Active Participation of Hsp90 in the Biogenesis of the Trimeric Reovirus Cell Attachment Protein ς1*

The reovirus cell attachment protein, ς1, is a lollipop-shaped homotrimer with an N-terminal fibrous tail and a C-terminal globular head. Biogenesis of this protein involves two trimerization events: N-terminal trimerization, which occurs cotranslationally and is Hsp70/ATP-independent, and C-terminal trimerization, which occurs posttranslationally and is Hsp70/ATP-dependent. To determine if Hsp90 also plays a role in ς1 biogenesis, we analyzed ς1 synthesized in rabbit reticulocyte lysate. Coprecipitation experiments using anti-Hsp90 antibodies revealed that Hsp90 was associated with immature ς1 trimers (hydra-like intermediates with assembled N termini and unassembled C termini) but not with mature trimers. The use of truncated ς1 further demonstrated that only the C-terminal half of ς1 associated with Hsp90. In the presence of the Hsp90 binding drug geldanamycin, N-terminal trimerization proceeded normally, but C-terminal trimerization was blocked. Geldanamycin did not inhibit the association of Hsp90 with ς 1 but prevented the subsequent release of Hsp90 from the immature ς1 complex. We also examined the status of p23, an Hsp90-associated cochaperone. Like Hsp90, p23 only associated with immature ς1 trimers, and this association was mapped to the C-terminal half of ς1. However, unlike Hsp90, p23 was released from the ς1 complex upon the addition of geldanamycin. These results highlight an all-or-none concept of chaperone involvement in different oligomerization domains within a single protein and suggest a possible common usage of chaperones in the regulation of general protein folding and of steroid receptor activation.

It is now known that the folding of nascent proteins in the cytosol is mediated by a group of proteins known as chaperones. These chaperones are believed to be present as large macromolecular complexes whose major roles appear to be the prevention of protein aggregation and the promotion of correct folding and assembly of newly synthesized proteins. Two candidate members of the chaperone family are the heat shock proteins Hsp70 and Hsp90, of which Hsp70 has been well characterized. Through its ATPase activity and associated bind and release cycles, Hsp70 assists in the folding of a wide spectrum of denatured and nascent proteins (see Refs. 1 and 2 for reviews). In contrast, Hsp90 does not exhibit any enzymatic activity and has not been extensively probed in terms of its possible role in the folding of nascent proteins. Indeed, its activity in vitro only approximates that of a true chaperone. Specifically, Hsp90 is able to maintain some denatured proteins in a state competent for refolding by Hsp70 and the cochaperone, Hip (p48), but Hsp90 alone cannot on its own produce refolded proteins (3).
Studies on Hsp90 have focused mainly on the role of this protein in the activation of several families of protein kinases and of steroid hormone receptors. Protein kinases associated with Hsp90 have included receptor tyrosine kinases such as erbB2 (4), nonreceptor tyrosine kinases such as Wee 1 (5) and v-src (6,7), ser/thr kinases such as Raf-1, Mek, and Cdk4 (8 -10) as well as the heme-regulated eukaryotic initiation factor kinase (HRI) 1 (11,12). However, the exact role of Hsp90 in the activation of these kinases remains unclear at present. In contrast, the nature of involvement of Hsp90 in the conformational maturation of hormone receptors is much better understood.
Accumulated evidence suggests that Hsp90 is part of a multiprotein chaperone complex that interacts with steroid receptors, keeping them in a state that is competent for binding substrate yet functionally inactive (for reviews, see Ref. [13][14][15]. In response to steroid binding, the chaperone complex is released, and the receptor becomes activated. In the case of the relatively well studied glucocorticoid receptor, maturation involves a series of complex but ordered interactions between a number of chaperones (16 -20). Reconstitution experiments have shown that an Hsp90⅐p60⅐Hsp70 complex first interacts with the receptor to convert the latter to a steroid binding conformation (19). However, this intermediate complex is highly unstable and requires the additional presence of another cochaperone, p23, for stabilization (20). p23 is also capable of stabilizing a receptor⅐Hsp90 heterocomplex from cytosol (21). Subsequently Hsp70 and p60 leave the complex and are replaced by any of the several immunophilins such as FKBP52 and CyP-40 (22). This complex then binds hormone, and the receptor is then released as an active transcription factor (18).
An invaluable tool in the study of Hsp90 has been the benzoquinone ansamycin, geldanamycin (GA). Originally characterized as an agent responsible for inactivation of select tyrosine kinases, GA has recently been shown to bind with high affinity to a specific binding pocket within Hsp90 (23). Treatment with GA abrogates formation of Hsp90/v-src complexes (24), inhibits the function of steroid hormone receptors (16,25,26), disrupts interaction of Hsp90 with the HRI (11), and targets denatured luciferase and glucocorticoid receptors for proteolytic degradation (25)(26)(27). Thus GA is considered as a very specific inhibitor of Hsp90 function, although the exact mechanism is unclear. Very recently, the GA binding site was found to co-localize with an ATP binding site on Hsp90 (at the N terminus) (23,28), and GA blocks p23 binding to Hsp90, presumably by inducing a conformational change in the p23 binding site (28). This has led to the suggestion that the ATP and GA binding site acts as a conformational switch to regulate the assembly of Hsp90-containing multichaperone complexes (28).
Although Hsp90 is generally accepted as a chaperone, its role in the folding of newly synthesized polypeptides has not been extensively probed until recently. Hartson et al. (29) studied the folding of the lymphoid cell kinase p56 lck translated in vitro and demonstrated the association of Hsp90 with newly synthesized p56 lck molecules. GA was found to disrupt this association and hence the proper folding of this kinase. A more recent study using the heme-regulated eIF-2␣ kinase translated in vitro shows that Hsp90 plays an obligatory role in this kinase acquiring and maintaining a conformation that is competent for transformation into an aggregation-resistant activable kinase (11). The in vitro translation system has also been used extensively in our laboratory to reveal the mechanisms of folding and oligomerization of the reovirus cell attachment protein 1, a trimeric protein positioned at the 12 vertices of the icosahedral virion (30 -36). The 1 trimer is highly asymmetric, with an N-terminal fibrous tail that is anchored to the virion, and a C-terminal globular head that interacts with the cell receptor (37)(38)(39)(40)(41)(42)(43). These two structurally distinct domains are separated by a protease-sensitive hinge region (42,44) and are generated by independent trimerization events (34), with Nterminal trimerization preceding C-terminal trimerization. The core of the N-terminal trimerization domain is the Nterminal one-third of the protein, which is highly ␣-helical and contains an extended heptad repeat of hydrophobic residues (45,46), endowing this region with the intrinsic propensity to form a triple coiled coil. During 1 biogenesis, assembly of three neighboring nascent chains occurs cotranslationally at the N terminus (35,36). This process does not involve Hsp70 or ATP and results in the generation of a loose triple coiled coil (36). This occurs at around the midpoint of the polysome where Hsp70 begins to interact with emerging residues, thereby sterically hindering tightening of the coiled coil. As the triplex moves down the polysome, more Hsp70 becomes associated with the elongating C termini, preventing their misfolding and aggregation. The immature trimer then leaves the polysome as a complex comprised of three 1 subunits, Hsp70, and possibly other chaperones. Subsequent ATP-dependent release of Hsp70 presumably provides the opportunity for the loose coiled coil to quickly snap together (tightening the coiled coil), whereas the remaining portions of the three C termini are available for continued interaction with Hsp70 and other chaperones. This structure, with a stably assembled N terminus and an unassembled C terminus, is called "hydra-like intermediate," and it migrates as a retarded trimer in SDS-PAGE under nondissociating conditions (34,35). Further ATP-dependent release and rebinding of Hsp-70 and other proposed chaperones leads to global assembly and folding of the C terminus, generating mature 1 with the characteristic "lollipop"shaped structure, which migrates as an unretarded trimer in SDS-PAGE under nondissociating conditions (34,35). We contend that the involvement of two mechanistically distinct oligomerization events for the same molecule, one cotranslational and one posttranslational, may represent a common approach to the generation of oligomeric proteins in the cytosol (35).
In the present study, we examined the possible participation of Hsp90 in 1 biogenesis in vitro. We demonstrate that Hsp90 associates with immature but not mature 1. This association occurs at the C-terminal half, but not the N-terminal half of 1. The cochaperone p23 also associates with the C terminus and immature 1. Geldanamycin treatment, which releases p23 but not Hsp90 from 1, has no effect on 1 N-terminal trimerization but effectively blocks C-terminal trimerization. These observations suggest that Hsp90 actively participates in the biogenesis of the trimeric 1 protein, and together with our previous data on Hsp70 involvement (35), are compatible with an all-or-none concept of chaperone involvement in different oligomerization domains within a single protein. They also suggest a possible common usage of chaperones in the regulation of general protein folding and assembly and of steroid receptor activation.

EXPERIMENTAL PROCEDURES
In Vitro Transcription-The plasmids encoding the full-length and various truncated 1 products have been described previously (34 -36, 44). All transcripts were generated in vitro using the MEGAscriptTM transcription kit (Ambion) for Sp6 RNA polymerase promoters. A typical transcription reaction involved incubation of 1 g of linearized plasmid with the prescribed contents of the Ambion transcription kit (total final volume of 20 l) for 5 h at 37°C. The mRNA product was isolated by LiCl precipitation followed by cleanup with the BIO 101, Inc., RNaid Kit. The purified mRNA was then resuspended in 0.1% DEPC-treated water to a final concentration of approximately 0.5 g/l and stored at Ϫ70°C for future use. DEPC is diethyl pyrocarbonate.
In Vitro Translation and Chase-Transcripts were translated in vitro in rabbit reticulocyte lysates (Promega) according to the manufacturer's specifications. Typically, 0.5-1.0 g of mRNA was incubated at 37°C with 7 Ci of [ 35 S]methionine (Amersham Pharmacia Biotech), 1 l of 1.0 mM methionine minus amino acids (Promega), and 18 l of rabbit reticulocyte lysate (Promega) for the duration indicated in figure legends. The labeled product was then analyzed by SDS-PAGE.
To follow the fate of the proteins synthesized, reaction mixtures were centrifuged at 35,000 rpm for 1.0 h at 4°C (Beckman TLA 100.1 rotor, TL-100 tabletop ultracentrifuge) to pellet ribosomes. The supernatants, which had no translation activity, were then incubated at 37°C for various durations and subsequently analyzed by SDS-PAGE as outlined in the figure legends.
Geldanamycin was prepared as a stock solution of 175 M in 20% Me 2 SO. It was added (to a final concentration of 7 M) to the reaction mixture at the onset of translation or to the postribosomal supernatant before the chase, as indicated in the figure legends. For control samples, an equal volume of 20% Me 2 SO was used.
SDS-PAGE-Discontinuous SDS-PAGE was performed using the protocol of Laemmli (47). Depending upon the size of 1 products analyzed, 10 or 12.5% polyacrylamide gels were used. Samples were incubated in protein sample buffer [final concentration: 50 mM Tris (pH 6.8), 1% SDS, 2% ␤-mercaptoethanol, 10% glycerol, and 0.01% bromphenol blue] for 30 min at either 37 or 4°C or, alternatively, boiled for 5 min before SDS-PAGE. The different SDS-PAGE conditions were used to differentiate between N-and C-terminal trimerization. Under dissociating conditions, where samples were boiled for 5 min before SDS-PAGE (carried out at room temperature), both N-and C-terminal trimers dissociated and migrated as monomers. Under nondissociating conditions, samples were either incubated for 30 min at 37°C, at which temperature only the trimeric N-terminal domain is stable, or incubated at 4°C, at which temperature both the N-and C-terminal trimeric domains are stable (SDS-PAGE was carried out at 4°C in both cases).
Immunoprecipitations were carried out on in vitro translations that were performed as outlined above. Typically, a 20-l translation reaction was stopped by the addition of four volumes of TEM buffer (20-mM Tris (pH 7.4), 5 mM EDTA, 10 mM ammonium molybdate, 50 mM NaCl). To this was added primary antibody at a dilution of between 1:35 and 1:150, as empirically determined for each antibody. For primary antibodies of IgG class, samples were incubated on ice for 1 h, and 50 l of IgGsorb (The Enzyme Center) was then added. If primary antibody was of the IgM class, then IgGsorb with preadsorbed IgG anti-IgM was used. After incubation for an additional 30 min with periodic shaking, the samples were microcentrifuged, and the pellets were washed three times with TEM-buffer containing 0.1% Triton-X. The pellets were then resuspended in protein sample buffer, then boiled for 5 min, and analyzed by SDS-PAGE and autoradiography.
For detection of trimeric 1 intermediates, immunoprecipitates were first released from IgGsorb under high pH conditions before nondissociating SDS-PAGE as described previously (32). Briefly, this involved resuspension of the pellet in 70 l of high pH release buffer (0.1% SDS, 2 mM dithiothreitol, 6.0 mM urea, 0.1 M H 3 PO 4 , 50 mM Tris, adjusted to pH 11.6 with NaOH), followed by incubation for 1 h at 37°C. The reaction was then neutralized with 3.5 l of a solution containing 0.9 M H 3 PO 4 , 1.0 M Tris (pH 7.4), and 0.3% SDS. After microcentrifugation, the supernatant was analyzed by SDS-PAGE under nondissociating conditions.
L Cell Binding Assay-The L cell binding assay was essentially the same as described previously (30,34). Monolayers of mouse L cells were grown to 90% confluency on 35-mm Corning tissue culture plates in Joklik's minimum essential medium containing 5% fetal calf serum. Media was aspirated from the plates, and the cells were washed once with ice-cold phosphate-buffered saline (PBS) (pH7.4) followed by incubation at 4°C for 20 min with an overlay of cold PBS. The PBS was then replaced with [ 35 S]methionine-labeled translation reactions diluted 10fold in PBS. After incubation at 4°C for 1 h with intermittent rocking, the monolayers were washed five times with cold PBS. The cells were then lysed with 200 l of lysis buffer (PBS containing 0.5% sodium deoxycholate, 1% Nonidet P-40, and 1 mM phenylmethylsulfonyl fluoride), and the nuclei were pelleted in a microcentrifuge for 2 min at 5000 ϫ g. The supernatant was mixed with protein sample buffer and analyzed by SDS-PAGE.

RESULTS
Association of Hsp90 with Immature but Not Mature 1-We previously reported that Hsp70 is associated with 1 intermediates and is actively involved in 1 biogenesis (35). To determine if Hsp90 was also part of the chaperone complex during 1 assembly, full-length S1 transcripts encoding 1 were translated in rabbit reticulocyte lysates for 12 min then immunoprecipitated with an anti-Hsp90 antibody. Subsequent SDS-PAGE analysis under dissociating conditions (Fig. 1A) revealed that 1 coprecipitated with the anti-Hsp90 antibody but not with a control antibody (anti-CD8), suggesting that at least some 1 was associated with Hsp90. To identify the 1 intermediates associated with Hsp90, similar immunoprecipitates were subjected to SDS-PAGE under nondissociating conditions. The results (Fig. 1B) show that both the apparent monomer and higher order forms of 1 (including the stable hydra form) were associated with Hsp90. Previously we demonstrated that the apparent monomer observed under these weakly denaturing conditions actually represents an early posttranslational form of the 1 trimer (unstable trimer form) that is SDS-sensitive (36). It is a direct presursor of the stable hydra form, which in turn is a direct precursor of mature 1 (compact form) (34,35). The association of Hsp90 with the two active intermediate forms of 1, but not with mature 1 (compact form), suggests that Hsp90 may be playing a dynamic role in 1 trimerization.
Mapping of Domains on 1 That Associate with Hsp90 -We have previously shown that the N-terminal coiled-coil domain of 1 can trimerize independently of ATP and that the triple coiled coil, after denaturation using guanidine hydrochloride, can spontaneously reassemble upon subsequent dialysis (36). This suggests that N-terminal trimerization is independent of chaperone involvement. In contrast, the C-terminal domain of 1 does not fold in the absence of ATP, nor does it spontaneously refold upon denaturation/renaturation (reviewed in Ref. 50). These data are consistent with the observation that Hsp70 only interacts with regions downstream of the ␣-helical coiled coil (35). It was therefore of interest to determine if Hsp90 has the same bias in terms of 1 domain association. To this end, various C-terminal-truncated 1 mutants were probed for their association with Hsp90 using the same coimmunoprecipitation approach as above. All the mutants examined have been previously characterized and found to form N-terminal stable trimers during in vitro translation (35,36). However, none of these trimers, with the exception of full-length 1, is functional for host cell binding because of the lack of C-terminal assembly (42). The results of the coprecipitation experiment are shown in Fig. 2. Only the full-length protein and d90 (lacking the Cterminal 90 amino acids) were efficiently coprecipitated with Hsp90. Some d204 was coprecipitated, but considerably less than was seen for d90; little or no coprecipitation was detected with further deletions. These findings demonstrate that as is the case with Hsp70, Hsp90 association is through the Cterminal half of 1. The lack of Hsp90 association at the Nterminal one-third of 1 is again congruent with the notion that formation of the triple coiled coil is a chaperone-independent process.
GA Interferes with C-terminal Trimerization but Not N-terminal Trimerization of 1-Our demonstration that Hsp90 was associated with 1 intermediates suggests that it probably plays a role in the 1 assembly pathway. That this association was confined to the C-terminal half of 1 further implicates its potential involvement in the formation of the globular head of 1. To determine if this was in fact the case, the effect of the Hsp90-specific inhibitor GA on 1 assembly was examined. Accordingly, protein 1 was translated in vitro in the presence of 7 M GA for 25 min, and the products were then analyzed by SDS-PAGE under nondissociating conditions. The results (Fig.  3A) show that 1 synthesis proceeded normally at 7 M GA (higher concentrations caused some inhibition of translation); however, the formation of the mature compact  1 with lane 2). This suggests that Hsp90 is not involved in the formation of the N-terminal coiled coil and is required only in the final step of 1 biogenesis, namely, formation of the globular head.
That GA had no effect on N-terminal trimerization was confirmed using the d294 mutant that represented the ␣-helical coiled coil region. This mutant was previously shown to undergo ATP-independent trimerization very efficiently when synthesized in vitro (36). We found that d294 trimerization was not impaired by GA (up to 10 M final concentration) that was added at the onset of translation (Fig. 3B). The lack of sensitivity to GA was therefore consistent with the absence of Hsp90 binding sites in this region.
GA Does Not Affect Hsp90 Association with 1 but Prevents Its Subsequent Release from 1-The interference of C-terminal assembly by GA could be because of the inability of GAbound Hsp90 to interact with the C-terminal domain on 1.
Alternatively, GA-bound Hsp90 could still associate with 1 but was unable to be subsequently released such that assembly of the globular head could ensue. To see if GA interfered with Hsp90-1 association, 1 synthesized in vitro in the presence or absence of GA was subjected to coprecipitation analysis using the anti-Hsp90 antibody. The results (Fig. 4) show that essentially the same amount of 1 coprecipitated with Hsp90 in the two reactions, suggesting that GA did not affect Hsp90 association with 1 and that it likely interfered with a subsequent 1 maturation step such as dissociation of Hsp90 from the 1 complex.
We then examined the effect of GA on preformed Hsp90⅐1 complexes. To this end, 1 was translated in vitro under normal conditions. Polysomes were then removed by ultracentrifugation, and the supernatant was chased at 37°C in the presence or absence of GA. Fig. 5A shows that in the absence of GA, the chase of monomeric 1 (unstable hydra) and stable hydra forms to the mature 1 trimer was readily observed (lanes 1-4) as previously reported (35). However, 1 maturation was severely hampered when the chase was carried out in the presence of GA (lanes 5-8). In fact, at later chase times (lane 8), some 1 degradation invariably occurred, an observation con-sistent with previous reports that GA also promotes ubiquitinmediated proteolysis (26,27). The above experiment therefore demonstrates that preformed Hsp90⅐1 complexes are also susceptible to the action of GA, again highlighting the effect of GA as post-(Hsp90⅐1) association.
To determine if subsequent dissociation of Hsp90 from the 1 complex could be blocked by GA, chase samples were subjected to coprecipitation analysis. A 10-min chase time was chosen, because GA-induced 1 degradation was negligible at this time point. The results (Fig. 5B) show that in the absence of GA, less 1 became associated with Hsp90 with time (compare lanes 1 and 2), consistent with the conversion of immature to mature (and presumably Hsp90-free) 1 during this chase period. However, when the chase was carried out in the presence of GA, dissociation of the preformed Hsp90⅐1 complex was not observed (compare lanes 1 and 2 with lanes 1 and 3). It therefore seemed reasonable to conclude that GA blocked 1 C-terminal trimerization by preventing Hsp90 from leaving the immature 1 complexes.
Association of the p23 with 1 Intermediates-p23 is an Hsp90-associated protein that is part of the chaperone system involved in steroid receptor maturation (20,21,51). Whether p23 represents a universal chaperone involved in the folding and assembly of newly synthesized cytosolic proteins in general remains to be seen. We therefore wished to determine if p23 was also part of the chaperone complex during 1 biogenesis. To this end, in vitro translated 1 was subjected to coprecipitation analysis using an anti-p23 antibody. The results (Fig.  6A) revealed a clear association of p23 with 1; however, the amount of 1 coprecipitated was consistently found to be some- what less than that using the anti-Hsp90 antibody (compare lanes 2 and 3). Although the reason for this is unclear at present, it is compatible with the binding of p23 to an existing Hsp90⅐1 complex at a late stage of 1 maturation. To see if p23, like Hsp90, only associated with intermediate 1 forms, immunoprecipitates were analyzed by SDS-PAGE under nondissociating conditions. Fig. 6B shows that indeed, like Hsp90, p23 associated with both the apparent monomer (unstable hydra) and the stable hydra form of 1 but not with mature 1.
Truncation mutants of 1 were then used to map p23 association domains on 1. Again, in agreement with the Hsp90 binding profile, p23 only associated with the full-length 1 and the d90 truncated form but not to a significant degree with the shorter C-terminal truncation mutants (Fig. 7). These results further suggest that Hsp90 and p23 may be present during a similar stage of the 1 folding pathway, and that there may well be coordination between these two proteins in 1 biogenesis.
GA Releases p23 from the 1 Complex-Recent studies have shown that the Hsp90-p23 association is disrupted in the presence of GA (28). This observation, coupled with our current demonstration that GA blocks the release of Hsp90 from the 1 complex (Fig. 5B), led to the interesting question as to the effect of GA on p23-1 association. Accordingly, 1 translated in the presence or absence of GA was subjected to coprecipitation analysis using the anti-p23 antibody. Interestingly, GA, which had no effect on Hsp90 association with 1, inhibited the association of p23 with the 1 complex (Fig. 8A). In view of the reported disruptive effect of GA on Hsp90-p23 association, our data suggest that the association of p23 with 1 is an indirect one, most likely via Hsp90.
The effect of GA on preformed p23⅐1 complexes was then examined. To this end, a postribosomal lysate from a normal 1 translation reaction was prepared and chased at 37°C in the presence or absence of GA. Chase samples were then subjected to coprecipitation analysis using the anti-p23 antibody. The results (Fig. 8B) show that GA efficiently released p23 from preformed p23⅐1 complexes within 5 min while preventing further maturation of 1 (see Fig. 5A). Thus GA manifested opposite effects on Hsp90 and p23 in terms of their association with the 1 complex, with the result being the inability of 1 to undergo the final maturation process, that of formation of the C-terminal globular head.

DISCUSSION
Protein 1 is a fiber-with-knob structure located at the 12 vertices of the reovirus icosahedron. The two morphologically distinct domains serve different functions: the C-terminal globular head contains a conformation-dependent receptor binding domain, whereas the N-terminal fibrous tail anchors 1 to the virion and serves as a stable extension, which presumably facilitates access of the globular head to the host cell receptor. Recent evidence indicates that binding of virion 1 to the cellular receptor (sialic acid) induces a conformational change in the globular head that progresses to the fibrous region of the protein and subsequently spreads to other capsid proteins (52). Thus, despite their pronounced structural and functional differences, there is communication between the two termini during the early stages of reovirus infection.
The discovery that 1 is a trimer (31), coupled with the demonstration that functional trimeric 1 can be generated in an in vitro translation system (42,53) has made 1 an interesting model system for the study of protein oligomerization and folding mechanisms. Generation of 1 involves two independent trimerization events (34). The first event, which leads to the formation of a triple coiled coiled at the N terminus, involves neighboring nascent chains that interact cotranslationally after the ribosomes associated with these chains have traversed past the mid-point of the S1 transcript (35). This event has previously been shown to be ATP-and Hsp70-independent and is believed to occur spontaneously (36). Indeed, truncated trimers representing the N-terminal one-third of 1, which had been dissociated using guanidine hydrochloride, renatured very efficiently upon subsequent dialysis (36). Furthermore, sucrose gradient analyses of in vitro translation products have revealed that, whereas full-length 1 intermediate forms are associated with a large molecular weight complex, no such association is demonstrable with the translation product representing the N-terminal one-third of 1. 2 The present study further demonstrates the lack of involvement of other chaperones such as Hsp90 or p23 in 1 N-terminal trimeriza-tion. It therefore seems safe to conclude that N-terminal trimerization of 1 is a spontaneous event that involves no other participants.
Formation of the C-terminal globular head follows a totally different strategy. First of all, it is global in nature and therefore necessarily occurs posttranslationally. The global nature of this process is suggested by the following observations. First, deletion of as few as four amino acids from the C terminus totally abrogates the cell binding function of 1 (42) as does the single substitution of certain conserved amino acids at the C-terminal half of 1 (53). In both cases, the N-terminal half of the protein remains intact (trimeric and protease-resistant), whereas the C-terminal half is grossly misfolded (unassembled and protease-sensitive). Second, 1 heterotrimers comprised of two wild-type subunits and a mutant subunit with deletions or substitutions at the C terminus are invariably nonfunctional and manifest C-terminal misfolding (34). These observations have led to the prediction that trimerization of the C terminus can proceed only when the C terminus of all three subunits are intact and is accordingly a posttranslational and global event. That this is in fact the case was subsequently demonstrated by following the fate of 1 intermediates in the postribosomal fractions (35). Overall C-terminal trimerization contrasts sharply with N-terminal trimerization in terms of temporality, stringency, and Hsp70 and ATP requirements.
In the present study, we provide evidence for the involvement of Hsp90 in 1 C-terminal but not N-terminal assembly. (i) Hsp90 is associated with 1 intermediates including the stable hydra form with a stably assembled N terminus and an unassembled C terminus, (ii) Hsp90 association sites are found on the C-terminal half but not the N-terminal half of 1, and (iii) the Hsp90 inhibitor geldanamycin blocks C-terminal trimerization but not N-terminal trimerization. The involvement of Hsp90 in 1 folding and assembly is therefore reminiscent of that of Hsp70 previously observed (35). Because Hsp70 and Hsp90 interaction sites on 1 overlap and because both chaperones are involved with a late stage of 1 maturation, it seems reasonable to suggest that they function cooperatively to generate mature 1. Although we do not yet have direct evidence that Hsp70 and Hsp90 interact during 1 biogenesis, the fact that they are often found as a complex in the cytosol suggests that physical association between these two proteins is likely and is part of their chaperoning function. In the case of steroid receptors, recent evidence from reconstitution experiments suggests that a foldosome (comprised of Hsp90, Hsp70, and an additional component called p60) is first formed that then associates with (and hence activates) the steroid receptor (19,20). Whether a similar foldosome is involved in 1 C-terminal assembly remains to be seen. In this regard, it would clearly be of interest to first determine if p60 (or other chaperone-associated proteins such as Hip and the immunophilins) is part of the 1 complex during 1 maturation, and if so, whether it also maps to the same region on 1 as Hsp70 and Hsp90. The involvement of a foldosome in 1 folding and assembly would have significant implications, because it would unify concepts pertaining to steroid receptor activation and those pertaining to chaperone-assisted folding of cytosolic proteins in general.
Another analogy to the steroid receptor activation mechanism is the association of p23 with 1 intermediates. Recent evidence suggests that p23 associates with Hsp90 in the glucocorticoid receptor-Hsp90-p60-Hsp70 assembly intermediate, thereby stabilizing a conformation of Hsp90 that mediates the steroid binding activity of the glucocorticoid receptor (20). Our present observation that p23 and Hsp90 association sites map to the same regions on 1 is compatible with the concept of p23 associating with 1 via Hsp90, although direct binding of 1 by 2 G. Leone and P. W. K. Lee, unpublished data.
FIG. 8. Effect of GA on p23 association with 1. A, full-length S1 mRNA was translated in vitro for 20 min in the presence or absence of 7 M GA. The reactions were then immunoprecipitated (IP) with the anti-p23 antibody and analyzed by denaturing SDS-PAGE. B, effect of GA on preformed p23⅐1 complexes. Full-length S1 mRNA was translated in vitro in the presence of [ 35 S]methionine for 10 min. The reaction was then subjected to ultracentrifugation to pellet the ribosomes, and the supernatant was incubated further at 37°C in the presence or absence of GA. At the times indicated, aliquots were immunoprecipitated with the anti-p23 antibody and analyzed by denaturing SDS-PAGE. p23 has not been ruled out. Whether p23 also plays a stabilizing role in the 1 assembly process remains to be seen. However, the fact that it is part of the 1 complex and that it associates only with the C terminus of 1 makes it likely that it is important for 1 maturation. The additional observation that GA causes release of p23 from the 1 complex while blocking C-terminal assembly is compatible with this view. At present we do not yet have any concrete information on the relative temporal aspects of Hsp90 and p23 involvement in 1 assembly. However, the observation that GA blocks the association of p23 but not the association of Hsp90 to 1 intermediates is compatible with the sequential binding of Hsp90 to the 1 complex followed by that of p23 (possibly to Hsp90). Preliminary data (not shown) from time course studies appear to concur with this notion. These results, if confirmed, would again be reminiscent of the steroid receptor system, further contributing to the concept of shared chaperoning mechanisms between biogenesis of cytosolic proteins and steroid receptor activation.