p50 cdc37 Binds Directly to the Catalytic Domain of Raf as Well as to a Site on hsp90 That Is Topologically Adjacent to the Tetratricopeptide Repeat Binding Site*

Several protein kinases (e.g.pp60 src , v-Raf) exist in heterocomplexes with hsp90 and a 50-kDa protein that is the mammalian homolog of the yeast cell cycle control protein Cdc37. In contrast, unliganded steroid receptors exist in heterocomplexes with hsp90 and a tetratricopeptide repeat (TPR) domain protein, such as an immunophilin. Although p50 cdc37 and TPR domain proteins bind directly to hsp90, p50 cdc37 is not present in native steroid receptor·hsp90 heterocomplexes. To obtain some insight as to how v-Raf selects predominantly hsp90·p50 cdc37 heterocomplexes, rather than hsp90·TPR protein heterocomplexes, we have examined the binding of p50 cdc37 to hsp90 and to Raf. We show that p50 cdc37 exists in separate hsp90 heterocomplexes from the TPR domain proteins and that intact TPR proteins compete for p50 cdc37 binding to hsp90 but a protein fragment containing a TPR domain does not. This suggests that the binding site for p50 cdc37 lies topologically adjacent to the TPR acceptor site on the surface of hsp90. Also, we show that p50 cdc37 binds directly to v-Raf, with the catalytic domain of Raf being sufficient. We propose that the combination of exclusive binding of p50 cdc37 versus a TPR domain protein to hsp90 plus direct binding of p50 cdc37 to Raf allows the protein kinase to select for the dominant hsp90·p50 cdc37 composition that is observed with a variety of protein kinase heterocomplexes immunoadsorbed from cytosols.

A variety of transcription factors and protein kinases have been recovered from cytosols in native heterocomplexes with the abundant, ubiquitous, and essential protein chaperone hsp90 1 (for review, see Refs. 1 and 2). Several other proteins, all of unknown function, have been recovered in steroid receptor⅐hsp90 and protein kinase⅐hsp90 heterocomplexes. Steroid receptor⅐hsp90 heterocomplexes contain one of several high molecular weight immunophilins or the protein serine/ threonine phosphatase PP5 (1). The protein kinase heterocomplexes contain a 50-kDa phosphoprotein that was originally identified as a component of the pp60 v-src ⅐hsp90 heterocomplex (for review, see Refs. 3 and 4).
We and others have recently cloned p50 and identified it as the vertebrate homolog of the yeast cell cycle control protein Cdc37 (5-7). 2 Genetic evidence suggests that Cdc37 is necessary for Src function (8) and for signaling via the sevenless receptor, a protein tyrosine kinase of Drosophila (9). The cyclindependent protein kinase Cdk4 is also recovered in heterocomplexes with hsp90 and p50 cdc37 (6,10), and we (10) and Stepanova et al. (6) have shown that p50 cdc37 binds directly to Cdk4 as well as to hsp90.
Although native receptor⅐hsp90 heterocomplexes contain one of the TPR domain proteins, they do not contain p50 cdc37 (30,31). In contrast, immune-isolated Src⅐hsp90 (3) and Cdk4⅐hsp90 (6) heterocomplexes contain p50 cdc37 , but no TPR protein has been identified. We have shown that v-Raf, a serine/threonine kinase involved in signal transduction, also exists in heterocomplexes with hsp90 and p50 cdc37 (31). Although v-Raf immune pellets have the ability to bind a small amount of [ 3 H]FK506 in a Raf⅐hsp90-specific manner (32), it seems clear that the majority of v-Raf⅐hsp90 heterocomplexes contain p50 cdc37 .
It is not known how the protein that is being chaperoned by hsp90 (i.e. steroid receptor or protein kinase) determines the composition of the heterocomplex. In this report, we provide evidence that p50 cdc37 binds to hsp90 at a site on its surface that is near the binding site for the TPR domain proteins.
Using FLAG-tagged p50 cdc37 and PP5, we show that p50 cdc37 exists in separate hsp90 heterocomplexes from the TPR proteins. In addition to binding to hsp90, p50 cdc37 binds directly to Raf. It is known that, during the process of Raf⅐hsp90 heterocomplex assembly, Raf is transiently associated with p60 (also called Hop) (33), which binds to hsp90 via its TPRs (34). p60/ Hop is required for assembly of hsp90 heterocomplexes (35), and we show here that p60/Hop competes for the binding of both TPR domain proteins and p50 cdc37 to hsp90. Our observations are consistent with a model in which dissociation of p60/ Hop from the newly formed Raf⅐hsp90 complex results in an open region on the surface of the hsp90 dimer that can be occupied by either p50 cdc37 or a TPR protein. With continued exchange binding of p50 cdc37 and TPR domain proteins to Rafassociated hsp90, Raf⅐hsp90⅐p50 cdc37 complexes are rapidly selected because p50 cdc37 also binds directly to Raf.

Materials
Untreated rabbit reticulocyte lysate was from Green Hectares (Oregon, WI). 125 I-Conjugated goat anti-mouse and anti-rabbit IgGs were from NEN Life Science Products. Goat anti-mouse IgG-horseradish peroxidase conjugate, monoclonal nonimmune IgG and IgM, purified rabbit IgG, monoclonal anti-glutathione S-transferase (GST) clone GST-2 ascites, and purified glutathione S-transferase were from Sigma. The AC88 monoclonal IgG against hsp90 was from StressGen (Victoria, British Columbia, Canada). The 3G3 monoclonal IgM against hsp90, and the anti-cyclophilin 40 (COOH-terminal peptide) antibody were from Affinity Bioreagents (Golden, CO). The anti-FLAG M2 monoclonal IgG, M2-agarose, and the FLAG peptide were from IBI (New Haven, CT). The C-12 rabbit anti-Raf-1 IgG was from Santa Cruz Biotechnology (Santa Cruz, CA). The anti-Raf antiserum prepared against the carboxyl-terminal 12 amino acids of human Raf-1 (34) was kindly provided by Dr. Richard Jove (Moffitt Cancer Center, Tampa, FL). The DS14F5 monoclonal antibody against p60/Hop (36) and Escherichia coli expressing human p60/Hop were kindly provided by Dr. David Smith (University of Nebraska, Omaha, NE). The XR recombinant pGEX-2T plasmid encoding GST-tagged rabbit FKBP52 (37) was kindly provided by Dr. Jack-Michel Renoir (University of Paris, France). The UPJ56 rabbit antiserum against hsp56 (38) was a kind gift from Dr. Karen Leach (The Upjohn Co., Kalamazoo, MI). The rabbit antiserum against hsp70 and hsp90 (39) was generously provided by Dr. Ettore Appella (National Cancer Institute). Rabit antiserum to PP5, purified FLAG-PP5, and the FLAG-tagged TPR domain of rat PP5 were prepared as described previously (26).

Methods
Cell Culture and Cytosol Preparation-Sf9 cells and 3Y1 rat fibroblasts stably transfected with DNA encoding v-Raf (31) were harvested, washed once, suspended in 1 volume of HE buffer (10 mM Hepes, pH 7.4, 1 mM EDTA), and ruptured by Dounce homogenization. Homogenates were centrifuged 15 min at 12,000 ϫ g.
Immunoadsorption-Native hsp90 heterocomplexes were immunoadsorbed from 150 l of rabbit reticulocyte lysate for 2 h at 4°C with 15 l of 3G3 antibody prebound to 12 l of protein A-Sepharose, as described previously (24). Native p60/Hop heterocomplexes were immunoadsorbed from 150 l of rabbit reticulocyte lysate with DS14F5 antibody against p60 (3%), and FLAG-PP5 or FLAG-p50 was immunoadsorbed with 6 g of M2 monoclonal antibody against the FLAG epitope. All immunopellets were washed three times by suspension in 1 ml of TEGM buffer (10 mM TES, 50 mM NaCl, 4 mM EDTA, 10% (w/v) glycerol, 20 mM sodium molybdate, pH 7.6), and proteins were resolved by SDS-polyacrylamide gel electrophoresis.
Western Blotting-Immunoblots were probed with 1 g/ml AC88 for hsp90 (or, in the case of insect hsp90, with 0.1% hsp70/hsp90 antiserum), 0.1% UPJ56 for hsp56, 0.1% PP5 antiserum for PP5, 1 g/ml M2 monoclonal for the FLAG-proteins, 0.1% DS14F5 for p60/Hop, 0.1% p50 antiserum for p50 cdc37 , 0.1% anti-Raf antiserum for v-Raf, 0.1% GST ascites for GST-Raf, or 0.1% anti-cyclophilin 40 for CyP-40. The immunoblots were developed with the appropriate horseradish peroxidaseconjugated and/or 125 I-conjugated counter antibody. Although immunoblots from individual immunoadsorption or competition binding experiments are presented, the experiments have been performed at least three times and corroborating results obtained by immunoadsorp-tion of, or competition by, other proteins are usually presented in other panels of the same figure.
Binding of Proteins to Purified hsp90 -Rabbit hsp90 was purified from brain cytosol as described by Hutchison et al. (40). Aliquots (30 l) of purified rabbit hsp90 (1 mg/ml) were immunoadsorbed to 12-l pellets of protein A-Sepharose precoupled with 15 l of 3G3 antibody. Pellets were washed twice with 1 ml of HE buffer and suspended in Hepes buffer, pH 7.4, plus 0.1% Nonidet P-40 in a final volume of 100 l, including 30 l of the pooled, hsp90-free hydroxylapatite fraction of rabbit brain cytosol containing p60/Hop, PP5, FKBP52, p50 cdc37 , and CyP-40 prepared exactly as described by Owens-Grillo et al. (29). In experiments where binding of proteins to hsp90 was competed with the PP5 TPR domain, 30 g of purified FLAG-tagged PP5 TPR in 30 l of 20 mM Hepes, 1 mM dithiothreitol, 150 mM NaCl were added, maintaining the same final incubation volume of 100 l. In experiments where binding of proteins to hsp90 was competed with bacterially expressed p60/Hop, Sf9-expressed FLAG-PP5 or FLAG-p50 cdc37 , bacterial lysate, or Sf9 cytosol was preincubated with the immunopellets in a final volume of 30 l on ice for 20 min with suspension of the pellets by shaking the tubes every 3 min. The hydroxylapatite pool was then added and reaction mixtures were brought up to a final volume of 100 l, and incubations were maintained on ice for 35 min with suspension of the pellets by shaking the tubes every 3 min. At the end of the incubation, the pellets were washed three times with 1 ml of HEG buffer (10 mM Hepes, pH 7.4, 1 mM EDTA, 10% glycerol), and proteins were resolved by SDS-PAGE and Western blotting.
Expression of p60 and GST-FKBP52 Fusion Protein-Bacterially expressed p60/Hop was prepared as described previously (35). For bacterial lysates containing GST-FKBP52, the expression plasmid containing the cDNA for the 59-kDa rabbit immunophilin subcloned into the SmaI site of pGEX-2T prepared by Le Bihan, et al. (37) was used to transform E. coli strain BL21(DE3). Purification of rabbit FKBP52 was performed by binding the GST-FKBP52 to GSH-agarose and incubation at 4°C with thrombin, which cleaves at a site between the GST domain and the FKBP52 domain.
Production of the Fusion Protein GST-Raf (COOH Terminus)-For bacterial expression of GST-Raf (COOH terminus), an in-frame deletion of amino acids 26 -309 of human c-Raf-1 following digestion with PvuII and BglI (41), was subcloned into the pGEX-2T bacterial expression vector and in-frame with the GST propeptide to generate pGEX⌬NRaf. The resulting construct was transformed into E. coli BL21(DE3). A control construct including GST in fusion with the first 25 amino acids of human c-Raf-1 behaved similarly as GST alone, in that it bound neither to p50 cdc37 nor to hsp90 (data not shown).
Binding of Purified FLAG-p50 cdc37 to Raf and GST-Raf (COOH Terminus)-Control E. coli and bacteria expressing the GST-tagged Raf (COOH terminus) were sonicated in phosphate-buffered saline, and 50 l of lysate were immobilized on 15 l of glutathione-cross-linked agarose. v-Raf was immunoabsorbed from 3Y1 cytosol (200 l) by rotation with the C-12 rabbit anti-Raf-1 IgG prebound to 8 l of protein A-Sepharose. The immune pellets were washed two times with 1 ml of TEG plus 0.1% Triton X-100, then two times with TEG (for native Raf heterocomplexes, 20 mM molybdate was present in the wash buffers). The pellets were then suspended in TEG buffer containing 0.5 M NaCl and stripped of Raf-associated hsp90 by heating for 1 h at 30°C followed by two buffer washes prior to incubation with 30 l of cytosol from Sf9 cells expressing FLAG-p50 cdc37 , 40 l of purified FLAG-p50 cdc37 , or 45 l of purified bacterially expressed rabbit FKBP52. Incubations were on ice for 35 min with suspension of the pellets by shaking the tubes every 3 min. At the end of the incubation, the pellets were washed three times with 1 ml of HEG, and proteins were resolved by SDS-PAGE and Western blotting.
Preparation of a Recombinant Baculovirus Expressing FLAG-tagged p50 cdc37 -The cDNA for p50 cdc37 , isolated from a human lymphocyte cDNA library through hybridization with the previously described chick cdc37 cDNA homolog (5, 10), 2 served as template to amplify by polymerase chain reaction the open reading frame, starting from codon 2 and including 285 base pairs of 3Ј-untranslated sequence. The amplified human p50 cdc37 cDNA was subcloned into the NotI site of pFastBAC1-FLAG, a modified version 3 of the baculoviral pFastBAC1 vector (Life Technologies, Inc.), in frame with a FLAG propeptide sequence. The resulting construct was verified by DNA sequencing and subsequently used to generate FLAG-p50 cdc37 encoding recombinant baculoviruses and high titer stocks, using the BAC-TO-BAC baculovirus expression system from Life Technologies, Inc.
Purification of FLAG-p50 cdc37 from Sf9 Cells-Sf9 cells (1.8 ϫ 10 7) were cultured into T-162 cm tissue culture flasks and infected with a baculovirus expressing FLAG-p50 cdc37 at a multiplicity of infection of 3, then incubated for 2 days at 27°C. Cytosol was prepared from infected cells and diluted 1:1 with TEG, the nonionic detergent Nonidet P-40 was added to 0.02%, and the diluted cytosol was rotated for 1 h at 4°C and centrifuged at 100,000 ϫ g. FLAG-tagged p50 cdc37 was then purified using M2-agarose beads (IBI) according to manufacturer's instructions.
Preparation of an Antibody against p50 cdc37 -Human p50 cdc37 (amino acids 2-378) expressed as GST fusion protein was purified by GSH-Sepharose chromatography and used to generate p50 cdc37 -specific antisera in rabbits. Although the rabbit anti-p50 cdc37 antiserum exhibits a wide reactivity for p50 cdc37 across species, it does not recognize the endogenous p50 cdc37 expressed in insect Sf9 host cells.

RESULTS
Competition for Binding of p50 cdc37 to hsp90 -In a previous study (29), we showed that a fragment containing the TPR domains of CyP-40 competed for the binding of FKBP52 and CyP-40 to hsp90. However, the binding of p60/Hop and p50 cdc37 was not inhibited by the highest achievable level of the CyP-40 TPR fragment. Subsequently, we found that the fragment of PP5 containing its four TPRs bound much more tightly to hsp90 and competed for p60/Hop binding (28). In Fig. 1, we use this tight binding PP5 TPR fragment to compete for the binding of p50 cdc37 and several TPR domain proteins to hsp90. In this experiment, an immune pellet alone (lane 1) or immune pellets prebound with purified hsp90 (lanes 2 and 3) were incubated with an hsp90-free hydroxylapatite pool of rabbit brain cytosol (29) that contains p50 cdc37 as well as p60/Hop, FKBP52, and CyP-40. As shown in lane 2 (Fig. 1), all four of these proteins bound to hsp90. However, in the presence of the PP5 TPR fragment (lane 3) binding of CyP-40 and FKBP52 was blocked and p60/Hop binding was inhibited. The p60/Hop band was probed with 125 I-labeled counter antibody, excised, and counted to determine the extent of inhibition. The PP5 TPR domain fragment (lane 3) reduced the binding of p60/Hop by 65% but it did not compete for the binding of p50 cdc37 to hsp90 (cf. lanes 2  and 3).
In contrast to the TPR domain fragment, intact TPR domain proteins do compete for binding of p50 cdc37 to hsp90. In the experiment of Fig. 2A, hsp90- Hop (lane 4). The pellets were then incubated with the rabbit brain hydroxylapatite pool, and binding of p50 cdc37 to hsp90 was assayed. It is clear from lane 4 that p60/Hop competes for the binding of p50 cdc37 to hsp90. In Fig. 2B, hsp90-bound immune pellets were preincubated with lysate from Sf9 cells expressing FLAG-p50 cdc37 and then incubated with the hydroxylapatite pool. In the presence of the Sf9-expressed p50 cdc37 (lane 4), the binding of p60/Hop to hsp90 was competed. As shown in Fig. 2C, Sf9-expressed FLAG-PP5 also competes for the binding of p50 cdc37 to hsp90.
The cloning and sequencing of p50 cdc37 showed that it does not possess a TPR domain (6,7), 2 yet intact TPR domain proteins compete for its binding to hsp90. As shown in the experiment of Fig. 1, we have occasionally observed an increase in the amount of p50 cdc37 binding to hsp90 when the PP5 TPR fragment is present. Such an increase in p50 cdc37 binding would occur if binding of TPR proteins to the TPR acceptor site on hsp90 prevented access of p50 cdc37 to its binding site, but the small PP5 TPR fragment did not. p50 cdc37 Does Not Exist in Native hsp90 Heterocomplexes with TPR Proteins-These competition data suggest that the binding site for p50 cdc37 may be close enough to the TPR binding site on the surface of hsp90 such that the binding of a protein to one site blocks access of the other protein to its binding site. If that is true, p50 cdc37 should not exist in a native hsp90⅐TPR protein complex unless there is a binding site for each of the proteins on each half of the hsp90 dimer. In which case, immunoadsorption of an hsp90-bound TPR protein should yield not only co-immunoadsorption of some p50 cdc37 but also of other TPR proteins. In the experiment of Fig. 3, either hsp90 or p60/Hop was immunoadsorbed from rabbit reticulocyte lysate and the washed immune pellets were assayed for coadsorbed proteins. Immunoadsorption of hsp90 (lane 2) yielded coadsorption of the four TPR proteins (p60/Hop, PP5, FKBP52, and CyP-40) as well as the non-TPR-containing p50 cdc37 . Immunoadsorption of p60/Hop (lane 4) yielded coadsorption of a substantial amount of hsp90 but no coadsorption of p50 cdc37 or of other TPR proteins. It is possible that p60/Hop is unique among TPR proteins in that it is present in hsp90 heterocomplexes free of p50 cdc37 . We were unable to test this possibility by coimmunoadsorption of hsp90 heterocomplexes with antibodies directed against p50 cdc37 or the immunophilins because of their substantial cross-reactivity. The antiserum against p50 cdc37 (␣-p50), for example, reacts on immunoblots with both PP5 and CyP-40 (data not shown). Given the cross-reactivity of the antisera, we used a monoclonal antibody against the FLAG epitope to immunoadsorb Sf9-expressed FLAG-p50 cdc37 and FLAG-PP5 and assayed for coimmunoadsorbed proteins. In the experiments of Fig. 4, a small amount of Sf9 cytosol with the expressed FLAGtagged protein was first incubated with rabbit reticulocyte lysate to ensure complete equilibration of the FLAG-p50 cdc37 and FLAG-PP5 with rabbit hsp90. The FLAG-tagged proteins were then immunoadsorbed with the M2 monoclonal anti-FLAG IgG, and coadsorbed proteins were assayed. It is clear that immunoadsorption of FLAG-p50 cdc37 yields coadsorption of hsp90, but there is no coadsorption of the rabbit TPR domain proteins PP5 or FKBP52. Similarly, immunoadsorption of FLAG-PP5 yielded coadsorption of hsp90, but there is no coadsorption of p50 cdc37 . Taken together, these coimmunoadsorption observations and the competition data of the previous section lead us to conclude that p50 cdc37 can bind to hsp90 when the TPR acceptor site is occupied by the TPR domain fragment of PP5 but not when the site is occupied by an intact TPR domain protein.
p50 cdc37 Binds Directly to Raf-The exclusive binding of a TPR domain protein or p50 cdc37 to hsp90 explains why there are separate heterocomplexes but not why the dominant Raf⅐hsp90 heterocomplex contains p50 cdc37 instead of an immunophilin. The experiments of Fig. 5 were performed to determine if p50 cdc37 also binds directly to Raf. In the experiment of Fig. 5A, v-Raf-1 was immunoadsorbed from rat 3Y1 cell cytosol, and the native heterocomplex of Raf with rat hsp90 and p50 cdc37 is shown in lane 2. Raf was stripped of its associated proteins (lane 4) and the stripped Raf immune pellet was incubated with purified FLAG-p50 cdc37 (lane 6). As shown in lanes 5 and 6 of Fig. 5A, FLAG-p50 cdc37 binds to the immune pellet in a manner that is specific for the presence of v-Raf-1. Fig. 5B shows that the catalytic domain of bacterially derived c-Raf is sufficient for direct p50 cdc37 binding. In this experiment the GST-tagged c-Raf COOH-terminal fragment expressed in E. coli was immobilized on glutathione-agarose and stripped with salt and heating (lane 2). When the immobilized c-Raf catalytic domain was incubated with cytosol from Sf9 cells expressing FLAG-p50 cdc37 , Raf⅐p50 cdc37 complexes were formed (Fig. 5B, lane 4). The immobilized c-Raf catalytic domain fragment also bound purified FLAG-p50 cdc37 (lane 6) but not purified FKBP52 (lane 8).
It is known that the catalytic domain of c-Raf is sufficient for forming the heterocomplex with hsp90 (31), and these data of Fig. 5 suggest a model in which p50 cdc37 may contact Raf as well as hsp90 when it is in the Raf⅐hsp90⅐p50 cdc37 heterocomplex. The direct binding of p50 cdc37 to Raf could allow the kinase to determine its existence in hsp90 heterocomplexes containing p50 cdc37 as opposed to immunophilins.

DISCUSSION
Previous studies have shown that FKBP52 and CyP-40 compete with each other for binding to hsp90 (21,24) and that these two immunophilins and the TPR-containing protein phosphatase, PP5, exist in separate heterocomplexes with hsp90 (28). In this work, we provide evidence that p50 cdc37 cannot bind to hsp90 when the TPR acceptor site on hsp90 is occupied by one of the TPR domain proteins, such as p60/Hop or PP5. However, p50 cdc37 does bind to hsp90 when the small TPR domain fragment of PP5 occupies the TPR acceptor site and prevents binding of the TPR domain proteins. These com-  4. p50 cdc37 , PP5 and FKBP52 exist in independent heterocomplexes with hsp90. Aliquots (100 l) of rabbit reticulocyte lysate were immunoadsorbed with nonimmune IgM or 3G3 antibody against hsp90. Other 100-l aliquots of reticulocyte lysate were incubated for 30 min at 30°C with 10 l of Sf9 cytosol overexpressing FLAG-p50 cdc37 or FLAG-PP5, and then immunoadsorbed with nonimmune IgG or the M2 monoclonal IgG against the FLAG epitope. Lane 1, immunoadsorption with nonimmune antibody, lane 2, immunoadsorption with the antibody indicated at the top of each pair of lanes. Note that the AC88 antibody used to blot hsp90 reacts with rabbit but not insect (Sf9) hsp90 (31). Because PP5 migrates close to FKBP52 on SDS-PAGE, any FKBP52 that might be present in the immune adsorbate would be obscured by the large amount of FLAG-PP5; thus, FKBP52 was not assayed. petition data suggest that the p50 cdc37 binds to a site on the surface of hsp90 that is close to the TPR binding site and that binding of a protein to one site may block binding of a protein to the other site.
It could be argued that binding of a protein, such as p60/Hop, PP5, or an immunophilin, to the TPR binding site on hsp90 influenced the conformation of hsp90 such that the affinity of a p50 cdc37 binding site located at some distance from the TPR binding site was reduced. However, the fact that binding of the PP5 TPR fragment to hsp90, if not augmenting, at least does not reduce the binding of p50 cdc37 argues against such an allosteric effect. Thus, we propose that p50 cdc37 binds to a site on hsp90 that is topologically adjacent to the TPR binding site, and at any instant in time, an hsp90 heterocomplex contains either p50 cdc37 or one of the TPR domain proteins.
Although hsp90 is present in cytosols as a dimer, it is likely that only one molecule of p50 cdc37 or TPR domain protein can be bound by the dimer. In the event that independent binding sites were available on each dimer, we should have recovered mixed complexes in which immunoadsorption of one TPR protein from cytosol yields coimmunoadsorption of other TPR proteins and p50 cdc37 . A stoichiometry in which one of these proteins is bound per hsp90 dimer is consistent with careful crosslinking studies of Gehring and his co-workers (42)(43)(44), who established a stoichiometry for untransformed steroid receptor heterocomplexes of one steroid-binding protein, two molecules of hsp90, and one molecule of immunophilin. However, it must be emphasized that the stoichiometry of hsp90⅐immunophilin and hsp90⅐p50 cdc37 complexes has not been determined directly in the absence of receptors or protein kinases, and the stoichiometry in two-protein versus the three-protein complexes could be different.
hsp90 has been found in complex with a confusing variety of proteins, and the model shown in Fig. 6 is presented to sort out established binding domains on the surface of hsp90. More than a dozen transcription factors and more than a dozen protein kinases have been reported to be in heterocomplex with hsp90 (see Table I in Ref. 1 for summary). These proteins are represented by the chaperoned protein in Fig. 6, and they must bind to a common domain (chaperoning domain) on hsp90 which appears to be located in its COOH-terminal half (45,46). Under nondenaturing conditions, hsp90 purifies as a dimer, with the dimerization site likely lying in a COOH-terminal region (47). The NH 2 -terminal domain (amino acids 1-221) of hsp90 contains a nucleotide binding site (48,49). Binding of p23 to the ATP-dependent conformation of hsp90 requires regions outside of the 1-221 domain, but on the basis of the observations of Toft and his co-workers (49,50), it is reasonable to predict that, in the three-dimensional structure of hsp90, the nucleotide binding domain (ATP/ADP switch domain), the p23 binding site, and the chaperoning domain are situated close to each other, forming an active center that determines a conformational change in the chaperoned protein.
The TPR binding domain of hsp90 is required for the binding p60/Hop (34), which in turn is required for steroid receptor⅐hsp90 heterocomplex assembly (35) and dissociates from hsp90 during the assembly process (51). Mature steroid receptor heterocomplexes have been reported to contain FKBP51, FKBP52, CyP-40, or PP5 bound to this TPR binding site (1,2). Only one of these TPR proteins exists in a receptor⅐hsp90 heterocomplex at any time (24,25). However, because binding of TPR proteins to the TPR binding site on hsp90 is a reversible process, over time, a single receptor⅐hsp90 heterocomplex may be associated with PP5 and any of the TPR domain immunophilins. A 38-kDa FKBP homolog with three TPR domains called ARA3 has been isolated with dioxin (Ah) receptor⅐hsp90 complexes (52). In addition to binding to hsp90, ARA3 appears to bind to the dioxin receptor directly (52), and there is indirect evidence that FKBP52 may contact the transformed glucocorticoid receptor (53). Thus, in Fig. 6, the TPR binding site on hsp90 has been placed such that the TPR protein that occupies the site may also contact the chaperoned protein.
The evidence of this study suggests that the p50 cdc37 component of protein kinase⅐hsp90 heterocomplexes binds, in vitro, to a site that is topologically adjacent to the TPR binding site on hsp90 but that p50 cdc37 and a TPR domain protein may not be able to bind to the same hsp90 dimer. The dashed borders of the TPR domain protein and p50 cdc37 in Fig. 6 indicate the overlapping space occupied by both proteins that accounts for their mutual competition for binding to hsp90. Because p50 cdc37 binds directly to Raf (Fig. 5) and to Cdk4 (6, 10), it has also been positioned such that it could contact the chaperoned protein as well as hsp90.
In the dynamic state when Raf⅐hsp90 complexes are being assembled, dissociation of the p60/Hop component of the assembly machinery would expose on hsp90 both the binding site for TPR domains and the adjacent binding site for p50 cdc37 . As both the TPR domain proteins and p50 cdc37 bind in a readily reversible manner to their respective sites on hsp90, simultaneous binding of p50 cdc37 directly to Raf should rapidly select for Raf⅐hsp90⅐p50 cdc37 complexes, which is the composition of native Raf⅐hsp90 heterocomplexes isolated from cytosols (31). Thus, the combination of exclusive binding of p50 cdc37 versus a TPR domain protein to hsp90 plus direct binding of p50 cdc37 to Raf allow the protein kinase to determine the dominant heterocomplex composition.
FIG. 6. Model of protein binding sites on hsp90. The chaperoned protein represents any of the many protein kinases or transcription factors that are recovered from cytosols in stable complexes with hsp90. The TPR protein can be p60/Hop, PP5, or any of the immunophilins that have been identified in transcription factor⅐hsp90 heterocomplexes. To date, p50 cdc37 has been recovered only with protein kinase⅐hsp90 heterocomplexes. Molybdate (MoO 4 ϭ ), which stabilizes hsp90 in its ATPdependent conformation (50), interacts with the nucleotide binding site.