THEJOURNAL OF BIOLOGICAL CHEMISTRY 0 1992 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 267, No. 7, Issue of March 5,pp. 4939-4948,1992 Printed in U.S.A.

Expression of Androgen Receptor in Insect Cells PURIFICATIONOF FUNCTIONS*

THE RECEPTOR AND RENATURATION OF ITS STEROID- AND DNA-BINDING

(Received for publication, July 25, 1991)

Yan-Bo Xie,Ya-Ping Sui, Li-Xin Shan, Jorma J. Palvimo, David M. Phillips, and Olli A. JanneS From the Population Council and the Rockefeller University, New York, New York 10021

A full-length rat androgen receptor cDNA was used to produce a recombinant baculovirus (AcrAR) by homologous recombination. Spodoptera frugiperda (Sf9) cells infected with this virus expressed a 110-kDa polypeptide that amounted up to about one-third of total cell protein. Studieswith AR antibodies confirmed that thisprotein was indeed rAR. Onlya minor portion of the recombinant AR was soluble in buffers without ionic detergents, but its complete solubilization was achieved in 6 M guanidine HCl(GdnHC1). Electron microscopy of cell pellets revealed that AR was localized to electron-dense cytoplasmic aggregates. The soluble cytosolic receptor was biologically active, in that it bound [3H]mibolerone with high affinity and specificity and interacted with an androgenresponsive element. The functions of the GdnHC1-solubilized AR were partially restored by a 20-50-fold dilution. The solubilized receptor was purified to an apparent homogeneity in a single step by gel filtration on a Sephacryl S-400 column in the presence of 6 M GdnHC1. The homogeneous AR protein could be renatured to bind [3H]mibolerone, interact specifically with a DNA element, and be recognized by receptor antibodies. Receptor-DNAinteraction was stabilized by an antibody directed againstthe N-terminal part and abolished by an antibody against the hinge region of the receptor. Zn2+ions were essential for the purified receptor to refold into a specific DNA-binding form during the renaturation, with the optimal ZnC12 concentration being 50-100 PM depending on the buffer conditions. Cd2+ions were also capable of restoring the receptor’s DNA-binding activity anddid so at concentrations 10-fold lower than those of the Zn2+ions.

Androgen receptor (AR)’ belongs to the family of ligand* This work wassupported by National Institutes of Health Grants DK37692 and HD13541, the Medical Research Council of the Academy of Finland, and the Finnish Life and Pension Insurance Companies. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “aduertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $ To whom correspondence should be addressed Dept. of Physiology, University of Helsinki, SF-00170 Helsinki, Finland.: Tel.: 3580-1918544; Fax: 358-0-1918681. The abbreviations and trivial names used are: AR, androgen receptor; ARE, androgen responsive element; GRE, glucocorticoid responsive element; AcNPV, A. californica nuclear polyhedrosis virus; GdnHC1, guanidine hydrochloride; Sf9, S. frugiperda cells; nt, nucleotide(s); NaDodSO.,, sodium dodecyl sulfate; mibolerone, 7a,17a-dimethyl-19-nortestosterone; dexamethasone, 9a-fluoro-l1(3,17a,21trihydroxy-16a-methyl-l,4-pregnadiene-3,2O-dione; PBS, phosphatebuffered saline; PMSF, phenylmethylsulfonyl fluoride; HEPES, 4-(2hydroxyethy1)-1-piperazineethanesulfonicacid.

induced nuclear transcription factors, which includes receptors for steroid and thyroid hormones, vitamin D, and retinoids (1-4). The primary physiological functions of the AR are regulation of the development and differentiation of male reproductive organs and maintenance of masculine phenotype and behavior (5-7). During the pastfew years, AR cDNA has been cloned by several groups, which has permitted deduction of the receptor’s primary amino acid sequence (8-12) and initial studies onits transactivation properties (13,14).Availability of purified AR protein would bebeneficial for a variety of structural andfunctional experiments; however, the receptor protein is present in androgen-target tissues at relatively low concentrations (15, 16), which, along with its labile nature, has made it very difficult to purify sufficient quantities of the protein. In view of this, expression of AR protein in a heterologous system appears a logical step to generate large amounts of this transcriptionfactor. The baculovirus (Autographa californica)expression system developed by Smith et al. (17,18) has been used for successful expression of a wide variety of heterologous eukaryotic genes (19, 20). The major advantages of this system are the high yield of the expressed protein and the immunological and functional similarity of these proteins to those present under physiological conditions in mammalian cells. Overexpression of full-length estrogen and glucocorticoid receptors in insect cells employing the baculovirus system has been recently reported (21-24); however, only a truncated AR protein has been overexpressed in a prokaryotic system (25-27). In this work, we have introduced rat androgen receptor (rAR) cDNA intothe genome of wild-type baculovirus by homologous recombination to express the AR protein in Spodoptera frugiperda (Sf9)cells. The transfected insectcells express recombinant AR protein in two forms: a minor portion is soluble and can be recovered in the low salt cytosol, whereas over 90% of the protein is initially insoluble in the form of electrondense cytoplasmic aggregates. By several criteria, the soluble form of the AR is functionally active. The insoluble form could be solubilized, purified to homogeneity in a single step, and partially renatured to a functionally active form. MATERIALSANDMETHODS

Cell Culture-S. frugiperda (Sf9) insect cells were purchased from Invitrogen Corp. (San Diego, CA). The cells were cultured as monolayers in T25 or T75 flasks a t 27‘C using Grace’s insect medium with supplements (Invitrogen Corp.) and 10% (vol/vol) fetal bovine serum (20). Gentamycin (50 pg/ml) and amphotericin B (2.5 pg/ml) were added to the culture medium. Construction of the Recombinant Transfer Vector and Recombinant Baculouirus-The plasmid containing a full-length rat ARcDNA insert between the EcoRI and PstI sites in the polylinker of pGEM42 was obtained from Dr. S. Liao, University of Chicago (8).Since there is an internal EcoRI site in the rAR cDNA sequence, a twostep cloning procedure was usedto insert the rAR cDNAdownstream

4939

4940

Baculovirus Expression

ofReceptor Androgen

of a polyhedrin promoter in the baculovirus transfer plasmid pVL1393 was22.5 nM and the sample volume 0.4ml. Steroid specificity of (Invitrogen Corp.) to form the recombinant transfer vector pVXrAR receptor binding was determined by including unlabeled mibolerone, (Fig. 1). To generate a recombinant baculovirus, 1 pgof wild-type testosterone, estradiol, progesterone, and dexamethasone at 225 nM baculovirus (AcNPV) DNA and 2 pg of pVXrAR DNA were cotrans- concentrations (10-fold molar excess) in the incubations containing fected into Sf9 cells by the calcium phosphate precipitation method 22.5 nM [3H]mibolerone. To determine AR concentration in Sf9 cell pellets after a 3-4-day (28). Four days after transfection, the virus-containing medium from Sf9 cells was serially diluted 1:100,1:1,000, and 1:5,000 with the infection, the cells were harvested bylow speed centrifugation at culture medium, and the dilutions were used to infect fresh mono- 2,500 X g for 5 min, washed once with 0.5 vol of NTE (0.15 M NaCl, layers of insect cells. Twoh afterthe infection, the cells were dispersed 10 mM Tris-HC1 (pH 7.5), and 1 mM EDTA). Cell pellet was then and 0.1-ml aliquots were seeded onto 96-well plastic culture plates, suspended in 0.5 vol of low salt buffer (10 mM KzHP04 (pH7.6) and which were incubated 3-4 days at 27 "C. After removing and storing 1 mM EDTA) and exposed to four cycles of freezing and thawing in the medium, the cells in each well were lysedin 0.5 M NaOH and the dry ice-ethanol and in water at 30 "C. Following centrifugation at neutralized lysate was bound to nitrocellulose membrane for screen- 2,500 X g for 5 min, the insoluble pellet was washed with 0.5 vol of ing by hybridization to 32P-labeledrAR cDNA. Standard DNA-DNA buffer TNNP (50 mM Tris-HC1 (pH 7.5), 500 mM NaC1, 1%Nonidethybridization conditions were employed (28). Samples with strong P-40, and 0.5 mM PMSF) twice and then two times with 0.5 vol of hybridization signals were subjected to a second round of dilution NTE buffer. The specimens were solubilized in 6 M GdnHCl (see and screening to isolate pure cultures of recombinant baculovirus. below) and assayed for steroid binding under conditions similar to The recombinant baculovirus vector used in all experiments was those for the cytosolic AR, with the final dilution of GdnHCl being 25-50-fold during the 20-h incubation. termed AcrAR. Production of AR Antibodies-In addition to two antisera against Solubilization of the Recombinant Androgen Receptor-Insoluble synthetic peptides corresponding to residues 479-495 (peptide 1)and pellets washed as above were suspended in a small volume of NTE 770-786 (peptide 2) of the human AR that we have described previ- and solubilized by adding 6.5 M GdnHCl buffer containing 1 mM ously (29), antiserawere developed against threeadditional synthetic EDTA, 40 mM Tris-HC1 (pH 7.5), and 1 mM dithiothreitol to give peptides. AR peptide 3 has the sequence Pro-Pro-Ser-Lys-Thr-Tyr- the final concentration of 6 M GdnHCl. After a thorough mixing, the Arg-Gly-Ala-Phe-Gln-Asn-Leu-Phe-Gln-Ser-Val-Arg-Glu (residues samples were kept at room temperature for 1 h with occasional 14-32); AR peptide 4 is Glu-Ala-Leu-Glu-His-Leu-Ser-Pro-Gly-Glushaking. Protein concentration in the specimens was determined by Gln-Leu-ArgGly-Asp-Cys-Met-Tyr-Ala (residues 250-268); and AR the Bio-Rad protein assay according to manufacturer's instructions, using bovine serum albumin as the standard. peptide 5 is Leu-Gly-Ala-Arg-Lys-Leu-Lys-Lys-Leu-Gly-Asn-LeuGel Retardation Studies of DNA-Receptor Interaction-To deterLys-Leu-Gln-Glu-Glu-Gly-Glu (residues 626-644). A Cys residue was included in the N terminus of each peptide to facilitate the coupling mine the ability of the recombinant AR to interact specifically with t o keyhole limpet hemocyanin. The peptides were synthesized by DNA, soluble proteins were extracted from Sf9 cells infected for 2 Multiple Peptide Systems, Inc. (San Diego, CA). Conjugation of the days either with the recombinant transfer vector or the wild-type peptides to keyhole limpet hemocyanin, immunization of New Zea- baculovirus by homogenization in TEDG buffer (lo' cells/ml) conland White rabbits, and evaluation of antiserum titers by enzyme- taining 0.4 M KC1,0.5 mMPMSF, and 50 nM 5a-dihydrotestosterone. linked immunosorbent assay techniques with immobilized free pep- After centrifugation at 100,000 X g for 45 min, the supernatant was tides were carried out as previously described (29). After immuniza- precipitated by ammonium sulfate (0.3 g/ml) for 30 min at 4 "C and tion for several months, the titersfor antisera against AR peptides 3, precipitated protein collected by centrifugation at 50,000 X g for 20 4, and 5 were 5 X lo6,7 X lo5, and 9 X lo6,respectively. The antisera min. The protein pellets were dissolved in one-tenth of the original cytosol volume in buffer D containing 20 mM HEPES (pH 7.9), 100 were further characterized by their ability to immunoprecipitate [%I methionine-labeled AR protein synthesized in uitro using capped AR mM KCl, 2 mM dithiothreitol, 0.2 mM EDTA, 20% glycerol, 0.05% mRNA transcribed from AR cDNA (29). Antisera with highest titers Nonidet P-40,0.5 mM PMSF, and 50 nM 5a-dihydrotestosterone and against the respective peptide immunogens were used in subsequent dialyzed against the same buffer for 2 h. The extracts were frozen in small aliquots in liquid nitrogen and stored at -70 "C. Receptor studies. IgG was isolated from serum by chromatography on protein preparations solubilized in 6 M GdnHCl, or purified to homogeneity A-Sepharose (Pharmacia LKB Biotechnology Inc.). by gel filtration in the presence of GdnHCl, were routinely renatured Polyacrylamide Gel Electrophoresis and Immunoblotting-Protein gel electrophoresis was carried out using 10% polyacrylamide slab by a rapid 10-fold dilution with buffer R containing 50 mM Tris-HC1 gels in the presence of 0.1% NaDodSOl (30). The gels were stained (pH 7.5), 0.8 M NaC1, 0.1 mM EDTA, 5 mM dithiothreitol, 0.12 mM routinely with Coomassie Blue, with the silver staining being used to ZnClz, 0.05% Nonidet P-40, and 20% glycerol. Diluted preparations verify the purity of receptor preparations. Following electrophoresis, were incubated at 4 "C for 1 h with occasional shaking and then the proteins were transferred (4 h, 0.6 A) to nitrocellulose paper in a dialyzed against buffer D containing 0.12 mM ZnCL for 3 h at 4 "C. buffer containing 20 mM Tris-HC1 (pH 8.3), 150 mM glycine, and In the experiments involving studies of divalent cation conditions, 20% (v/v) methanol. For immunoblot analyses, nitrocellulose sheets EDTA and ZnC12were omitted from buffer R. Oligonucleotides corresponding to both strands of the glucocortiwere first blocked with 10% (w/v) casein in 0.15 M phosphate-buffered saline (PBS), 0.05% (v/v) Tween 20 (PBS-Tween buffer, pH 7.2) for coid/androgen responsive element (GRE/ARE) of the rat tyrosine 30 min at 25 "C. Antiserum against AR peptide 3 (residues 14-32) aminotransferase gene promoter (32) were synthesized using a Pharwas diluted 1:800 in PBS-Tween buffer containing 5% casein and macia gene assembler (Pharmacia LKB BiotechnologyInc.). Compleincubated with the nitrocellulose sheet for 16-20 h at 4 "C. Normal mentary DNA strands of the responsive element (upper strand 5'rabbit serum or anti-AR peptide 3 antiserum preadsorbed with the GACCCTAGAGGATCTGTACAGGATGTTCTAGATCCAATTCGcorresponding free peptide (1 pg/ml) were used as negative controls. 3') were annealed at a concentration of 100 pg/ml in 10 mM TrisAfter five washings in PBS-Tween buffer, nitrocellulose sheets were HC1 (pH 7.5),150 mM NaCl by heating at 65 'C for 15 min and reblocked with 10% casein and incubated in a 1:1,000 dilution of goat cooling to 22 "C over a period of 2 h. Double-stranded oligonucleotide anti-rabbit F(ab')z-horseradish peroxidase conjugate (Sigma) in PBS- was labeled using T4 polynucleotide kinase with [y-32P]ATP.ComTween buffer containing 5% casein for 4 h at 22 "C. The membrane peting oligonucleotides prepared the same way had the following shown): C3(1), 5"GATCATAGwas washed again 5 times in PBS-Tween buffer, color developed with sequences (upper strandsare 3,3-diaminobenzidine in PBS-Tween buffer for 1-5 min, and finally TACGTGATGTTCTCAAGATC-3' (29); mutated ARE, muC3(1), (29); glucocortirinsed with distilled water. In some experiments, antiserum against 5'-GATCATAGTACGTGATTTTCTCAAGATC-3' AR peptide 5 (residues 626-644)wasemployed, essentially as de- coid responsive element (GRE),5"TCGACTGTACAGGATGTTCTAGCTACT-3'; and vitamin D responsive element, 5"TTGGTGACTscribed above for the AR peptide 3 antibody. Measurement of Androgen Receptor Concentration-ARconcentra- CACCGGGTGAACGGGGGCATT-3'. The binding reaction mixture (20 pl) contained typically 5 pg tion was determined by using [3H]miboleroneas the labeled ligand, essentially as previously described (31). Cytosol was prepared in (soluble AR) or 15-30 ng (reconstituted homogeneous AR) protein TEDGMo buffer (50 mM Tris-HC1 (pH 7.5), 0.1 mM EDTA, 5 mM with 0.2-2 pg of poly(d1-dC). (dI-dC) in 20 mM HEPES (pH 7.9), 50 1 mM dithiothreitol, 1 mM MgClz,0.4 mM EDTA, 10% dithiothreitol, 10 mM sodium molybdate, and 20% glycerol) and the mMKC1, hydroxylapatite adsorption method was used to separate bound and glycerol, and 0.05% Nonidet P-40. After a preincubation on ice for free radioactivity (31). Nonspecific binding of [3H]mibolerone was 10 min, 0.5 ng of a labeled oligonucleotide was added and the incuestimated from parallel sets of samples containing a 200-fold molar bation was continued at 22 "C for 20 min. The protein-bound DNA excess of the nonradioactive steroid. In routine single-point assays, complexes were separated from the free probe immediately after the final [3H]miboleroneconcentration in a 20-h incubation at 4 "C incubation on a 4% polyacrylamide gel (29:l) run in 0.25 X TBE (1

Baculovirus Expression ofReceptor Androgen

4941

X TBE: 90mM Tris base, 90 mM boric acid, and2 mM EDTA) in the DNAs originating from these samples did not contain30any presence of 0.1% Nonidet P-40. Gels were dried and autoradiographed kDa polyhedrin protein (see below),and appeared to represent under Kodak X-Omat AR films. Purification of the Recombinant Androgen Receptor by Gel Filtra- pure strains of recombinant AR-producing viruses. One of tion-The AR protein present inthe 6 M GdnHC1-solubilized Sf9 cell these strains (designatedAcrAR) was selected for subsequent pellet was purified by gel filtration on a Sephacryl S-400 HR (Phar- experiments. macia) column equilibrated and eluted with 6 M GdnHCl. Several Characterization of the AR Protein Expressed inInsect sizes of columns have been used for different sample volumes. In a Cells-In the initial characterization of AR produced in intypical experiment,a 1.5 X 69-cm column with a bed volume of 115 fected cells, total insect cell proteins were analyzed by polyml was used for a 2-2.5-m1 sample. The flow rate of the column was 5-7 ml/h, and 2.5-ml fractions were collected at 4 “C. Fractions were acrylamide gel electrophoresis under denaturing conditions. pooled and dialyzed against three changes of NTE buffer to remove This was accomplished by solubilizing washed Sf9 cells by GdnHCl, and aliquots were analyzed for purity by polyacrylamide gel boiling in the samplebuffer prior to electrophoresis. Compoelectrophoresis. Protein concentration was measured by the Bio-Rad sition of proteins in non-infected Sf9 cells and incells infected dye binding method. with wild-type viral (AcNPV) DNA and with the recombinant Ultrastructure of Sf9 Cells and Cell Pellets-Insect cells or cell pellets aftera 3-day infection withthe wild-type baculovirus DNAor virus is shown in Fig. 2. As expected, cells infected with the AcrAR DNA were washed as described above and prepared forelec- wild-type baculovirus contained a largeamount of the 30-kDa polyhedrin protein (Fig. 2 A , lane 2 ) . The cells infected with tron microscopy as previously described (33). AcrAR DNApossessedlarge quantities of a new 110-kDa RESULTS protein that,by densitometric scanning, amounted up to oneConstruction of the Recombinant Baculouirus-A full-length third of the total protein and, as expected from the way by rat AR cDNAwas insertedintothe baculovirus transfer which the recombinant transfer vector was constructed, no plasmid pVL1393 usingthe EcoRI andPstIsites of the polyhedrin protein was expressed in these cells (Fig. 2 A , lane multiple cloningsequence (Fig. 1).The resulting recombinant 3 ) . The above data provided strong circumstantialevidence for transfer vector (pVXrAR) contains32 nucleotides (nt) of the cells 5”untranslated region, the entire protein-codingsequence of the expression of a full-length AR proteinininsect 2706 n t for 902 amino acid residues, and 87 nt of the 3’- infected with the recombinant virusDNA. The total protein untranslated region of the AR cDNA. The orientationof the samples were subsequently analyzed by immunoblotting with 2388-nt EcoRIIEcoRI fragment (Fig. 1) in pVXrAR was de- androgen receptor antibodies,using several different antisera against AR peptides (12, 29) (see also “Materials and Methtermined by restriction enzyme mapping and partial DNA ods”). The data with all antiserawere very similar, and only sequencing. Sincetheinitiation codon forthepolyhedrin protein translation is mutated in pVL1393, translation of the results with antisera against AR peptide 3 (residues 14-32, RNA encoded by the recombinant transfervector initiates at “N-terminal antibody”) andAR peptide 5 (residues 626-644, “hinge region antibody”) are shown in Fig. 2B. The newly the authenticAUG of the AR coding sequence. A recombinant baculovirus expressing the AR protein in formed 110-kDa protein inSf9 cells infected withAcrAR was Sf9 cells was produced in vivo by homologous recombination recognized with the antisera, along with some smaller bands using co-transfectionof the insectcells with a wild-type viral that stained less intensively (Fig. 2B, lanes 2 and 3 ) . Immu(AcNPV) DNA andpVXrAR DNA. Recombinant viruses noreactivity was completely abolished bypreadsorption of the expressing theAR cDNA sequencewere initially identifiedby antiserum with thefree peptide (Fig. 2B, lane 4 ) or by using of antibodies againstAR peptides DNA hybridization with an AR cDNA clone (see “Materials normal rabbit serum instead and Methods”). The first round screening of the supernatant (lane 5 ) . The antiserum did not react with proteins in nonculture media at 1:5,000 dilution showed twostrongly positive infected Sf9 cells (Fig. 2B, lane 1) or those in cells infected hybridization signals (designated A4 and F9). By using F9 with the wild-type viral DNA (not shown). When human virus stock asstartingmaterialforthesecondround of hepatoma cell (HepG2) cytosol was analyzed by immunoblotinfection and screening, five strongly positive signals were ting with the antiserum against AR peptide 3, only a single 110-kDa band was detected (Fig. 2B, lane 6), in agreement identified. Lysatespreparedfrom cells infectedwithviral our previous results using other antisera against AR peptides (29). The amountof the 110-kDa band in the cytosol fraction of Sf9 cells infected with the recombinant virus was much lower than that in preparations of total cell proteins (Fig. 2 A,lanes 5 and 6), indicatingthatthereceptor was mainly in an insoluble form in the insect cells. Various means were used to extract the 110-kDa protein from thecells, including sonication, freezing and thawing, and treatments with buffers containing multiple detergents (50 mM Tris-HC1 (pH 8.0), 150 mM NaC1,0.02% sodiumazide, 0.1% (w/v) NaDodS04, 0.5 mM PMSF, 1% (v/v) Nonidet P-40, and 0.5% (w/v) sodium deoxycholate) or high salt concentrations (50 mM Tris-HC1 (pH 7.5), 500 mM NaC1, 1%NonidetP-40,and 0.5 mM FIG. 1. Construction of the recombinant transfer vector PMSF). In every case, the majority of the 110-kDa protein pVXrAR. The 2,388-nt-long EcoRI-EcoRI fragmentand the 437-nt- remained in thepellet fraction after low speed centrifugation long EcoRI-PstI fragment were cleaved from the rat ARcDNA in (2,500 X g for 5 min). Only 1% NaDodSOl and 6 M GdnHCl pCEM-4Z (prAR) and purified by agarose gel electrophoresis. The were able tosolubilize completely the pellet-associated recepEcoRI-PstI fragment was inserted into the EcoRI and PstI sites in tor. By washing the pelletsbefore solubilization, someloosely the polylinker of pVL1393 to form pVXrARc,after which the EcoRIEcoRI fragment was cloned into the EcoRI site of pVXrARc to yield associated proteins were removed from the sample (cf. Fig. the transfer vector pVXrAR. Abbreviations: R, EcoRI; P, PstI; PH, 2 A , lanes 3 and 7). After the high salt wash, AR comprised the polyhedrin promoter. pellet. AR solubilized in up to 50% of the total protein in the

4942

Baculovirus Expression

of Androgen Receptor

A. M 1

106

-

50

-

3 -1

3

4

5

6 7

M

-I

80

20

2

-

FIG. 3. U l t r a s t r u c t u r e of insect cells after infection with wild-type and recombinant baculovirus vectors. Plates a and b, Sf9 cells were collected after 3 days of infection with the wild-type baculovirus.Although virions ( V ) , fibrous material (F),and polyhedra ( P ) are seen in the nuclei of infected cells, no gross morphological changes are observed in the cytoplasm. Plates c and d, insect cells were infected with the recombinant baculovirus for 3 days prior to ultrastructural analyses. Aggregates of granular material ( C A ) are seen to occupy most of the cytoplasm of these cells. Mitochondria are often embedded in the granular material. Magnifications: X 6,000 (plates a and c) andX 21,000 (plates b and d ) .

FIG.2. Polyacrylamide gel electrophoretic and immunoblot analyses of proteins expressed in Sf9 cells infected with the recombinant baculovirus AcrAR. Toanalyze samples representing totalcellular proteins, the insectcells were harvested 3 days after infection, washed with NTE buffer, and solubilized directly in electrophoresis sample buffer. Specimens corresponding to the soluble cytosol and remaining pellet fractions were obtained by lysis of the cells with four cycles of freezing and thawing, followed by centrifugation a t 2,500 X g for 2 min. The samples were analyzed on 10% polyacrylamide, 0.1% NaDodS04 gels. A, gels stained with Coomassie Blue after electrophoresis. Lane I, total cell proteins (8 pg) from noninfected Sf9 cells; lane 2, wild-type baculovirus DNA-infected Sf9 cells, total cell proteins (10 pg); lane 3, AcrAR DNA-infected Sf9 cells, total cell proteins (10 pg); lane 4, proteins (0.4 pg) in the pellet of AcrAR-infected Sf9 cells after lysis by freezing and thawing; lane 5,soluble proteins (0.4 pg) from AcrAR-infected cellslysed by freezing and thawing; lane 6, proteins (0.2 pg) solubilized in high salt lysis buffer from recombinant baculovirus-infected cells lysed by freezing and thawing; lane 7, cell pellet proteins (0.4 pg) of AcrAR-infected cells washed in the presence of high salt. Molecular weight markers are shown in lanes M. B , immunoblotting with androgen receptor antibodies. After electrophoresis, the samples were transferred onto nitrocellulose membrane, and immunoreactive proteins detected with diluted (1:800) antisera against AR peptides 3 and 5 (see “Materials FIG.4. Electron micrographs of insoluble pellets of insect and Methods”).Lane I, noninfected Sf9 cells; lane 2, AcrAR-infected cells after infection with wild-type and recombinant baculocells, AR peptide 3 antiserum; lane 3, AcrAR-infected cells, AR v i r u s vectors. The insoluble pellets were prepared using Sf9 cells peptide 5 antiserum; lane 4, AcrAR-infected cells, AR peptide 3 after 3 days of infection as described under “Materials andMethods.” antiserum preadsorbed with the corresponding free peptide (1 pg/ Plates a and b, following infection with AcrAR DNA, the insoluble ml); lane 5, AcrAR-infected Sf9 cells, immunoblotting with normal material is composed of granular aggregates similar to those in the rabbit serum insteadof antisera against AR peptides; lane 6, human hepatoma cell (HepG2) cytosol, AR peptide 3 antiserum. All insect cytoplasmic compartment of intact cells (Fig. 3) infected with this may represent regionswhere cell specimens originated from washed pellets solubilized directly in virus. Empty spaces in the material mitochondria were extracted by washes with high salt and detergents. the electrophoresis sample buffer. The samples contained thefollowPlates c and d, micrographs of cells infected with the wild-type ing amounts of protein per lane: Sf9 cells, 32 pg; AcrAR-infected baculovirusshowonly the presence of material corresponding to cells, 4 pg; and HepG2 cells, 51 pg. polyhedra.Magnifications: X 6,000 (plates a and c) and X 21,000 (plates b and d ) .

6 M GdnHCl did not, however, bind [‘H]mibolerone without a prior dilution and/or removal of the chaotropic agent (see below). Ultrastructure of Sf9 Cells Infected with the Recombinant Baculouirw-Ultrastructure of insect cells infected with the wild-type baculovirus has been described by several investigators (34, 35). In agreement withprevious reports, we found intranuclear virions, inclusion bodies (polyhedra) andfibrous aggregates (Fig. 3a). The cytoplasm of infected cells was similar to thatof the uninfected cells except for the presence of occasional fibrous aggregates(Fig. 3b). The most prominent ultrastructural change in the AcrAR-infected cells, compared

with the wild-type baculovirus-infected cells, was the appearance of aggregated particles in thecytoplasm. These particles, which were 30-40 nmindiameterand formed irregular masses, occupied most of the cytoplasmic volume of the infected cells (Fig. 3, c and d ) . Mitochondria were frequently embedded in the dense aggregates. As mentioned above, extensively washed insoluble pellets of AcrAR-infected Sf9 cells contained AR proteinthat amounted to 3 0 4 0 % of total protein (Fig. 2 A , lane 7). This pellet material from AcrAR-infected cells was composed almost entirely of aggregates of the 30-40-nm particles (Fig. 4,

Baculovirus Expression a and b). Frequently,irregularaggregateswere observed around circular areas where presumably nuclearcontents had been extracted by the combination of high saltand detergents used for washing the insoluble pellet material (Fig. 4, a and b). No similar structures were seen i n the pellets isolated from Sf9 cells after wild-type virus infection,and this latter insoluble material was composed almost exclusivelyof polyhedra (Fig. 4, c and d ) . Collectively, the ultrastructural data suggest that the insolubleAR protein expressed in very high amounts in infected cells is located in aggregated 30-40-nm electrondense particles present i n the cytoplasm. Properties of the Soluble Receptor Protein-The majority of AR in insect cells was in the insoluble form; however, some receptor was always present i n the soluble cytosol prepared from AcrAR-infectedSf9 cells. The time course of appearance of [‘H]mibolerone binding (Kd = 5 nM) i n t h e cytosol of Sf9 cells is illustrated i n Fig. 5A; the concentration of AR, as calculated on the basis of the steroid binding capacity, was about 150,000 receptors/cell on day 3 after infection, which is at least 5 times higher than that i n rat prostate (36). Immunoblotting studies with the antiserum against AR peptide 3 indicated that the soluble cytosolic formof AR migrated ina

a

2

3

i

2

X

4943 I

I

1

2 3 4 5 COMPETING STEROID

6

FIG. 6. Steroidspecificity of therecombinantandrogen receptor in insect cell cytosol. The 105,000 X g cytosol of Sf9 cells infected with AcrAR DNA was prepared as described under “Materials and Methods” and in the legend to Fig. 5. Ten-fold molar excess of nonlabeled steroids (1, no competitor; 2, mibolerone; 3, testosterone; 4 , estradiol; 5, progesterone; and 6, dexamethasone) were incubated with samples containing 22.5 nM [3H]mibolerone.The values are mean of duplicate assays and represent specific [3H]mibolerone binding.

4

TIME AFTER INFECTION (DAYS)

1

of Receptor Androgen

3

106 80

50

33 28 FIG. 5. Time course of appearance of androgen bindingactivity in cytosol of Sf9 cells infected with the AR transfer vector AcrAR. A , Sf9 cells were infected with the transfer vector expressing rAR and harvested 2, 3, and 4 days after the infection. The cells were homogenized in TEGDMo buffer, and 105,000 X g soluble cytosolwas prepared as described under“Materials and Methods.” Androgen receptor concentration was measured by incubating the samples with 22.5 nM [3H]miboleronewith and without a 200-fold molar excess of the corresponding nonlabeled steroid. After the incubation, bound and free steroids were separated using the hydroxylapatite technique (33). The values shown represent specific [‘’Hlmiboleronebinding. Control samples (0 days) were from noninfected Sf9 cell cytosol. B, immunoblot analysis of the AR protein in the soluble cytosol fraction of Sf9 cells infected with AcrAR DNA and harvested 0 (lane 1, 14 pg of protein), 2 (lane 2, 5 pg of protein), and 3 (lane 3, 2 pg protein) days after infection. Antiserum against AR peptide 3 was usedthree experiments, as described in detail under “Materials and Methods” and in the legend to Fig. 2.

FIG. 7. Binding of the recombinant androgen receptor toa hormone responsive element. A , indicated amounts of soluble cytosolic AR from AcrAR-infected cells or of a similar extract prepared from wild-type virus-infected cells were incubated with a ”Plabeled 41-nt oligomer corresponding to the rattyrosine aminotransferase gene GRE/ARE, and protein-DNA complexes were separated from free DNA by nondenaturing gel electrophoresis as described under “Materials and Methods.” Lanes 1 and 2,4 and 8 pg of protein from wild-type virus infected cells, respectively; lanes 3 and 4 , 4 and 8 pg of protein from AcrAR-infected cells, respectively; lanes 5 and 6, the same sample as in lane 4 but in the presence of a 80-fold or 160-fold molar excess of nonlabeled DNA probe, respectively. The origin of the faster migrating band is not known. B, effect of anti-AR antisera on the DNA-binding activity of soluble cytosolic AR as tested by gel retardation assay. The following serum samples (0.4 pl in each case) were included in the preincubation (2 h at 4 “C) before addition of labeled GRE/ARE: lane 1, normal rabbit serum; lane 2, AR peptide 3 antiserum (N-terminal region); lane 3, AR peptide 5 antiserum (hinge region). C, binding of reconstituted AR to GRE/ ARE. Pellets from insect cells infected with the recombinant virus were solubilized in 6 M GdnHCl and diluted in buffer R (see “Materials and Methods”) or buffer R containing only 0.1 M NaCl (lanes 1 and 2, respectively). D, binding of homogeneous AR to GRE/ARE and influence of AR antibodies on this interaction. AR purified by gel filtration on Sephacryl S-400 was diluted and reconstituted as described under “Materials and Methods.” Prior to addition of labeled GRE/ARE oligomer, the samples were incubated with preimmune serum (lane I ) , AR peptide 3 antibody (lane 2 ) , and AR peptide 5 antibody (lane 3 ) , as described for experiments in panel B. fashion identical withthat of the insoluble form, i.e. as a 110k D a band (Fig. 5 B ) . Identical results were obtained whenthe antiserum against AR peptide 5 (hinge-region antibody) was employed in immunoblots. The steroid-binding specificityof the soluble AR expressed i n Sf9 cells was very similarto that of the receptor in androgen target tissuesof rodents (31,36), i n that the rank order of competition was androgen > estradiol

Baculovirus Expressionof Androgen Receptor

4944

TABLE I Reconstitution of the steroid-binding activity of the recombinant AR in GdnHCl by dilution under different conditions The final [3H]mibolerone concentration in each case was 22.5 nM, and the samples (1pg of protein) originating from solubilized pellets were incubated with the steroid for 16 h at 4 "C. Bound and free steroids were separated by the hydroxylapatitemethod (31). The values given represent mean specific binding in triplicate assays; nonspecific binding in each case was less than 5% of total binding. ExDerirnental conditions

r3Hlrnibolerone

Bound cpm

a. Ten-fold dilution of the sample (5 pl) with TEDG + steroid on ice for 15 min; additional 5-fold dilution with TEDG b. As in condition a, but theinitial dilution without the steroid c. Ten-fold dilution of the sample (5 p l ) with NTE without the steroid on ice for 15 min; additional 5-fold dilution with TEDG d. Fifty-fold dilution of the sample (5 pl) in TEDG steroid e. As in condition a, but buffer TEDG without 0.1 mM EDTA

24,720 26,970 7,720

+

19,350 43,250

0.20 0.15 0.10 0.05 -

0.25

A.

" " 0

10 15 20 Total [3H]Mibolerone, nM

5

25

B.

"."" ,

Bound, nM

FIG. 8. Binding of [3H]miboleroneto the solubilized recom7 binant androgen receptor produced in insect cells. A , satbation isotherms for steroid binding. Five-pl aliquots of the 6 M GdnHCl solubilized, AcrAR-infected cell pellets containing 1 pg (W), 0.67 pg (01,and 0.17 pg (0)of protein were first diluted 50-fold with TEGD buffer and the binding assays performed as described under "Materials and Methods." Nonspecific binding of ["H]mibolerone was determined by using a 200-fold molar excess of nonlabeled steroid in parallel samples. Only total binding of [3H]mibolerone is shown; nonspecific binding amounted for less than 5-10% for each point. B, specific binding data plotted according to the method of Scatchard. The plotis shown for the sample containing 1 pg of solubilized protein; similar data can be derived from experiments with the other two protein concentrations shown in panelA .

> progesterone > glucocorticoid (Fig. 6). An additional criterion used to evaluate the functional properties of the soluble AR form was its ability to interact with a specific DNA element. For this purpose, we employed a GRE of the tyrosine aminotransferase gene (32),which has been previously shown to confer androgen responsiveness to a reporter gene (37).A soluble extract from AcrAR-infected cells, but not from the cells infected with AcNPV DNA,

"."" . nnn-

0

I

.

. 10

.

.

20

.

.

.

30

II 111

.

.

40

FRACTION NUMBER

FIG. 9. Purification of the recombinant rat androgen receptor by gel filtration in the presence of 6 M guanidine hydrochloride. Three T75 flasks of confluent Sf9 cells were infected with the recombinant baculovirus vector AcrAR for 3 days, after which the insect cells (about 3 X lo7 cells) were harvested, lysed by freezing and thawing in high salt lysis buffer, and washed with NTE buffer, Pellets recovered by low speed centrifugation were solubilized in 6 M GdnHCl buffer (6 M guanidine hydrochloride, 1 mM EDTA, 40 mM Tris-HC1 (pH 7.5), and 1 mM dithiothreitol) a t room temperature for 1 h. An aliquot of the sample (2.4 ml, 928 pg of protein) was loaded onto a 1.5 X 69-cm Sephacryl S-400 HR column (bed volume: 115 ml), which was equilibrated and eluted with 6 M GdnHCl buffer. Three-ml fractions were collected and combined into three pools ( I , ZZ, and ZZZ), which were dialyzed against NTE buffer. After the dialysis, aliquots of the pools were analyzed by polyacrylamide gel electrophoresis and silver staining. Protein concentration in each fraction was measured by the Bio-Rad dye-binding assay reagents. Inset, analysis by electrophoresisona10% polyacrylamide, 0.1% NaDodSOl gel and silver staining of aliquots from pools I, 11, and 111.

interacted with a41-mer oligonucleotide containing the above GRE/ARE, as revealed by band-shift experiments (Fig. 7 A ) . Formation of a labeled DNA-AR complex was completely prevented by a 80-fold molar excess of the corresponding nonlabeled oligomer. Although 5a-dihydrotestosterone was included in buffers used for the preparation of cytosol extracts (see "Materials and Methods") and was present in the incubation mixture, samples prepared and incubated without the steroid yielded identical results (data not shown). The specificity of this interaction was further verified by using the antiserumagainst AR peptide 3, which recognizes the Nterminal region of the AR protein; the inclusion of this antiserum, but not preimmune serum, resulted in a further retardation and apparentstabilization of the GRE/ARE-AR complexes (Fig. 7 B ) . Interestingly, preincubation of the recombinant cytosolic AR with the hinge-region antibody (antiAR peptide 5) prior to the DNA-AR binding reaction almost completely abolished the formation of labeled GRE/ARE-AR complexes in band-shift assays (Fig. 7 B ) .

ofReceptor Androgen

Baculovirus Expression

-

4945

binding after renaturationwas about 8 nM, ascalculated from a Scatchard plot (Fig. 8).,A salientfeature in [3H]mibolerone binding was its very high capacity: when calculated on the basis of receptors/cell, each AcrAR-infected insect cell contained over lo7 receptor molecules, i.e. about 100 times more than in thesoluble cytosolic form (Fig. 5). In theexperiments shown in Fig. 8A, the highest proteinconcentration used corresponded to only about 20,000 infected Sf9 cells. The total concentration of AR molecules in insect cells must be even higher, since only 20-50%of the receptor molecules appear to have been renatured to a steroid-binding form under the assay conditions. These calculations are based on the assumption that theAR content in theGdnHC1-solubilized samples I is about 30% (cf. Fig.2) and thateach receptor molecule binds 1steroid molecule; on average, 1 pg of AR protein (10 pmol) I in our studies bound2-5 pmol of ["Hlmibolerone. FIG. 10. DNA-binding specificity of the homogeneous anTable I summarizes some of the experimentsin which drogen receptor. Homogeneous AR renatured by dilution in the different buffers were used to dilute 6 M GdnHCl, in which presence of Zn'+ ions (see "Materials and Methods") was used in all experiments. "P-Labeled 41-nt-long GRE/ARE of the tyrosine ami- Sf9 cell pellets were solubilized, in order to renature the notransferase gene (0.5 ng) was employed as the probe in the gel steroid-binding activity of AR. These studies indicated that retardation assays in the absence (Lane, Comp. -) or presence (+) of dilution of GdnHCl with a buffer containing 20% glycerol three different concentrations (expressed in molar excess) of nonyielded betterrenaturationthan one without glycerol (cf labeled completing oligomers. The sequences of the oligomers are Table I, conditions band c), but that diditnot matterwhether given under "Materials and Methods." Abbreviations: GRE, glucocorticoid responsive element; C3(1), and androgen responsive element the steroid was present during the initial 10-fold dilution ( A R E ) in the first intron of the C3(1) gene (26); muC3(1), mutated (Table I, conditions aand b). Interestingly,removal of EDTA C3(1)ARE (26); VDRE, vitamin D responsive element of the osteo- from the dilution buffer appeared to double the steroid-bindcalcin promoter. ing activity(Table I).When kept in 6 M GdnHCl, AR protein was stable for a t least 2 weeks, as judged by ['HH]mibolerone binding studies (resultsnot shown). In addition to being able to bind steroid upon dilution of 6 M GdnHCl, the solubilized and partially renatured AR also rl interacted with specific DNA sequences in band-shift assays k.4 (Fig. 7C). Although the extentof renaturation of this receptor function was not quantitated, our experience with the GRE/ ARE-AR interaction suggested that not all of the GdnHClsolubilized AR was renatured under the conditions employed. Renaturation of the receptor's DNA-binding ability was greater when the initial dilution was performed in a buffer containing 0.8 M NaCl than when 0.1 M NaCl was employed i . * (Fig. 7C, lanes 1 and 2). Identical results were obtained inthe presence and absence of 5a-dihydrotestosterone (data not FIG. 11. Requirement of divalent cationsfor the acquisition shown). methods were of DNA-bindingactivity by the homogeneous androgen recep-Purification of Recombinant AR-Several tor. The receptor protein purified to apparent homogeneity was tried to maintain the AR protein in a soluble form upon diluted in Zn'+-free buffer R containing the divalent cation concen- removal or dilution of 6 M GdnHCl. These attempts included tration specified in panels A-C and then dialyzed against buffer D 50- or 100-fold dilution of 6 M GdnHCl by adding the diluent containing the same cation concentration for 3 h (see "Materials and Methods"). A, optimization of the ZnC12 concentration. B and C, rapidly into the sample (38) or by adding the 6 M sample comparison of the efficacy of different divalent cations at 40 and 5 gradually to thediluent with stirring,and removal of GdnHCl from the sample by overnight dialysis (39) against NTE or p~ concentrations. The DNAprobe in all experiments was "P-labeled GRE/ARE of the tyrosine aminotransferase promoter. TEDG (50 mM Tris-HC1 (pH 7.5), 0.1 mM EDTA, 5 mM dithiothreitol, and 20% glycerol) buffers. These procedures Renaturation of the Insoluble AR Form-When the insolu- resulted invariably in the formation of slight to heavy precipble Sf9 cell pellet, prepared as described above, was used for itates or provided very large sample volumes. For this reason ['H]mibolerone binding, very little specific binding was meas- and due to the excellent stability of the AR in 6 M GdnHCl, urable (data notshown). On the basis of preliminary studies, we decided to purify the solubilized receptor without a prior the following procedure for the solubilization of the AR pro- dilution, using a Sephacryl S-400 HR column equilibrated teins was adopted; the cells were first broken by freezing and and eluted with abuffer containing 6 M GdnHCl (Fig. 9). The thawing, the pellets washed twice with each of TNNP buffer column fractions were routinely divided into three pools that were analyzed by 10% polyacrylamide, 0.1% NaDodS04 gel (50 mM Tris-HC1 (pH 7.5), 0.5 M NaC1, 1% Nonidet P-40, and 0.5 mM PMSF) and NTE buffer (0.15 M NaC1, 10 mM electrophoresis and the resolved proteins silver-stained (Fig. Tris-HC1 (pH 7.5), and 1mM EDTA), afterwhich the sample 9, inset). In theexperiment shown in Fig. 9, poolI correspondwas solubilized in 6 M GdnHCl for 1 h a t room temperature. ing to theascending part of the main peak contained only the This solubilization of the pellet followed by a 50-fold final 110-kDa protein, whereas pool I1 includes also some smaller dilution of 6 M GdnHCl yielded preparations with considera- molecular weight peptides. The protein concentration in pool ble [3H]mibolerone-binding activity (Table I). The equilib- I11 was much lower than in theother two pools, and it rium dissociation constant for the specific [3H]mibolerone contained mainly peptides of low molecular weight that are Comp.:

GRE

C3(1)

+ + + + + +

muCI(1)

VDRE

+ + + + + + -

"-

I__-

I

4946

Baculovirus Expression ofReceptor Androgen

poorly visible in Fig. 9. The largest scale purification procedure we have so far carried out included 12 T75 flasks of AcrAR-infected Sf9 cells, and it yielded about 2 mgof homogeneous AR protein after gel filtration on Sephacryl S400. Functional Properties of the Purified AR-Similar to the receptor present in cell pellets that were solubilized in 6 M GdnHCl and studies as such, the AR protein purified to homogeneity by Sephacryl S-400 gel filtration could be reconstituted to acquire three functions: steroid binding, specific DNA interaction, and antibody recognition. Under the conditions used in Fig. 8 for the solubilized cell pellet, significant [3H]mibolerone-binding activity (about 1-2 pmol of steroid bound/pg of receptor) with the same Kd as that of the renatured AR prior to gel filtration was measured for the purified protein (results not shown). These conditions may not be optimal, for example, due to much lower protein concentrations used in the assays with homogeneous preparations. As shown in Fig. 70, the homogeneous preparation could be reconstituted to acquire the ability for specific DNA binding. In these and other experiments with the purified AR, only about 15 ng of receptor protein (150 fmol) was used in a 20pl incubation volume, i.e. the final concentration ofAR in these DNA-binding experiments was 7.5 nM. This low protein concentration may explain, at least in part, why nonspecific proteins such as albumin and preimmune rabbit serum stabilized to some extent theGRE/ARE-AR complexes in bandshift studies (data not shown). The probe concentration in these experiments was about 1 nM (20 fmol). Similar to the soluble AR in the cytosol (Fig. 7 B ) , interaction of the pure AR with GRE/ARE was both stabilized and further retarded by the N-terminal antiserum, whereas antibody against the hinge region of AR inhibited this interaction (Fig. 70, lanes 2 and 3). The purified receptor protein interacted with the tyrosine aminotransferase GRE/ARE in a specific manner (Fig. lo), in that binding of the receptor to the labeled probe was completed for by an oligomer containing the GRE consensus sequence (2) and an ARE of the C3(1) gene (26), but not by a vitamin D responsive element of the osteocalcin gene or a mutated ARE in which the hexanucleotide 5’-TGTTCT-3’ is converted to 5’-TTTTCT-3’. This single-nucleotide mutation has also been shown to abolish specific interaction of the ARE of the C3(1)gene with a truncated AR protein produced in Escherichia coli (26). The presence of Zn2+ions in the dilution buffer used for reconstituting samples from 6 M GdnHCl was required for proper folding and acquisition of specific DNA binding of the recombinant AR protein. The initial experimentswere carried out using a buffer containing 0.2 mM EDTA, and theoptimal ZnClnconcentration was about 100p~ under those conditions (data not shown). In subsequent studies on Zn2+and other divalent cations,the final EDTA concentration in thesamples was reduced to 0.1 mM. Under these conditions, the optimal , higher concenZn2+ion concentration was about 50 p ~with trations being clearly inhibitory (Fig. 1lA). Itis also of note that very little DNA-binding activity was acquired when attempts torenature the purified receptor were performed in the absence ofZnC12 (Fig. 1 l A ) . Of other divalent cations tested, Cd2+ions were as active as Zn2+ions but at about 10fold lower concentration (5 p ~ Fig. , 1lC) and similar to ZnC12, higher CdClz concentrations prevented AR binding to the DNA element. In addition to Zn2+and Cd2+ions, some activation of DNA binding was also achieved with MnC12 (Fig. 11, B and C). In contrast to theAR purified in the presence of 6 M GdnHCl, specific DNA binding of the soluble cytosolic

AR in Sf9 cells was minimally influenced by different ZnC12 concentrations (data not shown), implying that this receptor form had been properly folded in vivo in theinsect cells. DISCUSSION

In the present work, we have established conditions for expression of the rat AR protein in S. frugiperda cells with a recombinant baculovirus vector to extremely high levels, up to 30% of total cellular protein. To thebest of our knowledge, this is the first report describing production of milligram quantities of a full-length AR protein in a heterologous system. Over 90% of the recombinant AR was recovered as an insoluble pellet; however, conditions were devised for its complete solubilization and purification to apparenthomogeneity. The insolubility of the expressed protein, although initially disappointing, subsequently facilitated acquisition of homogeneous AR by a very simple purification scheme for two main reasons: first, the insoluble pellet could be washed using buffers of high ionic strength andwith a non-ionic detergent without receptor loss and second, the solubilized receptor could be kept in6 M GdnHCl duringgel exclusion chromatography. We do not know whether other steroid receptors expressed in the baculovirus/insect cell system (21-24) have also been, at leastin part, in an insoluble form asthis possibility was not investigated in the previous studies. Behavior of the recombinant AR protein in insect cells was, however, very similar to that of protein phosphatase 1 (38), which was also expressed to very high levels (25% of total cellular protein) and mainly in aninsoluble form in Sf9 cells. Ultrastructural analysis of recombinant virus-infected cells (and pellets originating from these cells) revealed the presence of electron-dense aggregated particles that were 30-40 nm in diameter and possessed a relatively regular organization. Since these structures were not found in controlcells, or those infected with the wild-type virus, we assume that AR protein is contained in the electron-dense particles. With regard to the size of theseparticles, it is worth pointing out that previous electron microscopic measurements for tetrameric glucocorticoid and progesterone receptors have yielded diameters of about 20 nm (40, 41), implying that each particle should contain multiple AR molecules. The recombinant AR protein expressed in S. frugiperda cells in a soluble form appeared to behave ina fashion indistinguishable from the receptor in mammalian cells by at least two criteria: steroid binding and interaction with a specific DNA element. This result and similar findings on other steroid receptors, which have been expressed in insect cells (21-24), imply that these cells are capable of folding and processing eukaryotic steroid receptors in a proper way. In view of this, it is unlikely that aberrant synthesis, improper folding or incomplete processing of the AR were reasons for its initially insoluble form; in particular, since we were able to renature some of this latter receptor form to acquire functional characteristics indistinguishable from those of the soluble form. It is of particular note that both steroid binding and interaction with a specific DNA motif, GRE/ARE of the tyrosine aminotransferase gene (32), occurred with a homogeneous AR that was renatured, indicating that no accessory proteins may be needed for these two functions. The result on GRE/ARE-AR interaction is different from those reported for recombinant vitamin D receptor expressed in yeast and purified to homogeneity (42) and thyroid hormone (43) and retinoic acid (44) receptors produced by in vitro translation of the respective mRNAs, because each of the three latter receptors appeared to need nuclear accessory proteids) for specific DNA binding. Moreover, human progesterone and

Baculovirus Expression of Androgen Receptor estrogen receptors have been shown to bind better to their respective hormone response elements in the presence of an accessory protein (45, 46). The DNA-binding domain of the rat AR expressed in E. coli as a fusion protein and purified to apparent homogeneity did not, however, require additional proteins for a specific DNA interaction (26). The results on androgen receptor are not necessarily contradictory to those on otherreceptors, as itis highly likely that themajor role of these accessory proteins is to enhance the affinity and stability of the DNA-receptor complexes. Similar to our findings, other recombinant AR proteins containingthe ligand-binding site have been shown to maintain their high affinity for the cognate steroid (25, 27), even in theabsence of hsp90 (27), as was the case in our experimentswith the purified recombinant AR. The presence of the steroid was not, however, required for the soluble or reconstituted AR to interactwith a specific DNA motif. The DNA-binding domain of all steroid receptors contains two cysteine-rich "zinc-fingers," motifs that are required for binding to the DNA (1-4). When a 150-amino acid glucocorticoid receptor fragment encompassing the DNA-binding domain was expressed in E. coli and purified to virtual homogeneity, it was shown to ligate reversibly two Zn2+ ions. Release of the zinc ions from the protein by chelating agents yielded an apoprotein that failed to bind to its target DNA element; the binding was restored by a preincubation with Zn2+or Cd" ions (47). Our studies, performed with a fulllength AR purified to homogeneity in the presence of 6 M GdnHCl by gel filtration, agreed completely with these data, and the resultsindicated that Zn2+ ionsareessential for refolding the receptor from GdnHCl to acquire its specific DNA-binding function. In our experimentsthe optimal ZnC12 concentration was about 50 PM, i.e. 5 times lower than the optimum (250 PM) reported by Freedman et al. (47) for the DNA-binding domain of the recombinant glucocorticoid receptor. Interestingly, the optimal Cd" ion concentration for the full-length AR was the same as that for the DNAbinding domain of the glucocorticoid receptor (5 W M in both cases). It should be mentioned, however, thatSakaiand Gorski (48) in their studies on denaturation and renaturation of cytosolic estrogen receptors using 6 M GdnHCl reported reconstituting the receptor's ability to bind DNA even without added zinc ions. This latter study differed from our experiments in two important aspects: we used purified as opposed to crude cytosolic receptor preparations and studied receptor binding to a specific (GRE/ARE) DNA element instead of nonspecific (calf thymus) DNA. The two antisera against AR peptide 3 (N-terminal) and AR peptide 5 (hinge region) were very useful reagents in our studies. The former antibody corresponding to residues 1432 of the human AR reacted well with the recombinant receptor produced in S. frugiperda cells, indicating that the majority of the protein was full-length. Moreover, this antibody served also as a convenient specificity control in bandshift experiments and quantitatively up-shifted GRE/AREAR complexes. By contrast, the hinge-region antiserum inhibited the interaction of both soluble and renatured AR preparations with a specific DNA element. Two possible mechanisms may have been responsible for this effect: (i) direct masking by the antibody of the zinc-finger region in the AR that interactswith the DNA element or (ii) inhibition of receptor dimerization, which may be important for the formation of specific DNA-receptor complexes (1-4). It remains tobe elucidated whether this particular antiserum also inhibits AR-dependent transcription invitro, ina fashion similar to an antiserum directed against the DNA-binding

4947

domain of the estrogen receptor, which prevented both estrogen response element-receptor interactionand estrogen receptor-mediated transcription in a cell-free system (49). Acknowkdgrnents-We thank Drs. S. Liao and C. Chang (University of Chicago, Chicago, IL) for providing us with the ratAR cDNA.

1. 2. 3. 4. 5. 6. 7. 8.

9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33.

REFERENCES Evans, R. M. (1988) Science 240,889-895 Beato, M. (1989) Cell 5 6 , 335-344 Yamamoto, K. (1985) Annu. Reu. Genet. 19,209-252 O'Malley, B. (1990) Mol. Endocrinol. 4 , 363-369 Wilson, J. D., George, F. W., and Griffin, J. E. (1981) Science 211,1278-1284 Catterall, J. F., Kontula, K. K., Watson, C. S., Sepphen, P. J., Funkenstein, B., Melanitou, E., Hickok, N. J., Bardin, C. W., and Janne, 0. A. (1986) Recent Progr. Horm. Res. 42,71-109 Berger, F. G., and Watson, G. (1989) Annu. Rev. Physiol.51,5165 Chang, C., Kokontis, J., and Liao, S. (1988) Proc. Natl. Acad. Sci. U. S. A . 8 5 , 7211-7215 Lubahn, D. B., Joseph, D. R., Sar, M., Tan, J., Higgs, H. N., Larson, R. E., French, F. S., and Wilson, E. M. (1988) Mol. Endocrinol. 2 , 1265-1275 Tan, J. A., Joseph, D. R., Quarmby, V. E., Lubahn, D. B., Sar, M., French, F. S., and Wilson, E. M.(1988) Mol. Endocrinol. 2,1276-1285 Tilley, W. D., Marcelli, M., Wilson, J. D., and McPhaul, M. J. (1989) Proc. Natl. Acad. Sci. U. S. A . 86,327-331 Janne, 0. A., and Shan, L.-X. (1991) Ann. N. Y. Acad. Sci. 6 2 6 , 81-91 Rundlett, S. E., Wu, X.-P., and Miesfeld, R. L. (1990) Mol. Endocrinol. 4 , 708-714 Simental, J. A., Sar, M., Lane, M. V., French, F. S., and Wilson, E. M. (1991) J. Biol. Chem. 266,510-518 Chang, C. H., Rowley, D. R., and Tindall, D. J. (1983) Biochernistry 22,6170-6175 Johnson, M. P., Young, C. Y. F., Rowley, D. R., and Tindall, D. J. (1987) Biochemistry 26,3174-3182 Smith, G. E., Fraser, M. J., and Summers, M. D. (1983) J. Virol. 46,584-593 Smith, G. E., Summers, M. D., and Fraser, M. J . (1983) Mol. Cell. Biol. 3,2156-2165 Luckow, V.A., and Summers, M. D. (1988) BiolTechnology 6 , 47-55 Summers, M. D., and Smith, G. E. (1987) A Manual of Methods for BaculovirusVectorsandInsect Cell CultureProcedures, Texas Agricultural Experiment Station Bulletin No. 1555 Srinivasan, G., and Thompson, E. B. (1990) Mol. Endocriml. 4 , 209-216 Fawell, S. E., White, R., Hoare, S., Sydenham, M., Page, M., and Parker, M. G . (1990) Proc. Natl. Acad. Sci. U. S. A. 87, 68836887 Brown, M., and Sharp, P. A. (1990) J. Bwl. Chem. 2 6 5 , 1123811243 Alnemri, E. S., Maksymowych, A.B., Robertson, N. M., and Litwack, G. (1991) J . Bwl. Chern. 2 6 6 , 3925-3936 Young, C. Y.-F., Qiu, S., Prescott, J. L., and Tindall, D. J. (1990) Mol. Endocriml. 4 , 1841-1849 De Vos, P., Claessens, F., Winderickx, J., Van Dijck, P., Celis, L., Peeters, B., Rombauts, W., Heyns, W., and Verhoeven, G. (1991) J . Biol. Chem. 266,3439-3443 Ohara-Nemoto, Y., Nemoto, T., and Ota, M. (1991) J. Biochern. 109, 113-119 Ausubel. F. M.. Brent. R.. Kingston. R. E.. Moore. D. D.. Seidman. J. G., Smith; J. A.,'and StGhl, K. (1989) Current Protocols in Molecular Biology, John Wiley & Sons, New York Shan, L.-X., Rodriguez, M.C., and Janne, 0. A. (1990) Mol. Endocrinol. 4 , 1636-1646 Laemmli, U. K. (1970) Nature 227,680-685 Isomaa, V., Pajunen, A. E. I., Bardin, C.W., and Janne, 0. A. (1982) Endocrinology 1 11,833-843 Strahle, U., Klock, G., and Schutz, G . (1987) Proc. Natl. Acad. Sci. U. S. A . 8 4 , 7871-7875 Phillips, D. M., Zacharopoulos, V., and Perotti, M. E.(1990) Cell Tissue Res. 261,249-259

Baculovirus Expression of Androgen Receptor

4948

34. Krieg, A., and Huger, A. M. (1969) J. Znuert. Pathol. 13, 27242. 279 35. van der Wilk, F., van Lent, J. W.M., and Vlak, J. (1987) J. Gen. Virol. 68. 2615-2623 43. 36. Robel, P., Eychenne, B., Blondeau, J. P., Ingrid, J. T., Groyer, M. T., Christine, M. B., Hechter, O., Roux, C., and Dadoune, J. P. (1983) Hormone Res. 18,28-36 37. Denison, S. H., Sands, A., and Tindall, D. J. (1989) Endocrinology 124,1091-1093 38. Berndt, N., and Cohen, P. T. W. (1990) Eur. J. Biochem. 190, 291-297 39. Teipel, J. W. and Koshland, D. E., Jr. (1971) Biochemistry 10, 798-805 40. Payvar, F., DeFranco, D., Firestone, G. L., Edgar, B., Wrange, Okret, S., Gustafsson, J.-A., and Yamamoto, K. R. (1983) Cell 35,381-392 41. ThBveny, B., Bailly, A., Rauch, C., Rauch, M., Delain, E., and

o.,

44. 45. 46. 47. 48. 49.

Milgrom, E. (1987) Nature 329, 79-81 Sone, T., McDonnell, D. W.,P., O’Malley, B. and Pike, J. W. (1990) J. Biol. Chem. 265,21997-22003 Murray, M. B.,andTowle, H. C. (1989) Mol. Endocrinol. 3,14341442 Glass, C. K., Lipkin, S. M., Devary, 0. V., and Rosenfeld, M. G. (1989) Cell 59, 697-708 Edwards, D. P., Kuhnel, B., Estes, P. A., and Nordeen, S. K. (1989) Mol. Endocrinol. 3, 381-391 Mukheriee. R. and Chambon. P. (1990) Nucleic Acids Res. 18. 5713-*5716 Freedman. L. P.. Luisi. B.F.. Korszun. Z. R..BasavaDDa.R.. Sigler, P. B., and Yamamoto, K. R. (1988) Nature 334, 5431 546 Sakai, D. and Gorski, J. (1984) Endocrinology 116,2379-2383 Elliston, J. F., Fawell, S. E., Klein-Hitpass, L., Tsai, S. Y., Tsai, M.-J., Parker, M. G., and O’Malley, B. W. (1990) Mol. Cell. BWl. 10,6607-6612 ,

.

I