Cedex, France and the §Howard Hughes Medical Institute, Gene Expression Laboratory, Salk Institute,. San Diego ... and Baulieu, E. E. (1988) J. Biol. Chem.
THEJOURNAL OF BIOLOGICAL CHEMISTRY The American Society for Biochemistry and Molecular Biology, Inc.
Vol. 266, No. 9, Issue of March 25, pp. 5834-5841,1991 Printed in U.S.A.
( ~ 1991 1 by
Heat Shock Protein 90 as a Critical Factorin Maintaining Glucocorticosteroid Receptorin a Nonfunctional State* (Received for publication, November 5, 1990)
Franqoise Cadepondl, Ghislaine Schweizer-Groyerl, Ingrid Segard-Maurell, Nicole Jibard$, Stanley M. Hollenbergj, Vincent Giguerej, RonaldM. Evanss, and Etienne-Emile Baulieulll From the Slnstitut Nationalde la Sante et dela Recherche Medicale U33, Uniuersite Paris-Sud, Lab Hormones, 94275 Bichtre Cedex, France and the §Howard Hughes Medical Institute, Gene Expression Laboratory, Salk Institute, San Diego, California 92138
Previous work demonstrated that the ligand bindingSteroid hormone receptors are ligand-dependent transcripdomain (LBD)was required to determine the formation tion factors (Ringold, 1985; Yamamoto, 1985; Evans, 1988). of the cytosolic, untransformed, inactive, 8 S, heter- In the absence of their cognate ligand, steroid receptors are ooligomeric form of thehuman glucocorticosteroid recovered in the cytosol fraction of target cell homogenates receptor (hGR) which includes the 90-kDa heat shock in inactive untransformed heterooligomeric complexes sediprotein (hsp90) (Pratt,W. B., Jolly, D. J., Pratt, D. V., menting a t 8-9 S and devoid of DNA binding activity(review Hollenberg, S . M., Giguere, V., Cadepond, F. M., in Baulieu,1987). Upon hormone binding, the 8-9 S comSchweizer-Groyer, G., Catelli, M. G., Evans, R. M., plexes tend to dissociate. The steroid receptor molecule sediand Baulieu, E. E. (1988) J. Biol. Chem. 263, 267ments at -4 S and can bind with a high affinity to DNA 273). Truncationsof hGR deleting all oralmost all of the LBD give GR derivatives in the non-hsp90-inter- (Holbrook et al., 1983; Sherman et al., 1983; Sherman and acting 4 S form able to stimulate transcription in a Stevens, 1984; Vedeckis, 1983). The untransformed 8 S-glumolecular hormone-independent manner. To identify the LBD cocorticosteroid receptor (8 S-GR) has an apparent weight of -300,000 and contains one steroid-binding subunit subregion(s) involved in 8 S formation, we analyzed 1985; Okret et al., 1985) and a nonsteroid, the sedimentation behavior of hGR mutants with var- (Gehring and Arndt, ious LBD internal deletions and/or truncations tran- non-DNA-binding component (M, 90,000), identified as a siently expressed in cells that contain hsp90 but very heat shock protein (hsp9O) (Joab et al., 1984; Catelli et al., 1985; Sanchez et al., 1985; Schuh et al., 1985). hsp90 is low levels of endogenous GR, and we correlated the probably ina dimeric formin theheterooligomeric GR’ (Denis results with their transcriptional activity. LBD has been divided into threesubregions: two of et al., 1987, 1988b; Radanyi et al., 1989). The precise role of them, LBDl (between amino acids 551 and 626) and the association of the GR-binding entity with hsp90 is not LBD2(betweenaminoacids 627 and 696), include yet understoodclearly. However,analytical studiesperformed amino acid sequences highly conserved in the steroid after labeling of intact cells (Holbrook et al., 1983; Mendel et receptor superfamily, and LBD3 consists of the car- al., 1986; Lefebvre et al., 1988), 8 S stabilization by antagoboxyl-terminal part of the molecule (amino acids697- nistic ligand (Groyer et al., 1987; Raaka et al., 1989; Segnitz 777). Each of these subregions can be deleted withoutand Gehring, 1990), whole cell cross-linking (Rexin et al., impeding 8 S heterooligomer formation, and the cor- 1988), and in vitro translation in reticulocyte lysate (Denis responding receptors do not have transcriptional activand Gustafsson, 1989a; Dalman et al., 1989) strongly support ity in the absence as well as in the presence of hormone. the hypothesis that this association takes place in vivo and When linked to hGR mutants truncated after amino may be of physiological significance (see “Discussion”). acids 532 or 550, each subregiondoes separately proThe cloning of human(Hollenberg et al., 1985),mouse mote 8 S heterooligomeric complex formation and re(Danielsen et al., 1986), and rat (Miesfield et al., 1986) glupress the intrinsic constitutive transcriptional activity of the truncated receptors. These 8 S complexes con- cocorticosteroid receptors hasallowed the localization of funcof theprotein.Usingintact tain hsp9O. In a control experiment, the linkage of tionallyimportantdomains 1,017 amino acids of &galactosidase to the carboxyl- (hGRa) or mutated GR sequences expressed in COS-7 cells, Gigubre et al. (1986) and Hollenberg et al. (1987) defined a terminal of 1-532 hGR gave a hybrid receptor still a constitutively transcriptionally active which did not centralcysteine-rich DNA bindingdomain(DBD)and bind hsp9O. These results provide evidence that there ligand binding domain (LBD) localized in the carboxyl-teris a strong correlation between the association with minal third of the molecule. With the same technique, itwas hsp9O and theloss of GR functional properties andthat shown that the hGR LBD was required for the formation of high molecular weight 8 S heterooligomeric complexes (Pratt hsp9O may playa critical role in maintaining the receptor in a nonfunctional state. et al., 1988). Consistent with this result, the dataof Denis et al. (1988a) demonstrated that the tryptic 27,000-Da steroidbinding fragment interacts with hsp9O. We then sought to
-
* This investigation was supported by the Institut National de la Santk et de la RechercheMedicale and the Centre National de la Recherche Scientifique and by the Howard HughesMedical Institute. The costs of publication of this article were defrayed in part by the payment of page charges. Thisarticlemusttherefore be hereby marked “aduertisemenf” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 7 To whom correspondence should be addressed.
’ The abbreviations used are:GR, glucocorticosteroidreceptor; hGR,human glucocorticosteroid receptor;ER, estrogenreceptor; hER, human estrogen receptor; TA, triamcinolone acetonide; SDS, sodium dodecyl sulfate; DME medium, Dulbecco’s modified Eagle’s medium; LBD, ligand binding domain; DBD, DNA binding domain; hsp, heat shock protein; NL, nuclear localization; MOPS, 4-morpholinepropanesulfonic acid.
5834
Glucocorticosteroid Receptor hsp90 and
5835
delineate more precisely the regions involved in LBD-hspSO centrifuged for 5 min a t 600 x g, resuspended, and homogenized in interaction and to investigate possible the participation of the 1.5 volume (v/v) of GTED buffer (15% glycerol, 20 mM Tris-C1, 1 mM EDTA, 1 mM dithioerythritol, p H 7.4, at 4 "C) containing proDNA binding region and of the amino-terminal part of the teaseinhibitors (0.2 mM phenylmethylsulfonyl fluoride,5pg/ml receptor. leupeptin,and 5pg/ml antipain). Cytosolwas recovered after a Here we show that neither the amino-terminal region nor 105,000 X g centrifugation (1 h at 4 "C) and contained 3-15 mg/ml the DBDof hGR promotes the formation of heterooligomeric protein. Steroid Binding Assay-Cytosols were incubated with 40 nM [3H] complexes, which can be detected in the cytosol of hsp90containing cells transfected with mutatedcDNAs. No limited TA for 2 h a t 4 "C. Unbound hormone was extracted with 0.05% dextran-coated 0.5% charcoal. Nonspecific binding was evaluated in deletion precludes the formation of 8 S heterooligomer, and a parallel sample containing a 1,000-fold excess of nonradioactive three regions in the LBD are capable of ensuring the forma- ligand. tion of hsp90-containing complexes and the repression of Glycerol Gradients-150-pl aliquots were layered on top of 10-35% transcriptional activity when fused to constitutive mutants. glycerol gradients (4.2 ml) prepared in TED-Mobuffer (20 mM TrisThese results, extending the correlation between binding of C1, 1.5 mM EDTA, 1 mM dithioerythritol, and 50mM sodium molybhsp9O and lack of transcriptional activity, are consistent withdate, pH 7.4, a t 4 "C). 100 pg of glucose oxidase (7.9 S) and 10 pg of horseradish peroxidase (3.6 S) were added to each sample as internal the proposal that the heat shock protein90,000 is a negative standards. Gradients were centrifuged for 15 h at 49,000 rpm in a transcriptional regulator. The hormone-dependent release of Beckman SW 60 rotor. For radioactive counting, 2-drop fractions hsp90 or lack of hsp90 binding to truncated receptor is there-were collected intoscintillation vials andcounted.ForWestern blotting, 3-drop fractions were collected in cryotubes, and the tubes fore critical for transcriptional activity. were frozen a t -80 "C. Before radioactive counting or Western blot analysis, a 15-111aliquot was removed from each fraction for enzymatic assays. Forconvenience, we named "8 S-receptor"the complexes sedimentinginthe 7-9 S region, and"4s-receptor"the receptor Materials actually sedimenting between 3 and 4 S . The use of various deleted [1,2,4-:'H]Triamcinolone acetonide (TA;specific activity 1.07 TBq/ receptors explains thedifferences in sedimentation coefficients. mmol) was purchased from Amersham Corp. Glucose oxidase from Irnmunoblotting-Gel electrophoresis was performed in 7.5% polyA. niger and horseradishperoxidase were from Boehringer Mannheim acrylamide-SDSslab gels overnight a t room temperaturewith a and Sigma, respectively. Radioactive protein molecular weight mark- constant current of 7 mA/gel. Proteins were transferred from gels to ers were obtained from Amersham Corp. The Vectastain ABC im- nitrocellulose filters using an X-Blot apparatus (Ceralabo). The blot munoperoxidasesystem was purchasedfrom Vector Laboratories, was incubated overnight with 0.5% rabbit antiserum GR135 raised Inc., Burlingame, CA. against a peptide correspondingto amino acids144-172 of the hGRa (Hollenberget al., 1987) or PBL 135 antiserum raised against a Methods peptide corresponding to amino acids 1-30 of the hGRn (generously Plasmid Construction and Preparation-Except for the hGR muprovidedby Dr. Wylie Vale).GRbands were revealed withthe tants b550-600/1696* and A550-627/1696*, the constructionof plas- Vectastin ABC immunoperoxidase systemaccording to the manufacmidpRShGRncontainingthefull-length receptor cDNAand of turer's recommendations. plasmids containing deletion or truncation receptor mutants hasbeen Measurement of Transcriptional Actiuity-CV1 cells were homogdescribed previously (Hollenberg et al., 1987; 1989; Oro et al., 1988). enizedandchloramphenicolacetyltransferaseand 0-galactosidase A550-600/1696* was constructed by ligation of the 1.99-kilobase activities were measured using the methods of Gorman et al. (1982) KpnI-Saul fragment of A550-600, with the 0.83-kilobase Saul-XhoI and Herbomel et al. (1984),respectively. fragment of I696* and with the 3.5-kilobase KpnI-XhoI fragment of Detection of hsp90 in Heterooligorneric Complexes-Cytosol prehGRa or b491-515. Plasmid A550-627/1696* was constructed simipared in GTED (containing30 mM Na,MoO,) was diluted to 1.5 mg/ larly except that the 1.91-kilobase KpnI-Saul fragment of A550-627 ml with MEN medium + 10% G buffer (25 mM MOPS, 1 mM EDTA, was used instead of the corresponding fragment of A550-600. Com- 0.02% NaN:', pH 7.5, a t 0 "C + 10% (v/v) glycerol) with 10 pg/ml petent bacteria were transformed by ligation mixtures. Miniprepara- leupeptin, 10 pg/ml antipain, and 0.2 mM phenylmethylsulfonyl flutions of bacterial colonies were carried out and DNAs checked by oride and incubated for 2 h at 0 "C with 100 gl/ml of a solution of restriction endonuclease digestion. The selected clones were verified 8D3antibody(Perdew, 1988) or of control antibody (TEPC 183). by DNA sequencing. Plasmid DNA was prepared by a modification Solutions were layered on parallel 10-30% sucrose density gradients of the procedure of Birnboim and Doly (1979) followed by a two-step prepared in MEN + 10% G buffer and run for 120 min a t 70,000 rpm cesium chloride-ethidiumbromidegradientpurification procedure in a Beckman VTi-80 rotor. For mutantsI532* and I532* fused to pdescribed by Garger et al. (1983). The identity of each plasmid was galactosidase sequence (1532*/@-Gal),TED + 10% (v/v) G buffer was checked by restriction endonuclease mapping. used instead of MEN + 10% G buffer in the dilution of cytosol and Cell Culture and Transfection-COS-7 or CV1 cells were main- centrifugation steps to avoid the degradation of GR which occurred tained in monolayer culture in 10% CO, at 37 "C in DME medium with all 4 S forms tested in MENbuffer. After centrifugation, 3-drop containing 10% fetal calf serum, 50 pg/ml gentamicin, and 2 mM fractions were collected, and the tubewall opposite the rotation axis glutamine. COS-7cells were transfected byamodification of the was rinsed carefully. Each fraction was analyzed by SDS-polyacrylmethod of Sompayrac and Danna(1981). Briefly, one Petri dish was amide gel electrophoresis followed by Western blotting with PBL 135 expanded in three or four dishes on the daybefore transfection. The antibody. For hGRa liganded with ["HITA, aliquots were removed DNA to be transfected (10 pg/dish; typically 20 10-cm dishes/trans- for scintillation counting. fected plasmid) was mixed with 5 ml of transfection buffer (6.85 g/ liter Tris-CI, pH 7.0, 0.8 g/liter Tris base, and13.4 g/liter dry DME) RESULTS containing 50 mg of DEAE-dextran, preincubated30 min, then complemented with transfection buffer to a final volume corresponding The Amino-terminal Half and the DNA-binding Domain of to 5 ml/dish. Each culture dish was washed twice with 3 ml of DME medium, then incubated a t 37 "C with 5 ml of transfection mixture. the Glucocorticosteroid Receptor Are Not Directly Inuolued i n When the cellsbegan to detach (usually 5-8 h), the transfection Heterooligomer Formation mixture was removed, the cultures were washed with 3 ml of serumfree DME medium before being supplemented with DME medium When COS-7 cells were transfected either with the aminocontaining 10% fetal calf serum and allowed to grow for about 40 h terminal truncated mutantAO-417 (see Table11) or with GR before harvesting. mutants partially deleted in the DNA binding domain (Table CV1 cells were transfected by the calcium phosphate procedure I; A420-451 with deleted first zinc finger and A450-487 with (Wigler et al., 1979) with a mixture containing 10pg of hGR mutant, 5 p g of reporter plasmid (pMTVCAT) (GiguGre ef al., 1986), and 5 p g deleted secondzinc finger), 8 S steroid-labeledcomplexes were recovered as after transfection of hGRm (Fig. 1, a, b, c, of pCHllO expressing constitutively /%galactosidase as an internal standard for transfection efficiency. and d ) . Thus, neither the amino-terminal half of hGR nor the Cytosol Preparation-Cells were scraped in phosphate buffer saline,zinc fingers are necessary for the formation of the heterooliEXPERIMENTALPROCEDURES
5836
hsp90 and Glucocorticosteroid Receptor TABLEI Ability of hGR internal deletion mutants to form 8 S heterooligomers and to activate transcription Striped area, DNA binding domain; filled area, steroid binding domain; wt, wild-type glucocorticosteroid receptor (hGRa). 8 S heterooligomer formation, + +, indicates that most GR is recovered in low salt cytosol in the 8 S heterooligomeric form (see “Experimental Procedures” andlegends to Figs. 1 and 2 for details). Steroid binding activities (+, positive; -, negative) are from Hollenberg et al. (1987), from Pratt et al. (1988), or determined in this work. Transcriptional activities expressedas relative luciferase activitiesare from Hollenberg et al. (1989).Relative chloramphenicol acetyltransferase activities(*) are from Hollenberg e t al. (1987) and from Hollenberg and Evans
(1988). Presenceof 85 form 0
100
l
l
wt, A420-451
I
A450-487
I
A491-516
r
A488-533
I
A488-697
I
6515-551
L
A532-697
I
A550-600
L
200
300 l
400
500
l
600
0
Transcriptional activity
700
l
DNA
Steroid binding
I
- steroid
+steroid
+
