Repressors of Androgen and Progesterone Receptor Action*

Androgen and progesterone receptors (AR and PR) are two determining factors in gonadal differentiation that are highly expressed in developing and mature go-nads. Loss of AR results in XY sex reversal and muta-tions causing reduced AR activity lead to varying de-grees of defects in masculinization. Female PR knockout mice are infertile due to ovarian defects. While much has been discovered about positive regulation of these receptors by coactivators little is known about repression of the transcriptional activity of AR and PR in the presence of agonists. In this study we assessed the effect of SMRT and DAX-1 on AR and PR activity in the presence of both agonists and partial antagonists. We show that SMRT and DAX-1 repress agonist-dependent activity of both receptors, and the mechanism of repression includes disruption of the receptor dimer interactions rather than recruitment of histone deacetylases. We demonstrate that endogenous agonist-bound PR and DAX-1 in T47D breast cancer cells and endogenous AR and DAX-1 in LNCaP prostate cancer cells can be coimmunoprecipitated suggesting that the interaction is physiological. Surprisingly, although DAX-1 represses partial antagonist activity of AR, it was ineffective in repressing partial antagonist induced activity of PR. In contrast to most reported repressors, the expression of DAX-1 is restricted. We found that although DAX-1 is expressed in normal human prostate, its expression is strongly reduced in benign prostatic hyperplasia suggesting that DAX-1 plays a role

Androgen and progesterone receptors (AR and PR) are two determining factors in gonadal differentiation that are highly expressed in developing and mature gonads. Loss of AR results in XY sex reversal and mutations causing reduced AR activity lead to varying degrees of defects in masculinization. Female PR knockout mice are infertile due to ovarian defects. While much has been discovered about positive regulation of these receptors by coactivators little is known about repression of the transcriptional activity of AR and PR in the presence of agonists. In this study we assessed the effect of SMRT and DAX-1 on AR and PR activity in the presence of both agonists and partial antagonists. We show that SMRT and DAX-1 repress agonist-dependent activity of both receptors, and the mechanism of repression includes disruption of the receptor dimer interactions rather than recruitment of histone deacetylases. We demonstrate that endogenous agonist-bound PR and DAX-1 in T47D breast cancer cells and endogenous AR and DAX-1 in LNCaP prostate cancer cells can be coimmunoprecipitated suggesting that the interaction is physiological. Surprisingly, although DAX-1 represses partial antagonist activity of AR, it was ineffective in repressing partial antagonist induced activity of PR. In contrast to most reported repressors, the expression of DAX-1 is restricted. We found that although DAX-1 is expressed in normal human prostate, its expression is strongly reduced in benign prostatic hyperplasia suggesting that DAX-1 plays a role in limiting AR activity in prostate.
Nuclear receptors are regulated both by coactivators and by corepressors. Although steroid receptor coactivators have been studied extensively, less is known about corepressors of agonist activated steroid receptors. Androgens, acting through AR, 1 play a role in both benign prostatic hyperplasia (BPH) and prostate cancer. In both cases, reduction in AR activity is an important component of treatment, although fully effective treatments are not yet available. Two well characterized corepressors of thyroid receptor (TR) activity, NCoR (nuclear receptor co-repressor) and SMRT (silencing mediator for retinoid acid receptor (RAR) and TR), have been identified previously (1,2). They appear to work through binding to TR aporeceptor, recruiting complexes containing histone deacetylases (HDACs); they dissociate from the receptor upon agonist binding allowing coactivator complexes to form (3). In the case of estrogen receptor (ER), these repressors interact with the antagonist bound receptor through a region that largely overlaps with the coactivator binding interface, and recruit HDACs (4). An orphan nuclear receptor DAX-1 (dosage sex reversal, adrenal hypoplasia congenita critical region on the X chromosome, gene 1) has been reported to inhibit steroidogenic factor 1 (SF-1) and ER activity (5,6). DAX-1 is an atypical nuclear receptor containing an ssDNA/RNA binding domain in its N terminus and a multihelical C-terminal domain, a putative ligand binding domain (LBD) (7). DAX-1 is an X chromosomal gene, amplification of which results in XY sex reversal, a phenotype similar to AR inactivation (8). Opposing effects of the gene dose of DAX-1 and AR as well as a similar pattern of expression suggested that DAX-1 might be an AR repressor. AR shares many common features with other steroid receptors but differs in some aspects. First, whereas ER and PR form parallel homodimers through interactions within their LBD (9,10), AR forms antiparallel dimers that are stabilized by interaction between the N and C termini (11). Second, while most steroid receptors such as TR or ER rely strongly on their activation function-2 (AF-2) in the receptor hormone binding domain for recruitment of coactivators (12), the N-terminal AF-1 function of AR is more important (13,14). We have studied the functional and physical interactions of corepressors with AR and PR. We found that SMRT and DAX-1 were effective repressors of agonist-dependent activity of AR and PR, but DAX-1 did not repress the activity of the vitamin D receptor (VDR), a class II nuclear receptor. Moreover, endogenous agonist-bound PR and AR each interact with DAX-1. To better understand the mechanism of repression by candidate corepressors we compared their effects on AR and PR. In this study we show that both AR and PR are repressed by DAX-1 in the presence of their respective agonists, but exhibit opposite coregulation patterns when bound to their common partial antagonist RU486. In contrast, SMRT represses both AR and PR in the presence of either agonist or antagonist. During the preparation of this article, Holter et al. (15) reported that DAX-1 represses AR activity and suggested that DAX-1 can cause mislocalization of AR to the cytoplasm in the presence of agonist. However, our studies using a promoter interference assay, suggest that DAX-1 does not disrupt the interaction of either AR or PR with DNA. We find that the DAX-1 mechanism of action includes interference with N/C-terminal interactions in AR and hormone binding domain interactions in PR.

EXPERIMENTAL PROCEDURES
Constructs and Supplies-GRE 2 -E1b-Luc reporter was provided by Dr. Carolyn Smith, Baylor College of Medicine (16). This reporter contains 2 AR response elements inserted in front of the coding region for luciferase in the pBL3 basic vector (Promega, Madison, WI). VDRE Luc was obtained from Dr. Elizabeth Allegretto, Ligand Pharmaceuticals (17). Full-length DAX-1 cloned in pSG5 was obtained from Dr. Eckardt Treuter, Karolinska Institutet, Sweden (18), and subsequently recloned into a mammalian expression vector pCR3.1 (Invitrogen, Carlsbad, CA). The N terminus of DAX-1 consisting of amino acids 1-260 and the C terminus of DAX-1 consisting of amino acids 236 -470 with an alanine introduced before amino acid 236 were cloned into the pCR3.  (23). Trichostatin A was obtained from Sigma Aldrich, mifepristone (RU486) from Siniwest Holdings (San Diego, CA), Methyltrienolone (R1881) and Promegestone (R5020) from Perkin-Elmer Life Sciences (Boston, MA). Monkey kidney COS-1 cells and HeLa (ATCC, Manassas, VA) were maintained in Dulbecco's modified Eagle's medium in the presence of 5% fetal calf serum (Invitrogen).
Transient Cotransfection Experiments-Transient cotransfections were performed essentially as was described previously (24). Briefly, HeLa or COS-1 cells were plated the day before transfection at 50 -60% confluency in DME media (Invitrogen) supplemented with 5% charcoalstripped fetal bovine serum. The next day, the indicated amounts of DNA were transfected using polylysine-coupled adenovirus for 2 h in serum-free Dulbecco's modified Eagle's medium (Invitrogen). An equal volume of 10% charcoal-stripped fetal bovine serum was added to give a final concentration of 5% and cells were then treated with the indicated ligand or not for 24 h. Cells were harvested and assayed for reporter activity or used for Western blot analysis. For activity assays each data point was done in triplicate and the average and standard deviation calculated. Each experiment was performed a minimum of three times and a representative example is shown.
Immunoprecipitations-LNCaP (for AR interactions) or T47D (for PR interactions) cells were grown to 70% confluence on 10 cm dishes. Cells were treated for 1 h with the indicated hormone, harvested and protein extracted by 3 freeze/thaw cycles in TESH buffer (10 mM TRIS, 1 mM EDTA, 12 mM thioglycerol, pH 7.7) supplemented with 0.3 M NaCl and protease inhibitors (25). Protein extracts were spun at 100,000 rpm in a TLA 100.3 rotor (540,000 ϫ g) for 10 min at 4°C. Fat was aspirated and the insoluble material discarded. Clarified protein lysate was dialyzed against TESH buffer for 2 h at 4°C, centrifuged at 100,000 rpm for 10 min at 4°C, and precleared (incubated) with 70 l of a 1:1 protein A-Sepharose suspension in TESH for 30 min at 4°C. Sepharose beads were spun and discarded. For SMRT interaction, protein A-Sepharose beads were incubated with either 3 g of SMRT-1542 monoclonal an-tibody (Gene Tex, San Antonio, TX) and 5 g of rabbit anti-mouse IgG (RAM) (Zymed Laboratories Inc., San Francisco, CA) or RAM alone, and 1-4 mg of precleared protein extract overnight at 4°C. For DAX-1 interaction, protein A-Sepharose beads were incubated with 3 g of DAX-1 K17 rabbit polyclonal antibody or 3 g of RAM (Santa Cruz Biotechnology), and 1-4 mg of precleared protein extract overnight at 4°C. Beads were collected by brief centrifugation in an Eppendorf centrifuge and were washed three times with TESH/0.1 M NaCl. The samples were transferred to fresh tubes prior to the last centrifugation, centrifuged, and the beads eluted with 2ϫ Laemmli SDS sample buffer. Eluates were resolved by electrophoresis on 6.5% SDS-PAGE gels, transferred to nitrocellulose, and proteins detected by Western blot analysis.
In Vitro Transcription/Translation and in Vitro Interaction-DAX-1 was translated in vitro using TNT Quick Coupled Transcription/Translation System (Promega, Madison, WI) and [ 35 S]methionine (Amersham Biosciences) exactly as described in the manufacturer's protocol. 75 l of protein A-Sepharose slurry (1:1) and 100 l of TESH/0.1 M NaCl were incubated with 3 g of either DAX-1 or rabbit antimouse IgG for 1 h on the rocker at 4°C. To measure binding of PR to DAX-1, beads were washed twice with TESH/0.1 M NaCl, and 50 l of DAX-1, 100 l of TESH/0.1 M NaCl, and 1 g of purified baculovirus-expressed polyhistidine-tagged PR (25) were added. The mix was incubated for 2 h on a rocker at 4°C, washed twice and proteins eluted. Protein eluates were then resolved on 6.5% SDS-PAGE and analyzed for PR by Western blotting as described below. To measure DAX-1 binding to PR, 100 l of a 1:1 protein G-Sepharose (Amersham Biosciences) slurry in TESH and 100 l of TESH/0.1 M NaCl were incubated with 5 g of PR6 antibody (a mouse monoclonal antibody that recognizes the B specific region of PR) (26) for 1 h at 4°C. Protein G-Sepharose beads were washed twice with TESH/0.1 M NaCl, and 100 l of TESH/0.1 M NaCl, 50 l of DAX-1 with or without 2 g of purified polyhistidine-tagged PR were added. After a 2-h incubation on a rocker at 4°C, samples were washed once in TESH/0.1 M NaCl, eluted with 2ϫ Laemmli SDS sample buffer, and resolved on SDS-PAGE electrophoresis. DAX-1 coimmunoprecipitated with PR antibody was visualized following transfer to nitrocellulose by exposing the membrane to Kodak X-OMAT AR Film (Fisher, Pittsburgh, PA).
Western Blot Analysis-PR Western blot analysis was performed using the 1294 antibody (generously provided by Dr. Dean Edwards, University of Colorado Health Sciences Center, Denver, CO) as described previously (27). AR Western blots were done with AR 441 antibody as described in Nazareth et al. (24). Actin Western blots were performed as was previously described (28).
Promoter Interference Assay-Cells were transfected with 15 ng of CMV GRE 3 CAT, and 100 ng of GRE 2 -E1b-Luc reporters, the indicated receptor or not, and increasing concentrations of either DAX-1 or SMRT. Hormone (3 nM R1881 for experiments with AR and 10 nM R5020 in experiments with PR) was added 2 h after transfection, cells were harvested 24 h later, and cellular lysates were assayed for CAT and luciferase activity.
Luciferase Assay-Cells were washed with phosphate-buffered saline once and 200 l of lysis buffer (provided by the manufacturer) added to each well in a 6-well plate. After a 15-min incubation, 20 l of the lysate was used to measure luciferase activity as recommended by the manufacturer (Promega).
CAT Assay-CAT assays were carried out as was described previously (29). Briefly, cells were harvested in TEN buffer (Tris, EDTA, NaCl), spun, and the buffer was removed. Cells were then lysed in 0.25 M Tris (pH 7.5), supplemented with 0.4 M NaCl if the lysates were later used for western as well as CAT assay, by freezing and thawing. 10 -20 g of the protein were used to determine CAT activity.
␤-Gal Assay-␤-Galactosidase activity was measured to normalize activity of the receptors. The ␤-galactosidase gene cloned in the same expression vector (pCR 3.1) as PR and AR was used to aid in controlling both for transfection efficiency as well as any effects of treatment on the activity of the expression plasmid promoter. Ten percent of cellular lysates prepared for the luciferase assay was used to perform the ␤-galactosidase activity assay as previously described (30).
Tissue Microarray-Tissue arrays were made in the Baylor Prostate SPORE. Samples were harvested as previously described (31). Tissue arrays containing cores from normal peripheral zone and areas of BPH were analyzed for intensity and labeling frequency. The intensity of the grading scale ranged from no detectable signal (0) to a strong signal seen at low power (3). Two corresponds to a moderate signal seen at low to intermediate power, while one corresponds to a weak signal seen only at intermediate to high power. Labeling frequency was scored as 0 (0%), 1 (1-33%), 2 (34 -66%), or 3 (67-100%).
Immunohistochemistry-Briefly, the slides were deparaffinized, rehydrated, and then heated in 10 mM citrate buffer pH 6.0 for 30 min using a steamer. The slides were blocked with 10% normal rabbit serum for 30 min. After washing in PBS, the slides were incubated with the rabbit anti-DAX-1 polyclonal antibody (1:100) for 1 h at room temperature. Then the secondary biotinylated anti-rabbit IgG was applied for 30 min followed by 30 min of incubation with streptavidin peroxidase (LSAB kit, DAKO). After rinsing, slides were visualized by diaminobenzidine chromogen solution (DAKO) and counterstained with routine hematoxylin.

DAX-1, SMRT, and NCoR Repress Both Agonist-and Partial
Antagonist-dependent AR Activity-To characterize the effects of DAX-1, SMRT, and NCoR on AR activity, we transiently transfected HeLa cells with a constant amount of receptor and reporter expression plasmids and increasing amounts of repressor plasmid balanced with the corresponding vector control. As shown in Fig. 1A, DAX-1 inhibits agonist (R1881)-dependent activity of AR as recently reported (15) as does SMRT. Both DAX-1 and SMRT also repress the activity of AR in the presence of the partial antagonist RU486 (Fig. 1B). Both of these cofactors repressed AR activity in HeLa and COS-1 cells to a similar extent (data not shown). Surprisingly, NCoR repressed AR activity in the presence of RU486 and R1881 but only in a cell-specific context. Under identical transfection conditions AR is repressed by NCoR in COS-1 ( Fig. 1D) but not in HeLa cells (Fig. 1C). To determine if changes in receptor activity are due to changes in receptor levels we performed Western blots using samples from the transrepression experiments. As previously reported (32), addition of agonist increases AR expression through stabilization of the protein. However, expression of repressors did not change levels of AR expression (Fig. 1E).
DAX-1 and SMRT Repress PR Activity-It has been reported previously that SMRT preferentially represses partial antagonist bound PR activity (33). However, under our transfection conditions PR was repressed by SMRT in the presence of either an agonist or a partial antagonist to a similar extent in both HeLa (Fig. 2, A and B) and COS-1 (data not shown) cell lines. The effect of DAX-1 on PR activity has not been reported previously. Interestingly, DAX-1 appeared to be a much stronger repressor of PR than AR in transient cotransfection assays, since much less DAX-1 DNA was necessary to repress the agonist-dependent PR activity although repression by SMRT was similar to AR ( Fig. 2A). While SMRT represses PR antagonist dependent activity as expected, most intriguing was the effect observed with increasing concentrations of DAX-1 on the PR bound to partial antagonists (Fig. 2B). DAX-1 slightly increased the RU486-dependent activity of PR. Note that although DAX-1 potentiates RU486-dependent activity of PR, the maximal activity is still less than that of agonist-bound PR treated with DAX-1. As seen from Fig. 2C, the expression of RU486-bound PR is slightly higher in the presence of increasing concentrations of DAX-1, which may explain the increase in activity of partial antagonist bound PR (compare Fig. 2, B and  A). The opposite effect of DAX-1 on PR in the presence of different ligands suggests that its mechanism of repression may be different from that for AR. In contrast to AR, the expression levels of PR are reduced in response to agonist (34). As in the case of AR, levels of agonist-bound PR are not reduced by the addition of increasing concentrations of DAX-1 and SMRT (Fig. 2C), and therefore this does not contribute to decreasing PR activity.
DAX-1 Does Not Repress VDR and p53 Activity-To determine if DAX-1 shows specificity in repression of transcription factors we asked whether DAX-1 would repress the ubiquitously expressed transcription factor p53, or the activity of the vitamin D receptor (VDR) another member of the nuclear receptor family. As shown in Fig. 3A DAX-1 does not repress p53 activity (Fig. 3A). Interestingly, VDR which lacks an N-terminal domain and thus AF-1 is practically unaffected by increas-ing concentrations of DAX-1 with or without 1,25 D (Fig. 3B). We used protein levels to normalize the activity of VDR and p53, because VDR was endogenously expressed and p53 was in a different vector from that of ␤-Gal. However, as we observed from our previous experiments, neither DAX-1 nor SMRT affected ␤-galactosidase expression at any DNA level tested and therefore either protein concentration or ␤-galactosidase activity can be used to normalize the experiments.
AR and PR Interact with DAX-1 and SMRT-Although other investigators have also shown functional interactions between AR and SMRT or DAX-1 and we have shown functional interactions between agonist-bound PR and both SMRT and DAX-1, the physical interactions are not as well characterized. Holter et al. (15) has shown in vitro interaction between AR and DAX-1, Dotzlaw et al. (35) has shown in vitro interaction between AR and SMRT and Liao et al. (36) has shown an in vivo interaction between AR and SMRT. In contrast, only the in vitro interaction between PR and SMRT has been shown (33). To confirm that a physical interaction occurs between endogenous AR and DAX-1 in vivo we performed coimmunoprecipitation experiments using the LNCaP cell line. As shown in Fig.  4A DAX-1 but not rabbit anti-mouse IgG (RAM) antibody immunoprecipitated AR. To determine if DAX-1 and PR are involved in direct protein-protein interactions we first incubated in vitro translated DAX-1 and purified polyhisidine-tagged PRB and performed coimmunoprecipitation with either DAX-1 or PR antibody. As shown in Fig. 4B, PRB was co-precipitated much better by DAX-1 antibody than by RAM antibody. Conversely, DAX-1 was coimmunoprecipitated by PR antibody preferentially in the presence of PR (Fig. 4C). To confirm the physical interaction of endogenous PR with DAX-1 and SMRT in vivo we performed coimmunoprecipitation experiments using T47D cells. As shown in Fig. 4D DAX-1 but not RAM antibody precipitates PR. We were able to detect modest SMRT interaction with PR in extracts from both R5020 and RU486treated cells relative to the control RAM (Fig. 4E). The fact that only the A isoform of PR is detectable is probably in part a result of a much lower level of isoform B expression in our T47D strain, as seen from the input lanes.
TSA Does Not Relieve Repression of AR by SMRT-Since SMRT has been reported to physically interact with multiple HDACs and repression has been attributed to the activity of HDACs (3, 37) we asked whether the HDAC inhibitor, tricho-statin A (TSA) would block SMRT-mediated repression. As shown in Fig. 5A, the fold repression of AR by SMRT remains unchanged in the presence of R1881 alone or in combination with TSA, although TSA increases overall activity. Thus, TSAsensitive HDACs do not appear to be involved. Even though DAX-1 has neither HDAC catalytic activity nor a putative HDAC binding site, there is a report that DAX-1 can interact weakly with NCoR, a protein that does interact with HDACs (18,38). Thus, we asked whether an HDAC inhibitor, TSA, would relieve repression by DAX-1. Similar to SMRT, although the overall activity increased, no difference in repression level was observed for DAX-1 (data not shown).  35 S-labeled DAX-1 in vitro translation mix (I), transferred to nitrocellulose, and the membrane was exposed to X-OMAT AR film for 30 min to detect 35 S-labeled DAX-1. D, 3 mg of cleared protein extract from T47D cells treated with 100 nM R5020 were used in an immunoprecipitation with either DAX-1 or RAM antibody. Proteins were extracted as described in methods, run on a 6.5% SDS-PAGE gel and PR detected by Western blot using the 1294 antibody. I, 5 g T47D extract. E, 2 mg of precleared T47D cell extract prepared from cells treated with either 100 nM R5020 (left panel) or 10 nM RU486 (right panel) were used for immunoprecipitation with either SMRT and RAM antibody or RAM antibody only. On the left of each gel are 5 and 1 g of protein extracts from T47D cells treated with the indicated ligand. On the right are the immunoprecipitated proteins. Precipitated PR was detected by Western blot analysis using 1294 PR antibody.

DAX-1 Repression Is Reversed by Coactivators
That Act Primarily through AF-1 (SRC-1e) or AF-2 (ARA70)-Next we wished to determine whether AR repression by DAX-1 could be competed by AF-1 or AF-2 interacting coactivators. We tested SRC-1e, a splice variant of the more commonly studied SRC-1a, that lacks the C-terminal LXXLL motif of SRC-1a and is thought to interact with AR exclusively through AF-1 (39) and ARA70, which interacts through the ligand binding domain (40) for their ability to relieve repression of AR by DAX-1. As shown in Fig. 5B, repression by DAX-1 can be reversed by increasing concentrations of either ARA70 or SRC-1e coactivators. Two explanations are possible for the coactivator reversal of repression of AR by DAX-1. First, coactivators could be competing for the same binding interface on the AR dimer. Alternatively, the corepressors and coactivators can act independently of each other and the final activity of the receptor is a result of the balance between coactivators and corepressors.
DAX-1 Does Not Disrupt the Interaction of AR and the Coactivator SRC-1a in the Mammalian Two-hybrid Assay-To determine whether DAX-1 competes with the AF-1 coactivator binding site we measured the ability of DAX-1 to interrupt interaction between the AR N terminus and SRC-1a in a mammalian two-hybrid assay. The schematic of the constructs used in the interaction assay is shown in Fig. 6D. As shown in Fig.  6A no disruption of the interaction between the N terminus and SRC-1 was observed. This suggests that DAX-1 and possibly other AF-1-interacting coactivators have distinct binding sites. Steroid receptors interact with coactivators in part through LXXLL motifs, which interact with AF-2. Since DAX-1 has one LXXLL motif and 2 LXXLL-like motifs, it is possible that it competes with the coactivators for binding to the LBD of the receptor. However, no interference by DAX-1 was detected in a mammalian two-hybrid assay between SRC-1a and the C terminus of AR (Fig. 6B). Moreover, DAX-1 did not affect interaction between full-length SRC-1a and full-length AR in a mammalian two-hybrid assay showing that the combined AF-1 and AF-2 interaction of AR with SRC-1a is not compromised by DAX-1 (Fig. 6C). Note that there is a significant hormoneindependent interaction between full-length AR and SRC-1, presumably through AF-1 and that this is increased by hormone.

Both SMRT and DAX-1 disrupt the N-C-terminal
Interaction of AR-AR dimers are formed in an antiparallel orientation unlike other steroid receptors (11). Interaction between the N and C-terminal domains of the receptor stabilizes the dimer. To determine whether DAX-1 and SMRT disrupt the contact between the N and C terminus of AR we performed mammalian two hybrid experiments with Bind ARDH and Act ARABC (See Fig. 6D for construct design) alone or with increasing concentrations of DAX-1 or SMRT. As shown in Fig. 7, A and B, both SMRT and DAX-1 disrupt interaction between N-and C-terminal fragments of AR.
DAX-1 Affects PR Dimer Interactions but Not Interaction with SRC-1-Recent reports suggest that the LXXLL motifs of coactivators such as SRC-1 bind to agonist-bound receptors in a region that overlaps with the site in antagonist bound receptor that is recognized by CoRNR box motifs of SMRT and NCoR (41). Although DAX-1 has no perfect consensus CoRNR motifs we asked if its mechanism of action is similar. Interestingly, DAX-1 did not compete with SRC-1 for PR interaction (Fig. 7C). PR is a parallel dimer stabilized by interaction between LBDs and possibly hinge domains that can be detected in a mammalian two-hybrid assay (9). DAX-1 disrupted PR intramolecular interactions between full-length PR (Act PR) and the DBD LBD portion of PR (Bind PRLBD) (Fig. 7D).

DAX-1 Represses the Activity of the AR⌬LBD Mutant but Not of the PR⌬LBD Mutant, While SMRT Represses Both PR and
AR ⌬LBD Fragments-To identify the regions of AR and PR necessary for functional interaction with DAX-1 and SMRT we looked at DAX-1 and SMRT repression of constitutively active AR and PR mutants lacking the LBD (amino acids 1-660 for AR and 1-684 for PR) and therefore AF-2. As shown in Fig. 8A truncated AR is repressed by both DAX-1 and SMRT though possibly to a slightly lesser extent than wild-type AR. However, PR lacking the LBD is repressed only by SMRT (Fig. 8D) and not by DAX-1 (Fig. 8C) although both (Fig. 8B) repress fulllength PR.
The N Terminus of DAX-1 Is Sufficient to Repress PR Activity but Not AR Activity-Nuclear receptor-interacting domains in SMRT have been mapped previously (36). To determine which part of DAX-1 is necessary for AR and PR repression, we cloned the N-terminal and helical C-terminal domains into the mam- malian expression vector pCR3.1 and used them in the transrepression assay. As shown in Fig. 8E only full-length DAX-1 is capable of repressing AR. On the other hand both full-length and the N terminus of DAX-1 repress PR. It is evident however, that full-length DAX-1 is much more effective in repressing PR (Fig. 8F).
Neither DAX-1 nor SMRT Reduce the Ability of AR and PR to Bind DNA in a Promoter Interference Assay-Holter et al. (15) have suggested that DAX-1 may prevent AR from localizing to the nucleus and thus binding to DNA. To test whether DAX-1 or SMRT reduce DNA binding of AR or PR, we performed simultaneous promoter interference assays using a CAT reporter and a transrepression assay using a luciferase reporter. The two different reporters allowed us to measure in the same cell whether amounts of DAX-1 and SMRT sufficient for repression (inhibition of luciferase activity) would remove receptors from the DNA in the promoter interference assay (thus bringing the CAT activity to the level measured in the absence of activated AR). The model for this experiment is shown in Fig.  9. The promoter interference assay is based on the fact that the CMV promoter is much stronger than the hormone responsive promoter formed by the three ARE/PRE elements that are located between the CMV promoter and the CAT transcription start site. The CMV ARE (PRE) CAT promoter is constitutively active in the absence of AR or PR (Fig. 9A). Binding of liganded PR or AR to the ARE/PRE response elements partially blocks transcription from the stronger upstream CMV promoter (Fig.  9B). If DAX-1 or SMRT remove the receptor from the DNA, the CAT activity will return to the no receptor level (Fig. 9C). However, if SMRT or DAX-1 interacts with receptors on the DNA, the activity will remain reduced (Fig. 9D). In contrast, in the absence of hormone the positively regulated ARE Luc promoter will be inactive (Fig. 9A) and addition of hormone will stimulate luciferase activity (Fig. 9B). Whether the repressor causes release from DNA (Fig. 9C) or simply inhibits the activity of the DNA-bound receptor (Fig. 9D), luciferase activity FIG. 6. The SRC-1 coactivator interacting site differs from the DAX-1 interacting site in AR. A, 400 ng of 17-mer Luc and the indicated combinations of either 500 ng of Act or Act ARABC, 100 ng of Bind or Bind SRC-1, and increasing concentrations of DAX-1 balanced with pCR3.1 to a total of 100 ng were transfected; 24 h post-transfection cells were assayed for luciferase activity. RLU values were normalized to total protein concentration. B, cells were transfected as in A except Bind ARDH and Act SRC-1 were used and cells were either treated with 10 nM R1881 (filled bars) or left untreated (empty bars). C, cells were transfected as in A except Act AR and Bind SRC-1 were used. After transfection cells were either left untreated (empty bars) or treated with 10 nM R1881. 24 h later cells were harvested and assayed for luciferase activity that was normalized for protein concentration. D, structure of the constructs used in the interaction assays as previously reported (11). Full-length AR was fused to Act (an activation domain of VP16). Act ARABC contains the N terminus, DNA binding domain and hinge regions (dotted area) fused to the VP16 activation domain (Act). Bind ARDH contains the DNA binding domain, hinge, and LBD of AR fused to Bind (DNA4 DBD). Act PR contains the VP16 activation domain fused to full-length PRB. Bind PRLBD chimera has the GAL4 DBD fused to the hinge and LBD of PR. Bind SRC-1a is a full-length SRC-1a with the GAL4 DBD fused to the N terminus. will be reduced in the presence of repressor. As seen in Fig. 10 under conditions where DAX-1 repressed AR (Fig. 10A, lower  panel) and PR (Fig. 10B, lower panel) activity, both receptors continued to interfere with the stronger CMV promoter similarly to the receptor without repressors. This shows that inhibition by DAX-1 is not due to elimination of DNA binding. The statistically insignificant difference in the activity of the CMV CAT reporter in the presence of pSG5 and pSG5 DAX-1 plasmids in the absence of ligand (Fig. 10A, upper panel) may be due to a slight difference in squelching of these constructs. Note that in the case of PR (Fig. 10B, upper panel) no difference is observed since much less pSG5 DAX-1 DNA is needed to repress PR activity. Alternatively, DAX-1 may facilitate the binding of unliganded AR to the DNA. Colocalization of SMRT and AR on a target gene has been reported previously (42). In agreement with this we see that addition of SMRT sufficient for repression in the same assay (data not shown) does not relieve AR (Fig. 10C) and PR (Fig. 10D) interference with the CMV promoter. The identical behavior of AR and PR in this assay further supports the idea that DAX-1 represses their activity by transforming an active receptor complex on the promoter into an inactive one.

DAX-1 Is Expressed in Prostate and Its Expression Is Diminished in Benign Prostatic
Hyperplasia-Although the studies shown here as well as previously published studies, suggest that corepressors have the potential to modulate the activity of receptors in vivo, little is known about their roles in receptor function. If corepressors are important in the negative regulation of AR activity, then we predict that we would find reduced expression of corepressor under conditions of aberrant androgendependent growth such as in BPH. To test this, we used prostate tissue microarrays to measure the expression of DAX-1   4 and 1.4). While the vast majority of cells in normal tissues expressed DAX-1 with very high intensity (68.5%), the majority of the cells in BPH were negative for DAX-1 or had low to medium levels of expression (Fig. 11). 71.5% of the normal prostate samples have medium to high DAX-1 expression, while 51.5% of BPH samples have negative to low DAX-1 expression. DISCUSSION In this study, we have shown that corepressors differentially repress the activities of nuclear receptors and that the same corepressor may utilize different means and domains to repress the activities of nuclear receptors. As reported recently, DAX-1 represses the activity of agonist bound AR (15) as well as the previously described SF-1 orphan receptor (7) and ER (18). We show here that DAX-1 also represses the activity of PR, but does not repress the activity of endogenous VDR in HeLa cells. Thus, DAX-1 is not a universal repressor for nuclear receptors. Remarkably, DAX-1 represses RU486-dependent activity of AR, but does not repress RU486-dependent activity of PR. We find that SMRT, which has previously been considered to be a repressor of antagonist-bound receptors (4,33), also represses the agonist activity of both AR and PR (Figs. 1 and 2). Although Dotzlaw et al. (35) detected both agonist and partial antagonistdependent interaction between the C-terminal of SMRT and AR using a modified mammalian two-hybrid assay, SMRT was not very effective in repressing agonist bound AR activity in CV1 cells. However, Liao et al. (36) have just reported SMRT repression of agonist bound AR in 293 cells. Thus, the ability of SMRT to repress agonist bound AR may be cell type-dependent. Our finding that SMRT inhibits constitutively active AR lacking the LBD clearly demonstrates that an antagonist is not required for functional interaction between AR and SMRT. We find that both agonist-and partial antagonist-dependent activ-ities of PR are inhibited by SMRT with SMRT being slightly more effective in inhibiting antagonist-dependent activity. NCoR and SMRT have been reported to function in a similar manner and both inhibit the activity of unliganded TR (1,2). Surprisingly, NCoR does not appear to inhibit AR activity in HeLa cells, although it does inhibit AR activity in COS-1 cells (Fig. 1). Cheng et al. (43) recently reported that NCoR can inhibit the activity of AR in transfected CV-1 cells. The finding that the interaction between NCoR and the AR requires the LBD of AR (43) whereas the SMRT interaction site has been localized to the N terminus of AR (35) supports our finding of differential functional interactions in HeLa cells.
Although in vitro interactions between DAX-1 and AR have been reported (15), there have been no reports of interactions of endogenous DAX-1 and AR. Neither functional nor physical interactions between PR and DAX-1 have been reported. Using an antibody to DAX-1, we precipitated endogenous PR from T47D cells and endogenous AR from LNCaP cells demonstrating an in vivo interaction. Although the coimmunoprecipitation was antibody-specific (Fig. 4), the percent of receptor precipitated was rather low. This is, perhaps, not surprising as both cell lines express relatively high levels of receptor and based on our Western blotting of cell extracts (data not shown), either the levels of DAX-1 protein are low or the DAX-1 antibody is not a high affinity antibody. Unfortunately, the molecular mass of DAX-1 (51.7 kDa) precludes the reverse experiment as it comigrates with the heavy chain of IgG. In vitro (35) and in vivo (36,42) interactions of AR with SMRT have been reported previously. The in vivo physical interaction between AR and SMRT is stronger in the presence of the antagonist, flutamide than with DHT (36). We observe an interaction between either RU486-or R5020-bound PR and SMRT in the T47D cell line although the interaction relative to nonspecific was not as clean as in the case of DAX-1 (Fig. 4). Therefore, we scanned the films from multiple experiments to better assess the binding. In the course of seven experiments the intensity of the R5020 bound PR band pulled down by the SMRT antibody was on average 2.2-fold greater than that obtained with RAM antibody. In contrast, cells treated with RU486 produced an average difference of 5.4-fold between specific and nonspecific FIG. 9. Schematic of the promoter interference assay. A, in the absence of liganded receptor the constitutively active CMV promoter induces CAT transcription and the (ARE) 2 Luciferase promoter is silent. B, when hormone is added, activity of the stronger CMV promoter is compromised by 3 interfering AR or PR dimers, while the GRE 2 Luciferase promoter is now active. C, if the corepressor takes receptor off DNA, then activity of the CAT reporter returns to the initial level when no receptor is interfering with the CMV promoter, while luciferase reporter activity is reduced. D, if corepressor-receptor complexes remain on DNA, the CMV promoter activity remains reduced because receptor dimers remain between the CMV promoter and the CAT coding sequence and the activity of the GRE 2 Luciferase promoter is reduced due to repression of the AR transcriptional activity by corepressors. HSP-heat shock protein complex.
binding (five experiments) as determined by densitometry scans. Thus, we believe these results reflect an in vivo interaction.
Although in some cases SMRT represses transcription by recruiting HDACs to the promoter, the histone deacetylase inhibitor, TSA, does not block SMRT-dependent repression of AR demonstrating that the inhibitory mechanism for AR differs. Our finding that SMRT inhibits N/C-terminal interactions implies that SMRT bound to the N terminus prevents the N/C-terminal interaction presumably through steric interference. That SMRT inhibits the activity of partial antagonistbound AR, which does not induce N/C-terminal interactions (44), and AR lacking its hormone binding domain suggests that binding of SMRT also interferes with the formation of productive AR coactivator complexes. Our observation that SMRT blocks the activity of PR lacking its hormone binding domain or treated with the appropriate agonist for each receptor (filled bars, 3 nM R1881 for AR and 10 nM R5020 for PR). Cells were harvested, lysed, and assayed for CAT and luciferase activity, which was normalized for protein levels. Simultaneous measurements of CAT and luciferase activity show that amounts of DAX-1 that are optimal for AR and PR repression (A and B) do not take off receptors from the identical binding sites between the CMV promoter and the CAT coding sequence. also suggests that there is a region distinct from the hormone binding domain that interacts with SMRT. The previous study by Wagner et al. (33) showing PR interacting with the Cterminal portion of SMRT only in the presence of antagonists utilized the PR LBD linked to GAL4 DBD. Thus, there is likely an additional region of PR that also interacts with SMRT and induces the agonist-dependent repression.
DAX-1 is the least well characterized of the repressors we have examined. Although DAX-1 inhibits agonist-dependent activity of both PR and AR, several other features suggest that DAX-1 interactions with the two receptors differ substantially. DAX-1 inhibits the activity of the N terminus of AR but has no effect on the activity of the N terminus of PR. This is consistent with its failure to block antagonist-induced activity of PR, which is mediated through its N terminus. Both the N and C termini of DAX-1 appear to be necessary for AR repression, since neither the putative helical ligand binding domain nor the N-terminal RNA/ssDNA binding motif alone are sufficient to repress AR function. Since the N terminus has 3 LXXLL-like motifs, it is possible that this is the interacting portion of DAX-1, although the C terminus is required for repression. Holter et al. (15) found that the N terminus of DAX-1 was sufficient to cause aberrant localization of AR in the cytoplasm implicating this region in a physical interaction. In contrast, the N terminus of DAX-1 is sufficient to partially repress the activity of PR albeit to a lesser extent than the full-length (Fig.  7F). Thus, although the activities of both PR and AR are re-pressed by DAX-1, the functional interactions differ substantially. Whereas DAX-1 inhibits both the agonist-and antagonistdependent activity of AR as well as inhibiting an N-terminal fragment of AR, only the agonist-dependent activity of PR is inhibited and the N-terminal portion of DAX-1 is sufficient to induce partial inhibition.
Holter et al. (15) had found that expressing high levels of DAX-1 caused mislocalization of hormone bound GFP-AR to the cytoplasm, leading them to propose that DAX-1 acts in part by blocking nuclear localization and, thus, DNA binding of AR. However, our promoter interference studies show that DAX-1 does not reduce the DNA binding-dependent AR-or PR-mediated repression of a constitutive promoter while, in the same cell, preventing transactivation of a receptor responsive luciferase reporter (Fig. 10). Thus AR remains associated with the DNA and the mislocalization may be an artifact of overexpression. Certainly, DAX-1 is predominantly nuclear in prostate (Fig. 11) suggesting that this is an unlikely mechanism.
The finding that addition of high levels of coactivators stimulates the activity of AR in the presence of DAX-1 suggested that DAX-1 might act in part by competing with coactivators for binding to AR. Because DAX-1 inhibits the activity of AR lacking its hormone binding domain, we used a mammalian two hybrid assay to determine whether DAX-1 can block the interaction between SRC-1 and the N terminus of AR, but found no inhibition of binding. Moreover, DAX-1 was unable to inhibit the binding of SRC-1 to either the AR LBD or the FIG. 11. Comparison of DAX-1 expression in normal prostate and BPH. A, DAX-1 expression was detected using diaminobenzidine (brown/black), and the tissue was counterstained with hematoxylin (purple) as described in "Experimental Procedures." DAX-1 expression in normal prostatic tissue is shown in the left portion of the panel, while expression in BPH is shown on the right. Note the much higher intensity and percentage of cells expressing nuclear DAX-1 in both the epithelial cells lining the glands as well as the stromal cells in the normal tissue. (Immunohistochemistry upper panels, ϫ40; lower panels, ϫ100). B, 552 cores of normal prostate were graded for frequency of high intensity of DAX-1 expression and plotted against frequency. 0, no cells with high intensity staining for DAX-1; 1, 1-33% of cells with high intensity DAX-1 staining for DAX-1; 2, 34-66%, and 3, 67-100%. The results were mean, 2.6; S.D., 0.72. C, 554 BPH cores were analyzed as in B, and the number of cores with high intensity DAX-1 staining was plotted against the frequency. Mean, 1.4; S.D., 1.18. full-length receptor. Similarly, DAX-1 did not block interactions between PR and SRC-1. Although physical interactions with SRC-1 were not blocked, SRC-1 acts by recruiting a host of other factors including CBP (45) and P/CAF (46) and it is likely that the presence of DAX-1 prevents the assembly of the fully functional coactivator complex. In contrast, as reported previously (15) and shown here (Fig. 7), DAX-1 blocks N/C-terminal interactions in AR. Moreover, DAX-1 inhibits the interaction of PRLBD with PR (Fig. 7). Steroid receptors dimerize through multiple interaction surfaces including those measured in this assay as well as through interactions between DNA binding domains (9). Our finding that DAX-1 also inhibits the activity of AR lacking the hormone binding domain as well as antagonist-bound AR, which lack the N/C-terminal interaction, demonstrates that DAX-1 has multiple actions on AR. Taken together, these studies suggest that DAX-1 interacts at sites distinct from the SRC-1 binding site and inhibits activity by reducing dimer interactions and, in addition, likely interferes with optimal formation of complete coactivator complexes.
Although our studies and those of others have shown that DAX-1 has the potential to negatively regulate the activity of AR, the contribution to AR action in vivo is unknown. Interestingly, Gregory et al. (48) have shown that two coactivators SRC-1 and TIF2 are overexpressed in recurrent prostate cancer, another androgen-dependent disease, but found no evidence for overexpression in BPH. Our studies of the expression of DAX-1 in prostate and in BPH samples suggest that DAX-1, indeed, regulates AR activity. The BPH samples contain markedly lower levels of DAX-1 implying elevated AR activity. AR action contributes to BPH and treatment with a 5␣-reductase inhibitor, finasteride, which reduces levels of the potent androgen, dihydrotestosterone is a common treatment (49). Thus, the loss of DAX-1 expression may be a significant contributor to the development of BPH while overexpression of coactivators may be more important in prostate cancer.