Coordinated Action of Hypoxia-inducible Factor-1α and β-Catenin in Androgen Receptor Signaling*

Background: The mechanism that regulates androgen receptor (AR) function in castration-resistant prostate cancer (CRPC) remains unclear. Results: Hypoxia-inducible factor (HIF)-1α enhances β-catenin-mediated AR transactivation and promotes nuclear translocation of β-catenin. Conclusion: Coordinated action of HIF-1α and β-catenin contributes to AR function in CRPC. Significance: The novel functions of HIF-1α in AR signaling are important for understanding the development of CRPC. The androgen receptor (AR) acts as a ligand-dependent transcriptional factor and plays a critical role in the development and progression of androgen-dependent and castration-resistant prostate cancer. Castration results in hypoxia in prostate cancer cells, and hypoxia enhances transcriptional activity of AR through hypoxia-inducible factor (HIF)-1α at low serum androgen levels mimicking the castration-resistant stage. However, HIF-1α is necessary but not sufficient for hypoxia-activated AR transactivation, and the molecular mechanism that regulates AR function in castration-resistant prostate cancer remains unclear. Here, we report that β-catenin is required for HIF-1α-mediated AR transactivation in hypoxic LNCaP prostate cancer cells under low androgen conditions. HIF-1α and β-catenin coordinately enhanced AR N-terminal and C-terminal interaction. β-Catenin accumulated in the nucleus in the HIF-1α protein-positive cells of LNCaP xenografts in castrated mice. In LNCaP cells, when HIF-1α was knocked down or was exogenously expressed in the cytoplasm, hypoxia-induced nuclear localization of β-catenin was inhibited. β-Catenin formed a complex with HIF-1α both in the nucleus and in the cytoplasm. Hypoxia increased the amount of a complex composed of AR and β-catenin, and knockdown of HIF-1α attenuated the recruitment of AR and β-catenin to the androgen response elements (AREs) of androgen-responsive genes. Furthermore, together with β-catenin, HIF-1α bound to the AREs in the presence of androgen. These results demonstrate that (i) HIF-1α and β-catenin coordinately enhance AR transactivation by accelerating N-terminal and C-terminal interaction; (ii) HIF-1α promotes nuclear translocation of β-catenin in hypoxia; and (iii) AR, HIF-1α, and β-catenin form a ternary complex on AREs.

Androgens, principally testosterone and its more potent metabolite, 5␣-dihydrotestosterone (DHT), 2 bind to and activate the androgen receptor (AR), leading to the development and survival of normal and cancerous prostate tissues (1,2). Androgen ablation by medical or surgical castration, which is the most common treatment for patients with metastatic prostate cancer, reduces androgen levels (e.g. 0.1 nM DHT (3)) and concomitantly regresses tumor growth. However, most patients eventually relapse to castration-resistant prostate cancer (CRPC), which is referred to as androgen-refractory or androgen-independent prostate cancer, resulting in an increase in mortality related to prostate cancer (4 -6). In the low androgen environment, CRPC cells still express AR and grow in an AR-dependent manner. However, the mechanisms leading to the development and progression of CRPC remain poorly understood.
The AR is a member of the steroid hormone receptor superfamily, whose members share a common modular structure that is composed of an N-terminal domain (NTD), a central DNA-binding domain (DBD), and a ligand-binding domain (LBD) (7). Ligand-free AR exists in the cytoplasm in an inactive state. The binding of DHT to the LBD results in a conformational change of AR to an active form. The ligand-bound AR translocates to the nucleus and binds to specific androgen response elements (AREs) in the promoter regions of androgen-responsive genes, such as NKX3.1 and PMEPA-1, followed by transcriptional activation of these genes (8). Transcriptional activity of AR is enhanced by recruitment of coactivators, such as the p160 proteins (SRC-1 and TIF-2), ARA proteins (ARA24, ARA55, and ARA70), GAPDH, and ␤-catenin. The binding of ligand to AR causes the interaction between the N-terminal and C-terminal regions of AR (N-C interaction) (9). The N-C inter-action is a decisive event in causing AR transactivation and is enhanced by only a limited number of coactivators, such as SRC-1, ARA24 (10), and ␤-catenin (11).
Androgen ablation by castration causes hypoxia due to blockade of blood flow in the prostate tissue (12,13). Hypoxia is a key factor in tumorigenesis because solid tumors, including prostate cancer, grow in a hypoxic microenvironment that results from insufficient blood flow and dysfunctional tumor vessels (14). The transcriptional activation of hypoxia-inducible genes is mediated by hypoxia-inducible factor (HIF)-1 (15). HIF-1 is a heterodimeric protein composed of HIF-1␣ and HIF-1␤/ARNT subunits, both of which belong to basic helix-loophelix/PER-ARNT-SIM domain transcriptional factors. The HIF-1␤/ARNT protein is constitutively expressed in an oxygen-independent manner. In contrast, the HIF-1␣ protein is degraded by proteasome in an oxygen-dependent manner. HIF-1␣ has two transactivation domains, called the N-terminal and C-terminal activation domain (N-TAD and C-TAD, respectively). In normoxia, the HIF-1␣ subunit is hydroxylated on two conserved proline residues, leading to ubiquitination and proteasomal degradation (16). In hypoxia, however, the stabilized HIF-1␣ translocates to the nucleus and forms a heterodimer with HIF-1␤/ARNT. Subsequently, HIF-1 binds to hypoxia-responsive elements (HREs) in the promoter regions of hypoxia-inducible genes, resulting in their transcriptional activation (17). In CRPC, HIF-1␣ is expressed at a higher level than in benign prostate hyperplasia and normal tissue (18). We previously found that hypoxia enhances transcriptional activity of AR through HIF-1␣ at low androgen concentrations (0.05-0.1 nM DHT) mimicking the castration-resistant stage (19). Knockdown of HIF-1␣, but not HIF-1␤/ARNT, inhibits hypoxia-enhanced AR transactivation. However, exogenous expression of an O 2 -insensitive mutant of HIF-1␣, which is stably expressed even in normoxia, has no influence on AR transactivation in normoxia. These results indicate that HIF-1␣ is necessary but not sufficient for AR transactivation in hypoxia. Therefore, we hypothesized that HIF-1␣ needs to interact with certain proteins other than HIF-1␤/ARNT to enhance AR transactivation at a low DHT level in hypoxia. Recently, ␤-catenin, which is a ligand-dependent coactivator of AR, has been shown to interact with HIF-1␣ (20,21). ␤-Catenin is much more highly expressed in prostate cancer, compared with normal prostate tissue (4,23). In the present study, we investigated whether HIF-1␣ and ␤-catenin coordinately enhance the transcriptional activity of AR at a low DHT level in prostate cancer cells. Furthermore, we demonstrated that HIF-1␣ enhances ␤-catenin-activated N-C interaction and promotes the translocation of ␤-catenin to the nucleus. Our data show that AR forms a ternary complex with HIF-1␣ and ␤-catenin on AREs in the promoter region of androgen-responsive genes at a low DHT level in hypoxic prostate cancer cells.
Animals-Male nude (BALB/cS1c-nu/nu) mice were purchased from Japan SLC (Shizuoka, Japan). Mice were housed under temperature-and light-controlled (23 Ϯ 2°C; alternating light-dark cycles with 12 h of light and 12 h of darkness) conditions and had free access to water and food. All experimental procedures involving laboratory animals were approved by the Animal Care and Use Committee of Osaka Prefecture University.
Xenograft Tumor Implantation-Xenograft models of CRPC were generated with minor modification of the method described (24). In brief, male BALB/cS1c-nu/nu mice (5 weeks old) were castrated under anesthesia. Two weeks later, LNCaP cells (1 ϫ 10 7 ) were suspended in 100 l of RPMI1640 medium with 100 l of Matrigel and implanted in the same flank region of castrated mice under anesthesia. Tumors were harvested 7 days after inoculation of LNCaP cells.
Reporter Assay-Cells were grown on 48-well plates in phenol red-free RPMI 1640 medium supplemented with 10% dextran-coated charcoal-stripped fetal bovine serum (steroid-free RPMI 1640 medium) and transiently transfected with reporter vectors (p6xARE-TATA-Luc and pRL-SV40 (Promega)) using HilyMax reagent (Dojindo Laboratories, Kumamoto, Japan) for 24 h. After the medium was replaced with fresh steroid-free RPMI1640 medium, cells were incubated in the presence and absence of 0.1 nM DHT for 9 h in normoxia or hypoxia. The amount of DNA was kept constant by the addition of empty vector. Transfection efficiency was normalized using pRL-SV40 (control reporter vector). Cells were lysed, and firefly and Renilla luciferase activities were determined using the Dual-Luciferase reporter assay kit and GloMax 20/20 Luminometer (Promega). Data are expressed as relative light units (RLU; firefly luciferase activity divided by Renilla luciferase activity).
Mammalian Two-hybrid Assay-LNCaP and PC-3 cells were transiently transfected with pGL5-Luc (Promega), bait vector, and prey vector for 24 h, followed by incubation in the presence and absence of 0.1 nM DHT for 9 h in normoxia or hypoxia. The total amount of DNA was held by the addition of empty vector. Luciferase activity was determined, and each RLU was normalized by RLU obtained when VP16 mock vector was transfected in the absence of DHT.
Quantitative Real-time PCR-Total RNAs were extracted from the LNCaP cells. cDNAs were synthesized using reverse transcriptase and were subjected to quantitative real-time RT-PCR (qRT-PCR). The nucleotide sequences of primers for NKX3.1 and ␤-actin were described previously (19). The primers for ␤-catenin and PMEPA-1 were designed (see supplemental Table 1 for sequences). qRT-PCR was performed using GoTaq qPCR Master Mix (Promega) on a Thermal Cycler Dice real-time system (model TP-800, Takara Bio, Shiga, Japan). qRT-PCR was performed with a two-step PCR method. The relative amounts of each gene expression were calculated using the comparative Ct method, and data were normalized to the ␤-actin as an endogenous control.
Subcellular Fractionation-LNCaP cells were cultured in steroid-free RPMI 1640 medium for 72 h. The medium was replaced with fresh steroid-free RPMI 1640 medium, and cells were incubated in the presence of 0.1 nM DHT in normoxia or hypoxia for 9 h. PC-3 cells were transfected with ␤-catenin, HIF-1␣, or both expression vectors in steroid-free RPMI 1640 medium for 24 h, followed by incubation in fresh steroid-free RPMI 1640 medium supplemented with 0.1 nM DHT for 9 h in normoxia. Subcellular fractionation was performed as described previously (25). Proteins in each fraction were analyzed by Western blotting with anti-␤-catenin, anti-c-Myc, anti-HIF-1␣, anti-␣-tubulin, and anti-lamin B1 (a nuclear marker) (clone L-5, Zymed Laboratories Inc., San Francisco, CA) antibodies.
Immunofluorescent Microscopy-Immunofluorescent microscopy was performed as described previously (25). LNCaP cells were transfected with ␤-catenin, HIF-1␣, or both expression vectors in steroid-free RPMI 1640 medium for 24 h on round coverglasses on 24-well plates, followed by incubation in fresh steroid-free RPMI 1640 medium supplemented with 0.1 nM DHT for 9 h in normoxia. Cells were fixed in 4% paraformaldehyde in phosphate-buffered saline, permeated, and incubated with rabbit anti-c-Myc and mouse anti-HIF-1␣ antibodies, followed by incubation with Alexa Fluor 488-conjugated secondary anti-rabbit IgG and Alexa Fluor 594-conjugated secondary anti-mouse IgG, respectively. For the xenograft tumor sample, tissue sections were blocked in immunohistochemistry solution and incubated with rabbit anti-␤-catenin and mouse anti-HIF-1␣ antibodies, followed by incubation with Alexa Fluor 488-conjugated secondary anti-rabbit IgG and Alexa Fluor 594conjugated secondary anti-mouse IgG, respectively. The nuclei were stained with 4Ј,6-diamidino-2-phenylindole dihydrochloride (DAPI; 1 g/ml), followed by inspection using a fluorescence microscope (model BZ-9000, Keyence, Osaka, Japan).
Immunoprecipitation-LNCaP cells were cultured in steroid-free RPMI 1640 medium, followed by exposure to hypoxia in the presence of DHT (0.1 nM) for 9 h. PC-3 cells were transfected with AR, ␤-catenin, and HIF-1␣ expression vectors in steroid-free RPMI 1640 medium for 24 h, followed by incubation in fresh steroid-free RPMI 1640 medium supplemented with 0.1 nM DHT for 9 h in normoxia. Cells were lysed in hypotonic buffer (25), followed by incubation for 30 min on ice. The cell lysates were centrifuged at 20,000 ϫ g for 30 s. The precipitations were suspended in lysis buffer, followed by incubation for 20 min on ice. The cell suspension was centrifuged at 20,000 ϫ g for 20 min. The supernatant was incubated with rabbit polyclonal anti-AR IgG, anti-␤-catenin IgG, or control rabbit IgG overnight at 4°C, followed by reaction with 30 l of protein G-Sepharose (50% slurry) (GE Healthcare) resin for 2 h. The resin was washed five times with lysis buffer, and proteins bound to the resin were analyzed by Western blotting using anti-AR, anti-␤-catenin, anti-c-Myc, and anti-HIF-1␣ antibodies.
Chromatin Immunoprecipitation (ChIP)-LNCaP cells were exposed to hypoxia in the presence and absence of DHT (0.1 nM) for 9 h, followed by incubation with 1% paraformaldehyde for 15 min. Cells were lysed in 150 l of SDS buffer. For xenograft tumor sample, tissues were cut into small pieces and immediately fixed with 1% formaldehyde at room temperature for 10 min, followed by incubation with glycine to stop the cross-linking reaction. The tissues were homogenized in 500 l of SDS buffer. Cell lysates and tissue homogenates were sonicated, followed by centrifugation at 20,000 ϫ g for 3 min. The supernatants were diluted 10-fold with ChIP dilution buffer and incubated with rabbit polyclonal anti-AR IgG, anti-␤catenin IgG, anti-HIF-1␣ IgG, or control rabbit IgG overnight at 4°C, followed by incubation with 40 l of protein G-Sepharose resin (50% slurry) for an additional 2 h. Protein-DNA complexes were washed and eluted with fragments as described previously (28). The promoter regions of NKX3.1 and PMEPA-1 genes were amplified by PCR using primer sets (see supplemental Table 1 for sequences).
Re-ChIP-The protein-DNA complexes were immunoprecipitated with anti-AR IgG or anti-␤-catenin IgG according to the ChIP assay method described above and eluted with 20 l of elution buffer for 30 min at 37°C. The eluted complexes were diluted 50-fold with ChIP dilution buffer and incubated with anti-␤-catenin IgG or anti-HIF-1␣ IgG overnight at 4°C, followed by incubation with 20 l of protein G-Sepharose resin (50% slurry) for 2 h. DNA fragments were purified from complexes immunoprecipitated with anti-HIF-1␣ antibodies, and the promoter regions of NKX3.1 and PMEPA-1 genes were amplified by PCR.
In Vitro Immunoprecipitation and GST Pull-down Assay-For the immunoprecipitation assay, His-HIF-1␣DM, His-HIF-1␣⌬N-TAD, or His-HIF-1␣⌬C-TAD (2 g of each) were incubated with AR(NTD)-His (2 g) in the presence and absence of His-␤-catenin (2 g) in lysis buffer with anti-AR IgG or control rabbit IgG overnight and reacted with 30 l of protein G-Sepharose resin (50% slurry) for an additional 2 h at 4°C. The resin was washed with lysis buffer. For the GST pulldown assay, GST-AR(LBD) (2 g) was incubated with His-␤-catenin and His-HIF-1␣DM (2 g) with or without AR(NTD)-His (2 g) in GST binding buffer (50 mM Tris-HCl, pH 8.0, 100 mM NaCl, 10 mM MgCl 2 , 10% glycerol, 0.1% Nonidet P-40, 1 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride, 10 g/ml leupeptin, and 1 g/ml aprotinin) in the presence and absence of DHT (0.1 nM) overnight and reacted with 30 l of glutathione-Sepharose 4B resin (50% slurry) for an additional 2 h at 4°C. The resin was washed with GST binding buffer. Proteins bound to the resin were separated by SDS-PAGE, followed by Western blotting using anti-AR, anti-HIF-1␣, anti-␤-catenin, and anti-GST antibodies. The immunoreactive proteins were detected using the Immobilon Western Chemiluminescent HRP substrate and exposed to a luminescent image analyzer.
Statistics-Data were assessed by two-or three-way analyses of variance with Turkey's post hoc testing. Statistical analysis was performed using JMP statistical software ver-
␤-Catenin and HIF-1␣ Are Involved in AR N-C Interaction in Hypoxia-To determine whether hypoxia-induced AR transactivation results from enhancement of the N-C interaction of AR, VP16-AR(NTD) and GAL4DBD-AR(LBD) were co-expressed in LNCaP cells. Hypoxia enhanced the DHT-dependent N-C interaction of AR ( Fig. 2A). Furthermore, we assessed whether HIF-1␣ or ␤-catenin is required for the hypoxia-enhanced N-C interaction of AR. siRNA-mediated knockdown of either HIF-1␣ or ␤-catenin canceled the hypoxic induction of AR N-C interaction (Fig. 2B). In addition, when ␤-catenin and HIF-1␣ were targeted with each shRNA, hypoxic induction of N-C interaction was canceled (supplemental Fig. 2). Exogenous ␤-catenin, but not HIF-1␣DM, activated the N-C interaction of AR, and HIF-1␣DM enhanced ␤-catenin-activated AR N-C interaction (Fig. 2C). In GST pull-down assays, the physical interaction of GST-AR(LBD) with AR(NTD) occurred in the presence of DHT (Fig. 2D, left panels). The ligand-dependent interaction between AR(NTD) and GST-AR(LBD) was increased in the presence of ␤-catenin and was further increased in the presence of both ␤-catenin and HIF-1␣ (Fig.  2D, right panels). In contrast, HIF-1␣ had no influence on the ligand-dependent interaction in the absence of ␤-catenin. These results indicate that hypoxia induces the AR N-C interaction and that HIF-1␣ and ␤-catenin enhance the N-C interaction through physical interaction of AR with at least ␤-catenin.
HIF-1␣ Increases Nuclear Accumulation of ␤-Catenin in Hypoxia-To determine the effect of hypoxia on intracellular localization of ␤-catenin in vivo, LNCaP xenografts were generated in castrated nude mice. ␤-Catenin was localized in the nucleus in the HIF-1␣ protein-positive cells of LNCaP xenografts (Fig. 3A, top panels). On the other hand, ␤-catenin was distributed in the cytoplasm in the HIF-1␣ protein-negative cells (Fig. 3A, bottom panels). Next, we investigated the effect of HIF-1␣ on the localization of ␤-catenin in hypoxic LNCaP cells by subcellular fractionation. Hypoxia resulted in an increased accumulation of ␤-catenin in the nucleus, and knockdown of HIF-1␣ canceled hypoxia-induced nuclear accumulation of ␤-catenin (Fig. 3B). HIF-1␣ is synthesized in the cytosol and is translocated to the nucleus in hypoxia (15). To assess whether nuclear translocation of HIF-1␣ is involved in the increased nuclear accumulation of ␤-catenin in hypoxia, we constructed a mutant form of HIF-1␣, which was constitutively localized in the cytoplasm, termed HIF-1␣(K719T). Although HIF-1␣DM increased the expression level of ␤-catenin in the nucleus, HIF-1␣(K719T) had no influence on the nuclear accumulation of ␤-catenin (Fig. 3C). Immunofluorescence microscopy showed that exogenous ␤-catenin was distributed throughout the cytoplasm (Fig. 3D, top panels). However, when co-expressed with HIF-1␣DM, ␤-catenin accumulated in the nucleus (Fig. 3D, middle panels). Exogenous HIF-1␣(K719T) had no influence on the cytoplasmic localization of ␤-catenin (Fig. 3D, bottom  panels). An immunoprecipitation assay revealed that not only HIF-1␣DM but also HIF-1␣(K719T) formed a complex with ␤-catenin (Fig. 3E). Furthermore, HIF-1␣(K719T) inhibited hypoxia-enhanced AR transactivation (Fig. 3F). On the other hand, AR interacted with HIF-1␣DM, but not with HIF-1␣(K719T) (Fig. 3E), and hypoxia had no influence on the nuclear accumulation of AR (Fig. 3G). These results indicate that HIF-1␣ forms a complex with ␤-catenin in the cytosol and promotes the nuclear translocation of ␤-catenin, but not of AR, in hypoxia.
Hypoxia Increases the Amount of a Complex Composed of AR and ␤-Catenin in the Nucleus-To examine the effect of hypoxia on the association between AR and ␤-catenin or HIF-1␣ in the nucleus, the nuclear fraction was prepared from hypoxic LNCaP cells. Immunoprecipitation assays using anti-AR or anti-␤-catenin IgG showed that hypoxia increased the association between AR and ␤-catenin in the nucleus and that HIF-1␣ formed a complex with AR or ␤-catenin in hypoxia (Fig. 4A). Furthermore, to assess whether AR, ␤-catenin, or HIF-1␣ binds to AREs on the androgen-responsive genes (NKX3.1 and PMEPA-1), LNCaP cells were exposed to hypoxia, and ChIP assays were performed using four sets of PCR primers as shown in Fig. 4B. When the PCR products were amplified using the primer set P1 or P4, but not the primer set P2 or P3, hypoxia increased the level of AR or ␤-catenin on AREs and resulted in recruitment of HIF-1␣ to the AREs (Fig. 4C). To examine whether AR is involved in the association of ␤-catenin or HIF-1␣ with AREs on the NKX3.1 and PMEPA-1 genes, LNCaP cells were transfected with AR siRNA, followed by exposure to hypoxia. AR siRNA specifically reduced the AR level and had no influence on the ␤-catenin and HIF-1␣ levels (supplemental Fig. 3). Knockdown of AR abolished the association of ␤-catenin or HIF-1␣ with AREs in hypoxia (Fig. 4D,  left). Furthermore, siRNA-mediated knockdown of HIF-1␣ attenuated the association of AR or ␤-catenin with AREs in hypoxia (Fig. 4D, right). These results indicate that hypoxia increases the formation of a complex composed of AR and ␤-catenin in the nucleus and that HIF-1␣ is involved in an increased association of AR or ␤-catenin with AREs in hypoxia.
AR, HIF-1␣, and ␤-Catenin Form a Ternary Complex in Hypoxia-When recombinant AR(NTD) was incubated with ␤-catenin in the presence and absence of HIF-1␣, ␤-catenin co-immunoprecipitated with AR(NTD) in the presence but not in the absence of HIF-1␣DM (Fig. 6A). When recombinant GST-AR(LBD) was incubated with HIF-1␣DM in the presence and absence of ␤-catenin, ␤-catenin was pulled down with GST-AR(LBD) in the presence of DHT (Fig. 6B), consistent with the fact that ␤-catenin physically interacts with the LBD of AR in a ligand-dependent manner (21). On the other hand, recombinant HIF-1␣DM was pulled down with GST-AR(LBD) in the presence but not in the absence of ␤-catenin in a liganddependent manner. Next, we determined whether AR, ␤catenin, and HIF-1␣ forms a complex on the AREs by ChIP and re-ChIP assays. The ChIP assays showed that hypoxia increased the levels of AR and ␤-catenin on the AREs in the presence of DHT (Fig. 6C, left panels). Subsequently, a re-ChIP assay, which is a sequential ChIP assay with anti-HIF-1␣ IgG, was performed using anti-␤-catenin IgG-immunoprecipitated products. HIF-1␣ bound to the AREs with ␤-catenin in the presence of DHT in hypoxia (Fig. 6C, right panels). Furthermore, we assessed whether AR, ␤-catenin, and HIF-1␣ form a complex on AREs in LNCaP xenografts by ChIP and re-ChIP analyses. AR interacted with HIF-1␣ or ␤-catenin on the AREs (Fig. 6D, left panels), and ␤-catenin formed a complex with HIF-1␣ on the AREs (Fig. 6D, right panels). These results indicate that an endogenous ternary complex containing AR, ␤-catenin, and HIF-1␣ binds to the chromatin of androgen-responsive genes in the presence of DHT (0.1 nM) in hypoxia.

DISCUSSION
Prostate cancer eventually switches from a hormone-sensitive state to a hormone-refractory state after androgen ablation and consequently progresses to CRPC, which grows even in a low androgen environment (29). In CRPC, the AR signaling remains active and is proposed to be reactivated in several ways, including (i) increased AR expression, (ii) AR gene mutations leading to enhanced responsiveness to ligand or induced independence of ligand, and (iii) alterations of AR co-activators or co-repressors (30). However, the molecular mechanisms underlying the development and progression of CRPC are poorly understood. Our previous studies demonstrated that hypoxia enhances AR transactivation at low androgen levels and that HIF-1␣ is necessary but not sufficient for its AR transactivation (19). In the present study, knockdown of ␤-catenin canceled hypoxia-induced AR transactivation at a low level of DHT, and HIF-1␣ enhanced ␤-catenin-activated AR transactivation. ␤-Catenin is much more highly expressed in prostate cancer than normal prostate tissue and acts as a co-activator of AR (21,31). On the other hand, ␤-catenin acts with T-cell factor (TCF) to up-regulate the expression of certain genes like c-Myc and cyclin D1 that are related to tumor growth (32)(33)(34). Hypoxia inhibits the interaction of ␤-catenin with TCF, and consequently ␤-catenin binds to HIF-1␣ (20). Thus, hypoxia inhibits ␤-catenin-mediated TCF signaling. In addition, AR inhibits TCF activity in a ligand-dependent manner (35). These results indicate that, rather than activating the transcriptional activity of TCF in hypoxia, ␤-catenin together with HIF-1␣ enhance the transcriptional activity of AR under low androgen environmental conditions and that HIF-1␣ enhances the function of ␤-catenin in AR transactivation.
HIF-1␣ enhanced the ␤-catenin-activated N-C interaction of AR, and HIF-1␣ and ␤-catenin coordinately increased the physical interaction between NTD and LBD of AR. The N-C interaction is essential for the transcriptional activity of AR (9). Although some of the AR coactivators, such as SRC-1 and ARA24 (but not GAPDH and ARA70), enhance the AR N-C interaction, their mechanisms remain unclear (10,25,36). HIF-1␣, but not HIF-1␣⌬C-TAD, physically interacted with the NTD of AR, suggesting that HIF-1␣ binds to the NTD of AR through its C-TAD. ␤-Catenin interacted with the NTD of AR in the presence but not in the absence of HIF-1␣ (Fig. 6A). The six N-terminal armadillo repeat domains of ␤-catenin physically interact with the LBD of AR, and the C-terminal region of ␤-catenin physically interacts with the N-terminal region of HIF-1␣ (20,21). These results provide a model structure in which HIF-1␣ and ␤-catenin physically bind to the NTD and LBD of AR, respectively, to increase the N-C interaction of AR (Fig. 7).
Knockdown of HIF-1␣ canceled the hypoxia-increased accumulation of ␤-catenin in the nucleus. ␤-Catenin formed a complex with HIF-1␣DM and HIF-1␣(K719T), indicating that a complex composed of ␤-catenin and HIF-1␣ exists in both the nucleus and cytoplasm. Furthermore, ␤-catenin was located in the nucleus of cells expressing HIF-1␣ protein in LNCaP xenografts in castrated nude mice, providing clear evidence that supports the fact that castration-resistant cancer cells highly express HIF-1␣ or ␤-catenin proteins in the nucleus (15,37). On the other hand, AR interacted with HIF-1␣DM but not with HIF-1␣(K719T) (Fig. 3E), indicating that AR forms a complex with nuclear HIF-1␣ but not with cytoplasmic HIF-1␣. In hypoxia, endogenous HIF-1␣ interacts with AR in the presence of DHT but not in the absence of DHT in LNCaP cells (38). Because the ligand-bound and ligand-free ARs are localized in the nucleus and cytoplasm, respectively, AR and HIF-1␣ exhibit a different subcellular localization in the absence of DHT and meet up in the nucleus in the presence of DHT. In contrast, in normoxia, proteins such as LEF and AR support the transloca- Double-headed arrows indicate the DNA regions that were amplified using PCR primer sets designated as P1, P2, P3, and P4. C, LNCaP cells were exposed to normoxia (N) and hypoxia (H) in the presence of DHT, followed by cross-linking. Immunoprecipitated protein-DNA complexes were analyzed by PCR using each primer set. D, LNCaP cells were treated with siAR (left), siHIF-1␣ (right), or siControl and were exposed to hypoxia in the presence of DHT. Immunoprecipitated protein-DNA complexes were analyzed by PCR using each primer set. In all experiments, the result is representative of three independent experiments. tion of ␤-catenin from the cytosol to the nucleus (39,40), indicating that the mechanisms governing the nuclear translocation of ␤-catenin are different between normoxia and hypoxia. These results indicate that HIF-1␣ promotes the translocation of ␤-catenin from the cytosol to the nucleus under hypoxic conditions in CRPC and that AR forms a complex with ␤-catenin in the nucleus in the presence of HIF-1␣ (i.e. in hypoxia). In addition, because the N-C interaction of AR occurs as an intramolecular event within the AR monomer both in the cytosol and nucleus and as an intermolecular event in the AR homodimer in the nucleus (6,41), it is suggested that the com-plex composed of HIF-1␣ and ␤-catenin contributes to intraand/or intermolecular AR N-C interaction in the nucleus.
Hypoxia enhanced the recruitment of AR to AREs in the promoter regions of NKX3.1 and PMEPA-1 genes, consistent with the fact that hypoxia increases the ARE-binding activity of AR (42). Furthermore, in LNCaP xenografts, AR, ␤catenin, and HIF-1␣ were recruited on the AREs (Fig. 6D). Hypoxia increased the formation of a complex composed of AR and ␤-catenin in the nucleus, and the physical interaction of AR with ␤-catenin was increased in the presence of HIF-1␣ (Fig. 2D, right panels). In contrast, knockdown of HIF-1␣ attenuated the association of AR and ␤-catenin with the AREs in hypoxia. Furthermore, hypoxia enhanced the N-C interaction of AR. The N-C interaction is generally required for AR to bind to chromatin (43). These results suggest that hypoxia increases the binding of AR to the AREs by enhancing the N-C interaction. AR binds to ARE in a ligand-dependent manner, and ␤-catenin is recruited on the AREs through AR (21,22). Our data demonstrate that a ternary complex composed of AR, HIF-1␣, and ␤-catenin forms on AREs in a HIF-1␣-dependent manner in the presence of DHT.
The AR signaling pathway plays a dominant role in the progression of CRPC, and hypoxia enhances AR transactivation at a low androgen level. The present study reveals that HIF-1␣ enhances ␤-catenin-activated AR transactivation in hypoxia. HIF-1␣ increases the expression of hypoxia-responsive genes, including tumor growth-or survival-related genes in hypoxia. Therefore, HIF-1␣ acts not only as a transcriptional factor for HIF-1 activity but also as a cofactor that supports ␤-catenin in the transactivation of AR. This suggests that HIF-1␣ is a key factor in both the HIF-1 signaling and AR signaling pathways in CRPC.
Acknowledgment-We thank Yusuke Dairyo for helpful technical support. FIGURE 6. Formation of a ternary complex by AR, HIF-1␣, and ␤-catenin. A, recombinant AR(NTD)-His, His-HIF-1␣DM, and His-␤-catenin were incubated with anti-AR IgG or control rabbit IgG. Immunoprecipitated (IP) proteins were subjected to Western blotting. B, recombinant GST-AR(LBD), His-HIF-1␣DM, and His-␤-catenin were incubated in the presence and absence of DHT. Proteins pulled down with glutathione-Sepharose 4B resin were analyzed by Western blotting. C, LNCaP cells were exposed to normoxia (N) or hypoxia (H) in the presence and absence of DHT, followed by cross-linking. Soluble chromatin was immunoprecipitated with anti-AR IgG, anti-␤-catenin IgG, or control rabbit IgG (left). The protein-DNA complexes immunoprecipitated with anti-␤-catenin IgG were reimmunoprecipitated with anti-HIF-1␣ IgG or control mouse IgG (right). Reimmunoprecipitated protein-DNA complexes were analyzed by PCR. D, LNCaP xenografts were cross-linked. Soluble chromatin was immunoprecipitated with anti-AR IgG, anti-␤-catenin IgG, anti-HIF-1␣ IgG, or control rabbit IgG. The protein-DNA complexes immunoprecipitated with anti-AR IgG were reimmunoprecipitated with anti-␤-catenin, anti-HIF-1␣ IgG, or control mouse IgG (left), and the protein-DNA complexes immunoprecipitated with anti-␤-catenin IgG were reimmunoprecipitated with anti-HIF-1␣ IgG or control mouse IgG (right). Protein-DNA complexes were analyzed by PCR. In all experiments, the result is representative of three independent experiments.

ARE
Target genes FIGURE 7. Schematic model for the mechanism by which HIF-1␣ and ␤-catenin accelerate the N-C interaction of AR. In hypoxia, HIF-1␣ binds to ␤-catenin in the cytoplasm and promotes nuclear translocation of ␤-catenin. Subsequently, HIF-1␣ and ␤-catenin bind to NTD and LBD of AR, respectively, in the nucleus and enhance the ligand-induced N-C interaction of AR.