PIAS1 and PIASxα Function as SUMO-E3 Ligases toward Androgen Receptor and Repress Androgen Receptor-dependent Transcription*

The androgen receptor (AR) has been shown to be modified by SUMO-1, a ubiquitin-like protein. Recently we showed that PIAS family proteins function as SUMO-E3 ligases. Here we provide evidence that PIAS1 and PIASxα act as specific SUMO-E3 ligases for the AR. PIAS1 and PIASxα but not PIAS3 or PIASxβ enhanced the sumoylation of AR in intact cells and in vitro. PIAS1 and PIASxα bound Ubc9, the E2 enzyme for SUMO-1, in a RING finger-like domain-dependent manner. Consistent with previous studies (Kahyo, T., Nishida, T., and Yasuda, H. (2001)Mol. Cell 8, 713–718), the RING finger-like domain of the SUMO-E3 was required for ligase activity. The binding of a ligand, e.g. testosterone, to the AR was required for the sumoylation of AR in intact cells. Although AR-dependent transcription was enhanced by PIAS proteins without sumoylation of the receptor, PIAS1 and PIASxα repressed AR-dependent transcription in a manner dependent on the ectopic expression of SUMO-1 and their RING finger-like domain. Furthermore, the sumoylation sites of the AR were necessary for the full repressive effect on AR-dependent transactivation, indicating that the sumoylation of AR was crucial for the repression of transactivation of the AR. Thus, PIAS1 and PIASxα modulate the AR-dependent transactivation, which, at least in part, can be attributed to their SUMO-E3 activity toward AR.

The androgen receptor (AR) 1 belongs to the steroid hormone nuclear receptor superfamily (1). The AR remains in the cytoplasm until it is activated by ligand binding. Upon ligand binding, this receptor dissociates from its heat-shock protein chaperones, becomes dimerized, and is translocated into the nucleus where it binds to specific androgen response elements (AREs) to regulate, along with other transcription factors, the transcription of its target genes. The AR plays important roles in male sexual development, prostate cell proliferation, and the progression of prostate cancer.
The AR has been shown to be modified by small ubiquitinlike modifier 1, SUMO-1 (sumoylation), in vivo (2). Lysine residues at amino acid positions 386 and 520 of the human AR (hAR) are major sumoylation sites. Mutation of these residues blocks sumoylation and increases the transactivation ability of AR, suggesting that SUMO modification negatively regulates AR activity. However, the precise function of the sumoylation of AR remains unknown. Also SUMO-E3 ligase toward AR has not been identified.
SUMO-1 has been shown to conjugate to target proteins through an isopeptide linkage between a glycine residue in the terminus and the ⑀-amino group of a lysine residue of the target protein. Sumoylation has multiple functions that include involvement in protein targeting, stabilization, and transcriptional regulation. Sumoylation of RanGAP1 results in movement of the protein from the cytoplasm to the nuclear pore complex (3)(4)(5). In the case of PML (promyelocytic leukemia) proteins, sumoylation regulates their subnuclear localization to structures termed PML nuclear bodies (6 -9). Sumoylation of inhibitor of B␣ acts antagonistically to ubiquitinylation and protects the sumoylated molecule itself from ubiquitin-mediated proteolysis (10). In addition, sumoylation of p53 has been proposed to regulate the transcriptional activity of p53 (11,12).
Sumoylation of target proteins seems to occur in a manner analogous to the ubiquitinylation reaction. Heterodimeric E1 enzyme (Sua1/Uba2) and E2 enzyme (Ubc9) have already been detected by us and by other groups (13)(14)(15)(16)(17)(18), and recently we identified PIAS1 as a SUMO-E3 ligase toward p53 (19). Also, we and other groups have shown yeast Siz1/Ull1 protein to have SUMO-1/Smt3-E3 activity (20,21). PIAS1 and Siz1/Ull1 share significant homology in their critical RING finger-like domain (SP-RING) (22). More recently, PIASy was shown to markedly stimulate the sumoylation of LEF1 and multiple other proteins in vivo and to function as a SUMO-E3 ligase toward LEF1 in vitro (23). PIAS1 and PIAS3 were originally discovered as transcriptional co-regulators of the Janus kinase-STAT pathway (24,25). The human PIAS family consists of several homologous proteins, including PIAS1, PIAS3, PIASx␣, PIASx␤, PIASy, and a hypothetical protein (GenBank TM accession no. CAB66507); and all of them have a RING finger-like domain, where two consensus cysteine residues for the family seem to have been replaced by serine and asparatic acid (19,24,25). PIAS proteins interact with their target proteins, and the RING finger-like domain of each PIAS protein is necessary for the SUMO-E3 ligase activity (19,21,22). These findings suggest that PIAS family proteins act as SUMO-E3 enzymes toward cellular targets, although which member of this family is specific for which substrate remains to be elucidated. PIAS1 was shown to interact with the AR (26). Rat ARIP3 (androgen receptor interacting protein 3), a protein found to bind to AR (27), is an ortholog of human PIASx␣. These PIAS proteins influence AR-dependent transcription. Furthermore, PIAS1 and ARIP3/PIASx␣ are highly expressed in the testis (26,27). These findings raise the possibility that PIAS1 and ARIP3/ PIASx␣ may function as SUMO-E3 ligases toward the AR.
Here we present evidence that both PIAS1 and PIASx␣ act as specific SUMO-E3 ligases for AR in intact cells as well as in vitro and that sumoylation of the AR represses AR-dependent transcriptional activation.
Cell Culture and DNA Transfections-U2OS human osteosarcoma cells were maintained in Dullbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS). The cells were transfected by using FuGENE 6 reagent (Roche) according to the manufacturer's instructions. For immunoblotting, cells were plated in 35-mmdiameter culture plates (ϳ1 ϫ 10 5 cells/plate) 24 h before transfection, and 4 h before transfection the cells received fresh medium with 10% charcoal-stripped FBS. At 8 h after transfection, the medium was changed to DMEM containing 2% charcoal-stripped FBS in the presence or absence of 100 nM testosterone. The cells were harvested 24 h after transfection. For reporter activity assays, cells were plated in 12-well plates (ϳ3 ϫ 10 4 cells/well) 24 h before transfection, and 4 h before transfection the cells received fresh medium with 10% charcoalstripped FBS. Transfections were performed by using FuGENE 6 reagent. At 24 h after transfection, the medium was changed to DMEM containing 2% charcoal-stripped FBS with or without 100 nM testosterone. The cells were harvested an additional 24 h later.
Immunoblotting and Immunoprecipitation-Cells were washed with ice-cold phosphate-buffered saline and directly lysed with SDS-sample buffer. For immunoprecipitation, the cells were lysed in lysis buffer (50 mM Tris-HCl, pH 7.4, containing 150 mM NaCl, 1% sodium deoxycholate, 1% Triton X-100, 0.1% SDS, and 5 mM N-ethylmaleimide) supplemented with protease inhibitors (Roche). After brief sonication, the lysates were cleared by centrifugation at 4°C. Supernatants were incubated with mouse monoclonal anti-human AR antibody (Santa Cruz Biotechnology. Inc.) and protein A-Sepharose CL-4B (Amersham Biosciences) for 4 h at 4°C. The immunocomplexes were washed 5 times with the same buffer, and the immunoprecipitated proteins were removed from the protein A beads by boiling in SDS-sample buffer. Proteins were subjected to SDS-PAGE and then transferred to polyvinylidene difluoride membranes. Each membrane was treated with primary antibody and then with horseradish peroxidase-conjugated secondary antibody. The proteins that reacted with the primary antibody were detected by the ECL method (Amersham Biosciences) using x-ray film.
In vitro Sumoylation Assay-The in vitro sumoylation assay was performed in a 30-l reaction mixture containing 50 mM Tris-HCl, pH 7.4, 2 mM dithiothreitol, 3 mM ATP, 100 ng of purified GST-hAR (either the wild-type one or K386R/K520R), 0.36 g of purified E1 (GST-Sua1/ His-Uba2), 20 ng or 500 ng of Ubc9, 500 ng of SUMO-1, and 100 ng of GST-PIAS proteins. Ubc9 and SUMO-1 were expressed in Escherichia coli (BL21, LysS) by using a pET vector. E1, hAR, and PIAS proteins were expressed in Sf-9 cells by use of the baculovirus protein expression system (Invitrogen). Purification of enzymes was performed as described previously (14). After incubation for 30 min in the presence or absence of 1 M testosterone, the reaction mixture was subjected to SDS-PAGE and immunoblotting with a monoclonal anti-hAR antibody (Santa Cruz Biotechnology, Inc.).
Yeast Two-hybrid Assay-Wild-type PIAS1 and PIASx␣ and their mutants (C350A and C362A, respectively) were cloned into pGAD424 vector (Clontech) as GAL4 activation domain fusion constructs. SUMO-1, hUbc9, and full-length hAR cDNAs were cloned into pAS2-1 vector (Clontech) for fusion in-frame with the GAL4 DNA-binding domain. The two-hybrid assay was performed with Y190 cells transformed with a pGAD424-PIAS1 (wild-type or C350A) or pGAD424-PIASx␣ (wild-type or C362A) and a bait plasmid expressing the GAL4 DNAbinding domain fused to the indicated protein. All the transformants were plated on selective medium lacking tryptophan, leucine, and histidine but containing 25 mM 3-amino-1,2,4-triazole to test cell growth. The colonies positive on the histidine-deficient plate were further confirmed by using a ␤-galactosidase filter assay.
Luciferase Reporter Assays-Reporter plasmid pGV-ARE2-TATA-LUC expressing firefly luciferase under the control of the ARE (150 ng), pRL-CMV (Wako) expressing Renilla luciferase under the control of the cytomegalovirus (CMV) promoter (50 ng) as an internal control, pFLAG-hAR (wild-type or K386R/K520R; 100 ng), pFLAG-SUMO-1 (GG or ⌬GG; 200 ng), and pEGFP-PIAS1 (wild-type or C350A; 100 -500 ng) or pEGFP-PIASx␣ (wild-type or C362A; 100 -500 ng) were used for transfection. The total amount of transfected DNA was kept constant (1 g) by the addition of empty vectors. Luciferase activities were determined in extracts made from transfected cells by using a PicaGene Dual Luciferase Reporter Assay System (Toyo Ink) and a Lumat LB9501 luminometer (Berthold). The luciferase activities were normalized to the Renilla luciferase activity from the pRL-CMV vector. The results were presented as the average and standard deviation for relative luciferase activity from at least three independent experiments. The value observed for the expression of wild-type hAR alone in the presence of testosterone was assigned an arbitrary value of 1.

PIAS1 and PIASx␣ Specifically Enhanced the Sumoylation of AR in Intact Cells-
The hAR has been reported to interact with PIAS1 and PIASx␣ and to be modified by SUMO-1. Recently, we and other groups showed that PIAS family proteins have SUMO E3-like activity. Thus, the interaction between the AR and PIAS proteins raises the possibility that PIAS proteins might function as SUMO-E3 ligases for this receptor. Although PIAS proteins were shown to act as a co-regulator (26 -29), direct regulation of the sumoylation of the AR by PIAS proteins has not been reported. Therefore, we investigated whether the sumoylation of AR could be enhanced by PIAS proteins in intact cells. U2OS cells were made to co-express GFP-hAR, FLAGx2-tagged SUMO-1, and GFP-PIAS proteins in the absence or presence of testosterone. Then the sumoylation of AR was checked by immunoblotting or by immunoprecipitation of the cell lysate with an anti-AR antibody and incubating the blotted proteins with anti-AR or anti-FLAG antibody. Using cell extracts for immunoblotting with the anti-hAR, we detected GFP-hAR protein bands at 130 kDa, whereas the endogenous hAR was hardly detected in U2OS cells (data not shown). In the case of no expression of PIAS proteins, no bands representing modified AR were detected in the presence or absence of testosterone (Fig. 1A, lanes 1 and 2). In contrast, when PIAS1 or PIASx␣ was co-expressed, some additional slowermigrating bands representing the sumoylated AR appeared in the presence of testosterone. Only one faint slower-migrating band indicating the sumoylated hAR at one site was induced by PIAS1 in the absence of testosterone (Fig. 1A, lane 3). However, sumoylation of the hAR was strongly stimulated in a liganddependent and in an expression of PIAS1 or PIASx␣-dependent manner (Fig. 1A, lanes 4 and 8), whereas the expression of neither PIAS3 nor PIASx␤ enhanced the sumoylation of the receptor (Fig. 1A, lanes 5, 6, 9, and 10). Consistent with previously published results (2), the hAR mainly was sumoylated at two sites. However, we also noted other less intense bands. Next, to confirm that these AR antibody-reactive, slower-migrating bands were indeed sumoylated hARs, we transfected U2OS cells with the same combination of plasmids as in the previous experiment. Immunoprecipitation of the lysates prepared from the transfected cells was performed with anti-AR antibody, and then immunoblotting was conducted with anti-FLAG antibody or anti-AR antibody (Fig. 1, B and C, respectively). When anti-FLAG antibody was used for the detection, FLAGx2-SUMO-1-conjugated GFP-hAR bands were detected at the same place on the gel as in Fig. 1A. Therefore, we considered these bands to represent sumoylated hARs. In this experiment, additional bands of hAR that had shifted to a higher molecular weight than the former band, representing hAR sumoylated at two sites, were also detected. Overexpressed PIAS1 or PIASx␣ may induce sumoylation at other sites or polysumoylation, analogous to polyubiquitination. When cells expressed K386R/K520R, the hAR with 2 mutated sumoylation sites, no sumoylated bands were detected in the absence or presence of testosterone (Fig. 1C, lanes 7-10). These data thus indicate that the slower-migrating bands were sumoylated hARs. Under the condition of no ectopical expression of SUMO-1, the sumoylation of the hAR was also enhanced in the presence of testosterone, although the enhancement was weaker than that in the presence of FLAGx2-SUMO-1 (Fig. 1B,  lanes 8 and 10). Furthermore, when cells expressed the PIAS1 mutant (C350A) or PIASx␣ mutant (C362A), the sumoylation of hAR was only a little enhanced in the presence of C350A or C362A compared with that in the absence of PIAS proteins. The enhancement was very much smaller than that in the presence of wild-type PIAS1 or PIASx␣ ( Fig. 2A). This result is much the same as that found previously with p53 used as the substrate for sumoylation. Thus, mutations that interfere with Ubc9 prevent the enhancement of the sumoylation (described below). Taken together, the above data obtained from the transfection assay system clearly indicate that PIAS1 and PI-ASx␣ enhanced the sumoylation of hAR in intact cells in a ligand-dependent and RING finger-like domain-dependent manner. PIASx␣ and PIASx␤ are thought to be alternatively spliced forms from the same gene because their amino acids 1-550 are identical but their C termini are different from each other. The C-terminal deletion mutant PIASx (1-550) could not enhance the sumoylation of hAR (Fig. 2B), indicating that the C-terminal domain (amino acids 551-572) of PIASx␣ is essential for its SUMO-E3 ligase activity toward hAR.
PIAS1 and PIASx␣ Act as SUMO-E3 Ligases Toward AR in Vitro-Next, we conducted in vitro sumoylation by using purified recombinant hAR proteins as the substrate (Fig. 3). To confirm that PIAS1 and PIASx␣ indeed have SUMO-E3 ligase activity toward the hAR, we prepared and purified the recombinant proteins and assayed their activities. GST-tagged PIAS1, PIASx␣, and hAR (either the wild-type or the K386R/ K520R mutant) were expressed in Sf-9 cells, purified by use of glutathione-Sepharose 4B, and then used for the sumoylation assay in the presence of purified recombinant E1 (Sua1 and hUba2 heterodimer) and E2 (hUbc9). In the absence of PIAS proteins, no sumoylated hAR bands were detected. When a 25ϫ greater amount of hUbc9 than normal was used, only one sumoylated band was detected slightly in this assay system (Fig. 3, lane 2). In contrast, the addition of PIAS1 or PIASx␣ to the assay system enhanced the sumoylation of the hAR. Interestingly, these reactions were independent of the addition of testosterone, indicating that the reaction does not require a conformational change in the receptor by ligand binding in vitro. When reactions were performed in the absence of hUbc9, no enhancement of the sumoylation was observed (Fig. 3, lanes  5 and 8) 1. PIAS1 and PIASx␣ enhance the sumoylation of AR in intact cells. U2OS cells were co-transfected with expression plasmids encoding GFP-hAR (wild-type or K386R/K520R mutant), FLAGx2-SUMO-1, and the indicated GFP-PIAS proteins in the presence or absence of 100 nM testosterone. Twenty-four hours post-transfection the cells were harvested. A, the cells were lysed with SDS-PAGE sample buffer, and the lysates were resolved by SDS-PAGE and analyzed by immunoblotting using a monoclonal anti-AR (upper) or a polyclonal anti-GFP antibody (lower). B, the cells were lysed in lysis buffer, and the lysates were immunoprecipitated with a monoclonal anti-AR antibody. The immunocomplexes were then analyzed by immunoblotting with a polyclonal FLAG antibody (upper) or a monoclonal AR antibody (lower). C, the cells were lysed and subjected to immunoblotting as in A. seemed to enhance the sumoylation at the same sites both in vivo and in vitro.
These data described above strongly suggest that PIAS1 and PIASx␣ function as SUMO-E3 ligases toward the hAR.
Wild-type PIAS1 and PIASx␣, but Not Their RING Fingerlike Domain Mutants, Interact with Ubc9 in the Yeast Twohybrid Assay-By use of the yeast two-hybrid assay, we previously showed that PIAS1 bound to hUbc9 but that when a cysteine residue in the RING finger-like domain was mutated to alanine, the mutant could not bind to hUbc9. Therefore, we also examined the interaction between the wild-type or RING finger-like domain mutant of PIASx␣ with hUbc9 by using the same assay. As shown in Fig. 4, wild-type PIASx␣ interacted with hUbc9. In contrast, the C362A mutant could not bind to hUbc9.
These data suggest that the interaction between PIAS proteins and Ubc9 through the RING finger-like domain of these proteins was necessary for the SUMO-E3 ligase activity of PIAS.
PIAS1 and PIASx␣ Specifically Repress the AR-dependent Transactivation in the Presence of Ectopically Expressed SUMO-1-PIAS family proteins have been shown to influence AR-dependent transcription. It has been suggested that the sumoylation negatively regulates AR activity because the mutation of sumoylation sites increases the transactivation ability of AR. To investigate the influence of the sumoylation mediated by PIAS1 or PIASx␣ on AR-dependent transactivation, we performed transient transfection assays using U2OS cells transfected with a luciferase reporter having AREs in the promoter. As shown in Fig. 5A, in the absence of PIAS proteins the wild-type hAR increased the reporter gene expression level by 10-fold upon the addition of testosterone. When PIAS1 was co-expressed with the wild-type hAR, this ligand-dependent activation was slightly increased. Co-expression of PIASx␣ with wild-type hAR further enhanced the luciferase activity. In contrast, co-expression of the wild-type hAR with PIAS1 or PIASx␣ and SUMO-1(GG) strongly reduced the ligand-dependent activation. However, co-expression of SUMO-1(⌬GG), which could not be conjugated to target proteins, had no effect on the ligand-dependent hAR activation in the presence of PIAS1 or PIASx␣. PIAS3 or PIASx␤, which had shown no SUMO-E3 ligase activity toward hAR, strongly enhanced the transactivation activity in a dose-dependent manner. Interestingly, co-expression of SUMO-1(GG) did not reduce the activation by PIAS3 or PIASx␤ (Fig. 5A, right panel). Ectopic expression of SUMO-1 enhanced the sumoylation of hAR by PIAS1 and PIASx␣ in U2OS cells (Fig. 1B). These data suggest that the sumoylation of hAR may repress the ligand-dependent hAR activity.
Sumoylation Sites of hAR Are Required for the Repressive Effect on Ligand-dependent Activation-To further investigate the role of sumoylation of the AR, we used the K386R/K520R mutant, lacking the two sumoylation sites of hAR, in the same reporter assay. As shown in Fig. 5B, in the absence of PIAS proteins, the K386R/K520R mutant hAR increased the reporter gene expression level 2-fold over that of the wild-type hAR upon the addition of testosterone. This result is consistent with a previously published finding (2). At the amounts indicated in Fig. 5, PIAS1 or PIASx␣ alone did not influence this ligand-dependent transactivation. In contrast to the results obtained with the wild-type hAR, however, co-expression of PIAS1 or PIASx␣ with SUMO-1(GG) still repressed the transactivation of the K386R/K520R mutant hAR. The repression level of the transactivation of K386R/K520R, however, was weaker than that of the wild-type hAR. These results suggest that sumoylation is required for the full repression of the AR-mediated transactivation.
RING Finger-like Domain Is Required for the Repression of the AR-dependent Transactivation-Next, we examined whether the RING finger-like domain mutant of PIAS1 or PIASx␣, which did not enhance the sumoylation of hAR in intact cells, could repress the ligand-dependent hAR activation. As shown in Fig. 5C, co-expression of the wild-type hAR with the PIAS1 C350A or PIASx␣ C362A mutant could also enhance the ligand-dependent transactivation. Interestingly, PIAS1 C350A decreased this enhancement in a dose-dependent manner. Co-expression of the wild-type hAR with the PIAS1 C350A or PIASx␣ C362A mutant and FLAG-SUMO-1(GG) had no effect on the ligand-dependent activation. These data suggest that PIAS1 and PIASx␣ repress AR-dependent transactivation in a manner dependent on their RING finger-like domains, which are required for their SUMO-E3 ligase activity. When we performed the same reporter gene assays as above by using COS-7 cells, essentially similar results were obtained (data not shown).
Taken together, these results strongly suggest that PIAS1 and PIASx␣ repressed the AR-dependent transactivation through the sumoylation of AR by their SUMO-E3 ligase activity. DISCUSSION Recently, we and other groups identified PIAS family proteins as SUMO-E3 ligases toward p53 or LEF1. These findings suggested that PIAS family proteins might act as SUMO-E3 enzymes for other substrates as well. The AR has been shown to be sumoylated and to interact with PIAS1 and ARIP3/PI-ASx␣. Therefore, in the present study we sought to obtain evidence that PIAS1 and PIASx␣ have SUMO-E3 ligase activity toward the hAR. In this study PIAS1 and PIASx␣, but not PIAS3 and PIASx␤, could enhance the sumoylation of hAR. PIAS3 and PIASx␤ were shown to interact with the AR (28, 29), but they could not enhance the sumoylation of hAR. The protein interaction with PIAS proteins seems to be necessary for, but not sufficient to enhance, the sumoylation by PIAS proteins. A recent report showed that PIAS1 and PIASx␣ also acted as SUMO-E3 ligases for p53 and c-Jun (30) but not for other substrates such as Sp100. Taken together, these observations suggest that PIAS family proteins have substrate specificity. Because the PIAS family consists of a small number of family members, one of them may catalyze different substrates, but also multiple PIAS proteins may react with the same substrate. However, we cannot rule out the possible existence of yet undiscovered types of SUMO-E3 ligase. In fact, more recently it was shown that RanBP2, which is a component of the nucleocytoplasmic protein transport machinery, also had SUMO-E3 ligase activity (31). RanBP2 has no structural homology to the RING domain or homologous to E6-AP C terminus domain of ubiquitin-E3 ligase. Our yeast two-hybrid analysis revealed that PIAS1 and PIASx␣ also interacted via their RING finger-like domain with Ubc9, an E2 enzyme for the sumoylation. Because PIAS proteins also bind to substrate and SUMO-1, they may have a role in determining the precise conformation of the E2-substrate complex. As regards the substrate conformation, ligand binding induces conformational changes in the AR, causing the receptor to be transferred from the cytoplasm to the nucleus. Our results show that PIAS1 and PIASx␣ enhanced the sumoylation of hAR in a ligand-dependent manner in intact cells. In contrast, in the in vitro sumoylation assay, the sumoylation of hAR by PIAS1 or PIASx␣ was independent of the presence of the ligand. This result indicates that the sumoylation reaction does not require a conformational change of the AR by ligand binding in vitro. An earlier study (28) also showed the interactions of the PIAS proteins with the AR in vitro to be ligand-independent. However, the interaction in intact cells by use of a yeast or mammalian two-hybrid system was highly ligand-dependent. The ligand-dependent sumoylation of hAR in intact cells may be explained, at least in part, by the translocation of the receptor into the nucleus. The AR must be transferred into the nucleus to become sumoylated because PIAS proteins are localized exclusively in the nucleus (23,27,32,33). 2 Further experiments are needed to define whether ligand binding facilitates the interaction of the receptor with PIAS proteins and to determine the physiological consequence of sumoylation besides the accumulation of ARs in the nucleus.
In the luciferase reporter assay used to assess the transcriptional activity of the hAR, all PIAS proteins used in this study increased the ligand-dependent hAR transactivation. Importantly, however, PIAS1 and PIASx␣, which showed SUMO-E3 ligase activity toward the hAR, repressed specifically the hAR transactivation when co-expressed with SUMO-1. Interestingly, co-expression of the K386R/K520R mutant receptor with SUMO-1 still repressed the transactivation. This effect might have been due to the sumoylation of another regulator of AR. However, sumoylation sites of hAR were necessary for the full repressive effect on AR-dependent transactivation. We also found that the RING finger-like domain of PIAS1 or PIASx␣ 2 T. Nishida and H. Yasuda, unpublished data. was required to repress the hAR-dependent transactivation because a RING finger-like domain mutant of PIAS1 or PIASx␣ could not repress the ligand-dependent hAR activation in the presence of ectopically expressed SUMO-1. This domain in other proteins of the PIAS family also seems to be important for their biological activities. PIASy targets LEF1 to nuclear bodies and represses the transcriptional activation of LEF1, and this repression requires the RING finger-like domain of PIASy. Because the RING finger-like domain mutant failed to target both itself and LEF1 to nuclear bodies, this subnuclear sequestration by PIASy correlated with the repression of LEF1 (23). Similarly, RING finger-like domain mutant PIASx␤ slightly induced the transactivation of p53, although wild-type PIASx␤ repressed the transcriptional activity of p53 (30). In contrast, another group reported that PIAS1 stimulated p53-dependent transcriptional activation and that a PIAS1 mutant lacking the RING finger-like domain also activated p53-mediated gene expression as efficiently as did the wild-type (34). These findings suggest that the RING finger-like domain of PIAS proteins may not only play a role in the sumoylation of substrates but also in another function, such as protein-protein or protein-DNA interaction. For instance, Ubc9 interacts with the AR and activates AR-dependent transcription independently of its SUMO-E2 enzyme activity (35). The RING finger-like domain of PIAS proteins may recruit Ubc9 to the AR and/or stimulate the co-regulatory activity of Ubc9 toward the receptor.
The effects of PIAS proteins on AR-dependent transcription activity are yet obscure because they seem to depend on the cell type, the reporter gene promoter context, and expression levels of PIAS protein in a cell. For instance, when the minimal ARE2 TATA promoter or probasin promoter were used as the reporter, PIAS3 and ARIP3/PIASx␣ decreased AR-dependent transcription activity in a dose-dependent manner in HeLa, CV-1, and COS-1 cells; whereas in HepG2 cells, all of the PIASs increased AR-dependent transcription in a dose-dependent manner (28,36). It has been proposed that PIAS1 and PIASx␤ possess an inherent transcription-activating function but that ARIP3/PIASx␣ and PIAS3 are devoid of this activity (28). In prostate cancer cells, however, PIAS1 and PIAS3 enhanced the transcriptional activity of AR, but PIASy acted as a potent inhibitor of AR (37). These different effects of PIAS proteins may be caused by different expression levels of other cellular factors binding to the AR and/or PIAS proteins involved in AR-dependent transcriptional regulation. For example, PIAS1 interacts with other members of the PIAS family in a cell, for co-expression of PIAS1 with other members of the PIAS proteins influenced the effect of PIAS1 on AR-dependent transcription (38). Our present data provide evidence that PIAS1 and PIASx␣ act as SUMO-E3 ligases for the AR and repress AR-dependent transactivation, at least in part, through the sumoylation of the AR by their SUMO-E3 ligase activity. However, the mechanisms by which several PIAS proteins regulate AR-dependent transactivation are not yet clear. Some PIAS proteins may sequester ARs from the transcriptional complex and target them to the nuclear matrix, as in the case of LEF1 by PIASy. In our preliminary immunofluorescence microscopy experiments, although hARs were not co-localized to the nuclear matrix with PIAS when cells were co-transfected with PIAS1 or PIASx␣, the results from overexpression studies sometimes do not reflect the true physiological subcellular localization. It is also important to know what degree of sumoylation is needed to alter the properties of the AR. Sumoylation of the AR may alter the association of this receptor with other transcriptional coregulators, its DNA binding ability, and its stability. Further studies will be needed to address these issues.