The Hsp90 Cochaperone p23 Is the Limiting Component of the Multiprotein Hsp90/Hsp70-based Chaperone System in Vivo Where It Acts to Stabilize the Client Protein (cid:1) Hsp90 Complex*

A variety of signaling proteins form heterocomplexes with and are regulated by the heat shock protein chaperone hsp90. These complexes are formed by a multiprotein machinery, including hsp90 and hsp70 as essential and abundant components and Hop, hsp40, and p23 as non-essential cochaperones that are present in much lower abundance in cells. Overexpression of signaling proteins can overwhelm the capacity of this machinery to properly assemble heterocomplexes with hsp90. Here, we show that the limiting component of this assembly machinery in vitro in reticulocyte lysate and in vivo in Sf9 cells is p23. Only a fraction of glucocorticoid receptors (GR) overexpressed in Sf9 cells are in heterocomplex with hsp90 and have steroid binding activity, with the majority of the receptors present as both insoluble and cytosolic GR aggregates. Coexpression of p23 with the GR increases the proportion of cytosolic receptors that are in stable GR (cid:1) hsp90 heterocomplexes with steroid binding activity, a strictly hsp90-dependent activity for the GR. Coexpression of p23 eliminates the insoluble GR aggregates and shifts the cytosolic receptor from very large aggregates without steroid binding activity to (cid:1) 600-kDa heterocomplexes with steroid binding activity. These data lead us to conclude

efficient receptor⅐hsp90 heterocomplex assembly (5,6). It is important to realize that this purified five-protein system is a minimum system for achieving stable heterocomplex assembly. In various cell compartments or under conditions of stress, other proteins, such as the hsp70 cochaperones BAG-1 (Bcl-2associated gene product-1) (7,8) and Hip (Hsc70-interacting protein) (9) and the hsp90 cochaperone Aha (activator of hsp90 ATPase) (10), may also be involved in hsp90 heterocomplex assembly.
We have used glucocorticoid receptor (GR)⅐hsp90 heterocomplex assembly as a model system for studying the assembly process. Hsp90 binds to the ligand binding domain (LBD) of the GR, which must be in heterocomplex with hsp90 for it to bind steroid (1). The steroids bind deep in a hydrophobic cleft that appears to be collapsed in the absence of ligand (11), and the hsp90/hsp70-based chaperone machinery opens the binding cleft of the GR LBD such that it can be accessed by steroid (Ref. 12 and references therein). Because hsp90 is required for high affinity steroid binding in vivo, the generation of steroid binding activity by the five-protein system demonstrates that a physiologically appropriate folding change has occurred upon heterocomplex assembly in vitro. Among the five chaperones in the purified assembly system, we have determined that hsp90 and hsp70 are both essential for opening the steroid binding cleft in the GR LBD, and hsp40, Hop (hsp70/hsp90 organizing protein), and p23 act as cochaperones to increase the rate or extent of GR⅐hsp90 heterocomplex assembly (13). The cochaperones are much less abundant in cell lysates than the essential chaperones, and both Hop and hsp40 (we use the yeast ortholog YDJ-1) are optimally active in the five-protein assembly system when present at less than one-tenth the concentration of hsp70 and hsp90. In contrast, roughly stoichiometric levels of p23 are required for peak activity in the five-protein assembly system. Here, we investigate the consequences of increasing the concentration of each of the cochaperones on GR⅐hsp90 heterocomplex assembly in vivo in Sf9 cells.
When a protein is overexpressed, there is always the possibility that so much of it is produced that it overwhelms the cellular protein folding systems required to generate the physiologically active state. This is the case with baculovirus-mediated expression of the GR in Sf9 cells. Alnemri and Litwack (14) showed that only a small fraction of the total expressed GR was present in the cytosolic fraction as GR⅐hsp90 heterocomplexes, and the majority of the receptor was present as insoluble aggregates, which did not form when a truncated GR lacking the LBD was expressed. Coexpression of the full-length GR with hsp90 or hsp70 did not increase formation of the steroid binding receptor heterocomplex, but incubation of GR partially purified from the aggregates with reticulocyte lysate resulted in reconstitution of GR⅐hsp90 heterocomplexes and restoration * This work was supported by National Institutes of Health Grants DK31573 and CA28010 (to W. B. P.) and ES08365 (to Y. O.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom to correspondence should be addressed:  2 In this paper, we will use the term hsp70 collectively to refer to both the heat shock-induced hsp70 and the constitutively expressed heat shock cognate hsc70. of full steroid binding activity (14). This led to the conclusion that factors other than hsp90 and hsp70 are limiting in insect cells but can be supplied in vitro by reticulocyte lysate.
Here we coexpress the GR in Sf9 cells with Hop, YDJ-1, or p23 and show that p23 is the limiting component of the hsp90/ hsp70-based chaperone system. Increasing p23 increases the fraction of cytosolic receptors that are stably bound to hsp90, and p23 coexpression also eliminates the formation of receptor aggregates. These observations argue strongly that the effects of p23 expression on a client protein function must be interpreted in terms of the ability of p23 to enhance the activity of hsp90 as opposed to direct chaperone effects in which p23 interacts with a partially denatured protein to affect its folding state. From a protein engineering viewpoint, coexpression of p23 may prove to be very useful in the production of the physiologically regulated state of client proteins that normally are present in persistent heterocomplexes with hsp90.

Materials
Untreated rabbit reticulocyte lysate was purchased from Green Hectares (Oregon, WI). [6, H]Dexamethasone (40 Ci/mmol) and 125 I-conjugated goat anti-mouse and goat anti-rabbit IgGs were obtained from PerkinElmer Life Sciences. Sephacryl S-300 was from Amersham Biosciences. Goat anti-mouse and goat anti-rabbit horseradish peroxidase conjugates were from Sigma. The BuGR2 monoclonal IgG antibody against the GR was from Affinity Bioreagents (Golden, CO). The AC88 monoclonal IgG against hsp90, the N27F3-4 anti-72/73-kDa hsp monoclonal IgG (anti-hsp70), and anti-hsp40 polyclonal IgG were from StressGen Biotechnologies (Victoria, BC, Canada). The JJ3 monoclonal IgG against p23 and human p23 cDNA were provided by Dr. David Toft (The Mayo Clinic, Rochester, MN), and the DS14F5 monoclonal IgG against Hop and human Hop cDNA were provided by Dr. David Smith (Mayo Clinic, Scottsdale, AZ). Yeast YDJ-1 cDNA was a gift from Dr. Avrom Caplan (Mount Sinai School of Medicine). The rabbit antiserum against hsp70 and hsp90 used to immunoblot the insect chaperones was a gift from Dr. Ettore Appella (National Cancer Institute). Hybridoma cells producing the FiGR monoclonal IgG used to immunoadsorb the GR were generously provided by Dr. Jack Bodwell (Dartmouth Medical School, Lebanon, NH). Construction of the baculovirus for expression of the mouse GR was described previously (15).

Methods
Construction of Baculoviruses for p23, Hop, and YDJ-1-Human p23 cDNA was excised from the pET23 plasmid (16) with XbaI and XhoI and a 650-bp fragment was inserted into the multiple cloning site of the pFASTBAC vector to make pFB-p23. Human Hop cDNA was excised from the pET23c plasmid (17) with EcoRI and NotI and inserted into the multiple cloning site of the pFASTBAC vector to make pFB-Hop. Yeast YDJ-1 was excised from pET9d plasmid (18) with NcoI and BamHI and then treated with the Klenow fragment of DNA polymerase to create blunt ends. The 1218-bp fragment was inserted into the multiple cloning site of the pFASTBAC vector with StuI to make pFB-YDJ-1.
Overexpression of proteins in Sf9 cells was achieved according to the BAC-TO-BAC transfection kit protocol of Invitrogen. The pFB transfer vectors were transformed into DH10BAC competent cells containing the bacmid DNA and a helper plasmid. The pFASTBAC plasmid can then transpose with the bacmid by the transposition proteins on the helper plasmid. Colonies containing recombinant bacmids were identified by selection of blue/white colonies caused by disruption of the lacZ␣ gene. The bacmid DNA was prepared and then transfected into Sf9 cells with CellFECTIN (Invitrogen) to make recombinant baculovirus according to the manufacturer's protocol.
Expression of Proteins in Sf9 Cells-Sf9 cells were grown in Sf-900 II serum-free medium (Invitrogen) supplemented with Cytomax (Kemp Biotechnology, Rockville, MD) in suspension cultures with continuous shaking (150 rpm). Cultures were infected in log phase of growth with recombinant baculovirus at the multiplicity of infection (m.o.i.) indicated in the legends to each figure. Cultures were supplemented with 0.1% glucose at infection and 24 h post-infection as described by Srinivasan et al. (19). Cells were harvested after 48 h of infection, washed in Hanks' buffered saline solution, resuspended in 1.5 volumes of HEM buffer (10 mM Hepes, pH 7.5, 1 mM EDTA, 20 mM sodium molybdate) containing 1 mM phenylmethylsulfonyl fluoride and Complete Mini (Roche Applied Science) protease inhibitor mixture (1 tablet/7 ml), and ruptured by Dounce homogenization. The lysate was then centrifuged at 100,000 ϫ g for 30 min at 4°C and the supernatant was collected as the cytosolic fraction. In some experiments, aggregated GR in the 100,000 ϫ g pellet was also assayed by immunoblotting. HEM buffer containing phenylmethylsulfonyl fluoride was added to the pellet to the original volume of the lysate, and the pellet was suspended by vigorous Vortex mixing, prior to sonication for 1 min at a setting of 5 with 50% duty cycle. When pellets and cytosols were compared, the same volume of cytosolic and particulate fraction was loaded for gel electrophoresis and immunoblotting.
GR⅐Hsp90 Heterocomplex Reconstitution-Receptors were immunoadsorbed from 50-l aliquots of Sf9 cytosol by rotation for 2 h at 4°C with 14 l of protein A-Sepharose precoupled to 7 l of FiGR ascites suspended in 200 l of TEG buffer (10 mM TES, pH 7.6, 50 mM NaCl, 4 mM EDTA, 10% glycerol). Before incubation with reticulocyte lysate, immunoadsorbed receptors were stripped of associated hsp90 by incubating the immunopellet an additional 2 h at 4°C with 350 l of 0.5 M NaCl in TEG. The pellets were then washed once with 1 ml of TEG, followed by a second wash with 1 ml of Hepes buffer (10 mM Hepes, pH 7.4). FiGR immunopellets containing GR stripped of chaperones were incubated with 50 l of rabbit reticulocyte lysate and 5 l of an ATPregenerating system (50 mM ATP, 250 mM creatine phosphate, 20 mM magnesium acetate, and 100 units/ml creatine phosphokinase). The assay mixtures were incubated for 20 min at 30°C with suspension of the pellets by shaking of the tubes every 2 min. At the end of the incubation, the pellets were washed twice with 1 ml of ice-cold TEGM buffer (TEG with 20 mM sodium molybdate) and assayed for steroid binding capacity and receptor-associated proteins.
Assay of Steroid Binding Capacity-Immune pellets to be assayed for steroid binding were incubated overnight at 4°C in 50 l of HEM buffer plus 50 nM [ 3 H]dexamethasone. Samples were then washed three times with 1 ml of TEGM and counted by liquid scintillation spectrometry. The steroid binding is expressed as counts/min of [ 3 H]dexamethasone bound/FiGR immunopellet prepared from 50 l of cytosol. Steroid binding activity in Sf9 cytosol was assayed by incubating 100-l aliquots of cytosol overnight with 50 nM [ 3 H]dexamethasone in the presence or absence of competing non-radioactive dexamethasone. The incubations were mixed with 1.5 volumes of a dextran-coated charcoal suspension (1% charcoal and 0.2% dextran in 10 mM Hepes, pH 7.4) for 10 min at 0°C. The radioactivity remaining in the charcoal supernatant was assayed, and specific binding was determined by subtracting radioactivity obtained in the presence of competing dexamethasone from that in its absence.
Western Blotting-To assay GR and associated proteins, immune pellets were resolved on 12% SDS-polyacrylamide gels and transferred to Immobilon-P membranes. The membranes were probed with 0.25 g/ml BuGR for GR, 1 g/ml AC88 for rabbit hsp90, 0.05% rabbit antiserum for insect hsp70 and hsp90, 1 g/ml N27F3-4 for rabbit hsp70, 0.1% DS14F5 mouse ascites for human Hop, 0.5% anti-hsp40 for YDJ-1 and rabbit hsp40, or 0.1% JJ3 mouse ascites for human p23. The immunoblots were then incubated a second time with the appropriate 125 I-conjugated or horseradish peroxidase-conjugated counterantibody to visualize the immunoreactive bands.
Protein Purification-Hsp90 and hsp70 were purified from rabbit reticulocyte lysate by sequential chromatography on DE52, hydroxylapatite, and ATP-agarose as described previously (4). Human p23 (16) was purified from 10 ml of bacterial lysate by chromatography on DE52, followed by hydroxylapatite chromatography as described (5). For purification of YDJ-1, bacterial sonicates were cleared by centrifugation, and YDJ-1 was purified by sequential chromatography on DE52 and Relative Abundance of Proteins in Reticulocyte Lysate-To determine the concentrations of the five proteins of the chaperone system in reticulocyte lysate, aliquots of reticulocyte lysate were electrophoresed on SDS-polyacrylamide gels that also contained various amounts of purified hsp90, hsp70, Hop, YDJ-1, and p23 to provide a standard curve for each protein. Immunoblots were prepared and probed with mono-clonal IgGs against each protein, followed by incubation with 125 Ilabeled anti-IgG counterantibody. Samples and purified standards were then excised and counted to permit calculation of the concentration of each protein.
Fractionation of Sf9 Cytosol-Sf9 cytosol (500 l) was incubated overnight with 50 nM [ 3 H]dexamethasone and then applied to a column (1.5 ϫ 113 cm) of Sephacryl S-300. The column was eluted with HEM buffer, and 0.5-ml aliquots of each 2.75-ml fraction were assayed for radioactivity. Aliquots of 100 l were assayed for GR and p23 by SDS-polyacrylamide gel electrophoresis and immunoblotting.  (Table I). After washing, the immunopellets were incubated with [ 3 H]dexamethasone to assay steroid binding activity. The data are the mean Ϯ S.E. for three experiments. C, concentration dependence of the p23 effect. Stripped GR immunopellets were incubated with reticulocyte lysate plus various amounts of purified human p23, and steroid binding activity was assayed.

Relative
p23 and Glucocorticoid Receptor Folding the corresponding insect chaperones in Sf9 cytosol, the relative abundance of the five proteins of the chaperone machinery was assayed in rabbit reticulocyte lysate. Table I and Fig. 1A show the concentrations of each protein in reticulocyte lysate, and Table I presents a comparison with the optimal concentration of the protein in the purified five-protein assembly system. In reticulocyte lysate, the cochaperones Hop, hsp40, and p23 are present at only a fraction of the concentration of the essential chaperones. In the purified system, Hop and hsp40 are optimally active at similarly low molar ratios with respect to hsp90 and hsp70, but p23 is optimally active when it is present at roughly the same concentration as hsp90. Addition of purified hsp90, hsp70, Hop, or YDJ-1 to reticulocyte lysate does not enhance its ability to activate the GR to the steroid binding state, but addition of purified p23 yields a substantial increase in GR⅐hsp90 heterocomplexes with steroid binding activity (Fig. 1B). Maximum activation of GR steroid binding activity is achieved when 4 -8 g of p23 are added to the 50-l incubation mixture (Fig. 1C), a concentration of p23 that is ϳ10-fold higher than that in reticulocyte lysate.
Effect of Coexpression of Each Cochaperone on GR Steroid Binding Activity in Sf9 Cytosols-Before beginning coexpression experiments, we expressed each of the cochaperones in Sf9 cells, purified them, and demonstrated that they had the same activity in the purified five-protein GR⅐hsp90 assembly assay as the bacterially expressed proteins (data not shown). Also, to minimize toxicity related to infection with too much baculovirus, the baculovirus for GR expression was titrated to deter-mine the amount of virus just sufficient to produce the maximum amount of cytosolic GR in 48 h of infection. An m.o.i. of 0.1 was sufficient, and that amount of baculovirus was used in all of the experiments in this work.
The effect of coexpression of each of the cochaperones on GR steroid binding activity in Sf9 cytosol is shown in Fig. 2. In Fig.  2A, it can be seen that p23 can be expressed in quite high levels without affecting the amount of GR, hsp90, or hsp70 in cytosol and without affecting steroid binding activity. Within the range of m.o.i. shown for YDJ-1 expression, there is a slight increase in steroid binding activity that is consistently seen at the lowest level of expression followed by a decrease in binding activity at higher levels of expression (Fig. 2B). Coexpression of Hop does not affect the amount of cytosolic GR, hsp90, or hsp70 or the level of steroid binding activity (Fig. 2C).
Coexpression of p23 Increases GR⅐Hsp90 Heterocomplex Stability in Vivo-We have shown previously that p23 stabilizes GR⅐hsp90 heterocomplexes formed in vitro (20). Receptor⅐hsp90 heterocomplex assembly in cells or reticulocyte lysate is a very dynamic process in which assembly and disassembly are occurring simultaneously (21). In the purified assembly system, it has been shown that the steroid binding cleft is opened to access by steroid at the same rate with or without the presence of p23, but p23 is necessary to produce stable GR⅐hsp90 heterocomplexes that can be immunopurified, washed, and then bound with steroid (20). The receptor-bound hsp90 must achieve its ATP-bound conformation for the receptor to have steroid binding activity (22), and this is the conformation of  that is bound by p23 (23). In binding to the ATP-dependent conformation of receptor-bound hsp90, p23 stabilizes the heterocomplex.
The data of Fig. 3 show that p23 is also stabilizing the GR⅐hsp90 heterocomplexes formed in vivo in Sf9 cells. In these experiments, Sf9 cells were coinfected with GR baculovirus and various amounts of p23 baculovirus. After 48 h of infection, cytosol was prepared and immunoadsorbed with antibody against the GR. Fig. 3A shows the insect hsp90 and human p23 coimmunoadsorbed with the native GR heterocomplexes, as well as duplicate GR immune pellets that have been stripped of insect hsp90. Fig. 3B shows the stripped GR immune pellets after they were incubated with reticulocyte lysate, washed, and immunoblotted for rabbit hsp90 and p23 in the reconstituted heterocomplexes. Fig. 3C shows the steroid binding activity of the immunoadsorbed native GR⅐hsp90 heterocomplexes (solid bars) and the GR⅐hsp90 heterocomplexes reconstituted in reticulocyte lysate (open bars). It can be seen that the steroid binding activity of the immunoadsorbed native GR⅐hsp90 heterocomplexes increases with increasing expression of p23 in the Sf9 cells until it reaches the level of binding achieved by GR⅐hsp90 heterocomplex reconstitution in reticulocyte lysate.

p23 and Glucocorticoid Receptor Folding
The increase in steroid binding activity in native heterocomplexes (Fig. 3C) is accompanied by an increase in GR-bound hsp90 (Fig. 3A). The relationship between GR-bound hsp90 and steroid binding activity with increasing coexpression of p23 is shown in the graph of Fig. 3D.
In that hsp90 is required for steroid binding activity of the GR and increasing p23 coexpression causes increased recovery of cytosolic GR⅐hsp90 heterocomplexes, the effect of p23 in vivo is to stabilize the client protein⅐hsp90 complex. p23 stabilizes the GR⅐hsp90 complex in a dynamic manner (20), and when cytosol is prepared and cytosolic steroid binding is assayed at 0 -4°C, there is the same binding in the absence of p23 coexpression as there is with various levels of coexpression ( Fig.  2A). This suggests that there is sufficient endogenous insect p23 to maintain a limited amount of the overexpressed GR in cytosolic heterocomplex with hsp90, but at any instant, only p23 and Glucocorticoid Receptor Folding ϳ30% of the heterocomplexes are stable GR⅐hsp90⅐p23 complexes that survive immunopurification and washing with retention of steroid binding activity (Fig. 3C). The remainder are unstable GR⅐hsp90 heterocomplexes that do not survive immunopurification and washing but that can be reactivated to the steroid binding state by reticulocyte lysate (Fig. 3C). As the level of p23 is increased in the Sf9 cell by coexpression of the human protein, the percentage of stable cytosolic GR⅐hsp90⅐p23 complexes increases and the fraction of unstable GR⅐hsp90 complexes declines.
To examine the effect of coexpression of p23 on the size of GR heterocomplexes, receptors in Sf9 cytosol were bound with [ 3 H]dexamethasone and the cytosol was fractionated on a column of Sephacryl S-300. As shown in Fig. 4A, the bound steroid elutes in a complex that peaks at fraction 35 (ϳ600 kDa), regardless of whether the Sf9 cells were coexpressing human p23 or not. Fig. 4B shows that in the absence of p23 coexpression, most of the cytosolic GR is present in very large complexes, and coexpression of p23 shifts the GR to the peak at fraction 35. It is possible that the large, non-steroid-binding complexes in cytosol are GR self-aggregates that do not form when GR⅐hsp90 heterocomplexes are stabilized by coexpression of p23.
Coexpression of p23 Reduces GR Aggregate Formation-Alnemri and Litwack (14) found the majority of the overexpressed GR in Sf9 cells to be present in self-aggregates that remain in the pellet when the cell lysate is centrifuged to prepare cytosol. Thus, we asked whether p23 coexpression would affect aggregate formation. As shown in Fig. 5, in the absence of human p23 coexpression, the majority of the GR is in the pellet fraction. However, at higher levels of human p23 expression, there is a marked decrease in the amount of GR in the pellet without substantial change in the amount of cytosolic GR.
Of the three cochaperones in the hsp90/hsp70-based chaperone system, only coexpression of p23 reduces GR aggregate formation. Although the production of aggregates is reduced, the amount of GR in the cytosolic fraction does not increase in a compensatory manner; thus, it is likely that GR degradation is taking place. The GR is known to be degraded by the ubiquitin-proteasome pathway of proteolysis (24), but treatment of Sf9 cells with the proteasome inhibitors lactacystin and MG132 did not change the amount of cytosolic GR or affect the loss of total GR that occurred with high levels of p23 coexpression (data not shown). However, the GR aggregates are formed over 48 h, and it is possible that inhibition of GR⅐hsp90 disassembly by p23 keeps the receptors from aggregating such that they can be shunted into alternative proteolytic pathways. We do not know what limits the amount of cytosolic GR, but we do not achieve higher levels by infection with more GR baculovirus or by coexpression of p23 ( Fig. 2A).
Coexpression of YDJ-1 and Hop-At the lowest multiplicity of infection, coexpression of YDJ-1 modestly increases cytosolic steroid binding activity (Fig. 2B), the amount of GR in heterocomplex with hsp90 (Fig. 6A), and the steroid binding activity of immunoadsorbed receptor (Fig. 6C). At higher levels of YDJ-1 coexpression, there is a progressive decrease in steroid binding activity, and at the highest level, there is a decrease in the amount of receptor protein in cytosol (Fig.  2B). Coexpression of YDJ-1 does not affect the relative fraction of GR recovered in the cytosolic and particulate fractions (data not shown). The fact that the level of steroid binding activity decreases as cytosolic YDJ-1 increases suggests that the level of endogenous insect hsp40 in Sf9 cells is essentially optimal for the proper functioning of the multiprotein chaperone machinery in vivo.
Coexpression of Hop at the multiplicity of infections in Figs. 2C and 7 does not affect the steroid binding activity of GR in cytosol (Fig. 2C) or the proportion of cytosolic receptors in stable heterocomplexes with hsp90 that survive immunoadsorption and washing (Fig. 7). Coexpression of Hop also does not affect the relative amount of GR recovered in cytosolic and particulate fractions (data not shown). Thus, the level of endogenous insect Hop in Sf9 cells appears to be sufficient for the proper functioning of the multiprotein chaperone machinery in vivo.

DISCUSSION
Studies of the mechanism of p23 action on GR⅐hsp90 heterocomplexes derived from in vitro experiments have led to two opposing, but perhaps not exclusive, models. In Dittmar et al. FIG. 5. Coexpression of p23 inhibits GR aggregate formation. p23 was coexpressed with the GR in Sf9 cells, and after 48 h, cells were homogenized and centrifuged at 100,000 ϫ g. A, cytosol (C) and pellet (P) fractions were prepared as described under "Methods," and the same fraction of each preparation was resolved by SDS-polyacrylamide gel electrophoresis and immunoblotting for GR and p23. B, the GR bands labeled with anti-mouse 125 I-IgG were excised and counted for radioactivity to determine the relative amounts of GR in the cytosol (open bars) and pellet (solid bars) fractions compared with the amount of GR present in the corresponding fraction prepared from the control that was not expressing p23.
p23 and Glucocorticoid Receptor Folding (20), we showed that purified p23 stabilized immunopurified native and reconstituted GR⅐hsp90 heterocomplexes. Inhibition of hsp90 dissociation from the receptor was accompanied by stabilization of steroid binding activity. Young and Hartl (25) utilized a rat GR fragment (amino acids 518 -795) comprising the LBD that was expressed in Escherichia coli inclusion bodies and solubilized in guanidine hydrochloride or detergent. The solubilized LBD fragment was partially refolded by slow removal of the denaturant, and LBD⅐hsp90 complexes were formed by incubating this non-native GR LBD with reticulocyte lysate. Addition of purified p23 to these complexes promoted ATP-dependent release of hsp90 from the non-native GR LBD.
Thus, in one model (20), p23 binds to the GR-bound hsp90 in its ATP-dependent conformation and stabilizes the client protein⅐hsp90 complex, whereas in the other model (25), p23 functions as a client protein release factor for hsp90. The two models may not contradict each other but may reflect the study of p23 effects on hsp90 binding to native GR (20) versus an extensively denatured LBD fragment (25). Here, we have seen that coexpression of p23 with the GR results in increased recovery of stable GR⅐hsp90 heterocomplexes with a concomitant increase in steroid binding activity (Fig. 3). Thus, in vivo in Sf9 cells, p23 stabilizes the client protein⅐hsp90 interaction.
In addition to its binding to hsp90 and affecting hsp90 function, p23 has been shown to directly interact with denatured protein to maintain a non-native folding-competent intermediate (26). Some effects of p23 in vivo have been attributed to such a direct passive chaperoning activity. In particular, p23 has been shown to promote the disassembly of transcription factors (including the GR) from their appropriate response elements (27,28). Because overexpression of hsp90 yielded only a slight activity, the much stronger effect of p23 has been assumed to reflect its intrinsic chaperone activity, rather than its ability to affect hsp90 function (27,28). However, with respect to the GR, there is good reason to focus on a p23 effect on hsp90. It was shown many years ago in in vitro experiments that the GR does not have DNA binding activity when it is in heterocomplex with hsp90 (29) and that the chaperone machinery in reticulocyte lysate acts on DNA-bound, hormone-free GR to convert the receptor to the non-DNA binding state and restore steroid binding activity (3). In permeabilized cells, Liu and DeFranco (31) have shown that, during hormone withdrawal, both release of GR from chromatin and generation of high affinity steroid binding activity in the nucleus are inhibited by geldanamycin, a specific inhibitor of the hsp90 family of chaperones.

p23 and Glucocorticoid Receptor Folding
Here, we have shown that expression of p23 results in increased stable hsp90 binding to the GR, as well as increased steroid binding activity, an hsp90-dependent activity for this receptor. Our conclusion is that p23 affects client protein function in vivo by stabilizing hsp90 association with the client protein. The notion that p23 effects in vivo reflect its action on hsp90 is supported in a recent study by Oxelmark et al. (32) on p23 effects on signal transduction by the estrogen receptor. A variety of mutations in the yeast ortholog of p23 were selected on the basis of their ability to affect estrogen receptor-dependent transcriptional activation. The ability of p23 mutants to increase or decrease estrogen receptor signal transduction correlated with their ability to bind to hsp90 (32). Taken together, the observations of this paper and those of Oxelmark et al. (32) argue strongly that p23 effects in vivo do not reflect a direct chaperoning interaction with the client protein of hsp90; rather, they reflect a direct interaction with hsp90 to stabilize its association with the client protein.
Genetic studies in yeast have shown that p23 is not essential for the action of hsp90 client proteins in vivo (30,33). Here, we have shown that p23 is the component of the five-protein hsp90/hsp70-based chaperone system that is limiting for the production of stable client protein⅐hsp90 heterocomplexes in vivo. Although p23 is present at only ϳ1/6th the concentration of hsp90 (Table I), that level is probably optimal for the proper functioning of the multiprotein chaperone system in vivo. A dynamic cycle of heterocomplex assembly with hsp90 and dissociation from hsp90 is probably very important for hsp90 regulation of client protein function. Although expression of more p23 can decrease the rate of hsp90 heterocomplex disassembly that may be detrimental to hsp90-mediated regulation of signal transduction overall. Thus, when we say that p23 is the limiting component of the chaperone machinery, we mean that it is limiting for producing the stable client protein⅐hsp90 heterocomplexes and hsp90-dependent effects that we as investigators would like to study. From the viewpoint of the researcher it is important that p23 is limiting, because overexpression of p23 presents us with another way to study effects of