Evidence That the β-Isoform of the Human Glucocorticoid Receptor Does Not Act as a Physiologically Significant Repressor*

Alternative splicing of the human glucocorticoid receptor (hGR) primary transcript generates two receptor isoforms, hGRα and hGRβ, with different carboxyl termini diverging at amino acid 727. By reverse transcriptase-polymerase chain reactions it was previously demonstrated that the hGRβ message had a widespread tissue distribution. To demonstrate the presence of hGRβ as protein we produced specific rabbit antisera to hGRβ, as well as a hGRβ-specific mouse monoclonal IgM antibody, by peptide immunizations. By SDS-polyacrylamide gel electrophoresis and Western immunoblotting we showed that hGRβ is endogenously expressed at the protein level in HeLa cells and human lymphatic leukemia cells. Using an antibody directed against an epitope shared by both isoforms we showed a relatively lower expression of the hGRβ form. We also showed that hGRβ bound to hsp90 by immunoprecipitation of in vitro translated hGRβ in reticulocyte lysate with hsp90-specific antibodies, a coprecipitation occurring also in the presence of dexamethasone. We could not demonstrate that hGRβ inhibited the effects of dexamethasone-activated hGRα on a glucocorticoid-responsive reporter gene. In conclusion, low hGRβ expression levels and hGRβ-hsp90 interaction maintained in the presence of ligand and lack of inhibition of hormone-activated hGRα effects challenge the concept of the hGRβ isoform as a proposed dominant negative inhibitor of hGRα activity.

Alternative splicing of the human glucocorticoid receptor (hGR) primary transcript generates two receptor isoforms, hGR␣ and hGR␤, with different carboxyl termini diverging at amino acid 727. By reverse transcriptase-polymerase chain reactions it was previously demonstrated that the hGR␤ message had a widespread tissue distribution. To demonstrate the presence of hGR␤ as protein we produced specific rabbit antisera to hGR␤, as well as a hGR␤-specific mouse monoclonal IgM antibody, by peptide immunizations. By SDS-polyacrylamide gel electrophoresis and Western immunoblotting we showed that hGR␤ is endogenously expressed at the protein level in HeLa cells and human lymphatic leukemia cells. Using an antibody directed against an epitope shared by both isoforms we showed a relatively lower expression of the hGR␤ form. We also showed that hGR␤ bound to hsp90 by immunoprecipitation of in vitro translated hGR␤ in reticulocyte lysate with hsp90-specific antibodies, a coprecipitation occurring also in the presence of dexamethasone. We could not demonstrate that hGR␤ inhibited the effects of dexamethasone-activated hGR␣ on a glucocorticoid-responsive reporter gene. In conclusion, low hGR␤ expression levels and hGR␤-hsp90 interaction maintained in the presence of ligand and lack of inhibition of hormone-activated hGR␣ effects challenge the concept of the hGR␤ isoform as a proposed dominant negative inhibitor of hGR␣ activity.
The cloning of the full-length human glucocorticoid receptor (hGR) 1 in 1985 predicted the existence of two receptor isoforms differing at their carboxyl termini (1). Characterization of the genomic structure of the hGR gene suggested two alternative exons for the COOH-terminal part of GR and that alternative splicing of exons 9␣ and 9␤ was the mechanism responsible for generating the two receptor isoforms (2). However, Oakley et al. (3) have recently shown that the hGR gene is processed as one single exon 9 containing both exon 9␣ and 9␤. Both hGR isoforms share the same amino acid sequence through amino acid 727 but diverge beyond this position with hGR␣ having an additional 50 amino acids and hGR␤ with an additional nonhomologous 15 amino acids.
Human GR in similarity to the other members of the steroid receptor superfamily consists of a poorly conserved aminoterminal region containing a major transactivation domain responsible for gene activation, a highly conserved cysteinerich central DNA-binding domain, and a relatively well conserved carboxyl-terminal domain important for hormone binding (4). In addition to steroid binding, the ligand-binding domain also has a number of other functions including dimerization, heat shock protein 90 (hsp90) binding and transactivation (5). Most previous studies have not distinguished between hGR␣ and hGR␤ and have mainly taken into consideration hGR␣ which is widely expressed and functions as a ligand-dependent transcription factor. Without making any distinction between hGR isoforms, earlier studies could be summarized to have shown that hGR resides both in the cytoplasm and in the cell nucleus and forms a heteroligomeric complex containing one steroid binding hGR and several nonsteroid binding components (6), including a dimer of hsp90 and monomers of hsp70 and hsp56. Hsp90 has been suggested to maintain hGR in a conformation that is unsuitable for DNA (GRE) binding but favorable for ligand binding. Once hormone binds to the receptor, a conformational change of the receptor results in the dissociation of hsp90 and some of the other associated proteins. Upon ligand binding GR is preferentially located in the nucleus, where it binds as a homodimer to GREs usually localized in the promoter regions of glucocorticoid-responsive genes (7). It has also been demonstrated that hGR, in addition to either enhancing or repressing transcription of a specific gene by binding to its promoter or to further upstream regions, may modulate gene expression by interacting with other transcription factors such as AP-1 (8) and NF-B (9).
Only a few recent studies have addressed the function of the ␤-isoform of hGR (3,10,11) and thus many aspects of the possible physiological role of hGR␤ remain to be further elucidated. Oakley et al. (3) have shown that hGR␤ does not bind dexamethasone or RU486 (3). hGR␤ has been assigned a role as a dominant negative inhibitor of hGR␣ activity (3,10). In the absence of hGR␣, however, hGR␤ seems to be transcriptionally inactive (3), although it can bind to a GRE as demonstrated in a gel mobility shift assay. The presence of hGR␣ and ␤ at the mRNA level has also been demonstrated (3,10) and a relatively lower amount of hGR␤ message (0.2-0.3% of total GR mRNA) has been suggested (3). A recent study by de Castro et al. (11) has implicated that hGR␤ binds hsp90, but has not demonstrated this in a direct fashion. In the same study this group also demonstrated hGR␤ protein in various tissues. However, the relative levels of hGR␤ as compared with hGR␣ remain to be unequivocally established. Reports of the intracellular localization of hGR␤ are somewhat conflicting and it also remains to be established under which conditions, normal and/or pathological, glucocorticoid-dependent repression or activation of gene transcription may be modulated by the interaction of the two hGR isoforms. The exact mechanism of such an effect also warrants further study.
To demonstrate the presence of hGR␤ as an expressed protein and to enable further studies of its function, we have performed peptide immunizations of rabbits and mice and produced specific rabbit antisera to hGR␣ and hGR␤ as well as an hGR␤-specific monoclonal IgM antibody. We confirm the expression of specific hGR␤ transcripts in several human tissues and extend these findings to the demonstration of the expression of the hGR␤ protein in two cell lines. We also show a faster relative mobility of hGR␤ in SDS-PAGE/Western blotting as compared with hGR␣. Furthermore, our data indicate a lower relative amount of hGR␤ protein expression as compared with hGR␣. Finally, we demonstrate that hGR␤ can bind directly to hsp90, that this binding is not affected by ligand in the same way as hGR␣-hsp90 binding and, based upon cotransfection studies in COS-7 cells, we challenge the current belief that hGR␤ is a general, negative modulator of hGR␣ function.

EXPERIMENTAL PROCEDURES
Screening of a Human cDNA Library by PCR-A cDNA library panel (CLONTECH) of 8 different human 5Ј-Stretch Plus TM cDNA libraries was screened by PCR for hGR␤ specific expression. The sense primer corresponded to the ␣ and ␤ common exon 8 hGR sequence bp 2219 -2238, i.e. 5Ј-AGCTAGGAAAAGCCATTGTC-3Ј. The specific hGR␤ primer corresponded to the exon 9␤ sequence bp 2317-2335 generating the antisense primer 5Ј-CTGGTTTTAACCACATAAC-3Ј. PCR reactions were performed in the presence of 3 M of the primer set, 3 mM MgCl 2 , 0.8 g of DNA, 12.5 mM nucleotide mixture, and 0.5 units of Taq polymerase (Promega). After a 2-min denaturing step at 95°C, 28 PCR cycles were run using an annealing temperature of 45°C for 1 min and an extension temperature of 72°C for 2 min. A PCR product of approximately 117 bp was generated in all tissues tested, i.e. heart, pancreas, lung, liver, brain, placenta, kidney, and skeletal muscle (data not shown).
Plasmid Construction and Plasmids Used for in Vitro Transcription/ Translation-Plasmid pGem7SP6hGR␣ (RiboProbe System, Promega) contained the full-length human GR ␣-isoform under control of the SP6 promoter (generously provided by Dr. Sam Okret, Karolinska Institute). For construction of plasmid pGEM7SP6hGR␤, a ClaI-XbaI fragment was amplified from pRShGR␤ (kindly provided by Dr. R. Evans, Salk Institute, La Jolla, CA) by PCR. The NH 2 -terminal primer, GR1, 5Ј-CCATCGATAAAATTCGAAGA-3Ј, contained an internal ClaI site corresponding to bp 1525 within the GR gene (1) and the COOHterminal antisense primer, GR8, 5Ј-GCTCTAGAGCTGGCCAGATAA-CACATACA-3Ј, corresponding to bp 2609 -2627 contained an added 5Ј XbaI site (1). The GR1-GR8 fragment was amplified by PCR using Vent-Polymerase (New England Biolabs) and the following program: 25 cycles of 45 s at 94°C, 60 s at 44°C, 90 s at 72°C, and a final elongation step of 10 min at 72°C. The amplified segment was subcloned as a ClaI-XbaI fragment into pBluescript (Stratagene), resulting in plasmid p8-1 and subsequently sequenced. pGem7SP6hGR␤ was finally constructed by replacing the COOH-terminal ClaI-XbaI fragment in pGem7SP6hGR␣ (RiboProbe System, Promega) with fragment p8-1. GR␤ was produced under the control of the SP6 promoter.
Plasmids Used in Transfection Experiments-GR␣ was expressed from pMT-GR␣ (12) and GR␤ from pRShGR␤ (13). Control plasmid pRShGR␤⌬SalI was constructed by cutting pRShGR␤ with SalI, at a unique restriction site within the 1 sequence in the GR-coding region. The protruding ends were blunt ended by mung bean nuclease treatment and the plasmid was religated. This generates a 4-bp deletion (confirmed by sequencing) and a change in reading frame after R 131. The truncated protein produced from this construct corresponded to the NH 2 terminus of the hGR and was 168 amino acids long, with the last 37 amino acids differing from the wild type protein. This protein did not activate the GRE-dependent reporter gene used in our transfection experiments. The reporter plasmid pSALP contained the gene for secreted placental alkaline phosphatase under the control of the GR inducible MMTV promoter. Plasmid pAP contained the same reporter gene under control of a non-inducible promoter. Both plasmids were kindly provided by KaroBio, Huddinge, Sweden. Plasmids used in transfection experiments were prepared from Escherichia coli XL-1 Blue on Sepharose columns (Qiagen, KEBO, Sweden).
Growth and Transfection of COS-7 Cells-Monolayer cultures of COS-7 cells were grown in Dulbecco's modified Eagle's medium containing 4.5 mg/ml glucose; Life Technologies, Inc.) supplemented with heatinactivated 10% fetal bovine serum, 2 mM glutamine, penicillin (5 units/ml), and streptomycin (50 g/ml). Transient transfection was performed using DOTAP as a transfection agent according to the manufacturer's recommendation (Boehringer Mannheim). Cells were grown in 35-mm dishes. 1 g of reporter plasmid (pSALP) was cotransfected with various amounts of expression plasmid pMT-GR␣, pRShGR␤, and pRShGR␤⌬SalI, respectively, for 6 h. Cells were induced by adding fresh medium containing dexamethasone (Sigma), 24 h after transfection. 48 h after dexamethasone induction, alkaline phosphatase activity was assayed in cellular supernatants. Concentrations of plasmids and dexamethasone are as indicated in the figures.
Determination of Alkaline Phosphatase Activity-Transient transfections were assayed for alkaline phosphatase activity. For this purpose, 1 ml of medium of induced cells was cleared by centrifugation, the supernatant was transferred to a fresh tube and incubated at 65°C for 30 min and thereafter recentrifuged. 250 l of the supernatant was added to 80 l of alkaline phosphatase assay solution (5 l/ml Sigma 104 phosphatase substrate in 1.12 M NaCl, 2 mM MgCl 2 , 0.8 M Tris-HCl, pH 8.8) in microtiter plates and incubated at room temperature. Enzymatic activity was monitored by following the rate of change of absorption of p-nitrophenyl phosphate at 405 nm.
Immunization-Peptides were selected based on the predicted antigenicity or based on unique sequences of the two hGR receptor isoforms. Peptide E17P corresponds to amino acids 510 -526 in both hGR␣ and hGR␤. Peptide K15K corresponds to the 15 COOH-terminal amino acids of hGR␣ and N15I to the 15 COOH-terminal amino acids of hGR␤. The peptides were ordered from Neosystem Laboratoire (Strasbourg, France). Peptides were conjugated to keyhole limpet hemocyanin by the one-step glutaraldehyde method. Rabbits were immunized with 50 g of peptide-keyhole limpet hemocyanin initially in Freund's complete adjuvant and thereafter in incomplete adjuvant or phosphate-buffered saline. Immunization was repeated monthly and after several boosters, the animals were bled and the antisera tested in enzyme-linked immunosorbent assay against the relevant peptide. Positive sera were further tested in enzyme-linked immunosorbent assay against hGR purified by immunoaffinity chromatography (cf. below) from HeLa cells or a baculovirus-expressed hGR (kindly provided from KaroBio, Huddinge, Sweden). Monoclonal antibodies were prepared according to standard procedures.
Immunoaffinity Purification of hGR-A monoclonal antibody previously generated against the rat GR with a known cross-reactivity to hGR (mAb5, also called 293 (14 -16)) was coupled to cyanogen bromideactivated Sepharose 4B (Pharmacia, Uppsala, Sweden) according to a procedure described by the manufacturer. Cytosol from approximately 5 ml of HeLa cell pellets, stored frozen at Ϫ70°C, was prepared by homogenization in low salt buffer as described previously (16). The cytosol was slowly passed over the column, washed with 10 column volumes of EPG (1 mM EDTA, 20 mM sodium phosphate, pH 7.0, 10% (w/v) glycerol, 10 mM dithiothreitol) containing 50 mM NaCl and further with 10 column volumes of EPG with 1 M NaCl. Human GR was then eluted with 0.1 M sodium citrate buffer, pH 3.0. A similar procedure was used for hGR purification from plasmapheresis-enriched lymphocytes from patients with chronic lymphatic leukemia (a kind gift from Dr. Adam Smolovitz, Dept. of Hematology, Karolinska Hospital, Stockholm, Sweden) or from Sf1 cells expressing hGR obtained from KaroBio (Huddinge, Sweden).
In Vitro Transcription/Translation-The radiolabeled glucocorticoid receptors GR␣ and GR␤, respectively, were generated by in vitro transcription/translation of plasmids phGR␣ and pGEM7SP6hGR␤, in the presence of [ 35 S]methionine (Amersham) in rabbit reticulocyte lysate using a coupled in vitro transcription/translation kit (Promega) according to the manufacturer's recommendation.
Immunoprecipitation-Immunoprecipitation of [ 35 S]methionine-labeled glucocorticoid receptor-bound to hsp90 by monoclonal IgM antibodies against hsp90 (antibody 3G3, purchased from Affinity Bioreagents) and subsequent SDS-polyacrylamide electrophoresis was carried out as described previously (17), with an additional step to block nonspecific binding using 3% fat-free milk. The SDS gels were dried under vacuum and subjected to autoradiography. The same protocol was used with polyclonal anti-hGR antibodies, with the modification that samples were blotted to nitrocellulose before autoradiography.
Western Immunoblotting-SDS-PAGE and Western blotting to nitrocellulose of immunoaffinity-purified hGR were performed according to standard procedures. To allow discrimination between hGR␣ and hGR␤, strips were cut from single lanes of purified hGR subjected to SDS-PAGE and Western blotting, marked to ensure proper alignment, and probed with peptide-specific antibody and the relevant secondary antisera coupled to alkaline phosphatase (DAKO-Patts, Denmark). Apart from primary antibodies described in this paper we also used previously derived antibodies to rat GR, mAb5, and mAb7 (14).

RESULTS
We confirmed that hGR␤ is expressed at the mRNA level in human tissues by PCR screening of a commercial human cDNA library panel using an exon 9␤-specific primer together with a common internal GR primer corresponding to an exon 8-specific sequence. With all tissues investigated a PCR product corresponding to the expected size, 117 bp, was generated indicating the widespread presence of hGR␤-mRNA transcripts (data not shown), similar to what has been demonstrated previously (3,10). Primer specificity was tested by incubating the ␤-primers with plasmids containing the gene for hGR␣ and hGR␤, respectively. No PCR product was obtained with the combination of ␤-primers/␣-plasmid (data not shown).
To study the expression of the hGR␤ protein we produced and tested a number of anti-hGR␤ antibodies. Upon extensive immunization the three selected peptides, K15K, N15I, and E17P, corresponding to hGR␣, hGR␤, and a peptide common to both isoforms, elicited immune responses in rabbits. We also produced a monoclonal antibody to hGR␤. To ensure that the antisera were specific for each isoform we tested for crossreactivity. Proteins corresponding to each of the isoforms, hGR␣ and hGR␤, were specifically expressed in reticulocyte lysate. Fig. 1, shows that antibodies raised against peptides only recognized the cognate receptor isoform, i.e. antibodies raised against the hGR␣ specific peptide K15K precipitated hGR␣ but not hGR␤ and antibodies raised against N15I precipitated hGR␤ but not hGR␣.
The expression of hGR␤ was studied in HeLa and CLL cells. Cytosol from HeLa cells grown in culture or from lymphocytes obtained from plasmapheresis of patients with chronic lymphatic leukemia (CLL) was enriched for hGR using a monoclonal antibody directed against the NH 2 -terminal part of rat GR, mAb5, that cross-reacts with hGR and presumably should recognize both GR isoforms equally. Fig. 2 shows strips cut from a nitrocellulose filter after SDS-PAGE and Western immunoblotting. Strips from one large lane of cytosol were cut into several narrower strips and probed with antibodies and the relevant secondary antisera coupled to alkaline phosphatase. In Fig. 2A, hGR-enriched HeLa cell cytosol, when probed with a hGR␤ specific antibody raised against peptide N15I, was shown to contain a major immunoreactive band of slightly lower relative molecular weight than that of the predominant band, which was seen when the blot was probed with the antiserum raised against peptide E17P, which was common to both hGR isoforms. A similar result for CLL cells is shown in Fig. 2B. These results clearly showed that hGR␤ was expressed at the protein level in two different cell types of human origin. The expression levels of hGR␤, however, seemed to be significantly lower as compared with hGR␣ expression levels indicated by the fact that the intensity of the ␤-band was much lower when an antibody recognizing an epitope situated approximately 200 amino acids more NH 2 terminally in the ligand-binding domain and common for both receptor isoforms was used. This was also suggested by experiments where mAb7, recognizing an epitope within the NH 2 -terminal domain of hGR, separate from the mAb5 epitope (15) at increasing concentrations and recognized an additional band of lower molecular weight, corresponding in size to the hGR␤ isoform, as tested on CLL cytosol enriched for both GR isoforms by mAb5 immunoaffinity chromatography as described above (Fig. 3). Thus, the results in Figs. 2 and 3, although not directly quantitative, indicate a lower relative amount of hGR␤ protein in both HeLa and CLL cells.
To test whether hGR␤ is present as a heterocomplex with hsp90 and, if so, whether hsp90 could be released by addition of dexamethasone, in vitro translation of hGR␣ or hGR␤ in reticulocyte lysate in either the absence or presence of 100 M dexamethasone was carried out. Aliquots of the lysate were divided and immunoprecipitated using either Sepharose-coupled monoclonal hsp90 antibodies (cf. below) or, as a control, unspecific IgM antibodies (Fig. 4, lanes 2, 5, 9, and 12). As indicated in Fig. 4, 35 S-labeled protein was detected in lysate

FIG. 2. Expression of hGR␤ protein in HeLa and CLL cells.
Cytosols derived from two cell types of human origin, HeLa cells and CLL cells, were enriched for hGR by immunoaffinity chromatography using anti-ratGR antibody mAb5 directed against epitopes within the NH 2 -terminal domain of rGR and previously shown to be cross-reactive with the hGR NH 2 terminus (15). containing phGR␣ (lanes 6 and 13) and pGem7SP6hGR␤ (lanes  3 and 10), respectively. hGR␤, which is 35 amino acids shorter than hGR␣, runs slightly ahead, demonstrating the difference in size between the two hGR isoforms (85 and 81 kDa). Furthermore, both hGR isoforms were detected in immunoprecipitates using hsp90 antibodies, indicating that hGR␣ (Fig. 4, lane  4) as well as hGR␤ (Fig. 4, lane 1) is bound to hsp90. To test whether hsp90 was released in the presence of hormone, hGR␣ or hGR␤ were synthesized in reticulocyte lysate in the presence of 100 M dexamethasone and immunoprecipitated with hsp90 antibodies in the presence of dexamethasone. As shown in Fig.  4 (lane 11), hGR␣ was not immunoprecipitated with hsp90 antibodies in the presence of dexamethasone. In contrast, hGR␤ (Fig. 4, lane 8) was immunoprecipitated with hsp90 antibodies in the presence of dexamethasone. hGR␤ coprecipitated with hsp90 to the same extent in both the absence (Fig. 4,  lane 1) and presence (Fig. 4, lane 8) of dexamethasone.
To test whether hGR␤ had any effect on the hormone-induced hGR␣-mediated stimulation of gene expression, COS-7 cells were transfected with pSALP as a reporter gene and constant amounts of hGR␣ or hGR␤ plasmid. In hGR␣ transfected cells, dexamethasone induced alkaline phosphatase activity in a dose-dependent fashion, whereas no induction was observed in hGR␤ transfected cells (Fig. 5). In a second set of experiments, cells containing pSALP were transfected with increasing amounts of hGR␤ and a constant amount of hGR␣. As indicated in Fig. 6, panel A, alkaline phosphatase activity decreased accordingly. However, we obtained the same effect when cells were transfected with a a constant amount of a GR independent reporter gene (pAP) and increasing amounts of hGR␤ (Fig. 6, panel B), indicating that the hGR␤ effect on the GR-inducible pSALP activity might be due to unspecific squelching and not to a specific inhibitory effect of hGR␤. A caveat in this set of experiments is that the total plasmid concentration was changing. Therefore to keep the total concentration of transfected plasmid DNA constant, cells were transfected with increasing amounts of hGR␤ in the presence of hGR␤⌬. As shown in Fig. 7 (panel B), the problem with unspecific squelching in the GR independent reporter gene system was eliminated. However, we did not obtain a significant inhibitory effect of hGR␤ on dexamethasone-induced hGR␣-mediated stimulation of the pSALP reporter gene activity (Fig. 7,  panel A). DISCUSSION Based on characterization of multiple receptor cDNA clones and receptor protein analysis by immunoblotting, where only hGR␣ was demonstrated, it was initially concluded that the predominant physiological form of hGR is hGR␣ (1). Results have recently been published demonstrating expression of hGR␤ transcripts in a variety of human tissues and a potential role for hGR␤ as a dominant negative inhibitor of hGR␣ activity (3,10). In contrast to the well characterized hGR␣ isoform, very little is known about the hGR␤ splice variant. In this report, we examined the expression of the hGR␤ transcript and protein, association of hGR␤ with hsp90 and physiological function of hGR␤. By PCR we confirmed that hGR␣ and hGR␤ mRNA transcripts were co-expressed in several human tissues.
We have previously produced anti-rat GR antibodies that  1-3 and hGR␣ in lanes 4 -6) were immunoprecipitated with either anti-hsp90 monoclonal antibody 3G3 (lanes 1 and 4) or with an irrelevant monoclonal IgM TEPC 183 antibody (lanes 2 and 5). The precipitates were separated by SDS-PAGE and coimmunoprecipitated hGR␣ and -␤, respectively, were visualized by autoradiography (cf. "Experimental Procedures"). Lanes 3 and 6 corresponds to the total input amount of labeled hGR␤ or hGR␣ lysate, subjected to immunoprecipitation. Expression in the presence of 100 M dexamethasone of hGR␤ (lanes 8 -10) and hGR␣ (lanes [11][12][13], in vitro was also performed. Lanes 10 and 13 represent the total input amount of labeled hGR␤ and hGR␣ lysate. Aliquots of the lysates in the presence of 100 M dexamethasone were immunoprecipitated with either anti-hsp90 monoclonal antibody 3G3 (lanes 8 and 11) or with an irrelevant monoclonal IgM TEPC 183 antibody (lanes 9 and 12). Lane 7 contained a nonradioactive molecular mass standard. The arrow indicates the position of hGR␣ at 85 kDa. cross-react with hGR and recognize epitopes in the aminoterminal domain, thus recognizing both the hGR␣ and hGR␤ (14). In this report, we have produced isoform-specific anti-hGR␣ and hGR␤ polyclonal antibodies in rabbits, which are noncross-reactive in immunoprecipitation experiments and which specifically recognize the hGR␣ and hGR␤ proteins in Western blotting. In this study we also produced a monoclonal antibody against hGR␤, raised against the 15 unique COOHterminal amino acids of hGR␤ here called peptide N15I. This antibody also recognizes a specific immunoreactive band in CLL cell cytosol (Fig. 2B) as well as in HeLa cell cytosol (not shown), demonstrating that hGR␤ was expressed at the protein level in human cells. As compared with hGR␣, hGR␤ expression levels seemed to be significantly lower, as indicated by results using two different antibodies recognizing epitopes common for both isoforms, one in the NH 2 -terminal part of hGR and the other in the ligand-binding domain, approximately 200 amino acids NH 2 terminally of the diverging point of hGR␣ and -␤. A low hGR␤ expression is in better agreement with a recent report by Oakley et al. (3), where hGR␤ mRNA levels are estimated to be only 0.2-0.3% of total mRNA. These data and ours indicate that hGR␤ may not necessarily be of significant importance under normal physiological conditions. However, a recent study by de Castro et al. (11) suggests a high level of hGR␤ protein expression, in most cases exceeding hGR␣ expression, based on quantitation in Western blotting experiments, using peptides coupled to albumin to create standard curves. It is unclear to what extent coupling efficiency for the different peptides is controlled in this experiment and whether a quantitative comparison between the two isoforms in this fashion really is valid. Furthermore, we noted that the antisera raised by de Castro et al. (11) did not differentiate between the sizes of the two hGR isoforms. We believe that the use of an antibody that recognizes a common epitope in the two hGR isoforms, as described in this paper, is better suited for quantitative comparison of the isoforms as no difference in affinity to this epitope between isoforms is to be expected in Western blotting. This enables a direct comparison of relative levels of the proteins. Whether there exists a varying expression of the different hGR isoforms in normal tissues as well as in pathological tissues remains to be further studied.
In addition to steroid binding, the ligand-binding domain also harbors other functions including dimerization, hsp90 binding, and transactivation (6). It is well established that the hGR␣ receptor isoform translocates from the cytoplasm to the nucleus in a hormone-dependent manner and that, in the absence of hormone, the association of hsp90 with hGR␣ appears to inactivate the nuclear localization signal (18,19). The hormone-dependent dissociation of hsp90 from hGR␣ is probably important in the nuclear translocation of hGR␣. In this report we showed that hGR␤ is also associated with hsp90, but in contrast to hGR␣ still maintains the hsp90 association in the presence of ligand. Receptor derivatives of rat GR terminating at amino acids 766 (hGR 748) and 671 (hGR 653) were found to coprecipitate together with hsp90, whereas further truncation at the COOH-terminal end interfered with this interaction (20). Thus the site of hsp90 interaction appears to lie within a common region of the two hGR isoforms located at least 75 residues N-terminal of the diverging point of hGR␣ and hGR␤. Relating to the involvement of hsp90 in determining the intracellular localization of GR, a recent report by Oakley et al. (3) demonstrates that hGR␤ resides primarily in the nucleus of transfected cells independent of hormone treatment. However, this is in contrast to another study showing essentially the same distribution pattern for hGR␣ as for hGR␤ (11), i.e. the intracellular localization of hGR␤ and its relation to hsp90 interaction call for further studies. In our experimental system, and in contrast to previous reports (3, 10), we were not able to demonstrate that hGR␤ inhibits the effect of hormone-activated hGR␣ on a glucocorticoid-responsive reporter gene in COS-7 cells. In cotransfection experiments, using the reporter gene pSALP containing an  MMTV promoter positively regulated by glucocorticoids and hGR␣, alkaline phosphatase activity was clearly induced by dexamethasone in a dose-dependent manner, whereas no induction was obtained in cells transfected with pSALP and hGR␤. When COS-7 cells were transfected with a constant amount of hGR␣-plasmid and increasing concentrations of hGR␤-plasmid, hGR␣-mediated activation of the MMTV promoter was inhibited. However, the increasing expression of hGR␤ inhibited a glucocorticoid-independent constitutive reporter gene to the same extent, indicating that this effect was due to nonspecific squelching. When hGR␣ and hGR␤ were expressed in the same cell and transfected DNA was kept constant by adding the truncated pRSV-GR␤⌬SalI plasmid, we did not obtain a significant hGR␤ inhibition of glucocorticoidinduced hGR␣-mediated activation of the MMTV promoter. Thus, we conclude that the suggested hGR␤-mediated repression of hGR␣ is not a universal phenomen and also that the interaction between hGR␣ and hGR␤ may be more complex than previously suggested and warrants further studies. In the case of repression occurring in systems other than ours, it also remains to be determined to what extent hGR␣/hGR␤-heterodimers or hGR␤/hGR␤-homodimers participate in the occupation of GRE sequences. In addition, interactions with other steroid hormone receptors and other proteins and transcription factors, such as AP-1 (8) and NFB (9), may further contribute to the complexity of hGR␣ and hGR␤ regulation of gene expression.
Studies of progesterone receptor isoforms in different animal models have identified variations in the levels of progesterone receptor-A and -B as a consequence of endocrine manipulations as well as during development (21). Despite the issues raised above regarding hGR␣-hGR␤ interaction, the possible resulting effects on specific gene expression and the conflicting data with regard to the absolute and relative levels of hGR␣ and hGR␤ expressed in various human tissues, it may still be possible that during specific circumstances an altered ratio of these GR receptor isoforms may result in an alteration of hormonal responses.
In conclusion, there are a number of important issues yet to be addressed with regard to the physiological significance of hGR␤ as a modulatory receptor isoform. Studies by Oakley et al. (3) and de Castro et al. (11) demonstrate that the hGR␤ protein indeed is expressed in several tissues at the mRNA level. Based on experiments using antibodies detecting both the hGR␣ and hGR␤ isoform, however, GR␣ was suggested as the major form expressed in our system. We also found that hGR␤ was associated with hsp90, and our study indicated that ligand does not result in a significant release of hGR␤ from hsp90. These results and the fact that hGR␤ did not have a dominant negative action in a glucocorticoid-driven reporter gene system, but rather a nonspecific squelching effect, warrant further studies of the role of GR isoforms in human, and indicate that hGR␤, under normal physiological conditions probably does not have a significant function at observed expression levels.