Evidence that the peptidylprolyl isomerase domain of the hsp90-binding immunophilin FKBP52 is involved in both dynein interaction and glucocorticoid receptor movement to the nucleus.

We have previously shown that immunoadsorption of the FKBP52 immunophilin component of steroid receptor.hsp90 heterocomplexes is accompanied by coadsorption of cytoplasmic dynein, a motor protein involved in retrograde transport of vesicles toward the nucleus. Coimmunoadsorption of dynein is competed by an expressed fragment of FKBP52 comprising its peptidylprolyl isomerase (PPIase) domain (Silverstein, A. M., Galigniana, M. D., Kanelakis, K. C., Radanyi, C., Renoir, J.-M., and Pratt, W. B. (1999) J. Biol. Chem. 52, 36980-36986). Here we show that cotransfection of 3T3 cells with the FKBP52 PPIase domain and a green fluorescent protein (GFP) glucocorticoid receptor (GR) chimera inhibits dexamethasone-dependent movement of the GFP-GR from the cytoplasm to the nucleus. Cotransfection with FKBP12 does not affect GFP-GR movement. Inhibition of movement by the FKBP52 PPIase domain is abrogated in cells treated with colcemid to eliminate microtubules prior to steroid addition. After withdrawal of colcemid, microtubules reform, and PPIase inhibition of GFP-GR movement is restored. These observations are consistent with the notion that FKBP52 targets retrograde movement of the GFP-GR along microtubules by linking the receptor to the dynein motor. Here, we also show that native GR.hsp90 heterocomplexes immunoadsorbed from L cell cytosol contain dynein and that GR.hsp90 heterocomplexes assembled in reticulocyte lysate contain cytoplasmic dynein in a manner that is competed by the PPIase domain of FKBP52.

Steroid receptors move continuously into and out of the nucleus (Refs. 1-4; for review, see Ref. 5), and, depending upon the receptor, the hormone-free, untransformed receptor may have a predominantly nuclear or cytoplasmic localization. The hormone-free glucocorticoid receptor (GR) 1 is localized to the cytoplasm of most cells, and after steroid binding and transformation, it translocates to the nucleus (6 -8). Several studies with inhibitors suggest that the multiprotein hsp90-based chaperone system and the hsp90-binding immunophilin FKBP52 are involved in movement of the GR along microtubular tracks to the nucleus (for review, see Ref. 9).
Assembly of receptors into heterocomplexes with hsp90 is a dynamic process (10), and it has been shown that the GR and hsp90 can move together from the cytoplasm to the nucleus (11). A couple of observations suggest that the role of hsp90 in receptor movement is likely to involve dynamic assembly and disassembly of GR⅐hsp90 heterocomplexes. For example, Yang and DeFranco (12) showed that molybdate, which binds to hsp90 and stabilizes GR⅐hsp90 heterocomplexes in vivo (13), traps the GR in the cytoplasm of cells continuously exposed to hormone. Molybdate in this case was thought to inhibit reimport of the GR into the nucleus by inhibiting the dynamic cycling of receptors into and out of their complexes with the hsp90 chaperone. Also, geldanamycin, an antibiotic that binds to the nucleotide binding site on hsp90 (14) and prevents formation of normal receptor⅐hsp90 heterocomplexes (15), impedes steroid-induced movement of the GR from the cytoplasm to the nucleus (16,17).
Some localization studies have shown the untransformed GR to colocalize with microtubules (for review, see Ref. 18), but the evidence supporting movement along microtubular tracks is indirect. Although microtubule disrupting agents, such as colcemid, do not affect the overall rate of steroid-dependent receptor translocation to the nucleus (8,19), they eliminate the hsp90-dependent mode of receptor movement (17). Using a fusion protein of murine GR with Aequorea green fluorescent protein (GFP), we found that steroid-dependent GFP-GR translocation to the nucleus is rapid (t1 ⁄2 ϭ ϳ5 min) both in cells with intact cytoskeleton and in cells with disrupted cytoskeletal networks (17). However, in cells with normal cytoskeleton, the hsp90 inhibitor geldanamycin slowed translocation of the GFP-GR by an order of magnitude (t1 ⁄2 ϭ ϳ45 min), whereas in cells with colcemid-disrupted microtubules, geldanamycin had no effect on the translocation rate (t1 ⁄2 ϭ ϳ5 min). This suggests two mechanisms of GR movement. Under physiological conditions where the cytoskeleton is intact, diffusion is limited, and the GFP-GR utilizes a movement machinery in which the hsp90 heterocomplex assembly machinery plays a role. In cells where the microtubules are disrupted with colcemid, movement is still steroid-dependent, but the transformed GR moves through the cytoplasm by diffusion (17,20).
In normal cells, the retrograde movement of vesicles toward the nucleus is known to occur on cytoskeletal tracks in a process requiring the molecular motor protein cytoplasmic dynein (for review, see Refs. 21 and 22). However, like GFP-GR translocation to the nucleus, short range vesicle movement still occurs by diffusion when microtubules are disrupted (23). Using retrograde vesicular movement as a model, we have looked for potential links between the GR and cytoplasmic dynein. In addition to hsp90, steroid receptor heterocomplexes contain hsp90-binding immunophilins (for review, see Ref. 24), and we have shown that cytoplasmic dynein is coimmunoadsorbed from cytosols with FKBP52, one of the major immunophilins in GR⅐hsp90 heterocomplexes.
The immunophilins are proteins possessing peptidylprolyl isomerase (PPIase) domains that bind immunosuppressant drugs of the FK506 group (the FKBPs) or of the cyclosporin A group (the cyclosporin A-binding proteins). Three high molecular weight immunophilins, FKBP52, FKBP51, and cyclosporin A-binding protein-40, exist in receptor⅐hsp90 heterocomplexes (24). In addition to their PPIase domains, these high molecular weight immunophilins contain three tetratricopeptide repeats (TPRs), which are degenerative sequences of 34 amino acids that determine their binding to a common TPR acceptor site on hsp90 (25)(26)(27). We have shown that FKBP52 also binds directly to the GR and that a 35-amino acid segment spanning the proto-nuclear localization signals that comprise the NL1 nuclear localization signal of the GR (see Ref. 5 for NL1 composition) is sufficient for FKBP52 binding (28). FKBP52 binds either directly or via a linker protein with cytoplasmic dynein (29), and coimmunoadsorption of dynein is competed by a fragment of FKBP52 comprising its PPIase domain (28).
These binding studies support the notion (30, 31) that FKBP52, which itself colocalizes with microtubules (29,32), may target receptor movement toward the nucleus by determining its attachment to the retrograde dynein motor protein.
To date, the only in vivo observation linking FKBP52 to GR movement is the demonstration that microinjection of an antibody raised against a conserved negative sequence (6 of 8 amino acids) between domains I (PPIase) and II of FKBP52 (33) impedes steroid-dependent cytoplasmic-nuclear translocation of the endogenous GR in L cells (8).
In this work, we show that cotransfection of 3T3 fibroblasts with GFP-GR and the PPIase domain of FKBP52 slows subsequent steroid-dependent translocation of the GFP-GR to the nucleus (t1 ⁄2 ϭ ϳ45 min) to the same extent as treatment with the hsp90 inhibitor, geldanamycin. The effect is specific, inasmuch as cotransfection with FKBP12 does not affect GFP-GR movement. If cells are treated with colcemid to disrupt microtubules prior to steroid addition, the presence of competing PPIase domain does not inhibit movement. But, when colcemid is removed and microtubules reform, GFP-GR translocation is again slowed by the presence of the PPIase domain. The observations are consistent with an uncoupling of the receptor from movement along microtubules by competition for the binding of FKBP52 to the dynein motor.

Materials
NIH 3T3 mouse embryo fibroblasts were purchased from the American Type Culture Collection (Manassas, VA). Phenol red-free Dulbecco's modified Eagle's medium (DMEM) was from Life Technologies, Inc. Charcoal-stripped calf serum and colcemid were from Sigma. The Trans-Fast kit for cell transfection was from Promega (Madison, WI). The mouse monoclonal IgG (MAB1618) against the 74-kDa intermedi-ate chain subunit of mammalian (bovine brain) cytoplasmic dynein was purchased from Chemicon Intl. (Temecula, CA). Rabbit reticulocyte lysate was from Green Hectares (Oregon, WI), and geldanamycin was obtained from the Drug Synthesis and Chemistry branch of the Developmental Therapeutics Program, National Cancer Institute. Complete-Mini protease inhibitor tablets were from Roche Molecular Biochemicals. The EC1 anti-FKBP52 mouse monoclonal IgG was kindly provided by Dr. Lee Faber (Medical College of Ohio, Toledo, OH), and the UPJ56 antiserum against FKBP52 (34) was a gift from Dr, Karen Leach (Pharmacia and Upjohn, Inc., Kalamazoo, MI). The BuGR2 mouse monoclonal IgG used for immunoblotting the GR was from Affinity Bioreagents (Golden, CO), and the FiGR monoclonal IgG used for GR immunoadsorption was generously provided by Dr. Jack Bodwell (Dartmouth Medical School). The AC88 mouse monoclonal IgG against hsp90 was from StressGen (Victoria, Canada), and 125 I-conjugated goat antimouse IgG was from PerkinElmer Life Sciences. Construction of the GFP-GR was described previously (17). The cDNAs for rabbit FKBP52, FKBP52 Met 6 -Gly 148 (Domain I plus hinge), FKBP52 Gly 32 -Lys 138 (Domain I core), and human FKBP12 were described previously (35,36) and were subcloned into the pSG5PL mammalian expression vector from Stratagene (La Jolla, CA). The baculovirus for mouse GR was obtained from Dr. Edwin Sanchez (Medical College of Ohio, Toledo, OH) and was described previously (37). The baculovirus for the FLAGtagged TPR domain of rat PP5 (38) was kindly provided by Dr. Michael Chinkers (University of South Alabama, Mobile, AL).

Methods
Cell Culture and Transfection-NIH 3T3 mouse fibroblasts were grown on 11 ϫ 22-mm coverslips placed in 35-mm culture plates in 2 ml of DMEM containing 10% bovine calf serum, under a 5% CO 2 atmosphere at 37°C. When the cells were ϳ60% confluent, the medium was replaced by 0.75 ml of DMEM with 5% calf serum, and the incubation was continued for 1 h. Cells were then transfected with 2.4 g of plasmid encoding GFP-GR and 16 g of pSG5PL-FKBP52 domain I plus hinge (Met 6 -Gly 148 ), or pSG5PL-FKBP52 core domain I (Gly 32 -Lys 138 ), or pSG5PL-FKBP12, or pSG5PL vector alone. The cotransfection mixture was prepared by preincubating plasmids with 4.5 l of Trans-Fast reagent/g of total DNA and 0.25 ml of DMEM for 10 min at room temperature. The mixture was then added to the cell cultures and incubated for 1 h, prior to adding 4 ml of DMEM with 10% calf serum without washing out the mixture used for transfection. After 18 h, the medium was removed, cells were washed four times, 2 ml of phenol red-free DMEM with 10% charcoal-stripped calf serum were added, and cells were incubated for an additional 24 -30 h.
GFP-GR Translocation-Cells transfected as described above were incubated with 1 M dexamethasone, and at various times, coverslips were removed, and cells were rinsed with ice-cold phosphate-buffered saline and fixed by immersion in Ϫ20°C methanol for at least 15 min. Cells were rinsed again with phosphate-buffered saline, and the coverslips were inverted onto a slide with 5 l of mounting solution (1 mg/ml p-phenylenediamine in 10% phosphate-buffered saline, 90% glycerol, pH 9.0). Cells were visualized with a Leitz Aristoplan epiillumination microscope.
Cells were scored for GFP-GR translocation as we have described previously (17), using a score of 4 for nuclear fluorescence much greater than cytoplasmic fluorescence, 3 for nuclear fluorescence greater than cytoplasmic fluorescence, 2 for nuclear fluorescence equal to cytoplasmic fluorescence, 1 for nuclear fluorescence less than cytoplasmic fluorescence, and 0 for nuclear fluorescence much less than cytoplasmic fluorescence. The translocation scores represent the means Ϯ S.E. from three experiments in which Ͼ100 cells per condition per experiment were scored. Significance of differences was measured by analysis of variance followed by Bonferroni t test.
Immunoadsorption of FKBP52-For immunoadsorption of FKBP52, aliquots (150 l) of rabbit reticulocyte lysate were immunoadsorbed for 3 h at 4°C to 14 l of protein A-Sepharose with 7 l of UPJ56 antiserum against FKBP52. Immune pellets were washed three times by suspension in 1 ml of TEG buffer (10 mM TES, pH 7.6, 50 mM NaCl, 4 mM EDTA, 10% (w/v) glycerol) with 20 mM sodium molybdate prior to gel electrophoresis and immunoblotting using 0.1% EC1 to probe for FKBP52.
Glucocorticoid Receptor Heterocomplex Reconstitution-Mouse GR was overexpressed in Sf9 cells as described (37), and cytosol was prepared by Dounce homogenization in 1.5 volumes of buffer (10 mM Hepes, pH 7.4, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 1 tablet of Complete-Mini protease inhibitor mix/3 ml of buffer). The lysate was centrifuged at 100,000 ϫ g for 30 min, and the supernatant was stored at Ϫ70°C. Receptors were immunoadsorbed from 60-l aliquots of cytosol by rotation for 2 h at 4°C with 14 l of protein A-Sepharose and 7 l of FiGR ascites, and the immune pellets were washed once with 1 ml of TEG buffer. Prior to incubation with reticulocyte lysate, immunoadsorbed receptors were stripped of hsp90 by incubating for 2 h with 300 l of TEG buffer containing 0.7 M NaCl and 0.2% Nonidet P-40. The immune pellets were then washed three times with 1 ml of TEG buffer containing 0.7 M NaCl and twice with 10 mM Hepes, pH 7.4. GR⅐hsp90 heterocomplexes were assembled by incubating these stripped GR immune pellets with 50 l of rabbit reticulocyte lysate and 5 l of an ATP-regenerating system (50 mM ATP, 250 mM creatine phosphate, 20 mM magnesium acetate, and 100 units/ml of creatine phosphokinase). The assay mixtures were incubated for 30 min at 30°C, with suspension of the pellets by shaking the tubes every 2 min. At the end of the incubation, the pellets were washed three times with 1 ml of ice-cold TEGM buffer (TEG with 20 mM sodium molybdate), and boiled in SDS sample buffer.
Western Blotting-To assay GR and associated proteins, immune pellets were resolved on 10% SDS-polyacrylamide gels, and proteins were transferred to Immobilon-P membranes. The membranes were probed with 0.2 g/ml BuGR for GR, 1 g/ml AC88 for hsp90, 0.1% UPJ56 for FKBP52, and 0.1% of MAB1618 for dynein. The immunoblots were then incubated a second time with the appropriate 125 I-conjugated counter antibody to visualize immunoreactive bands.
Immunoadsorption of Native GR⅐hsp90 Heterocomplexes-Cytosol was prepared from L929 mouse fibroblasts by Dounce homogenization in 1.5 volumes of buffer (10 mM Hepes, pH 7.4, 1 mM EDTA, 20 mM molybdate, and 1 tablet of Complete-Mini protease inhibitor mix/3 ml of buffer) and centrifugation at 100,000 ϫ g for 30 min. Receptors were immunoadsorbed from 250-ml aliquots of cytosol by rotation for 2.5 h at 4°C with 14 l of protein A-Sepharose and 7 l of FiGR ascites, and the immune pellets were washed three times with 1 ml of TEG buffer with 20 mM molybdate.
Purification of FKBP52 Domain I Core-pGEX T plasmid expressing GST-FKBP52 Gly 32 -Lys 138 was used to transform Escherichia coli strain UT5600. The fusion protein was purified by binding to GSHagarose beads, and the core PPIase domain I of FKBP52 was released by incubation at 4°C with thrombin.

FKBP52 Domain I Inhibits Steroid-dependent GFP-GR
Translocation-Because the purified FKBP52 domain I (Met 6 -Gly 148 ) fragment competes for FKBP52 association with cytoplasmic dynein in vitro (28), we asked whether coexpression of domain I would alter the rate of GFP-GR movement from the cytoplasm to the nucleus in vivo. In the absence of steroid, the GFP-GR expressed in 3T3 cells is predominantly localized to the cytoplasm (Fig. 1, A, C, and E), and upon exposure to dexamethasone, it moves to the nucleus (Fig. 1B). As we reported previously (17), nuclear translocation of the GFP-GR is inhibited when hsp90 function is inhibited by geldanamycin (Fig. 1F). As shown in Fig. 1D, cotransfection with FKBP52 domain I also inhibits GFP-GR translocation. The bar graphs on the right in Fig. 1 summarize the translocation scores from several hundred cells per condition, where a value of zero reflects a completely cytoplasmic localization and a value of four a completely nuclear localization, a scoring system we have used previously with the GFP-GR (17,20). After the 20-min interval of steroid treatment, nuclear translocation of the GFP-GR is inhibited to the same extent by coexpression of FKBP52 domain I as it is by treatment with geldanamycin.
Inhibition of GFP-GR Translocation Requires Microtubules-Geldanamycin slows the rate of GFP-GR translocation, but by 60 min of steroid exposure, essentially all of the receptors have moved to the nucleus (Ref. 17 and Fig. 2A). As shown in Fig. 2A, cotransfection with FKBP52 domain I slows GFP-GR translocation to the same extent as geldanamycin, and geldanamycin treatment of cells coexpressing FKBP52 domain I does not yield a slower rate of translocation than that seen with either condition alone. As shown in Fig. 2B, neither coexpression of FKBP52 domain I nor treatment with geldana- mycin slows GFP-GR translocation in cells that have been treated with colcemid to disrupt microtubules.
We have shown previously, that treatment with colcemid under these conditions yields complete disruption of microtubules in 3T3 cells (17). When colcemid-treated cells are incubated in medium without colcemid for 1 h, the microtubular network is restored, and geldanamycin again inhibits GFP-GR movement (17). In the experiment of Fig. 3A, cells that were treated with colcemid, such that steroid-dependent GFP-GR movement was not inhibited by FKBP52 domain I (condition 3) or geldanamycin (condition 4), were withdrawn from colcemid for 1 h prior to steroid addition. In the colcemid-withdrawn cells, GFP-GR translocation was inhibited by FKBP52 domain I (condition 5) and by geldanamycim (condition 6). These observations strongly suggest that, under normal physiological conditions where the microtubular network is intact, the GFP-GR is linked via an associated immunophilin, such as FKBP52, to a system for movement along microtubular tracks.
Specificity of Inhibition of GFP-GR Translocation-The FKBP52 domain I (Met 6 -Gly 148 ) expressed in the experiments of Figs. 1, 2, and 3A contains all of domain I and a short, negatively charged hinge region that connects domains I and II. To be sure that the PPIase domain was responsible for inhibition of receptor movement, we coexpressed the GFP-GR with FKBP52 core domain I (Gly 32 -Lys 138 ), which does not contain the hinge segment. As shown in Fig. 3B, GFP-GR translocation is inhibited to the same extent in cells expressing FKBP52 core domain I (condition 4) as in cells expressing FKBP52 domain I plus hinge (condition 3). We determined by immunoblotting that the levels of both FKBP52 domain I (Met 6 -Gly 148 ) and core domain I (Gly 32 -Lys 138 ) were 35-fold higher than the endogenous FKBP52 in nontransfected cells (data not shown).
Although domain I alone has the same PPIase activity as the full-length FKBP52 (35), that enzymatic activity is not related to inhibition of GFP-GR movement. This is deduced from the fact that the immunosuppressant drug FK506, which binds in the active site and blocks PPIase activity, does not affect either the rate of steroid-dependent GFP-GR translocation or inhibition of the rate of translocation by FKBP52 domain I (data not shown). We have also shown previously that FK506 does not affect FKBP52 binding to cytoplasmic dynein (28).

FIG. 2. Expression of FKBP52 domain I inhibits GFP-GR movement only when microtubules are present. A, inhibition of GFP-GR movement by domain I and geldanamycin. 3T3 cells transfected with GFP-GR and either pSG5PL-FKBP52 domain I (Met 6 -
Gly 148 ) (Domain I) or empty pSG5PL vector (Vector). After 2 days, the cells were placed in ice and incubated for 90 min with 1 M DEX, 10 M geldanamycin (GA) was then added, and the incubation on ice was continued for 30 min. The temperature was then shifted to 37°C to permit nuclear translocation, and immunofluorescence was assayed at various times. Each value represents the translocation score (mean Ϯ S.E.) of three independent experiments. The values for vector ϩ geldanamycin, Domain I, and Domain I ϩ geldanamycin are all significantly different from vector alone at p Ͻ 0.01 from 5 min through 40 min and are not different from each other. B, colcemid disruption of microtubules eliminates inhibition of GFP-GR movement by both FKBP52 domain I and geldanamycin. 3T3 cells transfected as above were incubated for 1 h with 0.6 g/ml colcemid to disrupt microtubules prior to incubating them on ice with DEX and geldanamycin and shifting them to 37°C to permit nuclear translocation. The apparent slight lag in the kinetics at the first time point for two of the conditions is unique and has not been observed with geldanamycin in other experiments (17).

FIG. 3. The colcemid blockade of inhibition is reversible, and there is no inhibition of GFP-GR movement by FKBP12.
A, the effect of colcemid on the inhibition of GFP-GR translocation rate is reversible. 3T3 cells transfected with GFP-GR and either vector or FKBP52 domain I were treated for 1 h with 0.6 g/ml colcemid (conditions 1-6) prior to assay of DEX-dependent translocation to the nucleus. Column 1, transfected with vector, no DEX treatment; column 2, cells transfected with vector and treated with DEX; column 3, cells expressing FKBP52 core domain I treated with DEX; column 4, cells transfected with vector treated with DEX and geldanamycin; column 5, cells expressing FKBP52 core domain I and treated with colcemid were incubated in colcemid-free medium for 1 h to permit microtubule repolymerization prior to treatment with DEX; column 6, cells transfected with vector and treated with colcemid were incubated in colcemid-free medium for 1 h prior to treatment with DEX and geldanamycin. Conditions in columns 5 and 6 are significantly different from conditions in columns 3 and 4, respectively, at p Ͻ 0.001. B, domain I core inhibits GFP-GR translocation but FKBP12 does not. 3T3 cells were transfected with GFP-GR and pSG5PL-FKBP52 domain I plus hinge (Met 6 -Gly 148 ), or pSG5PL-FKBP52 core domain I (Gly 32 -Lys 138 ), or pSG5PL-FKBP12, or empty pSG5PL vector. Two days later, DEX was added for 20 min, and fluorescence localization was determined. Column 1, transfected with vector, no DEX treatment; column 2, cells transfected with vector and treated with DEX; column 3, cells expressing FKBP52 domain I treated with DEX; column 4, cells expressing FKBP52 core domain I treated with DEX; column 5, cells expressing FKBP12 treated with DEX; column 6, cells transfected with vector and treated with DEX and geldanamycin as described under Fig. 2. The data represent the mean translocation score Ϯ S.E. for 150 cells/condition. Conditions in columns 3, 4, and 6 are significantly different from conditions in column 2 at p Ͻ 0.001.
Inhibition of GFP-GR movement by FKBP52 domain I does not reflect an inhibition by PPIase domains in general. We have shown previously that purified FKBP12 does not affect FKBP52 binding to cytoplasmic dynein (28). In Fig. 3B (condition 5), it is shown that cotransfection of 3T3 cells with human FKBP12 does not affect steroid-dependent GFP-GR translocation. The level of FKBP12 in transfected cells was determined by [ 3 H]FK506 binding to be 40-fold higher than all endogenous [ 3 H]FK506 binding activity in nontransfected cells.
GR⅐hsp90 Heterocomplexes Containing Cytoplasmic Dynein-The experiment of Fig. 4 presents three treatments that selectively affect the binding of FKBP52 to hsp90 or dynein. In this experiment, aliquots of rabbit reticulocyte lysate were incubated with various additions at 30°C, and then FKBP52 was immunoadsorbed. Immunoadsorption of FKBP52 is accompanied by coadsorption of both hsp90 and cytoplasmic dynein (lane 2). The TPR domain fragment of rat PP5 markedly reduces FKBP52 binding to hsp90 but not to dynein (lane 3), whereas FKBP52 core domain I eliminates FKBP52 binding to dynein but does not affect its binding to hsp90 (lane 4). FKBP52 binding to dynein is not affected by FK506 (lane 6), which again suggests that the binding is independent of PPIase activity (28). Importantly, pretreatment of reticulocyte lysate with geldanamycin markedly reduces FKBP52 binding to hsp90 (lane 8).
Reticulocyte lysate contains the multi-protein hsp90/hsp70based chaperone machinery that assembles GR⅐hsp90 heterocomplexes (reviewed in Ref. 24). In the experiment of Fig. 5A, we asked whether the GR⅐hsp90 heterocomplex formed in reticulocyte contained dynein. Immunoimmobilized mouse GR that was stripped of its associated hsp90 was incubated with rabbit reticulocyte lysate, and after washing the immunopellet, the GR and associated rabbit proteins were resolved by SDSpolyacrylamide gel electrophoresis and immunoblotting. As shown in lane 3, the GR is assembled into heterocomplexes containing hsp90, FKBP52, and cytoplasmic dynein. If the lysate is preincubated with geldanamycin before it is incubated with the GR, then there is a marked decrease in hsp90, and there is no FKBP52 or dynein (lane 4). FKBP52 is bound to hsp90 via its TPR domain (25), and preincubation with a TPR domain fragment of PP5 yields a GR⅐hsp90 complex that does not contain FKBP52 or dynein (lane 5). As shown in Fig. 5B, preincubation with FKBP52 core domain I, eliminates dynein without affecting the amount of hsp90 or FKBP52 in the GR heterocomplex. These observations are consistent with a model in which the hsp90-bound GR is linked to the motor protein via an immunophilin, such as FKBP52.
In Fig. 5C, the GR was immunoadsorbed from cytosol prepared from mouse L cells. Cytoplasmic dynein was coimmuno-adsorbed, showing that it is a component of native GR⅐hsp90 heterocomplexes formed in the cell.

DISCUSSION
Inhibition of GFP-GR translocation by expression of the PPIase domain of FKBP52 is consistent with a model in which rapid receptor movement in cells with intact cytoskeleton requires participation of one or more of the high molecular weight, hsp90-binding immunophilins (8). We have previously reported that the rate of GR translocation is inhibited by geldanamycin (16,17), which blocks GR⅐hsp90 heterocomplex assembly, but the way in which dynamic GR⅐hsp90 heterocomplex assembly might promote GR movement along cytoskeletal tracts was unknown. FKBP52 provides a link between the receptor heterocomplex and cytoplasmic dynein, a motor protein for retrograde movement along microtubules. Treatment of reticulocyte lysate with geldanamycin inhibits FKBP52 binding to hsp90 (Fig. 4) and inhibits GR⅐hsp90 assembly, yielding receptors that are not associated with FKBP52 and dynein (Fig. 5A). A similar effect in vivo could explain why rapid GR movement in normal cells with intact microtubules is inhibited by geldanamycin. The ability of both geldanamycin treatment and FKBP52 domain I expression to disrupt GR linkage to dynein could explain why the two manipulations slow the rate of GFP-GR movement to the same extent and why there is no further slowing with the two manipulations together ( Fig. 2A).
In this work, we show that dynein is present in native GR⅐hsp90 heterocomplexes immunoadsorbed from L cell cytosol and that dynein-containing GR⅐hsp90 heterocomplexes are formed in reticulocyte lysate. These complexes contain FKBP52 (Fig. 5); however, complexes containing other immunophilins, such as FKBP51, are also formed (39). Assembly of GR⅐hsp90 heterocomplexes in the presence of the TPR domain fragment blocks the binding of all of the immunophilins (26). Thus, at this time, we can only say that the presence of dynein in the GR⅐hsp90 heterocomplex requires a TPR domain immunophilin, with FKBP52 providing at least part, and perhaps all, of that linker function. FIG. 5. GR hsp90 heterocomplexes assembled in reticulocyte lysate contain cytoplasmic dynein. A, a TPR domain protein, such as FKBP52, is required for dynein presence in the GR⅐hsp90 heterocomplex. Stripped GR immune pellets were incubated for 30 min at 30°C with reticulocyte lysate that was preincubated for 30 min at 30°C under conditions indicated below. The pellets were then washed, and the GR and associated proteins were assayed by Western blotting as described under "Methods." Lane 1, nonimmune pellet incubated with reticulocyte lysate; The pellets were then washed, and the GR and associated proteins were assayed by Western blotting.
The common feature of immunophilins is the presence of their PPIase domains, and the PPIase domain of FKBP52 possesses PPIase activity even when it is expressed as a domain fragment (35). However, the PPIase inhibitor FK506 does not affect FKBP52 binding to cytoplasmic dynein ( Fig. 4 and Ref. 28), or the rate of GFP-GR translocation, or inhibition of translocation by the FKBP52 PPIase domain. Thus, the PPIase activity of FKBP52 does not seem to be involved in the interaction with cytoplasmic dynein or retrograde movement of GFP-GR. The PPIase domain of FKBP52 functions as a dynein interaction domain independent of binding by FK506. In this respect, it differs from the interaction of FKBP12 with calcineurin, where binding requires the FK506-bound immunophilin (40).
It is clear that the PPIase domains of immunophilins are involved in different protein-protein interactions and effects. For example, the low molecular weight immunophilin FKBP12 does not affect GFP-GR movement, whereas the PPIase domain of FKBP52 does (Fig. 3B). FKBP12 also does not compete for FKBP52 binding to cytoplasmic dynein, whereas the PPIase domain of FKBP52 does (28). Although FK506-bound FKBP12 binds to and inhibits the calcineurin phosphatase (40), neither FK506-bound FKBP52 nor its expressed PPIase domain fragment inhibits calcineurin activity in vitro (41). Our impression is that the PPIase domains may constitute a family of proteinprotein interaction domains, with no overlap of binding targets between FKBP12 and the high molecular weight, hsp90-binding imunophilins.
Because we see the same rate of GFP-GR nuclear translocation in the presence and absence of microtubules (Fig. 2), it may be that nuclear import and not receptor movement through the cytoplasm is rate-limiting for nuclear accumulation under both conditions. When the microtubules are intact, it seems that receptor movement occurs predominantly on a microtubulebased movement system that requires dynamic interaction of the receptor with hsp90 and the immunophilin. When cytoskeleton is disrupted, the GR may move through the cytoplasm by diffusion. Although movement of the receptor through the cytoplasm is not the rate-limiting step in cytoplasmic-nuclear translocation under physiological conditions in cells with intact cytoskeleton, it becomes rate-limiting when geldanamycin or the PPIase domain of FKBP52 are present to uncouple the receptor from the movement system. Thus, in the normal cell, rapid movement of protein solutes by diffusion may be limited by the organized cytoskeleton, and attachment to a movement system permits more rapid delivery to their sites of action. Such a movement system may not be essential for targeted protein movement in non-neuronal cells. But in axons, movement solely by diffusion is not possible, and a movement system, such as that described here, would be essential.
In this work, we have provided the first in vivo evidence that the immunophilin PPIase domain links the glucocorticoid receptor to a retrograde movement system. We also show for the first time that GR⅐hsp90⅐immunophilin heterocomplexes are bound to the retrograde motor protein cytoplasmic dynein via the PPIase domain of the immunophilin component.