Protein chaperones often engage in rather transient associations with their target proteins, preventing an analysis of these associations by the conventional two-hybrid system. This study though revealed that many associations of Hsp90 can be stabilized in vivo using a mutation that inhibits the essential ATPase reaction of the Hsp90 chaperone cycle. The E33A mutation in the Hsp82-BD two-hybrid bait strongly reinforced the interactions of this bait with about 3% of the yeast proteome, the latter expressed as AD fusions (Table ). That many of the interactors are known cochaperones or clients of the yeast Hsp90 system (Fig. ) is a strong indication that this E33A mutant Hsp82-BD bait is allowing the detection of bona fide Hsp90 associations. This screening approach is therefore potentially of use in finding new binding partners of Hsp90, though it is improbable that it could reveal every Hsp90 interactor in the cell (false negatives arise for a large number of reasons in two-hybrid screens) (3
). Indeed, while our screen succeeded in identifying some of the known Hsp90 clients of yeast (10
), it also failed to identify others (1
Table lists the protein fusions selected by the E33A mutant Hsp82-BD bait. For most of these, further work is needed to confirm whether or not the two-hybrid association is meaningful in terms of a biologically relevant Hsp90 complex. It is noteworthy that a few other chaperones and chaperonins of yeast were identified as putative Hsp90 interactors (Table ). They include Hsp60/Hsp10, the functional equivalent of the bacterial GroEL/ES in mitochondria. Since yeast Hsp90 is localized to the cytosol (and probably also the nucleus), the detected interactions (also the interaction with the Grx5p mitochondrial glutaredoxin) might be to the precursor forms of these proteins that exist prior to mitochondrial import, especially since the presence of the Gal4p AD at the N termini of these protein fusions may interfere with the recognition of signal sequences for mitochondrial import. In mammalian systems the Hsp70/Hsp90 chaperone system delivers mitochondrial protein precursors to the preprotein translocase of the outer mitochondrial membrane, but it appears that this process is not Hsp90 dependent in yeast (70
). Nevertheless, a component of the multisubunit mitochondrial import receptor, Tom22p, was identified as a potential Hsp90 interactor in this screen.
A few cytoskeletal proteins (actin, tropomyosin, and Cdc12p) and chaperones/cochaperones involved in cytoskeletal function (Cct4p, Hsp42, and She4p) were identified as potential Hsp90 binding partners (Table ). Hsp90 has long been known to possess actin-binding properties (41
). There is also a constant requirement for Hsp90 function in the organization of actin microfilaments, as indicated by an almost immediate delocalization of rhodamine-phalloidin staining and loss of polarized growth in yeast cells treated with Hsp90 inhibitor drugs (S. H. Millson, unpublished observations). Certain Rab/Ras small GTPase protein family members were also apparent Hsp90 interactors (Table ). This is consistent with the recently discovered role of a Rab-recycling, membrane-associated Hsp90 chaperone complex in operation of the alphaGDP-dissociation inhibitor. The latter acts in the Rab-mediated targeting of vesicles to an acceptor compartment, coordinating the Ca2+
-dependent events that trigger the hydrolysis of Rab-bound GTP with the retrieval of the product of this reaction, Rab-GDP, from vesicle membranes to the cytosol (54
). Among the other potential Hsp90 interactors were 28 known or potential membrane transporter proteins, including no less than six of the plasma membrane permeases for amino acids (Table ). This is unexpected, since the two-hybrid system requires BD- and AD-containing fusions to associate noncovalently in the yeast nucleus. Possibly the fusion of these membrane proteins to the Gal4p AD, a domain with a nuclear import signal, has allowed their mislocalization to the nucleus. Only further work will tell if this apparent Hsp90 association with membrane transporters is of biological significance.
The E33A mutation inhibits the final, essential ATPase step of the Hsp90 chaperone cycle and the ensuing release of the activated client protein (43
). Client proteins interacting with this mutant Hsp90 should therefore accumulate as a late-stage chaperone complex, stabilized in association with the ATP-bound chaperone. When expressed at normal cellular Hsp90 levels, the E33A mutant Hsp82 cannot provide the essential Hsp90 function in yeast (43
). This mutant Hsp82 can though allow very slow growth when highly overexpressed as the sole Hsp90 of yeast cells (unpublished observations). An E33A mutant Hsp82 must therefore be capable of very slow progression through the chaperone cycle. It might be argued that the two-hybrid interactions detected on the basis of this E33A mutation are artifacts of such slow cycle progression and the resultant alterations to the stoichiometry of the different Hsp90 complexes in the yeast. It was necessary therefore to validate this approach for revealing new Hsp90 binding partners by confirming that an interaction reinforced by the E33A mutation involves a hitherto-unidentified Hsp90 client (Fig. to ).
Just 6 of the 117 yeast protein kinases exhibited a reinforced two-hybrid interaction dependent on the E33A mutation in the Hsp82-BD bait (Table ). Nevertheless, these included two of the kinases already identified as Hsp90 clients, as well as three of the five MAP kinases of yeast. We were particularly intrigued by the latter finding since, while an Hsp90 dependence has been shown for a number of MAP kinase pathway signaling events (see below), the MAP kinase family kinases that are the targets of this signaling are not generally regarded as Hsp90 clients. Indeed, it would seem that Hsp90 is not required for the activity of many of these kinases (for example, recombinant ERK2 and p38γ can be obtained in active states by Escherichia coli
expression in the absence of eukaryotic forms of Hsp90 [4
]). We therefore sought to establish whether one of the MAP kinases identified in the screen was Hsp90 binding and Hsp90 dependent in its activity. As shown in Fig. and , Slt2p interaction is specific for the ATP-bound form of Hsp82 and the dually Thr190
-phosphorylated, stress-activated state of Slt2p, most probably as a late-stage Hsp90 chaperone complex. The phenotype of the T22Ihsp82
mutant indicates, in turn, that Hsp90 function is essential for this dually phosphorylated Slt2p to activate one of its targets, the Rlm1p trans
activator of cell wall genes (Fig. ). It is probable therefore that the Slt2p-Hsp82 interaction, identified and characterized in this study (Fig. to ), is essential for Slt2p MAP kinase activity.
This Hsp90 requirement in the action of Slt2p may represent the Hsp90 machine participating in the formation of the active Slt2p MAP kinase in response to the structural changes induced in this MAP kinase as a consequence of the activating Thr/Tyr phosphorylation. Alternatively, a phosphorylation-induced dimerization of Slt2p (as in mammalian ERK2 [28
]) may be an essential step before Hsp90 (also a dimeric protein) is able to bind to Slt2p and promote its interaction with the MAP kinase docking site (D domain) on Rlm1p. A third possibility is that the Hsp90 function is required for the ability of Slt2p to phosphorylate its targets after the docking interaction. Future work should readily distinguish which of these events in the operation of this client protein require the Hsp90 chaperone.
A number of the signaling events of MAP kinase pathway activation are known to be Hsp90 dependent. In Schizosaccharomyces pombe
the Hsp90 cochaperone Cdc37p is involved in Spc1p stress-activated MAP kinase signaling and cdc37
mutations affect both Spc1p level and Spc1p phosphorylation by the Wis1p stress-activated MAP kinase kinase (62
). Hsp90 is also needed for the activity of Ste11p, a MAP kinase kinase kinase of S. cerevisiae
that signals to no less than two MAP kinases, Kss1p and Hog1p (31
). In addition, Hsp90 binding stabilizes mammalian Mok1p, a protein kinase that is structurally moderately related to conventional MAP kinases and activated during spermatogenesis (39
). Nevertheless stress-activated MAP kinases, such as the Hog1p and Slt2p of S. cerevisiae
, the Sty1p of S. pombe
, and the p38 of mammalian systems, have not previously been considered as Hsp90-dependent activities. As a MAP kinase with well-established genetics (5
), Slt2p should be an ideal model protein with which to investigate the Hsp90 dependence of protein kinase activity.