Specific association of a set of molecular chaperones including HSP90 and Cdc37 with MOK, a member of the mitogen-activated protein kinase superfamily.

We have recently identified and cloned a novel member of mitogen-activated protein kinase superfamily protein, MOK (Miyata, Y., Akashi, M., and Nishida, E. (1999) Genes Cells 4, 299-309). To address its regulatory mechanisms, we searched for cellular proteins that specifically associate with MOK by co-immunoprecipitation experiments. Several cellular proteins including a major 90-kDa molecular chaperone HSP90 were found associated with MOK. Treatment of cells with geldanamycin, an HSP90-specific inhibitor, rapidly decreased the protein level of MOK, and the decrease was attributed to enhanced degradation of MOK through proteasome-dependent pathways. Our data suggest that the association with HSP90 may regulate intracellular protein stability and solubility of MOK. Experiments with a series of deletion mutants of MOK indicated that the region encompassing the protein kinase catalytic subdomains I-IV is required for HSP90 binding. Closely related protein kinases (male germ cell-associated kinase and male germ cell-associated kinase-related kinase) were also found to associate with HSP90, whereas conventional mitogen-activated protein kinases (extracellular signal-regulated kinase, p38, and c-Jun N-terminal kinase/stress-activated protein kinase) were not associated with HSP90. In addition, we found that other molecular chaperones including Cdc37, HSC70, HSP70, and HSP60 but not GRP94, FKBP52, or Hop were detected specifically in the MOK-HSP90 immunocomplexes. These results taken together suggest a role of a specific set of molecular chaperones in the stability of signal-transducing protein kinases.

MAP 1 kinases have been known to be activated by various stresses and growth stimuli and to be involved in transmitting signals from outside of cells into cell nuclei (1)(2)(3)(4)(5)(6)(7)(8). In mamma-lian systems, three major types of MAP kinases, namely ERK, p38, and JNK/SAPK, have been identified and characterized extensively. Recently, striking progress in molecular cloning techniques and in bioinformatics has greatly facilitated homology-based discoveries of novel protein kinases that are more or less closely related to conventional MAP kinases (9). Among them, protein kinase catalytic domains of ERK5 (10,11) and ERK7 (12) are highly homologous to those of conventional MAP kinases, and activation mechanisms and downstream targets of ERK5 have been described (for review, see Ref. 13).
We have recently identified and cloned a novel member of the MAP kinase superfamily, MOK, by a computer search of the expressed sequence tags database followed by screening of human and mouse cDNA (14). MOK is most similar to previously identified protein kinases, male germ cell-associated kinase (MAK) (15) and MAK-related kinase (MRK) (16). The kinase catalytic domains of MOK, MAK, and MRK are highly homologous to that of conventional MAP kinases, especially at their activation loops where all of them encode the MAP kinase characteristic TXY motifs (9). MRK is abundantly expressed in ovary, whereas MAK and MOK are abundant in testis; hence they are postulated to play roles in regulating the meiotic cell division in reproductive tissues. Previously we observed that MAP kinase-stimulating treatment of cells including serum addition, anisomycin treatment, and osmotic shock did not significantly activate MOK (14), suggesting that specific upstream signal-transducing molecules are involved in the activation processes of MOK. Treatment of cells with tumor promoter 12-O-tetradecanoylphorbol-13-acetate activated MOK, and the activity of MOK was dependent on its phosphorylation status (14), but the precise physiological functions and cellular regulatory factors of MOK remain unknown.
Catalytic subunits of certain protein kinases are associated with cellular proteins. cAMP-dependent protein kinase catalytic subunit is complexed with a regulatory subunit (for a review, see Ref. 17). CK2 (casein kinase II) is composed of catalytic ␣ and ␣Ј subunits and a regulatory ␤ subunit (18,19). Cell cycle-regulated Cdk protein kinases are accompanied by their specific activating subunits, cyclins (20 -22). MAP kinases are known to physically associate with upstream activating kinases either directly (23,24) or via common tethering proteins called molecular scaffold (reviewed in Refs. 25 and 26). The specific associations of noncatalytic proteins are important for catalytic subunits in their regulation such as inhibitory control, signal-dependent activation, substrate recognition, or cellular localization.
More generally, a number of protein kinases are known to be associated with cellular molecular chaperones. Among many molecular chaperones, HSP90 and Cdc37 are often observed to be complexed with signal-regulated protein kinases such as pp60 v-src (27,28), Raf1 (29), Cdk4 (30 -32), and CK2 (33)(34)(35). HSP90 is ubiquitously and abundantly expressed in all organisms and tissues and associates with various kinds of important cellular proteins including transcription factors and protein kinases, thus assisting correct folding and functions of target proteins (36 -39). On the other hand, Cdc37 is thought to be a unique molecular chaperone that is involved specifically in the folding and regulation of protein kinases (30, 35, 40 -42).
To elucidate the physiological function and regulation of MOK, we have biochemically searched for cellular binding proteins of MOK in this report and identified several cellular MOK-associating proteins. The results have revealed that the associated proteins consist of a specific set of molecular chaperones including HSP90 and Cdc37. Functional importance of the association of molecular chaperones on the stability of MOK has been presented and discussed.

EXPERIMENTAL PROCEDURES
Plasmids-Plasmids encoding HA-tagged MOK (wild type and kinase-dead) were described previously (14). Plasmids encoding FLAGtagged MOK were constructed by inserting a BamHI fragment including the entire coding region of MOK into a BamHI site of pFLAG-CMV2 (Kodak Scientific Imaging Systems). MRK cDNA was described previously (16), and mammalian expression vectors with or without tags (C-terminal Myc tag or N-terminal HA tag) were constructed and kindly provided by Drs. S. Abe and H. Ellinger-Ziegelbauer. Expression vectors encoding two alternative splicing forms of rat MAK (MAK2.6 and MAK3.8) were described (15). Mammalian expression vectors encoding conventional MAP kinases (pSR␣-HA-MAP kinase for HA-tagged ERKtype MAP kinase from Xenopus, pSR␣-HA-p38 from human, and pSR␣-HA-SAPK/JNK from rat) are described previously (43,44). Rat Cdc37 cDNA was originally isolated by Dr. S. Sakiyama and colleagues (45). An expression plasmid of rat Cdc37 in pSR␣-HA1 was kindly provided by Dr. S. Matsuda (46). A BglII fragment encoding the Cdc37 coding region was excised and inserted into a BamHI site of pFLAG-CMV2 to make mammalian FLAG-Cdc37 expression vector.
Other Reagents-Geldanamycin was purchased from Life Technologies, Inc., and stock solution was prepared at 10 mM in Me 2 SO. MG-132 was purchased from Calbiochem, and stock solution was made at 10 mM in Me 2 SO. Tran 35  Series of Deletion Mutants of MOK-A series of deletion mutants of MOK was produced by in vitro mutagenesis of plasmids as described below. C-terminal deletion mutants were made by introducing stop codons at the indicated sites to make truncated proteins. For example, MOK(N107) encodes amino acids 1-107 of MOK. N-terminal deletion mutants were made by introducing a new MfeI restriction enzyme site by in vitro mutagenesis at the indicated positions. The wild type mouse MOK coding region possesses a unique MfeI site near its N-terminal end (amino acid 7); thus MfeI treatment and self-ligation of plasmids after introducing an additional MfeI site resulted in the N-terminal deletion mutants. For example, MOK(D8 -195) encodes a protein with amino acids 1-7 fused to the sequence from amino acid 196 to the end of MOK, lacking amino acids 8 -195. Site-directed in vitro mutagenesis was performed using mutagenic primers and DpnI digestion essentially as described (14). The sense sequences of mutagenic primers used were 5Ј-GAGAAAAAGATTATG-CTCTAGATGTATCAGCTG-3Ј for MOK(N107), 5Ј-CACCGACGGGTT-CTAGACATACAAGATGGACC-3Ј for MOK(N178), 5Ј-CAGCACCCCT-ATTTCTAGATGCAGAGGGCAGC-3Ј for MOK(N285), 5Ј-CAGAAAAT-CTGGTTCAATTGCACTAATATG-3Ј for MOK(D8 -078), 5Ј-CTGCGTG-TTCTACGCAATTGCCAGCCTGCAGC-3Ј for MOK(D8 -195), and 5Ј-G-TCCCAAGTTCTCAATTGTGCCAGAATCATC-3Ј for MOK(D8 -309).
Expression in Mammalian Cells and Immunoprecipitation of Tagged MOK-COS7 cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal calf serum. Mammalian expression plasmids encoding various MOK constructs were transfected into COS7 cells by electroporation as described previously (14). Cell lysates were prepared, and the tagged protein was immunoprecipitated as described (14). For affinity purification of FLAG-tagged proteins, beads that had been covalently conjugated with anti-FLAG antibody were used. For large scale affinity purification of FLAG-MOK and its associated proteins, anti-FLAG affinity beads were packed into a column (inner diameter ϭ 0.5 cm), and cell extract was applied to the column three times followed by extensive washing with cell lysis buffer before elution.
Pulse-Chase Experiments with Geldanamycin Treatment of Cells-COS7 cells were transfected with HA-MOK plasmid as above. 48 h later, the culture medium was replaced with Met-and Cys-free Dulbecco's modified Eagle's medium containing 10% fetal calf serum, and cells were incubated with or without geldanamycin (5 M) for 2 h. Then cells were labeled with [ 35 S]Met/[ 35 S]Cys in fresh Met-and Cys-free medium with or without geldanamycin for 2 h. Next, radioactive medium was removed, and cells were further incubated for chase in complete medium containing 10 g/ml cycloheximide for various times indicated before lysis. When indicated, geldanamycin (5 M) was also included during the chase period. The amount of radiolabeled MOK was determined by immunoprecipitation and SDS-PAGE followed by fluorography of dried gels.
Treatment of Cells with Geldanamycin and MG132-MOK-transfected COS7 cells were treated with MG132 (25 M) or vehicle (Me 2 SO) for 3 h then with geldanamycin (5 M) or control Me 2 SO for 2 or 4 h. MG132 was also included in the medium during the incubation with geldanamycin. Cells were collected as described above and then centrifuged at 17,000 ϫ g for 60 min. The supernatants were recovered as soluble fractions, and the precipitates were solubilized by adding diluted (5:7) SDS sample buffer and then boiling for 5 min to obtain insoluble fractions.
Purification of MOK-associated Proteins-MOK-associated proteins were isolated by using anti-FLAG affinity beads as described above. The associated proteins were eluted by treating the beads twice with a high salt buffer containing 1.5 M NaCl, 2 mM EDTA, 50 mM Tris-HCl, pH 7.4. Eluted proteins were diluted 10ϫ in a buffer consisting of 50 mM Tris-HCl, 2 mM EDTA, pH 7.4, and applied on a MonoQ ion exchange column. The column was washed extensively with 100 mM NaCl, 2 mM EDTA, 50 mM Tris-Cl, pH 7.4, and bound proteins were eluted and purified by a liner salt gradient (100 mM to 1.5 M NaCl) in 2 mM EDTA, 50 mM Tris-Cl, pH 7.4. Finally protein samples were separated by SDS-PAGE, and protein bands were cut out from the gel. The gel pieces were equilibrated in a buffer (50 mM Tris-Cl, 150 mM NaCl, pH 8.0) and homogenized in diluted (5:7) SDS sample buffer, and then the gel suspensions were applied to a second SDS-PAGE for analysis.
Expression and Purification of HSP90 and Cdc37-Human HSP90␣ cDNA was originally from Dr. K. Yokoyama (51). The coding region was inserted into the EcoRI site of pMAL-cRI (New England Biolabs) by blunt ligation. The resulting expression plasmid encodes human HSP90␣ with maltose-binding protein in the N termini. The plasmid was introduced into Escherichia coli host strain TB1. After reaching an A 600 level of 0.4, 1 mM isopropyl-1-thio-␤-D-galactopyranoside was added, and the culture was further shaken for 2 h at 37°C. E. coli cells were collected by centrifugation 3300 ϫ g for 15 min at 2°C, washed with ice-cold phosphate-buffered saline, frozen at Ϫ80°C, and then resuspended in a lysis buffer (10 mM sodium phosphate, 30 mM NaCl, 0.25% Tween 20, 10 mM 2-mercaptoethanol, 10 mM EDTA, 10 mM EGTA, pH 7.0) supplemented with 10 g/ml leupeptin, 1.7 mM phenylmethylsulfonyl fluoride, and 0.1 trypsin inhibitor unit/ml aprotinin. Then the cells were ruptured by repeated freeze-thawing followed by sonication. NaCl concentration was adjusted to 500 mM, and the solution was centrifuged at 11,000 ϫ g for 30 min at 2°C to recover the supernatant. Purification of maltose-binding protein-HSP90 fusion protein by an amylose column and cleavage with Factor Xa protease were performed essentially as described in the manual of the manufacturer. Recombinant HSP90 was further purified by MonoQ column chromatography using a linear gradient (20 -2000 mM NaCl in 50 mM Tris-Cl, pH 7.4) and then desalted with PD10 column (Amersham Pharmacia Biotech) in a buffer containing 50 mM Tris-Cl, 0.1 mM EGTA, 0.5 mM dithiothreitol, 10 mM sodium azide, 10% glycerol, pH 7.4. HSP90 from mammalian sources was purified as described previously (33). Rat Cdc37 BglII fragment described above was inserted into a BamHI site of pGEX6P2 (Amersham Pharmacia Biotech). The resulting plasmid encoding glutathione S-transferase-Cdc37 was introduced into E. coli strain BL21-CodonPlus (DE3)-RIL (Stratagene). After reaching an A 600 level of 0.6, 1 mM isopropyl-1-thio-␤-D-galactopyranoside was added for induction, and the culture was further shaken for 4 h at 23°C. E. coli cells were collected by centrifugation at 3300 ϫ g for 15 min at 2°C, washed with ice-cold phosphate-buffered saline, frozen at Ϫ80°C, and resuspended in B-PER solution (Pierce) supplemented with 10 g/ml leupeptin, 1.7 mM 4-(2-aminoethyl)benzenesulfonyl fluoride, and 0.1 trypsin inhibitor unit/ml aprotinin. The extract was clarified by centrifuging at 18,000 ϫ g for 15 min at 2°C. Glutathione-Sepharose (Amersham Pharmacia Biotech) was mixed with the extract and rotated for 3 h at 4°C. The beads were washed extensively with phosphatebuffered saline containing 1% Triton X-100 and then with cleavage buffer (50 mM Tris-Cl, 150 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, pH 7.0). Suspension of beads in the cleavage buffer was incubated with PreScission protease (Amersham Pharmacia Biotech) with gentle agitation for 16 h at 4°C. Cleaved recombinant Cdc37 was recovered by centrifugation and then dialyzed extensively in a buffer (50 mM Tris-Cl, 150 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, 10% glycerol, pH 7.4).
In Vitro Protein Kinase Assay-MOK was expressed in COS7 cells and then immunoprecipitated as described (14). The immunoprecipitates were mixed with 1 g of substrate proteins and incubated at 30°C for 30 min with gentle mixing in phosphorylation buffer containing [␥-32 P]ATP as described (14).
Other Procedures-Silver staining of SDS-polyacrylamide gels was performed using 2D-Silver Stain II kit (Daiichi Pure Chemicals). Western blotting was performed using horseradish peroxidase-conjugated secondary antibodies, and detection was performed using the chemiluminescent system as described (14).

Identification of Proteins That
Are Co-immunoprecipitated with Transfected MOK-COS7 cells were transfected with a plasmid encoding HA-tagged MOK protein kinase, and cell lysates were prepared. MOK protein was immunoprecipitated from the cell lysates with anti-HA tag antibody, and the immunocomplexes were analyzed by SDS-PAGE and silver staining to visualize possible cellular MOK-associated proteins. As shown in Fig. 1A, at least one cellular protein with a molecular mass of ϳ90 kDa, marked with an asterisk, was specifically co-immunoprecipitated along with MOK (lane 4). Although a lot of nonspecifically attached proteins can be seen because of the high sensitivity of silver staining, specificity of the coimmunoprecipitation of this protein with MOK was confirmed as follows. First, the 90-kDa band could not be observed when lysates of control cells without HA-MOK transfection were used (Fig. 1A, lanes 1 and 3). In addition, the association of the 90-kDa protein was not observed when nonimmune antibody was used for immunoprecipitation (Fig. 1A, lanes 1 and 2). In fact, we confirmed immunoprecipitation of MOK only in lane 4 (Fig. 1B); thus MOK-associated proteins should also appear only in lane 4 as was the case for the 90-kDa protein. The molecular mass of expressed HA-MOK is ϳ47 kDa (Fig. 1B); thus the 90-kDa protein cannot be a degradation product of expressed MOK. Taken together, at least one (but not necessarily only one) cellular protein with a molecular mass of 90 kDa associates with MOK.
Identification of 90-kDa MOK-associated Protein as a Major Molecular Chaperone HSP90 -Several protein kinases includ- ing pp60 v-src , Raf1, Cdk4, and CK2 had been reported to be associated with a 90-kDa cellular protein, and this protein had been identified as a major cytosolic molecular chaperone/heat shock protein, HSP90. Because the molecular mass of the MOK-associated protein observed in Fig. 1A is similar to that of HSP90, we examined the possibility of an association between MOK and HSP90 directly by Western blotting. The same set of immunoprecipitates as in Fig. 1 was probed with antibody specific for HSP90. As clearly shown in Fig. 2A, HSP90 was detected in the immunocomplexes only in lane 4, which shows the anti-HA immunoprecipitate from HA-MOK-expressing cell lysates. HSP90 was not detected in control immunoprecipitates (lanes 1-3), confirming specificity of the association between HSP90 and MOK. The association of HSP90 with HA-MOK was not a result of binding of HSP90 to the HA tag moiety of the fusion protein because we also observed the specific co-immunoprecipitation between FLAG-MOK and HSP90 (Fig. 2B). Although these results may not rule out the possibility of other 90-kDa MOK-associated proteins, they clearly indicate that MOK is associated with HSP90 in mammalian cells and that HSP90 can be isolated by co-immunoprecipitation with MOK.
Treatment of Cells with a Specific HSP90 Function Inhibitor, Geldanamycin, Induced Destabilization of MOK-Geldanamycin, a low molecular mass compound that is a derivative of ansamycin-type benzoquinoid (52), has been shown to inhibit the molecular chaperone function of HSP90 (53). After observing the physical association of HSP90 with MOK as above, we examined the effect of geldanamycin treatment of cells on MOK. COS7 cells had been transfected with HA-MOK and treated with 5 M geldanamycin for the indicated time periods up to 6 h. Cell lysates were prepared, and changes of the intracellular amount of MOK after geldanamycin treatment were examined by Western blotting with antibody against MOK. As shown in Fig. 3A, the amount of MOK was dramatically and rapidly decreased after the addition of geldanamycin. An identical result was obtained when we used anti-HA antibody for Western blotting (data not shown). This result strongly suggested that MOK requires the molecular chaperone function of HSP90 for its intracellular stability. As a control, ERK2, a conventional member of the MAP kinases, did not respond to geldanamycin treatment of cells (Fig. 3B), indicating that the sensitivity to geldanamycin treatment is not a general feature of MAP kinase superfamily members. The time course of the decrease of MOK was similar to that of endogenous Raf1 (data not shown), a relatively unstable HSP90 target (29). The decrease of HSP90-associated proteins with much longer cellular half-lives such as SV40 large T antigen required a much longer incubation with geldanamycin (54). These facts suggested that the intracellular half-life of MOK should be relatively short in the absence of HSP90 function.
To check directly whether the half-life of MOK in cells is shortened by geldanamycin treatment, we performed pulsechase experiments. COS7 cells transfected with HA-MOK were pulse-labeled with [ 35 S]Met/[ 35 S]Cys for 2 h and then chased in complete medium up to 2 h. Geldanamycin (5 M) had been added to cells 2 h before the start of pulse-labeling and also included during both the pulse and the chase periods. Several time points after the start of chase, cell extracts were prepared, MOK was immunoprecipitated, and the amounts of radioactivity of MOK were analyzed by SDS-PAGE and fluorography. As shown in Fig. 3C (upper panel), in the control (treated with vehicle Me 2 SO) cells, MOK was relatively stable, and only a minor decrease of synthesized MOK could be seen even after 2 h. On the other hand, in the presence of geldanamycin, MOK protein was very unstable, and the half-life of MOK in cells was shortened to less than 1 h (Fig. 3C, lower panel). In lane 1 of Fig. 3C, the amounts of radiolabeled MOK were almost equivalent between control and geldanamycin-treated cells. Moreover, we included the protein synthesis inhibitor cycloheximide during the chase period. Thus we concluded that geldanamycin treatment enhanced the net degradation of synthesized MOK without affecting the rate of protein synthesis of MOK. Taken together, these results indicate that function of HSP90 was required for the stable existence of MOK in cells after synthesis, and expressed MOK was degraded rapidly in the absence of HSP90 function.
Role of Proteasome-dependent Pathways in the Geldanamycin-induced Destabilization of MOK-Protein degradation can be achieved by several protease systems in cells. To examine the roles of proteasome-dependent degradation pathways in the instability of MOK in the absence of HSP90 function, we set up experiments with a specific inhibitor of proteasome, MG132. MOK-expressing COS7 cells were pretreated with MG132 (25 M) for 3 h. Then cells were incubated with or without geldanamycin for up to 4 h, and MG132 was also included when indicated during geldanamycin treatment. After 2 or 4 h of incubation, cells were ruptured, and the amounts of MOK in cell extracts were determined by Western blotting with anti-HA antibody. We anticipated that geldanamycin-induced degradation of MOK could be blocked by MG132 treatment if the degradation is mediated by proteasome-dependent pathways. Unexpectedly, the amounts of MOK in soluble cellular extracts decreased almost in the same manner even in the presence of MG132 as indicated in Fig. 4A (compare lanes 1-3  with lanes 4 -6). Then we examined the amounts of MOK in insoluble cell fractions recovered from centrifuge precipitates. MOK in the insoluble fractions also decreased by geldanamycin  (Fig. 4B, lanes 1-3) as in the case of soluble fractions. Interestingly, in the presence of MG132, treatment of cells with geldanamycin induced a significant increase and accumulation of MOK in the insoluble fractions (Fig. 4B, lanes  4 -6). This result indicates the essential role of proteasome-dependent pathways in the degradation of MOK in the presence of geldanamycin. In other words, loss of the HSP90 function induced destabilization and degradation of MOK, and if the degradation was blocked by proteasome inhibitor, undegradable MOK was aggregated and accumulated in the insoluble fraction.
Determination of an HSP90-binding Region in MOK-MOK consists of a kinase catalytic domain in the N terminus and a C-terminal region without known functional motifs that is not homologous to any other protein. To determine an HSP90binding region within MOK, we made a series of deletion mutants of MOK. Wild type mouse MOK consists of 419 amino acids, and MOK(N107), MOK(N178), and MOK(N285) encode C-terminal deletion mutants (the numbers in parentheses indicate the numbers of amino acids from the N terminus). MOK(D8 -078), MOK(D8 -195), and MOK(D8 -309) encode Nterminal deletion mutants but contain the first 7 amino acids of the N-terminal end (the numbers in the parentheses indicate the numbers of amino acids deleted; see scheme in Fig. 5D). We checked the expression of the deletion mutants by Western blotting and found all of the mutants expressed almost equally with expected molecular sizes (Fig. 5A, lanes 1 and 3-7), except a short N-terminal region of MOK, MOK(N107) (lane 2, see details below). These deletion mutants were examined for their HSP90 binding activity by co-immunoprecipitation experiments. The co-immunoprecipitation of HSP90 was revealed both by silver staining (Fig. 5B) and by Western blotting (Fig.  5C). could not associate with HSP90 (Fig. 5, B and C). The relationship between the HSP90 binding ability and the structures of the MOK deletion mutants was summarized in Fig. 5D. The results clearly indicate that the protein kinase consensus catalytic domain in the N-terminal region of MOK is essential for the HSP90 binding, whereas the nonconserved C-terminal tail is dispensable. More precisely, amino acids 8 -78 corresponding to kinase subdomains I-IV of MOK were very important for the HSP90 binding.
Interestingly, mutant MOK(N107) with a large C-terminal deletion, which was a short N-terminal segment of MOK, was hardly expressed in COS7 cells (Fig. 5A, lane 2), despite the fact that all of the mutants were in the same expression plasmid with the same promoter and 3Ј-untranslated sequences. We also tried to express this fragment in several other expression plasmids and with various lengths of the 3Ј-untranslated region, but we never succeeded in stably expressing this fragment in mammalian cells (data not shown). We postulate that this fragment requires HSP90 association but is not able to bind intracellular HSP90 and therefore is quite unstable and rapidly degraded in cells as in the case of wild type MOK in geldanamycin-treated cells as shown in Fig. 3. On the other hand, three C-terminal fragments were quite stably expressed even though they could not associate with HSP90 (Fig. 5, A-C,  lanes 5-7).
MAK and MRK Are Also Associated with HSP90 -We next examined whether or not other related protein kinases are also associated with HSP90. MOK, MRK, two splicing forms of MAK, ERK (Xenopus MAP kinase), p38, and SAPK/JNK were expressed in COS7 cells (tagged or without a tag) and immunoprecipitated with corresponding antibodies. Then the immunoprecipitates were revealed by Western blotting with anti- HSP90 antibody to examine the kinase-HSP90 association. As shown in Fig. 6A, MRK (lanes 3 and 4), both forms of MAK (lanes 5 and 6), along with MOK (lane 2), clearly did associate with HSP90. On the other hand, none of the conventional MAP kinases (ERK, p38, and JNK/SAPK) was associated with HSP90 (Fig. 6A, lanes 7-9). The expression and immunoprecipitation levels of these three conventional MAP kinases were equal to or rather higher than that of MOK and MRK (Fig. 6B, compare lanes 7-9 with lane 2 and 4), confirming the specificity of the HSP90 binding only to MOK, MRK, and MAK but not to conventional MAP kinases.
Association of Other Cellular Proteins with MOK-We next examined whether other proteins are included in the MOK-HSP90 complexes. We performed scale-up isolation of MOKassociated proteins from lysates of FLAG-MOK-expressing cells by using affinity beads that had been covalently conjugated with anti-FLAG antibody. Then we precisely analyzed what kinds of other cellular proteins were isolated along with MOK and HSP90. As indicated in Fig. 7A (lane 4), in addition to the 90-kDa protein (shown as band 1), three other proteins with approximate molecular masses of ϳ70 (doublet, band 2), ϳ60 (band 3), and ϳ50 kDa (band 4) were specifically isolated with MOK. All of these proteins could not be isolated when control lysates were used (Fig. 7A, lanes 1 and 2) or when control beads were used for affinity purification (Fig. 7A, lanes  1 and 3), confirming the specificity of the association of these proteins with MOK. As expected, the specific isolation of MOK by anti-FLAG affinity beads only from FLAG-MOK-transfected cell lysates (lane 4) was observed by Western blotting with anti-MOK antibody (Fig. 7B).
Association of Cdc37 with MOK-Several HSP90-associated protein kinases are known to be associated also with Cdc37, and mobility of mammalian Cdc37 on SDS-PAGE is ϳ50 kDa. We postulated that the MOK-associated 50-kDa protein shown in Fig. 7A (band 4) can be Cdc37 and examined the possibility by Western blotting with anti-Cdc37 antibody. The results in Fig. 7C clearly indicated the association of endogenous Cdc37 with affinity-purified MOK (lane 4). Cdc37 could not be isolated by the control affinity purifications with nontransfected cell lysates or with control affinity beads (Fig. 7C, lanes 1-3). In addition, we also detected co-immunoprecipitation of HA-MOK with FLAG-Cdc37 when they were co-expressed in COS7 cells (data not shown). Taken together, it is concluded that Cdc37 is specifically associated with MOK-HSP90 complexes in cells.
Association of HSP70 with MOK-The HSP70 family is another major molecular chaperone that has been often observed in HSP90 target protein complexes. We examined the association between HSP70 and MOK by Western blotting with antibody recognizing both HSC70 (constitutive form) and HSP70 (inducible form). MOK-associated proteins were isolated by affinity purification as described above, and HSP70 binding was detected (Fig. 7D, lane 4). The association of HSP70 was not observed by control affinity purifications (Fig. 7D, lanes  1-3). We concluded that HSP70/HSC70 is also a member of the MOK-HSP90-Cdc37 protein complexes. Because the 70-kDa protein in Fig. 7A (band 2) was observed as a closely migrating doublet, we speculated that the doublet consists of HSP70 and HSC70. We cut out separately the upper and lower migrating 70-kDa bands from the gel and examined the reactivity of eluted proteins with two antibodies that recognize specifically either HSP70 or HSC70. The upper (slower) migrating band was recognized by anti-HSC70 but not by anti-HSP70, whereas the lower (faster) migrating band was recognized by anti-HSP70 but not by anti-HSC70 (data not shown). The results clearly indicated that both HSC70 and HSP70 are associated with MOK.
Association of Other Molecular Chaperones with MOK-Next, the same set of affinity-purified samples as in Fig. 7 (A-D) was revealed by Western blotting with antibody specific for HSP60. The result clearly indicated that endogenous HSP60 was co-purified with MOK (Fig. 7E, lane 4). Again, HSP60 could not be isolated by control affinity purifications (Fig. 7E, lanes 1-3) showing the specificity of the association. FKBP52 (FKBP59) and Hop (STI1) have been observed often in complexes of HSP90-steroid hormone receptors, and both proteins possess molecular masses of ϳ60 kDa; however, we could not detect FKBP52 or Hop as MOK-associated proteins (data not shown).
Although we detected the above molecular chaperones in the MOK-associated protein fraction, a possibility exists that molecular chaperones account for only a minor part of the isolated protein bands seen in Fig. 7A. However, this possibility was excluded as follows. As described under "Experimental Procedures," we purified all of the MOK-associated proteins (Fig. 7A, bands 1-4) by affinity purification and MonoQ ion exchange column chromatography, and then finally the protein bands were excised after SDS-PAGE. No other protein with the same molecular masses was observed during the purification steps (data not shown). The purified (more than 95% pure) proteins were recognized by antibodies specific for HSP90, HSP70, HSC70, Cdc37, and HSP60, respectively (data not shown), indicating that the MOK-associated proteins observed in Fig. 7A were a set of molecular chaperones themselves and that no other protein was hindered at the same molecular masses.
Cdc37, but Not HSP90, Was Phosphorylated by MOK-As described above, MOK was found to be associated with molecular chaperones; thus a component(s) of the complexes could be phosphorylated by MOK protein kinase. We immunoprecipitated MOK from transfected cells and examined whether MOK phosphorylates purified HSP90 and/or Cdc37 in vitro. Cdc37 and HSP90 were expressed as glutathione S-transferase tag fusion proteins in E. coli and purified by tag-dependent affinity  (14), myelin basic protein is a good substrate in vitro for immunoprecipitated MOK (Fig. 8A, lane 4). Cdc37 was also significantly phosphorylated by immunoprecipitated MOK (Fig. 8B,  lane 4). On the other hand, phosphorylation of HSP90 was hardly observed in the same condition (Fig. 8C). We also checked MOK-dependent phosphorylation of purified HSP90 from several mammalian sources (COS7 cells, L5178Y cells, mouse liver, and porcine brain), but none of them was phosphorylated significantly by MOK (data not shown). These results indicated that Cdc37, but not HSP90, is an in vitro substrate of MOK. DISCUSSION In this report, we described identification of several cellular MOK-binding proteins by using a mammalian expression system and co-immunoprecipitation experiments. The data have clearly shown that the majority of the MOK-associated proteins is a set of molecular chaperones including HSP90, Cdc37, HSP70, HSC70, and HSP60, although we do not exclude a possibility of other MOK-associated proteins in cells. Previously, associations of HSP90 and Cdc37 with certain important protein kinases such as pp60 v-src (27,28), Raf1 (29), and Cdk4 (30 -32) have been described. Among a wide variety of proteins that are associated with HSP90, only certain protein kinases are associated with Cdc37 along with HSP90 (40). Thus the combination of Cdc37 and HSP90 can be considered as a kinase-specific chaperone complex (30,35,42). The results obtained here also support this idea.
It should be noted that not all of the protein kinases are associated with HSP90-Cdc37 complexes, i.e. there is specificity for the binding. In fact, our results demonstrated that conventional MAP kinases (ERK, p38, and JNK/SAPK) did not bind HSP90 despite the high homologies in the catalytic domains of conventional MAP kinases and MOK. MOK is most intimately related to MAK and MRK, two closely related protein kinases that show similarity to both the MAP kinases and the cyclin-dependent protein kinases (9), and we observed here that MAK and MRK also bind HSP90. It was reported that HSP90-Cdc37 binds Cdk4 and Cdk6 but not Cdc2, Cdk2, Cdk3, Cdk5, or CAK (31, 32, 55). Taken together, there seems to be a mechanism of HSP90-Cdc37 that discriminates subtle differ-ences between related kinases.
Our experiments with MOK deletion mutants clearly demonstrated that HSP90 requires the protein kinase catalytic domain of MOK for binding. It is also the case for other kinases such as pp60 v-src (56), Raf (29), Plk (57), and CK2 (34). Thus the HSP90 binding specificity seems to be determined by structures of kinase domains, and it will be interesting to identify a region that is responsible for the HSP90 binding specificity. However, a noncatalytic C-terminal part was reported to be important for HSP90 binding of Akt protein kinase (58); thus the rule may not be definitely generalized.
What is the physiological meaning of the association of these molecular chaperones with MOK? Previously we observed that MOK is activated by tumor promoter stimulation (14). In addition, most of the HSP90-associated proteins are classified into signal switching molecules (37). These facts imply that HSP90 is required for an activable structure of MOK that is essential for its function. In this article, we found that HSP90 is essential for the stability of MOK in cells. In other words, MOK was not able to stably exist in the absence of HSP90 and then rapidly degraded. Protein folding and protein stability are intimately interrelated, and once a protein fails to be correctly folded, the protein should be degraded. We have shown that MOK degradation in the absence of HSP90 function was carried out by the proteasome system and that MOK protein accumulated in the insoluble fraction if both HSP90 and the proteasome system were inactivated. Thus the balance between folding, degradation, and solubility of MOK should be controlled by HSP90-Cdc37 molecular chaperones and proteasome systems. This kind of link between protein folding and degradation is not restricted only for MOK but for many other HSP90 target proteins (59,60). The pharmacological effect of the anti-cancer drug geldanamycin was previously believed to be ascribed to the specific inhibition of growth-related proteintyrosine kinases. However, knowing that inhibition of HSP90 function leads to degradation of certain other kinases also as shown here, the effect of geldanamycin treatment should be more complicated.
HSP90 often associates with target proteins not alone but with several other molecular chaperones that are called "cochaperones" (61) or "HSP90 partner proteins" (38). In fact, we found here that HSP70 and HSC70 as well as Cdc37 are also included in the MOK-HSP90 complexes. For most of the steroid hormone receptors, high molecular mass FK506-binding immunophilin FKBP52 is associated with the receptor along with HSP90. Other co-chaperones such as p23, Hip, Hop (STI1), and HSP40 (eukaryotic DnaJ) are also observed in certain HSP90 target complexes. However, we could not detect FKBP52, Hop, GRP94, or HSP27 in the affinity-purified MOK-associated protein fraction (data not shown). Thus it seems that specificity exists in the selection of limited members of molecular chaperones for particular targets. Our previous immunofluorescent study revealed that MOK localizes mostly in the cytoplasm when expressed exogenously (14). On the other hand, most of mammalian HSP60 resides within mitochondria, and we do not know whether MOK associates with HSP60 within mitochondria or in the cytoplasm. Mitochondria might be ruptured during the cell lysis procedure; thus we cannot exclude the possibility that the HSP60 association with MOK may be observed only after cell lysis. In addition, by our co-immunoprecipitation experiments it is not clear whether MOK is associated with all of these molecular chaperones cooperatively in the same complexes or each molecular chaperone associates with MOK independently.
For efficient and specific signal transmission of the MAP kinase cascades, tethering proteins (also called molecular glues or molecular scaffolds) are often observed and thought to be important that associate with each of the members of the cascades (25,26). Thus a certain HSP90-associated protein kinase may be an upstream activator for MOK. Associated proteins can be also direct substrates for the kinase, and in fact molecular chaperones including HSP90, Cdc37, and FKBP52 have been reported to be phosphorylated by their associated protein kinases (33,35,50). Although Cdc37 was found to be phosphorylated by MOK in vitro, it is not obvious whether the phosphorylation of Cdc37 transmits a downstream signal from activated MOK.