Steroid Requirement for Androgen Receptor Dimerization and DNA Binding MODULATION BY INTRAMOLECULAR INTERACTIONS BETWEEN THE NH2-TERMINAL AND STEROID-BINDING DOMAINS*

Construction Transfer Vectors-Full-length and deletion mutants of human constructed in the 9.1-kb pAcC4 transfer vector. Polymerase chain reaction amplification of the human NcoIIAflII NHz-terminal fragment constructed Taq

roid receptors was shown to occur in Spodoptera frugiperda (Sf9)l cells infected with recombinant baculoviral vectors (2,3,5), and the expressed receptors were functional in transcriptional activation studies in vitro (6). Since transient transfection in monkey kidney COS cells yields androgen receptor (AR) in a predominantly insoluble form in limited amounts, baculovirus was used for overexpression of human AR.
An important question that remains in the functional analysis of AR is whether dimer formation requires hormone binding and occurs in association with DNA binding. Many transcription factors, including members of the steroid receptor family, undergo dimer formation in acquiring high affinity DNA binding (9). Dimerization of steroid receptors was demonstrated in the mobility shift assay using wild type and truncated forms of the progesterone, glucocorticoid (IO), and estrogen (11) receptors. Heterodimers were reported for the thyroid hormone, retinoic acid (12)(13)(14), and vitamin D receptors. The vitamin D receptor forms heterodimers with a ubiquitous 55-kDa protein that enhances DNA binding (15, 16). Heterodimer formation with coregulatory proteins stabilizes specific DNA interactions of this group of transcriptional regulatory proteins (14).
Construction of Transfer Vectors-Full-length and deletion mutants of human AR were constructed in the 9.1-kb pAcC4 transfer vector. Polymerase chain reaction (PCR) amplification of the human NcoIIAflII NHz-terminal fragment was constructed using Taq polymerase such that the 5' NcoI site contained the starting Met of AR coincident in position with the starting Met of the polyhedron protein.
All PCR-amplified regions were verified by sequence analysis using double-stranded sequencing with Sequenase. Triple ligations were performed using the transfer vector restricted with NcoI and BamHI, the NH2-terminal PCR-amplified NcoIIAflII fragments, and the AflIIlBamHI fragments containing the AR coding sequence. For dimerization studies, two major deletion mutants of human AR were constructed in the pAcC4 transfer vector. Deletion of the steroidbinding domain was achieved by restricting the full-length human AR transfer vector with TthlllI and XbaI to release the steroidbinding domain; the vector was made blunt ended with T4 DNA polymerase and self-ligated Deletion of the NHz-terminal domain was performed by restricting the full-length human AR transfer vector with NcoI and KpnI and blunt end-ligated; initiation of translation was from an internal methionine at position 507 (17).
Sfs CeU Cutture and Transfection"Sf9 cells were maintained in 60-100-ml spinner cultures in Grace medium containing 0.33% yeastolate, 0.33% lactalbumin hydrolysate, 10% fetal calf serum, 100 units/ml penicillin, 100 pg/ml streptomycin, and 70 pg/ml gentamicin, and were passaged every 3 days at an initial density of 0.5 X 10' cells/ml. Sf9 cells (4 X 10' cells/6-cm dish) were cotransfected with 1 pg of circular AcMNPV wild type viral DNA and 5 pg of AR transfer vector using 0.75 ml of Invitrogen transfection buffer in the presence of Grace medium that contained 10% serum but lacked supplements. After a 4-h incubation at 27 "C, medium was removed and the cells were incubated in complete Grace medium for 5 days. The transfection supernatants were subjected to serial dilution in 96-well microtiter plates containing 2 X lo' Sf9 cells/well, and recombinant viruses were identified by dot-blot hybridization as previously described (18). The radiolabeled probe was an [aS2P]dCTP random prime labeled HindIII/EcoRI DNA fragment of human AR that spans the DNAbinding domain and part of the steroid-binding domain. Three rounds of serial dilution of positive samples were followed by three rounds of plaque purification, where recombinant plaques were visualized using the methyl red overlay technique (19).
Phosphorylation of Baculovirus-expressed AR-Sf9 cells were infected with recombinant human AR baculovirus at a multiplicity of infection (m.0.i.) of 4 using 10' cells/lO-cm dish for 40-48 h. Medium was replaced with Excell 401 containing L-glutamine but lacking phosphate or L-methionine. After a 20-min incubation at 27 'C, 200 pci of [8zP]orthophosphate or 200 pCi of Trans%-label was added and incubated for 4 h at 27 "C. Cells were harvested in 4 "C PBS, pH 6.4, microfuged for 30 s, and resuspended in 800 pl of a buffer containing 1 mM EGTA, 1 mM EDTA, 12 mM monothioglycerol, 0.5 M NaCl, 50 mM NaF, 1 mM benuunidine, 1 mM iodoacetamide, 20 mM sodium molybdate, 100 nM R1881,0.5 mM phenylmethylsulfonyl fluoride, 1 pM leupeptin, 1 pM pepstatin, and 50 mM potassium phosphate buffer, pH 7.0. Cells were lysed using three cycles of freeze/ thaw and AR immunoprecipitated using antipeptide antibody AR52 and analyzed on an 8% SDS-polyacrylamide gel as previously described (20).
Immunoblot Analysis-In a control experiment, AR expressed from Sf9 cells was analyzed on immunoblots. Sf9 cells were plated at lo' cells/lO-cm dish, infected with recombinant virus for 40-48 h at m.0.i. 4, and treated for varying times with increasing concentrations of dihydrotestosterone. Cells were washed and harvested in PBS, pH 6.4, at 4 "C, pelleted, and resuspended in high salt extraction buffer (0.5 M NaC1,l mM EDTA, 1 mM dithiothreitol, 10% glycerol, 10 mM Tris, pH 7.4, containing the protease inhibitors, 0.5 mM phenylmethylsulfonyl fluoride, 0.1 p~ aprotinin, 1 pM leupeptin, 1 pM pepstatin A, and 1 mM benzamidine). Cells were exposed to three cycles of freeze/thaw, incubated on ice for 1 h, microfuged for 30 min, and the supernatant dialyzed against the above resuspension buffer containing 50 mM KCI. The dialyzed supernatant and the pellet were solubilized in SDS sample buffer (2% SDS, 10% glycerol, 10 mM Tris, pH 6.8) and analyzed on an 8% acrylamide minigel. Immunoblots were performed using antipeptide antibody AR52 as previously described (21). Zmmunaytochemistry-Sf9 or HighFivee insect cells were infected with recombinant AR baculovirus at m.0.i. 4 in serum free Excell 400 media for 40 h at 27 "C in the presence or absence of 100 nM R1881. Cells were washed in PBS, pH 6.4, and harvested in the same buffer. Cells were smeared on glass slides, air dried, and fixed in 95% alcohol at -20 "C for 10 min. After washing with PBS, pH 7.4, cells were incubated with primary antibody AR32 overnight as previously described (22-24). After incubation, cells were treated with fluorescein isothiocyanate-conjugated goat anti-rabbit IgG (1:400; Organon Technika, Cochranville, PA) for 1 h at room temperature. Slides were washed with PBS, pH 7.4, and mounted using 90% glycerol, 1 mM Tris, pH 7.6, and a cover glass placed on the cells. Slides were viewed using a Nikon-UFX-DA fluorescent microscope with a B-2A filter.
Steroid Binding Activity-Steroid binding was performed on intact Cells were harvested in ice-cold PBS, pH 6.4, and resuspended in TEDG buffer (0.5 M NaCl, 1 mM EDTA, 1 mM dithiothreitol, 10% glycerol, 10 mM Tris, pH 7.4, and protease inhibitors indicated above). Cells were lysed by three cycles of freeze/thaw and incubated on ice for 1 h with vortexing every 15 min. Cell lysates were microfuged for 30 min at 4 "C, and the supernatants were dialyzed for 3 h at 4 "C on a floating 0.025-pm pore, MF-Millipore cellulose membrane filter against TEDG buffer containing 50 mM KC1 to reduce the salt concentration, and microfuged for 1 min.
A 27-bp oligonucleotide (AAGCTl'AGTACGTGATGTTCTAA-GCTT) designated oligonucleotide C h a s i i c e derived from the 0.5-kb first intron region of the rat C3 prostatein gene (25). The underlined regions highlight the androgen response element sequence. In a control study for monomer binding, a 21-bp oligonucleotide (TCGACTGACTCAATTG"G) with flanking SalI sites was used that contained only the right half-site (underlined) of the ARE sequence, shown previously to be the high affinity site for the glucocorticoid receptor (26). Double-stranded oligonucleotides were labeled using [a-3ZP]dCTP and the Klenow fragment of DNA polymerase. The reaction buffer containing 4 pg of poly(dI-dC), 80 pg of bovine serum albumin, 15-30 pg of infected Sf9 whole cell-extraded protein, 40 mM KCl, 20% glycerol, 0.2 mM EDTA, 2 mM dithiothreitol, and 10 mM Tris, pH 7.5, in a total volume of 20 pl, was preincubated for 15 min on ice. Labeled oligonucleotides (0.2-0.3 ng, 10,ooO or 20,000 cpm) were added and incubated for 1 h on ice. The samples were incubated an additional 1 h in the absence or presence of 1 pg of antipeptide antibody AR52 as previously described (24). For mobility shift assays to demonstrate AR dimerization, the amounts of cell extracts ranged from 20-30 pg where half was either the full-length or deletion mutant AR extract. Samples were subjected to electro-phoresis on nondenaturing 4 or 5% acrylamide gels containing 0.5 X TBE (0.1 mM EDTA, 45 mM boric acid, and 45 mM Tris, pH 8.4). Gels were preelectrophoresed at 100 V for 1 h and samples separated at 150 V at 4 "C for 4-5 h or for 260 V for 2 h (25, 27,28). Gels were dried under vacuum and autoradiographed by exposure to Kodak X-OMAT film at -7OC for 24-48 h.

Construction of Full-length and Truncated
Human AR Recombinunt Baculouiruses-The full-length coding sequence of human AR was cloned into the NcoI and BamHI sites of the polylinker region of pAcC4 baculovirus transfer vector (Cetus). Cloning was achieved by ligation of the restricted vector with a 517-bp PCR amplified NHz-terminal NcoI/AflII AR fragment and a 2.24-kb AflIIIBamHI AR fragment from the mammalian expression vectors pCMVhAR (29). As shown in Fig. lA, the vector contains sequences required for homologous recombination with the wild type nuclear polyhedrosis virus from A. californica, AcMNPV, and a multiple cloning site that allows placement of the starting AR methionine coincident with that of the polyhedron protein. Two major deletion mutants of human AR were also constructed (see "Experimental Procedures") such that the coding sequence contained the NHz-terminal and DNA-binding domains (amino acids 1-660, designated AR1-660) or the DNA-binding, hinge, and steroid-binding domains (amino acids 507-919, designated AR507-919) of the 919-amino-acid human AR. Sf9 insect cells were cotransfected with the transfer vectors and wild type AcMNPV DNA. Viral stocks were collected from infected cells and recombinant plaques enriched by serial dilution and dot-blot hybridization. Recombinant plaque purification was achieved by visual screening for absence of polyhedrin-containing occlusion particles. Several plaques for each construction were selected for further characterization.
Properties  nonspecific band was observed when extracts from uninfected Sf9 cells were combined with ARE DNA, a protein-DNA complex that was unaffected by AR antibody addition (Fig. 3,  lanes 1 and 2). No additional DNA complexes were observed using extracts from cells infected with recombinant AR baculovirus in the absence of androgen (Fig. 3, lane 13). However, addition of dihydrotestosterone to cells expressing AR resulted in a dose-dependent formation of an AR -DNA complex that increased in intensity as the concentration increased from 10 to 50 nM dihydrotestosterone (Fig. 3, lanes 3,5, 7, 9, and 11 ). The most intense band observed in the absence of antibody occurred following 24-48 h of treatment with 50 nM dihydrotestosterone (Fig. 3, lane 14). Unlabeled ARE DNA in 100-fold excess competed for AR. DNA complex formation (data not shown) indicating that the DNA binding was specific.
Dose-dependent androgen-induced increases in AR. DNA complex formation were also evident in the presence of the AR antipeptide antibody AR52. In the absence of cell exposure to dihydrotestosterone, only weak intensity bands were observed with full-length AR when tested with AR52 IgG at increasing concentrations of Sf9 cell extracts (Fig. 3,

15-17).
Addition of 50 nM dihydrotestosterone to the Sf9 cell incubations 48 h prior to harvest and AR52 IgG to the binding reactions caused up to a 120-fold increase in band intensity (Fig. 3, lanes 18-20). A dose-dependent increase in complex formation with increasing androgen concentrations was also observed in the presence of AR antibody (Fig. 3, lanes 4,6,8 Fig. 4, hormone-induced ARE binding reflected the ability of that hormone to activate a reporter gene previously reported (20). Strongest DNA binding was noted with the androgens, R1881, dihydrotestosterone, and testosterone (Fig. 4, lanes 1, 3, and  5), and the addition of AR antibody further enhanced DNA binding (Fig. 4, lanes 2,4, and 6). Estrogen and progesterone promoted weak DNA binding only apparent in the presence of AR antibody (Fig. 4, lanes 7-10). Weakest activities were noted with dexamethasone and flutamide which do not bind AR with significant affinity at the concentrations tested. Somewhat to our surprise, the antiandrogen cyproterone acetate (Fig. 4, lanes 13 and 14) and the antiprogestin RU486 (Fig. 4, lanes 19 and 20) promoted relatively strong AR DNA binding, particularly in the presence of antibody. Both of these antihormones were shown previously to display agonist activity (20).
Of particular interest was hydroxyflutamide, since it binds  R1881 (lanes 1 and 21, DHT  (lanes 3 and 4 ) , or testosterone (T, lanes 5 and 6 ) , or 100 nM estradiol (E, lanes 7 and 8), progesterone (P, lanes 9 and I O ) , dexamethasone (Der, lanes 11 and 12), cyproterone acetate (CA, lanes 13 and 14), flutamide (FL, lanes 15 and 16), hydroxytlutamide (OH-FL, lanes 17  and 18), or RU486 (lanes 19 and 20). Even numbered lanes had 1 pg of AR52 antibody added during the DNA binding reactions. AR with moderate affinity, causes nuclear transport, but fails to transactivate an androgen-responsive reporter gene (20). Hydroxyflutamide was a poor activator of AR DNA binding (Fig. 4, lanes 17 and 18). Furthermore, including hydroxyflutamide together with DHT in the cell incubations markedly reduced AR DNA binding (Fig. 5). These results suggest that the antagonist activity of hydroxyflutamide results from its inability to invoke the necessary intracellular changes in AR to permit DNA binding.
Receptor Dimerization-AR dimer formation was investigated using two deletion mutants of human AR expressed in baculovirus: AR1-660 contained the NHn-terminal and DNAbinding domains and AR507-919 contained the DNA-, hinge, and steroid-binding domains (diagrammed in Fig. 7). Expression of the deletion mutants from baculovirus was verified by immunoblot analysis, revealing truncated receptor forms of 85 * 3 kDa (AR1-660, Fig. 6, lane 5, N ) and 43 f 1.5 kDa (AR507-919, Fig. 6, lane 6, C) relative to 118 k 4 kDa fulllength AR (Fig. 6, lane 4). The smaller sized bands reflect limited proteolytic cleavage prior to analysis. Dimerization was examined using 32P-labeled ARE DNA described above, and initially, receptors extracted from cells exposed to dihydrotestosterone.
Incubation of AR with ARE DNA resulted in a slowly migrating AR-DNA complex (Fig. 7, lane 7). A faster migrating complex was observed with AR507-919 (Fig. 7, lane 3), and AR1-660 formed a weak DNA complex migrating similar to full-length AR-DNA (Fig. 7, lane 5 ) . Addition of antipeptide antibody AR52 caused a further shift in each band, verifying the presence of AR in the protein-DNA complexes (Fig. 7, lanes 4, 6, and 8). No specific bands were observed using extracts of uninfected cells in the presence or absence of antibody (Fig. 7, lanes 1 and 2).
AR dimerization was evident in the mobility shift assay using combinations of truncated and full-length receptors. A major band of intermediate migration formed between deletion mutants AR1-660 and AR507-919 (Fig. 7, lanes 10-12). The intermediate band was 39-and 6-fold greater intensity than either mutant alone, respectively, suggesting a synergistic increase in DNA binding with dimerization. About 15% residual fragment AR507-919-DNA complex migrated in its original position (Fig. 7, lane 12 compared with lane 11). Dimer formation between the dissimilar deletion mutants was therefore stronger than between either mutant alone.
Combining full-length AR and AR507-919 also resulted in strong dimer formation, with an intermediate band (Fig. 7, lanes 14-16) migrating slightly slower than the AR1-660-AR507-919 DNA complex ( l a n e 12). Residual AR-DNA complexes remained detectable at their original positions of migration. The ratio of intensity of the three bands was 2:7:1 (Fig. 7, lane 16), corresponding to full-length AR, dimer, and AR507-919. Thus, dimerization between full-length AR and AR507-919 was preferred to dimerization of either with itself, assuming that AR. DNA complexes observed with each fragment alone represents dimerization (see below).
binding domains, the two NH2-terminal domains in the AR dimer inhibit DNA binding.
Intracellular Androgen Dependence of in Vitro Dimerization-We next determined whether androgen exposure of Sf9 cells expressing AR was required for AR507-919 dimerization with itself or with full-length AR. AR507-919 extracted from cells untreated with androgen produced a strong migrating DNA-protein complex (Fig. 8, lane 3 ) which migrated slightly faster than when extracted from androgen treated cells (Fig.   8, lane 1 ). The strong band intensities suggest that AR507-919 is capable of binding DNA independent of androgen exposure in Sf9 cells. The slight difference in migration may result from conformational differences in AR507-919 in the presence or absence of androgen binding. Both AR507-919-DNA complexes shifted with antibody (Fig. 8, hnes 2 and 4 ) .
Further evidence for dimerization of AR507-919 independent of androgen was obtained using an oligonucleotide containing a single right half-site ARE consensus sequence TGTTCT (see "Experimental Procedures"). A similar oligonucleotide was used previously to demonstrate monomer binding of the glucocorticoid receptor (30). Little or no detectable bands were observed using full-length AR507-919 with the right half-site oligonucleotide (Fig. 9, lanes 6 and 8). In a control experiment, full-length glucocorticoid receptor expressed from baculovirus displayed binding to both the halfsite and full ARE (Fig. 9, lanes 3 and 4 ) . Monomer binding of the glucocorticoid receptor was also evident by the change in mobility of DNA complexes formed between the dissimilar oligonucleotides ( Fig. 9, lanes 3 and 4 ) . The results suggest that AR, either in its full-length or NH2 terminally truncated form, binds DNA primarily as a dimer and that ARE binding of AR507-919 reflects dimer binding. Furthermore, AR507-919 dimerization and DNA binding occurs independent of androgen exposure. Full-length AR dimerization with AR507-919 was examined where either one or both were exposed to androgen in Sf9 cells. As shown above, dimerization occurred when both full-length AR and AR507-919 were exposed to androgen (Fig. 8, lanes 9-11). However, if either was untreated with androgen, no dimerization was observed (Fig. 8, lanes 12-20).
Thus, unlike AR507-919 dimerization with itself, dimerization of full-length AR with AR507-919 required that each be exposed to androgen in Sf9 cells. The results support the hypothesis that the AR NHp-terminal domain inhibits dimer formation in the absence of androgen and that androgen treatment relieves this inhibition.
Since AR1-660 and AR507-919 form a strong dimer band (see Fig. 7, lane 12), it was important to establish whether AR507-919 in this complex required androgen exposure as it did in dimerization with full-length AR but not with itself. Fig. 10 shows that androgen exposure of AR507-919 in Sf9 cells is required for dimerization with AR1-660. The results reaffirm the inhibitory influence of the NH2-terminal domain in dimer formation, even in a complex where one fragment contains the steroid-binding domain and the other contains the NHn-terminal domain. The results raise the possibility that regions required for strong dimer formation are located within the domains of these dissimilar AR fragments.
Finally, we tested the hormone specificity of dimerization. In control experiments in the absence of hormone, full-length AR fails to bind the ARE, AR507-919 binding is strong and dimerization between them is undetectable (Fig. 11, lanes 2-4 ) . Strong dimerization occurs with R1881 and DHT (Fig. 11,  lanes 5-10). Dimers are also detected with cyproterone acetate ( l a n e 13) and RU486 ( l a n e 16), but DNA binding and dimerization fails to occur with hydroxyflutamide (lanes [17][18][19]. It is noteworthy that the migration of AR507-919 changes markedly depending on hormone treatment. For example, AR507-919 migration with dihydrotestosterone (Fig. 11, lane 9 ) differs from that with R1881 ( l a n e 6) but is similar to that observed in the absence of hormone ( l a n e 3). Also, it is important to note that hydroxyflutamide inhibits ARE binding of both full-length AR ( l a n e 17) and AR507-919 (compare lanes 3 and 18). The change in AR507-919 DNA complex migration likely reflects changes in receptor conformation resulting from hormone binding.

DISCUSSION
Overexpression of recombinant steroid receptors provides an important approach with which to determine the molecular properties of this family of transcriptional regulatory proteins. By the criteria of androgen binding, subcellular localization,    (20). Particularly noteworthy was the differential response of the antihormones: those that display agonist activity in transient cotransfection assays (cyproterone acetate and RU486) induce AR DNA binding, while the pure antiandrogen, hydroxyflutamide, which lacks agonist activity, failed to potentiate AR DNA binding, and, in fact, inhibited AR DNA binding when added together with androgen. The antagonistic activity of hydroxyflutamide results, therefore, from its failure to induce changes in AR necessary for DNA binding.
To investigate further the molecular basis for the steroidmediated changes in AR required for DNA binding, two AR deletion mutants were expressed from baculovirus in Sf9 cells.
The results indicate that AR undergoes dimerization in association with ARE binding, and, as a monomer, fails to bind the ARE. Furthermore, by using different combinations of full-length and AR deletion mutants, it was established that dimerization was dependent on intracellular hormone exposure in order to overcome inhibition created by the AR NH2terminal domain. This conclusion was supported by the observation that deletion mutant AR507-919, which lacked the NH2-terminal region, dimerized and bound DNA efficiently in the absence of hormone, and that the presence of the NH2terminal domain blocked this hormone independent DNA binding. Finally, it is speculated that the hormone specificity of DNA binding reflects the ability of a given hormone to promote conformational changes in AR that overcome NHZterminal domain inhibition.
The critical role of hormone in steroid receptor action is believed to be activation of receptors to a DNA binding state. The demonstration of hormone-dependent DNA binding is sometimes complicated by isolation procedures that often cause artificial receptor activation so that receptors no longer require hormone for high affinity DNA binding. Sf9 cells expressing AR must be exposed to androgen intracellularly in order to render the receptor capable of binding DNA in vitro. In a recent study using baculovirus-expressed AR, binding to androgen response element DNA occurred independent of androgen exposure (7). An explanation for the differing results is that the AR used in the previous report was extracted using denaturing conditions and was subsequently renatured. Ligand-dependent DNA binding was reported, however, for baculovirus-expressed human progesterone receptor (5,6).
Antihormones bind AR with moderate affinity, cause nuclear transport (20), but, as shown in this report, differ in their ability to induce DNA binding. These observations suggest that hormone binding may impose different conformations on the receptor, which either allow or disallow DNA binding. That steroid hormones play a major role in receptor conformation is supported by the recent study of Allan et al. (31) where proteolytic cleavage products differed in size after hormone and antihormone exposure of the progesterone receptor. It was proposed that distinct receptor conformations are induced by hormone and antihormone binding. The primary site for conformational modification was thought to be the carboxyl-terminal tail (32). A monoclonal antibody raised against the 14 terminal amino acids of the progesterone receptor only recognized the progesterone receptor when unliganded or bound to the steroid antagonist, RU486, suggesting that agonist binding induces a unique receptor conformation (33). Our studies using hydroxyflutamide support the formation of an improper receptor conformation that interferes with DNA binding. This was evident even with the NH, terminally truncated mutant AR507-919, that without hormone, displayed strong DNA binding, but with hydroxyflutamide, showed inhibition of DNA binding.
Further evidence for the importance of the carboxyl-termind tail in conformational effects that differentiate agonist and antagonist activity comes from a switch in the activity of hydroxyflutamide from antagonist to agonist in an AR mutant. Agonist activity of hydroxyflutamide is undetectable with wild type AR but becomes significant with an AR mutant containing a single base mutation at amino acid residue 877, changing threonine to alanine, in the carboxyl-terminal region of the AR steroid-binding domain (34,35). It could be speculated that the single residue change contributed to a change in AR conformation when bound to hydroxyflutamide which potentiated DNA binding and thus, transactivation. Similarly, a single amino acid change in the progesterone receptor disallowed antagonist RU486, but not agonist binding (36).
Alternative mechanisms for the molecular basis of steroid antagonism include reduced nuclear translocation with concomitant inhibition of 10 S to 4 S conversion, as shown for the glucocorticoid receptor (37). All of the antihormones tested in the present report, including hydroxyflutamide, caused nuclear transport of AR (20) and preliminary studies with baculovirus-extracted receptor indicate that hormoneand nonhormone-treated receptors migrate as 5-6 S complexes on sucrose gradients. Studies on RU486 interaction with the human progesterone receptor indicate that binding to a specific response element was indistinguishable from that with the synthetic agonist, R5020 (38). Although the progesterone receptor binds its response element when associated with RU486, it apparently binds with a somewhat altered conformation reflected by differences in migration on sucrose gradients and in DNA mobility shift assays; failure to transactivate results therefore from structural alterations (39).
Using deletion mutants of the rabbit progesterone receptor, it was demonstrated that RU486-bound receptors interact with the same response element (40). In agreement with these studies, the AR-RU486 dimer appeared to migrate with slightly altered mobility. These studies support the important role of receptor conformation mediated by steroid binding in transcriptional activation. They point out that the molecular basis of hydroxyflutamide antagonism differs from that of most antihormones reported thus far, which bind with high affinity to their respective hormone response elements, as recently summarized (41,42). Hydroxyflutamide therefore belongs to a new class of "pure" antagonists that lack agonist activity due to their inability to promote specific DNA binding.
AR dimerization and the inhibitory role played by the NHzterminal domain was revealed using full-length and deletion baculovirus expressed mutants in the DNA mobility shift assay. AR1-660, that contains the NH2-terminal and DNAbinding domains, and AR507-919, the DNA-and steroidbinding domains, showed strong hormone-dependent dimerization. Thus, a major dimerization domain may reside within their overlapping regions in the DNA binding and hinge regions. Alternatively, dimerization may involve distinct regions in each of the dimer components. The DNA-binding domain of the glucocorticoid receptor contains a dimerization region (10, 11) within a so-called D box of the second zinc finger (43,44). Crystal structure analysis of the glucocorticoid receptor DNA-binding domain was linked directly to DNA binding-induced dimerization (45). A major dimerization domain of the estrogen receptor occurs within the steroidbinding domain (46), and the ligand-binding domain was proposed as a site for a major dimer interface that interacts with the same region in the dimer component (44). Such a mechanism does not comply with our results with AR. A dimerization region within the steroid-binding domain must necessarily interact with a region outside the steroid-binding domain in the other dimer component.
The intensity of DNA-receptor complex formation between AR1-660 and AR507-919 deletion mutants exceeded that observed with AR1-660 and the full-length receptor, both of which contain the NHz-terminal region. The results suggest that the AR NHz-terminal domain within the dimer weakens dimer strength. The results raise the possibility that efficient dimerization and DNA binding might occur when AR heterodimerizes with another protein. In this regard, it was recently reported that a truncated form of AR expressed in Escherichia coli complexes with an AR accessory factor that potentiates DNA binding (47).
After short incubation with androgen (1-3 h), AR displayed relatively weak DNA binding that was potentiated by antibody. AR antibody did not enhance DNA binding using extracts of cells untreated with androgen, even though the extraction buffer contained high salt. Furthermore, more prolonged androgen exposure of Sf9 cells (48 h) caused AR to bind DNA strongly independent of antibody. It was noted that addition of androgen to the in vitro DNA binding reaction without prior exposure of Sf3 cells was ineffective in promoting DNA binding, even though specific high affinity androgen binding was measured. These results, taken together, indicate that intracellular changes occur in AR in response to androgen that enable the receptor to bind DNA. The molecular basis for this sequential, time-dependent acquisition of DNA binding remains to be determined. Post-translational modification of AR such as phosphorylation occurs in Sf9 insect cells and could theoretically have a role in the induction of strong DNA binding. Antibody may stabilize receptor dimerization or a conformation conducive to DNA binding.
The results indicate that AR activation in Sf9 cells requires hormone-and time-dependent intracellular processes. Certain pure antiandrogens lack agonist activity because they fail to promote dimerization and DNA binding. Failure to promote DNA binding likely results from an altered conformation imposed on the ligand-binding domain. The AR NH2-terminal domain blocks DNA binding in the absence of the appropriate ligand by interfering with receptor dimerization.