INTRODUCTION

Little is known about how proteins that are not conveyed by a vesicle-based protein trafficking system move through the cytoplasm to arrive at their sites of action in organelles, such as the nucleus or mitochondria, or at a cellular locus like the internal surface of the plasma membrane. Steroid receptors are a useful model for studying such targeted protein movement. These ligand-regulated transcription factors must travel through the cytoplasm, traverse the nuclear pores, and then travel within the nucleus to their sites of action. Their localization is mediated by nuclear localization signal (NLS)¹ sequences and cytoplasmic nuclear shuttling of receptors occurs constantly (for review, see Ref. 1). In the case of the glucocorticoid receptor (GR), the NLS is under hormonal control (2), and localization to the nucleus occurs only after hormone binding. In contrast, estrogen and progesterone receptors have constitutively functional NLSs (3) such that the receptors are localized in the nuclei of hormone-free cells (4, 5). Despite their different localizations, all of these receptors are recovered in cytosols in the same receptor heterocomplex, which has been demonstrated by cross-linking to exist in intact cells (6, 7, 8). This ``core'' complex consists of the receptor bound to a dimer of hsp90 and one molecule of the immunophilin FKBP52/hsp56. Some hsp70 and an acidic 23-kDa protein (p23), both of which are required for assembly of the receptor·hsp90 complex (9, 10, 11, 12), may also be present (see Refs. 13 and 14 for review of receptor heterocomplex assembly and structure).

In 1992, we proposed that the receptors shuttle through the cytoplasm in the heterocomplex form, with hsp90 and the immunophilin acting as a protein transport unit or transportosome (15). This model of receptor movement was supported by experiments in which hsp90 was targeted to the nucleus by fusion to the nucleoplasmin NLS, and it was shown that coexpression of the hsp90 NLS and cytoplasmic receptor mutants devoid of an NLS resulted in complete nuclear localization of the receptors (16). It is important to note that the complex of steroid receptors and hsp90 is dynamic in the sense that assembly and disassembly occurs constantly (17), and it is possible that this dynamic cycling is required for receptor movement.

In 1993, we (18) proposed that the component of the receptor heterocomplex that targets receptor movement to the nucleus is FKBP52/hsp56. This protein is a member of the FK506- and rapamycin-binding class of immunophilins (19, 20, 21, 22), and it is a component of all steroid receptor heterocomplexes (23). FKBP52/hsp56 binds directly to the hsp90 component of the receptor heterocomplex (24, 25) via its 3 TPR (tetratricopeptide repeat)² domains (26). Cross-linking experiments suggest that FKBP52/hsp56 lies in close proximity to the receptor as well (27). FKBP52/hsp56 contains a sequence of 8 amino acids (EDLTDDED, rabbit (20)) with 6 negatively charged residues that is located in a short hinge segment between the first and second globular domains predicted by Callebaut et al. (28). This sequence, which is retained with conservative replacements in human and mouse FKBP52/hsp56 (22, 29), is electrostatically complementary to the receptor NLSs (e.g. the NL1 sequence RKTKKKIK of rat GR (2)). Recently, we showed that intracellular injection of an antibody directed against this conserved negative sequence of FKBP52/hsp56 impeded dexamethasone-mediated cytoplasmic nuclear trafficking of the GR (30). Also consistent with a role in targeted nuclear movement is the observation that the majority of FKBP52/hsp56 is nuclear, with the portion that is cytoplasmic being localized to microtubules (31, 32).

In addition to FKBP52/hsp56, a 40-kDa member of the cyclosporin A binding class of immunophilins, CyP-40, has been recovered with mammalian estrogen, progesterone, and glucocorticoid receptor heterocomplexes (33, 34, 35, 36). Because it is established that a portion of the hsp90 and a portion of the FKBP52/hsp56 in cytosols exist together in a multiprotein complex independent of the presence of steroid receptors (37, 38, 39), we asked if CyP-40·hsp90 complexes also existed. We showed that CyP-40 binds to hsp90 in a manner that is competed by FKBP52/hsp56 and that the two immunophilins exist in independent cytosolic heterocomplexes with hsp90 and with the untransformed GR (36). CyP-40 also contains three TPR domains (33), and in this work, we examine the hsp90 binding of several TPR-containing proteins, including FKBP52/hsp56, CyP-40, p60, and Mas70p.

p60 is a protein that was originally observed in reconstituted progesterone complexes when ATP was limiting (40) or at early stages of assembly (17). It is a homolog of the nonessential yeast heat shock protein, Sti1 (41, 42), and like Sti1 (43), it contains six to eight TPR domains (41). p60 interacts with both hsp90 and hsp70 (43), and the three proteins are thought to interact in a cooperative manner in receptor heterocomplex assembly.

Mas70p (also called Tom70p) is a major protein of the yeast outer mitochondrial membrane (44) that binds nuclear-encoded mitochondrial proteins (45) and acts as a protein import receptor (46). The Mas70p protein, whose gene was cloned in 1983 (47), is anchored to the membrane by a 41-amino acid amino-terminal hydrophobic domain (48) and contains a 60-kDa hydrophilic domain that lies in the cytoplasm. Proteins whose import is accelerated by Mas70p bind to this hydrophilic domain, which can be removed by mild trypsin treatment of mitochondria (49). This 60-kDa cytoplasmic portion contains seven tandem TPR sequences (43), and we show here that it binds hsp90.

As soon as it is translated, the oncogenic tyrosine kinase pp60^v-src becomes associated with hsp90 and a 50-kDa protein of unknown function, p50 (50, 51). The pp60^v-src remains transiently associated with hsp90 and p50 in a cytosolic complex until the kinase localizes to the cell membrane, where it dissociates from the complex (52, 53). These findings were consistent with the notion that hsp90 and p50 were involved in the movement of pp60^v-src through the cytoplasm to the membrane (see Ref. 54 for review). We have shown that the serine/threonine kinase c-Raf, which is involved in normal mitogenic signal transduction, is also in a heterocomplex with p50 and hsp90 (55). Both the Src and Raf heterocomplexes can be assembled under cell-free conditions with the same reticulocyte lysate system that assembles steroid receptor heterocomplexes (55, 56, 57). Like the immunophilins FKBP52/hsp56 and CyP-40, p50 exists in cytosolic complexes with hsp90 (38, 39). However, the steroid receptors and protein kinases make different choices of hsp90-associated protein, in that native steroid receptor·hsp90 complexes contain FKBP52/hsp56 but not p50 (39), whereas the protein kinase heterocomplexes contain p50 but not FKBP52/hsp56 (55). The receptor and protein kinase heterocomplexes are similar in the respect that the presence of p50 in the kinase·hsp90 complexes (55, 56, 57) and immunophilin in receptor·hsp90 complexes (25, 36) is stabilized by molybdate, vanadate, and tungstate. We have proposed previously (55) that p50 may play a role in targeting movement of the protein kinases to their sites of action at the plasma membrane, much as FKBP52/hsp56 may participate in targeting movement of receptors to their sites of action in the cell nucleus.

In this paper we show that CyP-40 and FKBP52/hsp56 (each with three TPRs) form relatively weak complexes with purified hsp90 that are competed by a purified fragment containing the three TPR domains of human CyP-40. In contrast, p60 (with six to eight TPRs) and p50 bind very tightly to hsp90 and their binding is not competed by the CyP-40 TPR fragment at the concentrations we can achieve. Native p60·hsp90 complexes do not contain FKBP52/hsp56, CyP-40, or p50, and, consistent with a common binding site for all the proteins on hsp90, bacterially expressed human p60 inhibits the binding of each to purified hsp90. With its seven TPR domains, Mas70p binds tightly to purified hsp90 in a manner that is competed by the tight binder p60 but not by the weakly binding CyP-40 TPR fragment. From these data we predict that hsp90 has a universal TPR domain binding region that permits it to bind to multiple proteins. Although the gene for p50 is not yet cloned, we predict that, like the others, it will encode TPR domains. We show by indirect immunofluorescence that each of these hsp90-associated proteins localizes to different organelles in a manner that is consistent with their predicted role in targeted protein movement.

EXPERIMENTAL PROCEDURES

Materials

Untreated rabbit reticulocyte lysate was from Green Hectares (Oregon, WI). ¹²⁵I-Conjugated goat anti-mouse and anti-rabbit IgGs were from DuPont NEN. Iron-supplemented bovine calf serum was from HyClone Laboratories, Inc. (Logan, UT). Trypsin, powdered Dulbecco's modified Eagle's medium (high glucose), goat anti-mouse IgG-horseradish peroxidase conjugate, monoclonal nonimmune IgG and IgM, nonimmune rabbit serum, TUB2.1 monoclonal anti--tubulin IgG, anti-nucleolar antibody (nucleolar positive control) and the fluorescein isothiocyanate (FITC)-conjugated antihuman IgG were from Sigma. Actigel ALD (activated aldehyde agarose) affinity support for protein immobilization was purchased from Sterogene Biochemicals (San Gabriel, CA). Goat anti-mouse IgM, donkey anti-rabbit IgG-horseradish peroxidase conjugate, and protein A-agarose were from Pierce. The AC88 monoclonal IgG against hsp90 and the N27F3-4 anti-72/73-kDa heat shock protein monoclonal IgG (anti-hsp70) were from StressGen (Victoria, Canada). The anti-cyclophilin 40 (COOH-terminal peptide) antibody and the 3G3 monoclonal anti-hsp90 IgM were from Affinity BioReagents (Golden, CO). FITC-conjugated donkey anti-mouse IgG and IgM and rhodamine-conjugated donkey anti-rabbit IgG were from Jackson ImmunoResearch Laboratories (West Grove, PA). The UPJ56 rabbit antiserum against FKBP52/hsp56 (58) was a gift from Drs. Karen Leach and Martin Deibel (The Upjohn Co.). The JJ5 monoclonal antibody against p23 (59) was a gift from Dr. David Toft (Mayo Clinic, Rochester, MN) and the DS14F5 monoclonal antibody against p60 (42) was kindly provided by Dr. David Smith (University of Nebraska, Omaha, NE). The anti-Mas70p rabbit antiserum (44) was a gift from Dr. Gottfried Schatz (Biozentrum, University of Basel). The IgM monoclonal antibody against p50 (39) has been described previously. CyP-4059 is a bacterially expressed human CyP-40 COOH-terminal fragment containing the FKBP52-like TPR domain, but not the CyP-18-like domain, and it was purified by Ni²⁺ affinity chromatography and thrombin cleavage.³

Methods

Rat pulmonary endothelial cells (60) were cultured in T25 flasks in Dulbecco's modified Eagle medium with 10% iron-supplemented calf serum. At least 2 days prior to immunofluorescent staining, cells were lifted from the flasks using 0.05% trypsin, 0.5 mM EDTA in calcium-free and magnesium-free Hanks' buffered saline and plated onto 11 × 22-mm glass coverslips (10/100-mm dish) in medium containing 10% iron-supplemented serum.

Rat pulmonary endothelial cells were fixed for 1 h in 3.75% formaldehyde and permeabilized by 5 min incubation in 20 °C methanol prior to staining for hsp56, CyP-40, tubulin, or the nucleolar positive control. Prior to staining for p50 or Mas70p, cells were fixed and simultaneously permeabilized in 20 °C methanol and then incubated in 20 °C acetone for 1 min. All cells were washed with phosphate-buffered saline (PBS) and then incubated for 45-60 min with primary antibody or mixtures of primary antibodies as noted in the figure legends. The cells were washed again with PBS and incubated in secondary antibody or mixtures of secondary antibodies for 30 min. The secondary antibodies used were rhodamine-conjugated donkey anti-rabbit for labeling with preimmune rabbit serum, UPJ56, anti-CyP-40 or anti-Mas70p, FITC-conjugated donkey anti-mouse IgG for TUB2.1, rhodamine-conjugated donkey anti-mouse IgM for anti-p50, or FITC-conjugated goat anti-human IgG for anti-nucleolar antibody. The cells were washed with PBS a final time and the coverslips mounted on slides with p-phenylenediamine mounting medium (61). The cells were viewed on a Leitz Aristoplan epi-illumination fluorescence microscope equipped with a Leitz Vario-Orthomat camera and photographed with T-Max 400 film (Leitz, Rockleigh, NJ). Confocal microscopy was viewed on a Bio-Rad MRC-600 laser scanning confocal microscope (Bio-Rad).

Aliquots of rabbit reticulocyte lysate (100 µl), purified hsp90 (75 µl, 0.5 mg/ml), or rabbit brain cytosol (25 µl) were immunoadsorbed to 7.5-µl pellets of Actigel-ALD precoupled with either nonimmune mouse ascites or 3G3 anti-hsp90 IgM or to 8 µl of protein A-agarose prebound with DS14F5 antibody against p60 (5%), UPJ56 antiserum against FKBP52/hsp56 (4%), anti-CyP-40 (10%), JJ5 antibody against p23 (5%), or nonimmune mouse IgG (5%) or nonimmune rabbit serum (2.5%). Immunoadsorptions were performed with the samples rotating at 4 °C for 2 h. Immunopellets were washed twice by suspension in 1 ml of HEG buffer (10 mM Hepes, pH 7.4, 1 mM EDTA, 10% glycerol) and centrifugation.

Immunopellets were boiled in SDS sample buffer, and proteins were resolved on 10% SDS-polyacrylamide gels. Proteins were transferred to Immobilon-P membranes and probed with 1 µg/ml AC88 for hsp90, 1 µg/ml N27F3-4 for hsp70, 0.1% anti-Mas70p, 0.1% DS14F5 for p60, 0.1% UPJ56 for FKBP52/hsp56, 0.1% anti-p50, or 0.1% anti-CyP-40. The immunoblots were then incubated a second time with the appropriate ¹²⁵I-conjugated counterantibody to visualize the immunoreactive bands.

Rabbit brain cytosol (20 ml) was diluted with an equal volume of 10 mM K₂HPO₄, 1 mM EDTA, pH 7.4, and then chromatographed onto a 2 × 8-cm hydroxylapatite column equilibrated in the same K₂HPO₄ buffer, and proteins were eluted with a 300-ml gradient of 10-400 mM K₂PHO₄ buffer. hsp90, hsp70, p60, FKBP52/hsp56, p50, CyP-40, and p23 were detected by resolving an aliquot of every other fraction by SDS-PAGE and Western blotting with appropriate antibodies. Fractions free of hsp90, but containing the other proteins, were combined and contracted to original volume and dialyzed against HKD buffer (10 mM Hepes, 25 mM KCl, 2 mM DTT, pH 7.4).

Rabbit hsp90 was purified from brain cytosol by sequential chromatography over DE52, hydroxylapatite, and ATP-agarose exactly as described by Hutchison et al. (10). Aliquots (75 µl) of purified rabbit hsp90 (0.5 mg/ml) were immunoadsorbed to 7.5-µl pellets of Actigel precoupled with 75 µl of 3G3 antibody. Pellets were washed once with 1 ml of HE buffer and suspended in HE buffer plus or minus 50 mM KCl and 0.1% Nonidet P-40 (as indicated) in a final volume of 100 µl, including 25 µl of the hydroxylapatite protein pool. Incubations were rotated for 1 h at 4 °C, washed twice with 1 ml of HEG, and proteins were resolved by SDS-PAGE and Western blotting.

One fresh rabbit brain was homogenized in 3 volumes of ice-cold HE buffer (10 mM Hepes, 1 mM EDTA, pH 7.4) and centrifuged at 600 × g for 10 min. The soluble supernatant from this step was further centrifuged at 12,000 × g. The resulting supernatant was clarified for 1 h at 100,000 × g. The final 100,000 × g supernatant contained the soluble 60-kDa fragment of Mas70p and was used for immunoadsorption of native hsp90·Mas70p heterocomplexes.

For bacterial lysates containing p60, cDNA for the 60-kDa human protein (IEF SSP 3521) cloned by Honoré et al. (41), which is the homolog of the rabbit p60 (42), was subcloned into a pET23C vector (Novagen) using the EcoRI and NotI sites.⁴ This construct was used to transform Escherichia coli strain BL21 (DE3), which harbors an integrated T7 polymerase gene. Control E. coli and bacteria expressing p60 were grown to an A₆₀₀ of 0.6, induced with isopropyl--D-thiogalactopyrnoside for 3 h at 25 °C, and harvested. Bacterial lysates were prepared by sonication in phosphate-buffered saline, and aliquots were flash-frozen and stored at 70 °C.

Rabbit brain cytosol (20 ml) was adsorbed to a 2.5 × 20-cm column of DE52 equilibrated with HE buffer, the column was washed with 150 ml of HE buffer, and the proteins were eluted with a 400-ml gradient of 0-0.5 M KCl. hsp90, p60, FKBP52/hsp56, p50, CyP-40, and p23 were detected by resolving an aliquot of each fraction by SDS-PAGE and Western blotting with appropriate antibodies. Fractions containing p50 and hsp90, but not p60, FKBP52/hsp56, CyP-40 or p23, were pooled, contracted to 1 ml, and dialyzed against HKD buffer. The sample was adjusted to 0.5 M KCl and rotated at 4 °C for 2 h prior to chromatography through a 1.5 × 120-cm column of Sepharose CL-6B in HE buffer containing 0.5 M KCl. p50 binds tightly to hsp90 and 0.5 M KCl is required to dissociate the complex. Fractions containing hsp90 and p50 were identified by SDS-PAGE and Western blotting with appropriate antibodies. The p50-containing, hsp90-free fractions were pooled, contracted to 0.5 ml, dialyzed against HKD buffer, flash-frozen, and stored at 70 °C.

RESULTS

As shown in Fig. 1, immunoadsorption of hsp90 from rabbit reticulocyte lysate with the 3G3 monoclonal IgM is accompanied by coimmunoadsorption of hsp70, p60, FKBP52/hsp56, p50, CyP-40, and p23. We have indicated previously that p23 dissociates rather easily during washing of the immunopellet (12), but some immune-specific p23 is clearly detectable in the 3G3 immunoadsorption of Fig. 1. Immunoadsorption with antibodies against individual hsp90-bound proteins show the multiple heterocomplexes that exist in reticulocyte lysate. It was reported by Smith et al. (42), that hsp90, hsp70, and p60 form a major complex, and in Fig. 1 it is shown that immunoadsorption of p60 is not accompanied by any coimmunoadsorption of FKBP52/hsp56, p50, CyP-40, or p23. As we have reported previously (36), immunoadsorption of FKBP52/hsp56 does not yield coimmunoadsorption of CyP-40, but as shown in Fig. 1, there is also no coadsorption of p60. Similarly, immunoadsorption of CyP-40 does not yield coimmunoadsorption of FKBP52/hsp56 or p60. Thus, the three hsp90-associated proteins that are known to have TPR domains appear to exist in separate native heterocomplexes with hsp90.

p50 could not be immunoadsorbed from reticulocyte lysate, because the anti-p50 IgM only recognizes the denatured protein on Western blot. We reported previously (55) that immunoadsorption of FKBP52/hsp56 from reticulocyte lysate with the UPJ56 antiserum yields coimmunoadsorption of p50, leading us to conclude that the two proteins may exist in the same heterocomplex with hsp90. As shown in Fig. 1, immunoadsorption of either FKBP52/hsp56 or CyP-40 is accompanied by the immune-specific presence of some p50. Immunoblotting of aliquots of reticulocyte with the antibodies against FKBP52/hsp56, CyP-40, and p50 suggests a lack of cross-reactivity between them (Fig. 2A). However, both UPJ56 and anti-CyP-40 immunoadsorb some p50 after it has been separated from hsp90, hsp56, and CyP-40 (Fig. 2B), implying direct recognition of p50 by both antibodies. When p50 is concentrated by adsorption to hsp90 such that much more of the protein is present in each lane than was present in Fig. 2A, reaction with both UPJ56 and anti-CyP-40 can be demonstrated by immunoblotting (Fig. 2C). On Western blotting of the denatured proteins, the antibody against p50 does not recognize FKBP52/hsp56 or CyP-40, UPJ56 recognizes p50 but not CyP-40, and the antibody against the COOH-terminal peptide of CyP-40 recognizes both p50 and FKBP52/hsp56. This cross-reactivity suggests that there is some similarity between the two hsp90-associated immunophilins and p50. Also, it is likely that p50 is present in UPJ56 and anti-CyP-40 immunopellets as a result of direct immunoadsorption, rather than being present because it exists in the same hsp90 heterocomplex with each immunophilin and is coimmunoadsorbed with it. Thus, p60, FKBP52/hsp56, CyP-40, and p50 likely exist in separate native heterocomplexes with hsp90.

To assay directly the binding of each protein to purified hsp90, rabbit brain cytosol was chromatographed on hydroxylapatite to separate the hsp90-associated proteins from hsp90. Fractions containing each of the proteins, but not hsp90, were pooled as shown in Fig. 3, and aliquots of the hydroxylapatite pool were incubated on ice with 3G3-Actigel pellets prebound with purified hsp90. As shown in Fig. 4, there was no binding of hsp70 or p23 to hsp90. We have shown previously that purified hsp90 and hsp70 do not bind to each other unless another factor (or factors) in lysate is (are) present (25). p60 is undoubtedly necessary for hsp70 to be in a complex containing hsp90 (42), but other conditions have not been defined. Binding of p23 to hsp90 requires an ATP-dependent process (62). In contrast to hsp70 and p23, the other proteins in the hydroxylapatite pool (p60, FKBP52/hsp56, p50, and CyP-40) all bind to the purified hsp90.

To ask if TPR domains are involved in the protein binding to hsp90, in the experiment of Fig. 5A we added CyP-4059 to the hydroxylapatite pool prior to incubation with the purified hsp90. The CyP-4059 fragment contains the three TPR domains of CyP-40 and its COOH-terminal calmodulin binding domain (63). CyP-4059 prevents the binding of CyP-40 and almost completely inhibits the binding of FKBP52/hsp56 to hsp90 (cf. lane 2 and lane 3 of Fig. 5A). However, the binding of p60 and p50 to hsp90 is not inhibited by the 60 µg of CyP-4059 present in this experiment or by the highest concentrations we could achieve (300 µg, data not shown).

Because p60 has six to eight TPR domains (41) and bacterially expressed human protein was available, we asked if p60 could compete for the binding of p50 to purified hsp90. In the experiment of Fig. 5B, purified hsp90 was incubated with the hydroxylapatite pool alone (lane 2), the pool plus lysate from control bacteria (lane 3), or bacteria expressing p60 (lane 4). The human p60 competes for the binding of p50, as well as FKBP52/hsp56 and CyP-40, to purified hsp90. This is consistent with binding of all four proteins to the same region on hsp90.

Because of their flat shape and relatively large ratio of cytoplasmic to nuclear volume, rat pulmonary endothelial cells are an optimal system for examining cytoskeletal structure and detecting cytoplasmic organelles by indirect immunofluorescence (60). We have reported previously colocalization of cytoplasmic immunofluorescence by the UPJ56 antibody against FKBP52/hsp56 and the TUB2.1 antibody against -tubulin (31) and that observation is repeated in B and C of Fig. 6 to permit comparison with the localization of anti-CyP-40 and anti-p50 immunofluorescence in the same figure.

The localization of the three hsp90-binding proteins is quite different. The majority of FKBP52/hsp56 is localized in the nucleus, and the nucleolar shadows (Fig. 6B and Ref. 31) suggest that this immunophilin is not present, or is present at much lower concentration, in nucleoli. In contrast, virtually all of the nuclear CyP-40 immunofluorescence (Fig. 6E) is localized in nucleoli, as verified by colocalization with anti-nucleolar antibody (Fig. 6F). The cytoplasmic CyP-40 immunofluorescence localizes to small punctate, and often oblong, bodies located throughout the cytoplasm. Although these cytoplasmic bodies look like mitochondria, this has not been established.

In quite a different pattern, immunofluorescence due to the anti-p50 antibody extends on cytoskeletal fibrils from a perinuclear region of intense signal out to the cell periphery (Fig. 6H). To demonstrate that the sharp fluorescence defining the cell periphery is not an artifact of ruffling of the cell margins, we show a confocal image through a single plane of the cells in Fig. 6I. This sharply defined peripheral immunofluorescence is consistent with a localization of some of the p50 at the inner surface of the plasma membrane. We were unable to obtain any distinct pattern of immunofluorescence with the DS14F5 antibody against p60.

The immunofluorescence patterns shown in Fig. 6 were consistent with the notion that FKBP52/hsp56, CyP-40, and p50 might serve to target protein movement to different sites, such as the nucleus, nucleoli, the internal surface of the plasma membrane, and perhaps to mitochondria. In considering a model of targeted protein movement in which a protein moves to an organelle in a transport complex, there must be some way to ``hand-off'' the protein to the organelle. In this kind of a model, it might be important that the hsp90 component of the complex bind tightly to the protein import receptor upon arrival at the organelle. There is solid evidence that Mas70p is a component of the mitochondrial receptor machinery for protein import (46), and we asked if Mas70p would bind to hsp90.

Rabbit brain was used as a source of Mas70p, and as shown in Fig. 7A, a mitochondrial pellet prepared from a brain homogenate contains both 70- and 60-kDa bands (lane 1) that immunoblot with anti-Mas70p antiserum. Both species are also present in a detergent extract of the mitochondrial pellet (lane 3). When the anti-Mas70p serum is used for immunofluorescence in rat pulmonary endothelial cells, it localizes mitochondria in the cytoplasm (Fig. 7B), but it also produces a nuclear immunofluorescence, which may suggest cross-reaction of the antiserum with a nuclear protein (Fig. 7B). Fortunately, the high speed supernatant of brain homogenate contains a lot of the 60-kDa fragment of Mas70p (Fig. 7A, lane 4), and this cytosolic fraction could be directly immunoadsorbed with 3G3 antibody to determine if Mas70p was present in native complexes with hsp90. Because other organs of the rabbits were being used for intact physiological preparations, the brains were not removed and placed in ice until ~20 min after death. This delay before tissue cooling and homogenization may account for the extensive, but useful, cleavage of Mas70p to its 60 kDa cytosolic fragment. Immunoadsorption of brain cytosol with 3G3 antibody resulted in coadsorption of Mas70p 60-kDa fragment with hsp90 (Fig. 7C, lanes 1 and 2). Preincubation of the brain cytosol with the CyP-4059 fragment eliminates binding of CyP-40 to hsp90 but does not affect the amount of hsp90· Mas70p complex that is immunoadsorbed (cf. lanes 2 and 4 of Fig. 7C).

Like the immunophilins and p50, the Mas70p 60-kDa fragment is present in the hydroxylapatite pool of brain cytosol proteins, and we could assay its binding to purified hsp90 as we did with the other proteins in Fig. 4. In the experiment of Fig. 7D, aliquots of the hydroxylapatite pool were incubated with 3G3-Actigel (lane 1) or with 3G3-Actigel prebound with purified hsp90 (lane 2), and it was shown by immunoblotting that the TPR-containing Mas70p 60-kDa fragment bound to hsp90. As described above for p50, binding of Mas70p to purified hsp90 was not competed by 60 µg of CyP-4059 (data not shown). However, the bacterially expressed p60 does compete for binding of Mas70p (cf. lanes 3 and 4 of Fig. 7D), suggesting that these two TPR-containing proteins may bind to the same region on hsp90.

DISCUSSION

hsp90 has been reported to be in association with at least 9 transcription factors, 10 protein kinases, the G protein / subunit and some other regulatory proteins (see Ref. 64 for review). Additionally, hsp90 complexes containing receptors and protein kinases have been reported to contain various amounts of hsp70, p60, immunophilins, p50, and p23 in various combinations and depending upon the conditions of the assay. An ability to associate with multiple proteins is consistent with the role proposed for this ubiquitous, abundant, and conserved protein as a member of a cytosolic superchaperone system (see Ref. 65 for review). Although there are multiple protein interactions and it is likely that many more will be reported, there are potentially four categories of protein interaction sites on hsp90 that have been established or can be reasonably predicted.

There is a region of hsp90 that interacts with the proteins that are being chaperoned (e.g. steroid receptors, protein kinases). This region appears to be located in the COOH-terminal half of hsp90 (71, 72), and it binds proteins of different structure and function without any specific binding motif having been detected. It is also thought that there is a region of hsp90 that binds hsp70, which is required for at least some hsp90 chaperone functions, such as hsp90 binding to steroid receptors (9, 10). However, we have been unable to demonstrate direct binding of purified hsp90 to purified hsp70 (25), and when the hydroxylapatite pool of brain cytosol is incubated with purified hsp90, we do not recover an hsp90·hsp70 complex (Fig. 4). As shown previously by Smith et al. (42), p60 binds to both hsp90 and hsp70, and large amounts of hsp70 are only obtained in native hsp90 heterocomplexes when p60 is present (repeated in Fig. 1, p60 immunoadsorption). It is likely that the presence of p60, and perhaps other lysate factors, generates a direct interaction between hsp70 and hsp90. This predicted hsp70 interaction site on hsp90 could lie proximate to the region that interacts with the chaperoned proteins: this is inferred from the fact that hsp70 binds to steroid receptors during receptor·hsp90 heterocomplex assembly (see Refs. 13 and 14 for review of heterocomplex assembly).

A third protein interaction site on hsp90 is likely required for binding of p23, which binds to hsp90 by an ATP-dependent, but apparently not hydrolysis-dependent, mechanism (62). This p23 interaction is required for proper receptor heterocomplex assembly (11, 12). Preformed complexes that contain hsp90, hsp70, and p60 but have been washed free of a component we have identified as p23 (12) will form a glucocorticoid receptor·hsp90 heterocomplex, but that complex does not bind steroid unless p23 is present (66). The glucocorticoid receptor must be bound to hsp90 to be in high affinity steroid binding conformation (13); thus, in this instance at least, p23 would seem to be required for a ``conformationally productive'' receptor·hsp90 interaction to occur (66). As with the predicted hsp70 interaction site, we would predict that the p23 binding site might be located within an active chaperoning center of hsp90 involved in heterocomplex assembly with receptors and other proteins.

The fourth protein binding site on hsp90 may very well be shared by immunophilins p60 and p50. The fact that p60, FKBP52/hsp56, and CyP-40 exist in different native heterocomplexes with hsp90 (Fig. 1) and that p60 prevents binding of the immunophilins to hsp90 (Fig. 5B) is consistent with shared or overlapping sites. Because the TPR domains of FKBP52/hsp56 (26) and CyP-40 (67) are required for their binding to hsp90, which itself does not have TPR domains, there is likely a TPR acceptor site on hsp90. The affinity of binding to this site may at least in part reflect the number of repeats in the binding protein. FKBP52/hsp56 and CyP-40, both of which possess three TPR domains, are relatively weakly associated with hsp90 (25, 36), and the binding of both is readily competed by the three TPR-containing CyP-4059 fragment (Fig. 5A). In contrast, the binding of p60, with six to eight TPR domains, and of Mas70p, with seven TPR domains, is not competed by CyP-4059 under the same conditions of competition (Figs. 5A and 7C). Obviously, any of these proteins could also bind to regions of hsp90 outside of its TPR acceptor site, with that non-TPR-mediated binding contributing to the overall affinity. p50 is not yet cloned, but we would suggest from these binding competition studies that it may also possess TPR domains and that it binds to a general TPR acceptor region on hsp90.

It has been shown that the GR becomes bound to hsp90 at the termination of its translation (68) and that p60 dissociates from the progesterone receptor during the process of heterocomplex assembly (17). We propose here a general model in which, upon their translation, multiple proteins may be assembled into complexes with hsp90 by a process involving hsp70, p60, p23, and possibly other components of the chaperone system. When p60 dissociates during the assembly process, the TPR acceptor site on hsp90 is available to interact with a targeting protein, such as an immunophilin or p50. This TPR-containing targeting protein binds to a localization signal on the chaperoned protein, as proposed for the GR NLS binding of FKBP52/hsp56 (30). The TPR-containing protein would then, either directly or through additional protein interactions, determine association of the multiprotein complex to the machinery for movement in the appropriate anterograde or retrograde direction. This is not envisioned as a static ``piggyback'' movement of the chaperoned protein with hsp90, but as with the steroid receptors, as a dynamic process in which heterocomplex assembly and disassembly occur continuously (17). Proteins such as the Mas70p mitochondrial import protein may serve to accept the complex by binding the hsp90 chaperone at the site of protein delivery via a tight interaction with the TPR acceptor site. The yeast PAS10 protein, which contains seven TPR domains and is essential for the import of most matrix proteins into peroxisomes (69), might serve the same acceptor function as Mas70p in that organelle. The nuc2⁺ protein is a nuclear protein with TPR domains (43) that is thought to be associated with the nuclear scaffold, and it might serve as a candidate for performing a similar function to that we propose for Mas70p, in that it binds hsp90 when the chaperoned complex arrives at the termini of the nuclear movement machinery.

One question that must be asked is what role does CyP-40 play in such a model of movement for steroid receptors? It is possible that CyP-40 has no function with respect to steroid receptors. CyP-40 binds much more weakly to the GR·hsp90 heterocomplex than FKBP52/hsp56 (36), there is very little of it bound, and it localizes to nucleoli (Fig. 6E), whereas FKBP52/hsp56 and the GR colocalize in nuclei but are both excluded from nucleoli (31). It is possible that upon dissociation of p60 from hsp90 during assembly of the GR·hsp90 heterocomplex, FKBP52/hsp56, CyP-40, and possibly other immunophilins can bind to the TPR acceptor site. The only productive interaction with respect to receptor movement would be the binding of an immunophilin that could also bind to the targeting signal, i.e. the receptor NLS. FKBP52/hsp56 is a good candidate for the receptor targeting protein and CyP-40 is not. Indeed, it is possible the presence of CyP-40 in receptor heterocomplexes is irrelevant to receptor function.