively) and the MR/GR heterodimer (147 kDa). Thus GR, and MR preferentially associate as a heterodimer at the nGRE, consistent with observations that MR and GR heterodimerize with a positive GRE (44,45).
To assess GR/MR receptor interactions in vivo, COS-7 cells were transfected with expression plasmids for Flag-GR alone, GST-MR alone, or both, and receptor expression was assessed by EMSA performed on nuclear extracts using the nGRE oligonucleotide (Fig. 4D). The COS-7 cells are derived from CV-1 cells, which lack endogenous MR and GR and therefore provide a null background for expression of transfected receptors (38). Consistent with an absence of detectable MR/GR, no gel-retarded bands were present in extracts from COS-7 cells not transfected with MR or GR plasmids (lane 9). Nuclear extracts from COS-7 cells transfected with 2 pM GST-MR, 2 pM Flag-GR, or 1 pM each GST-MR and Flag-GR plasmids together resulted in specific protein-DNA complexes that were supershifted with the appropriate antibodies (anti-GST or anti-Flag). The predicted molecular masses of these full-length recombinant receptor constructs were 140 kDa for GST-MR and 95 kDa for Flag-GR; hence the electrophoretic mobilities of homodimers or heterodimers were distinguishable (MR, MR/GR, and GR were indicated by arrows 1, 2, and 3, respectively, Fig. 4D) in 4% nondenatured gel. These results indicate that the transfected MR and GR in COS-7 cells were functional and present at a similar density as detected by Western blot (data not shown) and that both associated with the nGRE of the 5-HT1A receptor gene. Furthermore, anti-GST or -Flag antibody independently supershifted binding of the nGRE-protein complex in cells transfected with GRϩMR (lanes 8,9), indicating that the GR/MR heterodimer forms the major species of the complex.
The binding affinity of recombinant MR, GR or MR/GR heterodimer to the nMRE/GRE was evaluated by EMSA using increasing concentrations of radiolabeled 39-oligo followed by quantification of bound and free oligonucleotide (Fig. 5). Scatchard analysis revealed that the dissociation constant (K d ) value was 0.13 ϫ 10 Ϫ9 M for GR, 0.12 ϫ 10 Ϫ9 M for MR, and 0.08 ϫ 10 Ϫ9 M for GR ϩ MR (Fig. 5B). These K d values are in the same range as for GR binding to the nGRE (0.25 ϫ 10 Ϫ9 M) in the pro-opiomelanocortin gene (30). The K d value of GR ϩ MR was significantly lower than that of MR or GR alone (p Ͻ 0.05). These results indicate that the binding affinity of MR/GR heterodimer to nMRE/GRE is significantly enhanced compared with homodimers, suggesting that MR/GR heterodimerization may play an important role in down-regulation of the 5-HT1A receptor gene.
MR and GR Coactivation for Glucocorticoid Transrepression at the 5-HT1A nGRE-The relative importance of MR, GR, or both receptors in transcriptional repression was assessed by cotransfection of MR, GR, or both expression plasmids with the 1186-luciferase 5-HT1A construct in COS-7 cells (Fig. 6A). A MR/GR molar ratio of 1:1 was found to be optimal for glucocorticoid transrepression (data not shown) and was used in these experiments. Dexamethasone and aldosterone had no effect on luciferase activity of the 1186-luc construct in the absence of MR or GR. Transrepression was greatest upon coactivation of both MR and GR (53% inhibition), less for MR alone (33% inhibition), and least for GR alone (26% inhibition). The mutated nGRE (1186 m-luc) was not responsive to corticosteroids in the presence of both MR and GR, confirming the results obtained in SN-48 cells (Fig. 1). The greater activity (53 versus 41% inhibition) of the GR/MR combination in COS-7 (Fig. 6A) versus SN-48 cells (Fig. 1A) may be due to higher levels of receptor expression in transfected COS-7 cells.
The activity of the nGRE to mediate glucocorticoid transrepression at a heterologous promoter was examined by placing one to three copies of the nGRE sequence upstream of the SV40 promoter in the pGL3 luciferase vector (Fig. 6, B and C). The SV40 promoter was unresponsive to corticosteroids, even in the presence of cotransfected MR and GR (data not shown). For each construct, no effect of corticosteroids was observed in the absence of cotransfected MR or GR, and the mutated nGRE was unresponsive to corticosteroids in the presence of both MR and GR, as observed for the 5-HT1A promoter. In each case, cotransfection with both MR/GR conferred significantly more corticosteroid responsiveness on the nGRE-containing constructs than that of transfection with MR or GR alone as seen for the 5-HT1A promoter (Fig. 6A). The presence of three copies of the nGRE did not enhance the corticosteroid responsiveness of the SV40 promoter (Fig. 6C), which is consistent with previous findings (30). Interestingly, the corticosteroid responsiveness of the 5-HT1A promoter (53% inhibition in COS-7 cells) was greater than for the SV40 promoter (41% inhibition). These results indicate that the nGRE independently confers glucocorticoid transrepression on a heterologous promoter, with preferential repression activity for coactivation of MR and GR compared with independent activation of either receptor.
To assess the importance of DNA binding in glucocorticoid action, the mutant L501P-GR, which incorporates a mutation in the zinc finger domain to inactivate DNA binding (37), was cotransfected with each of the reporter constructs (Fig. 6, A-C). This mutant GR was completely inactive in all cases, indicating that direct receptor-DNA interaction is required for glucocorticoidinduced repression at the nGRE of the 5-HT1A receptor gene.
Pharmacology and Receptor Dependence of the nGRE Response-The concentration dependence of Dex-or Ald-mediated dexamethasone-or aldosterone-mediated repression of the 3xnGRE/SV40 construct in COS-7 cells cotransfected with GR (Fig. 7A) or MR (Fig. 7B) 2, 4, 6, and 8) or 10 units/ reaction. (lanes 1, 3, 5, 7, and 9) at concentrations similar to those required for ligand binding to their respective receptors. The functional importance of coactivation of MR and GR in SN-48 and COS-7 cells was addressed further using a pharmacological approach (Fig. 7, C and D). Three treatment conditions were used to selectively activate, respectively, GR alone, MR alone, or both: dexamethasone ϩ MR antagonist spironolactone (Dex ϩ Spi); aldosterone (MR agonist) ϩ GR antagonist RU38486 (Ald ϩ RU); or Dex ϩ Ald. In SN-48 cells cotransfected with 1186-luc (Fig. 7C), treatment with Dex and Ald resulted in the greatest level of transrepression (40% of control), whereas the other treatments produced a smaller but significant repression. COS-7 cells cotransfected with MR and GR plasmids in an equal molar ratio, and 3xn-GRE/SV40 displayed glucocorticoid receptor selectivity similar to that in SN-48 cells (Fig. 7D). Corticosteroid-induced repression was greatest (40% of control) upon activation of both MR and GR, whereas activation of either receptor alone displayed significant less repression activity. Thus, coactivation of GR and MR is required for optimal transrepression at the nGRE in both neuronal SN-48 cells and COS-7 cells. DISCUSSION A Novel nGRE Regulates the 5-HT1A Receptor Gene-The level of expression of the 5-HT1A receptor has been implicated in mental illnesses such as major depression, anxiety, and related disorders. For example, gene knockout of the 5-HT1A receptor gene results in mice with increased anxiety-related behaviors (3)(4)(5). On the other hand, depressed suicides show elevated levels of 5-HT1A autoreceptors compared with nondepressed suicides (46). Dysregulation of glucocorticoids is observed in a large proportion of depressed patients, suggesting a link between regulation of the 5-HT1A receptor by glucocorticoids and the etiology or maintenance of depression.
One of the most powerful regulators of 5-HT1A receptor expression in the rat brain is corticosterone, which rapidly and completely inhibits adrenalectomy-induced expression of the 5-HT1A receptor gene. Transcriptional repression of the 5-HT1A receptor gene by glucocorticoids has been demonstrated in the rat hippocampus, septum, and frontal cortex (20,21,23). However, the precise site of glucocorticoid regulation at the 5-HT1A receptor gene has yet to be clarified. We have identified a novel nGRE that mediates direct transcriptional repression of the rat 5-HT1A receptor gene by corticosteroids in 5-HT1A-expressing septal SN-48 or raphe RN46A neuronal cells. The 5-HT1A promoter activity was reduced by 40 -60%, consistent with the extent of inhibition observed at other nGRE elements (29,31). Importantly, we demonstrate that transrepression of the 5-HT1A promoter or heterologous promoters is mediated by the nGRE and is transduced preferentially by MR over GR, but transrepression is strongest upon coactivation of both receptors. Thus the nGRE of the 5-HT1A receptor gene represents the first nMRE/GRE to be identified. The ligandand receptor-dependent properties of this novel nGRE correspond well with the known properties of regulation of the 5-HT1A receptor transcription in limbic areas of the brain in vivo. In particular, both pharmacologic and gene knockout studies have provided evidence that MR is involved in corticosteroid-induced repression of 5-HT1A transcription. For example, low concentrations of corticosteroids or synthetic ligands that selectively activate MR are sufficient to reverse the adrenalectomy-induced increase in 5-HT1A receptor level in the  5, 6, and 7) or with preimmune rabbit serum (lane 8), as indicated. B, the expression of MR and GR in SN48 cells were detected by Western blot (upper panel). Nuclear protein was extracted from SN48 cells and electrophoresed (50 g/lane) on sodium lauryl sulfate-containing 10% polyacrylamide gel, and MR or GR were detected in parallel with separate rabbit polyclonal antibodies (1/1000 dilution each). The intensity of bands was quantified by Microcomputer Imaging Device (Imaging Research Inc., St. Catherines, Ontario, Canada). Data represent the mean Ϯ S.E. of three independent experiments. C, dexamethasone enhances protein-nGRE complex formation. Nuclear extracts (10 g) were made from SN-48 cells pretreated with different concentrations of dexamethasone as indicated and analyzed by EMSA (inset). Quantification for nGRE binding data (inset) is presented as a semi-logarithmic plot of relative optical density of the nGRE-protein complex versus dexamethasone concentration. Each point represents the mean Ϯ S.E. of three independent experiments. The specific nGRE-protein complex increased upon pretreatment with dexamethasone.
Glucocorticoid-induced repression of the rat 5-HT1A receptor gene at the nGRE proceeded via direct protein-DNA interaction as opposed to indirect mechanisms involving protein-protein interactions (29). We demonstrate that both GR and MR bind to the nGRE and that mutations in the nGRE or the receptor that disrupt this interaction block repression. In particular, the point mutant L501P-GR, which lacks DNA binding capability, failed to mediate or alter GR/MR-induced transrepression. This is in contrast to an indirect mechanism proposed for corticosterone-mediated inhibition of NFB-induced, but not basal, transcription of the 5-HT1A receptor gene (48). In COS-1 cells, cotransfection of 5-HT1A reporter constructs with the p50 or p65 subunit of NFB activated transcription by over 5-fold and was mediated by two NFB elements located at Ϫ365 and Ϫ64 bp of the 5-HT1A receptor gene. Transcription induced by NFB was inhibited 50% by glucocorticoids upon cotransfection of GR, but not MR, which is different from the dependence on MR and GR observed in the hippocampus. The NF⌲B-dependent indirect mechanism differed from direct transrepression at the nGRE, which suppressed basal transcription of the 5-HT1A receptor promoter in neuronal and non-neuronal cells via activation of both MR and GR. The importance of the NFB-dependent mechanism was not assessed under basal conditions or in neuronal cells (48). In the present studies, constructs that contained both NFB sites but lacked the nGRE element (e.g. 1.145-, 1.098-, 995-, 800-and 618-luc constructs, Fig. 1A) did not mediate dexamethasone-induced inhibition of basal transcription of the 5-HT1A promoter. The results obtained with nGRE mutants 1519 m and 1186 m also rule out participation in corticosteroid action of an adjacent nGRE-like sequence of the 5-HT1A promoter (at Ϫ1149 to Ϫ1134 bp). Furthermore, inclusion of the nGRE element upstream of heterologous promoters conferred corticosteroid-induced inhibition by both GR and MR receptors, the same receptor preference observed for 5-HT1A gene regulation in hippocampal neurons. Thus, direct glucocorticoid-induced transrepression at the nGRE is likely to represent the dominant mechanism of basal regulation of the 5-HT1A gene.
The NFB-dependent indirect pathway of glucocorticoid regulation may provide a complementary mechanism to regulate 5-HT1A receptor expression, especially in fibroblast or immune cells where NFB is implicated in the stimulation of 5-HT1A receptor expression (49). For example, in non-neuronal Chinese hamster ovary cells, NFB mediates 5-HT1A agonist-induced expression of the 5-HT1A receptor (50), regulation that is opposite to the negative regulation of the 5-HT1A receptor gene by agonists in hippocampal neurons (51). In B and T lymphocytes, where mitogens induce NFB-dependent transcriptional activation of the 5-HT1A promoter (49,52,53), antagonism of NFB action by glucocorticoids could inhibit mitogen action.
Heterodimerization of MR and GR at the nGRE-Both MR and GR bound directly to the 5-HT1A nGRE, and when both were present formation of a heterodimeric complex was fa- vored. Consistent with this finding, the affinity for the nMRE/ GRE of the MRϩGR combination was significant higher than that of GR or MR alone. Corticosteroid repression of transcriptional activity of the rat 5-HT1A receptor gene proceeded through a mechanism involving preferential coactivation of both MR and GR. Cooperative interactions between MR and GR at a positive GRE have been reported in the nervous system (54), in which MR and GR activate transcription synergistically through heterodimer formation. It has been proposed that glucocorticoid receptor heterodimerization may play a crucial role in glucocorticoid action in the brain, particularly in tissues (e.g. hippocampus) that express MR and GR (45). The high affinity of MR for corticosterone would allow for high sensitivity to the actions of glucocorticoids in regions that express this receptor (55). The strong expression of 5-HT1A receptors in MR-enriched hippocampal CA1 pyramidal cells and the cooperative interactions between MR and GR at the 5-HT1A nGRE strongly suggest that heterodimerization is the key mechanism for inhibitory regulation by glucocorticoids of the 5-HT1A receptor gene in the brain.
Structure and Conservation of the nGRE-Negative regulation of gene transcription by glucocorticoids involves repression via an nGRE composed of one GRE half-site (56,57) or three half-sites (30,58). An nGRE-like element has been found in several genes including the following: the pro-opiomelanocortin gene (30,59,60), the corticotrophin-releasing hormone gene (58), the gonadotropin-releasing hormone gene (32), the human interleukin-1␤ gene (61), and the bovine prolactin gene (33,56). In the present study, the GRE-like sequence has a repeated 6-bp GRE half-site (TGTCCT) separated by 6 nucleotides, which is conserved (uppercase) in both human (Ϫ1249 TGTCCT-TTgnnn-TGTCCTTTA) and mouse (Ϫ1158 TGTCCT-ccAnnn-TGTCCTTTc) 5-HT1A receptor genes. Note that in all cases the direct repeat GRE half-site (TGTCCT) and 6-bp gap are absolutely conserved. Although the nGRE contains two GRE-like half-sites, suggesting the formation of receptor dimers, the exact number of receptors involved in the present interaction needs to be elucidated by x-ray crystallography. Nevertheless, a theoretical consideration of the topology of GR homodimer binding to the nGRE, based on the crystal structure of GR homodimer binding to a consensus GRE (62), suggests that two receptors would bind in a head-to-tail conformation on the same face of the double helix. The critical parameter for this arrangement is the 6-nucleotide spacing between the two conserved GRE half-sites, which is sufficient for a complete 360°rotation of the two sites into alignment. The absolute conservation among 5-HT1A gene homologues of the 6-bp spacing between two GRE half-sites suggests that this is a crucial element of the nGRE. Indeed, elimination of 3 bp of the 6-bp gap in the 5-HT1A nGRE to rotate the receptor dimer 180°converts the negative GRE into a positive GRE that mediates glucocorticoid-induced gene transcription of 3-4-fold the The indicated reporter constructs were cotransfected with expression plasmids MR or GR alone or MR/GR together as indicated. The reporter constructs used were: A, 1186-luc or 1186 m-luc, constructs of the 5-HT1A gene, as indicated; B, 1 ϫ nGRE/SV40, containing the 5-HT1A nGRE or mutant nGRE located upstream of the SV40 promoter; and C, 3 ϫ nGRE/SV40, containing three tandem copies of the 5-HT1A nGRE or mutant nGRE located upstream of the SV40 promoter (see "Materials and Methods"). Cells were transfected with equimolar concentrations of MR, GR, both (or neither), or DNA binding mutant L501P-GR and treated for 48 h with 20 nM Ald, 100 nM Dex, or both Ald and Dex, respectively. The activity of SV40 vectors cotransfected with MR/GR was not altered by treatment with Ald and Dex (not shown). Note that mutated nGRE or receptor (L501P-GR) failed to mediate corticosteroid repression. Data represent percent inhibition compared with vehicle control and are plotted as mean Ϯ S.E. of at least six independent transfections done with triplicate samples. *, p Ͻ 0.05 and **, p Ͻ 0.02 compared with control. basal level in SN48 or GR-transfected COS-7 cells (data not shown).
In conclusion, we have identified a novel nMRE/GRE that demonstrates the crucial role of MR/GR heterodimers in mediating responses to corticosteroids, not only for transcriptional activation but also for transcriptional repression. This is consistent with the general importance of heterodimerization of the steroid receptor family of nuclear receptors (44,63). Furthermore, our results provide a plausible mechanism for the sensitivity of the rat 5-HT1A receptor gene to corticosteroidinduced transrepression, particularly in the hippocampus where MR and GR are both expressed.