The Cdc37 protein kinase–binding domain is sufficient for protein kinase activity and cell viability

The role of Hsp90 in protein folding appears to be restricted to signal-transducing proteins, represented largely by steroid hormone receptors and protein kinases (Caplan, 1999). Although protein kinases represent a large and abundant protein family, much more is known about Hsp90 mechanism of action in folding of steroid hormone receptors (Pratt and Toft, 1997). The overall scheme involves sequential action by Hsp70/Hsp40 chaperones followed by Hsp90 and several of its cochaperones. It is thought that Hsp70 acts first by interacting with nascent polypeptide chains in association with Hsp40s. Should a client polypeptide require Hsp90, then it must be recruited. This function is likely to be performed by Hop, a cochaperone that interacts with Hsp90 and Hsp70 via distinct tetratricopeptide repeat sequences (Chen et al., 1996; Scheufler et al., 2000). Hop also regulates Hsp90s' ATPase by stabilizing the ADP-bound form of yeast Hsp90 and by inhibiting client-stimulated ATP hydrolysis by the human homologue (Prodromou et al., 1999; McLaughlin et al., 2002). Recruitment of Hsp90 via Hop leads to the establishment of an “intermediate” complex containing the client polypeptide in association with Hsp70/Hop–Hsp90. Hsp70 and Hop subsequently dissociate from Hsp90 and are replaced by p23 and TPR-containing immunophilins (Smith, 1993). The change from intermediate to “mature” complexes is associated with a change in nucleotide bound to Hsp90, since p23 binds only to Hsp90-ATP (Johnson and Toft, 1995; Fang et al., 1998; Grenert et al., 1999). These interactions ultimately result in establishment of a hormone-binding competent conformation by steroid receptors. Investigators that reconstituted this pathway with purified proteins have proposed that Hsp70 and Hsp90 are the only essential chaperone components for folding of glucocorticoid receptor (GR)^* and that other cochaperones such as Hsp40 or Hop play supporting roles that stimulate the reaction (Morishima et al., 2000; Rajapandi et al., 2000). Similar studies with the closely related progesterone receptor, however, showed that Hsp40, Hop, and p23 are also necessary for folding and hormone binding (Kosano et al., 1998). Similarly, studies with a viral reverse transcriptase indicated that Hop and Hsp40, in addition to Hsp90 and Hsp70, are required for activity, whereas p23 is not essential but stimulates the reaction (Hu et al., 2002). The requirement of cochaperones for folding has also been demonstrated in yeast, since deletion of YDJ1 (a yeast Hsp40) or STI1 (yeast Hop) strongly affects the activity of exogenous Hsp90 clients such as the androgen receptor and v-Src (Caplan et al., 1995; Dey et al., 1996; Chang et al., 1997; Johnson and Craig, 2000). It is likely that individual clients will have their own conformational needs that manifest in the requirement for different cochaperones in a form of tailoring.

Folding of protein kinases is assumed to follow a similar pathway as for steroid receptors, with the exception that another chaperone called Cdc37 is also required. Cdc37 does not interact with most steroid hormone receptors, with the exception of the androgen receptor and a viral reverse transcriptase (Fliss et al., 1997; Rao et al., 2001; Wang et al., 2002b). Cdc37 is an Hsp90-binding protein that can also interact with unfolded or misfolded proteins (Kimura et al., 1997). It binds to Hsp90 at a site distinct from Hop or other TPR-containing cochaperones and may be present in mature complexes that contain p23 and immunophilins (Silverstein et al., 1998; Hartson et al., 2000). Overexpression of Cdc37 in transgenic mice results in hyperplasia and cancer when it is specifically targeted to breast and prostate tissues (Stepanova et al., 2000a,b), suggesting that Cdc37 expression must be tightly controlled. This is supported by the finding that it is normally expressed at very low levels (Kimura et al., 1997) and is not inducible under conditions that up-regulate other chaperones and heat shock proteins.

For the present study, we adopted a genetic strategy to understand the relationship between different chaperones and cochaperones in the folding of a protein kinase. We used v-Src as a test case because it is dependent on Hsp90 and its cochaperones for folding and activity. Our strategy was to block formation of the “intermediate” complex by analyzing yeast deleted for STI1, the gene that encodes Hop, which results in the loss of v-Src activity (Chang et al., 1997) and then search for suppressors. Our results show that Cdc37 is an effective suppressor of the v-Src folding defect in sti1Δ cells.

STI1 encodes an S. cerevisiae homologue of Hop, the hsp-organizing protein that coordinates the combined action of Hsp70 and Hsp90 on unfolded or misfolded proteins (Frydman and Hohfeld, 1997). Deletion of STI1 has little effect on cell growth under normal conditions, but results in decreased growth rate at high temperatures (Nicolet and Craig, 1989). Even under normal growth conditions, however, STI1 is important for folding of the oncogenic tyrosine kinase, v-Src, when heterologously expressed in yeast (Chang et al., 1997). To further investigate chaperone function, we searched for suppressors of v-Src loss of activity in yeast cells deleted for STI1 (sti1Δ) using a reverse genetic approach. Three genes were chosen for initial studies: CDC37, SBA1, and CNS1. CDC37 was chosen since it is essential for v-Src folding yet its action within the pathway of chaperone action is unknown. CNS1 was chosen since it has some similarity with STI1 (Dolinski et al., 1998; Marsh et al., 1998). SBA1 is the yeast homologue of mammalian p23, which is thought to function downstsream of Hop in steroid receptor activation and displayed genetic interactions with STI1 (Smith, 1993; Bohen, 1998; Fang et al., 1998). Each of these genes was overexpressed in sti1Δ cells and v-Src induced by growth in galactose-containing medium for 6 h. The short induction period was necessary since prolonged expression of v-Src is lethal to wild-type cells (Florio et al., 1994). Since yeast has very little endogenous tyrosine kinase activity, it is possible to measure v-Src activity with a monoclonal antiphosphotyrosine on Western blots using whole cell extracts. Induction of v-Src in wild-type cells led to the appearance of multiple bands that reacted with antiphosphotyrosine, as shown in Fig. 1. The intensity of these bands was reduced in sti1Δ cells at 25°C, and almost disappeared when the cells were grown at 30°C. The difference in v-Src activity when cells were grown at the two temperatures partly correlates with the levels of v-Src, which were markedly lower at 30°C. The decreased levels of v-Src in the sti1Δ strain did not reflect the levels of all protein kinases in yeast, since we observed no decrease in the levels of phosphoglycerate kinase (Fig. 1, A and B). Overexpression of SBA1 or CNS1 had no effect on v-Src levels or activity in sti1Δ cells (not shown). However, CDC37 overexpression resulted in increased v-Src activity at close to wild-type levels. The suppression phenotype was more pronounced when the cells were grown at 30°C compared with those grown at 25°C (Fig. 1). CDC37 overexpression also lead to increased levels of v-Src, especially at 30°C, perhaps by stabilizing the protein.

Our hypothesis regarding Cdc37 action in suppressing v-Src loss of function in sti1Δ yeast was based on the known or proposed functions of both Hop and Cdc37 as recruitment factors for Hsp90 (Stepanova et al., 1996; Frydman and Hohfeld, 1997). Both proteins bind to Hsp90 (at distinct sites; Silverstein et al., 1998), although only Cdc37 has been characterized to have chaperone activity. Thus, sti1Δ deletion may affect Hsp90 recruitment to v-Src, and CDC37 overexpression may compensate for this loss by acting to bypass the Hop requirement for Hsp90 recruitment. We therefore analyzed whether increasing Cdc37 dosage in sti1Δ cells stimulated v-Src binding to Hsp90 as a method of testing this hypothesis. The experiment was performed by inducing v-Src expression in yeast cells expressing GST-Hsp90 in the absence or presence of overexpressed Cdc37. Extracts from these cells were used for pull-down assays to determine whether or not v-Src was binding to Hsp90 (Fig. 1 C). In wild-type cells, v-Src binding to Hsp90 was observed using the pull-down assay and GST-tagged Hsp90 with or without Cdc37 overexpression. By contrast, there was no observable binding of v-Src to GST-Hsp90 in extracts from sti1Δ cells (Fig. 1 C), and Cdc37 overexpression failed to stimulate this association. These data support the hypothesis that Sti1 functions to recruit Hsp90 to client proteins, since this interaction was lost in the sti1Δ strain. On the other hand, it seems clear that Cdc37 does not substitute for Sti1 by promoting stable recruitment of Hsp90 to v-Src in the sti1Δ strain.

The data shown in Fig. 1 C may be explained by one of two possibilities; either Cdc37 functions in association with Hsp90 but fails to promote stable association of Hsp90–v-Src complexes, or it functions independently of Hsp90. To differentiate between these two possibilities, we devised an experimental strategy that uncoupled the ability of Cdc37 to interact with v-Src from its ability to interact with Hsp90. This strategy was based on prior studies showing that the NH₂-terminal region of human Cdc37 interacts with protein kinases, whereas the Hsp90-binding site resides toward the middle portion of the protein (Grammatikakis et al., 1999; Scholz et al., 2001). We confirmed this for interaction between human Cdc37 and v-Src. In the experiment shown in Fig. 2 B, we used purified full-length human Cdc37, or a COOH-terminal truncation (Cdc37^1–173) that was deleted for the putative Hsp90-binding site. Incubation of these recombinant proteins in rabbit reticulocyte lysates resulted in binding of Hsp90 to the full-length Cdc37 only, whereas in vitro synthesized ³⁵S-labeled v-Src bound to both proteins. Based on these results, we generated a series of truncation mutants of yeast Cdc37 that spanned the Hsp90-binding site. Each contained an NH₂-terminal His₆ tag for subsequent detection by Western blot analysis. Of the eight mutants that were generated, three failed to express in yeast, whereas the rest were soluble (Fig. 2 D). We performed several experiments to assay for interaction between His-tagged Cdc37 or the truncation mutants with Hsp90 in cell extracts, although we failed to detect any physical interactions. This is consistent with studies showing that interaction between yeast Cdc37 and Hsp90 is very weak (Kimura et al., 1997; Siligardi et al., 2002).

We expressed each of the cdc37 truncation mutants in sti1Δ yeast and tested whether any of them could suppress the v-Src activity defect. These studies were performed at both 25°C and 30°C. Remarkably, the Cdc37^1–388 mutant was as proficient as full-length Cdc37 for suppressing the sti1Δ phenotype, and its expression restored v-Src activity at both temperatures. None of the other truncations was as effective in this regard, although even the Cdc37^1–148 mutant was capable of stabilizing v-Src protein when induced at 30°C in the sti1Δ background and restoring a small amount of v-Src activity (Fig. 3 B). None of the truncation mutants smaller than Cdc37^1–388 was very effective at suppressing the v-Src activity defect in sti1Δ at 25°C. This may reflect the relative stability of v-Src at this lower temperature even in the absence of Sti1 (Fig. 3, compare lane 2 of A with that of B). The modest suppression phenotype observed at 30°C by the two smaller truncations (Cdc37^1–235 and Cdc37^1–148) may therefore relate to their ability to stabilize v-Src levels. Since these smaller truncations were deleted for the putative Hsp90-binding site, their function in activating v-Src may be Hsp90 independent. However, there is a sharp drop in v-Src activity in cells expressing truncated forms of Cdc37 smaller than Cdc37^1–388, which correlates with loss of the Hsp90-binding site (Fig. 3 B, compare lanes 3–7). These combined data therefore suggest that Cdc37 cooperates with Hsp90 when it is able to do so, although it can function independently in a less efficient manner when it has to.

We performed further studies to address the relationship between Cdc37 and Hsp90 in v-Src activity. Our approach was to test whether full-length or truncated versions of Cdc37 could suppress v-Src activity defects arising from decreased levels of Hsp90. The rationale was to differentiate between suppression due to Hsp90 interaction with Cdc37 versus Cdc37 being able to function by itself. Previous studies showed that deletion of the yeast HSC82 gene caused a substantial reduction in the cellular concentration of Hsp90 (by >90%) (Borkovich et al., 1989; Xu and Lindquist, 1993), with just enough expression from a second gene (HSP82) to maintain cell viability. Previous studies also demonstrated that v-Src was inactive in an hsc82Δ strain (Xu and Lindquist, 1993), and that Cdc37 overexpression partially suppressed v-Src loss of activity in a conditional hsp82 mutant (Kimura et al., 1997). We first confirmed that v-Src activity was decreased in the hsc82 mutant (Fig. 4), and then assayed for the ability of each Cdc37 truncation mutant to suppress the v-Src activity defect. As shown in Fig. 4 A, the results of this experiment were very similar to those observed for the sti1Δ mutant, although overall levels of suppression by full-length Cdc37 were reduced compared with the sti1Δ strain (Fig. 3). The similarity between the results using the hsc82Δ and sti1Δ strains extended the findings that Cdc37^1–388 was as effective as full-length Cdc37, whereas the smaller Cdc37^1–335, Cdc37^1–239, and Cdc37^1–148 proteins were effective, but to a lesser degree. Also, each truncation mutant stabilized v-Src to a similar extent independently of its size. Since the hsc82Δ strain was limiting for Hsp90 to begin with, and the smaller truncations were deleted for the putative Hsp90-binding site, we conclude that Hsp90 is not essential for Cdc37 function in protein kinase folding, but that both chaperones cooperate to optimize the reaction. One further possibility that was addressed is whether Cdc37 stimulated intrinsic Hsp90 activity. This was tested by analyzing the effect of Cdc37 on the activity of a protein that depends on Hsp90 but is not a client of Cdc37. The GR represents such a client, and its activity was analyzed in wild-type and sti1Δ cells in which its activity was known to be reduced (Chang et al., 1997). GR activity was measured as deoxycorticosterone induced β-galactosidase activity (from a reporter plasmid under control of hormone responsive elements). In sti1Δ cells, the basal levels of β-galactosidase was increased relative to that found in wild-type cells, but the inducible levels were reduced (Fig. 4 B). The hormone-dependent induction of β-galactosidase was therefore greatly reduced in sti1Δ cells compared with the wild-type from 30-fold induction to approximately sixfold induction. Importantly, Cdc37 overexpression had very little effect on GR activity in either strain, suggesting that Cdc37 overexpression does not affect intrinsic Hsp90 activity.

These findings led us to test whether the protein kinase–binding domain of Cdc37 was sufficient for yeast cell viability. Yeast strains expressing truncated forms of Cdc37 constructed in a cdc37Δ deletion background isolated from sporulated diploids. The growth phenotype of each haploid was scored after tetrad dissection. The results from this study demonstrated that all the truncated forms of Cdc37 could support yeast cell growth with the exception of Cdc37^1–335. Cells expressing the two smallest truncations, Cdc37^1–239 and Cdc37^1–148, were slow growing and displayed a temperature sensitive lethal phenotype at 37°C (Fig. 5 A). Further studies addressed whether the viability of cells expressing the truncations was dose dependent. We prepared plasmid constructs that expressed full length, Cdc37^1–388, Cdc37^1–335, Cdc37^1–239, and Cdc37^1–148 from the same ADH promoter on low copy number plasmids. This led to a significant decrease in the amount of expressed protein (Fig. 5 B). However, the amount of full-length Cdc37 expressed from the ADH promoter from a low copy number plasmid was still higher than wild-type chromosomal Cdc37 expressed from its own promoter (not shown). Using a plasmid swap procedure with 5-fluoroorortic acid, we then determined whether these truncations in low copy number could support growth of cdc37Δ yeast cells. In low copy, only full-length Cdc37 and Cdc37^1–388 suppressed the lethality of cdc37Δ. In high copy, further suppression was observed by plasmids expressing Cdc37^1–239 and Cdc37^1–148, as expected based on the tetrad dissection analysis. The only truncation that failed to support growth in either high or low copy was Cdc37^1–335 (not shown). These data demonstrate that the protein kinase–binding domain of Cdc37 is sufficient for cell viability if present in a high enough concentration. We also tested whether these mutants could suppress the temperature-sensitive lethal phenotype of cdc37–34, which has a point mutation (serine 14 to leucine; Fliss et al., 1997). As expected, the small truncations failed to do so since they failed to promote growth of cdc37Δ at 37°C, although expression of Cdc37^1–388 led to viable cells (Fig. 5 D). Interestingly, Cdc37^1–335 expression had a negative effect on cdc37–34 growth at 25°C. These combined data indicate that the COOH-terminal 118 amino acids of Cdc37 are dispensable for normal function and that Cdc37^1–335 has abnormal function that is partially dominant–negative in cdc37–34.

The ability of Cdc37^1–148 to promote v-Src folding in sti1Δ and hsc82Δ strains supports the hypothesis that Cdc37 is sufficient for v-Src folding when Hsp90 is limiting or cannot be recruited to the client. We further tested how well Cdc37^1–148 could substitute for full-length Cdc37 in cells that were wild type for Hsp90 and Sti1. To this end, we induced v-Src in cdc37Δ yeast cells that expressed Cdc37^1–148 from a multicopy plasmid. As shown in Fig. 6 A, the activity of v-Src was much lower in the presence of Cdc37^1–148 compared with full length Cdc37, similar to the results observed in hsp82Δ and sti1Δ strains. However, a major difference was that the levels of v-Src protein were also markedly lower in the cdc37Δ background than in hsp82Δ and sti1Δ cells. We extended this analysis to an endogenous protein kinase pathway. The system used was signaling from the α-factor mating pheromone via a G-protein–coupled receptor and the MAP kinase pathway. Previous studies showed that one of the kinases in this pathway, Ste11, interacts with Cdc37 and Hsp90, and that signaling was impaired in cdc37 and hsp82 mutant strains (Abbas-Terki et al., 2000). Using β-galactosidase as a reporter for this pathway, we tested how well Cdc37^1–148 could substitute for full-length Cdc37. As shown in Fig. 6 B, there was very little signaling in the cdc37–34 mutant, consistent with a previous report (Abbas-Terki et al., 2000). However, the Cdc37^1–148 truncation facilitated signal transmission at a level that was ∼70% of the wild-type value. These data suggest that folding is optimal when Cdc37 and Hsp90 interact with each other, but is also quite strong in the presence of the isolated protein kinase–binding domain of Cdc37.

The role of Cdc37 in protein kinase folding has become firmly established over the past few years. Cdc37 interacts with many different protein kinases and facilitates their folding in association with Hsp90. Although the exact relationship between these two chaperones is unknown, it has been suggested that Cdc37 targets Hsp90 to misfolded protein kinases (Stepanova et al., 1996). In this capacity, Cdc37 would bear some homology with Hop, which also recruits Hsp90 to misfolded polypeptides bound to Hsp70 (Frydman, 2001). In this report we show that CDC37 overexpression suppressed the defect in v-Src activity resulting from deletion of STI1, the yeast Hop homologue.

The defect in v-Src activity in sti1Δ was temperature dependent, with the defect being greater when the cells were grown at 30°C compared with 25°C. This is correlated with decreased levels of v-Src when the cells were grown at 30°C. We interpret these findings in terms of v-Src being less stable at the higher temperature combined with its inability to undergo refolding in the sti1Δ mutant. This defect probably arises due to lack of Hsp90 recruitment to v-Src (Fig. 1 C). Although overexpression of CDC37 suppressed the loss of v-Src activity, it does so in a manner that is independent of any stable recruitment of v-Src to Hsp90; although this finding alone does not rule out the possibility that full-length Cdc37 acts via Hsp90 in the suppression process. Since Cdc37 and Hsp90 are only weakly interacting in yeast, it is possible that Cdc37 interacts with both v-Src and Hsp90 in sti1Δ, but in a complex that rapidly dissociates or is sensitive to our extraction procedure. Our model is that Cdc37 acts downstream of Sti1 in the folding pathway, and that Sti1 normally functions to recruit Hsp90 (and perhaps Cdc37) (Abbas-Terki et al., 2002) to misfolded clients already bound to Hsp70 (Fig. 7). The bypass function of overexpressed CDC37 probably reflects the ability of Cdc37 protein to interact with both Hsp90 and v-Src. Since it is normally expressed at low levels, Cdc37 probably represents a limiting factor for protein kinase folding that must be recruited to the client via interaction with Hsp90 and Sti1. However, our results are also consistent with Cdc37 being able to function in an Hsp90-independent manner. This is because Cdc37 truncations lacking the Hsp90-binding site were still functional for v-Src folding in sti1Δ and hsc82Δ cells, albeit at reduced efficiency. The isolated protein kinase–binding domain of Cdc37 was also sufficient for cell viability and signaling via the MAP kinase–signaling pathway (Figs. 5 and 6). However, whether Cdc37 can function in a completely Hsp90-independent manner remains unclear, since is still possible that both chaperones could function together even though they fail to interact with each other. Supporting this notion is the intriguing finding that a truncated version of Cdc37 that failed to interact directly with Hsp90 could still stimulate Hsp90 recruitment to a protein kinase in animal cells, although this was far less than the effect of full-length Cdc37 (Scholz et al., 2000). Thus, it is still possible that truncated versions of Cdc37 could permit association of a protein kinase with Hsp90, perhaps via allosteric effects on the client itself. However, since Cdc37^1–148 facilitated some v-Src activity under conditions where Hsp90 could not be recruited to the client (Fig. 3), or was already present in reduced amounts (Fig. 4), we propose that the protein kinase–binding domain of Cdc37 is sufficient for a limited amount of v-Src folding independently of Hsp90 function. We also conclude that the COOH-terminal 118 amino acids of Cdc37 are dispensable for function since Cdc37^1–388 behaved just like full-length Cdc37 for v-Src folding and cell viability (Figs. 3, 4, and 5). This is consistent with a previous finding showing that cells expressing Cdc37 truncated to amino acid 360 were viable but temperature-sensitive for growth (Gerber et al., 1995).

Although Cdc37 overexpression suppressed loss of function by Hop and Hsp90, we have thus far been unable to suppress defects in v-Src folding in YDJ1 mutant strains (Dey et al., 1996) in the same manner. YDJ1 encodes an Hsp40 protein that functions in association with Hsp70 probably at the earliest stages of protein folding (Cheetham and Caplan, 1998; Frydman, 2001). We have recently shown, however, that SSE1 overexpression suppressed loss of v-Src activity in a ydj1 mutant strain (Goeckeler et al., 2002). SSE1, a member of the Hsp110 family, is similar in structure to Hsp70 and also interacts with Hsp90 (Liu et al., 1999). It is therefore possible that SSE1 suppressed defects in the ydj1 mutant because it has homologous action to Hsp70. This suggests that Hsp70 function is an essential step in v-Src folding that cannot be bypassed by Cdc37, unlike the actions of Sti1 or Hsp90. The role of other cochaperones such p23 (encoded by SBA1) is more speculative, since its deletion does not strongly affect v-Src folding in yeast (Fang et al., 1998), although it is possible that Cdc37 and p23 form a complex with Hsp90 in animal cells (Hartson et al., 2000).

The ability of Cdc37 to bypass Hsp90 chaperone components may occur only under certain conditions such as its overexpression, which occurs in cancer cells (Stepanova et al., 2000b). Previous studies have shown that Cdc37 is oncogenic when overexpressed in mice, and its levels are greatly increased in human prostate cancer tissues (Stepanova et al., 2000a,b). These data suggest that Cdc37 levels must be tightly regulated. We propose that this reflects a form of regulation whereby Cdc37 is recruited only to those protein kinases that have been through a triage system involving other chaperones, such as Hsp70/Hsp40 and Hsp90. In this manner, the cell maintains tight control over exposure of client kinases to Cdc37, and prevents inappropriate stabilization or activation. Indeed, recent studies demonstrated that Cdk4 induced destabilization by C/EBPα correlates with dissociation of Cdc37/Hsp90 from Cdk4, suggesting that chaperone-dependent stabilization plays an important role in cell cycle progression (Wang et al., 2002a). It is possible that Cdc37 overexpression could override such controls and contribute to the transformation process.

The W3031b (MATα) strain background was used for the sti1Δ and hsc82Δ studies. The cdc37Δ mutants were derived from strain BY4743 as part of S. cerevisiae gene deletion project (Winzeler et al., 1999). Haploids containing the cdc37Δ gene were isolated from tetrads. Yeast cells were grown in selective media (0.67% yeast nitrogen base, 2% glucose supplemented with amino acids to complement auxotrophic markers). Yeast transformation was performed by the lithium acetate method as described previously (Gietz et al., 1995). Plasmid swap experiments were performed by plating cells onto media containing 5-fluoroorotic acid to counterselect for plasmids encoding URA3 as described previously (Sikorski and Boeke, 1991).

v-Src was induced from the Gal1–10 promoter by growth in media containing 2% galactose for 6 h. Cells were grown in 2% raffinose containing media before induction. Extracts were prepared by glass bead lysis in 20 mM HEPES, pH 7.5, 100 mM KCl, 0.1 mM EDTA plus protease inhibitors contained in the complete protease inhibitor cocktail by Boehringer. Cells were lysed with 0.4-mm glass beads using 3× 1-min pulses in a bead beater with 1-min rests in between at 4°C. Extracts were cleared at 14,000 g for 10 min. Lysates were resolved by denaturing gel electrophoresis and transferred to Immobilon-P membranes (Millipore). Membranes were blocked with 5% nonfat dry milk in TTBS (20 mM Tris HCl, pH 7.5, 0.5 M NaCl, 0.05% Tween-20) overnight. Levels of tyrosine kinase activity were determined using 4G-10 monoclonal antibodies (Upstate Biotechnology) in antibody dilution buffer (1× phosphate buffered saline, 3% bovine serum albumin, 0.05% Tween-20, 0.1% thimersal) for 2 h followed by 3 × 10 min washes in TTBS and incubation in horseradish peroxidase–conjugated goat anti–mouse at 1:10,000 dilution for 1 h. The filters were washed 3 × 5 min in TTBS and treated with a chemiluminescence reagent (Pierce Chemical Co.) and exposure to x-ray film. V-Src protein was detected using EC10 monoclonal antibodies (a gift from Dr. I. Gelman, Mount Sinai School of Medicine, New York, NY) using the same conditions.

Wild-type and sti1Δ yeast cells expressing GST-Hsp90 (from p2U/HSP82-GST; a gift from D. Picard, Universite de Geneve, Geneve, Switzerland) were grown in 200-ml 0.67% yeast nitrogen base plus 2% raffinose to A₆₀₀ = 0.2. Galactose was added to 2% to induce v-Src expression for an additional6 h. The cells were harvested, and cell extracts were prepared as described above, but using 20 mM TrisHCl, pH 8, 200 mM NaCl, 1 mM EDTA, pH 8, 0.5% NP-40, 25 μg/ml PMSF plus protease inhibitors contained in the complete protease inhibitor cocktail by Boehringer. The extracts were adjusted to 3 mg/ml in the same buffer and incubated with 50 μl glutathione-agarose resin (ICN Biomedicals) for 2 h at 4°C. The resin was collected by low speed centrifugation for 2 min, followed by four washes in the same buffer. Proteins retained by the resin were dissolved in sample buffer, resolved by gel electrophoresis, and analyzed by Western blot.

The plasmid-encoding poly-histidine (His₆) tagged yeast Cdc37 under control of the yeast ADH1 promoter, pPL2 (TRP1, 2 μm), was described previously (Rao et al., 2001). To construct a URA3 version of this plasmid, the insert and promoter region were excised with Kpn1 and Sac1, and ligated into pRS426 (Christianson et al., 1992) to generate pPL3. Truncated forms of His₆Cdc37 were generated by one of two different methods. Cdc37^1–388, Cdc37^1–274, Cdc37^1–239, Cdc37^1–207, and Cdc37^1–148 were generated by digesting pPL3 with Xb1/Kpn1, Stu1/Kpn1, EcoR1/Kpn1, Pst1/Kpn1, and Cla1/Kpn1, respectively. The digested plasmids were gel-purified and converted to blunt ends with DNA polymerase I klenow fragment. They were subsequently ligated with T4 DNA ligase and transformed into E. coli XL1-Blue. Cdc37^1–335 and Cdc37^1–282 were generated by PCR amplification of Cdc37 from pRSS2 (Fliss et al., 1997). The upstream primer 5′-AAGCTTATGCACCACCACCACCACCACGCCATTGATTACTCTAAG-3′ was used for both constructs. The downstream primer for Cdc37^1–335 was 5′-GGTACCCTACTTGTCCTTGTTGAAGAT-3′, and for Cdc37^1–282 5′-GGTACCCTAACCATGTAACTGATACTC-3′. PCR products were generated with Pfu1 polymerase in 20 cycles using 50°C annealing temperature. The products were gel-purified and subcloned first into pCR-Script and then using HindIII and Kpn1 into pRS426. Plasmids encoding Cdc37^1–388, Cdc37^1–335, Cdc37^1–239, and Cdc37^1–148 were subsequently subcloned into pRS423 (His3, 2 μm) and pRS313 (His3, CEN/ARS) after PvuII digestion. The plasmid encoding v-Src used in these studies was constructed by digestion of pRS316-vSrc (a gift from Dr. Frank Boschelli, Wyeth-Ayerst, Pearl River, NY) with Not1/Xho1, and ligation into pRS425 (Leu2, 2 μm). The plasmid encoding full-length His₆ human Cdc37 for expression in E. coli (pET15b.Cdc37) was described previously (Rao et al., 2001). The truncated His₆ human Cdc37^1–173 was prepared by PCR amplification from pET15b.Cdc37 using the primers 5′-CATATGGTGGACTACAGCGTG-3′ and 5′-GGATCCTCACTTTTGGCTGTCATCCCAGCG-3′ using Pfu1 in a 20-cycle reaction at 55°C annealing temperature. The product was gel-purified and sub-cloned into pCR-Script. The resulting plasmid was digested with Nde1 and BamH1 and the fragment ligated into pET15b.

The expression and purification of His₆Cdc37 and His₆Cdc37^1–173 and the binding reactions with v-Src translated in rabbit reticulocyte lysates were performed exactly as described previously (Rao et al., 2001).

Wild-type and cdc37Δ/Cdc37^1–148 yeast cells were transformed by pPRE-LacZ (a gift from K. Morano, University of Texas Medical School, Houston, TX) that expresses β-galactosidase under control of Ste12 response elements. These cells were grown to mid-log growth phase at 25°C and incubated with 5 μM α-factor (Sigma-Aldrich) for 3 h. Extracts were prepared as above in 20 mM HEPES, pH 7.5, 100 mM KCl, 0.1 mM EDTA plus protease inhibitors contained in the complete protease inhibitor cocktail by Boehringer, and β-galactosidase activity was measured exactly as described previously (Caplan et al., 1995). GR activity was assayed by a similar assay in the absence or presence of 1 μM deoxycorticosterone (Fliss et al., 1997) using the Tropix kit.

We thank Drs. E. Craig, D. Gross, K. Morano, and D. Picard for their gifts of yeast strains and plasmids, and I. Gelman for a gift from EC10 monoclonal antibodies.

A.J. Caplan was supported by an Irma T. Hirschl Career Scientist Award and National Institutes of Health (NIH) grant DK60598. P. Lee was supported by a predoctoral training grant NIH 5T32DK07645. S. Garrett was supported by grant NIH GM66644.