Threonine 22 phosphorylation attenuates Hsp90 interaction with co-chaperones and affects its chaperone activity

SUMMARY

Heat Shock Protein 90 (Hsp90) is an essential molecular chaperone whose activity is regulated not only by co-chaperones but also by distinct post-translational modifications. We report here that casein kinase 2 phosphorylates a conserved threonine residue (T22) in α-helix 1 of the yeast Hsp90 N-domain both in vitro and in vivo. This α-helix participates in a hydrophobic interaction with the catalytic loop in Hsp90's middle domain, helping to stabilize the chaperone's ATPase competent state. Phospho-mimetic mutation of this residue alters Hsp90 ATPase activity and chaperone function, and impacts interaction with the co-chaperones Aha1 and Cdc37. Over-expression of Aha1 stimulates the ATPase activity, restores co-chaperone interactions, and compensates for the functional defects of these Hsp90 mutants.

INTRODUCTION

Heat shock protein 90 (Hsp90) is an essential molecular chaperone in eukaryotic cells, and its activity depends on an ATP binding/hydrolysis cycle (Mayer et al., 2009; Wandinger et al., 2008). Its target proteins are referred to as “clients” and are important mediators of normal cellular function that, when deregulated, also support cancer cell proliferation, survival, and metastasis (Trepel et al., 2010). Several pan-Hsp90 inhibitors block ATP binding and are currently undergoing clinical evaluation in cancer patients (Kim et al., 2009), but a more detailed understanding of Hsp90 regulation in eukaryotic cells may provide additional and perhaps more effective therapeutic strategies (Mollapour et al., 2010; Taipale et al., 2010).

Based on numerous structural studies, Hsp90 is dimeric and each protomer can be divided into three domains: an N-terminal ATP binding domain, a middle domain that serves as a site of interaction for both clients and some co-chaperone proteins, and a C-terminal dynamic dimerization domain (Ratzke et al., 2010; Wandinger et al., 2008). The current model for conformational coupling of the ATPase cycle suggests that N-terminal dimerization is an initiating step in a series of conformational changes that ultimately result in ATPase competence (Pearl and Prodromou, 2006). These steps include inter- and intra-protomer interactions of a cluster of hydrophobic and polar amino acids derived from N- and M-domains (Ali et al., 2006; Cunningham et al., 2008). The N-domain residues include T22, V23, and Y24 from α-helix 1 in the N-domain. While recent work suggests that Hsp90 can sample many of these conformational states in the absence of nucleotides (Bron et al., 2008), ATP binding is thought to shift the conformational equilibrium by reducing the energy barrier between open and N-domain dimerized states (Hessling et al., 2009; Mickler et al., 2009; Southworth and Agard, 2008).

We recently reported that Y24 is phosphorylated by Swe1^Wee1 in yeast and human cells, and that this serves to modulate Hsp90 ATPase activity and chaperone function (Mollapour et al., 2010). More than a decade ago, in a random mutagenesis study of Hsp90, Lindquist's group identified the adjacent amino acid T22 as an important determinant of Hsp90 function (Nathan and Lindquist, 1995). Mutation of this residue to isoleucine caused a significant defect in Hsp90's ability to chaperone the client proteins glucocorticoid receptor and v-Src. More recently, the ATPase activity of Hsp90-T22I was reported to be severely deregulated compared to the wild type protein (Hawle et al., 2006; Prodromou et al., 2000). Taken together, these data emphasize the importance of α-helix 1 for Hsp90 function. In the current study, we show that T22 is phosphorylated by casein kinase 2 (CK2) in vitro and in vivo, with consequences for ATPase activity, co-chaperone associations, and Hsp90 function. Surprisingly, over-expression of Aha1 helped normalize the chaperone activity of Hsp90 T22 mutants. Our data suggest that multiple phosphorylations within α-helix 1 can directly impact the Hsp90 conformational cycle and fine tune Hsp90 function.

RESULTS

CK2 phosphorylates a conserved threonine residue in the yHsp90 N-domain

CK2 has been reported to phosphorylate two serine residues in the highly charged region linking N- and M-domains of human Hsp90 (Lees-Miller and Anderson, 1989; Miyata and Yahara, 1992). Therefore, we determined whether purified GST-CK2 (human) can phosphorylate His6-tagged yeast Hsp82 (yHsp90) in vitro. Using a pan anti-phospho-threonine antibody, we readily detected CK2-mediated Hsp90 threonine phosphorylation (Figure 1A, top). Next, after carrying out the kinase reaction we treated yHsp90-His₆ with PreScission protease. This protease cleaves a PreScission site that we inserted between the N-domain and the charged linker (see EXPERIMENTAL PROCEDURES), allowing us to unambiguously detect CK2-mediated threonine phosphorylation in the N-domain of yHsp90 (Figure 1A, bottom). The inserted PreScission site did not interfere with yHsp90 function (Figure S1). Since mutation of T22 to isoleucine was identified previously to impact Hsp90 function in yeast (Nathan and Lindquist, 1995), we determined whether T22 was phosphorylated by CK2. We found that mutation of T22 to alanine, while not impacting CK2-mediated phosphorylation of intact Hsp90, abrogated threonine phosphorylation of the Hsp90 N-domain in vitro (Figure 1A, compare “wt” to “T22A” lanes).

Figure 1. CK2 phosphorylates yHsp90-T22 in vitro and in vivo, (See Figure S1).

A) In vitro CK2-dependent threonine phosphorylation of full-length and N-domain WT and T22A yHsp90.

B) Serine phosphorylation of p50-Cdc37-GFP chimera in yeast expressing either WT or ck2-ts mutant. Phosphorylation of S13 was detected by anti-phospho-Ser13 antibody and p50-Cdc37-GFP was visualized with anti-GFP antibody (See Figure S1).

C) Schematic diagram of temperature-dependent CK2 enzyme inactivation in ck2-ts yeast mutant.

D) CK2 phosphorylates yHsp90-T22 in vivo. N-domain WT and T22A yHsp90 threonine phosphorylation was detected in WT (CK2) and ck2-ts yeast at 25°C and 34°C.

Next, we determined whether CK2 phosphorylated T22 in vivo. Since CK2 is essential in yeast, we could not delete this gene. Instead, we transformed both wild type yHsp90-His₆ and yHsp90-T22A-His₆ (both under GAL1 promoter and containing PreScission protease sites) into either a temperature-sensitive CK2 mutant yeast strain, cka2-13 (ck2-ts; strain YSB49) or a comparable CK2 wild type strain (strain YSB48). These strains were a kind gift of Dr. Claiborne V. C. Glover (The University of Georgia). We validated the temperature sensitivity of CK2 in ck2-ts cells by examining CK2-mediated phosphorylation of the serine 13 residue in p50^Cdc37 (Bandhakavi et al., 2003; Miyata and Nishida, 2007; Vaughan et al., 2008). We transformed wild type CK2 or ck2-ts cells with C-terminally GFP-tagged p50^Cdc37. We detected p50^Cdc37 expression with anti-GFP antibody and phosphorylation of S13 with anti-phospho-S13 antibody (Figure 1B). Growing the ck2-ts cells at 34°C for 2 hrs fully abrogated phosphorylation of p50^Cdc37 S13, confirming efficient inactivation of CK2 at this temperature.

As shown schematically in Figure 1C, cells containing either wild type yHsp90-His₆ or yHsp90-T22A-His₆ were initially grown to mid-log on raffinose media at 25°C. Cells were then shifted to 34°C for 2 hrs (to inactivate the temperature sensitive CK2 mutant), and yHsp90-His₆ and yHsp90-T22A-His₆ were then expressed for a further 2 hrs by addition of galactose to the media (see also Figure S1). N-domain threonine phosphorylation of wild type yHsp90 was unaffected in wild type CK2 cells grown at either temperature; likewise, N-domain threonine phosphorylation of yHsp90-T22A was equally reduced at both temperatures (Figure 1D). We observed a similar result when yHsp90-T22A was expressed in ck2-ts cells at either 25°C or 34°C. However, N-domain threonine phosphorylation of wild type yHsp90 was reduced to the same level as that of yHsp90-T22A when ck2-ts cells were grown at elevated temperature (Figure 1D). These data confirm that, while other threonine residues in the N-domain of yHsp90 are phosphorylated in vivo by kinases other than CK2, T22 remains a major CK2-dependent N-domain threonine phosphorylation site in vivo.

We explored the possibility that Ppt1 phosphatase can dephosphorylate T22, since this co-chaperone interacts with Hsp90 and has been shown to dephosphorylate yHsp90 in vitro (Wandinger et al., 2006). However, purified yeast Ppt1 was unable to dephosphorylate T22 in vitro (Figure S1). We validated the activity of Ppt1 in this assay by examining its ability to promote N-domain yHsp90 serine dephosphorylation (mediated by CK2). Addition of purified Ppt1 resulted in a nearly complete loss of phosphorylated serine residues in the yHsp90 N-domain (Figure S1). These data further confirm that the anti-phospho-threonine antibody used in these experiments also does not detect phospho-serine residues, nor does T22A mutation affect the ability of CK2 to promote yHsp90 N-domain serine phosphorylation.

Hsp90 T22 phospho-mutation alters ATPase activity in vitro

Hsp90 function is coupled to its ability to hydrolyze ATP (Pearl and Prodromou, 2006). Therefore, we measured the ATPase activity of the non-phsosphorylatable mutant yHsp90-T22A and of the phospho-mimetic mutant yHsp90-T22E. ATPase activity of yHsp90-T22A was similar to that of wild type yHsp90, whereas the ATPase activity of yHsp90-T22E was reduced by approximately 60 % (Figure 2A). Both mutants bound AMPPNP (a nonhydrolyzable ATP analog) with an equivalent K_d (Figure 2A), suggesting that the reduced ATPase activity of yHsp90-T22E was not due to reduced affinity for ATP (indeed, both mutants bound ATP significantly better than wild type yHsp90). Aha1 is a potent stimulator of Hsp90 ATPase activity (Panaretou et al., 2002). We determined whether Aha1 also stimulated the ATPase activity of these mutants in vitro. yAha1 stimulated the ATPase activity of both mutants to the same degree as wild type yHsp90, thus correcting the ATPase deficiency of yHsp90-T22E (Figure 2B). Both mutants form N-domain dimers in the presence of AMPPNP, although yHsp90-T22E appears less efficient in this regard (Figure 2C). This observation is consistent with this mutant's reduced ATPase activity.

Figure 2. Chaperone function of T22 phospho-mutants.

A) ATPase activity of purified WT yHsp90, T22A and T22E mutants. Error bars represent standard deviation of three independent experiments. Kd of each protein for AMPPNP is shown in inset. The ATPase activity of WT yHsp90 (K_cat = 0.94 min⁻¹ +/− 0.03) is set to 100%.

B) Aha1 (20 μM) stimulates ATPase activity of purified WT, T22A and T22E yHsp90. Error bars represent standard deviation of three independent experiments. Data are expressed as fold increase in Hsp90 ATPase activity in the presence of Aha1 compared to activity of the same Hsp90 protein in the absence of Aha1.

C) AMPPNP (PNP)-stabilized N-domain dimerization of WT, T22A, and T22E yHsp90 proteins was determined after crosslinking and polyacrylamide gel electrophoresis. N-domain dimerized species run with an apparent molecular weight of approximately 190 kDa.

Hsp90 T22 phospho-mutation affects chaperone activity

We examined the ability of Hsp90 T22 phospho-mutants to chaperone several Hsp90 clients in yeast. The v-Src oncoprotein is a well-established Hsp90 client and its expression in yeast is lethal (Nathan and Lindquist, 1995). Neither v-Src expression nor significant v-Src-mediated protein tyrosine phosphorylation was detectable in either yHsp90-T22A or yHsp90-T22E mutants, in contrast to wild type cells (Figure 3A). Consistent with these observations, neither Hsp90 mutant allowed v-Src-induced yeast lethality (Figure 3B).

Figure 3. Chaperone function of T22 phospho-mutants in yeast, (See Figure S2).

A) Yeast expressing WT, T22A, or T22E yHsp90 were transformed with v-SRC, and total cellular phosphotyrosine & v-Src expression were analyzed by immunoblotting; α-tubulin was used as loading control.

B) Growth of the same strains on glucose- or galactose-containing media.

C) Yeast expressing WT, T22A, or T22E yHsp90, and containing Ste11ΔN-His₆, were grown on glucose- or galactose-containing media. Ste11ΔN-His₆ expression was detected by immunoblotting; α-tubulin was used as loading control.

D) GR-lacZ activity was assessed in the same strains. Data are shown as percentage of WT activity, and are depicted as mean +/− standard deviation derived from four independent experiments. Lysates from yeast expressing His-tagged WT, T22A, or T22E yHsp90 were precipitated by Ni-NTA, and associating GR was detected by immunoblotting.

Next we assessed the stability of the endogenous kinase client Ste11 (a Raf-1 ortholog) (Flom et al., 2008; Louvion et al., 1998). Ste11ΔN, a constitutively active variant expressed under GAL1 promoter, was readily detectable in wild type cells (Figure 3C). However, Ste11ΔN protein was moderately reduced in cells expressing yHsp90-T22A and was undetectable in cells expressing yHsp90-T22E (Figure 3C).

The glucocorticoid receptor (GR) is another well-characterized Hsp90 client that provides a sensitive assay for Hsp90 function in yeast (Pratt et al., 2004). To assess the impact of T22 mutation on GR activity, we transformed yHsp90-T22A and yHsp90-T22E mutants, and wild type cells, with a GR expression plasmid also carrying a glucocorticoid-regulated LacZ reporter gene. GR activity in T22A mutants was reduced to less than half that seen in WT cells. Unexpectedly, GR activity was increased approximately 4- fold (compared to wild type) in yHsp90-T22E mutants (Figure 3D). We also examined the interaction of both yHsp90-T22A and yHsp90-T22E with GR. Yeast lysates were subjected to affinity pull-down with Ni-NTA agarose, and GR interaction with Hsp90 proteins was assessed by Western blot. In contrast to their disparate impact on GR activity, GR interaction with both yHsp90-T22A and yHsp90-T22E mutants was slightly but uniformly reduced (Figure 3D). Thus, the discrepant effects of T22 mutation on GR activity cannot be explained solely by differential association with Hsp90.

To obtain greater insight into the functional consequences of T22 mutation, we also assessed their impact on the stability of the cystic fibrosis transmembrane conductance regulator protein (CFTR), an Hsp90 client (Loo et al., 1998; Youker et al., 2004). We expressed HA-tagged CFTR under the control of a constitutive promoter in wild type yHsp90, yHsp90-T22A, and yHsp90-T22E yeast cells. The rate of CFTR protein degradation was determined by cycloheximide (CHX) chase analysis. We found CFTR stability to be similar in Hsp90 wild type and T22A mutant cells, whereas CFTR stability was significantly improved in yHsp90-T22E mutant cells (Figure S2).

Chaperone activity/client interaction of human Hsp90 is affected by T36 phospho-mutation

Threonine 22 is highly conserved in eukaryotic Hsp90 proteins, including human Hsp90α (hHsp90α), where the equivalent residue is T36. However, because CK2 phosphorylates numerous threonine residues in the N-domain of hHsp90α, it is impossible to specifically observe T36 phosphorylation in hHsp90α using a pan anti-phosphothreonine antibody (Lees-Miller and Anderson, 1989; Rose et al., 1987). Instead, we assessed the impact of T36A and T36E mutation on interaction of hHsp90α with an array of client proteins. First, we transiently expressed FLAG-tagged wild type (WT) hHsp90α, hHsp90α-T36A, or hHsp90α-T36E mutants in v-Src-transformed NIH-3T3 fibroblasts. Following immunoprecipitation of v-Src, we detected its interaction with FLAG-tagged WT hHsp90α; v-Src interaction with hHsp90α-T36A was significantly reduced and its interaction with hHsp90α-T36E was completely lost (Figure 4A).

Figure 4. Chaperone function of T36 phospho-mutants in mammalian cells.

A) NIH 3T3 cells stably expressing v-Src were transfected with empty vector pcDNA3 (c), FLAG-tagged WT hHsp90α (wt), hHsp90α-T36A (T36A), and hHsp90α-T36E (T36E) mutants. After v-Src IP, FLAG-hHsp90α interaction was detected by immunoblotting.

B) Empty vector pcDNA3 (c), FLAG-tagged WT hHsp90α (wt), hHsp90α-T36A (T36A), and hHsp90α-T36E (T36E) mutants were expressed in COS7 cells. After 24 hrs, interaction of endogenous Raf-1, ErbB2 and CDK4 with FLAG IPs were monitored by immunoblotting.

C) COS7 cells were transfected with GR and indicated FLAG-hHsp90α constructs. After 24 hrs, lysates were immunoprecipitated with FLAG antibody-conjugated agarose; co-precipitating GR was detected by immunoblotting.

D) HEK293 cells were transfected with CFTR and indicated FLAG-hHsp90α constructs. After 24 hrs CFTR and FLAG-hHsp90α were detected by immunoblotting. α-tubulin was used as loading control.

Next we transiently expressed FLAG-tagged WT hHsp90α, hHsp90α-T36A, or hHsp90α-T36E in COS7 cells. Following immunoprecipitation with anti-FLAG antibody, we observed interaction of endogenous Raf-1, ErbB2 and CDK4 with WT hHsp90α; interaction of these kinase clients with hHsp90α-T36A was significantly reduced while interaction with hHsp90α-T36E was completely lost (Figure 4B). Similar results were observed for GR (Figure 4C).

To gain insight into the functional activity of these Hsp90 mutants in mammalian cells, we examined their ability to affect steady-state expression of CFTR. Because of its propensity to unfold in mammalian cells, steady-state expression of mature CFTR can be significantly improved by co-expression of exogenous Hsp90. HEK293 cells were transiently co-transfected with CFTR and either empty vector, FLAG-tagged WT hHsp90α, hHsp90α-T36A, or hHsp90α-T36E. The more mature, highly glycosylated (higher molecular weight) forms of CFTR were expressed to a greater level in cells that were co-transfected with exogenous Hsp90, but cells expressing hHsp90α-T36E displayed more abundant CFTR protein compared to cells expressing either WT hHsp90α or hHsp90α-T36A (Figure 4D). Thus, while phospho-mimetic mutation of either T22 or T36 abrogates Hsp90 association with kinase clients and interferes with chaperoning of several kinases, the same mutation enhances GR activity and CFTR expression.

Impact of T22 phospho-mutation on yHsp90 binding to co-chaperones

Yeast Hsp90-T22A has normal ATPase activity in vitro, yet its ability to chaperone kinases and GR in cells is clearly affected. The ATPase activity of yHsp90-T22E is significantly impaired in vitro and its chaperone defect in cells is generally more severe (with the exception of GR and CFTR). Therefore, we examined the interaction of both T22A and T22E yHsp90 mutants with several key regulatory Hsp90 co-chaperones. Lysates of yeast expressing WT yHsp90, yHsp90-T22A, or yHsp90-T22E were subjected to affinity pull-down with Ni-NTA agarose, and interacting co-chaperones were examined by Western blot. Neither mutation significantly impacted Sti1^Hop or Sba1^p23 interaction compared to WT (Figure S3). In contrast, Cdc37^p50 association with both mutants was reduced and yAha1 interaction with these mutants was completely lost (Figure 5A, B). We saw similar, although not identical, results when we examined the interaction of hHsp90α-T36A and hHsp90α-T36E mutants with mammalian cochaperones in COS7 cells (Figure 5C, D and Figure S3). While hAha1 interaction with both T36A and T36E mutants was completely abolished, p50^Cdc37 interaction with hHsp90α-T36Awas not as severely impacted as was its interaction with hHsp90α-T36E. Likewise, although p60^Hop association was not affected by T36 mutation, p23 association with both T36A and T36E mutants was slightly decreased.

Figure 5. Reduced interaction of Aha1 and Cdc37^p50 co-chaperones with T22T36 phospho-mutants, (See Figure S3).

(A) Lysates from yeast expressing WT, T22A, or T22E yHsp90 were precipitated by Ni-NTA, and associating Cdc37^p50 and;

(B) yAha1 were detected by immunoblotting.

(C) COS7 cells were transfected with indicated FLAG-hHsp90α constructs. After FLAG-hHsp90α IP, associated p50^Cdc37 and;

(D) hAha1 were detected by immunoblotting.

(E) The yeast strains used above were transformed with either yAha1-FLAG or empty plasmid. Interaction of T22A and T22E yHsp90 with Cdc37^p50 and yAha1-FLAG was examined by immunoblotting.

(F) COS7 cells were co-transfected as in (C), and also with either an empty plasmid or with hAha1 plasmid. FLAG-hHsp90α was immunoprecipitated and associated p50^Cdc37 and hAha1 was detected by immunoblotting.

Since Aha1 interaction with both yeast and human Hsp90 mutants was completely abolished, we examined whether over-expression of this co-chaperone in either background might ameliorate the observed defects in co-chaperone associations. Yeast Aha1-FLAG was expressed under its native promoter on a multi-copy (2μ) plasmid. Over-expression of yAha1-FLAG restored yHsp90-T22A and yHsp90-T22E association with both Cdc37^p50 and Aha1 (Figure 5E). Likewise, over-expression of human Aha1 in COS7 cells restored Aha1 and p50^Cdc37 association with hHsp90α-T36A and hHsp90α-T36E mutants (Figure 5F).

Aha1 over-expression compensates for the chaperone defects of T22 phospho-mutants

Aha1 is a potent stimulator of Hsp90 ATPase activity (Panaretou et al., 2002) and its over-expression restored interaction of T22 mutants with both yAha1 and Cdc37^p50. yAha1 over-expression also restored the ability of both yHsp90-T22A and yHsp90-T22E mutants to productively chaperone v-Src and Ste11 kinases (Figure 6A, B). MAP kinase Mpk1/Slt2 mediates a cell wall integrity pathway in yeast and requires Hsp90 to maintain its activity (Millson et al., 2005). We used the down-stream reporter gene YIL117c-LacZ to monitor Mpk1/Slt2 activity in yeast expressing WT Hsp90, yHsp90-T22A, or yHsp90-T22E. Using caffeine to activate the reporter (Millson et al., 2005), we found that reporter activity was significantly reduced in both yHsp90-T22 mutants, with the T22E mutant displaying the more severe chaperone defect (Figure 6C). Yeast Aha1 over-expression corrected the cell wall integrity pathway signaling defect of both mutants (Figure 6C).

Figure 6. Aha1 over-expression compensates for defective chaperone function of yHsp90-T22 phospho-mutants.

A) Yeast expressing WT, T22A, or T22E yHsp90 were transformed with v-SRC and with either yAha1-FLAG or empty plasmid as shown. Total cellular phosphotyrosine and v-Src expression were analyzed by immunoblotting; α-tubulin was used as loading control.

B) Yeast cells expressing WT, T22A, or T22E yHsp90, and also Ste11ΔN-His₆, were transformed with either yAha1-FLAG or empty plasmid as shown. Ste11ΔN-His₆ expression was detected by immunoblotting; α-tubulin was used as loading control.

C) YIL117c-LacZ activity was measured in yeast expressing WT yHsp90-His₆, yHsp90-T22A, or yHsp90-T22E, with (+) or without (−) yAha1-FLAG. Cells were grown to mid-log and then stressed with 8 mM caffeine for 4 h. Data are depicted as mean +/− standard deviation derived from three independent experiments. yAha1-FLAG was visualized by immunoblotting and α-tubulin was used as loading control.

D) GR activity was measured in yeast cells expressing WT, T22A, or T22E yHsp90. Cells were transformed with either yAha1-FLAG or empty plasmid as shown. Data are displayed as % of WT activity. Bars represent mean +/− standard deviation derived from four independent experiments. Lysates from yeast expressing WT, T22A, or T22E yHsp90 were precipitated by Ni-NTA, and associating GR was detected by immunoblotting. yAha1 was detected by anti-FLAG antibody and α-tubulin was used as loading control.

E) HEK293 cells were transfected with wild type CFTR, the indicated FLAG-hHsp90α constructs, and either an empty plasmid or hAha1 plasmid. After 24 hrs, CFTR, FLAG-hHsp90α and hAha1 were detected by immunoblotting. α-tubulin was used as loading control.

yAha1 over-expression enhanced GR interaction with yHsp90-T22A and yHsp90-T22E mutants. However, unlike the case of the kinases described above, the impact of yAha1 over-expression on GR activity was not uniform (Figure 6D). yAha1 over-expression increased the chaperone activity of previously GR chaperone-deficient yHsp90-T22A (i.e., GR activity greater than WT) while it decreased the chaperone activity of previously GR chaperone-hyperactive yHsp90-T22E (i.e., GR activity reduced 4-fold to WT level). It is noteworthy that yAha1 over-expression did not affect GR activity of WT yHsp90 (98% GR activity of yeast containing empty plasmid), Figure 6D). Finally, since CFTR expression was optimally enhanced in mammalian cells exogenously expressing hHsp90α-T36E (see Figure 4D), we examined the impact of hAha1 over-expression in this context as well. hAha1 over-expression also reduced CFTR levels in HEK293 cells exogenously expressing any of the Hsp90 proteins, including hHsp90α-T36E (Figure 6E).

DISCUSSION

We have shown conclusively that CK2 phosphorylates a conserved threonine residue (T22) in the N-domain of yHsp90 both in vitro and in vivo. Because additional Hsp90 N-domain threonine residues are constitutively phosphorylated in vivo in a CK2-independent manner, we chose to utilize non-phospho- and phospho-mimetic mutations of T22 (alanine and glutamic acid, respectively) to investigate the affect of T22 phosphorylation on chaperone function. Mutation of this residue to alanine did not significantly affect Hsp90 ATPase activity, while phospho-mimetic mutation to glutamic acid reduced ATPase activity by 60%, although both mutants were able to support yeast growth when present as the sole Hsp90 in vivo. Consistent with its reduced ATPase activity, N-domain dimerization of yHsp90-T22E in the presence of AMPPNP was less efficient than that of either wild type yHsp90 or yHsp90-T22A.

These data are in general agreement with studies showing that α-helix 1 in the Hsp90 N-domain, including T22 and several adjacent amino acids, participates in an important hydrophobic interaction with the catalytic loop in the Hsp90 middle domain that is initiated by ATP-stabilized N-domain dimerization and helps to stabilize the chaperone's ATPase competent state (Cunningham et al., 2008). We recently reported that mutation of an adjacent residue, Y24, to glutamic acid markedly reduced Hsp90 ATPase activity (Mollapour et al., 2010). It is not surprising that the negative charge introduced by glutamic acid mutation (or phosphorylation) of either Y24 or T22 would disfavor this hydrophobic interaction. Our data therefore suggest that α-helix 1 phosphorylation may provide an effective means to post-translationally regulate Hsp90 activity.

Neither non-phosphorylatable nor phospho-mimetic T22 mutants were able to fully chaperone several kinase clients in yeast, including v-Src, Ste11, and Mpk1/Slt2. In general, yHsp90-T22E displayed the more severe chaperone defect. In mammalian cells, mutation of the equivalent residue, T36, interfered with the Hsp90 association of several kinases, including v-Src, Raf-1, ErbB2, and Cdk4. Again, the functional impact of T36E mutation was more severe than that of T36A. Since neither non-phospho- nor phospho-mimetic Hsp90 mutants are competent to chaperone v-Src, our data suggest that conformational changes provided by dynamic phosphorylation and dephosphorylation of T22 are necessary for optimal Hsp90 chaperoning of this client. While the other kinases examined may have similar requirements for optimal activity, it is also possible that they can be effectively chaperoned by Hsp90 containing unphosphorylated T22/T36.

In contrast, Hsp90 T22E mutation supported greater than 4-fold more GR activity compared to WT yHsp90, while mutation toT22A resulted in less than half the GR chaperoning activity of WT Hsp90. Importantly, GR association with yHsp90 mutants in vivo was not predictive of chaperone activity. Like GR, stability/maturation of the highly dependent Hsp90 client CFTR was enhanced in both yeast and mammalian cells expressing Hsp90 containing a phosphomimetic mutation at this location. Thus some clients, including GR and CFTR, seem to be optimally chaperoned by Hsp90 that is phosphorylated at T22/T36. The discrepancy between the impact of T22A and T22E mutation on the chaperoning of kinases and non-kinase clients remains to be explained but likely reflects both distinct co-chaperone requirements and discrepant ATPase dependence of these structurally and functionally diverse Hsp90 clients.

Mutation of T22/T36 to either alanine or glutamic acid disrupted Hsp90 association with the co-chaperones Cdc37^p50 and Aha1. It is highly likely that this contributes to the chaperone defects seen with these mutants as Aha1 over-expression in vivo not only restored association of Cdc37^p50 and Aha1 with Hsp90 but also normalized (or at least impacted) Hsp90 chaperone activity. In support of this in vivo phenotype, we found that a molar excess of Aha1 effectively stimulated the in vitro ATPase activity of both Hsp90-T22A and Hsp90-T22E. Importantly, neither Cdc37^p50 interaction with wild type Hsp90 nor wild type Hsp90 function was affected by Aha1 over-expression.

Although not affecting the chaperone activity of wild type yHsp90, Aha1 over-expression in yeast fully restored the v-Src chaperoning activity of both yHsp90-T22A and yHsp90-T22E. The ability of these mutants to chaperone Ste11 and Mpk1/Slt2 was also recovered by Aha1 over-expression. In the case of GR, Aha1 over-expression reduced the chaperone activity of yHsp90-T22E to the level of wild type yHsp90, while increasing GR activity in the context of yHsp90-T22A to twice that seen in cells expressing wild type yHsp90. While we cannot yet explain this response, the tendency of Aha1 over-expression to act homeostatically with respect to GR chaperoning (in contrast to the chaperoning of kinases) is unexpected but intriguing. Aha1 over-expression also reduced CFTR levels in HEK293 cells exogenously expressing any of the Hsp90 proteins, including hHsp90α-T36E. Although these results are in general agreement with the recent work of Balch's laboratory (Koulov et al., 2010) suggesting that Aha1 interaction with Hsp90 is deleterious for its chaperoning of CFTR, we have consistently observed that hHsp90α-T36E is most proficient in chaperoning CFTR while the activity of hHsp90α-T36A is similar to that of wild type, even though neither the T36E nor the T36A Hsp90 mutant associates detectably with the endogenous Aha1 in HEK293 cells. These data suggest, therefore, that hHsp90-T36E is an inherently better chaperone of CFTR than either its non-phospho counterpart or wild type Hsp90.

In summary, CK2-dependent phosphorylation of T22/T36 in α-helix 1 of Hsp90's N-domain significantly affects chaperone function in yeast and mammalian cells, in part by modulating interaction of the co-chaperones Cdc37^p50 and Aha1. Phosphorylation of α-helix 1 allows Hsp90 to discriminate among its extensive clientele and may provide a mechanism for the chaperoning of specific Hsp90-dependent proteins.

EXPERIMENTAL PROCEDURES

Plasmids and yeast growth media

The yeast strain pp30 (MAT a, trp1-289, leu2-3, 112, his3-200, ura3-52, ade2-101, lys2-801, hsc82KANMX4, hsp82KANMX4), (Panaretou et al., 1998). Temperature sensitive ck2-13 mutant (YSB49) and its parent strain (YSB48) were kind gift of Dr. Claiborne V. C. Glover (Bandhakavi et al., 2003). Plasmids were constructed as previously described (Panaretou et al., 1998). Detailed procedures, list of primers (Table S1), and media conditions for both yeast and mammalian cells are provided in the Supplemental information.

Protein analysis, immunoprecipitation and immunoblotting

Total protein extracts were prepared and analyzed by immunoblotting, as previously described (Panaretou and Piper, 1996). Detailed immunoprecipitation and Western blotting are provided in the Supplemental information.

In vitro kinase assay

Yeast Hsp90 and the mutant T22A were N-terminally His6 tagged using pRSETA plasmid. They were then expressed in bacteria and 2 mg of protein extract were incubated with 50 μl of Ni-NTA agarose (Qiagen). 200 ng of baculovirus expressed and purified active recombinant human CK2-GST (Stressgen) was used in the kinase assay. CK2 kinase reactions were carried out as previously described (Lees-Miller and Anderson, 1989).

CK2 inactivation assay in yeast

Temperature sensitive ck2-13 mutant (YSB49) and its parent strain (YSB48), (kind gift of Dr. C. V. C. Glover) were transformed with either yHsp90-His ₆ and yHsp90-T22A-His₆ (both under GAL1 promoter and with PreScission protease sites) or p50-Cdc37-GFP plasmids and grown on YPRaf (2% (wt/vol) Bacto peptone, 1% yeast extract, 2% raffinose, 20 mg/liter adenine), at 25°C for 12hrs. Cells were then shifted to 34°C for 2hrs and then 2% (final concentration) galactose was added to cells and incubated at 34°C for further 2hrs. Total protein extracts were prepared and analyzed by immunoblotting, as previously described (Panaretou and Piper, 1996).

v-Src, GR, Ste11ΔN, and Rlm1 activation assays

Ste11ΔN induction was analyzed as previously described (Louvion et al., 1998). v-Src induction and activation were analyzed as previously described (Murphy et al., 1993; Xu and Lindquist, 1993). Measurements of YIL117c-LacZ and GR-LacZ activity were carried out as previously described (Garabedian and Yamamoto, 1992; Millson et al., 2005), respectively.

CFTR degradation assay in yeast

Wild type or phopho mutants yeasts were transformed with HA tagged CFTR plasmid under constitutive promoter (kind gift of Dr J. L. Brodsky). Cells were grown on YPD to mid-log and protein synthesis was stopped by the addition of cycloheximide to a final concentration of 50 μg/ml. Samples were collected and different time intervals and total protein extracts were prepared and analyzed by immunoblotting. CFTR was detected by anti-HA moclonal antibody (Covance).

Hsp90 ATPase Activity and Cross-Linking Assays

Hsp90 ATPase activity and dimerization were measured as previously described (Panaretou et al., 1998). Aha1 was added to a final concentration of 20 μM to stimulate ATPase activity. Data shown are averages of three separate measurements.

Isothermal Titration Calorimetry (ITC) and K_d Determinations

ITC and K_d determinations were performed as previously described (Panaretou et al., 1998). Heat of interaction was measured on a MSC system (Microcal), with a cell volume of 1.458 ml, in a buffer containing 20 mM Tris (pH 8.0), 1 mM EDTA, 5 mM NaCl and 7 mM MgCl₂ at 30°C. For AMPPNP interactions, 20 aliquots of 14.8 μl of 1 mM AMPPNP were injected into 50 μM WT or mutant T22A or T22E yeast Hsp90. Heat of dilution was determined in a separate experiment by diluting protein into buffer, and the corrected data fitted using a nonlinear least square curve-fitting algorithm (Microcal Origin) with three floating variables: stoichiometry, binding constant and change in enthalpy of interaction.