|Home | About | Journals | Submit | Contact Us | Français|
The molecular chaperone Hsp90 assists a subset of cellular proteins and is essential in eukaryotes. A cohort of cochaperones contributes to and regulates the multicomponent Hsp90 machine. Unlike the biochemical activities of the cochaperone p23, its in vivo functions and the structure-function relationship remain poorly understood, even in the genetically tractable model organism Saccharomyces cerevisiae. The SBA1 gene that encodes the p23 ortholog in this species is not an essential gene. We found that in the absence of p23/Sba1p, yeast and mammalian cells are hypersensitive to Hsp90 inhibitors. This protective function of Sba1p depends on its abilities to bind Hsp90 and to block the Hsp90 ATPase and inhibitor binding. In contrast, the protective function of Sba1p does not require the Hsp90-independent molecular chaperone activity of Sba1p. The structure-function analysis suggests that Sba1p undergoes considerable structural rearrangements upon binding Hsp90 and that the large size of the p23/Sba1p-Hsp90 interaction surface facilitates maintenance of high affinity despite sequence divergence during evolution. The large interface may also contribute to preserving a protective function in an environment in which Hsp90 inhibitory compounds can be produced by various microorganisms.
Sba1p is the ortholog (4, 15) of the Hsp90 cochaperone p23, a small acidic eukaryote-specific protein, in the budding yeast Saccharomyces cerevisiae (reviewed in references 16 and 46). The molecular chaperone Hsp90 is a highly conserved and abundant cytosolic and nuclear protein that is required for folding, assembly, and maintenance of a subset of proteins (23, 44-46, 62). The activity of its N-terminal ATPase domain is regulated by several cochaperones. Although the biochemical function of ATP hydrolysis for Hsp90 substrates is not understood, genetic experiments in budding yeast unambiguously demonstrated that it must be important for at least some substrates that are essential for viability (42). p23 binds the ATP-bound conformation of the molecular chaperone Hsp90, inhibits ATP hydrolysis, and, as a result of stabilizing the ATP-bound state, prolongs the interaction between Hsp90 and many of its substrates (11, 24, 26, 32, 33, 50-52, 56, 58, 60). The effects of Hsp90 inhibitors such as geldanamycin (GA) and radicicol, which compete with ATP for binding, are compounded by interfering with the binding of p23/Sba1p (15, 26). The very recently reported crystal structure of the Sba1p-Hsp90 complex shows intimate contacts involving multiple regions of Sba1p and both the N-terminal and middle domains of Hsp90. In the complex, which consists of two Sba1p monomers per Hsp90 dimer, Sba1p favors an Hsp90 conformation with the lid of the ATP binding pocket in its closed conformation, providing an explanation for the stabilizing effects of Sba1p (2).
Despite the regulatory interactions between p23/Sba1p and Hsp90, only Hsp90 is absolutely essential for viability. Deletion mutants of the p23 orthologs in budding and fission yeasts are viable (4, 15, 38). Similarly, p23-null mice initially develop relatively normally before dying at birth because of retarded lung development (22). Overall, the in vivo functions of p23/Sba1p remain poorly understood. For S. cerevisiae, early reports of a role for SBA1 in the “general control response” to amino acid starvation (14) and in maintaining chromosome stability (39) were not further investigated. Most of the reported defects of Δsba1 cells relate to the functions of vertebrate substrates of Hsp90 assayed in this heterologous environment (4, 7, 8, 13, 15, 17, 20, 28, 40). Indeed, the very name of the gene SBA1 (sensitivity to benzoquinone ansamycins) derives from the fact that mammalian steroid receptors are more sensitive to inhibition by Hsp90 inhibitors of this chemical class when tested in a Δsba1 strain, but genuine yeast functions or proliferation were not examined in this initial report (4). An essential role of Sba1p in maintaining telomeres by promoting dynamic interactions between the telomerase and telomeric repeats has only very recently been recognized (59). This might explain the previously reported chromosome instability in cells overexpressing Sba1p. However, it is not understood why this Sba1p requirement, while manifested immediately following the deletion of the SBA1 gene, is somehow compensated for seemingly well upon more long-term establishment of Δsba1 strains.
Hence, the role of Sba1p for yeast biology itself and the contributions of different Sba1p domains and functions remain poorly understood. For example, the relevance of the Hsp90-independent molecular chaperone function, which has been described for human p23 (5, 19, 61), remains unclear. It may contribute to the maturation of specific Hsp90 substrates (40), but its importance for endogenous yeast processes has not been addressed. We therefore set out to identify a new phenotype for S. cerevisiae strains lacking Sba1p and to characterize the role of Hsp90 regulation and Sba1p chaperone activity for this phenotype.
The strain BY4742 (MATα his3Δ1 leu2Δ0 lys2Δ0 ura3Δ0) (6) was used as the wild-type strain. This strain, and its Δsba1 (MATα his3Δ1 leu2Δ0 lys2Δ0 ura3Δ0 sba1Δ::kanMX4), Δhsc82, and Δhsp82 derivatives were obtained from Research Genetics and used to generate the Δpdr5 derivatives BYP1 (BY4742 pdr5Δ::HIS3) and BYP2 (BY4742 pdr5Δ::HIS3 sba1Δ::kanMX4), respectively. To produce a PCR product to substitute the HIS3 coding region for that of the PDR5 gene by homologous recombination, HIS3 sequences were amplified from plasmid pRS313 using the two primers 5′-agacccttttaagttttcgtatccgctcgttcgaaagactttagacaaaaatgACAGAGCAGAAAGCCCT-3′ and 5′-tctttcggacattgaactttgatttatcagagctggtaaaACTCAACCCTATCTCGGTCT-3′. The lowercase letters represent flanking sequences of the PDR5 open reading frame, including the ATG (boldface) on the 5′ side and nucleotides +4686 to +4725 located about 160 bp 3′ of the stop codon. Uppercase letters correspond to sequences matching the beginning of the HIS3 open reading frame and sequences downstream of it in plasmid pRS313.
Wild-type Sba1p and its truncation and point mutants were expressed in yeast as Flag-tagged derivatives with plasmid pYES/Flag (1). This plasmid is a URA3-marked 2μm episome with the galactose-inducible promoter GAL1. SBA1 sequences generated by PCR were inserted between the BamHI and EcoRI sites of pYES/Flag. For the introduction of single point mutations and deletion of codons 102 to 113, SBA1 sequences were first cloned into BluescriptM13+ and mutagenized by the QuikChange site-directed mutagenesis method (Stratagene) as directed by the manufacturer.
For growth assays, strains were precultured in minimal selective medium with 2% glucose. Cultures were then washed and diluted to an optical density at 600 nm of 0.1 and serially diluted four times. Four microliters of each dilution was spotted in a row on plates with minimal selective medium containing 2% galactose. Plates were supplemented with radicicol or GA from 10 mM stocks in ethanol and dimethyl sulfoxide (DMSO), respectively, or with solvent alone. Images were acquired after growth for 5 days at 30°C.
All Flag-tagged variants of Sba1p were expressed in strain BYP2. Transformants were precultured overnight in selective minimal medium with 2% glucose, washed once with water, and then incubated for 6 h in yeast extract-peptone (YEP) with 2% galactose. Cell pellets of 150 μl were resuspended in the same volume of lysis buffer (20 mM Tris-HCl [pH 8.0], 1 mM EDTA, 100 mM NaCl, 20 mM Na2MoO4, 20% glycerol) containing a yeast protease inhibitor cocktail as recommended by the supplier (Sigma). Cells were mechanically broken with a Mini-BeadBeater (BioSpec Products) using 500 μl of 0.5-mm zirconia-silica beads. For cell lysis, tubes were placed in a precooled Mini-BeadBeater aluminum rack and violently shaken for 5 min at 4°C. The crude extracts were cleared by centrifugation at 45,000 rpm for 30 min at 4°C, and protein concentrations were adjusted to 1, 3, and 5 mg/ml with lysis buffer for the experiments of Fig. Fig.2B2B and and3,3, ,2C,2C, and and6,6, respectively.
Flag-tagged proteins were immunoprecipitated from crude yeast extracts as follows. One milligram (experiments of Fig. Fig.2B2B and and3)3) or 5 mg of protein extracts was supplemented with ATP to 10 mM and Triton X-100 to 0.2% and incubated with 20 μl of packed EZview red anti-Flag M2 affinity gel (Sigma) for 3 h at 4°C. The resin was collected and washed four times with the same buffer (lysis buffer with protease inhibitors, ATP, Triton X-100) unless indicated otherwise. Wash steps included one change of tubes to ensure removal of Hsp90 that might be nonspecifically bound to the plastic wall. The resin-bound immunoprecipitates were incubated with 1× LDS protein sample buffer (Invitrogen) for 10 min at 50°C before eluted protein supernatants were adjusted to 100 mM dithiothreitol (DTT) and boiled for 10 min. The coimmunoprecipitation (co-IP) experiment for testing the release of Hsp90 in the presence of radicicol was done as follows. The resin-bound immunoprecipitates were split into three parts after the third wash: one part was eluted and processed as mentioned above, and the two other parts were incubated in IP buffer containing either 10 mM ATP or 10 μM radicicol at 4°C for 20 min and then washed three more times without ATP. Protein samples were resolved on 4 to 12% NuPAGE gels (Invitrogen) with morpholineethanesulfonic acid (MES) buffer and analyzed by immunoblotting either with a mouse monoclonal anti-Flag M2 antibody (Sigma) or with a chicken polyclonal antibody preparation raised against yeast Hsp82p (30).
The wild-type and point mutant Sba1p proteins were expressed without any tag in Rosetta (Novagen) bacteria using pET expression constructs. The proteins were purified essentially as described for human p23 (19). Briefly, crude clarified protein extracts were resolved over DEAE FF resin (GE Healthcare) using a linear gradient of 200 to 650 mM NaCl. The appropriate fractions were pooled, diluted threefold with TEN0.0 buffer (10 mM Tris-HCl [pH 6.9], 1.0 mM EDTA, 1.0 mM phenylmethylsulfonyl fluoride [PMSF], 1.0 μg/ml pepstatin, 1.0 μg/ml leupeptin, 0.25 μM aprotinin), and resolved over ResourceQ resin (GE Healthcare) with a linear gradient of 100 to 400 mM NaCl. The appropriate fractions were pooled, the total volume was reduced to ~200 μl using a microconcentrator device with a 10-kDa cutoff (Amicon Ultrafree-4; Millipore), and resolved over a Superdex-200 gel filtration column (GE Healthcare) equilibrated in TEN0.1 buffer (10 mM Tris-HCl [pH 6.9], 1.0 mM EDTA, 100 mM NaCl).
An N-terminal polyhistidine-tagged Hsp82p fusion protein was expressed in Rosetta (Novagen) bacteria using a pET expression construct (pET-His6HSP82), which was a kind gift of J. Buchner (Technische Universität, Garching, Germany). To purify the protein to near homogeneity, a crude clarified protein extract was batch adsorbed to Talon metal affinity resin (Clontech). The resin was washed with 1× Talon binding/wash buffer (50 mM sodium phosphate [pH 7.0], 300 mM NaCl, 10% glycerol, 1.0 mM PMSF, 1.0 μg/ml pepstatin, 1.0 μg/ml leupeptin, 0.25 μM aprotinin), and adsorbed protein was eluted with 1× Talon elution buffer (50 mM sodium phosphate [pH 7.0], 150 mM imidazole, 300 mM NaCl, 10% glycerol, 1.0 mM PMSF, 1.0 μg/ml pepstatin, 1.0 μg/ml leupeptin, 0.25 μM aprotinin). As indicated above for Sba1p, the eluted sample was diluted threefold with TEN0.0 buffer, resolved over ResourceQ resin, concentrated, and put over a Superdex-200 gel filtration column in TEN0.1 buffer with protease inhibitors.
The protein-refolding assays were performed as described previously (18). Briefly, firefly luciferase (Sigma) was prepared at 4 mg/ml in glycylglycine (pH 7.4) and denatured at 37°C for 30 min after dilution into unfolding buffer (25 mM HEPES [pH 7.5], 50 mM KCl, 5.0 mM MgCl2, 5.0 mM mercaptoethanol, 6 M guanidinium-HCl). For the molecular chaperone activity assays, a 1.0-μl aliquot (312.5 ng) of the denatured luciferase solution was added to 124 μl of refolding buffer (25 mM HEPES [pH 7.5], 50 mM KCl, 5.0 mM MgCl2, 10 mM DTT, 1.0 mM ATP) supplemented with either 3.2 μM bovine serum albumin (BSA) or wild-type or point mutant Sba1p. The reactions were incubated for 30 min at 37°C. Refolding was triggered by the addition of Hsp70 and Hdj1 to 1.6 μM and 3.2 μM, respectively, and allowed to proceed for 2 h. Aliquots (1.0 μl) were removed at the indicated time points and mixed with 49 μl of Bright-Glo luciferase assay reagent (Promega), and luciferase activity was determined using an Ultra Evolution plate reader (Tecan). The percent luciferase activity was calculated using an equivalent amount of native luciferase diluted to the same extent in refolding buffer supplemented with 3.2 μM BSA.
The ATPase activity of Hsp82p was measured using the EnzChek phosphate assay kit (Molecular Probes) as previously described (9, 52). Briefly, 1 μM of Hsp82p per monomer was supplemented with the indicated levels of Sba1p or point mutants along with 1 mM ATP in the presence or absence of 60 μM radicicol in buffer C (40 mM HEPES [pH 7.5], 150 mM KCl, 5 mM MgCl2, 2 mM DTT) in a total volume of 150 μl. The reactions were incubated at 37°C for 90 min and then diluted fivefold into a 100-μl phosphate assay as per the manufacturer's instructions. A standard curve was generated using the inorganic phosphate solution provided by the manufacturer. The radicicol control samples serve to confirm that phosphate release is Hsp90 dependent and, of course, Sba1p by itself displays no ATPase activity at all.
Limited proteolytic digestion of the Sba1p proteins was accomplished by incubating them (10 μg each) with 50 ng of trypsin in TEN0.1 buffer (10 mM Tris-HCl [pH 6.9], 1.0 mM EDTA, and 100 mM NaCl). Aliquots were removed after 0 and 15 min and resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and the products were visualized by Coomassie blue staining.
Increasing levels of recombinant Hsp82p (1 to 1,000 nM) were incubated alone or with the indicated Sba1p protein (16 μM) in 12.5 μl binding buffer (40 mM HEPES-KOH [pH 7.5], 150 mM KCl, 5.0 mM MgCl2, 2.0 mM DTT, 200 μM ATP). The chaperone reactions were mixed with 12.5 μl fluorescein isothiocyanate (FITC)-GA (InvivoGen) (8.0 nM FITC-GA, 40 mM HEPES-KOH [pH 7.5], 150 mM KCl, 5.0 mM MgCl2) in a low-binding, black, 96-well plate (HE Microplate; Molecular Devices) and incubated at 4°C for 120 min. The anisotropy values were determined using an Ultra Evolution plate reader (Tecan, Inc.). The apparent Hsp82 affinities for GA in the absence or presence of Sba1p protein derivatives were determined by fitting a curve to the data of four replicates according to the formula where m1 = calculated dissociation constant (Kd); m2 = total [FITC-GA]; and m3 = the maximal change in anisotropy, which is set for each experiment (59).
The isolation and culture of mouse embryo fibroblasts (MEFs) from the wild type and corresponding p23 gene disruption mouse line were described before (22). For proliferation assays, cells were seeded into six-well plates in triplicates at a density of 100,000 cells per well (about 20% confluence). About 12 h later, on day 0, the number of cells was verified with a CASY cell counter model TT and 40 nM GA was added. Cells were counted 12 h later (day 0.5) and thereafter at 24-h intervals. To determine the effects of GA on the levels of Raf-1, cells were seeded at 4 × 106 cells per well of a six-well plate 4 h prior to treatment with 60 nM GA. After several time intervals, cells were lysed in a mixture containing 10 mM Tris-HCl (pH 8.0), 137 mM NaCl, 2 mM EDTA, 10% glycerol, 1% Triton X-100, and protease inhibitors. Equal amounts of total protein were analyzed by immunoblotting with an anti-Raf-1 polyclonal antibody (C-12 from Santa Cruz). The same filter was probed with the monoclonal antibody ab8245 against GAPDH (glyceraldehyde-3-phosphate dehydrogenase; Abcam), which served as a loading control.
To explore the in vivo functions of Sba1p in the budding yeast S. cerevisiae, we screened for specific growth defects of a Δsba1 strain on a panel of different growth media and in the presence of different drugs. The only defect that we could observe with this simple assay was a hypersensitivity of the Δsba1 strain to the Hsp90 inhibitors radicicol and GA (Fig. (Fig.1A;1A; see also Fig. Fig.3D).3D). High concentrations of radicicol are also inhibitory, albeit weakly, to the isogenic wild-type strain, but the deletion of the PDR5 gene, which encodes a membrane-associated drug export pump, further accentuates the hypersensitivity of cells that lack Sba1p. In yeast, radicicol consistently seems to be slightly more potent than GA (48; data not shown). In all subsequent in vivo experiments with yeast, we therefore used a Δpdr5 strain and increased radicicol sensitivity as the Δsba1 phenotype.
In view of the extensive conservation of the Hsp90 chaperone machine, we wished to assess the generality of the protective function of p23/Sba1p newly identified in yeast. Having previously generated a p23-null mouse line (22), we had access to both wild-type and p23-null MEFs. Cells were seeded at a low density (about 20%) to allow for vigorous growth over a period of several days. In the absence of GA, p23-null cells grew slightly more slowly than wild-type cells as previously reported by others (31). However, in the presence of a relatively low dose of GA, p23-null cells rapidly stopped growing and died, whereas wild-type p23 cells even continued to grow for about 2 days and survived much longer (Fig. (Fig.2).2). Numerous studies have been published demonstrating that Hsp90 inhibitors induce apoptosis in a wide variety of different cell types (see, for example, references 10, 35, and 62). It is therefore very likely that our MEFs died by apoptosis, but we have not confirmed this assumption directly. The dramatic difference between the effects of GA on wild-type versus p23-null cells supports the notion that p23 fulfills a protective role against Hsp90 inhibitors in mammalian cells.
S. cerevisiae expresses two Hsp90 isoforms, Hsp82p and Hsc82p, which needs to be considered in exploring the phenotype of Δsba1 cells. The amino acid sequences of the two isoforms are 95% identical (674 out of 709 amino acids). Although half of the differences map to the evolutionarily poorly conserved and dispensable (30) charged domain between the N-terminal and middle domains, there is no direct evidence in the literature that Sba1p interacts with Hsc82p, with the exception of a reference to unpublished results (15). Therefore, we decided to verify experimentally whether Sba1p interacts with both isoforms. Sba1p with an N-terminal Flag tag (Fig. (Fig.3A)3A) was expressed in a wild-type strain and in strains lacking either one or the other Hsp90 isoform. A co-IP experiment revealed that Sba1p interacts similarly with both Hsc82p and Hsp82p (Fig. (Fig.3B3B).
To investigate the mode of action of Sba1p in vivo, we generated a panel of Sba1p truncation mutants and analyzed them for Hsp90 binding and complementation of radicicol hypersensitivity. Both wild-type and mutant proteins were expressed with an N-terminal Flag tag under the control of the galactose-inducible GAL1 promoter. Neither the Flag tag nor overexpression has any effect on complementation. Note that we do not see any toxic effects of overexpression of Sba1p on growth in the strains used in this study (data not shown), despite a previous report of others working with strains in a different genetic background at slightly elevated temperatures (40). Of the four mutants shown in Fig. Fig.3A,3A, only the Δ131-216 C-terminal truncation mutant is stably expressed. This mutant lacks the poorly conserved and apparently unstructured, highly acidic tail, which may carry or contribute to the chaperone activity (60, 61). As judged by the co-IP experiment of Fig. Fig.3C,3C, the portion of Sba1p comprising the N-terminal 130 amino acids is sufficient to bind Hsp90. This observation agrees with previous reports, in which the C-terminal tail of the human Sba1p ortholog p23 was found to be required for chaperone activity but not Hsp90 binding (36, 60, 61). Removal of the N-terminal 15 amino acids of Sba1p or of the internal p23 signature motif WPRLTKEK severely destabilizes the protein. Even though the low levels of these truncation mutants do not allow a direct comparison to the full-length Sba1 construct in the co-IP experiment, the results do suggest that the N-terminal truncation and internal deletion mutants may also be defective in Hsp90 binding. The growth assay displayed in Fig. Fig.3D3D demonstrates that the N-terminal 130 amino acids of Sba1 (sba1Δ131-216) are sufficient for conferring radicicol resistance to a Δsba1 strain. This is evident both at normal temperature (30°C) and under mild heat stress (37°C), which sensitizes the cells to Hsp90 inhibitors. Indeed, the N-terminal domain consistently complements even better than the wild type in the presence of the Hsp90 inhibitors radicicol and GA.
The aforementioned truncation experiments suggested that the N-terminal 15 residues and the signature motif WPRLTKEK might be important for folding and/or activity of Sba1p. We decided to mutate systematically Sba1p residues in these two conserved patches of the N-terminal domain and a few additional conserved ones just C terminal to the signature motif and to assess their ability to bind Hsp90 and to complement a Δsba1 strain (Fig. (Fig.4).4). The two proline residues within these regions were not altered to avoid major structural perturbations. In contrast to some of the truncation mutants, the point mutants are expressed at similar levels compared to wild-type Sba1p, with the exception of W104A, which is consistently slightly less abundant (Fig. (Fig.4A)4A) (data not shown). Co-IP experiments were performed to explore the interaction with Hsp90 in whole-cell extracts. In the yeast system, a co-IP of Hsp90 with Sba1p can only be observed in the presence of ATP (15, 25; data not shown). During the incubation with anti-Flag resin, ATP was therefore present in the buffer to favor assembly and maintenance of Sba1p-Hsp90 complexes. Immunoprecipitates were then washed with a buffer either without ATP (Fig. (Fig.4A,4A, upper panels labeled IP1) or with ATP (lower panels labeled IP2). Two Sba1p mutants (W12A and W104A) are severely defective in Hsp90 binding. In the absence of ATP during washing, two more mutants (A13S and K113A) lose Hsp90 almost quantitatively, whereas four other mutants (Q14A, R15A, R106A, and L107S) are intermediate in retaining Hsp90.
When the same mutants were tested for their ability to complement a Δsba1 strain, several classes became apparent (Fig. (Fig.44 and see Table S1 in the supplemental material). Those that are unable to bind Hsp90 with high affinity (W12A and W104A) failed to confer radicicol resistance. L107S and Y116A bind (albeit L107S binds somewhat less well) and yet do not complement. A13S and K113A, which display some reduction in Hsp90 binding, along with R106A, which binds Hsp90 almost like the wild type, appear to complement slightly better than the wild type. No complementing mutant was found that could not bind Hsp90 at all. We tentatively concluded at this point that some Hsp90 binding is necessary but not sufficient for protection against radicicol.
In an attempt to correlate the biochemical and in vivo complementation activities of the Sba1p point mutants, we performed in vitro assays with recombinant proteins. We selected a set of mutants that represent the three most distinct classes of mutants (see Table S1 in the supplemental material). Wild-type and mutant Sba1p proteins were produced in bacteria, purified to homogeneity, and analyzed for two known p23 family functions—in vitro chaperone activity and inhibition of the Hsp82p ATPase. The molecular chaperone capacities of the various Sba1 proteins were determined utilizing an established in vitro assay. It assesses the ability of the test protein to maintain chemically denatured firefly luciferase in a folding-competent state for refolding by the molecular chaperone couple Hsp70-Hdj1 (18). Thirty-eight percent of the total luciferase activity could be recovered in the presence of wild-type Sba1p (Fig. (Fig.5A).5A). Note that C-terminal truncation mutants of Sba1p are completely defective for this function and no better than BSA at maintaining luciferase for refolding (36, 60, 61; data not shown) and that this assay is done in the complete absence of Hsp90. In contrast, the point mutants retain some, albeit reduced, chaperone activity and fall into two general activity classes: W12A, A13S, W104A, and Y116A show ~60% of wild-type activity, whereas R106A and K113A only retain ~32%. It is interesting that all of these alterations reduce chaperone activity of Sba1p despite the fact that they map to predicted surface residues located outside of the acidic tail, which is required for chaperone activity. However, the reduced chaperone activities of the point mutants do not correlate with their in vivo activities.
The experiments of Fig. Fig.4A4A indicated that several Sba1p point mutants are reduced in their ability to co-IP Hsp90 from crude extracts, albeit to different degrees. The relative interaction affinities of W12A, A13S, W104A, R106A, and K113A were further explored with purified components by monitoring the Sba1p-dependent inhibition of the ATPase activity of Hsp82p. As expected, we found that wild-type Sba1p reduces the ATP hydrolysis rate of Hsp82p to ~50% (Fig. (Fig.5B),5B), which is consistent with a stabilization of the ATP-bound form (9, 33, 52, 56). Using the observed ATPase rates and Sba1p titration, we calculated an Sba1p-Hsp82p dissociation constant of ~2.2 μM, which is in good agreement with a prior report of ~1.8 μM (56). Three of the point mutants, A13S, R106A, and K113A, functionally interact with Hsp82p, albeit with lower affinities of ~33.1, 21.5, and 4.2 μM, respectively (Fig. (Fig.5B).5B). In contrast, W12A and W104A are apparently unable to inhibit the ATPase activity of Hsp82p (Fig. (Fig.5C).5C). In fact, increasing concentrations of these two mutants even stimulated the ATPase rate, suggesting an alternative mode of interaction with or effect on Hsp82p. Thus, the ability to bind and to regulate the Hsp90 ATPase activity is necessary for complementation of the Δsba1 phenotype. The A13S, R106A, and K113A mutants both inhibit the Hsp90 ATPase in a test tube and protect against high radicicol concentrations in living cells. The mutant Y116A, which has an affinity that is even slightly higher (1.1 μM) than that of wild-type Sba1p, appears to be a special case and will be further discussed below. Importantly, the Y116A results point out that binding and inhibition of the ATPase are not sufficient for in vivo complementation.
Unlike other mutants, W12A and W104A had proved to be defective for binding and regulating Hsp90 and for conferring radicicol resistance to cells, despite the fact that W12 and W104 do not appear to be part of the recently published Sba1p-Hsp82p interface (2). In order to further our understanding of the differences between the two classes of mutants, their global structures were investigated by limited proteolysis. With light trypsin treatment, we observed a slight digestion of wild-type Sba1p and the point mutants A13S, R106A, and K113A and the appearance of a faster-migrating species (~25 kDa) (Fig. (Fig.5D).5D). In contrast, W12A and W104A were almost completely digested and only small digestion products (<5 kDa) were apparent. Hence, the W12A and W104A mutants adopt distinct or loosened conformations relative to the other Sba1p mutants. Because of their loosened conformation, Hsp90 might now recognize them as substrates, albeit with a poor binding affinity, which may explain the stimulatory effect on the ATPase. It is noteworthy that W12A and W104A are clearly not grossly misfolded since they accumulate to appreciable levels both in Escherichia coli and in yeast and retain chaperone activity. Their structural and biochemical defects, however, appear to compromise their ability to complement Δsba1 cells.
It appeared at this point in our study that Sba1p could protect Hsp90 from inhibitors. Complex formation but not the intrinsic chaperone activity of Sba1p appeared to be necessary, albeit not sufficient, for this function. This called for a direct test of whether Sba1p prevents an inhibitor from binding Hsp90. The effect of various Sba1p derivatives on the binding affinity of Hsp82p for a fluorescein-labeled version of GA (FITC-GA) was therefore determined using fluorescence anisotropy (Fig. (Fig.6).6). Strikingly, wild-type Sba1p and the complementing mutants A13S, R106A, and K113A impair GA binding by twofold or more, whereas the complementation-defective mutants L107S and Y116A barely have any effect at all.
Intriguingly, despite the fact that Y116A is able to bind Hsp90 with high affinity, Y116A fails to protect Hsp90 against GA in vitro and radicicol in vivo, and its interaction with Hsp90 is insensitive to radicicol. Y116A can readily be coimmunoprecipitated with Hsp90 from crude yeast extracts supplemented with ATP, but then, unlike wild-type Sba1p, retains Hsp90 even after addition of radicicol (Fig. (Fig.7)7) or GA and various GA derivatives (data not shown). All of these biochemical experiments suggest that this unusual mutant, while being able to inhibit the Hsp90 ATPase, cannot prevent inhibitors from binding and possibly even binds in the presence of inhibitors. Whether the other mutant, L107S, which retains some ability to bind Hsp90 but does not complement, shares additional biochemical characteristics with Y116A remains to be further analyzed.
A prediction from all of these experiments is that Hsp90 substrates should be more vulnerable to GA-induced degradation in the absence of p23/Sba1p. While these experiments are difficult to perform because of the large-scale effects of GA treatment on cell growth and survival, we decided to try to illustrate this point by examining at least one model substrate. Figure Figure88 shows that the endogenous Hsp90 substrate Raf-1 (54) is degraded more rapidly at a given GA concentration in MEF cells devoid of p23.
We have revisited the phenotype of S. cerevisiae cells lacking the SBA1 gene and discovered that they display an increased sensitivity to Hsp90 inhibitors. Using this phenotype, we have determined which of the known Sba1p activities are important for complementation. Hsp90 binding and inhibition of the ATPase activity are required, whereas the intrinsic Sba1p chaperone function is dispensable for conferring relative resistance to Hsp90 inhibitors. The results with one particular Sba1p point mutant, Y116A, indicate that the abilities to bind and to inhibit the ATPase are not sufficient for protection and that inhibition of ATPase and protection against Hsp90 inhibitors are two distinct activities of p23/Sba1p. Thus, with the exception of the intrinsic chaperone activity of Sba1p, all of the assessed biochemical activities are necessary to protect cells. While all protective mutants qualitatively retain these activities, there are some quantitative differences that may be interesting to follow up. Alterations of other components of the Hsp90 complex, notably Hsp90 itself, Sti1, Cpr7m, and Cdc37, had previously been reported to render cells more sensitive to Hsp90 inhibitors (3, 12, 48, 49, 57). In contrast to these cases, we show that the hypersensitivity of cells lacking Sba1p/p23 is not merely due to a synthetic effect but can also be attributed to the loss of a direct protective effect of cochaperone binding.
S. cerevisiae Δsba1 strains can be established and appear to be fully viable. Although the loss of SBA1 compromises the maintenance of telomeres (59) and may lead to genomic instability (39) and a defective response to amino acid starvation (14), possibly in a strain-specific fashion, cells eventually seem to adapt to life without Sba1p and grow as well as wild-type cells (4, 15) (Fig. (Fig.1).1). In the fission yeast Schizosaccharomyces pombe, the SBA1 ortholog WOS2 is required for thermotolerance (38), but there is no indication that this is the case in S. cerevisiae. A previous survey of factors that might influence the sensitivity to Hsp90 inhibitors had missed Sba1p (48), most likely because their Δsba1 strain was not deleted for PDR5. Similarly, of two large-scale chemical genetic screens with Hsp90 inhibitors, one missed Δsba1 (64), whereas the other did score Δsba1 as a hit, but did not further highlight or discuss this finding (43). Although we have observed this hypersensitivity in two different strain backgrounds (data not shown), we cannot exclude that there is a strain-specific penetrance of this trait. The deletion of the gene encoding the export pump Pdr5p clearly sensitized our strains to Hsp90 inhibitors and facilitated the display of differential sensitivities (Fig. (Fig.1).1). Remarkably, we found that p23 plays a similar protective role in MEFs, suggesting that this may be a conserved function of p23 beyond its regulatory functions as an Hsp90 cochaperone.
In contrast to the molecular chaperone function of Sba1p, the ability of Sba1p to bind Hsp90 and to inhibit its ATPase seems to be crucial, albeit not sufficient, for protecting cells against Hsp90 inhibitors. The biochemical correlate of this phenomenon is likely to be the equilibrium between the Sba1p-Hsp90 complex that is favored by ATP binding of Hsp90 and the inhibitor-bound Hsp90 that is unable to bind Sba1p (24, 26, 50, 60). This suggests that the binding of Sba1p to the Hsp90-ATP complex might protect Hsp90 from binding an inhibitor molecule, which is compatible with the tight and intimate nature of the ternary complex (2) and the ability of human p23 to stabilize the nucleotide-bound state of Hsp90 (58). Most importantly, we have demonstrated that Sba1p directly reduces the affinity of Hsp90 for GA. Thus, if Sba1p reduces the affinity of Hsp90 for inhibitors and pushes the equilibrium toward the ATP-bound form, it might explain how Sba1p can exert this effect despite being about 10-fold less abundant than Hsp90. In the case of yeast, it is interesting to speculate that this protective function of Sba1p might be of importance in a natural environment where Hsp90 inhibitors such as radicicol and GA are produced by other microorganisms. There is indeed evidence for this type of pharmacological interaction between species. A rhizosphere fungus has been shown to produce large amounts of Hsp90 inhibitors that can influence Hsp90 and the stress response in higher plants associated with this fungus (34).
In an apparent contradiction to our findings with yeast, it has been suggested that cancer cells have a higher affinity for GA because their Hsp90 complexes are more extensively associated with cochaperones such as p23 (27). However, this view has recently been questioned (21), and, therefore, in the light of our findings with yeast and mammalian cells, the influence of p23 levels on the GA sensitivity of normal as well as cancer cells would be interesting to reexamine. Intriguingly, cancer cells, and in particular metastatic cells, were reported to have higher levels of p23 (29, 37), and the overexpression of p23 in MCF7 breast cancer cells stimulates their ability to invade a fibronectin membrane (41). It is tempting to speculate that cancer cells up-regulate p23 both to cope with an increased demand for handling Hsp90 substrates and to protect one of their Achilles heels, Hsp90, from inhibitors that might be produced by metabolism or taken up with food. It is also conceivable that p23 functions including this protective one could be posttranslationally regulated or influenced by the nature of the other cochaperones or even of the substrate in the Hsp90 complex. The phenotype of the Y116A mutant is an indication that even subtle changes of this kind might be able to affect the ability of wild-type p23 to protect Hsp90. In the context of cancer therapy, one might predict that tumor cells could be rendered even more sensitive to GA as an anticancer drug by p23-targeted drugs or small interfering RNA.
The increased growth inhibition and/or induction of cell death in response to Hsp90 inhibitors must be due to one or several Hsp90 substrates that fail. When Bohen originally reported that the deletion of the SBA1 gene renders the mammalian glucocorticoid receptor assayed in such a yeast strain hypersensitive to an Hsp90 inhibitor, he found that the absence of Sba1p had no impact on the accumulation of this exogenous substrate (4). We later confirmed this observation with p23-null MEFs (22). Despite being functionally defective, the levels of the glucocorticoid receptor were found to be similar in p23-null and wild-type cells. Thus, we suspect that the loss of full functionality of Hsp90 substrates in the absence of p23/Sba1p and their GA-induced degradation may be mechanistically distinct events. Nevertheless, our analysis of Raf-1 suggests that Hsp90 substrates are indirectly protected from Hsp90 inhibitor-induced degradation through the protective effect of p23/Sba1p for Hsp90.
Our findings reveal Sba1p biochemical functions and protein surfaces that are required for in vivo complementation. This is the first time that such an extensive mutagenesis analysis has been performed. The reason for focusing our mutagenesis on the two separate regions shown in Fig. Fig.9C9C was that these residues seemed to lie in the same patch on the surface of free p23 (60) (Fig. (Fig.9A).9A). Moreover, previous publications had pinpointed three more residues right next to this patch as being important for Hsp90 binding: I117 in Sba1p (40) and the equivalent of F121 and W124 in human p23 (63).
Although the published crystal structures for the free form of human p23 (60) and Hsp90-bound Sba1p (2) show a largely identical overall fold, there are significant differences in the details that call for a reinterpretation of the currently available structural information (Fig. (Fig.9).9). We have modeled the Sba1p polypeptide sequence onto the human p23 structure to compare the free and bound forms. The comparison between the two structures suggests that Sba1p undergoes structural rearrangements upon binding Hsp90, presumably to favor and to stabilize the tight packing that is apparent in the complex. This is particularly pronounced in the N-terminal portion beginning with W12. In the free form, residues 12 to 15 (in red in Fig. Fig.9)9) pack against the surface in the vicinity of the N-terminal side of the signature motif WPRLTKE, with the remainder looping out. Upon binding Hsp90, these residues are “pulled through” the interior of the structure so that A13 to R15 come to lie far away from W104 and now form part of a large interface, along with K113 and Y116, that packs against the lid of Hsp90. Part of the loop comes to lie inside the structure and even contributes a small β-sheet.
The two structures are interesting to consider in interpreting the defects of the two point mutants W12A and W104A. Whereas the aromatic side chains of the two tryptophan residues are in close proximity in the free form, they move far apart when Sba1p binds Hsp90. In the Sba1p-Hsp90 complex, W12 constitutes the very N terminus of the defined structure, protrudes away from its core, and is not involved in any protein-protein contacts. The fact that the bulky side chain of W104 points inwards might account for the increased protease sensitivity observed with the mutant W104A. The behavior of the W12A mutant, which is similarly loosened and unable to bind Hsp90, may be rationalized differently. It is conceivable that the N-terminal portion of the mutant is not properly tethered to the core of the protein through residue 12, normally in close proximity to hydrophobic L107, and is therefore unable to undergo the large structural rearrangement upon binding Hsp90. Despite a somewhat loosened structure, both W12A and W104A retain chaperone activity. Although this activity depends on the C-terminal domain of Sba1, it is not known whether that domain is sufficient for chaperone activity and to what extent the N-terminal domain contributes.
Surprisingly, our Sba1p mutations that do map to the known Sba1p-Hsp90 interface have either no effect or only a minor effect on Hsp90 binding. A13, Q14, R15, K113, and Y116 are all localized in the large interface that packs against the lid of Hsp90 and consists of Sba1p residues 13 to 16 and 113 to 118 (2), and yet only A13S and K113A have slightly reduced Hsp90 binding both in vitro and in IP experiments performed under certain conditions. Single point mutants may be unable to weaken the interaction sufficiently because of the large size of this particular interface. The specific characteristics of the amino acid substitution in mutant I117N may explain why it is severely defective for Hsp90 binding (40). Y116 is part of the interface, but its aromatic side chain faces away from Hsp90. While this may explain why the Y116A mutant still binds Hsp90, its unusual binding characteristics in the presence of Hsp90 inhibitors could only be understood with more detailed structural information.
Interestingly, none of the residues in this interface is absolutely conserved (Fig. (Fig.9C)9C) (data not shown). While the yeast Sba1p mutant A13S is partially defective for Hsp90 binding, a serine is the natural residue in Drosophila melanogaster. The human p23 mutant K95A, which is equivalent to the Sba1p mutant K113A, is completely defective for Hsp90 binding (47). Moreover, Sba1p can bind both yeast and mammalian Hsp90, but yeast Hsp90 cannot bind mammalian p23 (53). This suggests that a large interface as well as its coordinate evolution may contribute to ensuring sufficient affinity of this protein-protein interaction and functionality for this molecular chaperone complex.
We are very grateful to Natasha Kralli and Keith Yamamoto for yeast strains. We also thank Marc Fischer and Olivier Donzé for their efforts early on in this project and Martin Hessling for preliminary biochemical experiments with one mutant. We thank an anonymous reviewer for thorough editing of the manuscript.
The Freeman lab at the University of Illinois was supported by a grant from the National Institutes of Health (DK074270). Support for the Picard lab at the University came from the Canton de Genève, the Swiss National Science Foundation, and the Fondation Medic.
Published ahead of print on 24 March 2008.
‡Supplemental material for this article may be found at http://mcb.asm.org/.