Cytoplasmic 8 S Glucocorticoid Receptor Binds to Actin Filaments through the 90-kDa Heat Shock Protein Moiety*

The glucocorticoid receptor exists in the cytoplasm of hormone-untreated cells as a complex with the 90-kDa heat shock protein (HSPSO). Glucocorticoids in- duce dissociation of the glucocorticoid binding protein from HSPSO and translocation of the receptor to the nucleus. HSPSO binds to actin filaments, and calmodulin or tropomyosin inhibits the binding. We present here evidence that the HSPSO-containing glucocorticoid receptor complexes (8 S receptor) bind to filamen- tous actin in vitro while the HSPSO-free form of the receptor does not. The binding was detectable for both the crude cytosolic fractions and the partially purified 8 S glucocorticoid receptor. Purified HSPSO or tropomyosin completely abolished the binding. Calmodulin also inhibited the binding in a Ca2+-dependent manner. From these results, we conclude that the glucocorticoid receptor complex is able to bind actin filaments via the HSPSO moiety. The binding may provide an anchoring mechanism for the glucocorticoid receptor in the cytoplasm. The receptor (GCR)’

intrinsic glucocorticoid binding protein and a dimer of the associated protein (Joab et al., 1984;Housley et al., 1985;Schuh et al., 1985;Mendel et al., 1986;Denis et al., 1987) that has been identified as a heat shock protein (HSP), HSP9O (Catelli et al., 1985;Sanchez et al., 1985Sanchez et al., , 1987. During the activation process after the hormone binding, HSP9O dissociates from the receptor, and the GCR complex is converted from the 8 S non-DNA binding form to the 4 S DNA binding species (for review, see Pratt et al., 1989;Denis and Gustafsson, 1989). Even after the binding with the ligands, the 8 S ' The abbreviations used are: GCR(s), glucocorticoid receptor(s); HSP, heat shock protein; TAA, triamcinolone acetonide; Hepes, 4-(2-hydroxyethy1)-1-piperazineethanesulfonic acid; EGTA, [ethylenebis(oxyethylenenitrilo)]tetraacetic acid SDS, sodium dodecyl sulfate; PAGE, polyacrylamide gel electrophoresis. GCR is stabilized in a relatively low ionic strength buffer containing transition metal oxyanions such as molybdate (Dahmer et al., 1984). Current evidence suggests that the 8 S GCR is really present in the cytoplasm of living cells (Howard and Distelhorst, 1988), implying that the association of the 4 S GCR with HSP9O is physiologically significant. In fact, the association has been suggested to be necessary for high affinity binding of the GCR with glucocorticoids (Bresnick et al., 1989).
HSP90, a 90-kDa heat shock protein, is induced in various organisms by heat shock or other environmental stresses, but is also constitutively expressed in unstressed cells (for review, see Lindquist and Craig, 1988;Schlesinger, 1990). HSP9O exists as a dimeric form under physiological conditions and is distributed throughout the cytoplasm Lindquist and Craig, 1988). We have noted that ruffling membranes of cultured cells were brightly stained with antibodies against HSP9O and also with phalloidin . In addition, a heat shock-resistant Chinese hamster ovary variant expressing HSP9O to a relatively high degree showed the high level of cell motility . These observations raised the possibility that HSP9O interacts with actin filaments. This turned out to be the case, and we have found that HSP9O cross-links actin filaments Nishida et al., 1986). The binding of HSP9O to actin filaments is inhibited by tropomyosin or by Ca2+calmodulin . Taking this finding together with the fact that 8 S GCR consists of a dimeric HSP90, we have hypothesized that 8 S GCR interacts with actin filaments through its HSP9O moiety. In this study, we have examined this hypothesis and clearly showed that 8 S GCR but not 4 S GCR binds actin filaments and that the binding is specifically mediated by HSP9O. scribed (Nishida et al., 1979). Rabbit skeletal muscle G-actin was Proteins-Calmodulin was purified from porcine brains as deprepared by the method of Spudich and Watt (1971) and further purified by gel filtration on a Superdex-200 HiLoad column equilibrated with a buffer solution containing 0.1 mM CaCl,, 0.2 mM ATP, 0.1 mM dithiothreitol, 0.01% N&, and 2 mM Hepes, pH 7.8. HSP9O

8779
Binding of Steroid Receptor-HSPSO Complex to Actin was purified from mouse lymphoma L5178Y cells as previously described . Tropomyosin from rabbit skeletal muscle was a generous gift from Dr. E. Nishida (University of Tokyo). Rabbit antibodies raised against purified mouse HSP9O was prepared as described (Yonezawa et al., 1988). Ligand Binding Assay-A hydroxylapatite adsorption assay was used to quantitate the receptors (Nemoto et al., 1990). An aliquot of samples was incubated with 20 nM [3H]TAA in the presence or absence of a 500-fold molar excess of radioinert TAA. After the of hydroxylapatite suspension (50% v/v in HEDG + Mo buffer) for incubation for 3 h on ice, the sample was mixed with an equal volume 20 min with occasional mixing. Hydroxylapatite was precipitated by a brief centrifugation, the pelleted material was washed with 0.5 ml of HEDG + Mo buffer three times, and bound ['H]TAA.GCR complexes were extracted with 0.2 ml of 0.4 M potassium phosphate, pH 7.4, for 15 min on ice with occasional vortexing for the counting of radioactivity.
Preparation of Actin-free Glucocorticoid Receptors-Actin-free and ligand-free GCRs were prepared according to the method described by Nemoto et al. (1990) with modifications. All procedures were carried out at 0-4 "C. Briefly, rat livers were perfused with ice-cold phosphate-buffered saline, pH 7.4, washed with HEDG + Mo buffer, ground, homogenized in 2 volumes of HEDG + Mo buffer containing protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 2 pg/ml leupeptin, 1 pg/ml pepstatin A, 2 pg/ml antipain, and 10 pg/ml aprotinin), and clarified by centrifugation (75,000 X g, 40 rnin). The lipid layer was discarded, and the supernatant fluid was mixed with a half-volume of phosphocellulose PI1 equilibrated with HEDG + Mo buffer and gently agitated for 1 h. The unbound materials were collected by a centrifugation (3,000 X g, 10 min) and applied on a DEAE-Sepharose Fast Flow column (16 mm X 20 cm) equilibrated with HEDG + Mo buffer containing 80 mM NaCl, and the column was washed extensively with the same buffer until the absorbance at DEAE-Sepharose column were eluted with 0.2 M NaCl in HEDG + 280 nm reached a stable base line value. Proteins retained on the Mo buffer and subsequently separated by gel filtration chromatography on a Superdex-200 HiLoad column (16 mm X 60 cm) which had been equilibrated with HEDG + Mo buffer without NaC1. Proteins were monitored by the absorbance at 280 nm, and the glucocorticoid binding was assayed as described above. HSP9O is one of the major proteins in the preparation of partially purified 8 S GCR. The glucocorticoid binding activity was eluted as a single peak from the gel filtration column. The stannous chloride assay (Strelow and Bothora, 1967) showed that the GCR fractions contained molybdate. An addition of 500-fold nonradioactive TAA to the peak fraction completely abolished the binding of ['HITAA indicating that the observed [3H]TAA binding peak contains specific GCR. The fractions contained -2 mg/ml proteins and were free of actin as shown by SDS-PAGE (data not shown).
The partially purified GCR labeled with [3H]TAA was analyzed by centrifugation in 5-20% sucrose density gradients in low salt buffer containing molybdate. The partially purified GCR had a sedimentation coefficient of 8.8 S. An incubation of the partially purified rat GCR with polyclonal anti-mouse HSP9O IgG significantly increased the sedimentation coefficient of the GCR to 9.6 S (data not shown). These results are consistent with those previously reported (Nemoto et al., 1990).
Preparation of Cytosol-Hepa-1 cells were scraped and homogenized in HEDG + Mo buffer containing protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 2 pg/ml leupeptin, 1 pg/ml pepstatin A, 2 pg/ml antipain, and 10 pg/ml aprotinin). The homogenate was then centrifuged at 22,000 X g for 20 min at 2 "C, the supernatant was further centrifuged at 100,000 X g for 60 min at 2 "C, and the resulting supernatant was used as Hepa-1 cytosol. For the preparation of cytosol containing 4 S receptors, sodium molybdate was omitted during the processes, and 0.4 M KC1 was added to the buffer. Labeling the Glucocorticoid Receptors with fH]TAA-Samples containing the receptors were incubated with 20 nM ['HITAA for 4 h on ice. The unbound ligands were adsorbed to a '/5 volume of a dextrancharcoal suspension (100 mg/ml charcoal and 10 mg/ml dextran in HEDG + Mo buffer) at 4 "C for 20 min and were removed by centrifuging the mixture at 4000 X g for 10 min and by filtrating the supernatant through a 0.22-pm centrifuge filter. When indicated, the ["HITAA-labeled samples were incubated with anti-HSP9O IgG on ice for 3 h.
Sucrose Density Gradient Centrifugation Analysis-200-p1 aliquots of ['HITAA-labeled samples were layered onto 5.2-ml linear sucrose density gradients ( 5 2 0 % ) prepared in HEDG + Mo buffer. Gradients were centrifuged at 45,000 rpm for 16 h at 2 "C in a Beckman SW 50.1 rotor. 18O-gl fractions were collected, and the radioactivity in each fraction was determined.
Assay of Binding of Steroid Hormone Receptors to Actin Filaments-Hepa-1 cytosols or partially purified GCRs were labeled with ['HITAA as described above. Various concentrations of actin were polymerized by incubating in a buffer containing 20 mM Hepes, 100 mM KCl, 2.5 mM MgC12, 0.03 mM ATP, 1 mM 2-mercaptoethanol, 0.02 mM dithiothreitol, and 0.02 mM imidazole, pH 7.5, in the presence of 2 mM Ca2+ or 2 mM EGTA for 60 min at room temperature, and a further 20 min on ice, and resulting actin filaments were incubated with samples containing labeled receptors on ice for 90 min. When indicated, various concentrations of purified calmodulin, tropomyosin, or HSP9O were added in the mixture before the addition of the receptors. The mixture was then centrifuged at 100,000 X g for 90 min at 2 "C so that polymerized actin and associated proteins were precipitated, and the radioactivities in the pellet fractions solved in 1% SDS were counted.

RESULTS
Co-precipitation of Labeled GCRs in Crude Cell Extracts with Endogenous or Exogenous Actin Filaments-GCRs in Hepa-1 cytosols were labeled with [3H]TAA, mixed with or without exogenous actin filaments, and centrifuged so as to pellet down the actin filaments together with bound proteins. The binding was determined at various concentrations of cytosol fractions in the presence or absence of exogenously added actin. All the experimental results were summarized in Fig. 1, indicating that exogenously added actin filaments significantly increased the precipitated radioactivity in all preparations. We have confirmed that the [3H]TAA-labeled Hepa-1 cytosol used for these binding experiments contained exclusively the 8 S form of GCR (data not shown). These results suggest that [3H]TAA-bound 8 S GCR was co-precipitated with actin filaments. The cytosolic fraction contains high concentrations of actin. This explains the fact that significant amounts of [3H]TAA-bound receptors were precipitated in the absence of the exogenous actin. We have observed that calmodulin reduced the amount of [3H]TAA-bound 8 S GCR co-precipitated with actin filaments in a Ca2+-dependent manner (see below).
To compare the binding activity to actin filaments of HSPSO-containing 8 S GCR and that of HSPSO-free 4 S receptor, the cytosolic fraction containing HSPSO-free 4 S receptor was prepared by extracting the cells in the absence of molybdate, followed by treatment in 0.4 M KC1. We have confirmed by sucrose density gradient centrifugation analysis that the cytosol fraction prepared under these conditions contained 4 S GCR (data not shown). The cytosol fraction containing 8 S receptor and that containing 4 S receptor were separately incubated with polymerized actin in the same buffer condition, and co-precipitation of receptors was determined (Fig. 2). The results clearly indicated that the HSPSOcontaining 8 S GCR was co-precipitated with both endogenous and exogenous actin filaments, whereas the HSPSO-free 4 S GCR was not co-precipitated at all. Purified GCR Binds to Polymerized Actin Filaments-To elucidate the binding of 8 S GCR with actin filaments more precisely, we partially purified the 8 S GCR from rat livers to form an actin-free preparation. The purification was performed using phosphocellulose and DEAE-Sepharose followed by gel filtration according to the methods described by Nemoto et al. (1990) with modifications. Partially purified 8 S GCR was tested for its actin binding activity. Aliquots of partially purified 8 S GCR were incubated with 20 nM [3H] TAA, and the unbound ligand was adsorbed to dextrancharcoal. About one-fourth of the unadsorbed radioactivity was considered to be free from GCR because this radioactivity did not bind to hydroxylapatite. The results clearly showed that purified 8 S GCR bound to filamentous actin in an actin dose-dependent manner (Fig. 3,O). The binding of 8 S GCR was saturable as the concentration of actin increased. The amount of 8 S GCR bound to actin filaments was correlated with that of the bulk of HSPSO co-precipitated with actin filaments as revealed by SDS-PAGE of the precipitated proteins (Fig. 3, 0). 40-50% of the total ['HITAA was coprecipitated with an excess amount of polymerized actin. These results suggest that 53-67% of the total GCR bound actin filaments. It is not clear, however, why the remaining GCR did not bind actin filaments.
Effects of Purified HSPSO, Tropomyosin, and Ca2'-Calmodulin on the Binding of Partially Purified GCR with Actin Filaments-We tested whether or not the binding to actin filaments of partially purified 8 S GCR is similar to that of HSPSO. First, we examined the effect of purified HSPSO on the binding of 8 S GCR to filamentous actin. Purified HSPSO inhibited the binding in a dose-dependent manner (Fig. 4a). Secondly we tested the effect of purified tropomyosin on the Hepa-1 cytosols containing 8 S GCR or 4 S GCR were prepared as described under "Experimental Procedures." The cytosol fractions were labeled with 20 nM ['HITAA, and co-precipitated GCR in the presence or absence of exogenous actin (0.4 mg/ml) was determined. The total radioactivity of ["HITAA including the receptor-free TAA was 1241 cpm, and the radioactivity of the total GCR determined by the hydroxylapatite receptor assay was 820 cpm in each aliquot. The mixtures were incubated on ice for 90 min and then centrifuged 100,000 X g for 90 min at 2 "C so that the polymerized actin filaments and the associated proteins were precipitated. The amounts of co-precipitated HSPSO were determined by SDS-PAGE and densitometry (0) and the radioactivities in the precipitated fractions were counted (0). The radioactivity of the total GCR-bound ["HITAA in each aliquot was estimated to be 1490 cpm determined by the hydroxylapatite assay. TAA-labeled GCR were incubated with polymerized actin (0.2 mg/ ml) with various concentrations of purified HSP9O (a) or purified tropomyosin ( b ) , and bound GCR were determined as described in Fig. 3. The radioactivity of the total GCR-bound [3H]TAA determined by the hydroxylapatite receptor assay was 1160 cpm (a) or 1044 cpm ( b ) . HSP9O or tropomyosin did not affect the degree ofpolymerization of actin (data not shown).
binding for tropomyosin is known to inhibit the binding of purified HSPSO to filamentous actin in vitro . The results clearly showed that tropomyosin inhibits the binding (Fig. 4b). The inhibition by 50% was observed with about 0.25 mg/ml tropomyosin, which is comparable to the inhibitory concentration of tropomyosin on the binding of purified HSPSO to actin filaments in vitro . Finally, we have found that calmodulin inhibits the binding in the presence of Ca2+ (Fig. 5), which is consistent with the results obtained with purified HSPSO and actin. Neither Ca2+ without calmodulin nor calmodulin without Ca2+ affected the binding at all. These results again strongly suggest that 8 S GCR binds to filamentous actin via its HSPSO moiety.

DISCUSSION
The results presented in this paper argue strongly that 8 S GCR binds to actin filaments but that 4 S GCR does not. [3H]TAA-labeled GCR were incubated with or without exogenous actin filaments (0.2 mg/ml) in the presence or absence of 1.5 mg/ml calmodulin, with 2 mM Ca2+ or 2 mM EGTA, and bound GCR were determined as described in Fig.  3. The radioactivity of the total GCR-bound [3H]TAA in each aliquot determined by the hydroxylapatite assay was 1359 cpm. Neither Ca2+ nor calmodulin affected the degree of polymerization of actin (data not shown).
presence of Ca2+. Taken together with our previous findings on the in vitro binding of HSP9O to actin Nishida et al., 1986), we conclude that the binding is specifically and exclusively mediated by the HSP9O moiety.
A body of evidence indicates that 8 S GCR is present in the cytoplasm as the non-DNA binding form and that binding of glucocorticoids induces dissociation of the GCR. HSP9O complex and confers on the receptor the ability to bind DNA (for review, see Pratt et al., 1989;Denis and Gustafsson, 1989). During this activation process of GCR, the receptor is translocated into the nucleus (Wikstrom et al., 1987). The GCR polypeptide has been shown to contain a nuclear location signal sequence and, therefore, to be intrinsically located in the nucleus (Picard and Yamamoto, 1987). Nevertheless, 8 S GCR is present in the cytoplasm. One explanation for this fact is that the nuclear location signal of GCR becomes cryptic when it associates with HSP9O. Another plausible explanation was raised in this study; that is, 8 S GCR anchors on actin filaments and, therefore, is restricted to translocate into the nucleus although the nuclear location signal is exposed and functional.
Receptors for androgen (Joab et al., 1984), estrogen (Redeuilh et al., 1987), progesterone (Kost et al., 1989), mineralocorticoid (Rafestin-Oblin et al., 1989), and dioxin (Perdew, 1988;Denis et al., 1988) are similar to GCR in their complex formation with HSP9O. All of these receptor complexes might be reasonably expected to interact with actin filaments through their HSP9O moieties. The cytosolic and nuclear distribution of these receptor complexes are somewhat controversial, however. Recently, it was shown that the progesterone receptor. HSP9O complex is predominantly distributed in the nucleus (Renoir et al., 1990) suggesting that the nuclear location signal is exposed in this complex. As we have previously described, the interaction between HSP9O and actin filaments is relatively weak (KD = M) . In addition, the interaction was regulated by Ca2+ and calmodulin . All these properties of the interaction have been shown in this study for that between 8 S GCR and actin filaments. Some steroid hormone receptors or a portion of them may be free from the anchoring system postulated herein.
Other members belonging to the steroid hormone receptor superfamily or the nuclear receptor superfamily (Evans, 1988), including the receptors for thyroid hormone (Pascual et al., 1982) and retinoids (Petkovich et al., 1987;Giguere et al., 1987), exist as molecular forms free from HSP9O and they reside exclusively in the nucleus even in the absence of the corresponding ligands. The receptors of this type may be translocated to the nucleus soon after they are synthesized possibly because they do not form complexes with HSP9O. In fact, newly translated thyroid receptors do not form complexes with HSP9O in an in vitro system, whereas newly translated GCR does (Dalman et al., 1990). The intracellular distribution of the steroid hormone receptor superfamily proteins appears to be determined by the association with HSP9O and by the ability to interact with the cytoskeletal system through the HSP9O moieties.
Several physiologically important proteins other than the cytosolic steroid hormone receptors and the dioxin receptors have also been shown to be associated with HSP9O. They include tyrosine kinase oncogene products such as pp60""" (Brugge et al., 1981;Opperman et al., 1981;Schuh et al., 1985), and eIF2a kinase (Rose et al., 1987;Matts and Hurst, 1989), a serine/threonine-specific protein kinase. These facts suggest that HSP9O serves as a common carrier protein for biologically key functional molecules. It would be of interest to examine whether the complexes of the above kinases with HSP9O interact with actin filaments.