The Human Glucocorticoid Receptor Isoform

Alternative splicing of the human glucocorticoid receptor (hGR) primary transcript produces two receptor isoforms, hGRα and hGRβ, which differ at their carboxyl termini. The hGRα isoform conveys endocrine information to target tissues by altering patterns of gene expression in a hormone-dependent fashion. In contrast to hGRα, very little is known about the hGRβ splice variant. Using hGRα- and hGRβ-specific riboprobes on human multiple tissue Northern blots, we show that the hGRβ message has a widespread tissue distribution. We also prove by reverse transcriptase-polymerase chain reaction that the alternative splicing event underlying the formation of the hGRβ message occurs in these tissues. Because the hGRβ protein differs from hGRα at the extreme COOH terminus, we investigated several of the biochemical properties of hGRβ expressed in transfected cells. hGRβ does not bind the glucocorticoid agonist dexamethasone nor the glucocorticoid antagonist RU38486 in vivo. Moreover, in contrast to hGRα, hGRβ is located primarily in the nucleus of transfected cells independent of hormone administration. Finally, in the absence of hGRα, hGRβ is transcriptionally inactive on a glucocorticoid-responsive enhancer. However, when both isoforms are expressed in the same cell, hGRβ inhibits the hormone-induced, hGRα-mediated stimulation of gene expression. Thus, hGRβ potentially functions as a dominant negative inhibitor of hGRα activity.

Alternative splicing of the human glucocorticoid receptor (hGR) primary transcript produces two receptor isoforms, hGR␣ and hGR␤, which differ at their carboxyl termini. The hGR␣ isoform conveys endocrine information to target tissues by altering patterns of gene expression in a hormone-dependent fashion. In contrast to hGR␣, very little is known about the hGR␤ splice variant. Using hGR␣and hGR␤-specific riboprobes on human multiple tissue Northern blots, we show that the hGR␤ message has a widespread tissue distribution. We also prove by reverse transcriptase-polymerase chain reaction that the alternative splicing event underlying the formation of the hGR␤ message occurs in these tissues. Because the hGR␤ protein differs from hGR␣ at the extreme COOH terminus, we investigated several of the biochemical properties of hGR␤ expressed in transfected cells. hGR␤ does not bind the glucocorticoid agonist dexamethasone nor the glucocorticoid antagonist RU38486 in vivo. Moreover, in contrast to hGR␣, hGR␤ is located primarily in the nucleus of transfected cells independent of hormone administration. Finally, in the absence of hGR␣, hGR␤ is transcriptionally inactive on a glucocorticoid-responsive enhancer. However, when both isoforms are expressed in the same cell, hGR␤ inhibits the hormone-induced, hGR␣-mediated stimulation of gene expression. Thus, hGR␤ potentially functions as a dominant negative inhibitor of hGR␣ activity.
Two human glucocorticoid receptor (hGR) 1 cDNA clones, termed hGR␣ and hGR␤, were isolated in 1985 that predicted the existence of two receptor isoforms differing at their carboxyl termini (1). Amino acid sequence analysis revealed that the hGR␣ and hGR␤ isoforms were identical through amino acid 727 but diverged beyond this position with hGR␣ having an additional 50 amino acids and hGR␤ an additional, nonhomologous 15 amino acids. Exons 1-8 of the hGR gene contain the 5Ј noncoding and coding sequences common to the hGR␣ and hGR␤ cDNAs, and exons 9␣ and 9␤ contain the coding and 3Ј noncoding sequences specific to the hGR␣ and hGR␤ cDNAs (2). Because the hGR␣and hGR␤-specific sequences are located on the same gene, alternative splicing of exons 9␣ and 9␤ was speculated to be the mechanism responsible for generating the two receptor isoforms. However, initial Western blot analysis detected only the larger 94-kDa hGR␣ isoform, and only hGR␣ appeared to bind hormone and induce expression of a glucocorticoid-responsive reporter plasmid in a hormone-dependent manner (1,3). Because of its predominant expression, ligand binding properties, and transcriptional activity, hGR␣ became the primary focus of subsequent research. As a result, its expression, biochemical properties, and physiological function have been well characterized.
hGR␣ is expressed in most human tissues and cell lines and belongs to the superfamily of steroid/thyroid/retinoic acid receptor proteins that function as ligand-dependent transcription factors (for reviews see Refs. 4 -6). Members of this family are organized into structurally and functionally defined domains. Specifically, hGR␣ is comprised of a unique amino-terminal variable region that includes a transactivation domain that is important for regulation of gene expression. hGR␣ also contains a central DNA-binding domain crucial for specific interaction of the receptor with DNA sequences containing glucocorticoid receptor responsive elements (GRE). The carboxyl terminus of the hGR␣ protein contains the hormone-binding domain as well as sequences important for interaction with heat shock protein 90 (hsp90) (7), nuclear translocation (8), receptor dimerization (9), and transactivation (10).
In the absence of hormone, hGR␣ resides predominantly in the cytoplasm of cells, where it exists as a large multiprotein complex (for reviews see Refs. 11 and 12). This complex appears to consist of the receptor polypeptide, two molecules of hsp90, and several additional proteins. The association of hsp90 with the receptor is believed to maintain the receptor in a high affinity hormone binding state and sequester the receptor in the cytoplasm by inactivating the nuclear localization signals (NLS). Once hormone binds the receptor, a conformational change ensues resulting in the dissociation of hsp90 and the other associated proteins. In its new conformation hGR␣ translocates into the nucleus, where it binds as a homodimer to GREs that are usually found in the promoter regions of steroidresponsive genes. The receptor then communicates with the basal transcription machinery to either enhance or repress transcription of the linked gene. hGR␣ can also modulate gene expression by physically interacting with other nuclear proteins such as AP-1 (13)(14)(15) and NF-B (16).
In contrast to hGR␣, very little is known about the physiological significance of hGR␤. We demonstrate here that a mRNA transcript consistent in size with the hGR␤ cDNA is expressed in various human adult and fetal tissues and in * 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 U.S.C. Section 1734 solely to indicate this fact. several transformed human cell lines. We also confirm that the alternative splicing event underlying the formation of the hGR␤ message occurs in these tissues. In addition, we show that the unique COOH-terminal end of hGR␤ influences several key biochemical properties of this isoform that distinguishes it from hGR␣. We demonstrate that hGR␤ does not bind glucocorticoids or antiglucocorticoids in vivo, resides in the nucleus independent of hormone administration, and in the absence of hGR␣ is transcriptionally inactive on a glucocorticoid-responsive enhancer. It was recently reported that transfected hGR␤ inhibits transfected hGR␣-mediated induction of the mouse mammary tumor virus (MMTV) promoter (17). We extend these findings by demonstrating that hGR␤ represses the activity of endogenous hGR␣ and that this hGR␤-mediated repression is a general phenomenon of glucocorticoid-responsive promoters. Thus, the physiological significance of hGR␤ may reside in its ability to antagonize the function of hGR␣. Recombinant Plasmids-The hGR␣ expression vector pCYGR (by Y. Itoh-Lindstrom and J. A. Cidlowski) served as the vector for the expression plasmids pCMVhGR␣ and pCMVhGR␤ utilized in this study. pCYGR was constructed by isolating the 3.0-kb KpnI-XhoI cDNA fragment of pRShGR␣ (3). This fragment (which includes the entire hGR␣ coding region and the first 384 bp of the hGR␣ 3ЈUTR) was then cloned downstream of the human cytomegalovirus major intermediate early gene promoter region in the plasmid pCMV5 (18). The hGR␣ expression vector pCMVhGR␣ was constructed by isolating the 3.3-kb ClaI-BamHI cDNA fragment from the hGR␣ clone OB7 (1). This fragment (which contains the distal 940 bp of the hGR␣ coding region and the 2322 bp hGR␣ 3ЈUTR) was then cloned into the ClaI-BamHI sites in pCYGR. The hGR␤ expression vector pCMVhGR␤ was constructed by isolating the 2.3-kb ClaI-BamHI cDNA fragment from the hGR␤ clone OB10 (1). This fragment (which contains the hGR␤-specific coding sequences as well as the 1430-bp hGR␤ 3ЈUTR) was then cloned into the ClaI-BamHI sites in pCYGR. Plasmids pGMCS (19), pGRE2CAT (20), and pBLCAT2 (21) have been previously described.

Materials-Dexamethasone
Cell Culture and Transfections-HeLa S 3 and COS-1 cells were grown as described previously (22,23). CV-1 cells were grown in Dulbecco's minimum essential medium supplemented with 2 mM glutamine and 10% (v/v) heat-inactivated fetal calf serum. CEM-C7 cells were grown in suspension in RPMI 1640 medium supplemented with 2 mM glutamine and 10% (v/v) heat-inactivated fetal calf serum. All cultures were maintained in a 5% CO 2 humidified atmosphere at 37°C and were passaged every 3-4 days. HeLa S 3 and COS-1 cells were transfected essentially as described previously (22). Briefly, 4 h before transfection, medium was replaced with fresh Dulbecco's minimum essential medium containing 3% serum. Plasmid DNA was prepared as a calcium phosphate precipitate and incubated with cells for 5 h followed by a 30-s shock with 15% glycerol. Cells were then refed supplemented medium.
RT-PCR-Human liver and lung total RNA were kindly provided by Dr. Darryl Zeldin (NIEHS). Human heart, brain, and skeletal muscle total RNA was purchased from ClonTech. Total RNA from HeLa S 3 and CEM-C7 cells was isolated using TRIzol Reagent (Life Technologies, Inc.) according to the manufacturer's instructions. cDNA was prepared in a buffer containing 10 mM Tris-HCl, pH 8.3, 50 mM KCl, 5 mM MgCl 2 , 1 mM each dGTP, dATP, dTTP, and dCTP, 20 units RNase inhibitor, 50 units murine leukemia virus reverse transcriptase, 2.5 M random hexamers, 1.0 g of sample RNA, in a final volume of 20 l. This mixture was incubated for 10 min at RT, 15 min at 42°C, and 5 min at 99°C and then used for amplification of specific cDNAs by PCR. The buffer for PCR contained 10 mM Tris-HCl, pH 8.3, 50 mM KCl, 2 mM MgCl 2 , 2.5 units AmpliTaq DNA polymerase, 0.2 M upstream and downstream primers, in a final volume of 100 l. After an initial incubation for 1 min at 95°C, samples were subjected to 45 cycles of 30 s at 95°C, 30 s at 62°C, and 30 s at 72°C. This was followed by a final extension step at 72°C for 10 min. The primers used for the amplification of the 7.0-kb hGR message were as follows: 5Ј-GCATTCATACAG-GCAGCGAT-3Ј (upstream) and 5Ј-CCACGTATCCTAAAAGGGCAC-3Ј (downstream) corresponding to nucleotides 4221-4240 and 2503-2523 of the hGR␣ and hGR␤ cDNAs, respectively (1). The primers used for the amplification of the hGR␣ message were as follows: 5Ј-CCTAAG-GACGGTCTGAAGAGC-3Ј (upstream) and 5Ј-GCCAAGTCTTGGC-CCTCTAT-3Ј (downstream), corresponding to nucleotides 2158 -2178 and 2616 -2635 of the hGR␣ cDNA (1). The primers used for the amplification of the hGR␤ message were as follows: 5Ј-CCTAAGGACG-GTCTGAAGAGC-3Ј (upstream) and 5Ј-CCACGTATCCTAAAAGGG-CAC-3Ј (downstream), corresponding to nucleotides 2158 -2178 and 2503-2523 of the hGR␤ cDNA (1). Amplified DNA fragments were electrophoretically fractionated on 1.75% agarose gels. Restriction enzyme analysis of the RT-PCR products amplified by the hGR␣-and hGR␤-specific primers confirmed that these fragments contained the appropriate hGR␣ and hGR␤ cDNA sequences. Restriction enzymes employed included NsiI (specific to exon 8), SspI (specific to exon 9␣), and HaeII (specific to exon 9␤).
Quantitative RT-PCR-Human RNA (0.5 g) was reverse transcribed, and the resulting cDNA was amplified as described above. 5 l of the PCR reaction was removed at 2-cycle intervals and electrophoresed on 1.75% agarose gels stained with ethidium bromide. The intensity of the ethidium bromide fluorescence was measured densitometrically and plotted as a function of cycle number to generate amplification curves for the hGR␣ and hGR␤ PCR fragments. Regression equations of the form y ϭ a ϫ b n , where y is the intensity and n is the number of cycles, were fitted to the data in the linear portion of the semi-logarithmic graphs. The amplification efficiencies (E) for the hGR␣ and hGR␤ fragments, while nearly identical within tissues, varied slightly across tissues: E␣ ϭ 0.71 and E␤ ϭ 0.71 for CEM-C7 cells but E␣ ϭ 0.82 and E␤ ϭ 0.81 for human lung. Amplification curves were also generated (and regression equations fitted) for external standards containing linearized pCMVhGR␤ and pCMVhGR␣ at ratios of 1:1000, 10:1000, 100: 1000, and 1000:1000, respectively. A standard curve was generated by plotting the difference in the number of cycles required to amplify (during the exponential phase of each reaction) an identical amount of the hGR␣ and hGR␤ PCR products (3000 densitometric units) as a function of the pCMVhGR␤/pCMVhGR␣ ratio. Similarly, the difference in the number of cycles required to amplify (during the exponential phase of each reaction) the same amount of hGR␣ and hGR␤ PCR products (3000 densitometric units) for each human sample was calculated. Using the standard curve regression equation, the approximate hGR␤/hGR␣ cDNA (and hence mRNA) ratio was determined for each human tissue and cell line.
Sucrose Density Gradients and Western Blots-Subconfluent COS-1 cells (4 ϫ 10 6 ) were transfected by the calcium phosphate method as described above with equimolar amounts of pCMVhGR␣ (40 g), pCM-VhGR␤ (36.4 g), or pCMV5 (20.4 g). Each transfection was standardized to 40 g of DNA using pBR322. Cells were harvested 48 h posttransfection, resuspended in unsupplemented Dulbecco's minimum essential medium containing 100 nM [ 3 H]DEX or 20 nM [ 3 H]RU486, and incubated for 2 h on ice with gentle agitation. Whole cell extracts were prepared and processed on sucrose density gradients essentially as described by Tully and Cidlowski (24). To verify equivalent expression of the hGR␣ and hGR␤ proteins, an aliquot of each extract was analyzed by Western blotting. Proteins (125 g) were resolved by electrophoresis through 7.5% polyacryalmide gels and electrophoretically transferred to nitrocellulose. After incubating the membrane with epitope-purified polyclonal anti-hGR antiserum #57 (25) at a dilution of 1:2000, immunoreactivity was visualized using enhanced chemiluminescence accord-ing to the manufacturer's instructions (ECL, Amersham Corp.).
Immunohistochemistry-Subconfluent COS-1 cells (2 ϫ 10 5 ) were transfected with equimolar amounts of pCMVhGR␣ (20 g) or pCM-VhGR␤ (18.2 g) by the calcium phosphate method as described above or by the DEAE method of Sompayrac and Danna (26) as modified by Gorman (27). Each transfection was standardized to 20 g of DNA using pBR322. Following transfection, the cells were incubated for 24 h in Dulbecco's minimum essential medium supplemented with 10% dextran-coated charcoal-treated serum before plating in two-chamber glass slides. After an additional 24-h incubation, transfected cells were treated for 2 h with DEX (100 nM) or control vehicle and processed for immunohistochemical staining as described previously (25). Immunoreactivity was visualized by staining with avidin-biotin-peroxidase or Texas red fluorescent dye.
CAT Activity-Subconfluent COS-1, CV-1, and HeLa S 3 cells were transfected as indicated in the appropriate figure legends by the calcium phosphate method described above. 16 h post-transfection, medium was removed and replaced with control medium or medium containing DEX or RU486. Cells were harvested 24 h later and CAT assays were performed as described previously (22).

Northern Blot Analysis of the hGR␤ Message in Human Tissues and Cell
Lines-Northern blot analysis of RNA isolated from various human cell lines routinely shows multiple hGR mRNA transcripts approximately 7.0 and 5.0 kb in size (1, 28 -33). These transcripts are thought to arise from the use of alternative polyadenylation signals within the 3ЈUTR of exon 9␣ and are presumed to encode the hGR␣ isoform. Because early studies utilized probes that recognized a region common to the hGR␣ and hGR␤ messages, one or more of these transcripts might actually be the hGR␤ splice variant. In order to discriminate between hGR␣ and hGR␤ messages and to assess the relative amounts of these transcripts in a given tissue, we sequentially hybridized poly(A) ϩ RNA isolated from various human adult and fetal tissues with hGR␤and hGR␣-specific cRNA riboprobes. The hGR␤ riboprobe was designed to recognize a 578-nucleotide segment spanning the coding and proximal 3Ј noncoding regions of exon 9␤. The hGR␣ riboprobe was designed to recognize a 534-nucleotide segment in the distal 3ЈUTR of exon 9␣.
As shown in Fig. 1A (upper and lower panels), the hGR␤ probe hybridizes with an abundant message in all tissues migrating slightly below the 7.5-kb RNA marker. Additionally, a faint band slightly below the 4.4-kb RNA marker is observed in the adult heart, brain, placenta, lung, liver, skeletal muscle, and pancreas RNA samples and in the fetal brain, lung, and liver RNA samples. The length of the hGR␤ cDNA predicts the hGR␤ message to be at least 4.1 kb in size, thus the lower hybridization signal (approximately 4.3 kb in size) may correspond to the hGR␤ mRNA transcript. In contrast to the lower hybridization signal, the abundant message approximately 7.0 kb in size cross-reacts with the hGR␣ riboprobe (Fig. 1B, upper  and lower panels). The hGR␣ riboprobe also hybridizes with a message approximately 5.5 kb in size in many tissues and with a less abundant 4.4 kb message in several tissues, and these two transcripts do not appear to cross-react with the hGR␤ probe (Fig. 1B, upper and lower panels). Finally, the hGR mRNA transcripts approximately 7.0, 5.5, and 4.3 kb in size are all recognized by a probe made to the common coding region of the hGR␣ and hGR␤ cDNAs (data not shown). Similar hybridization patterns are also observed on Northern blots of RNA isolated from HeLa S 3 cells (a human cervical carcinoma cell line) and CEM-C7 cells (a human lymphoid cell line) (data not shown).
Characterization of the 7.0-kb hGR Message-Because the 7.0-kb hGR message is recognized by both the hGR␣ and hGR␤ riboprobes, it must contain information from both the 9␣ and 9␤ exons. These exons, therefore, do not appear to be mutually exclusive as previously reported (2). This observation made it important to determine whether the 7.0-kb message encodes the hGR␣ or hGR␤ isoform. Various data suggested that it would encode hGR␣. Exon 9␣ precedes exon 9␤ in the linear organization of the hGR gene (2). If this order is maintained in the mature message, exon 9␣ would form the distal coding and proximal 3Ј noncoding regions, and exon 9␤ would comprise the distal 3Ј noncoding region. A hGR message with this arrangement at its 3Ј end would be expected to encode the hGR␣ protein. Sequence conservation between hGR exons 9␣ and 9␤ (as well as the 155-bp intron separating these two exons) and the distal coding and 3Ј noncoding regions of the 6.5-kb rat GR message ( Fig. 2A) further suggested that the 7.0-kb hGR mRNA transcript contains (moving 5Ј to 3Ј) exon 9␣, intron J, and exon 9␤ sequences at its distal 3Ј end. If all 10 exons and intron J of the hGR gene are represented in a mature hGR message, this message would be approximately 6.8 kb in size, consistent with the size of the hybridization signal observed on our Northern blots.
To test whether the 3Ј end of the 7.0-kb hGR message is organized in this fashion, we performed RT-PCR using a sense 5Ј primer specific to the distal 3ЈUTR of exon 9␣ and an antisense 3Ј primer specific to the proximal 3ЈUTR of exon 9␤. If the 3Ј end of the 7.0-kb hGR message consists of sequences from exon 9␣, intron J, and exon 9␤, these primers will amplify a 933-bp PCR product. When total RNA extracted from HeLa S 3 cells is used for RT-PCR, a PCR product of this size is generated (Fig. 2B, lane 2). The PCR fragment is not produced when the reverse transcriptase is omitted from the reaction, demonstrating that contaminating DNA is not present (Fig. 2B, lane 1). In addition, restriction enzymes that cleave sites specific to exon 9␣, intron J, and exon 9␤ were used to confirm the sequence of the 933-bp PCR fragment (Fig. 2B, lanes 3-5). Therefore, exon 9␣ makes up the distal coding and proximal 3Ј noncoding regions of the 7.0-kb hGR message and both intron J and exon 9␤ form the distal 3Ј noncoding region of the 7.0-kb hGR message. This message would be expected to encode the hGR␣ isoform.
These RT-PCR results also demonstrate that sequences previously identified as intron J (2) are actually exonic sequences separating the 9␣ and 9␤ exonic sequences. In agreement with this finding, an oligonucleotide probe specific to intron J hybridizes on Northern blots with the 7.0-kb hGR␣ message (data not shown). Therefore, we propose that the hGR sequences formerly identified as exon 9␣, intron J, and exon 9␤ comprise one large terminal exon (exon 9) approximately 4.1-kb in size and that the hGR gene is organized into nine exons rather than the previously reported ten (2).
RT-PCR Analysis of the hGR␤ Message in Human Tissues and Cell Lines-The hGR␤-specific riboprobe also hybridizes with a faint message approximately 4.3 kb in size on the Northern blots shown in Fig. 1. The size of this message suggests that it might be the hGR␤ mRNA transcript. However, because the 7.0-kb hGR␣ message contains the 9␤ sequences at its 3Ј end, the 4.3-kb signal might instead be a degradation product of the 7.0-kb hGR␣ message. Alternatively, the 4.3-kb signal might represent nonspecific hybridization of the hGR␤ probe with 28 S rRNA. Recently, it was reported using RT-PCR that the alternative splicing event underlying the formation of the hGR␤ mRNA transcript occurs in many different human tissues (17). However, these experiments employed a very large number of PCR cycles, and these workers apparently did not control for potential amplification of contaminating DNA. Therefore, to confirm that the hGR␤ message is expressed, we performed RT-PCR on RNA isolated from various human tissues and cell lines using primers that hybridize on either side of the alternatively spliced region of the 4.3-kb hGR␤ message.
A sense 5Ј primer specific to exon 8 and an antisense 3Ј primer specific to the 9␤ sequences were utilized in the PCR reaction. If the alternative splicing event underlying the formation of the 4.3-kb hGR␤ message occurs (in which the end of exon 8 is linked to the 9␤ sequences located in the distal portion of exon 9) the hGR␤-specific primers will produce a PCR product 366-bp in length. If these primers hybridize with the 7.0-kb hGR␣ message (which also contains the 9␤ sequences at its far 3Ј end), they will generate a PCR product approximately 3000 bp in length. Conditions of our PCR amplification reaction did not favor production of this large PCR fragment, and it was never observed. For parallel analysis of the hGR␣ mRNA transcripts, an antisense 3Ј primer specific to the 9␣ sequences was used in combination with the same sense 5Ј primer. If the default splicing event underlying the formation of the hGR␣ messages occurs (in which the end of exon 8 is linked to the 9␣ sequences at the beginning of exon 9) these hGR␣-specific primers will produce a PCR product 477 bp in length.
When total RNA extracted from human heart, brain, lung, liver, and skeletal muscle is used for RT-PCR, a 366-bp PCR product is generated by the hGR␤-specific primers, suggesting that the hGR␤ message is present in these human tissues (Fig. 3A). In addition, the hGR␣-specific primers amplify the expected 477-bp PCR product in each tissue (Fig. 3B). When the reverse transcriptase is omitted from the RT-PCR reaction, the expected PCR fragments are not produced, indicating that only cDNA produced by the RT step is serving as template for the correctly sized PCR product. RT-PCR analysis was also performed on RNA isolated from HeLa S 3 and CEM-C7 cells (Fig. 3, A and B). Again, a 366-bp PCR fragment is produced by the hGR␤-specific primers, suggesting that the hGR␤ message is present in these transformed human cell lines. Consistent with our Northern blot data, the hGR␤ message appears to have a widespread tissue distribution.
Together, the Northern blot and RT-PCR analyses indicate that the hGR mRNA heterogeneity observed in human tissues and cell lines includes both hGR␣ and hGR␤ messages. The more abundant transcripts are approximately 7.0 and 5.5 kb in size and are expected to encode the hGR␣ isoform. Consensus polyadenylation signals are located at the end of the 9␣ sequences in exon 9, and use of these signals would generate a hGR␣ message approximately 1.6 kb shorter than the fulllength 7.0-kb hGR␣ message. These consensus signals are functional because they terminate transcription of the hGR␣ cDNA cloned into an expression vector lacking other polyadenylation signals (data not shown). Therefore, the 5.5-kb hGR␣ message appears to originate from alternative polyadenylation at these consensus sites. The less abundant hGR message recognized by the hGR␤-specific probe and approxi- The regions of greatest similarity and the percentage of identity between the two aligned sequences are indicated. Triangles identify consensus polyadenylation signals, and arrows indicate the location of PCR primers utilized in B. B, total RNA (1.0 g) from HeLa S 3 cells was reverse transcribed using random hexamers, and first strand cDNA was subsequently amplified with the addition of an upstream primer specific to the distal 3ЈUTR of exon 9␣ and a downstream primer specific to the proximal 3ЈUTR of exon 9␤. The resulting RT-PCR products were then analyzed by agarose gel electrophoresis. The reverse transcriptase was omitted in lane 1 but included in lane 2. In lanes 3-5, the PCR product was digested with restriction enzymes that cut specifically in exon 9␣ (Acc65I), intron J (HpaII), or exon 9␤ (HaeII). Sizes (in bp) of DNA markers (M) are indicated in the left margin. mately 4.3 kb in size is expected to encode the hGR␤ isoform. This mRNA transcript results from alternative splicing in which a 3Ј acceptor site preceding the 9␤ sequences in exon 9 is utilized by the splicing machinery rather than the normal 3Ј acceptor site preceding the 9␣ sequences in exon 9. The model shown in Fig. 4 summarizes the predicted structure at the 3Ј end of the hGR gene, primary transcript, and mature hGR␣ and hGR␤ mRNAs; the processing events underlying the formation of these mature messages; and the predicted translation products of these transcripts.
Based on the intensity of the Northern blot signals, the two hGR␣ messages (7.0 and 5.5 kb) are much more abundant than the 4.3-kb hGR␤ message (see Fig. 1). To more accurately assess the relative levels of the hGR␣ and hGR␤ mRNA transcripts, we performed quantitative RT-PCR on RNA isolated from adult lung, adult liver, HeLa S 3 cells, and CEM-C7 cells. Reaction cycle intensity curves for the 477-bp hGR␣ and 366-bp hGR␤ PCR products are shown for each tissue and cell line in Fig. 5A. For estimation of the hGR␤/hGR␣ mRNA ratio, regression equations were fitted to the linear portion of each amplification curve, and the difference in the number of cycles required to amplify an equal amount of hGR␣ and hGR␤ PCR product was calculated. Similarly, the difference in cycle number required to amplify an equal amount of hGR␣ and hGR␤ PCR product was determined for a series of external standards containing known hGR␤/hGR␣ cDNA ratios. Using the standard curve shown in Fig. 5B, the hGR␤/hGR␣ cDNA (and hence mRNA) ratio for each human sample was calculated and is as follows: 0.34% for lung, 0.21% for liver, 0.21% for HeLa S 3 cells, and 0.22% for CEM-C7 cells. Although these values reflect a large difference in expression levels, one should bear in mind that the amount of the 477-bp hGR␣ fragment is derived from two hGR␣ messages (7.0 and 5.5 kb), whereas the amount of 366-bp hGR␤ fragment comes only from the 4.3-kb hGR␤ message. In addition, our approach assumes that the efficiency of the RT reaction is the same for both the hGR␣ and hGR␤ mRNA transcripts. This may not be the case.
Ligand Binding Analysis of hGR␤-Operating under the

FIG. 3. RT-PCR analysis of RNA isolated from human tissues and cell lines using hGR␤and hGR␣-specific primers.
Total RNA (0.5 g) isolated from various human adult tissues (heart, brain, lung, liver, and skeletal muscle) and cell lines (HeLa S 3 and CEM-C7 cells) was reverse transcribed using random hexamers. The resulting cDNA was amplified using either hGR␤-specific (A) or hGR␣-specific (B) primers. For each set of primers, the reverse transcriptase was omitted in lanes 2, 4, 6, 8, 10, 12, and 14 but included in lanes 3,5,7,9,11,13,and 15. No  FIG. 4. Predicted structure of the hGR gene and gene products. hGR sequences formerly identified as exon 9␣, intron J, and exon 9␤ comprise one large exon (exon 9). Alternative processing of exon 9 generates multiple hGR messages. Specifically, splicing event #1 (default splicing pathway) in which the end of exon 8 is linked to beginning of exon 9 is predicted to generate the 7.0-and 5.5-kb hGR␣ messages, which differ in size due to the use of alternative polyadenylation signals. Splicing event #2 (alternative splicing pathway) in which the end of exon 8 is linked to the beginning of the 9␤ sequences is predicted to generate the 4.3-kb hGR␤ message. Translation of the messages produces the hGR␣ and hGR␤ isoforms, which are identical through amino acid 727 but then diverge. The functional domains and the putative site of hsp90 interaction are indicated for each isoform. Exons and introns (not to scale) are designated by boxes and lines, respectively. The arrows along the primary transcript identify the location of consensus polyadenylation signals. Splicing of introns A-G is not shown. The hGR␣and hGR␤-specific cRNA probes used in this study are indicated by solid lines.
assumption that the endogenous hGR␤ message is translated into the endogenous hGR␤ protein, we began investigating the biochemical properties of hGR␤ when overexpressed in transfected COS-1 cells. COS-1 cells contain undetectable levels of endogenous GR (3) and therefore provide a model system for studying the biochemical properties of hGR␤ in the absence of hGR␣. The carboxyl-terminal 50 amino acids of hGR␣ have been replaced in hGR␤ with 15 unique amino acids. To investigate the effect this has on the ability of hGR␤ to bind hormone or antihormone, we transfected COS-1 cells with a hGR␣ expression vector (pCMVhGR␣), a hGR␤ expression vector (pC-MVhGR␤), or the expression vector backbone (pCMV5). The  (Fig. 6A). In contrast, hGR␤ does not appear to bind either of these ligands, because the sedimentation profiles for the pCM-VhGR␤and pCMV5-transfected cells are superimposable. Western blot analysis of the extracts analyzed for hormone binding demonstrates that both the 94-kDa hGR␣ and 90-kDa hGR␤ isoforms were synthesized at similar levels (Fig. 6B). Therefore, lack of steroid binding by hGR␤ is not due to insufficient expression of the hGR␤ protein. Interestingly, the hGR␣ expression vector does not contain the hGR␤-specific sequences, yet a 90-kDa protein is observed in the pCMVhGR␣transfected cells (Fig. 6B). This protein, which cannot be hGR␤, appears to be a degradation product or post-translational modification of the 94-kDa hGR␣ isoform.
Subcellular Distribution of hGR␤-Numerous studies have shown that hGR␣ resides in the cytoplasm of cells in the absence of hormone and translocates to the nucleus in a hormonedependent manner (11). To determine where hGR␤ is localized within a given cell, COS-1 cells were transiently transfected with the hGR␣ or hGR␤ expression vectors. The plasmids were delivered into the cells using either the DEAE-dextran or calcium phosphate transfection method to rule out potential artifacts due to the transfection procedure. After treating the cells for 2 h with or without DEX, immunohistochemistry was performed, and immunoreactivity was visualized by staining with Texas red fluorescent dye or avidin-biotin-peroxidase. In the absence of hormone hGR␣ is found in the cytoplasm of the transfected cells, but following hormone administration it translocates and becomes predominantly nuclear (Fig. 7, left  panel). In contrast, the hGR␤ protein resides primarily in the nucleus of the transfected cells independent of hormone treatment (Fig. 7, right panel). These findings were consistent regardless of the transfection method or immunohistochemical staining procedure employed.
Transcriptional Activity of hGR␤ in the Absence of hGR␣-hGR␤ does not appear to bind hormone or antihormone, but it resides primarily in the nucleus of transfected COS-1 cells (and CV-1 cells) 2 and has been reported to bind a consensus GRE in vitro (17). Therefore, we investigated the ability of hGR␤ to activate or repress the glucocorticoid-responsive MMTV enhancer in the presence or the absence of steroid. For these experiments, COS-1 cells were cotransfected with an MMTV-CAT reporter plasmid (pGMCS) and either pCMV5, pCM-VhGR␣, or pCMVhGR␤. In response to DEX, hGR␣ induces a 4-fold increase in CAT expression over that observed in the vehicle-treated cells, whereas RU486 alone has little or no effect (Fig. 8A, lanes 4 -6). In contrast, CAT expression is unchanged in the hGR␤-containing cells treated with steroid, which is consistent with the inability of hGR␤ to bind these ligands (Fig. 8A, lanes 7-9). Moreover, hGR␤ does not appear to be constitutively active as an enhancer or repressor because levels of CAT activity in the vehicle-treated cells are similar whether they contain the CMV-vector backbone or pCMVhGR␤ (Fig. 8A, compare lanes 1 and 7).
The hGR␣-mediated induction of CAT activity in transfected COS-1 cells was only 4-fold. In cell lines where hGR␣ inductions are much greater, hGR␤ might display partial transcriptional activity. Therefore, we investigated the transcriptional activity of hGR␤ on the MMTV enhancer in receptor negative CV-1 cells. In response to DEX, hGR␣ induces a 74-fold increase in CAT expression (Fig. 8B, lanes 4 and 5). Interestingly, hGR␣ induces a 29-fold increase in CAT expression in response to RU486 (Fig. 8B, lanes 4 and 6). This partial agonist activity of RU486 has been reported previously and appears to be cell type-specific (37). Consistent with our findings in COS-1 cells, CAT expression is unchanged in CV-1 cells transfected with hGR␤ and treated with steroid (Fig. 8B, lanes 7-9). In addition, hGR␤ does not appear to be constitutively active (Fig.  8B, compare lanes 1 and 7). Thus, we conclude that in the absence of hGR␣, hGR␤ is transcriptionally inactive on the glucocorticoid-responsive MMTV enhancer.
Transcriptional Activity of hGR␤ in the Presence of hGR␣-In the absence of hGR␣, hGR␤ is unable to directly enhance or repress transcription of the MMTV-CAT reporter plasmid. To determine if hGR␤ influences gene expression indirectly by modulating the transactivation capacity of hGR␣, HeLa  amounts of pCMVhGR␤ (Fig. 9A). When pGMCS alone is transfected into the cells, DEX treatment results in a 50-fold induction of CAT activity. However, as increasing amounts of pCM-VhGR␤ are transfected into the cells, the glucocorticoidinduced, hGR␣-mediated activation of the MMTV enhancer is repressed in a dose-dependent manner. We observe a 78 and 96% inhibition of CAT expression when 1.0 and 3.0 g, respectively, of pCMVhGR␤ is used in the transfection mixture. Transfection of a 2-fold molar excess of the CMV-vector backbone (pCMV5) has no effect on hGR␣-mediated induction of CAT expression, indicating that hGR␤ is responsible for the repression.
We next evaluated whether the dominant negative activity of hGR␤ on hGR␣-mediated transcription is restricted to the MMTV promoter or is a general property of glucocorticoidresponsive promoters. The pGRE2CAT reporter plasmid contains two copies of the GRE consensus sequence derived from the tyrosine aminotransferase gene and a TATA box element upstream of the CAT gene (20). In contrast to MMTV, this "minimal promoter" does not contain binding sites for other ancillary transcription factors. When HeLa S 3 cells are cotransfected with pGRE2CAT and various amounts of pCMVhGR␤ (Fig. 9B), the hGR␣-mediated stimulation of CAT expression is inhibited in a dose-dependent manner, and the repression of hGR␣ activity is similar to that observed on the MMTV pro-moter. The inhibitory action of hGR␤ is restricted to glucocorticoid-responsive promoters because the hGR␤ protein has no effect on the constitutively active, nonglucocorticoid-responsive thymidine kinase CAT reporter plasmid (pBLCAT2) (data not shown). In sum, these results suggest that hGR␤ represses the function of hGR␣ by specifically inhibiting GRE-mediated transcription.

DISCUSSION
Alternative splicing of the hGR primary transcript produces two highly homologous isoforms, termed hGR␣ and hGR␤, which differ at their carboxyl termini (1,2). In contrast to the well characterized hGR␣ isoform, very little is known about the hGR␤ splice variant. In this report, we examine the expression, biochemical properties, and physiological function of hGR␤. Northern blot analysis with a hGR␤-specific riboprobe demonstrates the presence of a message approximately 4.3 kb in size (consistent with the length of the hGR␤ cDNA) in many different human tissues. We subsequently confirmed by RT-PCR that the alternative splicing event proposed to underlie the formation of the hGR␤ mRNA transcript occurs in these tissues as well as in several transformed human cell lines. Together, these results indicate that the hGR␤ message is endogenous to a variety of cells. Because both the hGR␣ and hGR␤ messages are co-expressed in many of the same tissues, previous studies investigating hGR␣ expression with probes that did not discriminate between hGR␣ and hGR␤ may be in error. The hGR␤ mRNA transcript generated in vivo from the pCMVhGR␤ expression vector is efficiently translated into the 90-kDa hGR␤ protein in transfected COS-1, CV-1, and HeLa S 3 cells, suggest- ing that the endogenous hGR␤ message can also serve as a template for protein synthesis. However, whether the endogenous hGR␤ message is actually translated into the hGR␤ isoform is unknown. Anti-hGR antibodies made to date in several different laboratories have epitopes in the amino terminus and thus recognize both the hGR␣ and hGR␤ proteins. The small difference in size between the two isoforms and the potential for hGR␣ to be post-translationally modified or degraded into a smaller protein make this cross-reactivity undesirable. To test directly for the expression of the hGR␤ protein, we are presently making a hGR␤-specific antibody.
During our investigation of the expression of the hGR␤ message, we made several observations that provide new insights both into the structure of the hGR gene and hGR␣ mRNAs and into the expression of the hGR␣ messages. The hGR gene has been previously reported to consist of 10 exons (2). Results from our Northern blot and RT-PCR analyses suggest that the last two exons, 9␣ and 9␤, and the intronic sequences separating these two exons (intron J) together form one large terminal exon (exon 9). Exon 9 encodes the 3Ј end of the hormonebinding domain of the hGR␣ protein (under normal splicing conditions) and contains approximately 4.0 kb of 3ЈUTR. It is interesting to note that the genes for the human androgen receptor, human estrogen receptor, and chicken progesterone receptor show a similar organization to that we have proposed here for the hGR. In each case, the most 3Ј exon (exon 8) encodes the COOH-terminal portion of the hormone-binding domain and specifies a very large 3ЈUTR (38 -42). The differential hybridization of the hGR␣-and hGR␤-specific riboprobes with the 7.0-and 5.5-kb hGR␣ messages suggests that they originate from the use of alternative polyadenylation signals located in the 4.0-kb 3ЈUTR of exon 9. Similar alternative polyadenylation events have been proposed to explain the GR mRNA heterogeneity observed in rat tissues (35). Within the 3ЈUTR of eukaryotic mRNA reside signals that influence mRNA localization, mRNA stability, and translation initiation (43,44). The 7.0-and 5.5-kb hGR␣ messages may differ in one or more of these properties. Both hGR␣ messages have a widespread tissue distribution, although the 5.5-kb hGR␣ message does not appear to be expressed at high levels in the brain. In addition, the 7.0-kb message is consistently more abundant than the 5.5-kb message.
The 7.0-and 5.5-kb hGR␣ messages are much more abundant than the 4.3-kb hGR␤ mRNA transcript. Quantitative RT-PCR analysis of RNA isolated from two adult tissues and two human cell lines suggests that there is 200 -500-fold more hGR␣. However, because alternative splicing is often regulated in a spatial and/or temporal fashion, the hGR␤ message may be expressed at higher levels in a tissue-specific and/or developmental stage-specific manner. In addition, the hGR␣/hGR␤ mRNA ratio may or may not reflect the hGR␣/hGR␤ protein ratio due to potential differences in stability and/or translation efficiency of the hGR␣ and hGR␤ messages and/or due to differences in protein half-life. Furthermore, because the 7.0-kb hGR␣ message has the 9␤ sequences at its far 3ЈUTR, we cannot exclude the possibility that it also encodes the hGR␤ protein. Development of hGR␣-and hGR␤-specific antibodies will provide insight into this issue. Interestingly, the 9␤ sequences are well conserved in the rat GR cDNA 3ЈUTR, suggesting that an hGR␤ homolog may exist in rat. RT-PCR analysis of rat liver (as well as a mouse lymphoma cell line) does indeed produce a PCR product that comigrates with the 366-bp PCR fragment derived from human cells. 2 Thus, although the hGR␤ message is expressed at low levels relative to the hGR␣ mRNA transcripts in the human tissues so far examined, its conservation across species suggests that it plays an important physiological role.
With few exceptions (45,46), modification of the GR hormone-binding domain results in a reduction or complete loss of hormone binding (3,(47)(48)(49)(50). The COOH-terminal 50 amino acids of hGR␣ have been replaced in hGR␤ with 15 unique amino acids. In agreement with previous reports (1,3,17), we show that this natural COOH-terminal modification prevents agonist binding to the hGR␤ protein. Recently, it was reported that a truncated version of the human progesterone receptor B form missing the COOH-terminal 42 amino acids did not bind progesterone or the synthetic agonist R5020 but still bound the antiprogestin RU486 (51). This finding suggested that amino acids at the extreme COOH terminus of the human progesterone receptor are critical for agonist but not antagonist binding. Because members of the steroid hormone receptor superfamily share many of the same properties, we tested the ability of hGR␤ to bind RU486 but found no evidence of binding. The 15 amino acids at the end of hGR␤ may prevent the association of RU486 with this isoform. Alternatively, the observation made for human progesterone receptor B form may not be conserved among other family members. At this time, hGR␤ is more aptly described as an orphan receptor whose natural ligand, if any, is unknown.
The hGR␣ receptor isoform translocates from the cytoplasm to the nucleus in a hormone-dependent manner (11). In the absence of hormone, the association of hsp90 with hGR␣ appears to inactivate the NLS (52,53). Once hormone binds hGR␣, hsp90 dissociates from the receptor resulting in the activation of the NLS and subsequent nuclear import of hGR␣ (8). In contrast to hGR␣, we demonstrate that hGR␤ resides primarily in the nucleus of transfected cells independent of hormone treatment. The amino acids necessary for interacting with hsp90 (7) are present in hGR␤, suggesting that hGR␤ may be in the nucleus in spite of its association with hsp90. Perhaps the unique COOH-terminal amino acids of hGR␤ delete sequences that inhibit the NLS or slightly alter the tertiary structure of the hGR␤/hsp90 complex such that the NLS are partially activated. This might account for our observation that most, but not all, of the hGR␤ protein is located in the nucleus. Further studies will be required to elucidate the precise mechanism(s) underlying the nuclear distribution of hGR␤.
In the absence of hGR␣, the hGR␤ isoform is transcriptionally inactive on the MMTV enhancer independent of steroid treatment. However, when hGR␣ and hGR␤ are expressed in the same cell, hGR␤ inhibits the glucocorticoid-induced, hGR␣mediated activation of the MMTV promoter. Although this dominant negative effect was first reported in COS-7 cells cotransfected with hGR␣ and hGR␤ expression vectors (17), we have extended this initial observation in several respects. First, we demonstrate that this dominant negative activity occurs in cells that have endogenous hGR␣ receptors. In addition, we show that the repression of hGR␣ activity occurs with the simple promoter pGRE2CAT. This indicates that the repression is a general phenomenon of glucocorticoid-responsive promoters and that it is GRE-mediated transcription that is actually inhibited.
The mechanisms responsible for the hGR␤-mediated repression of hGR␣ activity are unknown. The hGR␤ protein is primarily located in the nucleus of transfected cells, has an intact DNA-binding domain and has been reported to bind a consensus GRE in vitro (17). Therefore, it may compete with hGR␣ for binding to the GRE. Another possibility is that hGR␤ forms a heterodimer with hGR␣ that is transcriptionally inactive or less active than an hGR␣ homodimer. Alternatively, the hGR␤ isoform may inhibit the function of hGR␣ by interacting with and titrating out an essential cofactor needed by hGR␣ for full transcriptional activity. We are currently trying to identify the mechanism(s) responsible for the dominant negative activity of hGR␤ and to determine whether hGR␤ can inhibit hGR␣-mediated repression of gene expression. Moreover, we hope to discern if hGR␤ exhibits its dominant repressive effect on other members of the closely related subgroup of nuclear receptors that includes the progesterone, androgen, and mineralocorticoid receptors.
We have demonstrated that an alternatively spliced form of the hGR is present in many different tissues and is able to antagonize the physiological function of its predominant gene product. Alternative splicing plays a critical role in regulating the activity of several other members of the steroid/thyroid/ retinoic acid receptor superfamily. Most closely resembling that observed for hGR␣ and hGR␤ is the processing that occurs at the thyroid hormone receptor ␣ subunit (TR␣) locus. Alternative splicing of the last exon generates two receptor isoforms, TR␣1 and TR␣2, that differ at the carboxyl terminus (36). The TR␣2 isoform does not bind thyroid hormones, but it represses the transcriptional activity of TR␣1 by competing with TR␣1 for binding to the thyroid hormone receptor responsive elements (36). Clearly, the ability of steroid/thyroid/retinoic acid receptor genes to encode transcription factors with opposing biological activities adds another level of complexity to the regulation of the function of these receptors.