Comparative analysis of the impact of Heat shock protein 90 kDa or Cdc37 mutation on the yeast proteome

Abstract

Hsp90 is an abundant and essential molecular chaperone that is required for the folding and/or activity of up to 15%-20% of all yeast proteins. Hsp90 and its cochaperone Cdc37 are of interest due to their cooperative role in chaperoning oncogenic protein kinases. We previously grouped mutants in Saccharomyces cerevisiae Hsp90 into different categories based on the step of the adenosine triphosphate-dependent Hsp90 folding cycle that is disrupted. Here we used a quantitative proteomic approach to compare the impact of two different mutations that target the initial step of loading clients onto Hsp90, one in Hsp90 and one in Cdc37. We also included two Hsp90 mutants that target other steps in the folding pathway. We tested the effects at either 25°C or after a 1-h heat shock at 37°C. As expected, the Cdc37 mutant and the Hsp90 mutant that affects the loading step had similar effects, particularly on protein kinases, demonstrating the potential for using proteomics for testing the impact of alteration of the folding pathway. We also identified non-kinase proteins that may use a similar Cdc37-mediated pathway. Surprisingly, protein kinases were a small subset of the proteins impacted by Cdc37 alteration, challenging the notion that Cdc37 is a kinase-specific cochaperone. We also identified proteins that were most strongly affected by Hsc82 mutants that impact steps other than the loading step. Further analysis of how Hsp90 and cochaperones interact with these additional proteins may provide novel insights into alternative methods of client loading or other poorly understood steps in the folding cycle.

Background

Molecular chaperones maintain protein homeostasis by regulating the folding and activity of client proteins. Many of these chaperones function as cohesive networks that cooperate in the maturation of clients.¹ Heat shock protein 90 kDa (Hsp90) is an essential and abundant molecular chaperone first identified in complex with steroid hormone receptors and protein kinases. Over time, the list of Hsp90 clients has expanded to hundreds of structurally diverse proteins, representing up to 20% of the proteome.2, 3, 4 Hsp90 influences numerous aspects of cellular signaling pathways and has been described as a hub of protein homeostasis.⁵ Hsp90 clients function at multiple steps during the progression of cancer, and inhibitors have been sought for use as anticancer agents.6, 7 An Hsp90 inhibitor was recently approved for use in Japan to treat advanced gastrointestinal stromal tumor, although is not approved for use in the USA,8, 9 and more selective and/or less toxic inhibitors are needed.

Hsp90 is an adenosine triphosphate (ATP)-dependent chaperone that relies on a network of cochaperones to assist client folding.¹⁰ In an established folding pathway, Hsp90 progresses from an open conformation dimerized only at the carboxy-terminal domain to a closed ATP-bound conformation that is also dimerized at the amino-terminal domain. During early steps in the cycle, the cochaperones Sti1 or Cdc37 inhibit ATPase activity of Hsp90 by stabilizing an open conformation that is favorable for productive client loading.11, 12 Subsequent binding of ATP to Hsp90 promotes N-terminal dimerization.¹³ Other cochaperones such as Aha1 bind and stimulate ATPase activity or nucleotide exchange.10, 14, 15 We previously isolated a series of mutants of Hsc82, one of the two cytosolic yeast isoforms of Hsp90. When expressed in an hsc82hsp82 strain, the mutants cause temperature-sensitive growth and have differing impacts on the progression of the folding cycle. Loading mutants disrupt the early steps in the pathway that promote Hsp90-client interaction; reopening mutants disrupt steps associated with ATP hydrolysis and/or conformational changes; and “other” mutants have poorly defined defects. Even in cells grown at the optimal temperature of 30°C, these mutants have differing effects on the activity of specific clients as well as at the proteomic level.16, 17, 18

Cdc37 is an essential cochaperone that is critical for the stability of up to two-thirds of protein kinases in vivo.19, 20, 21, 22 Disrupting the Hsp90-Cdc37 interaction has been proposed as an anti-cancer therapy,²³ but has not yet been shown to be effective.²⁴ This may be due to direct interactions of protein kinases with both Hsp90 and Cdc37.²⁵ A few non-kinases have also been shown to be dependent on Cdc37, such as the androgen steroid hormone receptor and a human membrane receptor guanylyl cyclase receptor.26, 27 Overall, the interaction of “non-protein kinase” clients with Cdc37 is understudied and could reveal new cellular roles dependent on Cdc37 and/or Hsp90.

We recently used a proteomic approach to understand the impact of Hsp90 mutation on the yeast proteome at the optimal growth temperature.¹⁷ Here we used a similar quantitative proteomic approach to compare the impacts of mutations in genes encoding Cdc37 or Hsc82 before or after a 1-h heat shock. We compared three different Hsc82 mutants. One of the mutants (Hsc82-G309S) disrupts the loading step by altering a residue involved in Hsp90-Hsp70 interaction.²⁸ Cdc37 also functions during the loading step.²⁵ We tested one mutant in the reopening group (Hsc82-Q380K), which alters residues in a loop required for ATP hydrolysis.²⁹ For comparison, we added a mutant in the “other” category (Hsc82-G424D). Hsc82-G424D contains an alteration in a region linked to conformational changes,³⁰ and resulted in decreased interaction with Hsp70 and the cochaperone Sti1 relative to wild-type Hsc82.¹⁶ All three mutants have been shown to have distinct effects on clients in cells grown at 30°C,16, 18 and one question is whether those differences persist under stress conditions. The Cdc37 mutant we used (Cdc37-S14A) caused a temperature-sensitive growth defect and resulted in at least a 2-fold decrease in abundance of 51 of 65 protein kinases examined.²⁰ A goal of comparing the two sets was to understand whether the proteomic approach is feasible to compare the impact of mutants that alter the same step in the Hsp90 folding pathway. This could provide new data about whether Cdc37 mutants affect a subset of Hsp90 clients, consistent with the presence of independent folding pathways linked to one specific cochaperone. Another goal was to determine the extent of protein kinase and non-kinase proteins affected by the Cdc37 and Hsc82 mutants.

Results

Hsp90 inhibition or mutation results in loss of Hsp90 client function and/or client degradation, impacting up to 20% of the yeast genome and resulting in an altered transcriptional profile.2, 31, 32, 33 We previously found that yeast containing Hsc82 mutants that cause similar growth defects differentially impact the abundance of multiple cellular proteins, including known Hsp90 interactors.¹⁷ Figure 1a shows steps in the Hsp90 folding pathway affected by the mutants used in this study. Clients are delivered to Hsp90 by the general Hsp70 chaperone and a cochaperone such as Cdc37.¹⁰ Hsc82-G309S disrupts a contact site between Hsp90 and Hsp70,28, 34 disrupting client loading. Once ATP binds Hsp90, it undergoes a conformational change to form the closed complex. Hsc82-Q380K disrupts a residue within a loop in the middle domain that has roles in Hsp90 ATPase activity and associated conformational changes.²⁹ Multiple mutants in that loop disrupt ATPase activity and Hsp90-cochaperone interactions characteristic of the closed complex.16, 35 Hsc82-G424D does not share defects in Hsp70 or cochaperone interaction or client activity that have been identified for the other two mutants.16, 18 We compared the proteomes of yeast expressing wild-type Hsc82 or the three different mutants (G309S, Q380K, or G424D) to cells expressing wild-type Cdc37 or Cdc37-S14A. All mutants cause similar growth defects at 37°C (Figure 1b). Cdc37 directly binds protein kinases and assists loading onto Hsp90.21, 22, 25 In yeast, S14 is phosphorylated by casein kinase II.³⁶ The Cdc37-S14A alteration prevents phosphorylation and results in reduced Hsp90 interaction.³⁷ A prior study found that cells expressing Cdc37-S14A exhibited a 2-fold or greater decrease in abundance of 51 of 65 protein kinases examined.²⁰ We predicted that the Cdc37-S14A mutation would impact a subset of clients affected by the Hsc82-G309S mutation, since another cochaperone, Sti1 also functions to help load clients onto Hsp90.¹⁰ We also predicted the majority of proteins impacted by the Cdc37-S14A mutation would be protein kinases since Cdc37 is frequently referred to as a “kinase-specific” cochaperone.³⁸ Further, due to the growth defect of cells expressing mutant forms of Cdc37 or Hsc82, we expected to see a larger number of proteins affected at 37°C due to their misfolding and subsequent degradation and/or aggregation.

Fig. 1.

(a) Simplified model of the Hsp90 folding pathway. Client (dark gray oval) bound to a chaperone such as Cdc37 (light gray diamond) is targeted to the open conformation of Hsp90 (gray). After ATP binds, Hsp90 shifts to the closed conformation. ATP hydrolysis is associated with a return to the open conformation. The Hsc82-G309S and Cdc37-S14A mutants target early steps in the pathway associated with client loading (1, loading). The Hsc82-Q380K mutant targets reopening (2), and the G424D mutant targets an unknown step in the pathway (other). (b) Growth of yeast strains after 2 days at the indicated temperature on YPD media. ATP, adenosine triphosphate; YPD, yeast extract peptone dextrone.

Overview of proteomic data

In this study, we used data-independent acquisition mass spectrometry (DIA-MS) for global proteomic profiling, as it enables highly reproducible, comprehensive quantification of the proteome and outperforms traditional shotgun approaches in capturing proteome-wide changes.³⁹ We examined the effects of Cdc37 and Hsc82 alterations on the abundance of soluble proteins at both the permissive temperature (25°C), or after a 1-h heat shock (37°C). In the Cdc37 dataset a total of 3210 unique proteins were detected across the five different samples, representing ∼49% of the yeast proteome. In the Hsc82 mutant dataset, 4267 proteins were detected across the five different samples, reflecting ∼65% of the proteome. Table S1 lists all the proteins that were significantly affected (log₂ fold change ≥1, hereafter log₂FC). Notably, as shown in Table S1, the levels of Cdc37-S14A were elevated relative to wild-type Cdc37. In an independent study, there was also an apparent increase of Cdc37-S14A levels compared to wild-type Cdc37, although it was not quantified.³⁷ Regardless, the cells used in the study were confirmed to have a temperature-sensitive growth defect indicative of a Cdc37 defect (Figure 1b). The level of wildtype and mutant Hsc82 was not significantly different.

Figure 2a shows the overall effect of heat shock in each strain. Consistent with the established heat shock response, many proteins exhibited sharp increases in abundance upon heat shock.40, 41 A smaller number of proteins exhibited decreased abundance upon heat shock. For example, in the strain expressing wild-type Cdc37, the level of 178 proteins increased after heat shock, while the level of 41 proteins decreased. In cells expressing Cdc37-S14A,165 proteins increased after heat shock, and the levels of 64 proteins decreased. In comparison, in cells expressing wild-type Hsc82, the levels of 121 proteins increased after heat shock, and the levels of only 8 proteins decreased. The reason the overall reduced number of proteins with altered abundance in cells expressing wild-type Hsc82 versus wild-type Cdc37 is different is unknown. There were differences in the mass spectrometry methods used in the two datasets (see Materials and Methods), but a cross-platform comparison of 1214 proteins shared between the two studies showed that the heat shock response was highly correlated between the two datasets (r = 0.943; Figure S1). The transcription factors Hsf1 and Msn2/4 are known regulators of the heat shock response in yeast.40, 41 Hsf1 is responsive primarily to heat, while Msn2/4 induces a response after a variety of general stressors. To test whether Cdc37 or Hsc82 alteration impacted the ability of cells to mount a stress response, we assessed the quality of the heat shock by analyzing the changes in protein abundance of a few targets of Hsf1 (Aha1, Cpr6, Hsp78, and Hsp104) or Msn2/4 (Ald2, Ssa3, and Ssa4) after heat shock (Figure 2b).⁴⁰ In most cases, similar changes in abundance were observed across the different strains, indicating that the overall induction of the heat shock response was not affected by Cdc37 or Hsc82 alteration.

Fig. 2.

Summary of overall proteomic results. (a) Bar chart showing overall number of changes in protein abundance in each strain after a 1-h heat shock relative to the untreated strain (log₂FC ≥ 1.0). Hits above x-axis represent proteins with elevated abundance. Hits below the x-axis represent proteins with decreased abundance. (b) Bar chart showing overall log₂FC in select targets of Hsf1 or Msn2/4 in the different strains upon comparison of cells grown at 25°C versus after 1 h at 37°C. The dotted line represents the cutoff of log₂FC ≥ 1.0. Table S1 lists each protein and the relative change in abundance (log₂FC ≥ 1.0). For (b), Wild-type Cdc37 (gray), Cdc37-S14A (red), Hsc82-WT (dark grey), Hsc82-G309S (blue), Hsc82-Q380K (purple), and Hsc82-G424D (green).

We next examined the impact of the Cdc37 or Hsc82 mutation on the basal expression of stress-induced proteins. A genome-wide analysis identified 18 genes, most of which encode chaperones, that are dependent on Hsf1 for basal expression and heat shock induction.⁴⁰ Figure 3a and b compare the basal and induced expression, respectively, of a selection of five established Hsf1 targets in cells expressing the Cdc37 or Hsc82 mutants. Consistent with prior reports that mutation or inhibition of Hsp90 relieves repression of Hsf1,16, 42 all five of the proteins examined (Aha1, Cpr6, Hsp78, Ssa1, and Hsp104) exhibited increased abundance in cells expressing mutant Hsc82 relative to wild-type Hsc82. The increase was observed in both basal and induced conditions, but the log₂FC did not meet our cutoff. The Cdc37 mutation did not result in increased or basal levels of Hsf1 targets. We also examined the impact of Cdc37 or Hsc82 mutation on the abundance of Msn2/4 targets. The basal and/or induced levels of seven targets of Msn2/Msn4 were elevated in cells expressing Cdc37-S14A (Figure 3c and d), although the level of change in protein abundance was not consistent from protein to protein. However, in most cases, the change in abundance in cells expressing Cdc37-S 14A met or exceeded our cutoff. In contrast, the abundance of most Msn2/4 targets was not altered by the Hsc82 mutation.

Fig. 3.

Relative change in protein levels of select Hsf1 targets and Msn2/4 targets in different strains. Change in abundance of Hsf1 targets at 25°C (a) or after 1 h heat shock (37°C) (b). (c) and (d) Change in abundance of Msn2/Msn4 targets at 25°C (c) or after 1 h heat shock (37°C) (d). The dotted line represents the cutoff of log₂FC ≥ 1.0. Cdc37-S14A relative to wild-type Cdc37 under the same conditions (red). Indicated Hsc82 mutant versus wild-type Hsc82 under the same conditions: Hsc82-G309S (blue); Hsc82-Q380K (purple); and Hsc82-G424D (green).

Shared abundance changes in protein kinases across Cdc37-S14A and Hsc82 mutants

In the prior study that observed decreased levels of 2/3 of protein kinases impacted by Cdc37-S14A,²⁰ TAP-tagged protein kinases were expressed from the endogenous chromosomal location, and steady state protein kinase levels in cells expressing wild-type or mutant Cdc37 were compared after growth at 30°C. Yeast contain 129 protein kinases.⁴³ The Cdc37 and Hsc82 datasets contained data for 70 and 104 protein kinases, respectively (Table S2). Surprisingly, only 6 protein kinases showed statistically significant decreases (log₂FC ≥ 1) in abundance when either Cdc37 or Hsc82 was altered. The relative change in abundance of protein kinases before and after the 1-h heat shock are shown in Figure 4a, and b, respectively. Four kinases, Tpk3, Cdc28, Dun1, and Slt2, were affected by both Hsc82 and Cdc37 mutation, while two kinases, Tda1 and Ctk1, were impacted significantly only by the Hsp90 mutation. All of the kinases were previously shown to interact with Cdc37 or to be impacted by Cdc37-S14A,20, 44 and all of the kinases except Dun1 have previously been linked to Hsp90.2, 4, 44, 45, 46, 47 At 25°C, the level of Tpk3 was significantly reduced in cells expressing Cdc37-S14A and levels of Dun1 were reduced in cells expressing Hsc82-G309S. After heat shock, the abundance of all six kinases was reduced. Compared to the other two Hsc82 mutants, the Hsc82-G309S mutant had the largest impact on the abundance of protein kinases, consistent with a shared function with Cdc37 during the client loading step.16, 17 Tpk3 is a catalytic subunit of cAMP-dependent protein kinase,48, 49 and Cdc28 is an essential kinase that regulates cell cycle transitions.50, 51 Dun1 is a serine/threonine kinase involved in DNA damage response in yeast.⁵² Slt2 is a mitogen-activated protein kinase that regulates the maintenance of cell wall integrity.³⁷ Tda1 has a role in glucose signaling,⁵³ and Ctk1 is the catalytic subunit of the C-terminal domain kinase I.⁵⁴ Reduced levels of these proteins, particularly under prolonged stress, would impact diverse functions such as signaling pathways, the cell cycle, cell wall integrity, and transcription. A potential reason that we did not observe reduced levels of other kinases is that some may only need Cdc37 during initial synthesis. This is consistent with some of the results of Mandal et al, in which levels of Tpk2 were sensitive to the effects of the cdc37-S14A mutation in the early time points of a pulse chase experiment.²⁰ Alternatively, some kinases may maintain stability after only 1 h at the non-permissive temperature.

Fig. 4.

Relative change in protein levels of protein kinases in different strains at 25°C (a) or after heat shock (37°C) (b). The dotted line represents the cutoff of log₂FC decrease ≥ 1.0. Cdc37-S14A relative to wild-type Cdc37 under the same conditions (red). Indicated Hsc82 mutant versus wild-type Hsc82 under the same conditions: Hsc82-G309S (blue); Hsc82-Q380K (purple); and Hsc82-G424D (green).

Shared abundance changes in non-protein kinases across Cdc37-S14A and Hsc82 mutants

In addition to the protein kinases, two proteins, Cha1 and Frm2, were similarly impacted by the Hsc82 and Cdc37 mutants. Both were reduced at 25°C in cells expressing Cdc37-S14A or at least one Hsc82 mutant, while Cha1 demonstrated reduced levels after heat shock (Figure 5a and b, respectively). Cha1 is an L-serine deaminase⁵⁵ that has been shown to genetically or physically interact with Hsp90.⁴ Frm2 is a nitroreductase that uses NADH as a reductant,⁵⁶ and has not previously been linked to Hsp90 or Cdc37. The observation that these proteins are most strongly affected by Cdc37-S14A and Hsc82-G309S, similar to protein kinases, suggests that Cdc37 may assist loading of these proteins onto Hsp90. However, proteomic data alone cannot distinguish whether they reflect direct chaperone clients or downstream consequences, such as altered kinase signaling pathways.

Fig. 5.

Relative change in protein levels of non-kinase proteins affected by Cdc37 and Hsc82 alteration compared to respective wild-type strains. Proteins that exhibit reduced abundance at 25°C (a) or 37°C (b). Proteins that exhibit increased abundance at 25°C (c) or 37°C (d). The dotted line represents the cutoff of log₂FC ≥ 1.0. Cdc37-S14A relative to wild-type Cdc37 under same conditions (red). Indicated Hsc82 mutant versus wild-type Hsc82 under same conditions; Hsc82-G309S (blue); Hsc82-Q380K (purple); and Hsc82-G424D (green). Data for Frm2 was missing in the Cdc37-S14A dataset at 37°C.

A few proteins showed increased abundance in cells expressing Cdc37-S14A and/or Hsc82-G309S relative to the wild-type controls (Figure 5c and d). While not affected at 25°C, Pho5 was increased cells in expressing Cdc37-S14A or Hsc82-G309S after heat shock. Pho5 is induced by phosphate starvation.⁵⁷ At 25°C, increased levels of Ygp1, Pbi2, and Dal3 were observed in cells expressing Cdc37-S14A or Hsc82-G309S. Dur1,2 was only increased in cells expressing Hsc82-G309S. Elevated levels of all four proteins were also observed after heat shock: Ygp1, Pbi2, and Dal3 were increased in the cdc37-S14A strain, and Dur1,2 was increased in the hsc82-G309S strain. Ygp1 is induced by nutrient deprivation-associated growth arrest,⁵⁸ and Pbi2 abundance increases in response to DNA replication stress.⁵⁹ Dal3 and Dur1,2 are part of the allantoin degradation pathway, and their induction may signal nitrogen starvation.⁶⁰ All these proteins are induced by various stressors, suggesting the observed increase in abundance is due to transcriptional induction of stress responses. Although further evidence is needed to show the mechanism of these effects, this result provides a glimpse into the widespread effects of disrupting Hsc82 or Cdc37 function. As observed for protein kinases, cells expressing Cdc37-S14A or Hsc82-G309S had the most shared effects, with limited effects observed in cells expressing Hsc82-Q380K or Hsc82-G424D.

Identification of proteins impacted by Cdc37-S14A but not Hsc82-G309S

The abundance of 21 proteins decreased in cells expressing Cdc37-S14A in at least one of the two conditions without a corresponding change in cells expressing Hsc82 mutants (Figure 6a and b). The effect of the Hsc82-G309S mutant is shown for comparison. Some were affected primarily at 25°C, some were impacted primarily after heat shock, and some were impacted at both temperatures. We speculate that the proteins impacted at 25°C require Cdc37 for stability during synthesis or have a continual requirement for Cdc37 similar to v-src kinase.⁶¹ In contrast, the proteins impacted after heat shock may only require Cdc37 under conditions that lead to protein misfolding. These proteins have varied functions (Figure 6c). Some have DNA-related functions (Snf11, Msh1, Orc4, and Tod6), while many have enzymatic activities (Oye3, Erg4, Gtt2, Err1, Ecm4, Set2, and Dip5). A summary of prior evidence linking these proteins to Cdc37 or Hsp90 is shown in Figure 6c. Deletion of the genes encoding Erg4, Msh1, Gre2, and Get2 resulted in enhanced growth defects when combined with temperature sensitive cdc37 alleles.49, 62 Similarly, deletion of the genes encoding Oye3, Rcy1, Prm5, Snf11, Set2, Orc4, and Get2 resulted in enhanced growth defects upon genetic or pharmacologic inhibition of Hsp90.2, 49, 63, 64 Some of the hits were shown to crosslink to Hsp90 (Erg4, Ecm4, Hsp26, Gre2, Rsn1, and Get2).3, 4 Many of the proteins, as noted in Figure 6c, displayed reduced abundance in cells expressing mutant forms of Hsc82 in our prior study.¹⁷ Overall, 18/21 proteins in this list have been somehow linked to Hsp90 but were most strongly affected by Cdc37 alteration in this study. Additional studies are required to confirm that the proteins listed in Figure 6 are clients of Cdc37 and/or Hsp90 as compared to downstream effects of affected clients.°

Fig. 6.

Relative change in protein levels of non-kinase proteins affected only by Cdc37 mutation. Left: Proteins that exhibit reduced abundance 25°C (a) or 37°C (b). The dotted line represents the cutoff of log₂FC decrease ≥ 1.0. Cdc37-S14A (red) relative to wild-type Cdc37 under the same conditions. Hsc82-G309S (blue) under the same conditions shown for comparison. (c) Descriptions of affected protein functions and known interactions with Cdc37 or Hsp90 as shown on yeastgenome.org. For comparison, we also indicated whether proteins were impacted by a larger set of Hsc82 mutants in our prior study.¹⁷

Comparison of the effects of the different Hsc82 mutants

We further compared the effects of the different Hsc82 mutants. Relative to cells expressing Cdc37-S14A, there were moderate changes in cells expressing Hsc82-G309S using our standard log₂FC ≥ 1 cutoff, and even fewer changes observed in cells expressing Q380K or G424D compared to wild-type cells. Due to the low number of proteins that were impacted by the Hsc82 mutation, we reduced the log₂FC cutoff to allow a relative comparison of the effects of the different mutants. We first looked at proteins with reduced levels in cells expressing Hsc82-G309S using a log₂FC ≥0.75 cutoff (Figure 7a and b). Of the proteins impacted at 25°C, there was a sharp contrast between the effects of Hsc82-G309S and the other two Hsc82 mutants. After heat shock, most hits were specific to Hsc82-G309S, although some were affected to a lesser extent by G424D. Protein kinases that were impacted at this lower cutoff include Kin28, Hrk1, and Ypk3. We also observed reduced levels of the stress-induced proteins (Gre1 and Hsp42), the multi-drug transporter Pdr5, the cleavage/polyadenylation factor Pti1 and the phosphatase Sdp1. Additional hits include Bna7 and Fbp26, enzymes required for NAD biosynthesis and glucose metabolism, respectively. Other affected proteins are involved in RNA or DNA processing or maintenance (Sdc1, Rex2, Mhf2, and Raf1) and have less characterized function (YKL069W, YIL108W, Crg1, and Crh1). A subset of these proteins (Hbt1, Gre1, Pdr5, Hsp42, Kin28, Fbp26, Pti1, Hrk1, and Ctk2) have previously been linked to Hsp90 or Cdc37. Few of these proteins were similarly impacted by Cdc37-S14A, suggesting that cochaperones other than Cdc37 may be required for their function or that the proteins are less sensitive to the effects of the Cdc37-S14A alteration.

Fig. 7.

Proteins that exhibited the largest reduction in levels in Hsc82-G309S relative to the other two mutants at 25°C (a) or 37°C (b). The darker dotted line represents the cutoff of log₂FC increase ≥ 1.0. The lighter dotted line represents the reduced cutoff of log₂FC increase ≥ 0.75. Indicated mutant versus wild-type protein under same conditions. Cdc37-S14A (red); Hsc82-G309S (blue); Hsc82-Q380K (purple); and Hsc82-G424D (green). Cdc37 data was missing or incomplete for Gre1, Raf1, Mhf2, Ypk3, and Ctk2.

We next examined proteins impacted most strongly by Hsc82-Q380K or Hsc82-G424D. The modest effects of the Q380K and G424D mutants are consistent with our prior studies that show these mutants affect a restricted set of clients.16, 17, 18 We dropped to a cutoff of log₂FC ≥0.5, to gain preliminary understanding of whether there are client-specific effects of these mutants. The effect of the two mutants varied in a temperature-specific manner. Most of the effects of Hsc82-Q380K were observed at 25°C (Figure 8a). Reassuringly, we previously found that the folding and activity of Eft1 is sensitive to Hsc82-Q380K and other mutants in our reopening group,¹⁸ and that specificity was also observed here. Most impacts of the Hsc82-G424D mutants were observed only after the 1-h heat shock (Figure 8b). Proteins affected in Figure 8a have roles in nutrient acquisition and regulation. Pho5 and Pho11 are acid phosphatases that are induced under phosphate limitation, Zrt1 encodes the high-affinity zinc importer induced during zinc limitation, and Arg5,6, His3, and Aro5 (YGL117W) function in arginine, histidine, and aromatic amino acid biosynthesis, respectively. Proteins most sensitive to Hsc82-G424D in Figure 8b include three different methyltransferases. Trm12 and Ppm2 modify tRNA, and Rmt2, which modifies proteins. Many of the other hits also have enzymatic activity, such as Aad16, Ydr541C, Zwf1, Nma2, Faa4, Gde1, and Gad1. Further studies are needed to determine whether the proteins in Figure 8 are authentic Hsp90 clients that are specifically impacted by different mutant groups and whether more dramatic changes in growth temperature, such as a longer heat shock, lead to more significant changes in abundance.

Fig. 8.

Proteins that exhibited the largest reduction in levels in Hsc82-Q380K or Hsc82-G424D at 25°C (a) or 37°C (b). The darker dotted line represents the cutoff of log₂FC increase ≥ 1.0. The lighter dotted line represents the reduced cutoff of log₂FC increase ≥ 0.50. Indicated mutant versus wild-type protein under same conditions. Hsc82-G309S (blue); Hsc82-Q380K (purple); and Hsc82-G424D (green).

Discussion

Hsp90 partners with multiple cochaperones, and one hypothesis for the presence of the different cochaperones is they enable Hsp90 to interact with a range of clients with diverse structures.65, 66 Detailed structural information about how clients interact with Hsp90 and cochaperones in an ordered pathway is available for only one client, the glucocorticoid receptor.34, 67, 68 The structure of Hsp90 and Cdc37 in complex with the kinase domain of Cdk4 is also available.²⁵ Despite the lack of structural similarity, the glucocorticoid receptor and protein kinases interact with a similar surface of Hsp90.⁶⁸ Another client, tau, interacts with different surfaces on Hsp90.⁶⁹ One goal of these studies was to identify clients that are loaded onto Hsp90 with assistance from Cdc37. Another goal was to identify proteins that are highly dependent on steps that occur after client loading, potentially uncovering new mechanistic insights. This study builds on our earlier work that showed it is possible to use a proteomic approach to distinguish the effects of Hsp90 mutation that target distinct steps in the folding pathway.¹⁷ As expected, the Cdc37-S14A mutant and the Hsc82 mutant in the loading group had the strongest impacts on protein kinases. We also identified two additional proteins (Cha1 and Frm2) that were strongly impacted by Cdc37 mutation or Hsc82 mutation, suggesting their interactions with Hsp90 and Cdc37 are similar to those of protein kinases, but further tests are required to confirm these results.

Inhibition or mutation of Hsp90 is known to result in depression of the heat shock response, which undermines some of the therapeutic benefits.⁷⁰ Our results support prior evidence that Hsp90 mutation results in an increased level of proteins regulated by Hsf1,16, 42 although the effects were modest. Cdc37 mutation did not result in increased levels of Hsf1 targets. Similarly, in mammalian cells, a small molecule inhibitor that blocks Hsp90 interaction with Cdc37 did not trigger heat shock response.⁷¹ In prior studies, deletion of the gene encoding the Cpr7 cochaperone resulted in increased basal expression of an Hsf1 target.⁴² Here we showed that mutation of Cdc37 resulted in increased abundance of multiple targets of Msn2/4, which mediate the general stress response mediated by Msn2/4. Some Hsp90 mutations, as well as a mutation of the Sgt1 cochaperone, resulted in increased basal expression of genes regulated by Msn2/Msn4.17, 72 That effect is thought to be mediated by the adenylate cyclase Cyr1.⁷² Cdc37 is not known to interact with Cyr1, and the mechanism of activation of the general stress response is unknown, although it may be due to the role of the Hsp90-dependent kinase Slt2 in signal transduction.³⁷ However, our results indicate that Cdc37 alteration induces a general stress response under conditions that Hsc82 alteration does not, suggesting an Hsp90-independent function of Cdc37. The overall differences in how mutation of Cdc37, Cpr7, and Sgt1 cochaperones impact the stress response provides further evidence that cochaperones have different client specificity.⁶⁵ One unexpected finding is that levels of two different chaperones, Hsp26 and Hsp42, are decreased in cells expressing Hsc82 mutants, and that the Cdc37 mutation impacts levels of two different glutathione S-transferases, potentially resulting in further changes in the ability of yeast cells to respond to environmental stress. It is possible that reduced levels of soluble Hsp42 are due to formation of aggregates containing Hsp90-dependent kinases.⁷³

Cdc37 is an essential cochaperone known for its interaction with protein kinases.²² In yeast, Cdc37 controls cell-cycle progression and multiple signaling pathways by chaperoning the Cdc28, Slt2, Kin28, Ste11, Msp1, and Gcn2 kinases.37, 50, 74, 75, 76, 77 Due to its established functions in loading protein kinases onto Hsp90, we predicted the majority of the proteins with decreased abundance in cells expressing Cdc37-S14A would be protein kinases.19, 20, 21, 25 However, we only detected significant changes in the levels of four kinases, Cdc28, Slt2, Dun1 and Tpk3, all of which were previously shown to be sensitive to Cdc37 mutation.²⁰ Cdc37 has been known to interact with Cdc28 for over 30 years.50, 78 Recent studies suggest role for Cdc37, Cdc28, and Hsp90 in promoting morphological changes and the evolution of multicellularity.79, 80 One potential reason we did not detect additional kinases is that the kinases required Cdc37 only during initial synthesis or folding.²⁰ Protein kinases have been shown to have differing sensitivity to Hsp90 inhibitors,⁸¹ and it is possible that some kinases remained stable over this time course. All eukaryotic kinases are structurally similar, and studies suggested that destabilized kinase domains are more dependent on Hsp90 and/or Cdc37.¹⁹ A mutational analysis of v-Src and the homologous c-Src demonstrated a continuous spectrum of dependence that ranged from kinases dependent only during initial folding to clients constitutively dependent on Hsp90/Cdc37 for activity or stability.⁸² Two protein kinases, Ctk1 and Tda1 were significantly impacted by Hsp90 alteration but not Cdc37 alteration. It is possible that Ctk1 and Tda1 interact with Cdc37 but may be more sensitive to other mutations in Cdc37.⁸³ Further studies are needed to determine whether the activity of these two kinases is more sensitive to changes in Hsp90 than in Cdc37.

Although most prior studies focus on the interaction of Cdc37 with protein kinases, other studies have identified physical and functional interactions with non-kinase proteins. The structure of the human membrane receptor guanylyl cyclase GC-C in complex with Hsp90 and its co-chaperone Cdc37 shares similarities with the Hsp90-Cdc37-Cdk4 kinase structure.²⁶ Prior studies also suggest a broader role for Cdc37. A yeast two-hybrid assay identified Cdc37 interactions with transcription factors, proteins involved in vesicle trafficking or vacuolar protein sorting, nuclear import, chromosome segregation and others.⁸⁴ A separate synthetic lethal screen in yeast expressing the cdc37-S14A allele and identified 38 different proteins with genetic interactions, including components of topoisomerase and proteins that regulate chromatin complexes and modifications.⁶² Surprisingly, the only shared hit between our study and the prior studies is Get2, part of the guided entry of tail-anchored protein (GET) pathway.⁸⁵ Prior studies established a genetic link between the GET pathway and Hsp90,2, 63 but to our knowledge, a direct role for Cdc37 in the GET pathway has not been identified.

We previously reported that Hsp90 mutants in the different groups have differential effects on established clients. Here we identified additional proteins that may be selectively impacted by mutation, but further work is needed to demonstrate differential effects on client activity. The process of loading clients onto Hsp90 is complex, and our studies identified multiple proteins that appear to be specifically impacted by disruption of the loading step. The steps that occur after client loading are not well characterized, but prior studies have demonstrated that the dwell time in the different conformations is important.⁸⁶ We previously demonstrated that mutations in the reopening group result in destabilization of Eft1, which has a critical role in ribosome translocation,¹⁸ and our results here with Hsc82-Q380K reinforce those findings. Eft1 is known to interact with a different set of cochaperones (Cns1, Cpr7, and Hgh1),⁸⁷ and it is possible that Hsc82-Q380K and additional reopening mutants adopt a conformation that disrupt interactions with those cochaperones. Prior studies demonstrated that post-translational modifications impact Hsp90 conformation, cochaperone interaction, and the interactome,88, 89, 90 and further studies are required to understand how those changes impact client interaction. A recent article alternatively describes the Hsp90 machine as “a multifunctional Swiss Army knife, an automobile engine and its controls, a switchboard acting as a hub and directing signals, and an orchestra conductor setting the tempo of a symphony,” and many “known unknowns” remain.91, 92 Our results that Hsp90 mutants have different impacts on the proteome suggest that it may be possible to separate out some of these diverse functions of Hsp90. Further studies are needed to determine how Hsp90 or cochaperone mutation affects protein activity and/or aggregation, since our studies are limited to effects on the abundance of soluble proteins.

While further studies are required to validate that proteins that exhibit reduced accumulation in these studies are clients, our studies identified proteins that appear to be most strongly affected by Hsc82-Q380K or Hsc82-G424D. Further analysis of these proteins may identify new ways to target Hsp90 function and provide new mechanistic insights into how Hsp90 facilitates the folding of diverse clients. Some of the proteins differentially affected by the Hsc82 mutation in this study are essential genes or deletion causes temperature-sensitive growth defects. This suggests that the growth defects in the different mutant strains may be caused by loss of function of different proteins. For example, Hsc82-G309S and Cdc37-S14A caused reduced levels of essential Cdc28, and further inactivation would disrupt the cell cycle. The Cdc37-S14A mutation also caused decreased levels of Rrp7 and Orc4, which have essential roles in ribosome biogenesis and gene regulation. Hsc82-Q380K resulted in loss of Eft1, which, when combined with deletion of its paralog, Eft2, causes inviability due to translation defects.⁸⁷ In addition, three different proteins that play a role in mRNA cleavage and polyadenylation were identified in these studies (Ctk1, Ctk2, and Pti1).⁹³ We also identified multiple Cdc37-dependent proteins that bind NADH/NADPH, as well as proteins that are required for the biosynthesis of NAD (Nma2 and Bna7). Further studies are required to test whether cells expressing different mutations in Hsp90 have specific defects in these essential functions. However, it has been shown that amino acid changes throughout Hsp90 result in altered fitness in response to a range of stressors,⁹⁴ consistent with the concept that Hsp90 alteration selectively impacts function.

Conclusion

We used a combined genetic and proteomic approach to test whether disruption of the Hsp90-client loading step using mutation of two separate proteins results in similar proteomic effects. We observed similar effects in cells expressing Cdc37-S14A or Hsc82-G309S, suggesting that small molecules that target the loading step would elicit a similar proteomic response. We provide further evidence that Hsp90 mutants that disrupt different steps in the folding cycle have differential effects on the proteome. Further analysis of the effects of the other two Hsp90 mutants may provide new mechanistic insight into how Hsp90 interacts with a wide range of cellular proteins that share little sequence of structural similarity.

Materials and methods

Yeast strains and growth assays

Yeast were transformed by lithium acetate methods and were grown in either yeast extract peptone dextrone (YPD) (1% Bacto yeast extract, 2% peptone, and 2% dextrose) or defined synthetic complete media supplemented with 2% dextrose. Growth was examined by spotting 10-fold serial dilutions of yeast cultures on YPD, followed by incubation for 2 days at the indicated temperature. The JJ320 strain (cdc37:kan^r/URA3-CDC37) was transformed with pRS414-based vector expressing either Cdc37-WT or Cdc37-S14A under its own promoter. Strain JJ320 is isogenic to W303 and was derived from strain JJ114. Strain JJ114 and the plasmids expressing Cdc37 were previously described.⁷² Strain JJ816 (hsc82:LEU2 hsp82:LEU2/URA3-HSP82), which is also isogenic to W303, was transformed with plasmids expressing wild-type or mutant His-tagged Hsc82 in pRS313GPD.¹⁶ Transformants were grown on synthetic complete media containing 5-fluororotic acid (5-FOA, GoldBiotechnology) to counter-select for URA3-CDC37 or URA3-HSP82.

Sample preparation

Strain JJ320 expressing either WT Cdc37 or Cdc37-S14A or strain JJ816 expressing either WT Hsc82, Hsc82-G309S, Hsc82-Q380K, or Hsc82-G424D was grown overnight in 25 mL YPD media at 25°C. The following morning, each strain was diluted into two flasks containing 50 mL YPD to approximately 0.10 OD₆₀₀. Cultures were grown at 25°C until the mid-exponential growth phase was reached (0.45-0.65 OD₆₀₀). Unstressed control samples were immediately placed on ice to halt growth. For the heat shock samples, an equal volume of prewarmed YPD was added, and cultures placed in a shaking 37°C incubator for 1 h, then placed on ice. Cell pellets were harvested on ice, washed with cold water, and frozen at −80°. This heat shock was performed as described elsewhere,⁹⁵ except that cultures were prewarmed to 52°C rather than 49°C. Cell pellets were lysed in buffer (20 mM Tris, pH 7.5, 100 mM KCl, 5 mM MgCl₂) containing a protease inhibitor mixture (Roche Applied Science) in the presence of glass beads as described.¹⁷ Yeast extracts were analyzed by sodium dodecyl-sulfate-polyacrylamide gel electrophoresis and Coomassie blue staining to ensure sample quality. Quintuplet samples were shipped to the IDeA National Resource for Quantitative Proteomics at the University of Arkansas for Medical Sciences.

Cdc37 dataset: Orbitrap Eclipse DIA - instrument parameters

Total protein from each sample was reduced, alkylated, and purified by chloroform/methanol extraction prior to digestion with sequencing-grade modified porcine trypsin (Promega). Tryptic peptides were trapped and eluted on 3.5 µm CSH C18 resin (Waters) (4 x 75 µm) then separated by reverse phase XSelect CSH C18 2.5 µm resin (Waters) on an in-line 150 × 0.075 mm column using an UltiMate 3000 RSLCnano system (Thermo). Peptides were eluted at a flow rate of 0.300uL/min using a 60 min gradient from 98% Buffer A:2% Buffer B to 95:5 at 2.0 min to 80:20 at 39.0 min to 60:40 at 48.0 min to 10:90 at 49.0 min and hold until 53.0 min and then equilibrated back to 98:2 at 53.1 min until 60 min. (Buffer A = 0.1% formic acid, 0.5% acetonitrile; Buffer B = 0.1% formic acid, 99.9% acetonitrile). Eluted peptides were ionized by electrospray (2.4 kV) through a heated capillary (275°C) followed by data collection on an Orbitrap Eclipse mass spectrometer (Thermo Scientific).

Precursor spectra were acquired with a scan from 385 to 1015 Th at a resolution set to 60,000 with 100% AGC, max time of 50 msec, and an RF parameter at 40%. Data independent acquisition (DIA) was configured on the Eclipse to acquire 50 × 12 Th isolation windows at 15,000 resolution, normalized AGC target 500%, and maximum injection time 40 ms). A second DIA was acquired in a staggered window (12 Th) pattern with optimized window placements.

Hsc82 dataset: Orbitrap Astral DIA - 40 min - CME

For the Hsc82 dataset, we switched to a newer Orbitrap Astral instrument to increase proteome depth and quantitative reproducibility. Total protein from each sample was reduced, alkylated, and purified by chloroform/methanol extraction prior to digestion with sequencing-grade modified porcine trypsin (Promega). Tryptic peptides were then separated by a reverse-phase Ion-Opticks-TS analytical column (25 cm × 75 µm with 1.7 µm C18 resin) supported by an EASY-Spray nano-source and stabilized with a Heater THOR Controller (Ion-Opticks) at 60°C. Peptides were trapped and eluted from a (PepMap Neo, 300 µm x 5 mm Trap) using a Vanquish Neo UHPLC nano system (Thermo Scientific), which kept the samples at 11°C before injection. Peptides were eluted at a flow rate of 0.350 µL/min using a 35 min gradient from 98% Buffer A:2% Buffer B to 94.5:5.5 at 0.1 min to 56:44 at 27.1 min, followed by a column wash of 45:55 at 29.7 min to 1:99 at 35 min followed by equilibration back to 98:2. (Buffer A= 0.1% formic acid, 0.5% acetonitrile in water, Buffer B = 80% acetonitrile, 20% water, 0.1% formic acid). Eluted peptides were ionized by electrospray (2.5 kV) followed by mass spectrometric analysis on an Orbitrap Astral mass spectrometer (Thermo). Precursor spectra were acquired from 380 to 980 Th, 240,000 resolution, normalized AGC target 200%, and maximum injection time 3 ms. DIA acquisition on the Orbitrap Astral was configured to acquire 199, 3 Th window from 380 to 980 Th, 25% HCD Collision Energy, normalized AGC target 100%, maximum injection time 3 ms. Fragment (MS2) scan range from 150 to 2000 Th with an RF Lens (%) set to 40.

Data analysis: Spectronaut

Following data acquisition, data in the Cdc37 dataset were searched using Spectronaut (Biognosys version 19.1) against the UniProt Saccharomyces cerevisiae database (Proteome ID: UP000002311, Taxon ID: 559292, 4th version of 2024) using the directDIA method with an identification precursor and protein q-value cutoff of 1%, generate decoys set to true, the protein inference workflow set to maxLFQ, the inference algorithm set to IDPicker, the quantity level set to MS2, cross-run normalization set to false, and the protein grouping quantification set to median peptide and precursor quantity. Data in the Hsc82 dataset were analyzed similarly, except that the data were searched using Spectronaut (Biognosys version 20.1) against the UniProt S cerevisiae database (Proteome ID: UP000002311, Taxon ID: 559292, 2nd version of 2025). Fixed Modifications were set to Carbamidomethyl (C) and variable modifications were set to Acetyl (Protein N-term), Oxidation (M). Protein MS2 intensity values were assessed for quality using ProteiNorm.96, 97 The data was normalized using Cyclic Loess (Ritchie et al) and analyzed using proteoDA to perform statistical analysis using Linear Models for Microarray Data (limma) with empirical Bayes (eBayes) smoothing to the standard errors.98, 99, 100, 101, 102, 103, 104, 105 Proteins with an FDR-adjusted P value < .05 and a fold change > 2 (log₂FC > 1.0) were considered significant. Prism Graph Pad was used to generate graphs of expression and significance.

Cross-platform comparison of heat shock responses between the Hsc82 and Cdc37 datasets

To assess cross-platform comparability between the Cdc37 and Hsc82 datasets, we compared the strains expressing wild-type Cdc37 or Hsc82 after 1 h heat shock using the log₂FC values for each dataset. Proteins were matched between studies using UniProt identifiers. To increase confidence in the comparison and minimize potential artifacts arising from missing values or low-confidence quantification, analyses were restricted to proteins detected in at least three of five biological replicates within each dataset and with an adjusted P value < 0.055. Proteins meeting these criteria in both datasets were retained for analysis (n = 1214). The “cor.test(method = “pearson”)” function was used in R studio to perform the two-sided Pearson correlation test analysis to test if the null hypothesis that the true population correlation coefficient equals 0. Scatterplots were generated in R (ggplot2), with the line of perfect correlation (y = x) and linear regression fit shown for visual comparison.

Declaration of interest

The authors declare the following financial interests/personal relationships, which may be considered as potential competing interests. Jill L Johnson reports financial support was provided by the National Institutes of Health. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

Supplementary material Table S1. Lists of proteins in each column shown in Figure 2

.

Supplementary material Table S2. Yeast contain 129 protein kinases.⁴³ The Cdc37 and Hsc82 datasets contained data for 70 and 104 protein kinases, respectively

.

Supplementary material Fig. S1. Cross-platform comparison of heat shock responses between the Hsc82 and Cdc37 datasets. Scatterplot of modeled log₂FC values for the heat shock responses comparing the cells expressing wild-type Hsc82 (x-axis) or wild-type Cdc37 (y-axis). Proteins were restricted to those detected in at least three of five replicates in both conditions and with adjusted p-value < 0.055 to increase confidence in cross-dataset comparability. A total of 1214 proteins met these threshold criteria in both studies. The red line indicates the linear regression fit; the dashed gray line denotes perfect correlation (y = x) for comparison to the fit line. Black dashed vertical and horizontal lines mark the origin (log₂FC = 0). The five proteins with the largest positive and the five with the largest negative inter-dataset differences (Cdc37 − Hsc82) are annotated in bold red. Heat shock responses were highly correlated between datasets (Pearson product–moment correlation r = 0.943, p < 2 × 10⁻¹⁶)

.

Data availability

Data will be made available on request.