The Dominant Negative Activity of the Human Glucocorticoid Receptor β Isoform

Alternative splicing of the human glucocorticoid receptor gene generates a nonhormone binding splice variant (hGRβ) that differs from the wild-type receptor (hGRα) only at the carboxyl terminus. Previously we have shown that hGRβ inhibits the transcriptional activity of hGRα, which is consistent with reports of elevated hGRβ expression in patients with generalized and tissue-specific glucocorticoid resistance. The potential role of hGRβ in the regulation of target cell sensitivity to glucocorticoids prompted us to further evaluate its dominant negative activity in other model systems and to investigate its mode of action. We demonstrate in multiple cell types that hGRβ inhibits hGRα-mediated activation of the mouse mammary tumor virus promoter. In contrast, the ability of the progesterone and androgen receptors to activate this promoter is only weakly affected by hGRβ. hGRβ also inhibits hGRα-mediated repression of an NF-κB-responsive promoter but does not interfere with homologous down-regulation of hGRα. We show that hGRβ can associate with the heat shock protein hsp90 although with lower affinity than hGRα. In addition, hGRβ binds GRE-containing DNA with a greater capacity than hGRα in the absence of glucocorticoids. Glucocorticoid treatment enhances hGRα, but not hGRβ, binding to DNA. Moreover, we demonstrate that hGRα and hGRβ can physically associate with each other in a heterodimer. Finally, we show that the dominant negative activity of hGRβ resides within its unique carboxyl-terminal 15 amino acids. Taken together, our results suggest that formation of transcriptionally impaired hGRα-hGRβ heterodimers is an important component of the mechanism responsible for the dominant negative activity of hGRβ.

Alternative splicing of the human glucocorticoid receptor gene generates a nonhormone binding splice variant (hGR␤) that differs from the wild-type receptor (hGR␣) only at the carboxyl terminus. Previously we have shown that hGR␤ inhibits the transcriptional activity of hGR␣, which is consistent with reports of elevated hGR␤ expression in patients with generalized and tissue-specific glucocorticoid resistance. The potential role of hGR␤ in the regulation of target cell sensitivity to glucocorticoids prompted us to further evaluate its dominant negative activity in other model systems and to investigate its mode of action. We demonstrate in multiple cell types that hGR␤ inhibits hGR␣-mediated activation of the mouse mammary tumor virus promoter. In contrast, the ability of the progesterone and androgen receptors to activate this promoter is only weakly affected by hGR␤. hGR␤ also inhibits hGR␣-mediated repression of an NF-B-responsive promoter but does not interfere with homologous down-regulation of hGR␣. We show that hGR␤ can associate with the heat shock protein hsp90 although with lower affinity than hGR␣. In addition, hGR␤ binds GRE-containing DNA with a greater capacity than hGR␣ in the absence of glucocorticoids. Glucocorticoid treatment enhances hGR␣, but not hGR␤, binding to DNA. Moreover, we demonstrate that hGR␣ and hGR␤ can physically associate with each other in a heterodimer. Finally, we show that the dominant negative activity of hGR␤ resides within its unique carboxyl-terminal 15 amino acids. Taken together, our results suggest that formation of transcriptionally impaired hGR␣-hGR␤ heterodimers is an important component of the mechanism responsible for the dominant negative activity of hGR␤.
The ability of both natural and synthetic glucocorticoids to act on a target tissue and elicit specific biological responses is dependent on the presence of the ␣ isoform of the glucocorticoid receptor (GR␣). 1 GR␣ belongs to the superfamily of steroid/ thyroid/retinoic acid receptor proteins that function as liganddependent transcription factors (1). This family also includes receptors for vitamin D and a large group of proteins termed orphan receptors whose ligands and/or functions are unknown. Like other members of this receptor superfamily, GR␣ is composed of an amino-terminal domain that is involved in gene activation, a central DNA binding domain, and a carboxylterminal hormone binding domain (HBD).
GR␣ is expressed in almost all tissues and cells, and in the absence of hormone it resides in the cytoplasm of cells as part of a large multiprotein complex. This complex consists of the receptor polypeptide, two molecules of the heat shock protein hsp90, and several additional proteins (2,3). When hormone binds the receptor, a conformational change ensues, resulting in the dissociation of hsp90 and the other associated proteins. In its new conformation, GR␣ translocates into the nucleus, where it binds as a homodimer to glucocorticoid-responsive elements (GREs) located in the regulatory regions of target genes. GR␣ then communicates with the basal transcription machinery and, depending on the GRE sequence and promoter context, either positively or negatively regulates expression of the linked gene. The receptor can also modulate gene expression apart from DNA binding by physically interacting with other transcription factors such as AP-1 and NF-B (4 -7).
Sensitivity to glucocorticoids varies between tissues and even within the same tissue during different stages of development (8,9). In addition, the beneficial effects of glucocorticoids in the treatment of many immune and inflammatory diseases is often limited by the development of glucocorticoid resistance in the diseased tissue (10). The molecular basis for these variations in glucocorticoid responsiveness is poorly understood. As the sole effector molecule in the glucocorticoid signaling cascade, GR␣ is the primary target for regulatory events that modulate target cell sensitivity to glucocorticoids. Changes in GR␣ expression and/or the potency with which GR␣ functions as a ligand-dependent transcription factor will elicit corresponding changes in glucocorticoid responsiveness. Because glucocorticoids profoundly influence all aspects of biological function and are extensively used as therapeutic agents in the treatment of many diseases and disorders, understanding the factors that regulate GR␣ expression and/or activity, and hence glucocorticoid responsiveness, is an important goal of current research.
Alternative splicing of the human GR (hGR) gene produces a splice variant termed hGR␤. hGR␤ differs from the wild-type receptor (hGR␣) only at the carboxyl terminus (11)(12)(13). The two isoforms are identical through amino acid 727 but then diverge, with hGR␣ having an additional 50 amino acids and hGR␤ an additional, nonhomologous 15 amino acids. hGR␤ has been detected in many different tissues (13)(14)(15)(16)(17). Within tissues, hGR␤ is most abundant in certain epithelial cells (16). In contrast to hGR␣, hGR␤ resides in the nucleus of cells independent of glucocorticoid treatment and does not bind glucocorticoids or activate glucocorticoid-responsive genes (11,13,16,18). We and others have shown that hGR␤ functions as a dominant negative inhibitor of hGR␣ on both complex and simple glucocorticoid-responsive promoters (13,14).
The ability of hGR␤ to antagonize the function of hGR␣ suggests that hGR␤ will play a critical role in the regulation of target cell sensitivity to glucocorticoids. Indeed, recent studies have demonstrated that hGR␤ expression is elevated in patients suffering from generalized and tissue-specific glucocorticoid resistance (19,20). In the present study, we investigate the ability of hGR␤ to repress hGR␣ in multiple cell types and evaluate how much hGR␤ (relative to hGR␣) is necessary for the observed repression. hGR␤'s dominant negative activity is further explored on other closely related steroid hormone receptors and on genes negatively regulated by glucocorticoids. We also assess the ability of hGR␤ to associate with hsp90, bind GRE-containing DNA, and heterodimerize with hGR␣ in order to understand the mechanism responsible for the dominant negative activity of hGR␤. Finally, we investigate whether the hGR␤-specific amino acids mediate this repression of hGR␣ function.

EXPERIMENTAL PROCEDURES
Materials-Dexamethasone (DEX) and progesterone were purchased from Steraloids (Wilton, NH), and the synthetic androgen R1881 was obtained from NEN Life Science Products. [ 14 C]Chloramphenicol (40 -60 mCi/mmol) was obtained from NEN Life Science Products, and [␣-32 P]UTP and dCTP (3000 Ci/mmol) were supplied by ICN Radiochemicals (Irvine, CA). Acetyl coenzyme A and protease inhibitors were obtained from Roche Molecular Biochemicals. The peroxidase-labeled anti-rabbit and anti-mouse secondary antibodies and the chemiluminescent detection reagents were purchased from Amersham Pharmacia Biotech. Dithiobis(succinimidyl propionate) (DSP) was from Pierce. The TNT Coupled Reticulocyte Lysate Translation System was from Promega (Madison, WI).
Recombinant Plasmids-Construction of pCMVhGR␣ and pCM-VhGR␤ expression vectors as well as the human cytomegalovirus major intermediate early gene promoter expression vector backbone pCMV5 have been described previously (13). Plasmids pGST-hGR␣HBD and pGST-hGR␤HBD were generated by polymerase chain reaction cloning and express the hGR␣ HBD (amino acids 523-777) and hGR␤ HBD (amino acids 523-742), respectively, fused to the glutathione S-transferase (GST) protein. Site-directed mutagenesis by polymerase chain reaction was employed to generate the truncated receptor pCMVhGR728T. Nucleotides GTG, encoding valine at amino acid 728 of hGR␣, were replaced with TAG, which mutates the valine to a stop codon. The expressed hGR728T protein was recognized by the antipeptide hGR antibody 57 but not by the hGR␣-specific antibody AShGR nor the hGR␤-specific antibody BShGR (16,21,22). All constructs were confirmed by DNA sequencing. The expression vector CMV3.1AR (pC-MVhAR) encodes the human androgen receptor and was obtained from Dr. M. J. McPhaul (University of Texas Southwestern) (23). The expression vector phPR-B, kindly provided by Dr. D. P. McDonnell (Duke University), encodes the B form of the human progesterone receptor and is driven by the SV40 enhancer (24). The expression vector for the transcriptionally active p65 subunit of NF-B (pCMVp65) and the NF-B-responsive reporter 3XMHCCAT (MHC-CAT) were obtained from Dr. A. S. Baldwin (University of North Carolina, Chapel Hill, NC) (7). Plasmids pHHluc and pGMCS contain the mouse mammary tumor virus (MMTV) promoter cloned upstream of the luciferase and CAT genes, respectively (25,26). Plasmids pT7/T3-hGR␣ and pT7/T3-hGR␤, used for in vitro synthesis of the hGR␣ and hGR␤ proteins, have been described previously (16,27).
Cell Culture and Transfections-COS-1, CV-1, and COS-7 cells were grown in Dulbecco's minimum essential medium supplemented with 2 mM glutamine and 10% (v/v) of a mixture (1:1) of heat-inactivated fetal calf and calf serum. All cultures were maintained in a 5% CO 2 humidified atmosphere at 37°C and passaged every 3-4 days. For transfection of cells, medium was removed from subconfluent cells and replaced with fresh Dulbecco's minimum essential medium containing 3% serum. Plasmid DNA was prepared as a calcium phosphate precipitate, and the total amount of DNA in each transfection was kept constant by the addition of empty vector (pCMV5) and salmon sperm DNA. The precipitates were incubated with cells for 5 h. The medium was then removed, and the cells were shocked for 30 s with 15% glycerol and then refed with fully supplemented medium stripped of endogenous steroids.
Luciferase Assays-Cells were transfected as described above and in the appropriate figure legends. Immediately after the transfection, DEX or vehicle was added to the cells, which were then incubated an additional 18 h. Cells were harvested, lysates were prepared, and luciferase assays were performed according to the manufacturer's instructions (Analytical Luminescence Laboratory, San Diego, CA). Luciferase activity was calculated per g of protein for each sample.
CAT Assays-Cells were transfected as described above and in the appropriate figure legend. Immediately after the transfection, the appropriate steroid (DEX, progesterone, or R1881) or vehicle was added to the cells. After an 18-h incubation, cells were harvested, and CAT assays were performed as described previously (13). Briefly, a lysate was prepared and inactivated by heating at 68°C for 6 min. Equivalent amounts of protein were adjusted to 156 mM Tris, 1 mM acetyl coenzyme A, and 0.1 Ci of [ 14 C]chloramphenicol in a final volume of 150 l. The CAT reaction was then allowed to proceed for 16 h at 37°C. Samples were applied to a thin layer chromatography plate and chromatographed in chloroform/methanol (95:5). After autoradiography, CAT activity was quantitated by excising the appropriate area from the thin layer chromatography plate and counting the [ 14 C]chloramphenicol and acetylated derivatives in a Beckman LS 7000 scintillation counter.
Northern Blotting-Poly(A) ϩ RNA (5 g) from transfected cells was isolated, electrophoresed, and transferred as described previously (29). After transfer, the RNA was UV-cross-linked to the membrane and subsequently stained with methylene blue (0.04% methylene blue, 0.5 M NaOAc, pH 5.2) to examine both the integrity of the RNA and the completeness of the transfer. Membranes were prehybridized (3 h at 65°C) and hybridized (18 h at 65°C) in 50% formamide, 5ϫ saline/ sodium phosphate/EDTA, 5ϫ Denhardt's solution, 2% SDS, 200 g/ml yeast RNA, and 200 g/ml denatured, sheared salmon sperm DNA. The [␣-32 P]UTP-labeled hGR␣ or hGR␤-specific cRNA probes, which have been described previously (13), were included in the hybridization fluid (1 ϫ 10 6 cpm/ml). Following hybridization, blots were washed once at room temperature and four times at 65°C in 0.1ϫ saline/sodium phosphate/EDTA, 0.1% SDS. To control for loading differences in RNA, membranes were stripped (30-min incubation in 0.1ϫ saline/sodium phosphate/EDTA, 0.1% SDS heated to 100°C) and reprobed with a ␤-actin cRNA probe. The hGR␣ and hGR␤ mRNA signals were quanti-tated densitometrically using NIH Image analysis software and normalized to ␤-actin mRNA levels.
Immunoprecipitation of Receptor-hsp90 Complexes-For immunoprecipitation of receptor-hsp90 complexes from whole cells, transfected cells were lysed using a Dounce homogenizer in HEPES buffer (10 mM HEPES, pH 7.4, 1 mM EDTA, 20 mM sodium molybdate) containing protease inhibitors. The homogenate was centrifuged at 165,000 ϫ g in a Beckman 50 Ti rotor for 1 h at 4°C, and the supernatant was collected. Proteins (200 g) were precleared with preimmune serum and protein A-Sepharose for 1 h at 4°C. After centrifugation, the supernatant was removed and transferred to a new tube. The antipeptide hGR antibody 59 (1:50 dilution) or preimmune serum was added and each tube incubated for 2 h at 4°C with rotation (21). For binding immune complexes, 80 l of protein A-Sepharose beads (Sigma) were added, and the incubation continued an additional 30 min at 4°C with rotation. Protein A-Sepharose was prepared in TEGM buffer (10 mM TES, pH 7.6, 4 mM EDTA, 50 mM NaCl, 10% glycerol, 20 mM sodium molybdate). The protein A-Sepharose immune complexes were washed four times with 1 ml of TEGM buffer and then resuspended in sample buffer containing 10% glycerol, 2% SDS, 0.2 mg/ml bromphenol blue, 62.5 mM Tris-HCl, pH 6.8, and 5% 2-mercaptoethanol. Immunoprecipitated proteins were eluted from the protein A-Sepharose by boiling 5 min and resolved on SDS-polyacrylamide gels. After electrophoretically transferring the proteins to nitrocellulose, immunoblotting was carried out as described above. Immunoprecipitation of receptor-hsp90 complexes from reticulocyte lysates expressing equivalent amounts of 35 Slabeled hGR␣ or 35 S-labeled hGR␤ was performed using the anti-hsp90 monoclonal antibody 3G3 (Affinity BioReagents, Golden, CO) essentially as described (30).
DNA Binding Analysis of in Vitro Translated hGR␣ and hGR␤-DNA binding analysis was carried out essentially as described previously (27). The hGR␣ and hGR␤ proteins were synthesized in vitro using reticulocyte lysates. Translation products (5 fmol) were incubated with an equal volume of DNA binding buffer (20 mM NaHPO 4 , pH 7, 10% glycerol, 50 mM NaCl, 1 mM EDTA, and 2 mM 2-mercaptoethanol) for 2 h at 0°C with or without 200 nM DEX. Lysates were then heatactivated for 30 min at 25°C, after which they were transferred to tubes coated with 1% BSA. An 868-base pair ClaI/SphI fragment containing the MMTV long terminal repeat from plasmid pLTR190 or a 777-bp PstI/ClaI fragment from the plasmid pBR322 was labeled on one end with [␣-32 P]dCTP using Klenow to fill in the ClaI-generated 5Ј overhang. Labeled DNA fragments (10-fold molar excess) and nonspecific DNA (poly(dI-dC) at a 50-fold excess and Escherichia coli DNA at a 50-fold excess) were added to the lysates, and the incubation continued at 0°C for 2 h. For some experiments, unlabeled competitor DNA was added at a 50-fold excess. Antipeptide hGR antibody 57 (1:50 dilution) was then added to the lysates, and the incubation continued for 2 h at 0°C. Protein A-Sepharose (133 mg/ml DNA binding buffer) was then added to the lysates for an additional 1-h incubation, and the pellets were subsequently washed four times with 1 ml of DNA binding buffer. DNA fragments were recovered from the pellets by phenol chloroform extraction and analyzed on 8 M urea, 4% polyacrylamide gels. Gels were fixed in 30% methanol, 10% acetic acid for 15 min, dried for 1 h under vacuum at 60°C, and processed for autoradiography.
Immunoprecipitation of hGR␣-hGR␤ Heterodimers-The hGR␣ and hGR␤ proteins were synthesized in vitro using reticulocyte lysates. Thirty-five l of 35 S-labeled hGR␣ translation mix were added to 35 l of unlabeled hGR␤ translation mix or 35 l of unprogrammed translation mix and incubated for 30 min at room temperature in 20 mM HEPES, pH 7.9, 50 mM KCl, 2.5 mM MgCl 2 , 1.0 mM dithiothreitol, 200 nM DEX, and protease inhibitors. An aliquot was then removed from each tube to verify that equivalent amounts of hGR␣ were present in each mixture. Proteins were cross-linked by the addition of 2.5 mM DSP for 10 min at room temperature. Reactions were then terminated by the addition of 0.1 M ethanolamine. The cross-linked proteins were diluted in immunoprecipitation buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM EDTA, 0.5% sodium deoxycholate, 0.5% Nonidet P-40, 0.1 M ethanolamine, and protease inhibitors), precleared, and immunoprecipitated with the hGR␤-specific antibody BShGR (16). Immune complexes were recovered, washed extensively in the immunoprecipitation buffer, and resuspended in Laemmli buffer with or without ␤-mercaptoethanol. After proteins were resolved on SDS-polyacrylamide gels, gels were fixed in 30% methanol-10% acetic acid for 30 min, dried for 1 h under vacuum at 60°C, and processed for autoradiography using liquid EN 3 HANCE.
GST Pull-down Assays-GST pull-down assays were carried out essentially as described (31). In brief, the GST-hGR␣HBD and GST-hGR␤HBD fusion proteins were induced in DH5␣ E. coli by the addition of 0.1 mM isopropyl ␤-D-thiogalactoside to log phase cells. Following a 4-h incubation, the cells were harvested and lysed by sonication in 10 mM Tris-HCl, pH 8.0, 1 mM EDTA, 100 mM NaCl, 10% glycerol, 0.2 mg/ml lysozyme, and protease inhibitors. Following centrifugation, the soluble protein extract was incubated with phosphate-buffered salinewashed glutathione agarose in the presence of 1% Triton X-100 and 1 mM dithiothreitol. The immobilized fusion proteins were then incubated with full-length, in vitro translated 35 S-labeled hGR␣ or 35 S-labeled hGR␤ treated with or without 1 M DEX. Assays were carried out at 4°C with constant rotation in binding buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, 100 mM NaCl, 10% glycerol, 1 mM dithiothreitol, 0.01% Nonidet P-40, and protease inhibitors). Recovered proteins were resolved on SDS-polyacrylamide gels and processed as described above.

RESULTS
The Dominant Negative Activity of hGR␤ Occurs in Multiple Cell Types When hGR␤ Is More Abundant than hGR␣-We have demonstrated previously in HeLa S 3 cells that hGR␤ can inhibit the transcriptional activity of hGR␣ in a dominant negative fashion (13). This early study, however, did not investigate whether the dominant negative activity occurred in other cell types; nor did it determine how much hGR␤ (relative to hGR␣) was needed for the observed repression. To address these issues, we initially analyzed the ability of hGR␤ to repress hGR␣-mediated activation of the glucocorticoid-responsive MMTV promoter in transfected COS-1 cells. COS-1 cells (which are devoid of endogenous GR) were transfected with a fixed amount of the MMTV-luciferase reporter pHHluc, a fixed amount of the hGR␣ expression vector pCMVhGR␣, and increasing amounts of the hGR␤ expression vector pCMVhGR␤ corresponding to a 5-and 10-fold molar excess over the amount of transfected pCMVhGR␣. The cells were then treated with or without hormone for 18 h and analyzed for luciferase activity. As increasing amounts of pCMVhGR␤ were transfected into the cells, the glucocorticoid-induced, hGR␣-mediated activation of the MMTV promoter was diminished in a dose-dependent manner (Fig. 1A). Transfection of a 5-and 10-fold molar excess of pCMVhGR␤ resulted in a 36.7 and 63.7% decrease, respectively, in hGR␣ activity. Similar reductions in hGR␣ activity were also observed following a 48-h hormone treatment (data not shown).
Measuring the amount of hGR␣ and hGR␤ expressed in these cells with an antibody that recognizes both receptor isoforms is problematic because the 90-kDa hGR␤ protein comigrates with a 90-kDa hGR␣ degradation product (16). Therefore, the relative levels of the hGR␣ and hGR␤ proteins were assessed in the following manner. First, COS-1 cells were separately transfected with an equimolar amount of pCMVhGR␣ or pCMVhGR␤. A Western blot was performed on these cells with the anti-peptide hGR antibody 57, which recognizes the same epitope (amino acids 346 -367) in both receptor isoforms (Fig. 1B, lanes 1-2). hGR␤ was expressed at approximately 80% the level of hGR␣. Second, COS-1 cells were co-transfected with a fixed amount of pCMVhGR␣ and various amounts of pCM-VhGR␤ corresponding to a 1-, 5-, or 10-fold molar excess. Western blots were performed on these cells with isoform-specific antibodies (Fig. 1B, lanes 3-5). hGR␣ levels remained constant in these cells, and hGR␤ levels increased the expected 5-and 10-fold when a 5-and 10-fold molar excess of pCMVhGR␤ was included in the transfection mixture. Taken together, these blots indicate that the dose-dependent reduction in hGR␣ activity observed for COS-1 cells in Fig. 1A occurred when hGR␤ was expressed at levels 4-and 8-fold greater than hGR␣.
The ability of hGR␤ to function as a dominant negative inhibitor of hGR␣ was evaluated in two additional cell lines devoid of endogenous GR. COS-7 and CV-1 cells were each transfected with a fixed amount of pHHluc, a fixed amount of pCMVhGR␣, and a 5-or 10-fold molar excess of pCMVhGR␤. The cells were then treated with our without glucocorticoids for 18 h and analyzed for luciferase activity. Expression of hGR␤ inhibited hGR␣-mediated activation of the MMTV promoter in both cell types (Fig. 1C). The transcriptional activity of hGR␣ was reduced 55.6 and 50.4% in COS-7 cells and 25.6 and 69.0% in CV-1 cells when a 5-and 10-fold molar excess, respectively, of pCMVhGR␤ was included in the transfection mixture. Similar reductions in hGR␣ activity were also observed following a 48-h hormone treatment (data not shown). Western blots, performed exactly as described above for COS-1 cells, again showed that hGR␤ was expressed at levels approximately 4and 8-fold greater than hGR␣ (data not shown). Thus, the dominant negative activity of hGR␤ appears to be a general phenomenon that occurs in multiple cell types when hGR␤ is more abundant than hGR␣.
The Dominant Negative Activity of hGR␤ Is Selective for hGR␣-hGR␣ is a member of a subfamily of receptors that includes the mineralocorticoid receptor (MR), progesterone receptor (PR), and androgen receptor (AR). The DNA-binding domains of these receptors are highly conserved. As a consequence, each receptor activates the MMTV promoter by binding the same hormone response element (HRE) (32). Our observation that hGR␤ represses the transcriptional activity of hGR␣ on the MMTV promoter raised an important question. Does hGR␤ also inhibit the ability of these other closely related steroid hormone receptors to activate the MMTV promoter? To answer this question, COS-1 cells were transfected with a fixed amount of the MMTV-CAT reporter pGMCS; a fixed amount of the expression vector for hGR␣, the B-form of the human PR (hPRB), or the human AR (hAR); and a 10-fold molar excess of pCMV5 or pCMVhGR␤. The cells were then treated with or without the appropriate hormone for 18 h. Expression of hGR␤ inhibited the activity of hGR␣ by 60% in cells transfected with pCMVhGR␣ ( Fig. 2A, left panel). In cells transfected with the more weakly expressing pRSVhGR␣, expression of hGR␤ produced an even greater 75% reduction in the transcriptional activity of hGR␣ ( Fig. 2A, right panel). However, hGR␤ had no effect on the transcriptional activity of hPRB, and only a small inhibition (27%) was observed for hAR (Fig. 2, B and C). These results suggest that the dominant negative activity of hGR␤ is selective for hGR␣.
The Dominant Negative Activity of hGR␤ Occurs on Some Genes Negatively Regulated by hGR␣-hGR␣ enhances the expression of some genes, and it represses the expression of others. However, it is not known whether the dominant negative activity of hGR␤ occurs on genes negatively regulated by hGR␣. One gene that is repressed by hGR␣ in almost all cell types is the hGR gene itself by a process termed homologous down-regulation (33). In COS-1 cells expressing only hGR␣, the hGR␣ mRNA was down-regulated to approximately 20% of control levels following a 24-h glucocorticoid treatment (Fig.   FIG. 1. The dominant  3A). In contrast, no reduction in hGR␤ mRNA was observed in cells expressing only hGR␤ (Fig. 3A). The ability of hGR␤ to antagonize hGR␣-mediated down-regulation was assessed in COS-1 cells transfected with a fixed amount of pCMVhGR␣ and increasing amounts of pCMVhGR␤ corresponding to a 5and 10-fold molar excess. After treating the cells with or without glucocorticoids for 24 h, Western blots were performed with isoform-specific antibodies. Expression of hGR␤ (at levels approximately 4-and 8-fold higher than hGR␣) had no effect on the magnitude of hGR␣ down-regulation (Fig. 3B). hGR␣ not only fully down-regulated its own expression but also partially down-regulated the expression of hGR␤ (Fig. 3B). This partial down-regulation of hGR␤ was also observed at the mRNA level in cells expressing equivalent amounts of the two receptor isoforms (Fig. 3A).
hGR␣ also represses gene expression apart from DNA binding by physically associating with other transcription factors. For example, hGR␣ physically associates with NF-B, an activator of a broad class of immune system genes, and prevents NF-B from activating target genes (7). This is illustrated in Fig. 4A, where ligand-bound hGR␣ blocked the constitutively active p65 subunit of NF-B from activating the NF-B-responsive reporter MHC-CAT in COS-1 cells. The DNA-binding domain of hGR␣ is required for this repression (34). hGR␤ has an intact DNA-binding domain, but this receptor isoform did not repress the transcriptional activity of NF-B following hormone treatment (Fig. 4A). In addition, hGR␤ did not constitutively inhibit NF-B, since similar absolute amounts of CAT activity were measured in the pCMV5-and pCMVhGR␤-transfected cells (data not shown). We next investigated whether hGR␤ could inhibit hGR␣'s antagonism of NF-B. hGR␣ repressed the transcriptional activity of NF-B to approximately 15% of control levels in the absence of co-expressed hGR␤ (Fig.  4B). This response was reduced to 33% of control levels in cells transfected with a 5-fold molar excess of pCMVhGR␤ (Fig. 4B). However, no further inhibition was observed with greater amounts of hGR␤ (up to 10-fold), suggesting that the efficacy of hGR␤ as a dominant negative inhibitor is greater on genes activated by hGR␣ than on genes repressed by hGR␣.
hGR␤ Associates with hsp90 -The ability of hGR␤ to repress the transcriptional activity of hGR␣ requires hGR␤ to be expressed at levels 4 -8-fold greater than hGR␣. This requirement may indicate that a portion of the expressed hGR␤ molecules are held in an inactive state. In the absence of hormone, hGR␣ is held in an inactive state that cannot dimerize nor bind DNA by its association with the heat shock protein hsp90 (2, 3). To test whether hGR␤ associates with hsp90, we transfected COS-1 cells with equimolar amounts of pCMVhGR␣, pCM-VhGR␤, or the expression vector backbone pCMV5 (mock). Immunoprecipitations were performed with preimmune serum or the antipeptide hGR antibody 59, which recognizes an epitope common to both receptor isoforms. The immunoprecipitated proteins were then analyzed on a Western blot with the hsp90 antibody AC88. AC88 detected the 90-kDa hsp90 protein in the immunoprecipitates prepared from the hGR␣-and hGR␤-transfected cells but not the mock-transfected cells (Fig.  5, upper panel). The blot was then stripped and reprobed with the antipeptide hGR antibody 57 to verify that hGR␣ and hGR␤ were immunoprecipitated at comparable levels (Fig. 5, lower  panel). These results indicate that hGR␤ can associate with hsp90 and suggest that a portion of the expressed hGR␤ molecules may be unavailable for repressing ligand-bound hGR␣.
hGR␤ Binds GRE-containing DNA-The molecular mechanisms responsible for hGR␤'s antagonism of hGR␣ are currently unknown. Cellular sensitivity to glucocorticoids is regulated by changes in receptor expression and/or alterations in the efficacy with which the receptor functions as a ligand-dependent transcription factor. Because overexpression of hGR␤ had no effect on the expression of hGR␣ (Fig. 1B), hGR␤ may interfere with the ability of hGR␣ to function as a ligand-dependent transcription factor. hGR␤ has an intact DNA-binding domain and is located in the nucleus of cells independent of hormone treatment (13,16). Therefore, hGR␤ might compete with hGR␣ for GRE binding. To test whether hGR␤ can bind a GRE-containing piece of DNA, equivalent amounts of in vitro translated hGR␣ or hGR␤ (Fig. 6D) were heat-activated in the presence of glucocorticoids and incubated with a 10-fold molar excess of a radiolabeled DNA fragment. The DNA fragment utilized was a GRE-containing MMTV promoter fragment or a similar-sized piece of DNA from pBR322 that does not contain a GRE. Receptor-DNA complexes were immunoprecipitated using the antipeptide hGR antibody 57, and the DNA recovered from the immune complexes was analyzed on a denaturing gel. Recovery of radiolabeled DNA is indicative of receptor binding. As shown in Fig. 6A, both hGR␣ and hGR␤ bound the MMTV fragment but not the pBR322 fragment. hGR␣ bound the MMTV fragment with a greater capacity than hGR␤ as results from four independent experiments revealed a 3-4-fold difference in DNA binding between the two receptor isoforms. Binding of hGR␣ and hGR␤ to the MMTV fragment was also specific (Fig. 6B). In the presence of a 50-fold excess of unlabeled MMTV fragment, binding above background was eliminated for both receptor isoforms. In contrast, no reduction in binding was observed for either isoform in the presence of a 50-fold excess of the pBR322 fragment (Fig. 6B). In addition, a 50% reduction in DNA binding was observed for both hGR␣ and hGR␤ when a 100-fold excess of a GRE oligonucleotide was included in the reaction, whereas a non-GRE oligonucleotide had no effect (data not shown).
The 3-4-fold difference in the capacity of hGR␣ and hGR␤ to bind the MMTV fragment (Fig. 6A) might reflect differences in the affinity of the two receptor isoforms for GREs and/or differences in the number of hGR␣ and hGR␤ molecules free from hsp90 and available to bind DNA. Removal of hsp90 is necessary for DNA binding to occur, and glucocorticoid binding triggers the dissociation of hGR␣-hsp90 complexes (2, 3). However,  control; t test). B, cells were transfected with 2.5 g of the MHC-CAT reporter, 2.5 g of the p65 subunit of NF-B, 2.5 g of pCMVhGR␣, and various amounts of pCMV5 and/or pCMVhGR␤. Molar ratios of transfected plasmids are indicated. Cells were processed as described in A. CAT activity is plotted as a percentage of control and represents the mean Ϯ S.E. of five independent experiments (*, p Ͻ 0.05 versus percentage repression in absence of pCMVhGR␤; t test). hGR␤ does not bind glucocorticoids (11,13). Therefore, we next compared the binding of hGR␣ and hGR␤ to the GRE-containing MMTV fragment after heat activation in the absence of glucocorticoids. Surprisingly, the capacity for hGR␤ to bind the MMTV fragment under these conditions was greater than that for hGR␣ (Fig. 6C, left panel). Four independent experiments revealed a 2.5-3.5-fold difference in DNA binding. Lysates prepared in parallel and activated by heat in the presence of glucocorticoids revealed no significant change in the capacity of hGR␤ to bind the MMTV fragment but a marked increase in hGR␣ binding, consistent with hormone-mediated dissociation of the hGR␣-hsp90 complexes but not hGR␤-hsp90 complexes (Fig. 6C, right panel). These results suggest that in the absence of glucocorticoids the hGR␤-hsp90 interaction may be weaker than the hGR␣-hsp90 interaction; consequently, more hGR␤ molecules are free of hsp90 and available to bind DNA. To test this hypothesis, receptor-hsp90 complexes were immunoprecipitated from heat-activated lysates in the absence of glucocorticoids using the anti-hsp90 antibody 3G3. The amount of recovered hGR␤ was approximately 40% less than the amount of recovered hGR␣ (Fig. 6E), consistent with hGR␤-hsp90 complexes being less stable than hGR␣-hsp90 complexes.
hGR␤ Heterodimerizes with hGR␣-Homodimerization of hGR␣ and the presence of ligand on each receptor monomer is required for glucocorticoid induction of gene expression (35,36). Therefore, an hGR␣-hGR␤ heterodimer, in which ligand is bound to one partner (hGR␣) but not the other (hGR␤), would be transcriptionally impaired. Co-immunoprecipitation experiments were performed with in vitro translated hGR␣ and hGR␤ to test whether the two proteins could physically associate with each other. Equivalent amounts of 35 S-labeled hGR␣ were incubated with lysates containing unlabeled hGR␤ or lysates containing no receptor (mock) in the presence of hormone (Fig. 7A, left panel, lanes 1 and 2). After cross-linking with DSP, proteins were immunoprecipitated with the hGR␤specific antibody BShGR (16). 35 S-Labeled hGR␣ was recovered by BShGR from the hGR␣/hGR␤ mixture (Fig. 7A, middle  panel, lane 1). Moreover, the 35 S-labeled hGR␣ monomers were shifted to a size consistent with 35 S-labeled hGR␣-hGR␤ heterodimers when the cross-linker was not broken (Fig. 7A, right panel, lane 1). No shift was observed in the small amount of 35 S-labeled hGR␣ recovered nonspecifically from the hGR␣/ mock mixture (Fig. 7A, compare middle and right panels, lane  2). To further investigate the ability of hGR␣ and hGR␤ to physically associate as a heterodimer, GST fusion proteins were constructed with the HBD of hGR␣ (amino acids 523-777) and the HBD of hGR␤ (amino acids 523-742). In addition to binding glucocorticoids, this region of hGR␣ is thought to contain a dimerization interface (37). Equivalent amounts of in vitro translated 35 S-labeled hGR␣ or 35 S-labeled hGR␤, treated with or without glucocorticoids, were incubated with immobilized GST, GST-hGR␣HBD, and GST-hGR␤HBD. Full-length were heat-activated in the presence of 200 nM DEX and incubated with a 10-fold molar excess of a radiolabeled MMTV promoter fragment (MMTV) or a radiolabeled DNA fragment that does not contain a GRE (pBR322). Receptor-DNA complexes were immunoprecipitated with antibody 57. DNA recovered from the immune complexes was then analyzed on a 4% denaturing urea-polyacrylamide gel. B, in vitro translated hGR␣ or hGR␤ were heat-activated in the presence of 200 nM DEX and incubated with a 10-fold molar excess of the radiolabeled MMTV promoter fragment in the absence (None) or presence of a 50-fold excess of unlabeled MMTV or pBR322 and processed as described above. C, in vitro translated hGR␣, hGR␤, or unprogrammed lysates (mock) were heat-activated in the absence or presence of 200 nM DEX, incubated with a 10-fold molar excess of the radiolabeled MMTV promoter fragment, and processed as described above. D, Western blot analysis with antibody 57 of in vitro translated hGR␣, hGR␤, or unprogrammed lysates (mock) utilized in the reactions above. E, association of in vitro translated hGR␣ and hGR␤ with hsp90. Immunoprecipitations were performed with the anti-hsp90 antibody 3G3 on heat-activated reticulocyte lysates expressing equivalent amounts of 35 S-labeled hGR␣ or 35 S-labeled hGR␤ in the absence of glucocorticoids. Left panel, autoradiograph of receptor input (10% of total); right panel, autoradiograph of co-immunoprecipitated receptor. hGR␣ interacted specifically with both the HBD of hGR␣ and the HBD of hGR␤ independent of ligand (Fig. 7B, left panel). Fulllength hGR␤ also interacted specifically with the both the HBD of hGR␣ and the HBD of hGR␤ independent of ligand (Fig. 7B,  right panel). Thus, results from both co-immunoprecipitation experiments and GST pull-down assays indicate that hGR␣ and hGR␤ can physically associate with each other as a heterodimer.
The Dominant Negative Activity of hGR␤ Resides within the Unique 15 Amino Acids at the Carboxyl Terminus of hGR␤-The hGR␣ and hGR␤ proteins are identical through amino acid 727 but then diverge, with hGR␣ having an additional 50 amino acids and hGR␤ having an additional, nonhomologous 15 amino acids. What role, if any, these unique 15 amino acids play in the dominant negative activity of hGR␤ has never been explored. We investigated this issue by truncating the receptor after amino acid 727 (hGR728T). The hGR728T protein, like hGR␤, did not bind DEX. 2 COS-1 cells were transfected with a fixed amount of the MMTV luciferase reporter pHHluc, a fixed amount of pCMVhGR␣, and increasing amounts of pCMVhGR728T corresponding to a 5-or 10-fold molar excess over the amount of transfected pCMVhGR␣. The cells were then treated with or without glucocorticoids for 18 h and analyzed for luciferase activity. In contrast to our findings for hGR␤ (see Fig. 1A), transfection of increasing amounts of pCMVhGR728T did not diminish the transcriptional activity of hGR␣.
Western blots were then performed to examine the relative levels of the hGR728T and hGR␣ proteins. hGR728T was expressed at approximately 70% the level of hGR␣ in COS-1 cells separately transfected with an equimolar amount of pCM-VhGR␣ or pCMVhGR728T (Fig. 8B, lanes 1 and 2). In COS-1 cells co-transfected with a fixed amount of pCMVhGR␣ and various amounts of pCMVhGR728T corresponding to a 1, 5, or 10-fold molar excess, hGR␣ levels did not change, but hGR728T levels increased the expected 5-and 10-fold (Fig. 8B, lanes 3-5). Together, these blots indicate that hGR728T was expressed at levels 3.5-and 7-fold greater than hGR␣ in the experiments presented in Fig. 8A. We showed earlier that hGR␤ functioned as a dominant negative inhibitor when expressed at levels 4and 8-fold greater than hGR␣ (see Fig. 1, A and B). Therefore, the inability of hGR728T to repress hGR␣-mediated activation of the MMTV promoter was not due to insufficient expression of the truncated receptor. DISCUSSION In the present study, we have characterized the dominant negative activity of hGR␤. We show that hGR␤ inhibits the transcriptional activity of hGR␣ in multiple cell types. This antagonism is selective for hGR␣, since other closely related steroid hormone receptors are only weakly inhibited by hGR␤. In addition, we show that the dominant negative activity of hGR␤ occurs on some, but not all, genes negatively regulated by glucocorticoids. We demonstrate that hGR␤ can associate with hsp90 and provide evidence that hGR␤-hsp90 complexes are less stable than hGR␣-hsp90 complexes in the absence of hormone. We also show that hGR␤ can bind GRE-containing DNA and heterodimerize with hGR␣, suggesting that competition for GRE binding and/or the formation of transcriptionally impaired hGR␣-hGR␤ heterodimers might be responsible for the dominant negative activity of hGR␤. Finally, we demonstrate that the dominant negative activity of hGR␤ resides within the unique 15 amino acids at the carboxyl terminus of hGR␤.
Within the steroid/thyroid/retinoic acid receptor superfamily, the GR, MR, PR, and AR comprise a subfamily based on sequence conservation (1). The DNA-binding domains of these receptors are highly conserved; consequently, each receptor recognizes the same HRE (32). Specificity is governed by ligand availability, cell-specific receptor expression, chromatin structure, affinity for the response element, and interactions (or lack thereof) with cofactors. Although hGR␤ can bind HREs, it only weakly represses the ability of PR and AR to activate the MMTV promoter. This suggests that the dominant negative activity of hGR␤ does not result simply from hGR␤ competing with ligand-bound hGR␣ for HRE binding. Otherwise, hGR␤ would be expected to compete with PR and AR for DNA binding and effectively inhibit their transcriptional activity, particularly since these receptors have a lower affinity for the HRE than does hGR␣ (38).
Important to the dominant negative activity of hGR␤ may be the formation of transcriptionally impaired hGR␣-hGR␤ heterodimers. We have shown using co-immunoprecipitation and GST pull-down techniques that hGR␤ can physically associate with hGR␣. Regions of hGR␣ involved in homodimerization have been mapped to a five-amino acid segment at the base of the second zinc finger known as the D-box and to a region in the HBD (39,40). The latter domain is located near the carboxyl terminus and consists of a heptad repeat of hydrophobic residues and additional hydrophobic residues located in intermediate positions (37,41). Interestingly, hGR␣ and hGR␤ diverge in the middle of this dimerization domain. The proximal two residues of the hydrophobic repeat (leucine 718 and methionine 725) and the hydrophobic amino acids between them are intact in hGR␤. Studies on the estrogen receptor have shown that this proximal portion of the dimerization domain is most important, since these residues form part of the dimer interface (37). The distal two residues of the hydrophobic repeat are leucine 732 and threonine 739. Within its unique carboxyl terminus, hGR␤ has a leucine at position 732. Moreover, an intervening hydrophobic amino acid (valine 729) is perfectly conserved in hGR␤. The presence in hGR␤ of an intact D-box, an intact proximal half of the carboxyl-terminal dimerization domain, and a partially conserved distal half of the carboxyl-terminal dimerization domain suggests, on sequence conservation alone, that hGR␤ can dimerize with hGR␣.
Further support that heterodimerization between hGR␣ and hGR␤ is an important component of hGR␤'s dominant negative activity comes from our studies with hGR728T. hGR728T is missing the unique 15 amino acids at the carboxyl terminus of hGR␤, and this truncated receptor does not repress the transcriptional activity of hGR␣. Removal of the hGR␤-specific amino acids deletes the distal half of the carboxyl-terminal dimerization domain, and it is this portion of the dimerization domain that is thought to play an important role in stabilizing the dimer (37). Therefore, the inability of hGR728T to function as a dominant negative inhibitor of hGR␣ may result from its inability to form stable heterodimers with hGR␣. Recent studies have demonstrated that hGR␣ can heterodimerize with MR and AR (42)(43)(44). The regions of hGR␣ involved in this interac-tion are thought to be the same ones that mediate hGR␣ homodimerization. Therefore, hGR␤ might also heterodimerize with these receptors. Consistent with this hypothesis, we observed a small reduction in the transcriptional activity of AR in the presence of hGR␤. In addition, hGR␤ has been shown to repress the transcriptional activity of MR in COS-7 cells (45).
Formation of hGR␣-hGR␤ heterodimers might reduce the transcriptional activity of hGR␣ in several ways. If the heterodimers do not bind GREs, then hGR␣ will be effectively sequestered by hGR␤ and denied access to target DNA. Other inhibitors of glucocorticoid action such as AP-1, NF-B, and calreticulin function in a similar manner by forming complexes with activated hGR␣ that prevent it from interacting with GREs (7, 46 -50). However, we have shown that hGR␤ can bind GRE-containing DNA. Therefore, hGR␣-hGR␤ heterodimers might bind GREs just as well as hGR␣ homodimers but communicate poorly with the general transcription machinery. Ligand is absolutely required for hGR␣ trans-activation (36), and a heterodimer in which ligand is bound to one partner (hGR␣) but not the other (hGR␤) may be impaired in this capacity. Finally, in cases where hGR␤ inhibits hGR␣-mediated repression of NF-B-responsive genes, formation of hGR␣-hGR␤ heterodimers may decrease the pool of hGR␣ molecules available to interact with and repress NF-B.
In contrast to our findings implicating heterodimerization as an important component of the dominant negative activity of hGR␤, Hecht et al. (51) have proposed that hGR␤ represses hGR␣-mediated activation of the MMTV promoter in COS-7 cells by nonspecific squelching of a general transcription factor. Inconsistent with this mechanism is the finding that overexpression of hGR␤ does not diminish the transcriptional activity of constitutively active glucocorticoid-independent promoters (13,14). In addition, "self-squelching" is not observed when hGR␣ is overexpressed (14), and we show in the current study that hGR␤ does not repress PR-mediated activation of the MMTV promoter. However, we cannot exclude the possibility that squelching is responsible for the small hGR␤-mediated reduction in the transcriptional activity of the AR. The findings by Hecht et al. (51) might be attributable to the amount of expressed hGR␣ and hGR␤ protein (which were not evaluated), use of a longer hormone incubation time (48 h), and/or utilization of a less sensitive and indirect alkaline phosphatase assay for measuring reporter activity.
The ability of hGR␤ to regulate hGR␣ activity in vivo will depend largely on two factors: its expression level relative to hGR␣ and the strength of its association with hsp90. Unfortunately, very little is still known about the relative levels of endogenous hGR␣ and hGR␤. Semiquantitative reverse transcriptase-polymerase chain reaction analysis of whole tissues indicates that the hGR␤ mRNA is less abundant than the hGR␣ mRNA (13,17). Similar results are observed on immunoblots of whole tissues (16). These results, however, may not reflect the situation in individual cells, particularly since hGR␤ appears to be expressed in a cell type-specific pattern. Immunohistochemistry has been performed on whole tissue sections prepared from lung, thymus, and liver with the hGR␤-specific antibody BShGR (16). Results from these studies indicate that hGR␤ is very abundant in certain epithelial cells. These include the epithelial cells lining the terminal bronchiole of the lung, forming the outer layer of Hassall's corpuscle in the thymus, and lining the bile duct in the liver.
Cells with elevated levels of hGR␤ (relative to hGR␣) should be less responsive to glucocorticoids. This in fact appears to be the case, since two recent studies have shown that hGR␤ is elevated in patients with glucocorticoid resistance. Leung et al. (19) compared the expression of hGR␣ and hGR␤ in peripheral blood cells isolated from normal subjects, glucocorticoid-sensitive asthmatics, and glucocorticoid-resistant asthmatics. Comparable levels of hGR␣ were expressed in each of the three groups, but the levels of hGR␤ were significantly higher in the glucocorticoid-resistant asthmatics. The expression of hGR␤ was found to be induced by cytokines, and the elevated levels of hGR␤ led to a reduction in the ability of hGR␣ to bind GREs. An increase in hGR␤ and reduction in hGR␣ has also been reported in a patient with systemic glucocorticoid resistance and chronic lymphocytic leukemia (20). Thus, an imbalance in hGR␣ and hGR␤ expression may underlie the pathogenesis of many clinical conditions associated with glucocorticoid resistance.
The strength with which hGR␤ associates with hsp90 will be another critical parameter influencing the ability of hGR␤ to function as a dominant negative inhibitor of hGR␣. We have shown that hGR␤ can associate with hsp90, and in vitro analysis of this interaction suggests that hGR␤-hsp90 complexes are less stable than hGR␣-hsp90 complexes in the absence of glucocorticoids. Consistent with this interpretation is our finding that hGR␤ binds DNA with a greater capacity than hGR␣ in the absence of glucocorticoids. Factors that regulate the stability of hGR␤-hsp90 complexes will, by altering the pool of hGR␤ molecules that are free from hsp90 and available to dimerize with hGR␣, regulate the dominant negative activity of hGR␤. Stress in the form of trauma, infection, or fever may provide the thermal stimulus in vivo needed for the selective dissociation of hGR␤-hsp90 complexes. In addition, an endogenous ligand may exist for hGR␤ that upon binding promotes the dissociation of hGR␤-hsp90 complexes. Our finding that hGR␤ can associate with hsp90 is somewhat surprising in view of this receptor isoform's predominant nuclear localization (13,16). In the case of hGR␣, sequences within the amino-terminal third, middle third, and carboxyl-terminal third of the HBD have each been shown to interact with hsp90 (52), and this interaction appears to mask or inactivate hGR␣'s nuclear localization signals, resulting in its cytoplasmic distribution in the absence of hormone (53). Only the first two-thirds of hGR␣'s HBD are intact in hGR␤. Therefore, hGR␤ may exhibit an altered interaction with hsp90 such that its nuclear localization signals are partially exposed or activated even when associated with hsp90. In addition, it has been reported that sequences at the carboxyl terminus of hGR␣ exert an inhibitory influence on the nuclear localization signals (54). Deletion and/or modification of this sequence, as occurs in hGR␤, might alleviate this repression and allow hGR␤ to localize in the nucleus independent of its association with hsp90. Finally, if the hGR␤-hsp90 interaction is unstable in cells, then the hGR␤ molecules residing in the nucleus may in fact be free of hsp90.
In summary, hGR␤ functions as an inhibitor of hGR␣, and this antagonism appears to be mediated primarily by the formation of hGR␣-hGR␤ heterodimers. As a modulator of hGR␣ activity, hGR␤ will play a pivotal role in the regulation of target cell responsiveness to glucocorticoids. Non-hormone-binding splice variants of the thyroid hormone receptor and estrogen receptor also function as dominant negative inhibitors of their wild-type counterparts, suggesting that alternative splicing is a common mechanism by which many steroid receptor genes generate isoforms with opposing biological activity (55,56).