Cdc37 Interacts with the Glycine-Rich Loop of Hsp90 Client Kinases

Plasmid construction.

To produce the recombinant protein Cdc37(181-378) fused to glutathione S-transferase (GST) [GST-Cdc37(181-378)] in Escherichia coli, the BamHI-EcoRI fragment from the construct for Cdc37(181-378) (47) was inserted into a pGEX6P1 plasmid (Amersham Biosciences) that had been cut with both BamHI and EcoRI to yield pGEX6P1-Cdc37(181-378). The preparation of FLAG-tagged Cdc37, Cdc37(1-180), and Cdc37(181-378) [FLAG-Cdc37(181-378), FLAG-Cdc37(1-180), and FLAG-Cdc37(181-378), respectively] has been described previously (47). The plasmids used in this study, pcDL-SRα-456, pcDL-SRα-Myc-GST, SRα-Myc, and pcDNA3FLAG1, and mouse PKA cDNA were supplied by E. Nishida (Kyoto University, Japan). Oligonucleotides containing ATG (for Met), followed by three restriction sites in the order of BglII, EcoRI, and NotI, were introduced into the pcDL-SRα-456 plasmid to obtain pSRα-MCS.

The coding region of enhanced green fluorescent protein (EGFP) was amplified by PCR using a pEGFP-N3 plasmid (Clontech) as a template, with the addition of the oligonucleotide encoding a Myc epitope tag at the 3′ end; the resultant EGFP-Myc fragment was inserted at the NotI site of the pSRα-MCS plasmid to yield pSRα-EGFP-Myc. To obtain constructs for EGFP-fused peptides, oligonucleotides corresponding to peptide sequences were inserted into a pSRα-EGFP-Myc plasmid that had been cut with both BglII and EcoRI. The Myc-tagged and Myc-GST-fused Raf-1 kinase domains have been described previously (47). Constructs for the N lobes of Raf-1 (residues 349 to 423), Cdk2 (residues 1 to 82), and Cdk4 (residues 1 to 95) fused to Myc-tagged GST were produced according to the procedure for the N-terminal portion of Raf-1 (47). To produce a construct for Myc-tagged Cdk2 according to the procedure for Myc-tagged Cdk4 (47), the full-length cDNA of human Cdk2 was synthesized by PCR with reverse transcription using mRNA isolated from HeLa cells, with the addition of a BamHI site at the 5′ end and an EcoRI site following a stop codon (TGA) at the 3′ end.

To obtain cDNAs for chimeric mutants (Cdk2/4 and Cdk4/2), the complementary PCR products for the N-terminal portion of Cdk2 and the C-terminal portion of Cdk4 and those for the reciprocal pair were amplified by PCR. The resulting cDNAs were inserted into a pcDNA3Myc1 plasmid that had been cut with both BamHI and EcoRI; the pcDNA3Myc1 plasmid was made by inserting PCR products containing an oligonucleotide for two copies of Myc epitope flanked by BglII and BamHI-EcoRI sites into pcDNA3 (Invitrogen) that had been double digested with BamHI and EcoRI. Myc-tagged GST fused to the N-terminal portion of the Raf-1 kinase domain (subdomains I to IV) has been described previously (47). The C-terminal portion of the Raf-1 kinase domain (subdomains V to XI) was produced by PCR as described previously (47), and its BamHI/EcoRI fragment was inserted into an SRα-Myc plasmid digested with BglII and EcoRI to produce the Myc-tagged C-terminal portion of the Raf-1 kinase domain. The BamHI fragment of PKA cDNA was inserted into the BamHI site of a pcDNA3Myc1 or pcDNA3FLAG1 plasmid. Mutagenesis of the Raf-1 kinase domain (ASA), its N lobe (ASA), and PKA (TA) was carried out using a QuikChange site-directed mutagenesis kit (Stratagene). All constructs were confirmed by DNA sequencing.

Purification of bacterially expressed recombinant proteins.

To obtain the recombinant proteins GST and GST-Cdc37(181-378), the plasmids pGEX6P1 and pGEX6P1-Cdc37(181-378), respectively, were introduced into E. coli strain BL21(DE3)pLysS (Stratagene). Cultures were induced with 0.4 mM isopropyl-1-β-d-galactopyranoside after reaching an A₆₀₀ of 0.6 and were further incubated for 1 h at 30°C. Cells were harvested by centrifugation and washed with ice-cold phosphate-buffered saline (137 mM NaCl, 2.68 mM KCl, 8.1 mM Na₂HPO₄, 1.47 mM KH₂PO₄, pH 7.4). The cells were resuspended in phosphate-buffered saline containing 1 mM phenylmethylsulfonyl fluoride and disrupted by sonication; insoluble material was then removed by centrifugation. Recombinant proteins were purified using glutathione Sepharose 4 (Amersham Biosciences) according to the manufacturer's protocol and dialyzed against a buffer solution containing 20 mM Tris-HCl, pH 7.5, 2 mM EGTA, 1 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride.

Phage display technique.

A Ph.D.-12 phage display peptide library kit purchased from New England Biolabs was used for affinity panning with GST-Cdc37(181-378). Approximately 1.5 × 10¹¹ phage particles in Tris-buffered saline (TBS; 50 mM Tris-HCl, pH 7.5, 150 mM NaCl) were incubated with GST preadsorbed to glutathione Sepharose 4B for 2 h at 4°C to remove phage particles that were nonspecifically bound to GST, and after clarification by centrifugation, the supernatants were further incubated with GST-Cdc37(181-378)-preadsorbed beads for 2 h. The beads were washed five times with TBS plus 0.1% (vol/vol) Tween 20, and the bound phages were eluted with elution buffer containing 50 mM Tris-HCl, pH 8.0, 2 mM dithiothreitol, and 10 mM reduced glutathione. Recovered phage particles were amplified according to the manufacturer's protocol and affinity panned 3 more times as described above, except that the beads were washed 10 times with TBS plus 0.5% (vol/vol) Tween 20. After the third and fourth rounds of panning, randomly selected phage clones were subjected to DNA sequencing.

Protein identification by LC-MS/MS analysis.

FLAG-Cdc37(181-378)-associated complexes were digested with Achromobacter protease I, and the resulting peptides were subjected to analysis by a nanoscale LC-MS/MS system (29). The criteria for match acceptance that we used have been reported previously (29).

Cell culture and transfection.

COS7 cells were cultured at 37°C in Dulbecco's modified Eagle's medium containing 10% (vol/vol) fetal bovine serum. Cells were transfected with Lipofectamine Plus (Invitrogen), according to the manufacturer's protocol.

Immmunoprecipitation and immunoblotting.

Cell lysis and immunoprecipitation were carried out essentially as described previously (47). The proteins obtained were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and analyzed by immunoblotting. Unless otherwise indicated, the anti-Hsp90 antibody used was provided by Y. Miyata (Kyoto University). In the experiment whose results are shown in Fig. 5F, anti-Hsp90 antibody (SPA-830) purchased from Stressgen was used. Anti-Myc (A-14) and anti-Raf-1 (C12) polyclonal antibodies were obtained from Santa Cruz Biotechnology, anti-FLAG polyclonal antibody was obtained from Sigma, anti-Cdc37 polyclonal antibody was obtained from Neomarkers, and PKA C phospho (Thr197) polyclonal antibody was obtained from Cell Signaling.

In vitro binding experiments.

FLAG-Cdc37(181-378) and Myc-GST fused to either the kinase domain of Raf-1 or its N lobe were expressed in and purified from COS7 cells as described previously (47). The FLAG-Cdc37(181-378)-adsorbed beads were incubated with 100 μg/ml of FLAG peptide (Sigma) in TBS for 1 h at 4°C to release FLAG-Cdc37(181-378) proteins. The Myc-GST fusion proteins (i.e., the Raf-1 kinase domain and its N lobe) adsorbed to glutathione Sepharose 4B (Amersham Bioscience) were washed with TBS, and then the resin was incubated with the purified FLAG-Cdc37(181-378) proteins in TBS for 2 h at 4°C. The wild-type and ASA mutant forms of Raf-1 peptides (residues 352 to 368), to which six residues of histidine were C terminally added, were purchased from Operon Biotechnologies; briefly, the peptides were synthesized by standard solid phase methods using 9-fluorenylmethoxy carbonyl chemistry and purified by reversed-phase high-performance liquid chromatography. These His-tagged peptides (1 μg) were bound to 10 μl of Ni-NTA agarose (QIAGEN) for 30 min at room temperature, and then the resin was washed with TBS. The recombinant GST-Cdc37(181-378) protein (100 ng) was incubated with the peptide-bound beads in TBS for 2 h at 4°C, and the beads were washed with lysis buffer (20 mM HEPES, pH 7.5, 1 mM MgCl₂, 1 mM EGTA, 150 mM NaCl, 1% [vol/vol] Nonidet P-40, and 1% [vol/vol] proteinase inhibitor cocktail [Sigma]).

In vitro translation using rabbit reticulocyte lysate.

Using a PKA-inserted pcDNA3FLAG1 plasmid as a template, we amplified by PCR DNA fragments containing the T7 promoter and the PKA-coding regions either with or without a stop codon; the resultant PCR products were used as a template to synthesize PKA mRNAs in the T7 RiboMAX express large-scale RNA production system (Promega) according to the manufacturer's protocol. After purification of the synthesized mRNAs with an RNeasy minikit (QIAGEN), in vitro translation reactions were carried out by using the rabbit reticulocyte lysate system (Promega) for 20 min at 30°C. After protein synthesis, PKA released from the ribosome or the ribosome-bound PKA (corresponding to the DNA construct with or without a stop codon, respectively) was incubated for a further 30 min at 30°C in the presence of 0.2 mg/ml of RNase A (Nippon Gene) to terminate translation or 1 mM puromycin (Sigma) to release translation products from the ribosome, respectively.

Phosphatase treatment.

Myc-tagged PKA was immunoprecipitated as described above. The immunoprecipitates were washed with and resuspended in lambda protein phosphatase buffer (New England Biolabs) and then incubated with or without 100 U of lambda protein phosphatase (New England Biolabs) for 45 min at 30°C.

Identification of a Cdc37-interacting motif.

Although we have recently identified a client-binding site of Cdc37 that is responsible for its association with Hsp90 client kinases (47), the region or regions in protein kinases that contribute to interaction with Cdc37 remain uncertain. To determine the Cdc37-interacting region in Hsp90 client kinases, we adopted two different experimental approaches (phage display and LC-MS/MS analysis) using a Cdc37(181-378) protein containing the 20-residue client-binding site of Cdc37 as bait; our previous work showed that Cdc37(181-378) binds not only to Hsp90 client kinases (Raf-1, Akt1, Aurora B, and Cdk4) but also to nonclient kinases (Cdc2 and Cdk2) (47).

We examined whether Hsp90 is present in the complex that is formed between Cdc37(181-378) and the kinase domain of Raf-1, albeit Cdc37(181-378) loses the residues involved in its interaction with Hsp90 (i.e., M¹⁶⁴LRR¹⁶⁷ and Asp¹⁷⁰) (40). We found that Hsp90 was detected scarcely in the immunoprecipitates of FLAG-tagged Cdc37(181-378), although the Raf-1 kinase domain was efficiently coprecipitated (Fig. 1, lane 4); by contrast, full-length Cdc37 immunoprecipitated Hsp90 together with the Raf-1 kinase domain (Fig. 1, lane 3). In a reciprocal experiment of immunoprecipitation with anti-Myc antibodies (to pull down the Myc-tagged Raf-1 kinase domain), Hsp90 was coprecipitated even in the case of Cdc37(181-378) (Fig. 1, lane 6); however, it is conceivable that this Hsp90 protein interacts directly with the Raf-1 kinase domain that does not associate with Cdc37(181-378). Therefore, we conclude that protein kinases are bound to Cdc37(181-378) directly and not indirectly via Hsp90.

First, we screened a commercially obtained library of random 12-mer peptides fused to a minor coat protein of M13 phage by affinity panning with GST-Cdc37(181-378). After three and four rounds of panning, 11 and 12 clones, respectively, were randomly chosen for DNA sequencing (Table 1). Notably, although peptides of a single sequence or a few specific sequences were not enriched, peptides possessing the sequence GXF (where X is any amino acid) were in the majority (8/11 and 9/12 for the third and fourth rounds of panning, respectively); in addition, most of these peptides had glycine following the phenylalanine (i.e., GXFG).

Next, we attempted to identify proteins that physically interact with Cdc37(181-378) in cells. FLAG-Cdc37(181-378) was expressed in HEK293 cells and immunoprecipitated with anti-FLAG antibody; the immunoprecipitates were digested and then subjected to analysis by a nanoscale LC-MS/MS system (29). A total of 33 proteins met the match acceptance criteria (29) and were specifically identified as interacting with FLAG-Cdc37(181-378) (Table 2). Although two protein kinases (MEKK4 and STE20-like kinase MST4) were identified, it is more intriguing that the sequences GXF and GXY occur in 28 proteins at least once and twice or more in most cases; in addition, some of the sequences are followed by glycine (i.e., GXFG and GXYG).

The above-described two experiments afforded us mutually concordant results, that is, Cdc37(181-378) preferably binds a unique sequence, GXF(G). Out of 424 entries of human protein kinases (24), 267 kinases (63%) contain either GXFG (203 kinases) or GXF (64 kinases), and 86 kinases (20%) contain either GXYG (68 kinases) or GXY (18 kinases). Although the sequence GXY(G) was obtained from the pull-down/LC-MS/MS analysis, it was not detected by the phage display, implying that Cdc37(181-378) associates more strongly with GXF(G) than with GXY(G). Importantly, these sequences are located in the so-called glycine-rich loop (GXGXXG), which reportedly helps to anchor the nontransferable phosphates of ATP (12, 45). Furthermore, it is notable that the glycine-rich motif commonly appears in protein kinases regardless of their dependency on Hsp90/Cdc37 (12, 24, 45); thus, even though the glycine-rich motif physically interacts with Cdc37, it cannot be the sole determinant for stable binding of protein kinases to Cdc37.

The glycine-rich loop interacts with Cdc37.

The above observations strongly suggest that the glycine-rich loop of protein kinases interacts with Cdc37(181-378). To address this possibility, we examined whether Cdc37(181-378) binds to a Raf-1 peptide containing the Raf-1 glycine-rich loop (Fig. 2A, wt). Immunoprecipitation data indicated that Cdc37(181-378) bound to the Raf-1 peptide conjugated to EGFP-Myc but not to EGFP-Myc alone in COS7 cells (Fig. 2B, compare lanes 5 and 8 with lanes 4 and 7, respectively).

To verify that this interaction was specific and physiologically relevant, we introduced a mutation into Raf-1. A previous study has revealed the significance of the second and third glycines (i.e., Gly³⁵⁸ and Gly³⁶¹, respectively) of the glycine-rich motif of Raf-1 (G³⁵⁶SGSFG³⁶¹) in the interaction of Raf-1 with Cdc37 (58), and furthermore, our above finding suggests that GXFG is preferred by Cdc37; however, when we estimated the phi/psi angles for selected residues (Gly³⁵⁸, Phe³⁶⁰, and Gly³⁶¹) based on the three-dimensional structure of B-Raf (PDB entry 1UWH [48]), it was suggested that the last glycine (Gly³⁶¹) is not allowed to be replaced with alanine, although Gly³⁵⁸ and Phe³⁶⁰ are allowed. Therefore, the glycine-rich motif of Raf-1 was changed to GSASAG to abolish its Cdc37-binding ability (substitutions are underlined). Clearly, the mutant form of the Raf-1 kinase domain did not associate with Cdc37 in COS7 cells (Fig. 2C, ASA), whereas the wild-type Raf-1 kinase domain bound to Cdc37 (Fig. 2C, wt). In addition, the data clearly showed that Cdc37(181-378) bound the Raf-1 kinase domain but did not bind its mutant form (Fig. 2D, compare wt with ASA).

Next, we performed in vitro binding experiments to corroborate the data shown in Fig. 2D; to this end, FLAG-Cdc37(181-378), the Myc-GST-fused Raf-1 kinase domain, and the Myc-GST-fused N lobe of Raf-1 (residues 349 to 423) (45) were expressed in COS7 cells and purified from the cell lysates, as we were not able to bacterially express either the Raf-1 kinase domain or its N lobe in a soluble form (data not shown). A preparation of FLAG-Cdc37(181-378) was highly pure (Fig. 2E, lane 1), and no Hsp90 protein was detected (Fig. 2E, lane 8). Three unknown proteins were invariably present in all preparations of Myc-GST fusion proteins (Fig. 2E, lanes 2 to 6); both wild-type and mutant forms of the N lobe of Raf-1 were substantially pure except for these proteins (Fig. 2E, lanes 3 and 4) and, notably, did not contain Hsp90 (Fig. 2E, lanes 10 and 11). Cdc37(181-378) was pulled down efficiently by the wild-type N lobe of Raf-1, but its ASA mutant lost such an ability (Fig. 2F); thus, the N lobes of Raf-1 and Cdc37(181-378) on their own are sufficient for a stable interaction. As for the kinase domain of Raf-1, we obtained similar results (Fig. 2F): even though, besides its three contaminating bands, a small amount of Hsp90 existed in preparations of both wild-type and mutant forms of the kinase domain (Fig. 2E, lanes 5, 6, 12, and 13), it seems that a physical interaction between Cdc37(181-378) and the kinase domain of Raf-1, which is mediated via the glycine-rich loop and is therefore abrogated by the alanine mutation, was unaffected by this coexisting Hsp90 protein (Fig. 2F).

Finally, the same mutation was introduced into the Raf-1 peptide conjugated to EGFP-Myc (Fig. 2A, ASA); the mutant Raf-1 peptide was profoundly compromised in its ability to associate with Cdc37(181-378) (Fig. 2B, compare lanes 5 and 8 with lanes 6 and 9, respectively). Furthermore, we found that the GST-Cdc37(181-378) recombinant protein (which was used in the phage display) was bound to the His-tagged peptide of the glycine-rich loop of Raf-1 (residues 352 to 368) (Fig. 2A) but that its binding to the ASA mutant peptide was significantly reduced (Fig. 2G). Thus, these data convincingly demonstrate that the glycine-rich loop of Raf-1 is recognized and directly bound by Cdc37(181-378), as predicted from the results obtained by phage display and LC-MS/MS.

Taking these results altogether, we conclude that the glycine-rich loop of protein kinases as well as the client-binding site of Cdc37 (as has been revealed in our previous work [47]) is necessary for the interaction between Cdc37 and protein kinases.

A protein kinase region other than the glycine-rich loop is crucially required for binding to Cdc37.

Consistent with the in vitro binding experiments, Cdc37(181-378) was found to bind to the N lobe of Raf-1 fused to Myc-tagged GST in COS7 cells (Fig. 3A, lanes 8 and 12) as well as to its entire kinase domain (Fig. 3A, lanes 6 and 10), whereas Myc-tagged GST alone did not bind to Cdc37(181-378) (data not shown). However, full-length Cdc37 did not interact with the N lobe of Raf-1 (Fig. 3A, lanes 7 and 11), although it bound to the Raf-1 kinase domain (Fig. 3A, lanes 5 and 9). In addition, the Raf-1 peptide containing the glycine-rich loop (Fig. 2A, wt) did not bind to full-length Cdc37 despite its binding to Cdc37(181-378) (Fig. 3B). The fact that full-length Cdc37 can bind neither the N lobe of Raf-1 nor the Raf-1 peptide, both of which contain the glycine-rich loop, seems contradictory to the results presented above. Taken together, these observations seem to indicate that the N lobes of protein kinases alone may be inaccessible to Cdc37, even though they encompasses the Cdc37-interacting region (i.e., the glycine-rich loop); furthermore, it is highly plausible that the residual C lobe may be critically required for the association between Cdc37 and the N lobes of protein kinases.

In this context, it is notable that Cdk2 has been shown to bind to Cdc37(181-378), even though it did not bind to full-length Cdc37 in our previous study (47); on the other hand, its homolog, Cdk4, which is also an Hsp90 client kinase, associates with both Cdc37(181-378) (47) and Cdc37 (3, 20, 44, 47). As shown in Fig. 4A, the N lobes of both Cdk2 and Cdk4 (residues 1 to 82 and 1 to 95, respectively) bound equally well to Cdc37(181-378), in line with the fact that the glycine-rich loops of Cdk2 and Cdk4 have similar sequences (GEGTYG and GVGAYG, respectively).

Thus, we considered that the C lobe of Cdk4 might enable the otherwise incapable N lobe of Cdk2 to interact with Cdc37. Consequently, we generated chimeric mutants of Cdk2 and Cdk4 by swapping their N and C lobes, because these regions are thought to be folded in similar fashions (15, 17). The boundary between the two lobes resides between β-sheet 5 (β5) and α-helix D (αD) (45); however, neither of the two chimeras generated using this boundary associated with Cdc37 (data not shown). As shown in Fig. 4B, the boundary was changed to a more N-terminal position between αC and β4, such that the C-terminal portion corresponded to the C lobe plus two N-terminal β-sheets (β4 and β5). As a result, besides Cdk4 (Fig. 4C, lanes 5 and 9), a Cdk2/4 chimera (Fig. 4C, lanes 6 and 10), in which the N-terminal portion of Cdk2 was joined to the C-terminal portion of Cdk4, was found to interact with Cdc37; by contrast, neither a Cdk4/2 chimera (i.e., the inverse of the Cdk2/4 chimera) (Fig. 4C, lanes 7 and 11) nor Cdk2 (Fig. 4C, lanes 8 and 12) associated with Cdc37. Additionally, it seems intriguing that both Cdk4 and Cdk2/4 (Fig. 4C, lanes 9 and 10 [WB: α-Hsp90]) were associated with Hsp90, whereas Cdk2 and Cdk4/2 were not (Fig. 4C, lanes 11 and 12 [WB: α-Hsp90]). By contrast, all immunoprecipitates with anti-FLAG antibody (for the immunoprecipitation of FLAG-tagged Cdc37) contained Hsp90 in comparable amounts (Fig. 4C, lanes 5 to 8 [WB: α-Hsp90]), indicating that Cdc37 directly associates with Hsp90 regardless of its interaction with client kinases.

Taken together, these results suggest that the C-terminal portion (i.e., C lobe plus β4 and β5) of Cdk4 may be responsible for binding of the N-terminal portion, containing the glycine-rich loops of both Cdk2 and Cdk4, to Cdc37, whereas the corresponding C-terminal portion of Cdk2 has no such ability.

Although molecular mechanisms are uncertain at present, we can envisage the following potential scheme: the C-terminal portions of an Hsp90 client kinase (but not a nonclient kinase) directly and/or indirectly (e.g., via interaction with Hsp90, as suggested above) induce conformational alterations in Cdc37 and/or in its N-terminal portions, which in turn enable the N-terminal portion to gain access to Cdc37, thereby facilitating a stable interaction between Cdc37 and the client kinase. To gain more insight into the underlying mechanism, we carried out the following experiment, in which the N- and C-terminal portions of the Raf-1 kinase domain (subdomains I to IV and V to XI, respectively) were coexpressed in cells as separate molecules to explore the ability of the N-terminal portion to interact with Cdc37. Although it has already been shown that the N-terminal (but not the C-terminal) portion of the Raf-1 kinase domain binds to Cdc37(181-378) (47), neither of the regions was able to interact with full-length Cdc37 (Fig. 4D), consistent with the above results (Fig. 3). Of note, the presence of the C-terminal portion of the Raf-1 kinase domain enabled the N-terminal portion to bind to full-length Cdc37 (Fig. 4D). Because the C-terminal portion of the Raf-1 kinase domain was not detected in the resulting complex, the underlying mechanism remains obscure; however, we conclude that the C-terminal portion of the Raf-1 kinase domain can elicit its efficacy toward its N-terminal portion without a molecular linkage between the two regions.

Unphosphorylated PKA binds to Cdc37.

We found that PKA did not bind to full-length Cdc37 in cells (Fig. 5A, lanes 4 and 7), indicating that it is a Cdc37-independent kinase, in keeping with a previous study (1). Reminiscent of Cdk2 (Fig. 4), however, PKA associated with Cdc37(181-378) (Fig. 5A, lanes 6 and 9); by contrast, PKA did not interact with Cdc37(1-180), which lacks the client-binding site (Fig. 5A, lanes 5 and 8).

Interestingly, PKA exhibited a doublet band in cells coexpressing Cdc37(181-378) (Fig. 5A, lane 9 [WB: α-Myc]) but not in cells coexpressing full-length Cdc37 and Cdc37(1-180) (Fig. 5A, lanes 7 and 8, respectively [WB: α-Myc]); for unknown reasons, this doublet was undetectable in whole-cell lysates (Fig. 5A, lane 3 [WB: α-Myc]). We considered that the appearance of this doublet might be caused by phosphorylation of PKA, as many protein kinases, including PKA, require phosphorylation of their activation loop to adopt an active conformation (16, 30). None of the Myc-tagged PKA proteins that coimmunoprecipitated with Cdc37(181-378), as detected by anti-Myc antibody (Fig. 5B, lane 1 [WB: α-Myc]), reacted with antibody specifically recognizing PKA phosphorylated at Thr¹⁹⁷ of its activation loop (Fig. 5B, lane 1 [WB: α-pPKA]).

The lower band of the doublet seemed to represent the unphosphorylated form of PKA (Fig. 5A, lane 9, and B, lane 2 [WB: α-Myc]), indicating that approximately half of the total amount of PKA was not prevented from being phosphorylated even in the Cdc37(181-378)-expressing cells (Fig. 5B, lane 2). PKA expressed in cells without coexpression of Cdc37(181-378) was completely phosphorylated; therefore, only the upper band was detected (Fig. 5C, lane 1). By contrast, PKA coexpressed and coimmunoprecipitated with Cdc37(181-378) was unphosphorylated; therefore, only the lower band was visible (Fig. 5C, lane 4). Phosphatase treatment affected only the former type of PKA (Fig. 5C, lanes 2 and 3) and not Cdc37(181-378)-associated PKA (Fig. 5C, lanes 5 and 6) and resulted in the appearance of the lower unphosphorylated (i.e., dephosphorylated) band (Fig. 5C, lane 3). Taken together, these observations imply that Cdc37(181-378) prevents phosphorylation of PKA by its binding, leading us to assume that even full-length Cdc37 may be able to bind the unphosphorylated form of PKA. To address this issue, Thr¹⁹⁷ in the activation loop, which must be phosphorylated for activation of PKA (16, 30), was altered to alanine. As expected, the TA mutant form of PKA was strongly bound to Cdc37; however, the wild-type form of PKA was not (Fig. 5D, compare wt with TA). This finding suggests that during the biogenesis of PKA, Cdc37 may bind to PKA in an unphosphorylated state and, on phosphorylation at Thr¹⁹⁷, may subsequently dissociate from this kinase.

Consequently, we analyzed whether nascent polypeptide chains of PKA associate with Cdc37 using the reticulocyte lysate system; based on the fact that translation products of mRNAs lacking a stop codon remain ribosome bound as peptidyl-tRNAs (8), PKA was translated using mRNAs either with or without a stop codon. As shown in Fig. 5E, PKA proteins synthesized from mRNA containing a stop codon were phosphorylated, whereas those synthesized from mRNA lacking a stop codon were completely unphosphorylated, implying that phosphorylation occurred post- and not cotranslationally; therefore, completion of translation may be a prerequisite for phosphorylation of PKA at Thr¹⁹⁷. To address this issue, translation of PKA (containing a stop codon) was halted by RNase treatment and chased for a further 30 min to examine whether phosphorylation of PKA proteins increased posttranslationally. As expected, phosphorylated PKA significantly increased after 30 min of incubation (Fig. 5F, α-pPKA). Thus, because RNase destroys mRNAs and prevents further initiation/progression of translation, the increase in phosphorylation of PKA must represent posttranslational events. In accord with this idea, ribosome-bound PKA (lacking a stop codon), which was unphosphorylated as shown in Fig. 5E, was efficiently phosphorylated during a chase period after release from the ribosome by puromycin treatment (Fig. 5G). Taken together, these data show that PKA phosphorylation at Thr¹⁹⁷ is posttranslational and occurs after PKA leaves the ribosome.

Importantly, it was evident that apparently concurrent with PKA phosphorylation, Hsp90 and Cdc37, which associated with PKA before RNase treatment (Fig. 5F, RNase at 0 min), dissociated from PKA (Fig. 5F, RNase at 30 min). Thus, phosphorylation of PKA may induce a conformational alteration that eventually gives rise to the dissociation of Hsp90 and Cdc37, which are probably no longer necessary to maintain PKA folding; alternatively, dissociation of Hsp90 and Cdc37 from PKA may be needed prior to PKA phosphorylation, the mechanism of which is currently unknown. Although PKA has been regarded as a nonclient kinase thus far, our data suggest that its association with Cdc37 (and Hsp90) is required for correct folding during biosynthesis. As such, the standard discrimination between client and nonclient kinases may not be that strict and remains to be reevaluated.