|Home | About | Journals | Submit | Contact Us | Français|
The heat-shock protein 90 (Hsp90) is a cytosolic molecular chaperone that is highly abundant even at normal temperature. Specific functions for Hsp90 have been proposed based on the characterization of its interactions with certain transcription factors and kinases including Raf in vertebrates and flies. We therefore decided to address the role of Hsp90 for MAP kinase pathways in the budding yeast, an organism amenable to both genetic and biochemical analyses. We found that both basal and induced activities of the pheromone-signaling pathway depend on Hsp90. Signaling is defective in strains expressing low levels or point mutants of yeast Hsp90 (Hsp82), or human Hsp90β instead of the wild-type protein. Ste11, a yeast equivalent of Raf, forms complexes with wild-type Hsp90 and depends on Hsp90 function for accumulation. For budding yeast, Ste11 represents the first identified endogenous “substrate” of Hsp90. Moreover, Hsp90 functions in steroid receptor and pheromone signaling can be genetically separated as the Hsp82 point mutant T525I and the human Hsp90β are specifically defective for the former and the latter, respectively. These findings further corroborate the view that molecular chaperones must also be considered as transient or stable components of signal transduction pathways.
The 90-kDa heat-shock protein (Hsp90)1 (for reviews, see Jakob and Buchner, 1994 ; Csermely et al., 1998 ) is an ubiquitous and abundantly expressed cytosolic protein even at normal temperature. It is highly conserved from bacteria to mammals. Two genes encode closely related isoforms in mammals as well as in the budding yeast Saccharomyces cerevisiae. Deletion experiments in yeast have shown that the expression of at least one of the two Hsp90 isoforms, either Hsp82 or Hsc82, is essential for viability (Borkovich et al., 1989 ). Similarly, many mutant alleles of the Drosophila HSP90 homolog, HSP83, are embryonic lethals over a deficiency of the locus (van der Straten et al., 1997 ), whereas the Escherichia coli homolog of Hsp90, HtpG, appears to be dispensable (Bardwell and Craig, 1988 ). Hsp90 can act as a molecular chaperone in vitro to promote refolding of denatured proteins (Wiech et al., 1992 ; Yonehara et al., 1996 ; see also Shaknovich et al., 1992 ; Shue and Kohtz, 1994 ), to hold denatured proteins in a folding-competent state for other chaperones (Freeman and Morimoto, 1996 ; Yonehara et al., 1996 ) and to prevent protein unfolding and aggregation (Miyata and Yahara, 1992 ; Jakob et al., 1995a , 1995b ; Yonehara et al., 1996 ).
The interaction of Hsp90 with steroid receptors, which can be thought of as a signal transduction complex, has been the most extensively investigated. A variety of in vitro and in vivo studies have revealed that steroid receptors are complexed with Hsp90 and several other proteins in the absence of hormone (for review, see Pratt and Toft, 1997 ). Upon ligand binding, the hormone binding domain (HBD) undergoes a conformational change that results in the release of Hsp90 and the concomitant activation of the steroid receptor. Steroid receptors and many heterologous proteins fused to the HBD are maintained inactive in the absence of hormone. We have therefore hypothesized that the hormone-reversible inactivation function of the HBD is mediated by Hsp90, possibly by steric hindrance (Picard, 1993 , 1994 ). Further insights into the role of Hsp90 in the regulation of this particular signal transduction pathway come from studies made in yeast (reviewed in Picard, 1998 ). Vertebrate steroid receptors expressed in yeast strains with a low level (Picard et al., 1990 ; see also Holley and Yamamoto, 1995 ) or specific point mutants of Hsp82 (Bohen and Yamamoto, 1993 ; Bohen, 1995 ; Nathan and Lindquist, 1995 ; Fang et al., 1996 ) show a defective hormonal response that is due to a decrease in the ligand-binding affinity (Bohen, 1995 ; Fang et al., 1996 ). Thus, Hsp90 may have a dual role: it ensures that receptors are kept inactive in the absence of hormone and helps them to respond specifically and efficiently to ligand. This view is also corroborated by pharmacological in vivo experiments with geldanamycin (Whitesell et al., 1994 ), a compound that interferes with certain Hsp90 functions such as the proper maturation of steroid receptor–Hsp90 complexes (Smith et al., 1995 ; Whitesell and Cook, 1996 ; Bamberger et al., 1997 ; Czar et al., 1997 ; Segnitz and Gehring, 1997 ).
There is ample evidence for a role of Hsp90 in regulating the activity of several other signaling pathways, such as the xenobiotic response mediated by the dioxin receptor (see for example Pongratz et al., 1992 ; Carver et al., 1994 ; McGuire et al., 1994 ; Antonsson et al., 1995 ; Coumailleau et al., 1995 ; Whitelaw et al., 1995 ). Interaction of the dioxin receptor with Hsp90 is essential for ligand binding and for acquiring a DNA-binding conformation. Activation of the dioxin receptor depends on the release of Hsp90 upon ligand binding and heterodimerization with Arnt. A functional dependence on, and a direct interaction with, Hsp90 has also been described for kinases such as the fission yeast Wee1 (Aligue et al., 1994 ), the vertebrate v-Src (Schuh et al., 1985 ; Xu and Lindquist, 1993 ; Nathan and Lindquist, 1995 ), and the related kinase Lck (Hartson et al., 1996 ).
Hsp90 may also be required for growth factor signaling. 1) Raf-1, a serine/threonine kinase involved in mitogenic signal transduction in vertebrates, exists in a geldanamycin-sensitive heterocomplex with Hsp90 (Stancato et al., 1993 , 1994 ; Lovric et al., 1994 ; Wartmann and Davis, 1994 ; Schulte et al., 1995 , 1996 ; Stancato et al., 1997 ). 2) Mutations in Drosophila HSP83 reduce signaling by the torso (Doyle and Bishop, 1993 ) and sevenless receptors (Cutforth and Rubin, 1994 ; van der Straten et al., 1997 ), which may be due, at least in part, to a requirement for Hsp90 for Raf function (van der Straten et al., 1997 ). 3) The insulin receptor binds Hsp90, and antibodies to Hsp90 interfere with insulin signaling (Takata et al., 1997 ).
Comparable MAPK pathways also exist in yeast where they regulate the pheromone response, invasive growth, pseudohyphal development, osmoregulation, cell wall integrity, and sporulation (for reviews, see Herskowitz, 1995 ; Levin and Errede, 1995 ; Schultz et al., 1995 ; Leberer et al., 1997 ). The pheromone-signaling pathway has received a lot of attention over the past few years. Binding of the mating pheromones to transmembrane receptors elicits a series of events including the sequential activation of the kinases Ste11, Ste7, and Fus3, leading to morphological changes, a cell cycle arrest in G1, and the expression of specific genes required for mating. The kinase Ste11 from S. cerevisiae occupies a position analogous to that of Raf. This prompted us to test genetically whether Hsp90 plays a role in the pheromone pathway.
Wild-type Hsp82 (Hsp82 wt), Hsp82 G313N, and Hsp82 T525I were expressed from plasmids pTCA/Hsp82, pTCA/Hsp82 G313N, and pTCA/Hsp82 T525I, respectively (Bohen, 1995 ), or various derivatives thereof with other auxotrophic markers. Unless indicated, the strong constitutive promoter from the glyceraldehyde-3-phosphate dehydrogenase (GPD) gene TDH3 was used to drive expression. Plasmid pHCA/Hsp82 is the HIS3 version of pTCA/Hsp82 obtained by substituting the backbone of shuttle vector pRS313 for that of pRS314 (Sikorski and Hieter, 1989 ). Plasmid p2U/Hsp82, a 2μ-URA3 expression vector for Hsp82 has been described previously (Louvion et al., 1996 ).
To obtain reduced levels of Hsp82 (~10% of Hsp82 + Hsc82 in a wild-type strain), Hsp82 was expressed from a construct containing the leaky GAL1 promoter from strain GRS4 (Picard et al., 1990 ) fused to HSP82 coding sequences in plasmid pRS304 (Sikorski and Hieter, 1989 ). On medium with 2% glucose, repression of this mutant GAL1 promoter construct is only partial, and low levels of Hsp82 accumulate.
Plasmid p2TG/hHsp90β expressing human Hsp90β was constructed as follows. The coding sequence for human Hsp90β was excised as a SnaBI–SalI fragment from pKN1–3 (Rebbe et al., 1987 ) and cloned into the SmaI site of pSP64 to add a BamHI site at the 5′-end and a SacI site at the 3′-end. The BamHI–SacI fragment containing the human HSP90β sequence was fused to the GPD promoter in shuttle vector pRS304 (Sikorski and Hieter, 1989 ) with a 2μ replicon. Plasmid p2HG/hHsp90β is the HIS3 version based on expression vector p2HG (Picard et al., 1990 ).
Plasmids p2G/Hsp82, p2G/Hsp82 G313N, and p2G/hHsp90β are identical to plasmids p2HG/Hsp82 (Louvion et al., 1996 ), p2HG/Hsp82 G313N (the G313N derivative of p2HG/Hsp82), and p2HG/hHsp90β, respectively, except that they lack an internal HindIII fragment of the HIS3 marker. Thus, rather than an auxotrophic marker it is the Hsp90 function itself that provides the selectable marker for these plasmids.
Plasmid p2TG/flag.Hsp82wt serves to express Hsp82 with a FLAG epitope at the N terminus. The expression vector was derived from p2TG/hHsp90β. Sequences encoding the FLAG epitope (DYKDDDDK) were placed between the initiator codon and the second codon of the wild-type HSP82 sequences, following the introduction of a BglII site just upstream of the second nucleotide of the HSP82-coding sequence. FLAG epitope and second amino acid of Hsp82 are thus separated by the three extra amino acids EIL.
Plasmid pYES/Ste11ΔN encoding Ste11ΔN was generated as follows: the coding sequence for the catalytic domain of Ste11 was excised from plasmid pNC199 (a gift from B. Errede) as a DdeI–BglII fragment and subcloned into pSP72 to add a BamHI site at the 5′-end. This fragment was further subcloned as a BamHI–BglII fragment into a pUC18 derivative containing a stop codon in the proper reading frame followed by a SacI site. Finally, the sequence encoding the catalytic domain of Ste11 was introduced into plasmid pYES 2.0 (Invitrogen, San Diego, CA) as a BamHI–SacI fragment. pYES 2.0 is a yeast expression vector that contains the galactose-inducible GAL1 promoter, the 2μ replicon, and the URA3 selectable marker. Plasmid pYES/HA-Ste11 was constructed for galactose-inducible expression of full-length Ste11 with an influenza virus hemagglutinin (HA) epitope (Daro et al., 1996 ) at its N terminus; a KpnI–NdeI fragment with sequences encoding the HA epitope (MQDLPGNDNSTAG) was joined in-frame to a BamHI fragment carrying STE11-coding sequences from plasmid BB345 (mentioned as pYBS345 in Choi et al., 1994 ) and cloned into pYES 2.0 linearized with KpnI and NotI; noncomplementary sites were filled in or chewed back to allow ligation.
Plasmid pUCA/Ste7 M was used to express a myc-tagged Ste7 protein. It contains the CYC1 promoter and Ste7-coding sequences from plasmid pNC318 (Zhou et al., 1993 ) excised as a SalI–HindIII fragment and cloned into the SalI–SmaI linearized plasmid pRS316 (Sikorski and Hieter, 1989 ).
The yeast genomic library (a gift obtained via M. Collart) was a Sau 3A partial library cloned into the BamHI site of the 2μ-URA3 vector YEplac195 (Gietz and Sugino, 1988 ).
Plasmids p2U/GST-2 (Warth et al., 1997 ), p2U/GST-STE5, and pYes/Ste11ΔN.GST served to express glutathione-S-transferase (GST), GST fused to Ste5, and GST fused to Ste11ΔN, respectively. p2U/GST-STE5 was constructed by replacing the BamHI–BglII fragment at the 5′-end of HSP82 of p2U/Hsp82 with a BamHI fragment carrying GST-coding sequences fused in-frame to STE5 sequences; STE5 sequences lacking the first 24 codons were from plasmid BB192 (mentioned as pYBS146 in Choi et al., 1994 ).
The parent strains and some of the derivatives are listed in Table Table2.2. The related yeast strain backgrounds, HH1a and JC6a (gifts from S. Lindquist), were used to replace the endogenous Hsp82/Hsc82 with Hsp90 mutants by plasmid shuffling. Plasmids were introduced into yeast by the LiAc/PEG method and selected for on appropriate minimal media. Strain HH1a-pHCA/Hsp82wt is essentially the MATa version of the previously described strain HH1-KAT6 (see Palmer et al., 1995 ). It was obtained by tetrad dissection of a diploidized HH1-KAT6 and further plasmid shuffling.
The strain DP121 was obtained by substituting the HIS3 coding body for that of FUS1 in strain DP120 (see Table Table2)2) with the gene replacement construct pSL1497 (Stevenson et al., 1992 ). Plasmid p2U/Hsp82 was subsequently replaced by the Hsp90 expression vectors pTCA/Hsp82, pTCA/Hsp82 G313N, and p2TG/hHsp90β, to yield strains DP122, DP123, and DP124, respectively. In strains HH1a-p2G/Hsp82wt, HH1a-p2G/Hsp82 G313N, and HH1a-p2G/hHsp90, the Hsp90 derivatives themselves are used as selectable marker to maintain the episomes.
To monitor the cell cycle arrest in response to α-factor, cells were diluted to a density of 1.2 × 107 cells/ml and streaked or spotted onto YEPD plates containing 10 mM Na-citrate, pH 4.3, and, where indicated, 5 μM α-factor (Bachem, Torrance, CA). The FUS1-LacZ reporter plasmid pSB234 was used to measure the transcriptional output of the pheromone pathway (Trueheart et al., 1987 ). Wild-type and mutant strains were grown to early logarithmic phase and exposed to 5 μM α-factor for 2 h after addition of 10 mM Na-citrate, pH 4.3. Quantification of the LacZ expression was performed as described by Yocum et al. (1984) except that chlorophenol red-β-d-galactopyranoside was used as β-galactosidase substrate instead of O-nitrophenyl β-d-galactopyranoside for more sensitivity.
The levels of overexpressed Ste11 (yeast strains JC6a-Hsp82, JC6a-Hsp82 G313N, and JC6a-hHsp90β with plasmid pYES/HA-Ste11) were quantitated using crude extracts prepared by a rapid protein extraction protocol (Horvath and Riezman, 1994 ) and loaded onto 10% SDS-polyacrylamide gels. To confirm that equal amounts of protein had been loaded, proteins were stained with Ponceau S after transfer onto a nitrocellulose membrane before immunostaining.
JC6a strains expressing the Hsp90 derivatives were transformed with plasmid pUCA/Ste7 M. Transformants were grown to early logarithmic phase in 1% sucrose as a carbon source. After addition of 10 mM Na-citrate, pH 4.3, the cultures were exposed to 5 μM α-factor for 2 h. Cell extracts were prepared at 4°C by breaking the cells with glass beads in 10 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM DTT, 20 mM sodium molybdate, 15 mM MgCl2, 10% glycerol, 1 mM PMSF, the protease inhibitors aprotinin, leupeptin, and pepstatin A, and the phosphatase inhibitors okadaic acid (1 μM), Na2MoO4 (10 mM), Na3VO4 (0.1 mM), and NaF (5 mM). Samples were frozen in liquid nitrogen and stored at −70°C. Extracts, 10 μg each, as determined with the Bio-Rad (Richmond, CA) Bradford reagent, were boiled in SDS sample buffer for 5 min and loaded onto 7.5% SDS-polyacrylamide gels.
GST pull-down experiments were performed as follows. Yeast cells (strain RMY326 with plasmids pYes/Ste11ΔN.GST or p2U/GST-2) were washed once with water containing 1 mM DTT and 1 mM PMSF and once with TEG (25 mM Tris-HCl pH 7.4, 15 mM EGTA, 10% glycerol, 1 mM DTT, 1 mM PMSF, 3 μg/ml chymostatin, 1.5 μg/ml pepstatin A, 0.75 μg/ml leupeptin, 3.8 μg/ml antipain) containing 150 mM NaCl. Cell pellets were then resuspended in a small volume of the same buffer and broken with glass beads by two 30-s pulses at maximum speed in a Mini-BeadBeater-8 (Biospec Products, Bartlesville, OK) at 4°C. After centrifugation at 15,000 rpm in a table top centrifuge at 4°C, the supernatant was quantitated and adjusted to 0.1% Triton X-100. Glutathione-sepharose beads (Pharmacia, Piscataway, NJ) were added to the extracts, tumbled for 30–45 min at 4°C, washed three times with TEG containing 150 mM NaCl, 0.1% Triton X-100 and twice with TEG with 0.1% Triton X-100. Bound proteins were eluted with 7.5 mM reduced glutathione in 50 mM Tris-HCl, pH 8.0, and concentrated by trichloroacetic acid (TCA) precipitation, resuspended in SDS sample buffer, and loaded onto 10% SDS-polyacrylamide gels.
Coimmunoprecipitation experiments using the FLAG tag were done as follows. Extracts from strains HH1a-p2TG/flag.Hsp82wt and HH1a-p2G/hsp82wt with and without plasmid pYES/HA-Ste11 were prepared as described above for the GST pull-down experiments except that the buffer was 10 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM DTT, 10 mM sodium molybdate, 1 mM EDTA, 10% glycerol, 1 mM PMSF, 3 μg/ml chymostatin, 1.5 μg/ml pepstatin A, 0.75 μg/ml leupeptin, 3.8 μg/ml antipain. After adjusting the extracts to 0.1% Triton X-100, they were incubated at 4°C with the anti-FLAG monoclonal antibody M2 (Eastman Kodak, Rochester, NY) for 2 h followed by 1 h with Protein G-sepharose (Pharmacia). Immunoprecipitates were washed four times for 10 min at 4°C with the extraction buffer containing 0.1% Triton X-100, solubilized in SDS sample buffer, and loaded onto 10% SDS-polyacrylamide gels. The same protocol was used for immunoprecipitation by a Ste11-specific rabbit polyclonal antiserum (Cairns et al., 1992 ) of endogenous Ste11 from 0.5 mg of extracts from strains HH1a-p2G/Hsp82wt, HH1a-p2G/Hsp82 G313N, and HH1a-p2G/hHsp90β.
After transfer of proteins from SDS-polyacrylamide gels to nitrocellulose membranes, the membranes were blocked with Tris-buffered saline, 0.05% Tween-20 (TBST) containing 5% (wt/vol) milk powder and probed with appropriate antibodies in TBST + milk powder at room temperature for 1 h. Mouse anti-GST (Santa Cruz Biotechnology, Santa Cruz, CA), anti-HA (a gift from K. Matter; for references, see Daro et al., 1996 ), and anti-FLAG (Kodak) monoclonal antibodies, chicken anti-Hsp82 antibodies (Louvion et al., 1996 ), rabbit polyclonal anti-Hsp82 antiserum (a gift from S. Lindquist), and rabbit polyclonal anti-Ste11 antiserum (Cairns et al., 1992 ) were diluted 1:1000, 1:100, and to 10 μg/ml, 1:1000, 1:400, and 1:1000, respectively. Membranes were washed three times for 10 min with TBST. The secondary antibodies were alkaline phosphatase-conjugated goat anti-rabbit (Bio-Rad) or anti-chicken (Promega, Madison, WI), horseradish peroxidase-conjugated anti-mouse (Cappel, Cochranville, PA). They were used in TBST + milk powder at room temperature for 1 h. After three washes with TBST, the blots were developed either with the NBT/BCIP reagent for alkaline phosphatase or with the enhanced chemiluminescence reagent (Amersham, Arlington Heights, IL) for horseradish peroxidase.
A large variety of HSP90 mutants have been described that complement yeast strains carrying disruptions of the essential chromosomal HSP90 genes, HSP82 and HSC82 (Borkovich et al., 1989 ; Picard et al., 1990 ; Bohen and Yamamoto, 1993 ; Kimura et al., 1994 ; Minami et al., 1994 ; Bohen, 1995 ; Nathan and Lindquist, 1995 ; Palmer et al., 1995 ; Louvion et al., 1996 ). We examined pheromone signaling (Figure (Figure1A)1A) in three types of mutant strains: a strain with only 10% of the normal levels of Hsp82, a strain with human Hsp90β (hHsp90β; hereafter considered a Hsp90 mutant for yeast), and strains expressing specific Hsp82 point mutants. The latter had been found in a screen for defective steroid receptor signaling in yeast. The point mutants T525I and G313N are temperature sensitive for viability and show an impaired hormonal response of glucocorticoid, estrogen, progesterone, and mineralocorticoid receptors (Bohen and Yamamoto, 1993 ; Bohen, 1995 ). Hsp82 T525I and Hsp82 G313N are expressed at similar levels as the wild-type Hsp82 (Bohen and Yamamoto, 1993 ; Bohen, 1995 ) (see also Figure Figure4B).4B). We first tested the different mutant strains for their ability to arrest growth in response to the mating pheromone α-factor. As shown in Figure Figure1B,1B, low levels of Hsp82, point mutant Hsp82 G313N, and hHsp90β are not able to promote a substantial activation of the pheromone pathway as demonstrated by a poor growth arrest in the presence of pheromone. Interestingly, the point mutation T525I discriminates between two different functions of Hsp90, the pheromone and the steroid-signaling pathways being functional and defective, respectively. The other Hsp90 isoform of yeast, Hsc82, as well as the Trypanosoma cruzi Hsp83, which we have previously shown to complement defective yeast strains (Palmer et al., 1995 ), are also able to support pheromone signaling (our unpublished results).
The activation of the pheromone pathway also results in the induction of proteins required for cellular and nuclear fusion (mating). To test the Hsp90 requirement in the pheromone-dependent transactivation of mating genes such as FUS1, we assessed the induction of a FUS1-LacZ reporter gene (Trueheart et al., 1987 ) upon treatment of cells with α-factor. The results shown in Figure Figure1C1C indicate that, similarly to what was observed in the case of the G1 arrest, the activation of FUS1-LacZ in response to α-factor is strongly reduced in Hsp90 mutant strains compared with a strain with wild-type Hsp82.
The pheromone pathway exhibits low activity even in the absence of pheromones (Hagen et al., 1991 ). To determine whether Hsp90 is also required for this basal activity, we examined the activity of a more sensitive reporter gene, FUS1-HIS3, in a his3− strain. Any disruption in the pheromone signaling pathway, such as the complete absence of a component, abrogates the basal activity and prevents growth on medium lacking histidine (Stevenson et al., 1992 ). As shown in Figure Figure2,2, growth on selective medium is severely impaired for Hsp82 G313N and hHsp90β strains when compared with a strain with wild-type Hsp82. In this assay, the hHsp90β strain is reproducibly the most defective. These results indicate that Hsp90 is necessary for both the induced and the basal activity of the pheromone-signaling pathway.
In a first attempt toward determining the step(s) of the pheromone pathway (Figure (Figure1A)1A) that is dependent on Hsp90, we assayed growth arrest induced by a constitutively active Ste11 mutant. It has been shown that the deletion of the amino-terminal regulatory domain of Ste11 (Ste11ΔN) results in constitutive activation of this kinase and in pheromone-independent induction of the mating pathway (Cairns et al., 1992 ). We constructed such a dominant STE11 and placed it under the control of the conditional GAL1 promoter. When the expression is induced by growth on galactose, only the strain with wild-type Hsp82 exhibits a complete growth arrest (Figure (Figure3A).3A). Strains with either Hsp82 G313N or hHsp90β fail to be fully growth arrested. The same pattern was observed with equivalent strains of the opposite mating type (MATα) (our unpublished results). Thus, these experiments showed that the requirement for Hsp90 is independent of mating type and possibly at the level of Ste11 or downstream of it.
We performed a screen for high-copy suppressors of the signaling defect of Hsp90 mutant strains. Using the hHsp90β strain with the FUS1-HIS3 reporter, we selected suppressors that allow growth on plates lacking histidine in the absence of pheromone and screened them further for a restored sensitivity to α-factor. In a limited screen with a yeast genomic library in a high-copy vector, only one clone met the two criteria. Sequencing revealed that its genomic insert contains the STE5 gene. It starts 1020 bp upstream of the initiator codon ATG and presumably contains the complete STE5 promoter. At the 3′-end the STE5 sequence is truncated at codon 801 (of 917). The isolated plasmid, denoted Ste5ΔC, thus encodes the first 800 amino acids of Ste5 fused to 15 unrelated amino acids at the C terminus. Further experiments showed that Ste5ΔC is also able to suppress the HSP82 mutation G313N. Figure Figure3B3B shows the growth assays on plates lacking histidine and also illustrates that full-length Ste5, as a fusion protein with GST, retains suppressor activity. These data corroborate the tentative conclusion that Hsp90 may be required at the level of the MAPK module consisting of the kinases Ste11, Ste7, and Fus3 that are tethered together by Ste5 (reviewed by Elion, 1995 ; Leberer et al., 1997 ).
Certain client proteins of Hsp90, such as Raf-1, the glucocorticoid receptor, or luciferase, appear more susceptible to degradation when interaction with Hsp90 is blocked/altered pharmacologically (Schulte et al., 1995 –1997 ; Schneider et al., 1996 ; Whitesell and Cook, 1996 ; Czar et al., 1997 ; Segnitz and Gehring, 1997 ; Stancato et al., 1997 ). We therefore examined the accumulation of Ste11 in Hsp90 mutant strains. An epitope-tagged version of Ste11 was overexpressed under the control of the inducible GAL1 promoter and revealed by immunoblotting (Figure (Figure4A,4A, left panel). In strains with hHsp90β or Hsp82 G313N, Ste11 levels were severely reduced. In the Hsp82 G313N strain Ste11 levels were at the detection limit. At this point we speculated that the levels of the endogenous Ste11, which is difficult to detect, might mirror this pattern. To explore this possibility, we concentrated endogenous wild-type Ste11 by immunoprecipitation with a Ste11-specific antiserum and displayed it by immunoblotting with the same antiserum (Figure (Figure4A,4A, right panel). Despite a relatively high background, the identity of the Ste11 band could be confirmed unambiguously using an extract from a ste11− strain as a control sample (Figure (Figure4A,4A, lane Δ). As in the case of the overexpressed Ste11, accumulation of endogenous Ste11 is reduced in both mutant strains although Hsp82 G313N appears to have a less severe effect on the endogenous than on the overexpressed protein. Thus, reduced levels of Ste11 could, at least in part, explain the functional defects of the pheromone pathway in these strains.
The direct target of the Ste11 kinase is the kinase Ste7 (Figure (Figure1A).1A). Upon exposure to pheromone, Ste7 is activated by phosphorylation by Ste11 and becomes hyperphosphorylated in the presence of Fus3/Kss1 (Zhou et al., 1993 ; Neiman and Herskowitz, 1994 ). As shown in Figure Figure4B,4B, the degree of hyperphosphorylation of Ste7 as well as Ste7 protein levels is strongly reduced in mutant strains. The latter is particularly true for Hsp82 G313N. When compared with the wild-type strain, all the mutant strains show a three- to fivefold reduced hyperphosphorylation of Ste7 both in the absence (basal activity) and in the presence (induced activity) of pheromone. This experiment indicates that Hsp90 function is essential both for Ste7 accumulation and for efficient basal and induced phosphorylation of Ste7.
Several experiments described so far suggested that Hsp90 might interact with components of the MAPK module and Ste11 in particular. We performed coprecipitation experiments to examine this issue. Figure Figure5A5A shows that HA epitope-tagged Ste11 is specifically coprecipitated with FLAG-tagged Hsp82. The association of Ste11 and yeast Hsp90 (Hsp82) was confirmed by a GST pull-down experiment. GST alone or GST fused to the constitutive Ste11ΔN (Ste11ΔN.GST) was inducibly expressed under the GAL1 promoter in a wild-type strain. While wild-type Hsp82 (and Hsc82) does not associate with GST alone, it specifically coprecipitates with Ste11ΔN.GST (Figure (Figure5B).5B). These results establish that Ste11 exists in complexes with Hsp90 and that the regulatory N-terminal domain of Ste11 is dispensable for this interaction.
Surprisingly, the Hsp82 requirement for pheromone signaling exhibits a temperature dependence. The mutant phenotype of Hsp82 G313N and hHsp90β strains was not apparent in all assays at the lower temperature of 21°C (Table (Table1).1). This is particularly striking for G313N whose response in all our assays is only slightly reduced compared with wild-type Hsp82 at 21°C. In contrast, low amounts of wild-type Hsp82 are unable to support signaling in response to α-factor even at the lower temperature. Similarly, the basal and α-factor–induced activities of the pheromone pathway are defective in strains with hHsp90β at both temperatures. Interestingly, a full growth arrest is observed at 21°C with a hHsp90β strain when the pheromone pathway is activated with the constitutive Ste11ΔN. This suggests that Hsp90 function might be required differentially both “upstream” and “downstream” of Ste11. At low temperature, hHsp90β appears to be able to fulfill the downstream, but not the upstream requirement. Since none of the mutations are able to block the Ste11ΔN-induced cell cycle arrest at the lower temperature, we cannot formally rule out the possibility that Hsp90 function is not required at all for this particular response. However, this seems unlikely in view of the striking signaling defects at 30°C and may be due to the vast overexpression of Ste11ΔN in this assay and initiation of signaling at an intermediate level.
Using a series of Hsp90 mutants we have demonstrated that pheromone signaling through the MAPK cascade depends on Hsp90 function. Hsp90 is required both for the basal activity of this pathway in the absence of pheromone and for efficient induction upon exposure to pheromone. A combination of genetic and biochemical experiments pinpoints Ste11, a yeast equivalent of Raf, as a target of Hsp90. Since mammalian Raf-1 can substitute for Ste11 under certain circumstances (Freed et al., 1994 ; Irie et al., 1994 ), our results also set the stage for using yeast genetics to investigate the role of Hsp90 for Raf function and for mammalian MAPK signaling.
Our results support the conclusion that pheromone signaling depends on Hsp90 at the level of Ste11: 1) The constitutive Ste11 mutant (Ste11ΔN) fails to elicit a complete cell cycle arrest (at 30°C) in Hsp90 mutant strains; 2) The levels of both endogenous and overexpressed Ste11 are reduced in mutant strains; 3) The basal and induced phosphorylation of the Ste11 substrate Ste7 are reduced in mutant strains; 4) Ste11 and Hsp90 (Hsp82) are found in a complex.
The reduction of Ste11 protein levels are an indication that Hsp90 may be required to ensure the stability of Ste11. Since the plasmids that we used for overexpression of Ste11 contained exclusively the STE11 coding body, the rate of synthesis is likely to be similar. This leads to the tentative conclusion that it is the turnover of Ste11 that is increased in Hsp90 mutant strains. Whether the destabilization of Ste11 is due to misfolding and/or a failure to form complexes with other factors remains to be determined. While the effects on Ste11 protein levels could also be indirect, the finding that Hsp90 and Ste11 form complexes suggests that it is the altered nature of these complexes in Hsp90 mutant strains that leads to enhanced degradation of Ste11. The low levels of Ste11 in these strains have so far precluded experiments to determine whether Hsp90 mutants form complexes with Ste11 at all. In vitro experiments with purified components might allow assessment of whether the Ste11-Hsp90 interaction is direct and how it is affected by alterations of Hsp90. Further analyses will also have to establish the stoichiometry of the complex and the proportion of Ste11 that is associated with Hsp90 at any given time. Interestingly, the effects of mutating HSP90 in yeast are mirrored by pharmacological experiments with the Hsp90 “drug” geldanamycin (or herbimycin A, another ansamycin) in vertebrate cells. Raf-1 is degraded when cells are treated with this compound (Schulte et al., 1995 –1997 ; Schneider et al., 1996 ; Stancato et al., 1997 ). Similar effects have been reported for the glucocorticoid receptor, another Hsp90 substrate (Whitesell and Cook, 1996 ; Czar et al., 1997 ; Segnitz and Gehring, 1997 ). Although accumulation of Raf was apparently not affected in Drosophila strains with HSP83 mutations (van der Straten et al., 1997 ), it should be pointed out that the severity of the effect also depends on the mutation in our system. Recently, Errede and her collaborators have obtained results that support our conclusions. They could notably demonstrate with a temperature-sensitive Hsp82 mutant (Nathan and Lindquist, 1995 ) that the accumulation of newly synthesized Ste11 depends on continuous Hsp90 function (Buehrer, Rhodes, Rutherford, and Errede, unpublished data).
The reduced accumulation of Ste11 (and possibly Ste7) might be sufficient to account for the mutant phenotype. Since it is technically difficult to measure the specific activity of Ste11, we cannot exclude that Ste11 also requires Hsp90 to reach its full enzymatic activity. The residual number of Ste11 molecules in Hsp90 mutant strains might well be sufficient, but they may have a lower specific activity. In the case of geldanamycin-treated vertebrate cells, specific activity of Raf appears to remain unchanged (Stancato et al., 1997 ) whereas in Drosophila its specific activity appeared to be affected by HSP83 mutations (van der Straten et al., 1997 ).
The Hsp90 mutant strains that we have tested are not completely defective in Ste11 activity. Unlike ste11 deletion strains, they are able to form shmoos in response to pheromone, and they can mate albeit with reduced efficiency (our unpublished results). The hyperphosphorylation of Ste7 that occurs at a lower level even in Hsp90 mutant strains further corroborates that there is residual Ste11 activity. This is either due to a pathway that allows Ste11 maturation/stabilization to proceed partially in an Hsp90-independent manner or to residual activity of our panel of Hsp90 mutants. Indeed, Hsp90 mutants that are both viable and completely defective for this specific function may be difficult to find. Along with the fact that there are two genes for Hsp90 in S. cerevisiae, this residual Ste11 activity probably explains why HSP90 was never found in screens for sterile mutants.
Both biochemical evidence and results obtained with the yeast two-hybrid system have led to the view that Ste11, Ste7, Fus3/Kss1, and other components of the pheromone pathway are all tethered together by Ste5. Ste5 may serve as a scaffold to maintain the different kinases and their substrates in a macromolecular signal transduction complex, thereby ensuring specificity and efficiency (for reviews, see Elion, 1995 ; Leberer et al., 1997 ). This illustrates that the notion of a linear signal transduction from upstream to downstream components, as derived from genetic epistasis experiments, is too simplistic. Moreover, it does not take into account that additional factors such as molecular chaperones could be required for the maturation of the individual components and/or the multiprotein complex. Two linked hypotheses are worth considering in this context: 1) Hsp90 chaperones the dynamic assembly of this multiprotein signaling complex; 2) Hsp90 is required for the maturation/stabilization of additional signaling molecules. Note that Hsp90 does not have to be a stable component of these complexes; it might only transiently interact with Ste11 and/or other proteins.
To address the first hypothesis the tools have yet to be developed. Using the yeast two-hybrid assay that relies on interactions of chimeras in the nucleus, we have not seen any differences in Hsp90 mutant strains for the interactions of Ste5 with Ste11 or Ste7 (our unpublished results). However, it will ultimately be necessary to characterize the complex formed of the endogenous wild-type proteins, a technically daunting task.
Regarding the second hypothesis, the reduced Ste7 levels are compatible, but not more, with an interaction of Ste7 with Hsp90. The differential behavior of certain Hsp90 mutants, notably hHsp90β, at different temperatures in different assays (see Table Table1)1) might also support such an assumption. While hHsp90β allows signaling by the constitutive Ste11ΔN at low temperature, it fails to allow pheromone to signal through the complete pathway. Interestingly, Ste5 overexpression suppresses the signaling defect of Hsp90 mutant strains, but only biochemical experiments will be able to elucidate how this increases the efficiency of the signaling complex. Taking these observations as guidelines, the interaction of Hsp90 with signaling molecules both upstream and downstream of Ste11 as well as Ste5 will have to be examined directly.
Other MAPK-signaling pathways in yeast may also depend on Hsp90. While the cell wall integrity pathway does not appear to be affected in our Hsp90 mutant strains (our unpublished results), other pathways await examination. This will be particularly interesting for the three other pathways that are known to share Ste11: one of the two osmoregulatory pathways (Posas and Saito, 1997 ), the invasive growth response of haploid cells, and pseudohyphal development of diploids (see Herskowitz, 1995 ; Levin and Errede, 1995 ; Schultz et al., 1995 ). In this context it is noteworthy that the growth arrest/“toxicity” induced by Ste11ΔN appears to be due to its functions in both the pheromone and the high osmolarity response pathways (Posas and Saito, 1997 ). Since Hsp90 mutant strains are at least partially refractory to the Ste11ΔN toxicity, we speculate that Hsp90 may be required for Ste11 function in both pathways.
Previous studies had demonstrated that it is possible to selectively abolish specific dispensable functions of Hsp90 without compromising its ability to ensure viability in yeast; specifically, a variety of HSP82 mutations result in a defect in the regulation of steroid receptors or v-Src or folding of p53 in yeast (Picard et al., 1990 ; Bohen and Yamamoto, 1993 ; Xu and Lindquist, 1993 ; Bohen, 1995 ; Nathan and Lindquist, 1995 ; Blagosklonny et al., 1996 ; Fang et al., 1996 ; Nathan et al., 1997 ). We have now considerably extended this theme by showing that even subtle point mutations can discriminate between the Hsp90 requirements in two different signaling pathways. Some mutants, like the Hsp82 point mutant G313N, are defective in both steroid receptor and pheromone signaling. Another point mutant, T525I, is only defective in steroid receptor signaling while the converse is true for human Hsp90 (this article and our unpublished results). Moreover, G313N has a different temperature sensitivity for several Hsp90 functions: at room temperature, only hormone binding of steroid receptors is defective (Bohen, 1995 ) whereas viability (Bohen and Yamamoto, 1993 ) and pheromone signaling are only lost upon increasing the temperature to 37°C and 30°C, respectively. A deletion analysis of HSP82 has proven of limited use in assigning specific functions to individual domains of Hsp90 (Louvion et al., 1996 ). Only two regions, the eukaryote-specific N-terminal charged domain and the C-terminal conserved pentapetide, could be deleted without affecting viability. These two portions of Hsp82 are also dispensable for Hsp90 function in pheromone signaling (Louvion et al., 1996 ). By ensuring viability with human Hsp90 (hHsp90β), which cannot promote pheromone signaling, it might nevertheless be possible to map the domains of Hsp82 that are specifically required for its function in pheromone signaling. In such a system, even coexpressed Hsp82 mutants, which fail to provide the viability function, might be able to restore pheromone signaling. Additional insights could be gained by examining a series of chimeras between yeast Hsp82 and human Hsp90β.
We thank S.P. Bohen and K.R. Yamamoto, B. Cairns, M. Collart, E.A. Elion, B. Errede, G.R. Fink, T. Kreis, S. Lindquist, K. Matter, R. Movva, G. Sprague, and D.O. Toft for plasmids, strains, antibodies, and other reagents. We acknowledge the sequencing services of S. Antonorakis and of the Department of Molecular Biology. We are grateful to M. Strubin and J. Geiselmann for critical comments on an early version of the manuscript. We thank B. Errede for her thoughtful comments and for communicating unpublished results. This work was supported by the Swiss National Science Foundation and the Canton de Genève.