Unliganded and hormone-bound glucocorticoid receptors interact with distinct hydrophobic sites in the Hsp90 C-terminal domain

Abstract

Unlike most chaperones, heat-shock protein 90 (Hsp90) interacts with a select group of “client proteins” that regulate essential biological processes. Little is known about how Hsp90 recognizes and binds these proteins. The glucocorticoid receptor (GR) is a well characterized Hsp90 client protein, whose hormone binding, nuclear-cytoplasmic trafficking, and transcriptional activity are regulated by Hsp90. Here, we provide evidence that unliganded and hormone-bound GR interact with two distinct, solvent-exposed hydrophobic sites in the Hsp90 C-terminal domain that contain the sequences “MxxIM” (HM10) and “L/MxxIL” (HM9). Our results indicate that binding of Hsp90 HM10 to unliganded GR stabilizes the unliganded ligand-binding pocket of GR indirectly by promoting an intramolecular interaction between the C-terminal α-helix (H12) and a solvent-exposed hydrophobic groove in the GR ligand binding domain. In the presence of hormone, Hsp90 appears to bind the hydrophobic groove of GR directly by mimicking the interactions of GR with transcriptional coactivators. The identified interactions provide insights into the mechanisms that enable Hsp90 to regulate the activity of both unliganded and hormone-bound GR and to sharpen the cellular response to hormone.

Hsp90 is an unusual chaperone that assists the folding and function of a restricted yet diverse set of structurally distinct regulatory proteins involved in cell cycle regulation, DNA processing, intercellular communication, protein trafficking, and protein turnover (1). Hsp90 is essential for the viability of eukaryotic cells and has become a prime therapeutic target for a variety of human diseases (2).

Among the best studied client proteins of Hsp90 are steroid hormone receptors (SRs), such as the glucocorticoid receptor (GR), which belong to the large nuclear receptor (NR) family (3). NRs are hormone-activated transcription factors that bind hormones through a ligand-binding pocket that is deeply buried within the ligand-binding domain (LBD) (4). High-affinity hormone binding by SRs requires the interaction of the LBD with Hsp90 and associated cochaperones (5). The unliganded (apo)-GR-Hsp90 heterocomplex has been extensively studied and reconstituted by using purified proteins (6). These studies revealed that binding of Hsp90 to apo-GR is ATP-dependent and initially mediated by the Hsp90 cochaperone Hop and a GR-bound Hsp70 assembly complex. In a later step, Hsp90 binds GR directly, and the GR-Hsp90 heterocomplex becomes hormone-binding competent. This transition is accompanied by the ATP-dependent replacement of Hop, Hsp70, and Hsp70 cochaperones by the Hsp90 cochaperone p23 and one of several tetratricopeptide repeat domain proteins (e.g., FKBP51). The Hsp90 inhibitor geldanamycin (GA) prevents the direct but not the Hsp70- and Hop-mediated interaction between GR and Hsp90 and triggers the proteolytic degradation of GR (7).

In addition to assisting hormone binding, Hsp90 also regulates signaling events initiated by hormone-bound GR (holo-GR) (8–9). Hormone binding by GR induces the replacement of the GR-Hsp90 heterocomplex component FKBP51 by FKBP52, whose interaction with the motor protein dynein has been implicated in the nuclear import of holo-GR (9–10). Hsp90 and p23 are present at glucocorticoid response elements (GREs) and regulate the interaction of GR with GREs and transcriptional coregulators (11–13).

Despite many efforts, little is known about the structural features of GR and Hsp90 required for their interaction. Mutational analyses and peptide-competition studies revealed that formation of apo-GR-Hsp90 heterocomplexes that are hormone-binding competent depends on several GR subregions that cluster around the ligand-binding pocket and on three regions within the Hsp90 middle and C-terminal domain (CTD) (5, 14–16). Because these Hsp90 domains are also engaged in interactions with cochaperones, it has been uncertain whether these regions contain direct binding sites for GR. The binding sites of holo-GR and Hsp90 have yet to be characterized.

Because ectopically expressed GR assembles with yeast Hsp90 and cochaperones into hormone-responsive GR-Hsp90 heterocomplexes, yeast has been a very useful model system to study the composition and function of these complexes (17). Using a functional screen in yeast, we have identified four rat GR LBD mutants (Y616N, F620S, M622T, and M770I), which are less dependent on Hsp90 but bind hormone with similar affinity as WT GR (D.R., U. Hostick, L.F., K. Yamamoto, and B.D., unpublished work). The amino acids that are replaced in these GR mutants belong to a conserved allosteric network that transmits hormone-induced structural changes in the buried hormone-binding pocket to a solvent-exposed hydrophobic groove in the GR LBD, which interacts with transcriptional coregulators (18–20) (Fig. 1). Because this mechanism should operate in both directions, these mutants suggested that Hsp90 might stabilize the ligand binding pocket of apo-GR through interactions with the hydrophobic groove. The goal of this study was to test this hypothesis.

Fig. 1.

Identification of the hydrophobic groove of GR as a potential Hsp90-binding site. In the NR LBD, an allosteric network (green) transmits hormone-induced structural changes within the buried ligand-binding pocket (dark gray) into changes in a solvent-exposed hydrophobic groove (yellow) that interacts with amphipathic α-helices of coregulators (red) (18–19). We found that, in GR, replacements of residues within this allosteric network alter the dependence of GR on Hsp90 (D.R., U. Hostick, L.F., K. Yamamoto, and B.D., unpublished work, indicating that Hsp90 may stabilize the ligand-binding pocket of apo-GR through interactions with this groove.

Results

Identification of Potential Hydrophobic Groove-Binding Sites in Hsp90.

Transcriptional coregulators bind the hydrophobic groove of NRs through amphipathic α-helices that contain the sequence LxxLL (x, any amino acid) (21). In the case of GR, the leucine residues of this sequence can be replaced by other hydrophobic residues with only minor effects on the affinity of coregulators for GR (see Fig. 7, which is published as supporting information on the PNAS web site). Hypothesizing that Hsp90 interacts with GR by mimicking the binding of transcriptional coregulators with the hydrophobic groove in the GR LBD, we identified eleven “hxxhh” sequence motifs (called HM1,…, HM11; h, hydrophobic amino acid; x, any amino acid) within the middle and C-terminal domains of Hsp90 and replaced the hydrophobic residues of each motif by alanines (Fig. 2A). Introduced into a yeast strain (YBD100) in which expression of endogenous Hsp90 can be silenced, four of the resulting Hsp90 HM mutants (HM2m, HM9m, HM10m, and HM11m) did not or only modestly impair growth but significantly reduced the response of GR to hormone (Fig. 2 B and C; the phenotypes of the other mutants is shown in Fig. 8, which is published as supporting information on the PNAS web site). Immunoblot analysis revealed that all Hsp90 HM mutants are expressed, although the expression levels of HM9m and HM11m were ≈30% lower than WT Hsp90 (Fig. 2D). Suggesting that these Hsp90 mutants can be recruited by GR, we found that all four Hsp90 HM mutants coimmunoprecipitated with apo-GR (Fig. 2E). However, in the case of HM2m and HM10m, these interactions were at least partially due to nonspecific interactions with protein A/G agarose-bound antibodies. Moreover, these experiments did not resolve whether the interactions of GR with these Hsp90 mutants are direct or mediated by Hsp70. Yeast-expressing HM2m and HM10m showed a temperature-sensitive phenotype, suggesting that the structure of these mutants might be altered (Fig. 2B). Whereas trypsin digests with purified proteins at 25°C and 30°C confirmed differences in the structure of HM2m, in these studies as well as in unfolding experiments, WT Hsp90 and HM10m did not display any differences (see Fig. 9, which is published as supporting information on the PNAS web site).

Fig. 2.

Functional analysis of hydrophobic hxxhh sequences in Hsp90. (A) Identified hxxhh sequences (HM1,…, HM11) in the Hsp90 middle domain and CTD (yeast HSP82 numbering). Alanine replacements of selected hydrophobic residues (bold) within these sequences yielded the Hsp90 HM mutants (HM1m,…, HM11m). (B and C) Duplication times (B) and β-galactosidase activity (C) of yeast (YBD100) expressing GR and WT Hsp90, HM2m, HM9m, HM10m, or HM11m. Shown are the averages and standard deviations of at least three independent experiments performed in duplicate. (D) Immunoblot analysis of Hsp90 (WT/mutants) expression levels in YBD100 grown in the presence of either vehicle (ethanol, −H) or 10 μM cort (+H). Equal loading and transfer was monitored by Ponceau red staining of the immunoblot (data not shown). (E) Coimmunoprecipitation of Hsp90 and GR in yeast expressing either WT Hsp90 or the Hsp90 HM mutants HM2m, HM9m, HM10m, and HM11m. GR-Hsp90 heterocomplexes were immunoprecipitated with either a GR-specific antibody (FiGR) or unspecific IgGs (negative control) and monitored by immunoblot analysis. To reduce unspecific binding, in the case of HM10m, the amount of yeast lysates in the coimmunoprecipitation reactions was titrated.

HM9 and HM10 Have the Structural Features of Potential GR-Binding Sites.

Based on the recently solved structure of full-length yeast Hsp90 (22), the HM2 sequence is part of a buried β-sheet in the Hsp90 middle domain, and HM11 is buried within the CTD homodimer interface (Fig. 3A). Hence, these sequences are likely not engaged in direct interactions with GR. In contrast, most of the hydrophobic residues of HM9 (I557 and L558) and HM10 (M589, I592, and M593) are solvent exposed and part of hydrophobic surfaces in the Hsp90 CTD that could be engaged in protein–protein interactions (Fig. 3C). HM9 resides in α-helix 1 (H1) and HM10 in α-helix 2 (H2) of the Hsp90 CTD. Because of the location and flexibility of H2, it has already been suggested that this α-helix plays a role in client protein binding (23).

Fig. 3.

HM9 and HM10 are potential GR-binding sites. (A) Based on the structure of full-length yeast Hsp90 (22), HM9 (“L/MxxIL”; red) resides in the Hsp90 CTD α-helix H1, HM10 (“MxxIM”; orange) in H2 and HM11 (LxxLL; black) in H4. (B) Sequence comparisons of HM9 and HM10 of human (h) and yeast (y) Hsp90 with the NR-binding sites (NR boxes) of the p160 coactivator GRIP1 as well as with the α-helix H12 of selected SRs, respectively. MR, mineralcorticoid receptor; PR, progesterone receptor; AR, androgen receptor. Corresponding residues in Hsp90 and GRIP1 or H12 are colored. (C) Solvent accessibility of HM9, HM10, and HM11 in the full-length yeast Hsp90 dimer (22) [CTD, blue; N-terminal (N) and middle (M) domains, gray]. L558 and I557 of HM9 (red) form a solvent-exposed hydrophobic surface that is extended by Y647 (violet; H4) and flanked by K522, A553, G559, and D560 (purple). The solvent-exposed hydrophobic surface formed by the HM10 residues I592, M593, and M589 (all orange) is flanked by F583, W583 (both yellow), and Q596 (brown). HM11 is buried within the homodimer interface.

In the case of transcriptional coregulators, sequences amino- or carboxy-terminal to the conserved LxxLL motifs modulate the affinity and specificity for NRs (24). Supporting the role of HM9 and HM10 as potential GR-binding sites, Hsp90 residues adjacent to HM9 (K552, A553, G559, and D560) and HM10 (W585 and F586) are also solvent-exposed and homologous to the SR-binding site of p160 coactivators (NR-box 3) and to H12 of GR and other NRs, respectively (Fig. 3B and C). P160 coactivators bind the hydrophobic groove of agonist-bound GR. H12 interacts with the hydrophobic groove of GR bound to the mixed agonist/antagonist Ru486 (19).

HM9 and HM10 Differ in Their Interactions with apo- and holo-GR.

Motivated by these sequence similarities, we next determined whether the Hsp90 HM9- and HM10-binding sites imitate the interactions of coactivators or H12 of GR with the GR LBD. To this end, we studied the effects of peptides representing Hsp90 H1 (pHM9), H2, (pHM10), and NR-binding site (NR-box) 3 of the p160 coactivator GRIP1 (pGRIP1) on the interaction of H12 with the apo-GR LBD, the ability of apo-GR to bind hormone, and the interaction of GRIP1 with hormone-bound GR (Fig. 4).

Fig. 4.

HM9 and HM10 differ in their interactions with apo- and holo-GR. (A) Peptides representing NR-box 3 of the p160 coactivator GRIP1 and Hsp90 H1 (pHM9), H2 (pHM10), and a mutant H1 (pHM9m) (sequences correspond to hHsp90β CTD). The hxxhh motifs are labeled (∗). (B) Hormone-induced structural changes in the GR LBD based on structural studies (18, 19). In Ru486-bound GR, H12 binds within the hydrophobic groove and blocks the interaction of GR with coactivators. In dex-bound GR, H12 forms one wall of the hydrophobic groove and stabilizes binding of coactivators to this groove. (C) The conformational changes shown in B alter the accessibility of K781 at the C terminus of H12 to trypsin. A trypsin-resistant GR LBD fragment that is characteristic for Ru486-bound GR is labeled (∗). Autoradiogram of apo- and Ru486-bound GR (WT and K781N mutant) after incubation with trypsin (0, 75, 150, or 300 μg/ml) in the presence of vehicle (H₂O), pHM10, or pHM9 (300 μM each). (D) Binding of ³H-dex (4 nM; subsaturation) to in vitro-expressed GR (WT or E773R) ± pGRIP1, pHM9, pHM9m, or pHM10 (300 μM each). Data were normalized with respect to dex bound in the absence of peptides. (E) Interaction of GR with GST or GST-GRIP1 (amino acid 563-1121) (15 μM) ± pGRIP1 (30 μM), pHM9, pHM9m, or pHM10 (300 μM each). The fraction of GR bound to GST-GRIP1 [% Input] was quantified by using PhosphorImaging. The results shown in D and E represent the averages and standard deviations of at least four independent experiments.

The NR H12 is a structural switch that can bind the core LBD at different positions (4). In the presence of the agonist dexamethasone (dex), the GR H12 occupies a position that facilitates the binding of coactivators to the hydrophobic groove, whereas in Ru486-bound GR, H12 docks within the hydrophobic groove of GR and prevents the interaction of this groove with coactivators (18–19) (Fig. 4B). These GR LBD conformations can be distinguished by their different sensitivity to trypsin (25). Although unliganded GR is readily digested by trypsin, dex- and Ru486-bound GR display distinct trypsin-resistant LBD fragments that result from differences in the accessibility of K781 at the C terminus of H12 (25) (Fig. 4C). To determine whether Hsp90 affects the interaction of H12 with the core GR LBD, we analyzed the trypsin sensitivity of GR in the absence and presence of the pHM10 and pHM9 peptides. We found that in the presence of pHM10, apo-GR showed similar trypsin-resistant fragments to Ru486-bound GR (Fig. 4C). pHM10 had no effect on the ability of trypsin to cleave Ru486-bound GR, which demonstrates that the decreased sensitivity of apo-GR to trypsin is not caused by changes in the activity or specificity of trypsin (Fig. 4C). In contrast to pHM10, pHM9 slightly reduced the activity of trypsin but did not alter the trypsin-cleavage pattern of apo- or Ru486-bound GR (Fig. 4C). These results suggested that pHM10 stabilizes apo-GR in a conformation that is similar to that of Ru486-bound GR.

The interaction of coactivators with the hydrophobic groove of dex-bound GR reduces the mobility of H12 and the release of hormone. Consequently, in hormone-binding experiments, pGRIP1 increased binding of dex to GR by ≈70% (Fig. 4D). pHM9 also enhanced hormone binding (by 50%), whereas a mutant pHM9 peptide (pHM9m), in which the hydrophobic LxxIL motif has been replaced by “AxxAA,” had no effect (Fig. 4D). Contrary to pHM9, pHM10 reduced hormone binding by GR by ≈25%. The ability of coactivators to bind to the hydrophobic groove depends on an ionic interaction of E773 in GR H12 with the backbone of the coactivator LxxLL motif (18). Consequently, replacement of E773 by arginine impaired the stabilizing effect of pGRIP1 on GR hormone binding (Fig. 4D). Indicating that binding of pHM9 to the hydrophobic groove requires similar interactions, the E773R mutation also significantly curtailed the stabilizing effect of HM9 on GR hormone binding but did not affect the destabilizing effect of pHM10. These results indicate that HM9 interacts with dex-bound GR by mimicking the interactions of coactivators with the hydrophobic groove of GR. Consistent with this conclusion, we found that the GRIP1 NR-box 3 and the Hsp90 HM9 peptides, but not the mutant HM9 or the HM10 peptides, are able to compete with a GST-GRIP1 fusion protein for binding to dex-bound GR (Fig. 4E). However, whereas 30 μM pGRIP1 released >85% of GRIP1-bound GR, in these experiments 300 μM pHM9 was required to release about half of this amount (40%). Hence, the affinity of the HM9 peptide for hormone-bound GR is likely in the upper μM range. In conclusion, these in vitro assays indicated that Hsp90 HM10 binds apo-GR and stabilizes GR in a conformation similar to that of Ru486-bound GR, whereas the Hsp90 HM9 mimics the interaction of the coactivator LxxLL motifs with holo-GR.

Replacement of HM10 and Treatment with GA Have Similar Effects on the Expression, Nuclear Import, and Activity of GR in Yeast.

The Hsp90 inhibitor GA blocks the transition from Hop- and Hsp70-mediated interactions to direct interactions between GR and Hsp90, which arrests the GR-Hsp90 heterocomplex in a hormone-binding-incompetent state and ultimately triggers degradation of GR by the proteasome (7, 26). If the Hsp90 HM10 sequence is necessary for direct binding of apo-GR, we expect that HM10m yeast have similar phenotypes to yeast treated with GA. As shown in Fig. 2C, the Hsp90 mutant HM10m strongly impaired the transcriptional activity of GR. Moreover, immunoblot analyses revealed that like GA, the presence of HM10m, but not any of the other Hsp90 mutants, reduces the expression levels of apo-GR (Fig. 5A). In contrast to apo-GR, the expression of proteins that are not Hsp90 clients (e.g., the vacuolar H⁺-ATPase component Vma2) were unaffected by HM10m (data not shown).

Fig. 5.

Expression of HM10m and treatment of yeast with GA result in similar phenotypes. (A) Immunoblot analysis of GR protein levels in yeast (YBD100) expressing Hsp90 WT, HM9m, HM10m, or HM11m in the presence of either vehicle (ethanol, −H) or 10 μM cort (+H). Equal loading and transfer was monitored by Ponceau red staining of the immunoblot (data not shown). (B) Subcellular localization of GFP-GR in live yeast expressing Hsp90 WT, HM9m, HM10m, or HM11m. The vacuole and the nucleus of yeast cells are identified in one of the Hsp90 WT images. In the case of HM9m (−H), ≈25% of yeast cells displayed nuclear GFP staining. All other phenotypes are representative of >90% of cells of the respective cultures (evaluated cells, n ≥ 200).

Another function of the interaction of Hsp90 with GR is to prevent nuclear import of apo-GR. In the presence of Hsp90, WT and HM11m nuclear import of GFP-GR was hormone-dependent, whereas in ≈25% of yeast expressing HM9m GR was predominantly nuclear in the absence of hormone (Fig. 5B). In most HM10m yeast, nuclear GFP-GR staining was below the level of detection both in the absence and presence of hormone (Fig. 5B). Similar to GA-treated yeast and mammalian cells that express a GFP-GR fusion protein, we found that, in HM10m yeast, GR often displayed a “punctate” localization pattern that is likely formed by partially degraded and aggregated GFP-GR fusion proteins (27) (Fig. 5B). Consistent with this interpretation, immunoblot analysis of these cells using a GFP antibody showed high levels of GFP-domain-containing GR fragments but low levels of full-length GFP-GR fusion protein (data not shown). Demonstrating that this phenotype is specific for HM10m and not a general consequence of impaired Hsp90 activity, in yeast expressing HM2m, GR was constitutively nuclear and not aggregated (data not shown). In control experiments, we determined that fusion of GFP to either the N or C terminus of GR did not affect the functional interactions of GR with Hsp90 (data not shown). These results are consistent with the role of HM10 as a direct binding site of apo-GR.

Discussion

It is only by understanding the nature of the Hsp90–client protein complex that we can begin to comprehend the functional roles of this unusual chaperone. In this study, we provide evidence that Hsp90 interacts with apo- and holo-GR through two distinct binding sites in the Hsp90 CTD. The interaction of apo-GR with Hsp90 appears to be mediated by the Hsp90 α-helix 2, which contains the HM10 “MxxIM” sequence as well as other solvent-exposed hydrophobic residues whose contributions to the interaction of Hsp90 with GR remain to be investigated. H2 has sequence homologies to H12 at the C terminus of the SR LBD, which regulates the interactions of SRs with coregulators. We found that binding of an H2 peptide (pHM10) stabilizes apo-GR in a conformation in which H12 docks within the hydrophobic groove at the surface of the GR LBD. This conformation appears to be similar to that of GR bound to the antagonist Ru486, which explains previous observations indicating that Ru486 stabilizes GR-Hsp90 heterocomplexes (19, 28). These results indicate that Hsp90 interacts with a defined surface in the apo-GR LBD and stabilizes the unliganded ligand-binding pocket indirectly by promoting the interaction of H12 with the hydrophobic groove. These observations imply that the differences in the dependence of apo-SRs on Hsp90 may correlate with the ability of these receptors to adopt this conformation.

The similar in vivo phenotypes of HM10m yeast and yeast treated with GA suggest that the Hsp90 mutant HM10m traps the GR-Hsp90 complex in a premature, hormone-binding-incompetent state in which interactions between GR and Hsp90 are mediated by Hsp70 and Hop and are not stabilized by p23 and immunophilins. The indirect recruitment of Hsp90 by Hsp70-primed apo-GR may enable Hsp90 to overcome the low affinity of the Hsp90 HM10-binding site for GR (upper μM range) allowing the direct interaction between Hsp90 and GR to remain dynamic. In turn, a dynamic Hsp90-GR interaction would likely relieve inhibition of hormone access to the ligand-binding pocket by Hsp90. These predictions could be tested by analyzing the interactions of HM10m and of the other Hsp90 mutants with cochaperones involved in the assembly and maturation of GR-Hsp90 heterocomplexes.

Binding of hormone to the GR LBD enables H12 to replace Hsp90 and to provide contacts that stabilize the interaction of transcriptional coactivators with the hydrophobic groove. Our binding studies demonstrated that these hormone-induced structural changes in GR also facilitate interaction of GR with the Hsp90 HM9-binding site (Fig. 6A). Hence, hormone binding by GR might transfer GR from the Hsp90 HM10 to the HM9-binding site. In the crystal structure of yeast Hsp90, HM9 and HM10 are in close proximity to each other, suggesting that this transfer might be accomplished without the dissociation of the GR-Hsp90 heterocomplex. Because HM9 is near the binding site of the immunophilins FKBP51 and FKBP52 at the Hsp90 C terminus, transfer of GR from HM10 to HM9 might trigger the observed hormone-induced exchange of the immunophilins FKBP51 by FKBP52 (10). Moreover, the ability of the Hsp90 HM9-binding site to compete with the coactivator LxxLL motifs for binding to the hydrophobic groove provides a potential explanation for the observed ability of Hsp90 to inhibit the transcriptional activity of GR (11, 13). These interactions of Hsp90 with apo- and holo-GR appear to optimize the transition between inactive and active states of GR, resulting in a sharpening of the cellular response to hormone.

Fig. 6.

Proposed interactions of Hsp90 with apo- and holo-GR. (A) In apo-GR, H12 is flexible and can adopt various positions. By binding apo-GR at a position that, in holo-GR, is occupied by α-helix H12 of GR, Hsp90 HM10 foces GR H12 to dock within the hydrophobic groove, which stabilizes the unliganded ligand-binding pocket. Upon hormone binding, GR H12 replaces Hsp90 HM10 and provides the contacts that enable coactivators or Hsp90 HM9 to bind within the hydrophobic groove. (B) Location of known Hsp90 mutations that affect the formation of functional GR-Hsp90 heterocomplexes (black) (22, 29–30). An amphipathic loop that is bound by kinases is shown in green (31).

Although the yeast system is very well established and has been extensively used to characterize the activity of apo-GR-Hsp90 heterocomplexes, little is known about the functional roles of holo-GR-Hsp90 heterocomplexes in yeast. Moreover, some mechanisms that regulate the nuclear translocation and transcriptional activity of GR in mammalian cells are not used in yeast. In particular, yeast lack most of the known coactivators that interact with the hydrophobic groove in the GR LBD. Therefore, further studies on the interaction of GR with these putative Hsp90 interaction surfaces and evaluation of the contributions of these interactions to the cellular response to hormone need to be conducted in mammalian cells. Because the phenotypes of the Hsp90 HM mutants are not dominant negative, these studies require mammalian cell lines that allow the selective expression of Hsp90 mutants. Selective down-regulation of endogenous Hsp90s by RNA interference could be a possible strategy to obtain these lines. However, because of the extraordinarily high cellular concentration of Hsp90, the high sequence conservation of Hsp90, and the importance of Hsp90 for cell viability, the construction and functional analysis of these lines will not be easy.

Functional screens by others have identified various mutations in the middle and C-terminal domain of Hsp90 that disrupt the formation or function of GR-Hsp90 heterocomplexes (29–30). Based on the location of the replaced residues in full-length yeast Hsp90, some of these mutations likely impair the formation of the GR-Hsp90 heterocomplex indirectly by affecting either the N-terminal ATPase activity of Hsp90 or the recruitment of p23 and other cochaperones (Fig. 6B). Several mutations (A587T, T525I, and R579K) replace residues at the base of the flexible loop that contains HM10. Based on their location, it is unlikely that these residues bind GR directly. However, they might contribute to the interaction of Hsp90 with GR by affecting the conformation of H2. One of the few solvent-exposed residues identified by these mutational analyses is E431, which is located in the middle domain (Fig. 6B). Replacement of E431 by lysine selectively impairs the interaction of Hsp90 with GR (29), providing further evidence that GR binds Hsp90 at the boundary between the middle and C-terminal domain of Hsp90. This mutant also suggests that the interactions of GR with Hsp90 are not restricted to the Hsp90 HM9- and HM10-binding sites but involve other contacts within Hsp90. Because of the dependence on the Hsp70 assembly complex, the structural characterization of GR-Hsp90 heterocomplexes is likely beyond reach in the near future. However, the structural analysis of GR LBDs bound to HM9 or HM10 peptides may be feasible and could provide the entrance for the structural characterization of Hsp90–client protein interactions.

It has been shown that the kinase PKB/Akt binds a flexible, amphipathic loop in the Hsp90 middle domain that protrudes into a cleft formed by the Hsp90 homodimer (Fig. 6B) (31). Because this kinase-binding site is in close proximity to HM10, kinases and SRs may bind to similar regions of Hsp90, which is consistent with conclusions from previous mutational studies (30). Although it still remains to be shown experimentally, it is likely that this hydrophobic loop and the hydrophobic surfaces formed by HM9 and HM10 differ in their specificity for client proteins and function as distinct client protein-binding sites. Indicating that HM9 and HM10 may interact with SRs selectively, we found that, in yeast, HM9m and HM10m have very little effect on the ability of the estrogen receptor to activate transcription in response to hormone (L. Kelley and B.D., unpublished observation). The interaction of Hsp90 with so many physiologically important client proteins has been a major obstacle in the functional analysis and therapeutic manipulation of Hsp90. The identification of client-specific Hsp90-binding sites would provide the means for a comprehensive functional and structural analysis of Hsp90–client protein interactions and could lead to novel rationales for the therapeutic targeting of specific Hsp90 client pathways.

Materials and Methods

Yeast Strains and Transformations.

YBD100 (PDR5::LEU2::GT3Z, HSC82::URA3, HSP82::GAL1-HSP82::LEU2) was generated by crossing YNK410 (32) and 5CG2 (33). Yeast Hsp90 (WT and mutant Hsp82) and rat GR were expressed by using pRS316 (pTCA) and pRS313 (pHCA) expression vectors and a constitutive yeast glyceraldehyde-3-phosphate dehydrogenase (GPD) promoter. Transformation of YBD100 followed a standard lithium acetate protocol. If not noted otherwise, YBD100 (GR, Hsp90) transformants were grown at 25°C in minimal medium (S) containing amino acids (-his, -ura, -leu, -trp) and 2% glucose (D).

Growth Assay.

Saturated cultures of YBD100 (GR, Hsp90) were diluted 1:100 in SD-(his, ura, leu, trp) containing either 10 μM corticosterone (cort) or vehicle (ethanol). Cultures were incubated in side-arm flasks at 25°C and 30°C, and their optical density (OD₆₀₀) was monitored at 2-hour intervals until saturation.

β-Galactosidase Assays.

Saturated cultures of YBD100 (GR, Hsp90) were diluted 1:50 in SD-(his, ura, leu, trp) containing either vehicle (ethanol) or cort (10 nM-100 μM). Cultures were incubated in 96-well microtiter plates at 25°C until they reached OD₆₀₀ 0.6 (≈16 h). β-galactosidase activity (v_max) was measured as described (34).

Immunoblots.

Saturated YBD100 (GR, Hsp90) cultures were diluted 1:100 in 25 ml of SD-(his, ura, leu, trp) containing either vehicle (ethanol) or 10 μM cort. Cultures were incubated at 25°C to an OD₆₀₀ of 0.7–0.8 (≈12 h), harvested by centrifugation (600 × g at room temperature for 5 min) and resuspended in 10 mM Tris·HCl (pH 7.5), 50 mM NaCl, 10% glycerol, 2 mM DTT, 0.2 mM PMSF, 1 mM EDTA (pH 7.5), and protease inhibitors (1 μg/ml leupeptin, pepstatin, and aprotinin, and 1 mM PMSF Sigma, St. Louis, MO). Cells were lysed by vortexing with acid-washed glass beads (425–600 μm) (4°C for 40 min), cell lysates were cleared twice by centrifugation (20,000 × g at 4°C for 15 min), and protein concentrations were quantified (protein assay; Bio-Rad, Hercules, CA). Proteins (100 μg per lane) were separated by SDS/PAGE, and probed by using specific antibodies against GR (BUGR2; Abcam, Cambridge, U.K.), Hsp90 (gift from S. Lindquist, Massachusetts Institute of Technology, Cambridge, MA) or Vma2p (gift from T. Stevens, University of Oregon, Eugene, OR).

In Vitro Transcription-Translation.

The rat GR mutants E773R and K781N were constructed by QuickChange PCR (Stratagene, La Jolla, CA). Cloning of pSG5 rat GR has been described (26). A similar cloning strategy yielded pSG5 E773R, pSG5 K781N, and pSP6T Hsp82. GR and yeast Hsp90 were expressed in the absence or presence of l-[³⁵S]methionine (PerkinElmer, Wellesley, MA) by using a coupled reticulocyte lysate expression kit (TNT; Promega, Madison, WI). To obtain hormone-bound GR, 10 μM hormone (cort, dex, and Ru486) were added during synthesis.

GR-Hsp90 Coimmunoprecipitations.

Coimmunoprecipitation of GR-Hsp90 heterocomplexes followed the procedures described in ref. 35.

Hormone Binding.

Hormone-binding assays were performed as in ref. 36, except that our reactions (40 μl) contained 10 μl of in vitro-expressed GR (1–10 pmol), 0–300 μM coactivator or Hsp90 peptides, and 2–50 nM [1,2,4,6,7-³H]dex (Amersham Pharmacia, Piscataway, NJ).

Limited Proteolysis.

GRIP1 and Hsp90 peptides (for sequences see Fig. 5A) were purchased from GenScript (www.genscript.com) and dissolved in water at a concentration of 10 mM. In vitro-expressed, ³⁵S-labeled GR (WT or K781N) were incubated with Hsp90 HM9 or HM10 peptides (300 μM each) or vehicle (H₂O) at 30°C for 30 min. Two microliters of these reactions were digested with trypsin (0–300 μg/ml, Roche, Indianapolis, IN) in 20 mM Tris·HCl (pH 8.0), 0.1 M NaCl, and 10% glycerol (total volume 10 μl). Reactions were incubated on ice for 1 h and stopped by adding 2× SDS loading buffer. Proteolytic fragments were separated by PAGE and visualized by autoradiography.

Coactivator Peptide Competitions.

GST and GST-GRIP1 were expressed, purified, and bound to glutathione agarose as described (20). For peptide competition, 12.5 μl of reticulocyte lysate containing 500,000 counts of in vitro-expressed, ³⁵S-labeled GR were incubated at 4°C for 2 h with 0–300 μM peptides and 20 μl of agarose-bound GST or GST-GRIP1 (15 μM) in a total volume of 50 μl of 20 mM Tris·HCl (pH 8.0), 0.1 M NaCl, 10% glycerol, 10 μM dex, 0.01% Nonidet P-40, 1 mM DTT, 20 μg/ml BSA, 0.1 mM PMSF, and protease inhibitors (Complete; Roche). Agarose-bound proteins were collected by centrifugation (1,000 × g for 1 min at 4°C), washed five times with 200 μl of binding buffer, eluted by boiling in 20 μl of SDS loading buffer, and analyzed by SDS/PAGE. Bound receptors were quantified by PhosphorImaging (Molecular Dynamics, Sunnyvale, CA).

Nuclear-Cytoplasmic Localization.

YBD100 were transformed with pTCA GFP-GR (kind gift of K. Yamamoto, University of California, San Francisco, CA) and pHCA Hsp82 (WT/mutants) and grown as outlined before. Saturated cultures were diluted 1:50 into SD-(his, ura, leu, trp) containing either cort (10 μM), GA (50 μM; Stressgen Bioreagents, Ann Arbor, MI), or vehicle (ethanol) and grown at 25°C to an OD₆₀₀ 0.5–1.0 (≈12 h). Cultures (10 μl) were spotted on a coverslip, and localization of GFP-GR was assessed by fluorescence microscopy (Axioplan 2; Zeiss, Thornwood, NY) by using a ×100 oil-immersion lens.

Abbreviations

cort

corticosterone

CTD

C-terminal domain

D

2% glucose

dex

dexamethasone

GA

geldanamycin

GR

glucocorticoid receptor

LBD

ligand-binding domain

NR

nuclear receptor

S

minimal medium

SR

steroid hormone receptor.