DJ-1 positively regulates the androgen receptor by impairing the binding of PIASx alpha to the receptor.

DJ-1 was first identified as a novel candidate of the oncogene product that transformed mouse NIH3T3 cells in cooperation with an activated ras. Later DJ-1 was also found to be an infertility-related protein that was reduced in rat sperm treated with sperm toxicants that cause infertility in rats. To determine the functions of DJ-1, cDNAs encoding DJ-1-binding proteins were screened by the yeast two-hybrid method. Of several proteins identified, PIASx alpha/ARIP3, a modulator of androgen receptor (AR), was first characterized as the DJ-1-binding protein in this study. DJ-1 directly bound to the AR-binding region of PIASx alpha by an in vitro coimmunoprecipitation assay and also bound to PIASx alpha in human 293T cells. Both proteins were co-localized in the nuclei. PIASx alpha inhibited the AR transcription activity in a dose-dependent manner in cotransfected monkey CV1 cells with an androgen responsive element-luciferase reporter. Introduction of DJ-1 into CV1 cells in a state of inhibition of AR activity by PIASx alpha restored AR transcription activity by absorbing PIASx alpha from the AR-PIASx alpha complex, while a DJ-1 mutant harboring an amino acid substitution at number 130 from lysine to arginine did not restore it. These results indicate that DJ-1 is a positive regulator of the androgen receptor.

DJ-1 was first identified by our group as a novel candidate of the oncogene product that transformed mouse NIH3T3 cells in cooperation with activated ras (1). Both human and mouse DJ-1 genomic DNAs comprise seven exons, and exons 2-7 code for DJ-1 proteins (2). The human DJ-1 gene is mapped at chromosome 1p36.2-p36. 3, where a hot spot of chromosome abnormalities has been reported in several tumors (2). DJ-1 is preferentially expressed in the testis and moderately in other tissues, and it is translocated from the cytoplasm to nuclei during the cell cycle after mitogen stimulation, suggesting that DJ-1 has a growth-related function (1). However, the mechanism by which cells are transformed has not been clarified. Another group has also identified RS, another name for human DJ-1, as a regulatory component of an RNA-binding protein complex (3). Furthermore, CAP-1 or SP22, a rat homolog of human DJ-1, has been identified by other laboratories as a key protein related to the infertility of male rats exposed to sperm toxicants such as ornidazole and epichlorohydrin in which DJ-1 in the sperm and epididymis decreased in parallel with the following infertility of rat (4 -7). It was shown that DJ-1/CAP-1/SP22 is a first protein clearly related to male infertility (4,5,7). In addition to its expression in spermatocytes and spermatids, DJ-1 is also expressed in the sperm head and is translocated to the cytoplasmic side of sperm after the sperm toxicant treatment in infertile rats, suggesting that DJ-1 plays a role in fertilization (7)(8)(9). The exact roles of DJ-1 in spermatogenesis and fertilization, however, have not been determined.
The androgen receptor (AR) 1 bound by androgen activates the genes essential for development and maintenance of male reproductive function, including spermatogenesis (10 -12). AR belongs to the family of nuclear receptor proteins that act as ligand-dependent transcription factors that bind to respective DNA elements (13)(14)(15)(16)(17)(18). The nuclear receptor is composed of at least three domains, a ligand-independent transcription activation (AF1) domain, a DNA-binding domain, and a liganddependent transactivation (AF2) domain, and the molecular mechanisms underlying the transactivation functions of nuclear receptors, especially the estrogen receptor, have been extensively studied. The N-terminal domains along with DNAbinding domains of AR (19 -22), estrogen receptor (23,24), progesterone receptor (25), and PPAR␥ (26) interact with the ligand-bound C-terminal domain of AF2. The fragment containing the N-terminal domains along with the DNA-binding domain of AR is constitutively active in transcription without a ligand, whereas that of AF2 along with the DNA-binding domain lacks transcription activation activity in either the absence or presence of androgen (22,27). Nuclear receptors bind DNA elements as homo-or heterodimers, and AR homodimerization is stimulated by the presence of an androgen-responsive element (ARE).
Activation of the transcription of genes targeted by the nuclear receptors is modulated by the interaction between the nuclear receptors and their coactivators that bridge the general transcription machinery, and several coactivators that modulate the activity of nuclear receptors have been identified (28 -33). These include CBP/p300, the p160 family, pCAF/GCN5, and TRAP/DRIP (34,35). A number of AR-associated proteins, such as ARA24 (36), ARA54 (37), ARA55 (38), ARA70 (39), ARA160 (40) FHL2/DRAL (41), and cyclin E (42), have also been reported to modulate AR transcription activity. Transcription repression of the nuclear receptor-specific genes is relieved through receptor binding of cofactors, including histone acetyltransferase, and other chromatin remodeling factors that increase the accessibility of nucleosome DNA to transcription complex (43,44). HB01, a protein first identified as a protein that binds to an origin-recognition complex (45), has been shown to have a putative histone acetyltransferase activity and to repress AR transcription activity (46). ARIP3/PIASx␣ belongs to the family of PIAS (protein inhibitor of activated STAT) proteins, which includes PIAS1/GBP, PIAS3, PIASy, and PIASx␤/Miz1. PIAS1 and PIAS3 have been suggested to inhibit cytokine signaling (47,48). ARIP3/PIASx␣ was found to modulate AR transcription activity in a manner dependent on cells or the target genes; ARIP3/PIASx␣ first activates and then represses ARE-E1b promoter (minimal promoter) activity while it represses the probasin promoter, a natural target promoter of AR, in a dose-dependent manner in monkey CosI cells (49,50). Recent studies have shown that PIAS1 and PIASx␤ possess inherited transcription activity, whereas PIASx␣ and PIAS3 lack such activity (50). It has also been reported that PIAS1 stimulated AR promoter activity (51), whereas PIAS3 inhibited it (52).
To try to determine the molecular mechanism of DJ-1 function, we screened cDNAs encoding DJ-1-binding proteins by the yeast two-hybrid method using a human testis cDNA library, and we identified PIASx␣ as a DJ-1-binding protein. PIASx␣ inhibited the AR minimal promoter activity in monkey CV1 cells, and DJ-1 antagonized the repression activity of PIASx␣ to AR by absorbing PIASx␣ from the AR-PIASx␣ complex.

EXPERIMENTAL PROCEDURES
Cells-Mouse TM4 cells were obtained from American Type Culture Collection (ATCC). Human 293T, human HepG2, mouse TM4, monkey CosI, and monkey CV1 cells were cultured in Dulbecco's modified Eagle's medium with 10% calf serum.
Plasmids-pSG5rAR, a expression vector for rat AR, and pARE2-TATA-Luc, a reporter plasmid for testing AR transcription activity, were provided from J.J. Palvimo, cDNAs of human PIASx␤ and mouse PIAS3 were provided from K. Shuai and H. Yokosawa, respectively. An expressed sequence tag clone, ID no. 2228314 corresponding to human PIASy cDNA, was purchased from Incyte Genomics. pGLex-DJ-1:DJ-1 cDNA starting first ATG was inserted in frame into the EcoRI-XhoI site of pGlex, a modified version of the LexA-derived bait vector for yeast two-hybrid screening (53). pcDNA3-F-DJ-1 and pcDNA3-F-AR:DJ-1 cDNA and AR cDNA starting first ATG were inserted into the EcoRI-XhoI sites of pCMV-F, a pcDNA3 containing FLAG-tag (53). pEF-PIASx␣-HA:PIASx␣ cDNA was inserted into the EcoRI-NotI sites of pEF-HA (53).
Cloning of DJ-1-binding Proteins by a Two-hybrid System-Saccharomyces cerevisiae L40 cells containing the lacZ gene driven by the GAL1 promoter were transformed first with pGLex-DJ-1, which did not activate lacZ transcription by itself. The transformant cells were subsequently transformed with human testis MATCHMAKER cDNA (CLONTECH), a cDNA library expressing the GAL4 activation domain (GALAD) fused to the cDNAs from human testis cells. Approximately 1.5 ϫ 10 6 colonies were screened for lacZ expression, which indicated the association of a GALAD-fused protein with the LexABD-fused DJ-1. The plasmid DNAs in the lacZ-positive cells were extracted by the procedure described in the protocols from CLONTECH. Nucleotide sequences of the plasmids derived from positive colonies were determined by using an ABI377 or Li-Cor Long Reader 4200 autosequencer.
In Vitro Binding Assay-35 S-labeled FLAG-tagged DJ-1 and PIASx␣ were synthesized in vitro using the reticulocyte lysate of the TnTtranscription-translation coupled system (Promega). Labeled proteins were mixed at 4°C for 120 min and then immunoprecipitated with a mouse anti-FLAG antibody (M2, Sigma) or with nonspecific mouse IgG in a buffer containing 150 mM NaCl, 5 mM EDTA, 50 mM Tris (pH 7.5), 1 mg/ml bovine serum albumin, and 1% Nonidet P-40. After washing with the same buffer, the precipitates were separated in a 12.5% of polyacrylamide gel containing SDS and visualized by fluorography.
In Vivo Binding Assay-Ten g of pcDNA3-F-DJ-1 together with 10 g of pEF-PIASx␣-HA was transfected to human 293T cells 60% confluent in a 10-cm dish by the calcium phosphate precipitation technique (54). Forty-eight h after transfection, the whole cell extract was prepared by the procedure as described previously (55). Approximately 500 g of the 293T cell proteins were first immunoprecipitated with a mouse anti-FLAG antibody (M2, Sigma) or with nonspecific mouse IgG under the same conditions as those of the in vitro binding assay as described above. After washing with the same buffer except for 0.05% Nonidet P-40 instead of 0.25%, the precipitates were separated in 10% of polyacrylamide gel containing SDS, blotted onto a nitrocellulose filter, and reacted with a mouse anti-HA antibody (12CA5, Roche Molecular Biochemicals) or with the mouse anti-FLAG antibody. Indirect Immunofluorescence-The monkey CosI cells were cotransfected with pcDNA3-F-DJ-and pEF-PIASx␣-HA by the calcium phosphate precipitation technique (54). Forty-eight h after transfection, the cells were fixed with a solution containing 4% paraformaldehyde and reacted with a mouse anti-FLAG monoclonal antibody (M2, Sigma) or anti-HA polyclonal antibody (Y-11, Santa Cruz). The cells were then reacted with an fluorescein isothiocyanate-conjugated anti-mouse IgG or rhodamine-conjugated anti-rabbit IgG and observed under a confocal laser fluorescent microscopy.
Luciferase Assay-Two g of a reporter plasmid and 0.5 g of pCMV-␤-gal, a ␤-galactosidase expression vector, were transfected to cells ϳ60% confluent in a 6-cm dish by the calcium phosphate method (54). Two days after transfection, whole cell extracts were prepared by the addition of the Triton X-100-containing solution from a Pica gene kit (Wako Pure Chemicals Co. Ltd., Kyoto, Japan) to the cells. About a one-fifth volume of the extract was used for the ␤-galactosidase assay to normalize the transfection efficiency as described previously (55), and the luciferase activity due to the reporter plasmid was determined using a Pica gene kit and a luminometer, Lumat LB 9507 (EG & G Berthold). The same experiments were repeated three to five times.
Gel Mobility Shift (Bandshift) Assay-Bandshift assays were carried out as described previously (56). Briefly, a reaction mixture containing 20 g of CV1 cell nuclear extract, 2 g/ml poly(dI-dC), 4 mg/ml bovine serum albumin, 16 mM Hepes (pH 7.9), 50 mM KCl, 1 mM EDTA, 1 mM dithiothreitol, 4% Ficoll 400, and 25 nM testosterone was incubated for 10 min at room temperature. A labeled probe of 1 ϫ 10 4 cpm was then added to the mixture, and the mixture was further incubated at room temperature for 30 min. DNA-protein complexes formed in the mixture were separated in a 4% polyacrylamide gel containing 0.25 ϫ TBE and visualized by autoradiography. Nucleotide sequences used for the bandshift assay were 5Ј-CGAGTAGTACGTGATGTTCTAG-3Ј and 5Ј-TC-GACTAGAACATCACGTACTACTCGAGCT-3Ј of the upper and lower strands, respectively.

Identification of PIASX␣/ARIP3 as a DJ-1-binding Protein
and Determination of the DJ-1-binding Region-To screen cDNAs encoding DJ-1-1-associating proteins, a full-size DJ-1 starting from amino acid number 1 was fused to the LexA DNA-binding domain and introduced into S. cerevisiae L40 cells. A human testis cDNA library cloned in pGADGH was then introduced into the transformant yeast cells, and the colonies resistant to a His marker followed by ␤-galactosidase expression were selected. Among a total of 5 ϫ 10 5 transformant cells, 10 colonies were His-and ␤-galactosidase-positive, and two of the 10 positive colonies were identified as PIASX␣/ ARIP3 after determination of their nucleotide sequences. PI-ASX␣/ARIP3 has been reported to be expressed preferentially in the testis and to modulate AR transcription activity (49). Because the longest clone contained amino acids 335-572 of PIASx␣ (see Fig. 1), the full-size PIASx␣ cDNA was amplified by polymerase chain reaction with the specific primers corresponding to the nucleotide sequence of human PIASx␣ using total testis cDNAs (Marathon cDNA, CLONTECH) as templates. PIASx␣ was comprised of at least two domains, leucine zipper-like and AR-interacting domains, and four LXXLL motifs.
To determine the DJ-1-binding region of PIASx␣, various deletion constructs fused to the GAL4 activation domain (GA-LAD) were used for a two-hybrid assay with DJ-1 as a bait (Fig.  1A). Neither the N-terminal fragment spanning amino acids 1-334 nor the C-terminal fragment spanning amino acids 494 -572 of PIASx␣ bound to DJ-1. All other fragments including the original clones isolated (335-572, 340 -572, and 433-572) bound to DJ-1 more strongly than the full-size PIASx␣ did. These results indicate that DJ-1 binds to the C-terminal region of PIASx␣ spanning amino acids 433-493 that includes the AR-interacting domain. The binding activity of DJ-1 to AR was then examined by the two-hybrid method. DJ-1 fused to the LexA DNA-binding domain was not able to bind to AR fused to GALAD (Fig. 1B).
Then, the interaction of DJ-1 with other PIAS family proteins, including PIAS1, PIAS3, PIASx␤, and PIASy, was tested by the yeast two-hybrid method, where DJ-1 and PIAS family were fused to LexA DNA-binding domain and GAL4 activation domain, respectively (Fig. 1C). Of the PIAS family proteins, DJ-1 strongly bound to PIAS3 and PIASx␣, weakly bound to PIASy, and did not bind to PIAS1. We therefore concentrated on PIASx␣ in this study.
Interaction of DJ-1 with PIASx␣ in Vitro and in Vivo-An in vitro binding assay was then performed by using 35 S-labeled FLAG-tagged DJ-1 and PIASx␣ synthesized in vitro. Both labeled proteins were mixed together, and the proteins were immunoprecipitated with an anti-FLAG antibody or nonspecific IgG. Then the precipitate(s) was separated on the gel and visualized by fluorography ( Fig. 2A). First, it was confirmed that the anti-FLAG antibody alone did not precipitate labeled PIASx␣ without DJ-1 (data not shown). PIASx␣ was coimmunoprecipitated with DJ-1 by the anti-FLAG antibody but not by IgG ( Fig. 2A, lanes 1 and 2, respectively), suggesting that there was a direct interaction between DJ-1 and PIASx␣.
To observe the complex formation of PIASx␣ with DJ-1 in vivo, expression vectors for FLAG-tagged DJ-1 and HA-tagged PIASx␣ together were transfected to human 293T cells. Fortyeight h after transfection, the cell extract was prepared, and the proteins in the extract were first immunoprecipitated with the anti-FLAG antibody or nonspecific IgG. The precipitates were immunoblotted against the anti-HA antibody (Fig. 2B). The anti-FLAG antibody did precipitate FLAG-DJ-1 (data not shown). PIASx␣-HA, on the other hand, was detected in the immunoprecipitate with the anti-HA antibody but not with IgG (Fig. 2B, lanes 3 and 2, respectively), indicating that PIASx␣ was associated with DJ-1 in ectopic-expressed 293T cells.
Co-localization of DJ-1 with PIASx␣ in Cells-Previous studies have shown that DJ-1 was localized both in the cytoplasm and nucleus in DJ-1-transfected human HeLa cells and translocated from the cytoplasm to nuclei during the S-phase of the cell cycle upon induction by mitogen (1) and that PIASx␣/ ARIP3 was located in nucleus (49). To determine the cellular localization of PIASx␣ and DJ-1, expression vectors for FLAGtagged DJ-1 and HA-tagged PIASx␣ were transfected into monkey CosI cells. Two days after transfection, the cells were stained with anti-FLAG and anti-HA antibodies, and the proteins were detected by fluorescein isothiocyanate-and rhodamine-conjugated second antibodies, respectively, and then visualized under confocal laser microscopy (Fig. 3). DJ-1 (green) and PIASx␣ (red) were co-localized in nuclei as shown by the yellow color (Fig. 3, Merge). Cell nuclei were identified by (4,6-diamidino-2-phenylindole) staining (data not shown).
Repression of AR Transcription Activity by PIASx␣-We have recently found that DJ-1 was conjugated with SUMO-1, a small ubiquitin-like modifier, at lysine of amino acid number 130, and that this modification is necessary for DJ-1 to manifest full activities for cell transformation in collaboration with activated ras and stimulation of cell growth. 2 DJ-1-K130RX is a point mutant where lysine at amino acid number 130 was changed to arginine and three amino acids at amino acid numbers 57, 96, and 126 were also changed from serine, glutamic acid, and histidine to arginine, glycine, and tyrosine, respectively. Because this mutant was found to lose full activities of DJ-1 and also binding activity to PIAS␣ (see Fig. 5C), we used K130RX 3 as a negative control of wild-type DJ-1 in this study.
It has been reported that PIASx␣/ARIP3 modulates AR transcription activity in a cell type or promoter-dependent manner (49,50). PIASx␣/ARIP3 first stimulates and then decreases AR activity with a dose of PIASx␣/ARIP3 introduced into the ARE-E1b promoter (minimal promoter), but it represses the activity of probasin promoter, a natural target of AR, in monkey CosI cells (50). Moreover, PIASx␣/ARIP3 lacks the inherent transcription activation function, whereas other PIAS family proteins, PIAS1/GBP and PIASx␤/Miz1, possess this function (50). To confirm this observation, we first cotransfected the expression vector for AR, PIASx␣ or DJ-1 together with the ARE minimal promoter-luciferase reporter (pARE2-TATA-Luc) into monkey CV1 cells in the presence or absence of testosterone, and then we measured the luciferase activity two days after transfection. In the absence of testosterone, a low level of transcription activity was observed without the AR expression vector (see Fig. 6), and this value was therefore set at 1 in all the transcription analyses. In the presence of testosterone, the luciferase activity was stimulated by the expression vector for AR in a dose-dependent manner, while the expression vector for PIASx␣ and DJ-1 did not stimulate the luciferase activity (Fig. 4A). When the expression vector for PIASx␣ was added to the point where AR activity was almost at maximal level (0.5 g of pSG5rAR, an AR expression vector), PIASx␣ repressed AR activity to 40% of the original level in a dose-dependent manner, whereas DJ-1 (pcDNA3-F-DJ-1) or its point mutant, pcDNA3-F-DJ-1-K130RX, did not (Fig. 4B). Constant amounts of AR or a dose-dependent increase in the amount of PIASx␣ in the transfected cell extracts were confirmed by Western blotting (Fig. 4C). Even small amounts (0.01, 0.05, and 0.1 g) of the expression vector for PIASx␣ also repressed AR activity in CV1 cells, and a marginal stimulation of AR activity was observed with small amounts (0.01 and 0.05 g), but not with amounts above 0.1 g, of the expression vector for PIASx␣ in CosI cells (data not shown). These results suggest that PIASx␣ modulates AR transcription activity in a cell type-specific manner but that the priority function of PIASx␣ is repression of AR activity.
Because DJ-1 binds to the AR-interacting domain of PIASx␣, it is thought that DJ-1 abrogates the binding of PIASx␣ to AR. To address this issue, HA-tagged DJ-1 was cotransfected into CV1 cells with constant amounts of FLAG-tagged AR and HAtagged PIASx␣. Two days after transfection, the proteins in the cell extract were precipitated with an anti-FLAG antibody, and the precipitates were immunoblotted with the anti-FLAG antibody to detect the precipitated FLAG-AR and an anti-HA antibody to detect the coimmunoprecipitated PIASx␣ (Fig. 5A). The results of the Western blotting showed that DJ-1 was expressed in the cells in a dose-dependent manner. While the amounts of the precipitated AR were almost constant over the ranges used, AR-bound PIASx␣ was reversibly decreased with the amounts of DJ-1. When the same experiment was carried out using HA-tagged DJ-1-K130RX instead of wild-type DJ-1, the relatively constant binding activity of PIASx␣ to AR was observed (Fig. 5B). The ratio of the amount of AR-bound PI-ASx␣ to that of the precipitated AR was changed little over the ranges of DJ-1-K130RX. The DJ-1-K130RX was found not to bind to PIASx␣ by the yeast two-hybrid assay (Fig. 5C). Because DJ-1 does not bind to AR, this result suggests that DJ-1 but not DJ-1-K130RX absorbs PIASx␣ from the PIASx␣-AR complex.
Abrogation of PIASx␣-suppressed AR Transcription Activity by DJ-1-To observe the effect of DJ-1 on the AR-transcription activity suppressed by PIASx␣, DJ-1 and its mutant K130RX 3 were introduced into CV1 cells in a combination of AR and PIASx␣. As reported previously, ARE-minimal promoter activ-

DJ-1 as an AR-positive Regulator
ity was enhanced by the AR in the presence of testosterone to 30 -40-fold of that without testosterone (Fig. 6A, lanes 2-4). As shown in Fig. 4B, PIASx␣ suppressed the AR activity in a dose-dependent manner (Fig. 6A, lanes 4 -8). Addition of DJ-1, on the other hand, reversibly antagonized AR activity suppressed by PIASx␣ in a dose-dependent manner while the K130RX 3 mutant of DJ-1 did not (Fig. 6A, lanes 9 -12 and  13-16, respectively). The amounts of the proteins in transfected cells were also examined by Western blotting (Fig. 6B). Under the same conditions as those of luciferase assays, both DJ-1 and K130RX 3 were expressed with the doses transfected into the cells, while AR and PIASx␣ were almost constant as expected (Fig. 6B).
To determine the molecular mechanisms by which PIASx␣ and DJ-1 affect AR transcription activity, the DNA binding activity of AR in the presence or absence of DJ-1 with PIASx␣ was examined. The labeled oligonucleotide corresponding to the sequence of the ARE inserted into the luciferase reporter construct was mixed with the extract of CV1 cells transfected with various combinations of expression vectors for AR, PI-ASx␣, DJ-1, or DJ-1-K130RX, and then the DNA-protein complex was separated on gel (Fig. 6C). Two bands of the DNA-protein complex appeared (Fig. 6C, lane 2), and the slower migrating band but not faster migrating one disappeared after the addition of the extra amounts of the homologous oligonucleotide but not the mutant one (data not shown), suggesting specificity of the slower migrating band. Consistent with the results of luciferase activity shown in Figs. 4 and 6A, the extract from the cells transfected with both PIASx␣ and AR gave a reduced intensity of the band (Fig. 6C, lane 3). Introduction of DJ-1 into this combination of AR and PIASx␣, on the other hand, restored the intensity of the band corresponding to the AR-DNA complex in a dose-dependent manner (Fig. 6C,  lanes 4 -6). DJ-1-K130RX, on the other hand, did not restore the intensity of the AR-DNA complex (Fig. 6C, lanes 7-9). These results suggest that both PIASx␣ and DJ-1 do not affect the complex status of AR on the AR-responsive element and that PIASx␣ interferes with AR-DNA binding activity by masking its region. These results together with those shown in Figs. 5 and 6 clearly indicate that DJ-1 antagonized PIASx␣ function toward AR by absorbing it and suggest that this function of DJ-1 necessitates SUMO-1 modification to DJ-1.
Cell Type-specific Effect of DJ-1 on AR Transcription Activity-Because other reports showed that cell background and promoter context alter the effect of PIAS family proteins on AR-dependent transcription, the effect of DJ-1 on AR transcription activity was examined in human HepG2 and mouse TM4 cells in addition to monkey CV1 cells. In HepG2 hepatoma cells, ARIP3/PIASx␣ has been reported to potentate ARE-dependent transcription of AR (50). TM4 is an established Sertoli cell line in which AR is expressed (57). Expression vectors for AR, PIASx␣, DJ-1, or DJ-1-K130RX were transfected together with ARE-reporter into these cells, and their luciferase activities were measured two days after transfection (Fig. 7). In TM4 cells like CV1 cells, PIASx␣ inhibited AR transcription activity in a dose-dependent manner and DJ-1 reversibly antagonized AR activity suppressed by PIASx␣ in a dose-dependent manner, whereas the K130RX 3 mutant of DJ-1 did not (Fig. 7B). In HepG2 cells, on the other hand, PIASx␣ stimulated AR transcription activity in a dose-dependent manner as described previously (50), and DJ-1 had little effect of AR activity stimulated by PIASx␣, whereas the K130RX 3 mutant of DJ-1 rather inhibited it in a dose-dependent manner (Fig. 7A). These results suggest that cell background alters the effect of PIAS family proteins on AR-dependent transcription, and DJ-1 responds to suppressed AR activity. DISCUSSION DJ-1 was first identified as a novel oncogene product that transforms mouse NIH3T3 cells in collaboration with activated ras (1) and was later found to be a protein related to male rat infertility caused by exposure of rats to the male reproductive toxicants such as ornidazole or epichlorohydrin (4 -7). DJ-1 was found to be expressed in both somatic and germ cells. In germ cells, DJ-1 is expressed in stages from pachyten spermatocytes and finally expressed in the anterior-ventral region of the sperm head, which is thought to be essential for fertilization (7,8). DJ-1 is thus thought to have at least two functions: a function in somatic cells and a function in germ cells. In this study, we investigated the function of DJ-1 in somatic cells.
We found that DJ-1 acts as a positive regulator of AR by preventing PIASx␣ from binding to AR. DJ-1 bound to the AR-binding region of PIASx␣, thereby interfering with the binding of PIASx␣ to AR. PIASx␣ is a member of the PIAS family proteins that were first identified as inhibitors of STAT, a transcription factor employing cytokine signaling (4,8). The proteins belonging to the PIAS family are PIAS1/GBP, PIAS3, PIASy, PIASx␣/ARIP3, and PIASx␤/Miz1. As described in the Introduction, there seems to be two separate groups: proteins possessing inherit transcription activation activity and those lacking this activity. The former group includes PIAS1/GBP and PIASx␤, and the latter group includes PIAS3 and PIASx␣/ ARIP3 (50). DJ-1 was found to bind to the latter group of PIAS family proteins possessing inherit transcription suppression activity. DJ-1 also bound to PIASy, which has been recently reported to repress AR transcription activity in LNCaP prostate cancer cells (58). Although the transcription activities of these proteins toward AR appeared to reflect their intrinsic activities, discrepant results, depending on the cells or promoters used in the experiments, have been reported (50). In the CV1 cells in this study, we always observed inhibitory activities of PIASx␣ to AR, as shown in Figs. 4 and 6, even when only amounts of PIASx␣ had been transfected into the cells. In the HepG2 cells, we also observed the stimulating activity of PI-ASx␣ to AR transcription as described previously (50), and DJ-1 had no effect on AR transcription activity stimulated by PIASx␣. DJ-1 alone does not directly bind to AR and therefore FIG. 6. Restoration of AR-suppressed transcription activity by DJ-1. A, CV1 cells were transfected with various doses of the expression vector for FLAG-DJ-1, FLAG-K130RX mutant of DJ-1 (F-DJ-1-K130RX), or PIASx␣-HA together with pARE2-TATA-Luc and 0.5 g of pSG5rAR, an expression vector for AR. Two days after transfection, the luciferase activities were measured as described in Fig. 4A. B, proteins in an aliquot of the extract used in A were separated in 10% polyacrylamide gel and blotted with an anti-AR antibody (N-20, Santa Cruz), an anti-HA monoclonal antibody (12CA5, Roche Molecular Biochemicals), or an anti-actin antibody (C4, Roche Molecular Biochemicals). C, nuclear extracts were prepared from an aliquot of the transfected cells and used for a bandshift assay with a labeled oligonucleotide corresponding to the AR-responsive element as described under "Experimental Procedures." Free and AR complex indicate the positions of the free probe and AR-DNA complex, respectively.
does not respond to an AR-responsive element as shown in Fig.  4A. What then is the mechanism by which PIASx␣ inhibits the activity of AR? Because PIASx␣ bound to the DNA-binding region of AR as previously reported (49) and the results of bandshift analysis in this study using an AR-responsive element as a probe showed that PIASx␣ and DJ-1 were not able to form a tight complex with AR on DNA, PIASx␣ is thought to mask the DNA-binding region of AR and DJ-1 is thought to relieve its masked state.
Our recent results have shown that a lysine residue at amino acid number 130 of DJ-1 is conjugated with SUMO-1 and that the sumoylated state of DJ-1 may be necessary to show the full activity of DJ-1 in terms of cell transformation and cell growth stimulation 2 . We used two substitution mutants of DJ-1, K130R, in which the lysine at amino acid number 130 was changed to arginine, and K130RX 3 , in which three amino acids, in addition to the lysine at amino acid number 130, were changed, to determine the effects on PIASx␣ and AR. Because K130RX 3 was found to have the similar characters to those of K130R and to lose binding activity to PIASx␣, we therefore used K130RX 3 as a mutant to wild-type DJ-1 in this study. K130RX 3 did not antagonize the function of PIASx␣ to AR due to no binding activity to PIASx␣, confirming that DJ-1 is a positive regulator of AR. Although the K130R mutant of DJ-1 bound to PIASx␣ to the same extent as that of wild-type DJ-1, the antagonized transcriptional function of DJ-1 against PI-ASx␣ was hampered (data not shown). Although the molecular reason of this difference in the activities of wild-type and K130R DJ-1 is not clear at present, other factors associated with DJ-1 but not with K130R DJ-1 may be responsible for these different phenomena. These factors may determine the location of DJ-1 in cells where DJ-1 is active or, alternatively, may keep the proper conformation of the active form of DJ-1. As described in the Introduction, a number of AR-specific or nuclear receptor-binding proteins have been reported. We are now screening these proteins, including the cofactors stated in the Introduction, to determine which distinctly bind to wildtype DJ-1 and K130R mutant DJ-1.
The finding that DJ-1 is a positive regulator of AR in somatic cells may explain the relationship between the reduced amount of DJ-1 and infertility in rats in which the AR inactivated by PIASx␣ does not activate the genes essential for spermatogenesis.