The Androgen Receptor Represses Transforming Growth Factor- (cid:1) Signaling through Interaction with Smad3*

In the prostate, androgens negatively regulate the expression of transforming growth factor- (cid:1) (TGF- (cid:1) ) liga-nds and receptors and Smad activation through unknown mechanisms. We show that androgens (dihyd-rotestosterone and R1881) down-regulate TGF- (cid:1) 1-indu-ced expression of TGF- (cid:1) 1, c-Fos, and Egr-1 in the human prostate adenocarcinoma cell line, LNCaP. Moreover, 5 (cid:2) -dihydrotestosterone (DHT) inhibits TGF- (cid:1) 1 activation of three TGF- (cid:1) 1-responsive promoter constructs, 3TP-luciferase, AP-1-luciferase, and SBE4 BV -luciferase, in LNCaP cells either with or without enforced expression of TGF- (cid:1) receptors (T (cid:1) RI and T (cid:1) RII). Similarly, DHT inhibits the activation of Smad-binding element (SBE)4 BV -luciferase by either constitutively activated T (cid:1) RI (T204D) or constitutively activated Smad3

Normal prostatic epithelium depends on androgens for growth, development, secretory function, and survival (1)(2)(3)(4). Most remarkably, androgen ablation induces massive apoptosis of prostatic epithelium (2,(5)(6)(7)(8). Loss of androgen dependence occurs invariably during prostate carcinogenesis, accounting for poor long term success of androgen ablation therapy (9). Recent studies (10) show that acquisition of androgen autonomy occurs despite retention or elevated expression of the androgen receptor (AR) 1 in the majority of prostate tumors. AR, a 110-kDa zinc finger transcription factor belonging to the nuclear receptor superfamily, is activated by phosphorylation (11) and dimerization upon ligand binding. This promotes nuclear localization and binding of AR to androgenresponsive elements in the promoters of androgen-regulated genes. AR-mediated transcription is regulated by many ARinteracting proteins such as ARA 70 (AR-associated proteins) (12) and ARA 160 (13), along with cAMP-response element-binding protein (14), AP-1 (9,15), and Ets (16). The growing list of recently discovered AR transcriptional co-regulators supports the notion that complex networks of signals tightly regulate transcription by androgens. Understanding how these signals promote growth and maintain cell viability will certainly impact on the therapeutic strategies for the prevention and cure of prostate cancer.
Androgens negatively regulate TGF-␤1 ligand (17,33) and receptor expression (34,35), along with Smad expression and activation (36) in the prostate. Recent reports show AR associates with Smad3 and that this association may either enhance (37) or inhibit (38) AR-mediated transcription. Here we report the first in vitro demonstration that DHT inhibits TGF-␤ signaling in prostatic epithelial cells through interaction of AR with Smad3. Our results support that the binding of ligandbound AR to activated Smad3 inhibits TGF-␤ transcriptional responses by blocking the association of Smad3 with SBE.
Quantitation of TGF-␤1-3-A suspension of 250,000 cells/ml in DMEM/F-12, 10% dextran charcoal-stripped FBS was dispensed in polylysine-coated 6-well dishes. Following 48 h, wells were washed twice with DMEM/F-12 supplemented with 100 g/ml bovine serum albumin fraction V, 5.0 g/ml human transferrin, 20 ng/ml mouse epidermal growth factor, and 10 ng/ml sodium selenite and replaced with 2 ml of the above medium. After 2-3 h at 37°C, the medium was replaced as before, and various test factors were added. Following 48 -72 h of treatment, conditioned media from these cultures were harvested to assay for TGF-␤1-3 by sandwich enzyme-linked immunosorbent assay (40 -42). Samples were prepared as follows. Conditioned medium was treated with protease inhibitors (2 g/ml aprotinin, 2 g/ml leupeptin, and 0.2 mg/ml phenylmethylsulfonyl fluoride) and clarified at 13,000 ϫ g for 10 min. 55 l of a 100% solution of trichloroacetic acid was added per ml of conditioned medium, vortexed, set on ice for 30 min, and microcentrifuged at 14,000 ϫ g for 10 min. The pellets were washed with 1 ml of cold ether/ethanol (1:1, v/v), lyophilized for 10 min, and dissolved in 125 l of 4 mM HCl, 150 mM NaCl, and 0.5 mg/ml bovine serum albumin. Daily TGF-␤1 secretion rates were normalized to DNA content of producer cells. DNA was measured by fluorescence with 3323 Hoeschst (43).
Northern Blot Analysis-Total RNA was purified using RNeasy TM columns (Qiagen) as modified by Bonham and Danielpour (44). 10 g of total RNA was separated by electrophoresis through a 1% agarose, 0.66 M formaldehyde gel containing 0.72 g/ml ethidium bromide. Equal loading and even transfer were assessed by visualization of the 18 S and 28 S rRNAs. Gels were treated with 60 mM NaOH for 20 min, neutralized with 50 mM Tris-HCl (pH 7.4) and 10 mM NaCl for 20 min, and then blotted onto Nytran (0.45-m) for 16 -24 h with 10ϫ SSPE. After membranes were cross-linked with UV light, they were pre-hybridized for 2 h, hybridized overnight at 65°C, and washed at the same temperature under modified Church's hybridization conditions (45). The cDNA probes were labeled with [ 32 P]dCTP by a random priming reaction. Hybridization was done with 2-3 ϫ 10 6 cpm/ml.
GST fusions were purified as follows: Luria Burtani/Miller broth in 500-ml batches containing 100 g/ml of ampicillin were inoculated with 1-ml overnight cultures of BL21 (Stratagene) cells transformed with pGEX-Smads or pGEX-AR constructs. Following 1.5 h of incubation at 37°C, cultures were brought to room temperature and induced overnight with 50 g/ml isopropyl-1-thio-␤-D-galactopyranoside. Bacterial pellets were lysed by sonication using a microprobe (50 strokes, 50% duty cycle, setting 3) in 8 ml of cold TNE (10 mM Tris (pH 7.5), 150 mM NaCl, 5 mM EDTA) containing Complete Mini EDTA-free Protease Inhibitor Mixture (Roche Molecular Biochemicals), 0.2 mg/ml phenylmethylsulfonyl fluoride, and 2 mM dithiothreitol. Lysates were treated with 1% Nonidet P-40 for 5 min on ice, clarified at 20,000 ϫ g for 15 min at 4°C, and mixed overnight at 4°C with 0.4 ml of glutathione-Sepharose. Resin-containing fusion proteins were washed twice with 10 ml of TNE containing 1% Nonidet P-40 and 0.5 mM dithiothreitol and three times with TNE containing 0.5 mM dithiothreitol. Fusion proteins were either used directly on resin or eluted at 4°C by a 45-min incubation with 10 mM glutathione, 50 mM NaCl, and 50 mM Tris (pH 8.7) and then dialyzed extensively in EMSA buffer without glycerol. GST fusion proteins were quantified by Coomassie Blue staining of Tris glycine gels, using bovine serum albumin standards.
Transfection of NRP-154 Cells-Cells (6.0 ϫ 10 4 ) were plated overnight in 12-well dishes and transfected using a standard calcium phosphate co-precipitate method for 3 h in GM2 (39) made with dextran charcoal-stripped FBS. The calcium phosphate/DNA co-precipitate was washed away, and the cells were glycerol-shocked (15% glycerol in 1ϫ HEPES-buffered saline) for 90 s. Cells were washed twice with phosphate-buffered saline, allowed to recover overnight in GM3 (47) (made with dextran charcoal-stripped calf serum) Ϯ 10 nM DHT (Sigma) (or vehicle, 70% ethanol), and then treated with 10 ng/ml TGF-␤1 (R & D Systems) or vehicle (4 mM HCl, 1 mg/ml bovine serum albumin). Luciferase was measured 24 h later using Promega's Dual Luciferase Assay Kit and a Dynex ML3000 Microtiter Plate Luminometer.
Transfection of LNCaP Cells-Cells (2.5 ϫ 10 5 ) were plated overnight in 6-well polylysine-coated plates with 2 ml of media (DMEM/F-12 with 15 mM HEPES, 1% dextran charcoal-stripped FBS). Cells were transfected with LipofectAMINE Plus reagent according to the manufacturer's protocol (Invitrogen) for 3 h before replacing with above medium (ϩ20 ng/ml epidermal growth factor, except for SBE4 BV -luciferase assays) and Ϯ10 nM DHT. 10 ng/ml TGF-␤1 was added the next day, and the cells were harvested 24 -48 h later. Luciferase was measured using Promega's Dual Luciferase Assay Kit and a ML3000 Microtiter Plate Luminometer.
Immunoprecipitation-NRP-154 cells (8.0 ϫ 10 5 ) were plated overnight in 100-mm 2 dishes with 5 ml of GM2. Cells were transiently transfected with 5 g each pCMV2-FLAG-S3* and pCMV5-AR using a standard calcium phosphate co-precipitate method. 10 nM DHT (or vehicle) was added after the transfection and again 4 h prior to harvest. Cells were lysed at 4°C with 400 l of cold radioimmunoprecipitation (RIPA) buffer (containing Complete Mini-EDTA-free Protease Inhibitor Mixture and 0.2 mg/ml phenylmethylsulfonyl fluoride) using a 25gauge needle and then clarified at 10,000 rpm for 10 min (4°C). Lysates were precleared using 0.5 g of mouse IgG and 20 l of protein A/G Plus-agarose (Santa Cruz Biotechnology) for 1 h, immunoprecipitated with either 2 g of mouse anti-FLAG M2 (Sigma) or mouse anti-AR (Lab Vision) overnight in the presence of 10 nM DHT (or vehicle), and treated with 20 l of protein A/G Plus-agarose for 2 h at 4°C. The resin was washed four times with RIPA buffer and eluted with 50 l of 1ϫ SDS-PAGE loading buffer. Samples were separated by electrophoresis using 4 -12% NuPAGE BisTris gel (Invitrogen) with 1ϫ MOPS buffer (Invitrogen) and transferred to nitrocellulose (Invitrogen). Western blot detection was done as described (47). Antibodies for Western blot detection were from Santa Cruz Biotechnology (human androgen receptor C-19, (1:4000)) and Sigma (FLAG M2 (1:500)).
In Vitro Transcription/Translation and GST Pull Down-Fulllength AR cDNA (a.a. 1-919) was cloned into pcDNA3 (Invitrogen) and in vitro transcribed/translated using a T7 TnT® Coupled Reticulocyte Lysate System (Stratagene). 5 l of the 35 S-labeled TnT product was added to equal volumes of glutathione-Sepharose-GST-Smad fusions in buffer A (20 mM Tris (pH 7.8), 180 mM KCl, 0.5 mM EDTA, 5 mM MgCl 2 , 0.5 mM ZnCl 2 , 10% glycerol, 0.1% Nonidet P-40, 0.05% milk, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride) and rotated overnight at 4°C. The glutathione-Sepharose beads were then centrifuged (5 min at 3000 rpm) and washed 3 times with 500 l of cold buffer A. GST-Smad⅐AR complexes were eluted with 50 l of 1ϫ SDS-PAGE loading dye and heated to 70°C for 10 min prior to resolving the complexes in a 4 -12% NuPAGE BisTris gel with 1ϫ MES buffer (Invitrogen). The proteins were then transferred to nitrocellulose and detected by a Molecular Dynamics PhosphorImager.

Androgens Block the TGF-␤1-induced Autoinduction of TGF-␤1 Ligand and c-Fos and Egr-1 Expression in LNCaP
Cells-Androgens negatively regulate expression of TGF-␤ ligands in prostatic cells in both animals (17) and rat cell culture (48). We determined whether androgens could also down-regulate TGF-␤1 autoinduction (49) in the androgen-responsive human prostatic adenocarcinoma cell line, LNCaP, under serum-free conditions, using isoform-specific sandwich enzymelinked immunoadsorption assays for TGF-␤s. From a list of common hormones and growth factors added to LNCaP cells, only TGF-␤s substantially elevated protein levels of TGF-␤1 (data not shown; referred to as "TGF␤ autoinduction"). A physiological concentration of DHT (10 nM), the active metabolite of testosterone, inhibited the induced expression of TGF-␤1 protein (Fig. 1A) and mRNA (Fig. 1B). The stable androgen analogue, R1881, also blocked TGF-␤2-induced expression of TGF-␤1 (Fig.  1, B-D). We used R1881 to study the kinetics of TGF-␤1 mRNA loss following an initial (72 h) induction by TGF-␤1 (Fig. 1D). Changes in the expression of two transcription factors (c-Fos and Egr-1), shown to be induced by TGF-␤ and involved in TGF-␤ autoinduction (50 -53), were also determined (Fig. 1D). The induced expression of TGF-␤1 ligand, c-Fos, and Egr-1 mRNAs was inhibited by R1881 in a time-dependent manner. Of note, LNCaP cells treated with TGF-␤1 will maintain increased mRNA levels of TGF-␤1 ligand, c-Fos, and Egr-1 between 2 and 5 days of treatment (data not shown).
The decrease in TGF-␤1 expression by R1881 occurred through transcriptional repression and did not require de novo protein synthesis, as demonstrated by loss of TGF-␤1 mRNA expression following 24 h of treatment with R1881 in the presence of actinomycin D or cycloheximide (Fig. 1E, right panel). Quantification of the Northern blot in Fig. 1E is also presented (Fig. 1E, left panel). Actinomycin D caused the same loss of TGF-␤1 mRNA as did R1881, and co-treatment of R1881 and actinomycin D did not enhance loss by R1881 alone. In the presence of cycloheximide, R1881 caused a smaller loss of TGF-␤1 mRNA expression, perhaps due to a decrease in AR expression. Co-treatment of actinomycin D and cycloheximide had the same effect as actinomycin D alone. Together, these data suggest that AR functions through directly repressing transcription by TGF-␤.
DHT Inhibits the Transcriptional Activation of Several Response Elements Induced by TGF-␤1-We tested the above hypothesis by assaying the effect of DHT on TGF-␤1-induced transcriptional activation of various response elements. In LN-CaP cells, the TGF-␤1-induced 3TP-luciferase (54) activity was blocked by co-treatment with DHT ( Fig. 2A). Previous work supports that LNCaP cells express low levels of either T␤RI or T␤RII (55, 56), potentially accounting for their relatively weak response to TGF-␤1. Therefore, we co-transfected these cells with either T␤RI or T␤RII along with 3TP-luciferase. Co-trans- fection of T␤RII resulted in Ͼ50-fold enhanced activation of luciferase by TGF-␤1 (Fig. 2B). In contrast, T␤RI did not enhance TGF-␤1-induced 3TP-luciferase activity (data not shown), suggesting that T␤RII but not T␤RI was limiting in our LNCaP lineage. Overexpression of T␤RII did not blunt the ability of DHT to inhibit TGF-␤1-induced 3TP-luciferase (Fig.  2C). Additionally, the inhibition of TGF-␤1-induced 3TP-luciferase activity was dependent on DHT concentration (Fig. 2D). These data demonstrate that the levels of endogenous AR in LNCaP can fully repress TGF-␤1-induced 3TP-luciferase activity, even in the presence of overexpressed T␤RII, when ligand-bound.
We tested the above results and the requirement for AR in another prostatic cell line, NRP-154, which is exquisitely sensitive to TGF-␤1 but has undetectable levels of AR (39). Transfection of full-length AR (57) enabled DHT to similarly inhibit TGF-␤1-induced 3TP-luciferase activity in NRP-154 cells (Fig.  2E). 3TP-luciferase has a complex promoter, consisting of 3ϫ TRE elements upstream of a plasminogen activator inhibitor-1 promoter fragment. We co-transfected LNCaP cells with basic promoter constructs, AP-1-or SBE4 BV -luciferase (58), to define better the elements responsible for transcriptional inhibition by DHT. TGF-␤1 activated both AP-1-luciferase and SBE4 BVluciferase, and 10 nM DHT substantially inhibited these activ- ities, comparable with 3TP-luciferase (Fig. 2, F and G). Consistent with these data, activation of SBE4 BV -luciferase by a constitutively activated T␤RI, T204D (59), was also suppressed by DHT (Fig. 2H).
DHT Inhibition of TGF-␤1-induced SBE4 BV -Luciferase Activity Occurs through Smad3-The above data suggested Smads were a potential direct target of the observed DHT inhibition on TGF-␤1 responses. To begin examining this possibility, we transfected LNCaP and NRP-154 cells with constitutively active Smad 2 or 3 (S2* and S3*, respectively) developed by site-directed mutagenesis of their carboxyl-terminal serines to aspartic acids in the SSXS phosphorylation motifs. Transfection of S3* in both LNCaP (Fig. 3A, left panel) and NRP-154 cells (Fig. 3A, right panel) caused pronounced SBE4 BV -luciferase induction, whereas S2* did not induce SBE4 BV -luciferase activity in either cell line. The induction of SBE4 BV -luciferase by TGF-␤1 in LNCaP cells was blocked by co-transfection of DN-Smad4 or Smad7 (Fig. 3, B and C, respectively), implicating the dependence of Smads on this reporter. Although co-transfection of LNCaP cells with wild-type AR (ϪDHT) had little inhibitory effect on S3*-induced SBE4 BVluciferase activity, treatment of AR-transfected LNCaP cells with DHT completely blocked transcription by S3* (Fig. 3D). Due to the substantial S3*-induced SBE4 BV -luciferase activity in LN-CaP, additional AR was required to fully repress this response (however, endogenous AR could fully inhibit TGF-␤1-induced SBE4 BV -luciferase activity, Fig. 2G). Similarly, in NRP-154 cells, DHT inhibited S3*-induced SBE4 BV -luciferase activity only in AR co-transfected cells (Fig. 3E), demonstrating the dependence of both androgens and AR on this effect. Finally, to determine whether Smad3 was the major target of DHT-induced inhibition of TGF-␤1 signaling, we transfected LNCaP cells with wild-type Smad3 in an attempt to reverse the inhibition. As shown in Fig.  3F, overexpression of wild-type Smad3 fully reversed the inhibition of TGF-␤1-induced SBE4 BV -luciferase activity by AR and DHT.
The Androgen Receptor Associates with Smad3 Independent of Ligand-Recent evidence from other laboratories (37,38) suggests that AR associates with Smad3 in an androgen-independent manner. We confirmed these results in our system, along with determining if S3* was still able to bind AR due to mutation of its carboxyl terminus, by both GST pull-down assays and co-immunoprecipitation, respectively. NRP-154 cells were co-transfected with FLAG-S3* and AR and treated with DHT (or vehicle) before immunoprecipitation 24 h later. Fig. 4A demonstrates that immunoprecipitating against either FLAG or AR can capture S3*⅐AR complexes independent of DHT addition. Full-length (a.a. 1-919) in vitro transcribed/ translated 35 S-labeled AR associated specifically with GST-Smad3 and not with GST-Smad2 or -4 and weakly with GST-Smad1 (Fig. 4B). Thus, our results show that AR associates directly with Smad3 (and S3*) in the absence of DHT (Fig. 4, A  and B) but not with any other Smads (Smads 2 or 4) involved in TGF-␤1 signal transduction.
The Androgen Receptor DNA Binding Domain Is Not Required to Inhibit Smad3 Activity-To define better the inhibition of S3*-induced SBE4 BV -luciferase activity by AR, we tested various domains of AR for inhibitory function. NRP-154 cells were co-transfected with SBE4 BV -luciferase, S3*, and various AR expression constructs (60, 61) as follows: wild-type AR, ⌬538 -614 (no DNA binding domain (DBD)), DBD, and ABC (no hinge/no ligand binding domain (LBD)) (Fig. 5A). Maximum inhibition of AR on S3*-induced SBE4 BV -luciferase activity was observed by expression of wild-type AR or ⌬538 -614 in the presence of ligand (Fig. 5B). In contrast to full-length AR that had no inhibitory effect without DHT, ⌬538 -614 repressed the majority of S3* activity without ligand. Expression of DBD or ABC had little to no significant inhibitory effect on S3* activity; and as expected, DHT did not change the effectiveness of these (DBD or ABC) regions on inhibiting S3* activity (Fig. 5B). These data suggest that DHT promotes the inhibitory effect of AR on Smad3 in a DBD-independent manner that requires the LBD or some other region in the AR carboxyl terminus.
The Ligand Binding Domain of the AR Inhibits Smad3 Binding to the Smad-binding Element-Inhibition of Smad3 signaling by DHT could occur through several mechanisms. Upon TGF-␤1 treatment, Smad3 translocates to the nucleus where it induces target gene expression; similarly, AR is localized to the nucleus in the presence of DHT. Therefore, we examined the effect of AR on the association of Smad3 to SBE. Electrophoretic mobility shift assays were performed with 32 P-labeled SBE oligonucleotides and purified GST-Smad3 in the presence of different domains of AR ( Fig. 6A; amino-terminal GST-AR, GST-AR DBD, or GST-AR LBD). GST-Smad3 was able to bind to the SBE in the presence of the amino-terminal GST-AR and GST-AR DBD; however, GST-AR LBD inhibited this association (Fig. 6B). This result is consistent with the data presented in Fig. 5B, which suggests the carboxyl terminus (i.e. LBD) of AR confers inhibition of S3* activity. We next determined the effect of DHT on the association of GST-Smad3 binding to SBE in the presence of GST-AR LBD (200, 400, and 800 ng). The addition of 10 nM DHT enhanced GST-AR LBD-induced inhibition of GST-Smad3 binding to SBE (Fig. 6C). This inhibition was specific to GST-Smad3 because no loss of the GST-Smad4⅐SBE complex (Fig. 6D) in the presence of GST-AR LDB (800 ng) ϩ DHT was observed. Also, DHT alone did not abrogate GST-Smad3 binding to the SBE. Together, these results suggest that DHT binding to AR enhances or promotes inhibition of Smad3 association to SBE, thereby leading to inhibition of Smad3-mediated transcription. DISCUSSION Our data show that androgens can block TGF-␤ responses in prostate epithelial cells through an association of AR with Smad3, which inhibits the binding of Smad3 to SBEs in TGF-␤-responsive promoters. This mechanism of cross-talk provides both rapid and direct means by which androgens can inhibit TGF-␤ signaling.
Our data demonstrate that DHT is not necessary for the in vitro association of full-length AR to Smad3 (Fig. 4, A and B). However, the exact localization of ligand-free AR remains controversial and depends on cell type (60 -63). It is likely that the DHT-independent association of AR to Smad3 may not occur in intact NRP-154 cells due to cytosolic localization of ligand-free AR and nuclear localization of S3*. If so, the association of AR to Smad3 without DHT would have occurred after cell lysis. In view of the DHT requirement for the loss of TGF-␤1-induced transcription by AR, we believe this association in NRP-154 cells occurs only by nuclear co-localization of ligand-bound AR and active Smad3. Data presented in Fig. 6 suggest DHT may also promote a modification in the AR⅐Smad3 complex to inhibit further the association to SBE.
There is a definite requirement for Smads on all TGF-␤1responsive reporter constructs used in this study. We have investigated this requirement by observing the effects of DN-Smad4 and Smad7 on TGF-␤1-induced 3TP-luciferase and AP-1-luciferase. DN-Smad4 or Smad7 inhibit ϳ70 -80% of TGF-␤1-induced activity on both of these reporters (data not shown). As shown in Fig. 3, B and C, DN-Smad4 or Smad7 also inhibits TGF-␤1-induced SBE4 BV -luciferase by ϳ80%.
DHT is required for full-length AR to inhibit S3*-induced SBE4 BV -luciferase (Fig. 5B). ⌬538 -614 AR, which lacks the DBD, is able to inhibit ϳ70% of S3* activity independent of ligand and ϳ90% (similar to wild type) of S3* activity with ligand (Fig. 5B); therefore, the DBD may prevent DHT-independent inhibition of S3*. The mechanism by which ⌬538 -614 without DHT inhibits S3* activity is unclear; however, data show AR⌬DBD (no DBD) is more efficiently translocated to the nucleus compared with an AR nuclear localization mutant (64). This suggests that the nuclear localization signal of AR is bi-or tripartite and may be enhanced in the absence of DBD. Thus, we hypothesize a greater percentage of ⌬538 -614 may be localized to the nucleus in the absence of DHT, as compared with wild-type AR. The means by which AR DBD alone reduces S3* activity is not known, because the ⌬538 -614 AR data suggest this region is not necessary for complete inhibition. Furthermore, the EMSA results substantiate that the DBD is not essential for AR-mediated loss of S3* activity (Fig. 6A). AR LBD is able to inhibit GST-Smad3 binding to SBE without DHT; nevertheless, addition of ligand does enhance this effect by ϳ4-fold, perhaps through an AR LBD conformational change.
The above result parallels data from previous reports that demonstrated LBD alone can bind to Smad3 (37). Interestingly, one report (37) showed the AR DBD or AR LBD could bind to Smad3, although no functional consequence of these interactions was presented. The other study revealed the AR aminoterminal region (a.a. 1-563) associates with the MH2 domain of Smad3 to repress androgen receptor-mediated transcription of murine mammary tumor virus-luciferase (38), but the significance of this interaction (amino-terminal AR⅐Smad3) to TGF-␤ signaling was not examined. We are not able to explain the discrepancy that exists among several groups attempting to characterize the association of AR with Smad3. However, we believe our data are solid due to our EMSA assay observing a function of the AR⅐Smad3 association and not solely the physical interaction. It is important to note that although AR can inhibit Smad3, we cannot yet rule out that this inhibition is exclusive to Smad3 and does not also involve Smads 2 or 4 through indirect means.
The normal prostate requires an intact androgen signaling pathway for growth and function, whereas prostate carcinomas often escape from this dependence through changes likely to involve AR. AR receptor mutants that are unable to activate androgen-responsive genes or change the sensitivity of the receptor to circulating androgens (or other steroids) may suppress androgen dependence in the prostate (65)(66)(67). Also, mutations within the AR carboxyl terminus which decrease steroid affinity may produce a ligand-insensitive and yet transcriptionally active AR (68,69). An abundance of AR polymorphisms without functional significance have been characterized, especially within the amino-terminal region, which are linked to increased incidence of prostate cancer (70 -72). The expression of several ARAs have been implicated in altering AR activity in the prostate (12,13,(73)(74)(75), which may dysregulate AR⅐Smad3 interaction and function. Moreover, it is apparent that prostate tumors, both localized and metastatic, maintain or increase AR expression and sensitivity following androgen ablation therapy (10, 76 -79). The data presented here suggest a novel mechanism by which such aberrations in AR can directly antagonize TGF-␤ effects within the prostate and promote the development and progression of cancer.
Restoration of TGF-␤ receptor levels by overexpression of wild-type T␤RII in LNCaP cells was reported to promote TGF-␤ responsiveness and suppress tumor growth through reduced cell proliferation and the induction of apoptosis (80).
FIG. 6. LBD of AR inhibits Smad3 binding to SBE. A, diagram of different AR domain GST-fusions. B, an EMSA was performed using a 32 P-labeled SBE probe (50,000 cpm) with 400 ng of purified GST-Smad3 in the presence of various GST-AR domains (800 ng of amino terminus (N-term) of DBD or LBD). Complexes were resolved using a 6% DNA retardation gel and 0.5ϫ TBE running buffer. The gel was dried and visualized using a PhosphorImager. C, EMSA (as described in B) of the AR LBD (200, 400, and 800 ng) induced inhibition of GST-Smad3 binding to the SBE with and without 10 nM DHT. D, GST-Smad4 binding to SBE in the presence of 800 ng AR LBD ϩ DHT is shown as control. A representative gel of four independent experiments is shown.
Consistent with a tumor-suppressive role of TGF-␤ in the prostate, we have shown that DN-T␤RII promotes malignant transformation of two non-tumorigenic prostate epithelial cell lines (19). 2 Our data showing that TGF-␤ is a potent inducer of apoptosis in the above cell lines further support TGF-␤ may suppress prostate tumor growth through the induction of apoptosis (47,81). With this in mind, the current work proposes that constitutive or enhanced activation of AR, through means discussed earlier, may cause loss of androgen dependence (e.g. rescue from undergoing apoptosis) partly via loss of TGF-␤ signaling through inactivation of Smad3, because Smads were shown to be critical to the induction of apoptosis by TGF-␤ (82). This loss would allow for prostatic epithelial cells to escape growth inhibition and apoptosis by TGF-␤, contributing to carcinogenesis of the prostate.
In conclusion, our data demonstrate for the first time that the association of AR with Smad3 can inhibit the ability of Smad3 to bind SBE and activate transcription. Amplification of AR, or variances within the AR (or its signaling pathway) that promote deregulated and enhanced Smad3 binding, may counteract tumor suppression by TGF-␤. Moreover, these findings strengthen our hypothesis that androgens promote viability of prostatic epithelial cells, in part, by preventing TGF-␤-induced apoptosis. In view of the numerous binding partners for Smads and AR, understanding how such co-regulators affect the inhibition of Smad3 signaling is critical to our understanding of normal homeostatic mechanisms and carcinogenesis of the prostate.