The Protein-Protein Complex between pp60""" and hsp90 Is Stabilized by Molybdate, Vanadate, Tungstate, and an Endogenous Cytosolic Metal*

In a recent study demonstrating the cell-free recon- stitution of the pp60""rc-hsp90 complex, we found that the tyrosine kinase-hsp9O complex is stabilized by molybdate (Hutchison, K. A., Brott, B. K., De Leon, J. H., Perdew, G. H., Jove, R., and Pratt, W. B. (1992) J. Biol. Chern. 267, 2902-2908). In this paper, we ex-amine in detail the stabilization of this protein-protein interaction by transition metal oxyanions. The pp60" ""-hsp90 complex is stabilized by sodium molybdate with the same concentration dependence as the gluco- corticoid receptor-hsp9O complex. As with the steroid receptor heterocomplexes, vanadate and tungstate also stabilize the pp60""rc-hsp90 interaction. Passage of cytosol through a Chelex- 100 metal-chelating resin destabilizes the native pp60'""'"-hsp90 complex, sug- gesting that the complex is normally stabilized by an endogenous metal factor. Readdition of either the heat- stable components of cytosol or a partially purified endogenous metal factor stabilizes the metal-depleted complex. Molybdate also stabilizes the presence of p60, a known hsp90-associated protein, in the pp60"""

Ten years ago it was shown that pp60v-src, a tyrosine kinase encoded in Rous sarcoma virus, is associated in a cytosolic complex with hsp90l and a 50-kDa protein of unknown function, p50 (1,2; for review see Ref. 3). As soon as it is translated, pp60""" associates with hsp90 and p50, and it remains in the heteroprotein complex while it is transported to the cell membrane (45). In 1985 it was shown that unliganded steroid receptors are recovered in cytosolic complexes containing Grants DK31573 (to W. B. P.) and CA47809 (to R. J.). The costs of * This investigation was supported by National Institutes of Health publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "adver- hsp90 (6-9; for review see Ref. lo), and recently, other proteins, including hsp70, have been found in these heteromolecular structures (11-13; for review see Ref. 14). Like pp60""", the steroid receptors appear to bind to hsp90 at the termination of translation (15), and they remain in the heteroprotein complex while they are transported to and within the nucleus (13,14,16).
Recently it has been demonstrated that hsp90 and hsp7O exist together in cytosolic complexes that contain other proteins, including both heat shock proteins (e.g. hsp56) and non-heat shock proteins (e.g. p50) (17)(18)(19). It has also been demonstrated that rabbit reticulocyte lysate can reassociate the steroid receptors (20,21) or pp60""" (22) with the multiprotein complex in an energy-dependent reaction. In the case of the glucocorticoid receptor (GR), the reaction has been shown to yield a functional receptor-hsp90 complex (21).
One feature of the steroid receptor-hsp90 interaction which has permitted the study of these multiprotein complexes is its stabilization by the transition metal oxyanions molybdate, vanadate and tungstate (for review see Dahmer et al. (23)). Molybdate was first found to stabilize the unliganded GR in its steroid binding form (24) and to prevent receptor transformation to the DNA binding state (25,26). Subsequent studies showed that molybdate stabilizes steroid receptors in their 9 S form (23) and that it directly stabilizes the purified GR-hsp9O complex (27,28). Indeed, all of the effects of molybdate, vanadate, and tungstate on GR function can be explained by the ability of the metal oxyanions to stabilize the protein-protein interaction between the receptor and hsp90 (29)(30)(31).
In parallel with the study of molybdate effects, we found that cytosol preparations contain a ubiquitous, small, heatstable factor that stabilizes the GR in its high affinity steroid binding conformation and inhibits receptor transformation to the DNA binding state (32,33). Sat0 et al. (34) reported that such a dialyzable, heat-stable factor inhibited transformation of androgen and estrogen receptors as well. This endogenous factor has the same size, charge, and chelation properties of molybdate, and it produces all of the effects on GR structure and function as the transition metal oxyanions (31). The factor has been purified extensively from boiled rat liver cytosol in a four-step procedure terminating with high pressure liquid chromatography on an Ion-110 anion exchange column (35). The Ion-110 matrix is particularly useful for separating metal anions, and the heat-stable factor copurifies with endogenous molybdenum (35). It is clear that this endogenous metal normally stabilizes the GR-hsp9O complex in cytosol because removal of the factor by passage of cytosol through a metal chelating matrix (Chelex-100) makes the GR unstable, thus markedly facilitating its dissociation from hsp90 and its transformation to the DNA-binding state (36).
Until now, transition metal stabilization of the interaction of a protein with hsp90 has been thought to be unique to the steroid and dioxin receptors. In this work, however, we show that both transition metal oxyanions and the endogenous metal factor stabilize the pp60"""-hsp90 complex in the same way they stabilize the interaction between the GR and hsp90. Removal of the endogenous metal factor from cytosol destabilizes the interaction of both pp60""" and the GR with hsp90. However, removal of the factor does not affect the interaction of hsp7O with either pp60""" or the mouse GR overexpressed in Chinese hamster ovary cells. These observations led us to predict that there is a general mode of interaction between hsp9O and proteins of widely different primary sequence and structure and that transition metal anion stabilization of this protein-protein interaction is an important common feature that can now be examined in a more general light.

Materials
Iron-supplemented bovine calf serum was from HyClone Laboratories (Logan, UT). Chelex-100 chelating ion exchange resin was from Bio-Rad. 1251-Labeled goat anti-mouse and anti-rabbit IgGs were from Du Pont-New England Nuclear. Powdered Dulbecco's modified Eagle's medium (high glucose), protein A-Sepharose, and goat antimouse IgG, goat anti-rabbit IgG, and goat anti-mouse IgM horseradish peroxidase conjugates were from Sigma. The rat fibroblast 3Y1 cell line expressing the temperature-sensitive PA104 pp6OV~"" mutant (37) was generated by transfection of expression vectors containing the molecularly cloned src sequences2 and has been used to study pp60"""-hsp90 association previously (22). The BuGR monoclonal antibody prepared against the rat GR (38) was kindly provided by Drs. Robert Harrison 111 and William Hendry (University of Arkansas for Medical Science), the rabbit serum against hsp7O and hsp90 (39) was a generous gift from Dr. Ettore Appella (National Cancer Institute), and the IgM monoclonal antibody against p50 (19) was kindly provided by Dr. Gary Perdew (Purdue University). Anti-pp60"" N2-17 monoclonal antibody (directed against amino-terminal residues 2-17 of pp60"") and synthetic pp60""' N2-17 peptide were from Microbiological Associates (Bethesda, MD).

Methods
Cell Culture and Fractionation-Rat 3Y1 cells stably transfected with PA104 pp60""" were prepared essentially as described previously (40). All cells were grown in Falcon T150 flasks in Dulbecco's modified Eagle's medium plus 10% iron-supplemented calf serum (and for WCL2 cells (13),40 +g/ml proline and 4 +M methotrexate). Confluent 3Y1 cells were washed once on ice with 5 ml of ice-cold phosphatebuffered saline and then lysed in 2 ml of cold supplemented RIPA buffer (50 mM Tris, 150 mM NaC1, 1% Nonidet P-40, 0.25% sodium deoxycholate, 2 mM EGTA, 1 p M aprotinin, 1 p M leupeptin, 0.1 +M trasylol, pH 7.4) by incubation on ice for 15 min with rocking. Flasks were then tilted on their side to allow the lysates to collect, and the lysates were removed and centrifuged a t 4 "C for 15 min at 12,000 X g in a microcentrifuge. The supernatant was used as the cytosol preparation.
L cells are the L929 mouse fibroblast line routinely used in this laboratory (30), and WCLZ cells are Chinese hamster ovary cells cotransfected with a plasmid containing the cDNA for dihydrofolate reductase and a plasmid containing wild-type mouse GR cDNA and then selected for amplification by growth in methotrexate as originally described by Hirst et al. (41). Both cell lines were harvested by centrifugation, washed by resuspension in Earle's saline, and recentrifuged at 600 X g for 5 min. Cells were ruptured by resuspension in 1.5 volumes of 10 mM HEPES, 1 mM EDTA, pH 7.35, and Dounce homogenization. The homogenate was centrifuged at 100,000 X g for 1 h, and the resulting supernatant was used as cytosol.
Chelex Treatment of Cytosol-Chelex resin was washed extensively and equilibrated with 10 mM HEPES buffer, pH 7.2, loaded into a 10-ml syringe, and centrifuged inside a 30-ml Corex tube at 10,000 X ' B. Brott and R. Jove, unpublished data. g to remove buffer from within the beads. Cytosol was placed on top of the resin in a 1:4 ratio (cytosol volume/resin volume) and centrifuged through the resin immediately. This method allows removal of Chelex-bound metals from cytosol without dilution of the GR-or pp60"""-hsp90 complexes as described by Meshinchi et al. (36). To have virtually no loss of hsp90 from complexes at the start of the experiment, all experiments were performed with freshly prepared cytosol, treatment with Chelex and separation from Chelex resin was performed as rapidly as possible, and samples were kept rigorously cold prior to subsequent procedures.
GR was immunoadsorbed from 300-+l aliquots of L cell or WCL2 cell cytosol with the BuGR antireceptor antibody at 10% of the final volume, the mixture was incubated 4 h at 0 "C, and each sample was added to a 10-pl protein A-Sepharose pellet. Samples were mixed by rotation for 1 h at 4 "C, and protein A-Sepharose pellets were washed three times by resuspension in 1-ml aliquots of TEGM buffer.
Gel Electrophoresis and Immunoblotting-GR, pp60v-src, and their associated hsp90 and hsp7O proteins were assayed by SDS-polyacrylamide gel electrophoresis and immunoblotting. SDS-PAGE was performed in 7% (GR) or 10% (pp60'+") slab gels as describedpreviously (30). Immunoblotting was carried out by transferring proteins from acrylamide slab gels to Immobilon P transfer membranes followed by incubation for 2 h at room temperature with 2% BuGR2 monoclonal antibody to detect the GR, 0.1% N2-17 monoclonal antibody to detect pp60"", 0.1% antibody against p50, or 0.02% rabbit antiserum to detect hsp90 and hsp7O. Immobilon membranes were subsequently probed by the appropriate peroxidase-conjugated counterantibody followed by '"1-conjugated counterantibody. The relevant bands visualized by the peroxidase reaction were then excised from the membrane and assayed by scintillation counting.

RESULTS
Stabilization of pp60U~sm-hsp90 Complex by Molybdate-In our recent work utilizing the rabbit reticulocyte lysate system to reconstitute the pp60"""-hsp90 heterocomplex, we found that immunopurified native or reconstituted complexes would dissociate if they were suspended in a low ionic strength TEG buffer and incubated for 20 min at 37 "C (22). Dissociation of hsp90 from pp60'~"" was inhibited by 20 mM sodium molybdate (22). To study metal stabilization of pp60"""-hsp90 complexes, in this work we will utilize the PA104 variant of the Rous sarcoma virus pp60"-"" (37). In contrast to the wild-type pp60""", in which only a portion of the protein kinase is recovered in heterocomplex with hsp90, lesions in the catalytic domain of PA104 pp60"~"" result in it being present almost entirely as a soluble complex in association with hsp90 (42). In our previous study, we found that the stability of the PA104 mutant pp60"""-hsp90 complex in TEG buffer was the same as that of the wild-type pp60"""-hsp90 complex (22). Fig. 1 shows the stabilization of both the GR-hsp9O complex and the PA104 pp60"""-hsp90 complex by various concentrations of molybdate. Half-maximal stabilization of the GR-hsp90 complex is achieved at about 2 m M molybdate, consistent with the previously published ability of the metal to stabilize the steroid binding capacity of the immunoadsorbed complex (30). It is clear from Fig. 1 that the pp60"""-hsp90 complex is stabilized over the same range of molybdate concentrations as the GR-hsp9O complex.
Passage of Cytosol through Chelex-100 Destabilizes the pp60"-""-hp90 Complex-In a previous publication (36) we showed that rapid passage of L cell cytosol through a bed of Chelex-100 removes an endogenous metal-stabilizing factor, yielding GR-hsp9O complexes that are present at the same concentration as in whole cytosol but that are unstable. In the experiment of Fig. 2A, whole L cell cytosol and cytosol that had been depleted of metals by Chelex treatment were incubated at 20 "C. At various times, aliquots were removed, the GR was immunoadsorbed to protein A-Sepharose, and GR-associated hsp90 was assayed by Western blot. In the metal-depleted cytosol, hsp90 dissociated at a rapid rate, containing the GR were immunoadsorbed to protein A-Sepharose with anti-pp60"" or BuGR antibody, respectively. After washing in TEG buffer, the immune pellets were suspended in TEG buffer and incubated for 20 min at 37 "C with the indicated concentrations of sodium molybdate. At the end of the incubation, all samples were washed with buffer containing 20 mM molybdate, and pp6OV'""-associated and GR-associated hsp90 were resolved by SDS-PAGE and Western blotting. Immunoblotted bands of hsp9O labeled with '"Ilabeled counterantibody (shown above the graph) were excised, and the radioactivity was assayed. Maximum stabilization was achieved in samples containing 10 mM molybdate, and this value was set at 100% to permit normalization of data from several experiments (average of three experiments & S.E.). Solid dots represent hsp90 bound to pp60"""; open circles, hsp9O bound to GR. whereas the complex remained stable in whole cytosol containing the endogenous metals. An identical experiment with 3Y1 cytosol containing PA104 pp6OV.""-hsp90 complexes is shown in Fig. 2B. It is clear that Chelex treatment results in rapid dissociation of hsp90 from pp60""".
The results shown in Fig. 2 suggest that an endogenous metal component (or components) of cytosol normally stabilize the protein-protein interaction between pp60"sm and hsp90. In the experiment of Fig. 3, the transition metal oxyanions that are known to stabilize steroid receptor-hsp90 complexes (molybdate, vanadate, and tungstate) were added (at 10 mM) to the metal-depleted 3Y1 cytosol. Considerable stabilization of the pp6OV~""-hsp90 complex is achieved. Stabilization is also achieved by addition of the endogenous steroid receptor-stabilizing factor. The endogenous factor has been purified from boiled rat liver cytosol through the phosphocellulose P-11 step described in Meshinchi et al. (35), and its final concentration in the incubation is the same as its concentration in normal boiled liver cytosol. Readdition of boiled L cell or 3Y1 cell cytosol stabilizes in the same manner as the more purified factor preparation used here (data not shown). (22) that some hsp70 is also recovered in an immune-specific manner when cytosol from 3Y1 cells expressing PA104 pp60""" is immunoadsorbed with anti-pp6O"" antibody. We have also reported (13) that hsp70 is a component of the GR heteroprotein complex in WCLZ cells, a Chinese hamster ovary cell line that overexpresses the mouse GR. This stands in contrast to the GR heterocomplex in cytosols prepared from other cell types where hsp7O is not present (13,43). In the experiment of Fig. 4 we asked if Chelex treatment of WCLZ cell cytosol also promotes dissociation of hsp70 from the receptor.

Metal Effects on Other Proteins in the pp60"'"" Heterocomplex-We have recently reported
In Fig. 4A, cytosol from WCL2 cells was passed through Chelex and incubated at 20 "C. The receptor was immunoadsorbed and both receptor-associated hsp90 and hsp7O were

FIG. 2.
Rapid dissociation of hsp90 from GR or pp60"'" in metal-depleted cytosol. L cell or 3Y1 cell cytosois were passed rapidly through a bed of Chelex-100 to remove metals as described under "Methods." Portions of both Chelex-treated and whole cytosol (i.e. not passed through Chelex) were incubated at 20 "C. A t the indicated times, aliquots were removed, 20 mM sodium molybdate was added, and GR or pp60""" was immunoadsorbed to protein A-Sepharose. GR, pp60""" and the associated hsp90 were resolved by SDS-PAGE and immunoblotting using '"I-labeled counterantibodies. The radioactivity in each band was assayed, and the graphs show the relative amount of GR-or pp6OV~""-associated hsp90 in whole cytosol (0) or Chelex-treated cytosol (0) expressed as a percent of the zero time control. Panel A, G R panel B, p p 6 F . The Western blots above the diagrams show the '""I-labeled receptor, pp60'", and associated hsp90 bands as indicated. Effect of Chelex treatment on the hsp70 component of the GR and pp60""" heterocomplexes. WCL2 cell or 3Y1 cell cytosols were passed rapidly through a bed of Chelex-100 to remove metals as described under "Methods." The cytosols were then incubated at 20 "C, aliquots were removed, immunoadsorbed, and assayed for GR-associated and pp60v~""-associated hsp90 and hsp70 as described for Fig. 2. Autoradiogram of the hsp90 and hsp7O are shown above the graphs. A , GR immunopellets; R, pp60""" immunopellets. Lune I , immunoadsorbed with nonimmune IgG (panel A ) or with anti-pp6O"" in the presence of competing N2-17 peptide (panel B ) ; lane 2, immunopellet from Chelex-treated cytosol incubated at 0; lanes 3, 4, and 5 are Chelex-treated samples incubated at 20 "C for 15, 30, and 60 min, respectively; lane 6 in B was incubated a t 20 "C for 90 min; last lane, immunopellet from whole cytosol, 0 "C. The ' *' I radioactivity in each band was assayed, and the graphs show the relative amount of GR-or pp6OV~"~-associated hsp90 (0) and hsp70 (0). The hsp7O is expressed as a percent of the zero time control in lane 2, but because the Chelex treatment itself caused a loss of hsp90 from GR heterocomplexes in WCLB cell cytosol (dashed line), the hsp90 is expressed as a percent of the GR or pp6O"""-associated hsp90 in whole cytosol that was not Chelex-treated (last lane).
assayed. The rapid Chelex treatment of WCL2 cytosol causes loss of 50% of the receptor-associated hsp90 prior to the incubation (dashed line in Fig. 4A, cf. lanes 2 and 6 above the figure). In some experiments, even more of the hsp90 dissociated during Chelex treatment of WCLB cytosol, and we have the impression that this complex is somewhat less stable than the receptor heterocomplex in L cell cytosol where the rapid Chelex procedure itself does not cause significant dissociation of hsp90 ( Fig. 2A and ref. 36). It also appears to be less stable than the pp60'."" heterocomplex where there is no loss of hsp90 during Chelex treatment (Fig. 4B).
In contrast to hsp90, Chelex treatment does not affect the stability of hsp7O binding either to the GR or to pp60""" (Fig.  4). When immunopurified pp60""" heterocomplexes are incubated at 37 "C as in the experiments shown in Fig. 1, hsp7O dissociates but addition of sodium molybdate does not inhibit its dissociation (data not shown). The failure of hsp70 to dissociate is consistent with the conclusion of other studies that hsp70 remains bound directly to steroid receptors even after hsp90 has dissociated during receptor transformation (11)(12)(13). As shown in Fig. 5, incubation of immunoadsorbed pp60""" heterocomplexes at 37 "C results in dissociation of the 50-kDa pp60"""-associated protein and this dissociation of p50 is inhibited by molybdate.

DISCUSSION
We have shown here that the pp60"""-hsp90 complex is stabilized by metals in the same manner as the GR-hsp9O complex. Because cytosol from the 3Y1 cell, contains almost no GR (data not shown), we have compared the effects of metals on the PA104 pp60''""-hsp90 complex in 3Y1 cytosol to the effects of metals on the GR-hsp9O complex in L cell cytosol. Both complexes are stabilized by the same concentrations of molybdate (Fig. 1) and vanadate and tungstate also stabilize the pp6Ov~""-hsp9O interaction (Fig. 3), as was shown previously for steroid receptor-hsp9O complexes (see ref. 23 for review). In the same manner as the GR-hsp9O complex, the pp6OV'""-hsp90 complex is normally stabilized by an endogenous metal component of cytosol that is removed by Chelex treatment (Fig. 2). Readdition of either the heat-stable components of cytosol or the partially purified endogenous metal factor (Fig. 3) significantly restores the stability of the complex.
In contrast to the GR or pp60'"" complexes formed with hsp90, the complexes formed with hsp7O are not affected when the endogenous cytosolic metals are removed by Chelex treatment (Fig. 4). Also, addition of molybdate does not inhibit hsp7O dissociation when immunopurified heterocomplexes are incubated a t 37 "C (data not shown). It is known that transformation of progesterone or glucocorticoid receptors to the DNA binding state is accompanied by dissociation of hsp90 but not hsp7O (11)(12)(13), leading to the impression that there is a direct protein-protein interaction between hsp70 and the steroid receptors. Although much less hsp7O (relative to hsp90) is present in the PA104 p60'."" heterocomplex than in the WCLB cell GR heterocomplex (Fig. 4 and Ref. 22), the hsp70 remains bound after hsp90 dissociation (Fig. 4B), suggesting that there is a direct protein-protein interaction between hsp7O and pp6OV-"".
It is perhaps not surprising that molybdate inhibits dissociation of p50 from the pp60"'"" heterocomplex ( Fig. 5). Whitelaw et al. (19) have shown that p50 is coimmunoadsorbed from cytosols with hsp90, and it is thought that p50 is associated with hsp90. Thus, the molybdate effect on the p50 component of the complex may reflect metal stabilization of the protein-protein interaction between pp60'~"" and the hsp90 to which p50 is bound.

Metal Stabilization ofpp6PC-hsp90 Complexes
Although stabilization of steroid receptor-hsp90 interactions by molybdate has proven to be very useful in studies of both receptor structure and function, very little is known about the mechanism by which the protein-protein interaction is stabilized. It is known that hsp90 binds to the steroid binding domain of the GR (44, 45), and Simons et al. (46) have shown that a 16-kDa fragment (amino acids 537-673 of the rat GR) of the steroid binding domain binds steroid in a high affinity manner and is stabilized by molybdate. Chakraborti and Simons (47) have published evidence that this fragment is bound to hsp90. Although it is clear from the work of Cadepond et al. (48) that portions of the receptor to the COOH-terminal side of this fragment form part of the hsp90 binding site, any amino acids on the receptor that might be involved in metal stabilization of the protein-protein interaction must lie within the 16-kDa fragment. Molybdate has a well established avidity for sulfur (49), and it is clear that molybdate can directly interact with cysteine sulfhydryl groups located in the hormone binding domain of the receptor (50), but there is as yet no evidence that cysteine or any particular amino acid is involved in molybdate stabilization of the receptor-hsp90 interaction. Indeed, as we have noted previously (51), there is no evidence that the stabilizing effect of molybdate involves a direct interaction of the metal with the receptor. It is entirely possible that molybdate could interact solely with hsp90 to maintain it in a conformation that dissociates only very slowly from the receptor.
In contrast to steroid receptors, there are no direct studies that identify a region of pp60""" that is involved in binding hsp90. Indirect immunoadsorption studies suggest that the COOH terminus is involved (52,53), and Jove et al. (42) have shown that increased binding of hsp90 (as in the PA104 mutant used in this study) segregates with mutations in the catalytic domain, which is located in the COOH-terminal half of the kinase. Such a location would also be consistent with the suggestion that hsp90 attenuates the kinase activity of pp60""" (2)(3)(4). There is no information regarding potential sites for metal interaction with the kinase. An examination of the primary sequence of pp60""" does not reveal the presence of metal binding configurations, such as concentrations vicinal cysteines or histidines. As with the steroid receptors, we are presented with the possibility that molybdate and the other transition metal anions may stabilize the pp60'"""-hsp90 complex by binding solely to hsp90. hsp90 is a highly conserved, abundant, and ubiquitous protein that is required for cell survival and is thought to be involved in protein chaperone functions in the cell (14,54,55). If indeed hsp90 is performing a chaperone function, it must interact with a wide variety of proteins in the cell. The steroid receptors and pp60""" could then be considered as part of a subgroup of proteins that bind to hsp90 with higher affinity than other proteins and remain bound to it during their intracellular transport (3,14). Most chaperone proteins interact with a wide variety of polypeptides that do not share any primary sequence homology (for review, see Ref. 55), and it is perhaps not surprising that we do not find sequence homology between the GR and pp60""".
Because two such dissimilar groups of molecules as the steroid receptors and pp60""" are bound to hsp90 in a manner that is stabilized by the transition metal oxyanions, it seems very likely that it is the common component, i.e. the hsp90, that is the site of the metal interaction. If molybdate and the other metals act via conformational stabilization of hsp90, then it is possible that a wide variety of inherently weaker protein-hsp90 interactions in the cell require the metal binding for normal (or optimal) chaperone function.