Functional interaction of human Cdc37 with the androgen receptor but not with the glucocorticoid receptor.

Cdc37 is a molecular chaperone closely associated with the folding of protein kinases. Results from studies using a yeast model system showed that it was also important for activation of the human androgen receptor (AR). Based on results from the yeast model system (Fliss, A. E., Fang, Y., Boschelli, F., and Caplan, A. J. (1997) Mol. Biol. Cell 8, 2501-2509), we initiated studies to address whether AR and Cdc37 interact with each other in animal cell systems. Our results show that Cdc37 binds to AR but not to glucocorticoid receptors (GR) synthesized in rabbit reticulocyte lysates. This binding occurs via the ligand-binding domain of the AR in a manner that is partially dependent on Hsp90 and the presence of hormone. Further studies using the yeast system showed that Cdc37 is not interchangeable with Hsp90, suggesting that it functions at a distinct step in the activation pathway. Expression of a dominant negative form of Cdc37 in animal cells down-regulates full-length AR but has very little effect on an AR truncation lacking the ligand-binding domain or full-length GR. These results reveal differences in the mechanisms by which AR and GR become active transcription factors and strengthen the notion that Cdc37 has a wider range of polypeptide clients than was realized previously.

Cdc37 is a molecular chaperone closely associated with the folding of protein kinases. Results from studies using a yeast model system showed that it was also important for activation of the human androgen receptor (AR). Based on results from the yeast model system (Fliss, A. E., Fang, Y., Boschelli, F., and Caplan, A. J. (1997) Mol. Biol. Cell 8, 2501-2509), we initiated studies to address whether AR and Cdc37 interact with each other in animal cell systems. Our results show that Cdc37 binds to AR but not to glucocorticoid receptors (GR) synthesized in rabbit reticulocyte lysates. This binding occurs via the ligand-binding domain of the AR in a manner that is partially dependent on Hsp90 and the presence of hormone. Further studies using the yeast system showed that Cdc37 is not interchangeable with Hsp90, suggesting that it functions at a distinct step in the activation pathway. Expression of a dominant negative form of Cdc37 in animal cells down-regulates full-length AR but has very little effect on an AR truncation lacking the ligand-binding domain or fulllength GR. These results reveal differences in the mechanisms by which AR and GR become active transcription factors and strengthen the notion that Cdc37 has a wider range of polypeptide clients than was realized previously.
Steroid hormone receptors comprise a family of proteins that bind their steroid ligands inside cells, in either the cytosol or the nucleus. Their conformation is stabilized prior to ligand binding by a group of proteins known collectively as molecular chaperones. In addition to maintaining apo-receptors in a stable conformation, molecular chaperones also appear to prevent ligand-independent activation (1).
Many studies performed over the last 10 years have led to the identification of molecular chaperones that have a general function in the activation of steroid hormone receptors. Hsp90, for example, interacts with the androgen receptor (AR), 1 aryl hydrocarbon receptor, estrogen receptor, glucocorticoid receptor (GR), mineralocorticoid receptor, and the progesterone receptor (PR) and is important for ligand binding by all of these receptors (reviewed in Ref. 1). The pathway leading to Hsp90 association involves the action of several other chaperones and co-chaperones (or helpers). The minimal complement of chaperones and co-chaperones necessary for efficient folding includes Hsp70, Hsp40, Hop, and p23 in addition to Hsp90 (2,3). The pathway by which these chaperones act has common requirements among several receptors, but there are also some differences. Common requirements include the actions of Hsp70 and Hsp90 (4,5). The co-chaperone called Hop (Hsp organizing protein; Ref. 6) binds to both Hsp90 and Hsp70, forming a bridge between them. A complex containing the receptor, Hsp70, Hop, and Hsp90 may be replaced by complexes containing just Hsp90, p23, and immunophilins such as cyp40 or FKBP52. The function of p23 is unknown; it is important for hormone binding by GR and PR (7) (3) yet has been shown recently to inhibit the activity of AR (8). In a second case, AR activity was shown to be stimulated by the long form of Bag-1 (Bag1-L), a co-chaperone of Hsp70, whereas this same isoform inhibited hormone-dependent signaling by GR (9,10). The AR and GR also respond differently to deletion of yeast YDJ1, which encodes an Hsp40 co-chaperone. In a ⌬ydj1 yeast strain, GR exhibits increased levels of hormone-independent reporter gene expression (11,12), whereas AR does not (13,14). These examples suggest that different receptors, even those so closely related that they can bind to the same hormone-responsive DNA elements, have different molecular chaperone requirements for activation.
Another example of a differential chaperone requirement for steroid receptor activation involves Cdc37. Cdc37 is required for the folding of several protein kinases and was proposed to be kinase specific based on its failure to interact with aryl hydrocarbon receptor, estrogen receptor, GR, or PR in vitro (15)(16)(17). On the other hand, genetic studies using the yeast Saccharomyces cerevisiae as a model system showed that Cdc37 was important for hormone-dependent activity by AR but not by GR (18). These studies also revealed that Cdc37 functions at a later stage in the folding process compared with Hsp90 or Ydj1p. This was implied from the results of experiments showing that mutation in CDC37 affected transactivation by AR but not hormone binding, whereas in hsp82 and ydj1 mutants there was a strong correlation between loss of activity and decreased affinity for hormone (14,19).
Several studies have shown that Cdc37 can bind to protein kinases, both in association with and independently of Hsp90.
The protein kinase catalytic domain itself appears to be the target for interaction with Cdc37 (20), and it can bind to or affect the activity of a broad range of kinase subtypes, including casein kinase II (21), cyclin-dependent kinases (22) (23), the mitogen-activated protein kinase Ste11 (24), Mps1, a kinase required for spindle pole body duplication (25), the nonreceptor tyrosine kinase v-Src (26), and Raf (27). The role of Cdc37 in the folding process remains unclear, although recent studies have begun to address its mechanism of action. One suggestion was that Cdc37 targets Hsp90 for subsequent chaperone action (23). On the other hand, overexpression of yeast Cdc37 can suppress the defect in v-Src activity in an hsp82 mutant strain, suggesting that it can act independently as a molecular chaperone (28). Perhaps the most likely scenario is that Cdc37 has chaperone activity that is closely tied to the presence of Hsp90 and other co-chaperones. Recent studies have shown that the normally salt labile interaction between Hsp90 and Cdc37 is made salt-resistant in the presence of a client kinase. Furthermore, Cdc37 is present in multi-chaperone complexes containing p23 and immunophilins (29). This suggests that Cdc37 functions late in the folding pathway compared with other chaperones such as Hop/Hsp70/Hsp40, which may function further upstream, as they do in steroid hormone receptor folding.
Given that Cdc37 functions in association with protein kinases, it remained possible that the defect in AR transactivation that we observed in a yeast cdc37 mutant could have resulted from defective kinase activity. Support for this possibility was strengthened by the finding that the AR is regulated in mammalian cells, although only in part, by phosphorylation (30 -32). We therefore initiated a biochemical approach to assay for Cdc37 binding to the AR. The rationale for this approach was based on the hypothesis that interaction between these proteins would only occur if Cdc37 had a direct effect on the ability of AR to adopt the active state, as it does with v-Src but not with the GR, which were used as controls for the experiments described below.

EXPERIMENTAL PROCEDURES
Materials-Dihydrotestosterone (DHT) was from Sigma and was stored in ethanol at Ϫ20°C. 3 H-R1881 (methyltrienelone) and [ 35 S]methionine were obtained from PerkinElmer Life Sciences. Geldanamycin was obtained from the Drug Synthesis and Chemistry Branch, Developmental Therapeutics Program, Division of Cancer Treatment, National Cancer Institute, and stored in dimethyl sulfoxide at Ϫ20°C. Ni-NTA resin was obtained from Qiagen. Pfu polymerase and the pPCR-Script vector was from Stratagene. Isopropyl-␤-D-thiogalactopyranoside was from Roche Molecular Biochemicals and stored as a 100 mM solution in water at Ϫ20°C. Imidazole was obtained from Sigma. T3, T7, and Taq polymerases were obtained from Promega. Rabbit reticulocyte lysates (treated) were obtained from Promega. Antisera to Hsp90, Cdc37, and the AR were described previously (18,19). Nitrocellulose membranes were from Immobilon and polyvinylidene difluoride membranes were from Millipore.
Cloning, Expression, and Purification of human His 6 Cdc37-Human Cdc37 with an N-terminal six-histidine tag was prepared from human Cdc37 cDNA (Ref. 23; kind gift of Dr. J. Wade Harper). The Cdc37 gene was amplified by PCR using the primers: 5Ј-CATATGGTGGACTA-CAGCGTG-3Ј and 5Ј-GGATCCGCAGGTGGCGGTGGTAGC-3Ј. The recombinant Cdc37 gene was amplified by 20 cycles at 55°C annealing temperature and 72°C elongation temperature using Pfu polymerase. The 1140-base pair product was gel purified before subcloning into pcrScript. The Cdc37 gene was excised from pcrScript with NdeI and BamHI and ligated into similarly digested pET15b, which contains an in-frame six-histidine tag. His 6 Cdc37 was expressed in Escherichia coli strain BL21DE3. The strain was grown overnight in 100 ml of LB medium (with 100 g/ml ampicillin) at 37°C. The culture was diluted 1:10 with fresh LB and incubated at 37°C for an additional hour, and Cdc37 expression was induced by addition of isopropyl-␤-D-thiogalactopyranoside to a final concentration of 1 mM for 2 h. The cells were harvested by centrifugation and washed once with water and once with extraction buffer (20 mM Hepes pH 7.5, 100 mM KCl, 0.1 mM EDTA pH 8). The cells were resuspended in 5 ml of extraction buffer containing a protease inhibitor mixture (1 mM phenylmethylsulfonyl fluoride, aprotinin, chymostatin, leupeptin, and pepstatin each at 1 g/ml). The cells were lysed by sonication with 5-s bursts (and 1 min rests at 4°C). Cell debris was pelleted, and the supernatant was transferred to a fresh tube and spun at 100,000 ϫ g for 20 min at 4°C. The supernatant was transferred to a fresh tube and incubated with Ni-NTA resin at a ratio of 1 ml of packed resin/5 ml of extract. This was incubated for 30 min at 4°C on a nutator. The resin was washed three times in extraction buffer containing 10 mM imidazole before elution of bound His 6 Cdc37 with extraction buffer containing 150 mM imidazole (at 2:1 ratio over packed resin) for 10 min at 4°C. The elution was performed three times, and eluates were dialyzed overnight at 4°C against 20 mM Tris-HCl pH 7.5, 1 mM EDTA, 25 mM NaCl, and 0.5 mM phenylmethylsulfonyl fluoride. The dialyzed protein was stored at Ϫ80°C. The His 6 Cdc37 was typically purified to ϳ90% by this procedure. Protein concentration was determined by the Bradford assay using bovine serum albumin as a standard.
In Vitro Transcription, Translation, and His 6 Cdc37 Binding Reactions-A plasmid containing the human AR (pSP72hAR-1) used for in vitro transcription was the kind gift of Dr. E. Wilson. Plasmids encoding GR and v-Src used for in vitro transcription were constructed by subcloning each gene into pBluescript. The parent plasmid for rat GR was the kind gift of Dr. K. Yamamoto (pGN795), and that for v-Src was the kind gift of Dr. F. Boschelli (pRS316.v-Src). Plasmids encoding truncated versions of the AR were synthesized by PCR amplification from pSP72hAR-1. pJR20 (AR 1-625; TAD-DBD) was constructed using the primers: 5Ј-ATGGAAGTGCAGTTAGGG-3Ј and 5Ј-TCAAGTCATCCCT-GCTTCATA-3Ј. The product was subcloned into pPCR-Script and mRNA synthesized using BamHI linearized plasmid and T7 RNA polymerase. pJR21 (AR 624 -919; LBD) was synthesized using 5Ј-AT-GACTCTGGGAGCCCGG-3Ј and 5Ј-TCACTGGGTGTGGAAATA-3Ј. The product was subcloned into pPCR-Script and mRNA synthesized using EagI linearized plasmid and T3 RNA polymerase. Messenger RNAs for AR, GR, and v-Src were synthesized using T3 or T7 RNA polymerases. mRNAs for AR, AR truncations, GR, and v-Src were translated in rabbit reticulocyte lysates (RRL) according to instructions provided by the supplier (Promega). Geldanamycin (100 g/ml) or DHT (100 nM) were added prior to the start of translation reactions. Control reactions containing solvent alone (Me 2 SO for geldanamycin and ethanol for DHT) were performed at the same time. After 2-h translation reactions (typically 25 l), the lysates were diluted to 400 l with extraction buffer (20 mM Hepes pH 7.5, 100 mM KCl, 0.1 mM EDTA pH 8) and incubated with 80 g of His 6 Cdc37 that was prebound to Ni-NTA resin (15 l of packed beads/reaction). The binding reaction was incubated at 4°C for 1 h. The resin was pelleted, washed three times with extraction buffer containing 10 mM imidazole, and eluted with 0.5 ml extraction buffer containing 150 mM imidazole on ice for 10 min. Eluted proteins were precipitated with 10% trichloroacetic acid for 20 min on ice. Precipitates were resuspended in SDS-polyacrylamide gel electrophoresis sample buffer. The samples were resolved by denaturing gel electrophoresis. The gels were fixed in 20% methanol and 10% acetic acid, washed three times in water, and incubated with 1 M sodium salicylate for 20 min. The gels were dried and exposed to x-ray film. In each case, binding of 35 S-labeled proteins to His 6 Cdc37 was compared with the level of labeled protein contained in 1% of the translation reaction.
Expression of Yeast His 6 yCdc37 and Isolation of Complexes-An Nterminal six-histidine-tagged version of yeast Cdc37 was constructed by PCR amplification. The template for the reaction was pRSS2 (26), which contained wild type CDC37. This was amplified over 10 cycles at 37°C annealing temperature and 72°C elongation temperature using Taq polymerase. The primers were: 5Ј-AAGCTTATGCACCACCAC-CACCACCACGCCATTGATTACTCTAAG-3Ј and 5Ј-GGTACCGCTA-CATAAATTTCTA-3Ј. The 1.55-kilobase product was gel purified and ligated into pcrScript. The His6yCDC37 gene was ligated into a 2-m plasmid (pRS424; TRP1) that also contained the ADH1 promoter to form pPL2. This plasmid and pARH were transformed into wild type yeast strain W3031b. Transformants were grown in selective media, and extracts were prepared as described previously (33). His 6 yCdc37 was isolated after incubation of Ni-NTA resin with 0.5 ml of whole cell extract at 3 mg/ml for 1 h at 4°C on a nutator. The resin was washed three times in extraction buffer (as above) containing 10 mM imidazole, and bound proteins were eluted with extraction buffer containing 150 mM imidazole as described above. Eluted proteins were precipitated with 10% trichloroacetic acid and redissolved in SDS-polyacrylamide gel electrophoresis sample buffer. These samples were resolved by denaturing gel electrophoresis and transferred to polyvinylidene diflu-oride membranes. The membranes were processed for Western blot as described below.
In Vivo Hormone Binding Assays-Wild type yeast strain p82a (34) and the isogenic Hsp90 mutant strain G313N (33) were transformed by pARH, which constitutively expresses the human AR and pGalCDC37 (2 m, URA3; Ref. 18). The cells were grown in glucose or galactose containing selective medium to A 600 ϭ 0.2, and 1-ml aliquots were incubated with 100 nM 3 H-R1881 (diluted 1:5 with unlabeled R1881 of the same concentration), plus or minus 100-fold excess of cold R1881. The cells were incubated at 30°C for 1.5 h with shaking. The cells were pelleted and washed three times with ice-cold water. The cells were counted in 5 ml of scintillation fluid. Specific binding was calculated by subtracting the counts retained by cells incubated with 100-fold excess of unlabeled R1881.
Transfection-A plasmid encoding Cdc37 for expression in animal cells was constructed by subcloning the full-length Cdc37 gene as an EcoRI/XbaI fragment into pCMV5. The dominant negative Cdc37 (Cdc37 1-173 ) was isolated as a 500-base pair fragment by EcoI/RsaI digestion of the EcoRI/XbaI fragment and subcloned into pCMV5. CV1 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Hyclone). Cells were plated at 50% confluence for transfection by the calcium phosphate precipitation method, in Dulbecco's modified Eagle's medium containing 5% charcoal-treated NuSerum IV (Collaborative Research. The 3xHREtkCAT reporter and AR and GR expression vectors have been described (35,36). After overnight incubation with the DNA precipitate (1 g), cells were washed and refed media with or without hormone (10 Ϫ7 M dihydrotestosterone for AR and dexamethasone for GR). Cells were harvested after 2 days, and CAT activity was measured as described previously (35).
Western Blot-Proteins were transferred to nitrocellulose or PDVF membranes using a semi-dry Transblot apparatus. Filters were washed briefly with TTBS (20 mM Tris-HCl, pH 7.5, 0.5 M NaCl, 0.05% Tween 20) and blocked overnight with TTBS containing 5% nonfat dry milk. Filters were incubated with primary antibodies (usually diluted 1:1000 in antibody dilution buffer; 1ϫ phosphate-buffered saline, 3% bovine serum albumin, 0.05% Tween 20, and 0.1% thimerosal) for 2 h. Filters were washed three times for 10 min each in TTBS. Filters were incubated with secondary antibody (horseradish peroxidase-conjugated goat anti-mouse IgG, diluted 1:10,000 in antibody dilution buffer) for 1 h and were subsequently washed as for the primary antibody. Filters were then treated with the chemiluminescence reagent (Pierce) and exposed to x-ray film for detection.

RESULTS
Interaction of Human Cdc37 with AR-We initiated biochemical studies to address whether Cdc37 could interact with the AR. An N-terminal His 6 -tagged version of human Cdc37 (His 6 Cdc37) was constructed and overexpressed in E. coli. The recombinant protein was purified by one step affinity chromatography on Ni-NTA resin (Fig. 1A). Binding reactions were performed with 35 S-labeled AR translated in rabbit reticulocyte lysatesRRL after addition of recombinant His 6 Cdc37. Interaction between AR and His 6 Cdc37 was monitored after reisolation of His 6 Cdc37 on Ni-NTA resin, washing the complex in buffers containing a low concentration of imidazole, and elution of bound proteins with a high concentration of imidazole. Eluted proteins were resolved by denaturing gel electrophore-sis, and labeled binding partners of Cdc37 were detected by fluorography. As shown in Fig. 1B, the translated AR appears primarily as a single band of ϳ110 kDa, and this was pulleddown on Ni-NTA resin that was preadsorbed with His 6 Cdc37 (Fig. 1B, lane 3). There was very little binding of AR to Ni-NTA resin in the absence of preadsorbed His 6 Cdc37 (Fig. 1B, lane 2). The amount of AR that was captured in these assays was ϳ1% of the total amount translated (1% of the translation was loaded in lane 1 of Fig. 1B as a standard). Similar experiments were performed with v-Src, a well characterized Cdc37 binding protein (37), and ϳ1% of the translation product was also captured on His 6 Cdc37:Ni-NTA resin using the same conditions (Fig. 1B, lanes 7-9). In other experiments we tested whether the AR was binding to contaminants from E. coli that copurified with His 6 Cdc37. However, no AR was pulled-down on Ni-NTA resin that was preadsorbed with E. coli lysates not containing His 6 Cdc37 (not shown). The specificity with which Cdc37 interacted with the AR was examined further by testing whether His 6 Cdc37 could also interact with the GR. In vitro translation of rat GR resulted in the appearance of several bands, the largest of which was of the size expected for the full-length product (94 kDa). However, the GR did not associate with His 6 Cdc37 when assayed by itself (Fig. 1B) nor when mixed with the AR (data not shown). The results shown here for the GR and v-Src are consistent with those described previously by Whitelaw et al. (17), where Cdc37 was found to coimmunoprecipitate with v-Src but not GR in lysates from cultured animal cells (17).
The binding site for Cdc37 on the AR was examined next. Based on results using a yeast model system, we had predicted that Cdc37 would function via the LBD. The rationale for this was 2-fold. First was that defects in AR activity only occurred with the full-length molecule; AR truncations containing just the transactivation and DNA-binding domains (TAD-DBD) were fully functional in the cdc37 mutant (18). This suggested that Cdc37 was not involved in folding of the TAD-DBD truncation. Second, Hsp90 is known to interact with the AR-LBD, and if Hsp90 and Cdc37 functioned together (as discussed in more detail below), it seemed likely that this would occur via the LBD. We therefore tested whether His 6 Cdc37 could interact with two truncated forms of the AR: one containing the TAD-DBD (1-625) and a second containing the ligand-binding domain and hinge regions (LBD; 624 -919). These were translated in RRL and assayed for binding to His 6 Cdc37 as described above. The results of these experiments (Fig. 2) showed that His 6 Cdc37 does indeed interact with the LBD to a similar extent as that observed with full-length AR. Much lower, but reproducible, levels of His 6 Cdc37 bound to the TAD-DBD, although these were not much above the background levels.
The role of Hsp90 in His 6 Cdc37 binding to the LBD was analyzed by performing AR translation in the presence of geldanamycin, an Hsp90 inhibitor. Geldanamycin is known to prevent association of Hsp90⅐Cdc37 complexes with protein kinases and inhibit their maturation (27,38,39). Geldanamycin also inhibits maturation of steroid hormone receptors including the AR, although to a lesser extent than for estrogen receptor or GR (40). In the experiment shown in Fig. 3, geldanamycin addition (100 g/ml) to RRL did not affect the level of AR translation but did affect the ability of AR to interact with His 6 Cdc37 by 2-3 fold (Fig. 3A). Similar results were obtained for v-Src (Fig. 3B), which also migrated slightly faster in the gel after geldanamycin treatment. Further studies revealed that v-Src synthesized in the presence of 100 g/ml geldanamycin is catalytically inactive, 2 suggesting that the geldanamycin treatment is completely effective at the concentration used. The similarity with which geldanamycin treatment disrupted interaction of His 6 Cdc37 with AR and v-Src suggested that both complexes formed by a similar mechanism. In the case of v-Src and other kinases, Cdc37 associates in a complex with Hsp90 and independently of it (20) (27). Thus, the partial disruption of such complexes by geldanamycin suggested that His 6 Cdc37 interacted with the AR in a manner that is partially dependent on Hsp90 but also partially independent.
Further investigation of Cdc37 interaction with the AR was performed in the yeast system. For these studies, we constructed a His-tagged version of yeast Cdc37 (His 6 yCdc37). The His 6 yCdc37 protein was fully functional because it suppressed the temperature-sensitive growth phenotype of a cdc37 mutant strain (18) when constitutively expressed from a low copy number CEN plasmid (not shown). To assay for binding, His 6 yCdc37 was constitutively expressed from a multi-copy plasmid in a wild type yeast strain also expressing the AR (Fig.  4). Incubation of lysates from this strain with Ni-NTA resin led to adsorption of His 6 yCdc37, as detected by Western blot. Western blot analysis also revealed that AR was coadsorbed with His 6 yCdc37, although Hsp90 was not (Fig. 4A, middle and  lower panels). There was very little nonspecific adsorption of AR to the Ni-NTA resin in experiments using lysates from strains expressing the AR but not His 6 yCdc37 (Fig. 4A, lanes 1  and 3), suggesting that AR was specifically bound to His 6 yCdc37. In similar experiments the GR did not bind to His 6 yCdc37 (Fig. 4B). The lack of stable association of Hsp90 with His 6 yCdc37 was somewhat surprising because we and others have observed some yeast Cdc37 to be coadsorbed with His-tagged Hsp90 (28) (18). As pointed out by Kimura et al. (28), however, yeast Cdc37 and Hsp90 proteins do not form the same stable association as do their mammalian homologs.
Association of Hsp90 with steroid hormone receptors is important for generating a high affinity ligand binding conformation. Hormone binding facilitates subsequent steps in the activation pathway, such as coactivator binding, dimerization, and association with chromatin. The change from inactive to active states is therefore dependent on the hormone itself, which stimulates Hsp90 dissociation from the AR (41) or prohibits further interaction. We found that His 6 Cdc37 interaction with the AR was also hormone sensitive, because His 6 Cdc37 binding was reduced after DHT was added to RRL prior to AR translation (Fig. 5). However, this inhibition was relatively weak for full-length receptors, where a reduction in complex formation by ϳ2-fold was commonly observed. In experiments with the isolated LBD, however, the presence of hormone almost completely inhibited complex formation. Similar results were obtained by incubating the hormone with reactions after trans-2 P. Lee and A. J. Caplan, unpublished results.  4) and a control strain not expressing His 6 yCdc37 (lanes 1 and 3) were incubated with Ni-NTA resin (lanes 3 and 4). Western blot analysis was used to assay for the presence of yeast Cdc37, AR, and Hsp90 as labeled (B). Expression and affinity chromatography of His 6 yCdc37 in strains expressing GR are shown. lation but prior to incubation with His 6 Cdc37. Thus, hormone binding appears to alter AR conformation in a manner that affects Cdc37 interaction but does not necessarily inhibit its association.
Cdc37 Does Not Substitute for Hsp90 Function in Hormone Binding by AR-Recent studies have shown that overexpressed Cdc37 can partially compensate for loss of Hsp90 function in protein kinase folding (28). We therefore tested for AR function in an hsp82 yeast mutant in the presence or absence of overexpressed yeast Cdc37. Previous studies have shown that the AR has a reduced affinity for hormone in an hsp82 mutant yeast strain (19). For these studies, AR was expressed in wild type and hsp82 G313N mutant strains. The G313N mutant was chosen because its growth rate was reported to be modestly increased by Cdc37 overexpression (28). Western blot analysis revealed that reduced amounts of AR were recovered in the mutant strain compared with the wild type (Fig. 6A). This reduction, although quite variable in different experiments, was largely suppressed by overexpression of Cdc37 (compare Fig. 6A, lanes 2 and 4). Results of hormone binding studies correlated with the reduced AR levels in this strain (Fig. 6B) and were ϳ10-fold lower in the mutant compared with the wild type. These levels did not change in the hsp82 mutant upon overexpression of Cdc37, although increased amounts of AR protein were recovered from the strain. It is possible that overexpressed Cdc37 can stabilize AR in this hsp82 mutant, although it apparently fails to facilitate a folding reaction in the absence of functional Hsp90. Thus, binding of Hsp90 and Cdc37 to the same domain of the AR has different functional consequences.
A Dominant Negative Form of Cdc37 Inhibits AR Activity in Mammalian Cells-Previous studies (27) showed that a truncated version of Cdc37, containing the N-terminal half of the protein, acted in a dominant negative manner for Raf activation. We prepared a similar construct (Cdc37 1-173 ) and coexpressed it in CV1 cells in the presence of full-length AR, the TAD-DBD form of AR and the GR. A reporter construct containing consensus hormone response elements for both AR and GR regulating the CAT gene was also coexpressed to assay receptor dependent gene expression. As shown in Fig. 7, coexpression of Cdc37 1-173 led to a 2-fold decrease in hormone-dependent reporter gene expression by AR. By contrast, the same mutant had no inhibitory effect on the ability of a TAD-DBD construct to stimulate hormone-independent gene expression and was only mildly inhibitory on the activity of GR. These results are analogous to those previously described from the yeast system (16) where mutation in CDC37 affected hormonedependent gene expression by full-length AR but had no effect on the ability of the TAD-DBD form to function as a constitutive activator. Overexpression of wild type Cdc37 had no effect on the ability of AR to function as a transcriptional activator (data not shown).

DISCUSSION
The paradigm for molecular chaperone function in steroid hormone receptor activation evolved from studies on GR and PR. In this paradigm, Hsp90 is recruited to the receptor ligandbinding domain via the actions of an Hsp70/Hsp40 chaperone pair, working in association with Hop, a co-chaperone that indirectly binds Hsp90 to Hsp70. Hsp70/Hsp40 and Hop are subsequently replaced on the receptor by Hsp90 in a complex with p23 and one of several immunophilins, such as cyp40 or FKBP52 (4) (5). This paradigm does not include Cdc37 because it appears not to interact with the GR or PR. On the other hand, there is strong experimental support for Cdc37 function, in association with Hsp90, in protein kinase folding; Cdc37 binds to protein kinases in vitro and prevents them from denaturing. Mutation in yeast Cdc37 affects the folding and activity of several different protein kinases, and expression of a dominant negative form of Cdc37 inhibits Raf activity in animal cells (27,42). We used similar criteria to judge whether Cdc37 was also important for activation of the human AR. Mutation in yeast CDC37, for example, inhibits hormone-dependent activation of AR (18). AR also binds to Cdc37 in vitro ( Figs. 1 and 2), and a dominant negative form of Cdc37 partially inhibits AR activity in animal cells (Fig. 6).
Although it is unknown what makes the AR require Cdc37 for activity, there are some instances where it behaves quite differently from other steroid receptors. One relates to the interaction of AR with coactivators and its AF2 transcriptional activity. AR has a very weak AF2 activity compared with GR, although both interact with similar coactivator proteins. Recent studies suggest that interaction of AR with coactivators such as SRC1 occurs in a relatively unique manner (43,44). Previous studies characterized coactivator binding to steroid receptors as being via the C-terminal ligand-binding domain in a hormone-dependent manner (45). The change in conformation that occurs upon ligand binding creates a site of interaction for coactivators that stimulate AF2 function. In the case of AR, coactivators such as SRC1 bind to the N-terminal domain instead and in a manner that requires interaction between the N terminus and the C terminus (43) (44). Such conformational changes between the AR N and C termini occur in a hormonedependent manner (46,47) and may provide a basis for the Cdc37 requirement. This would be consistent with a function for Cdc37 that is downstream of hormone binding (18). Furthermore, Cdc37 association with full-length AR but not the LBD persisted in the presence of hormone. This could be interpreted in terms of Cdc37 association with ligand-bound AR requiring an intact molecule containing both N-and C-terminal domains. We propose, therefore, that Cdc37 facilitates conformational changes in the AR that occur upon hormone binding, as the N terminus interacts with the C terminus.
The possibility that chaperones function downstream of ligand binding in steroid hormone receptor activation has never been fully explored. There is evidence for persistent Hsp70 association with ligand bound receptors (48), although the functional significance of this interaction remains unknown. A new study on the Drosophila Ecdysone receptor, however, showed that Hsp90 and several other co-chaperones facilitate heterodimerization with RXR and DNA binding (49). Chaperone involvement with Ecdysone receptor occurs only at this late stage in the activation process, and they are not required for hormone binding. Cdc37 may have an analogous function in AR activation. In this case, however, Hsp90/Hsp70/Hsp40 are required for generating and maintaining the high affinity hormone binding state (14,19), and Cdc37 would function at a  , lanes 1 and 2) or the just the LBD (lower panels, lanes 1 and 2) were incubated in the presence of DHT (ϩ, lanes 2 and 4) or ethanol (Ϫ, lanes 1 and 3). Binding reactions with His 6 Cdc37 were performed as described (lanes 3  and 4).
later step to facilitate ligand-dependent changes in conformation. It seems likely that Hsp90 plays a role in complex formation between the AR and Cdc37, because geldanamycin treatment compromised this interaction (Fig. 3A). This effect may be a direct consequence of the interaction of Hsp90 with Cdc37, providing a targeting function. Alternatively, Hsp90-dependent folding may be affected in geldanamycin-treated lysates, and the AR might fail to adopt the conformation that accepts Cdc37. It also appears likely, however, that Cdc37 can stabilize misfolded AR, at least when overexpressed in yeast cells. As shown in Fig. 6, overexpression of CDC37 in the hsp82 G313N mutant strain had a stabilizing effect AR protein levels, which were substantially reduced in the mutant alone. This effect was restricted to AR protein levels, however, and CDC37 did not suppress the folding defect as measured by the ability of AR to bind hormone. This means that whereas Hsp90 and Cdc37 have a functional relationship, there exists circumstances where they function independently of each other.
The interaction of Cdc37 with the AR has implications for some human diseases, including prostate cancer and benign prostatic hyperplasia. Prostate gland growth and differentiation depend on androgens, and antiandrogens or drugs that affect DHT synthesis are effective therapeutic agents against both cancer and benign prostatic hyperplasia. Recent studies also indicate that Cdc37 is up-regulated in prostate tumors and induces benign prostatic hyperplasia after overexpression in prostate glands of transgenic mice (50). The effect of Cdc37 in these diseases probably reflects its action on several different signaling pathways involving protein kinases that regulate cell cycle progression. Whether similar changes occur in AR signaling is unknown, although our results suggest that Cdc37 plays a role in AR activation in mammalian cells because a dominant negative form of Cdc37 partially down-regulated hormone-dependent activation of AR (Fig. 7). Although fairly weak, the 2-fold down-regulation of full-length AR by Cdc37 1-173 was not observed for the TAD-DBD truncated form of AR. Furthermore, although a similar mutant effectively inhibited activation of Raf (27), it was ineffective against another Cdc37-dependent kinase, Hck, suggesting that the action of this mutant is complex (51). Regardless of how well the mutant inhibited AR, however, the finding that Cdc37 1-173 can affect full-length AR but not TAD-DBD further supports the argument that Cdc37 functions via the AR-LBD, as indicated by the data shown in Fig. 2.
In summary, Cdc37 appears to function in AR activation in animal cells. Cdc37 interacts specifically with the AR but not the GR ligand-binding domain in a manner that is at least partially dependent on Hsp90. Our results suggest that Cdc37 is not an exclusively kinase-specific chaperone, although kinases may represent the majority of its clients. This is reminiscent of the proteins that interact with cytosolic chaperonins (Tric/CCT) of eukaryotes. Tric/CCT is required for actin and tubulin folding and was believed initially to be a chaperone specific for these proteins. Recent studies (52) now show that Tric/CCT interacts with many newly synthesized proteins shortly after translation. In this manner, some chaperones may have a broad specificity that is partially obscured if their client base is represented by protein families with a large or abundant membership, such as cytoskeletal elements or protein kinases.