Interaction of Androgen Receptors with Androgen Response Element in Intact Cells

Promoter interference assay was employed to examine in intact cells the roles of the functional domains of androgen receptor (AR) and the ligand for specific DNA interactions using a cytomegalovirus-(androgen response element)-chloramphenicol acetyltransferase reporter (pCMV-ARE2-CAT). Native rat and human ARs interfered with pCMV-ARE2-CAT expression in a hormone-dependent fashion. Low steroid-independent interference seemed to occur because of the ligand binding domain (LBD), which was transcriptionally inhibitory also in a heterologous context. AR devoid of LBD (rARΔ641–902) decreased pCMV-ARE2-CAT activity by 50%. The rARΔ46–408 mutant devoid of the NH2-terminal transcription activation region exhibited ligand-dependent promoter interference of a similar magnitude. Ligand and DNA binding-deficient mutants (hARM807R and rARC562G, respectively) did not influence pCMV-ARE2-CAT expression, although hARM807R binds to ARE in vitro. Non-steroidal anti-androgens casodex and hydroxyflutamide antagonized agonist-dependent promoter interference, whereas cyproterone acetate, RU 56187, RU 57073, and RU 59063 were partial agonists/antagonists. Collectively, interaction of ARs with ARE in intact cells does not require the presence of the COOH-terminal or NH2-terminal domain and/or their interaction. In the context of native AR, however, the androgen-induced conformational change in LBD is mandatory for generation of a transcriptionally competent receptor that binds to DNA in intact cells.

Androgen receptor (AR) 1 belongs to the nuclear receptor superfamily of ligand-regulated transcription factors (1)(2)(3)(4). Although the interaction of AR with specific hormone-responsive DNA elements is usually required for androgen-dependent transcriptional activation, binding of AR to specific DNA motifs is not always necessary for the receptor's ability to downregulate gene expression (5)(6)(7). With regard to the ligand requirement for the recognition of specific DNA elements by AR, in vitro electrophoretic mobility shift assays have yielded conflicting results. AR protein expressed in reticulocyte lysate or produced insect cells is capable of binding to specific androgen response elements (AREs) in vitro even in the absence of androgen (8,9); however, the ligand requirement for the binding of hAR to AREs in vitro has also been reported (10). Very limited information is available on ARE occupancy by the receptor protein in intact cells. In vivo footprinting of an androgendependent enhancer of the mouse slp gene failed to reveal clear protection of hormone response elements by AR (11). While this work was in progress, Kuil and Mulder (12) reported that AR interaction with an "idealized" consensus ARE derived from polymerase chain reaction selection experiments is ligand-dependent in cultured cells.
The promoter interference assay developed by Hu and Davidson (13) is based on competition of DNA-binding proteins, such as nuclear receptors, for binding with essential transcription factors driving a constitutively active heterologous promoter. The interference is achieved by inserting an attachment sequence between the TATA box and the start site of transcription in the promoter. A promoter interference assay has been employed in a few reports to examine the interaction of estrogen receptor (ER) with its cognate response elements in mammalian and frog cells (14,15) and that of AR with AREs in Chinese hamster ovary cells (12). Because of the use of transiently transfected cells, the physiological chromatin environment is not achieved under these conditions. However, using a pCMV-PRE 2 -CAT construct stably integrated into T47D breast cancer cells, Gass et al. (16) recently reported that progesterone receptor (PR) is able to interfere with expression of this reporter also in the context of stable chromatin conformation, albeit less well than in transiently transfected cells.
In our previous studies, we constructed several NH 2 -terminal deletion mutants of AR to define domains crucial for the activation functions of this protein (5,8,9). One of the mutants, rAR⌬46 -408, devoid of a region mandatory for transcriptional activation (8), bound to ARE in vitro with somewhat lower affinity than the wild-type AR (9) but behaved as a dominant negative regulator of the native protein, possibly through forming transcriptionally inactive heterodimers with the latter (8,9). The ability of rAR⌬46 -408, or comparable mutants of the glucocorticoid receptor subfamily of nuclear receptors, to interact with specific DNA sequences in intact cells has, however, not been addressed previously. The COOH-terminal ligand binding domain (LBD) provides the means to regulate the function of native AR in a steroid-dependent fashion. Removal of this region generates a constitutively active receptor form (17,18), whose activity is dependent on the promoter context. 2 However, in vivo DNA binding characteristics of LBD-deficient AR forms or those of other members of the glucocorticoid receptor subfamily have not been examined.
In the present work we have used promoter interference assays to investigate whether binding of AR to specific DNA sequences is ligand-dependent in intact CV-1 cells. We have also examined which functional domains of the wild-type receptor, in addition to the DNA binding domain (DBD), are mandatory for specific DNA binding in vivo. To assess the role of the ligand, several non-steroidal and steroidal anti-androgens alone or together with an androgen agonist were used. Finally, immunocytochemical analysis of cultured cells was employed to rule out the possibility that some of the results were compromised by altered subcellular distribution of receptor proteins.
Immunocytochemistry-CV-1 cells were seeded on glass slips on 10-cm plastic plates and were transfected with 1 or 2.5 g of pSG5r/hAR DNA, 2 g of pCMV-ARE 2 -CAT, 3 g of pCH110 and filled to 10 g of total DNA with empty pSG5. Replacement of medium and exposure to steroids were conducted as in the promoter interference assay. Cells were fixed in methanol at Ϫ20°C for 10 min and washed three times with phosphate-buffered saline (150 mM NaCl, 20 mM Na 2 HPO 4 , pH 7.3) at 22°C for 10 min. The primary antibody K183 (diluted 1:200) is a polyclonal rabbit antiserum raised against full-length rAR purified to homogeneity using preparative polyacrylamide gel electrophoresis under denaturing conditions from Sf9 insect cells infected with a recombinant baculovirus encoding rAR (25). In short, the purification procedure was as follows. Two ϫ 10 7 cells were lysed in 2 ml of electrophoresis sample buffer (26), passed five times through a 25gauge needle, and soluble proteins were separated using Bio-Rad model 491 Prep Cell apparatus with a 37-mm (inner diameter) tube. Electrophoresis was carried out at 50 mA, proteins were eluted with 50 mM Tris/HCl, 5 mM ␤-mercaptoethanol, 0.02% (v/v) Nonidet P-40 (pH 8.0) at 0.45 ml/min, and 4.5-ml fractions were collected. Fractions containing rAR were pooled and concentrated using Amicon P-10 membrane (Amicon, Beverly, MA). Rabbits were injected with 50 g of rAR protein at each immunization. Diluted K183 antiserum (1:200) was incubated on cells for 1 h at 22°C, and fluorescein isothiocyanate-conjugated goat anti-rabbit secondary antibody (Jackson ImmunoResearch Laboratories, Philadelphia) was used (1:200 dilution, 30 min at 22°C) to detect the antigen.
Immunoblotting-Conditions for immunoblotting were as described previously (8), except that antiserum K183 (diluted 1:1,500) was used as the primary antibody.

Constructs for Promoter
Interference Assays-An AR-dependent promoter interference reporter plasmid, pCMV-ARE 2 -CAT, was constructed in a fashion similar to that described for ER (14) by inserting a duplicated high affinity ARE sequence between the TATA box and transcription start site of pCMV-0-CAT (Fig. 1A). Two AREs were inserted, as the presence of more than one palindromic response element facilitates AR-ARE interactions in vitro (9). pCMV-0-CAT devoid of ARE sequences was initially used as the control reporter, but its basal activity in CV-1 cells was several times higher than that of pCMV-ARE 2 -CAT. This result was in agreement with a previous report (13) that insertion of sequences into a similar location (between the TATA box and the start site of transcription) decreases CAT expression. For this reason, we used pCMV-ERE 2 -CAT reporter (Fig. 1A) as control to assess the specificity of AR-ARE interaction in intact cells. CV-1 cells were cotransfected with various AR expression plasmids (Fig.  1B) and pCMV-ARE 2 -CAT (or pCMV-ERE 2 -CAT) reporter driven by the constitutively active CMV promoter.
Specific Promoter Interference by the Native AR Protein Requires Intact Ligand and DNA Binding Domains-Native rAR and hAR interfered with pCMV-ARE 2 -CAT expression in such a way that both ligand-dependent and ligand-independent inhibition of the reporter activity was seen (Fig. 2, A and B). rAR showed somewhat higher ligand-independent promoter-interfering activity than hAR. Since expression levels of rAR and hAR proteins were not compared in these experiments, the reason for this species difference remains to be elucidated. Both ARs also elicited some inhibition of pCMV-ERE 2 -CAT expression which was, however, not increased by androgen (Fig. 2, C  and D). Even though the weak ligand-independent activity of AR was not specific for the hormone response element inserted in the promoter interference construct, it was specific for the promoter interference assay itself, as a similar androgen-independent activity was not observed in transactivation experiments. It is also worth pointing out that maximal promoter interference and transcriptional activation were achieved with comparable amounts of rAR or hAR expression plasmid (1 g DNA/plate). 3 However, ligand-occupied hAR was twice as active as rAR in these latter experiments (110-fold versus 50-fold increase over control, respectively 3 ), which is in contrast to the behavior of apo-hAR and apo-rAR in promoter interference assays (see Fig. 2, A and B).
AR forms incapable of DNA or hormone binding (rARC562G and hARM807R, respectively) were used to assess the importance of these regions for specific AR-ARE interaction in vivo. rARC562G with a Cys 3 Gly substitution at codon 562 does not bind to DNA in vitro and is transcriptionally inactive (22). hARM807R is incapable of hormone binding because of a Met 3 Arg substitution at codon 807 (21). In electrophoretic mobility shift assays under cell-free conditions, hARM807R bound to ARE with approximately the same affinity as the wild-type receptor in the absence of androgen (21), but exposure to androgen did not elicit a conformational change in the LBD of hARM807R in a fashion similar to that in the wild-type LBD (21,27). rARC562G and hARM807R mutants did not interfere with pCMV-ARE 2 -CAT activity in a steroid-dependent fashion, but they exhibited weak hormone-independent interference, essentially indistinguishable from that of the wild-type apo-AR (Fig.  3, A and B). The fact that rARC562G behaved like apo-AR both in the presence and absence of testosterone implies that specific binding of the wild-type AR to DNA in intact cells is totally dependent on the integrity of the first zinc finger and that the conformational change elicited by an active androgen in the LBD does not abolish the ability of this region to interfere weakly with pCMV-ARE 2 -CAT expression.

Influence of COOH-terminal and NH 2 -terminal Regions of AR on Receptor-ARE Interaction in Intact Cells-Previous
studies have shown that AR forms devoid of the entire LBD behave as constitutively active transactivators and are capable of binding to specific DNA elements in vitro (17,18). 2 The rAR⌬641-902 mutant that lacks the LBD reduced expression of pCMV-ARE 2 -CAT by 50% but did not influence that of pCMV-ERE 2 -CAT (Fig. 4A). It is of note that maximal promoter interference was achieved with the lowest amount of pSGrAR⌬641-902 expression vector used (0.5 g of DNA/ 10-cm plate), and no additional increase in the interference occurred when the amount of plasmid was augmented 5-fold (Fig. 4A) or 10-fold (data not shown). Interaction of rAR⌬641-902 with DNA was, as expected, independent of the presence of ligand, whether an agonist (Fig. 4A) or an antagonist (54 Ϯ 5% of control in cells transfected with 0.5 g of pSGrAR⌬641-902 in the presence of 1 M casodex). To examine the role of the NH 2 -terminal region, rAR⌬46 -408 and rAR⌬46 -408/C562G mutants were used. rAR⌬46 -408 expressed in insect cells binds to specific DNA sequences in vitro, albeit with a somewhat lower affinity than the wild-type AR (8,9). When expressed in CV-1 cells, rAR⌬46 -408 elicited promoter interference in an androgen-dependent fashion (Fig.  4B). This interference required the presence of ARE sequences, as a similar steroid-dependent inhibition of pCMV-ERE 2 -CAT expression failed to occur under the same conditions (Fig. 4B). Likewise, an intact DBD of the receptor was required, as judged by the inability of rAR⌬46 -408/C562G to interfere with pCMV-ARE 2 -CAT activity. 2 Some hormone-independent inhibition of pCMV-ARE 2 -CAT activity was observed with rAR⌬46 -408 when higher amounts of expression plasmid were used in transfections (Fig. 4B).
Taken together, interaction of AR with specific DNA sequences in intact cells can take place in the absence of LBD or most of the NH 2 -terminal region. DNA binding of the receptor, even in the presence of LBD and steroid, is not sufficient to elicit proper physiological responses, as the rAR⌬46 -408 mutant is transcriptionally inactive (8). Comparison of the doseresponse curves between rAR⌬641-902 and rAR⌬46 -408 suggests that LBD is responsible for the weak hormoneindependent interference of AR with pCMV-ARE 2 -CAT or pCMV-ERE 2 -CAT expression (Fig. 4, A and B). The results with full-length AR forms (Fig. 2) were in general agreement with those obtained with rAR⌬641-902 and rAR⌬46 -408 mutants.
Repression of Transcription at a Distance by LBD in a Heterologous System-The ability of LBD to interfere with transcription was examined by additional experiments, in which pG5SV40CAT reporter containing five GAL4 binding sites upstream of the SV40 promoter was used. CV-1 cells were cotransfected using this reporter or an appropriate control reporter (pSV40CAT) and expression vectors encoding GAL4-DBD, GAL4-LBD, or GAL4-Nterm fusion proteins. The NH 2terminal region of AR (residues 5-538) fused to GAL4-DBD did not influence the activity of the latter; however, when the fusion partner was LBD (residues 641-902), repression of transcriptional activity took place (Fig. 5A). It is of note that the ability of ligand-free LBD to repress transcription at a distance exceeded that of the ligand-occupied LBD in the GAL4-DNAbinding construct (Fig. 5A). The repressive action of LBDcontaining fusion protein was dependent on its tethering to DNA, as no inhibition was detected in the absence of GAL4binding elements (Fig. 5B).
Antagonistic and Agonistic Activities of Anti-androgens-Several anti-androgens (casodex, hydroxyflutamide, RU 56187, RU 57073, RU 59063, and cyproterone acetate) were examined by promoter interference assays using rAR⌬46 -408, as differentiation between ligand-dependent and -independent activities was more clear-cut with this mutant than with native ARs. All other anti-androgens examined but hydroxyflutamide were able to confer some promoter-interfering activity upon rAR⌬46 -408, with cyproterone acetate and the RU compounds being as potent as an agonist (10 nM testosterone) at the concentration used (1 M) (Table I). Additional studies with casodex revealed that its low agonist-like activity in the promoter interference assay was also detectable with wild-type rAR and at varying receptor levels (Fig. 6A). When present concomitantly with androgen, casodex inhibited the agonist-induced promoter interference significantly but not completely (Fig.   FIG. 3. Binding of rARC562G and hARM807R  6B). Previous studies have shown that casodex and hydroxyflutamide are completely devoid of inherent transactivation ability (22,28), whereas cyproterone acetate is a partial agonist/antagonist (29).
Three N-substituted arylthiohydantoin anti-androgens (RU 56187, RU 7073, and RU 59063) were as effective as testosterone in conferring promoter-interfering activity on rAR⌬46 -408 (Table I). This is in agreement with binding affinities of these compounds for rat and human ARs in vitro, which are 92-300% of that of testosterone (30). When CV-1 cells cotransfected with pSGrAR and pMMTV-CAT vectors were exposed to 1 M of each RU anti-androgen alone, CAT activity was 50% with RU 56187, 28% with RU 57073, and 35% with RU 59063 of that with 10 nM testosterone. Concomitant exposure of the cells to 10 nM testosterone and 1 M RU antiandrogens resulted in reporter gene activities that were 39 -66% of those achieved with androgen alone, indicating that the three RU compounds were partial agonists/antagonists under these conditions. As the RU compounds promoted AR-ARE interaction in CV-1 cells as well as testosterone (Table I), conformational requirements on LBD for transcriptional activation and recognition of specific DNA sequences (promoter interference) are not identical in intact cells.
Localization of Wild-type and Mutant ARs by Immunocytochemistry in Transfected Cells-The polyclonal antiserum (K183) used in these experiments was specific for AR proteins, as no extra bands were observed in mock-transfected cells, and only the receptor protein was detected in immunoblots of COS-1 cells transfected with some of the AR expression constructs used in promoter interference assays (Fig. 7). The antiserum recognized rAR⌬46 -408 and rAR⌬641-902 mutants as efficiently as the native receptor protein and, therefore, was useful for immunological studies of different AR forms in cultured cells.
CV-1 cells transfected with 1 g of pSGrAR DNA showed some immunoreactivity in the cytoplasm in the absence of androgen, whereas unliganded hAR was more clearly nuclear in its localization under the same conditions (Fig. 8, A and C). Both receptors became solely nuclear upon hormone exposure (Fig. 8, B and D). rARC562G and hARM807R exhibited both cytoplasmic and nuclear localization in the absence of testosterone (Fig. 8, E and F). The AR form devoid of most of the NH 2 -terminal region (rAR⌬46 -408) showed mainly nuclear localization even in the absence of androgen (Fig. 8G), and it was strictly nuclear after exposure to testosterone (data not shown). Likewise, the LBD-deficient mutant (rAR⌬641-902) was nuclear in the absence of the hormone (Fig. 8H). CV-1 cells transfected with pSGhAR expression vector and exposed to casodex or hydroxyflutamide exhibited only nuclear localization of the receptor protein (Fig. 8, I and J). Taken together, immunocytochemical studies of transfected CV-1 cells expressing wild-type and mutant AR proteins revealed that the receptor forms were mainly nuclear even in the absence of ligand, indicating that differences in subcellular localization do not explain their dissimilar behavior in the promoter interference assay.

DISCUSSION
The results of this work show that sequence-specific DNA binding of the native AR in cultured CV-1 cells requires the presence of ligand. In this respect, our data agree with a recent report by Kuil and Mulder (12), who also used the promoter interference assay to examine the interaction of AR with a synthetic consensus ARE in Chinese hamster ovary cells. Our data show further that neither LBD nor a major part of the NH 2 -terminal transactivation region of the receptor is needed for its interaction with specific DNA sequences in vivo. LBD is not only dispensable but also appears to be responsible for the ability of apo-AR to interfere weakly with pCMV-ARE 2 -CAT expression. In addition, unoccupied LBD fused to GAL4-DBD was capable of repressing transcriptional activity of a heterologous SV40 promoter at a distance. Should AR bind to AREs as a homodimer in intact cells, as it does in vitro (8,9), the results with rAR⌬641-902 illustrate that this dimerization utilizes sequences other than those residing in LBD. Moreover, the interaction between COOH-and NH 2 -terminal regions of AR, which has been shown to occur in mammalian cells (31), 2 is not mandatory for the ability of the receptor to recognize specific  confer ARE binding ability on AR as determined by promoter interference assay CV-1 cells were transfected with 1 g of pSGrAR⌬46 -408, 2 g of pCMV-ARE 2 -CAT, 3 g of pCH110 (␤-galactosidase), and 4 g of pSG5 vector DNA. Eighteen h after transfection, the cells received fresh medium with 10 nM testosterone or 1 M anti-androgen and were collected 30 h later for the measurement of CAT and ␤-galactosidase activities. The values (mean Ϯ S.E., n ϭ 3) are expressed relative to that in CV-1 cells transfected without pSGrAR⌬46 -408. DNA elements in vivo, as shown by appropriate deletion mutants in promoter interference assay. Studies with rAR⌬46 -408 that lacks most of the NH 2 -terminal region and with antiandrogens also provided compelling evidence that mere DNA binding in vivo is not sufficient for the AR to become a competent transcriptional activator.
The hARM807R mutant is incapable of hormone binding, totally inactive in transactivation assays, and the reason for complete androgen insensitivity of a patient (21). It could be speculated that inappropriate folding of this mutant's LBD may render it free of intracellular chaperones such as heat shock proteins (32), thereby permitting its binding to AREs even in vivo. However, our present data and those of Kuil and Mulder (12) demonstrate that specific DNA binding in living cells is not achieved without a ligand-induced allosteric change in the LBD. The findings that hARM807R binds well to AREs in vitro (21) and is localized in nuclei of transfected cells (this study) are interpreted to mean that the steroid-induced change in LBD conformation is primarily required for the release of AR from associated inhibitory protein complexes, rather than for generation of a receptor form capable of ARE recognition.
The mutant rAR⌬46 -408, which has a region mandatory for transcriptional activation deleted (8), recognized specific DNA sequences in CV-1 cells in a fashion indistinguishable from that of the wild-type receptor, implying that the NH 2 terminus is not essential for this interaction. This result is in agreement with our previous studies on AR-ARE interaction in vitro, in that NH 2 -terminal internal deletions compromised the affinity of rAR for AREs only to a moderate extent; the K D of interaction increased from 0.5-1.0 nM to 3.0 nM (8,9). AR protein produced in insect cells has been suggested to require intracellular hormone exposure to overcome the inhibition imposed by the NH 2 -terminal domain on dimerization and DNA binding of the aporeceptor (10). The present experiments showed, however, that unliganded rAR⌬46 -408 was incapable of sequencespecific DNA binding in CV-1 cells (Fig. 4B), implying that a major part of the NH 2 -terminal region is not critical for maintaining the receptor in a non-DNA-binding form.
Anti-androgens that were examined in this work can be divided into two groups on basis of their ability to modulate the binding of AR to AREs. The first group is not capable of converting the receptor to a form that binds to specific DNA sequences in CV-1 cells, or it does this poorly. When present concomitantly with an agonist, these compounds inhibit the agonist-dependent promoter interference. Non-steroidal antiandrogens casodex and hydroxyflutamide belong to this group. The second group includes the steroidal anti-androgen cyproterone acetate and three N-substituted arylthiohydantoin antiandrogens (RU 56187, RU 57073 and RU 59063). These compounds, at least at the concentrations used, converted AR to a form capable of ARE binding in CV-1 cells. Moreover, the compounds of the second group behaved as partial agonists/ antagonists in transactivation experiments. Kuil and Mulder (12) have also reported that in Chinese hamster ovary cells, casodex and hydroxyflutamide inhibited DNA binding of AR, but cyproterone acetate failed to do so.
Ligand-free human and rat ARs exhibited weak promoterinterfering activity. In this respect, our results on AR differed form those reported by Reese and Katzenellenbogen (14) and Xing and Shapiro (15) on human and frog ERs, respectively. In these latter cases, unoccupied wild-type ER showed moderate to high interfering activity, and some ligand-free ER mutants were even more potent than the estrogen-bound wild-type ER (14,15). Likewise, apo-PR elicited strong interference with pCMV-PRE-CAT expression, which was almost 50% of that of the progestin-bound PR (16). The observed differences in the behavior of nuclear receptors in promoter interference assays could result, at least in part, from dissimilar affinities of their LBDs for auxiliary proteins communicating with transcription machinery. A COOH-terminal region of PR has recently been reported to contain a transcriptional repressor domain that functions through a putative corepressor (33); however, a similar function has not so far been described for the COOH terminus of AR. The use of a heterologous GAL4 system supported the notion that LBD is responsible for the weak ligandindependent promoter-interfering activity of AR. The activity of this LBD in the context of GAL4 fusion protein was not dependent on the ligand-induced conformational change; rather, ligand-free GAL4-LBD bound to elements upstream of SV40 promoter exhibited stronger repressive action than the ligand-occupied form, suggesting that apo-LBD is the favored FIG. 8. Immunocytochemical localization of wild-type and mutated AR proteins in transfected CV-1 cells. The cells were plated on glass coverslips on 10-cm cell culture dishes, transfected, and treated as described under "Experimental Procedures." Detection by immunofluorescence was performed using K183 antiserum (dilution 1:200) as the primary antibody and fluorescein isothiocyanate-conjugated goat anti-rabbit secondary antibody. The panels correspond to the following receptor forms and treatments: rAR in the absence (A) and presence of 10 nM testosterone (B); hAR without (C) and with 10 nM testosterone (D); rARC562G (E), hARM807R (F), rAR⌬46 -408 (G), and rAR⌬641-902 (H) without androgen; and hAR in the presence of 1 M casodex (I) or 1 M hydroxyflutamide (J). Original magnification: A-H, ϫ 400; I and J, ϫ 630. conformation in recruitment of putative repressor proteins.
In conclusion, the results of this work indicate that interaction of AR with specific DNA sequences in CV-1 cells does not involve COOH-terminal or NH 2 -terminal regions of the receptor, even though the NH 2 -terminal domain contains sequences required for transcriptional activation. In the context of wildtype AR, the androgen-induced conformational change in the COOH terminus is, however, mandatory for generation of a transcriptionally competent receptor protein that binds to ARE sequences in intact cells, even though mere DNA binding in vivo is not sufficient for the AR to acquire competence in transcriptional activation. And finally, LBD appears to be responsible for the weak promoter-interfering activity of apo-AR.