Ubc9 Interacts with the Androgen Receptor and Activates Receptor-dependent Transcription*

Ubc9, a homologue of the class E2 ubiquitin-conjugating enzymes, has recently been shown to catalyze conjugation of a small ubiquitin-like molecule-1 (SUMO-1) to a variety of target proteins. SUMO-1 modifications have been implicated in the targeting of proteins to the nuclear envelope and certain intranuclear structures and in converting proteins resistant to ubiquitin-mediated degradation. In the present work, we find that Ubc9 interacts with the androgen receptor (AR), a member of the steroid receptor family of ligand-activated transcription factors. In transiently transfected COS-1 cells, AR-dependent but not basal transcription is enhanced by the coexpression of Ubc9. The N-terminal half of the AR hinge region containing the C-terminal part of the bipartite nuclear localization signal is essential for the interaction with Ubc9. Deletion of this part of the nuclear localization signal, which does not completely prevent the transfer of AR to the nucleus, abolishes the AR-Ubc9 interaction and attenuates the transcriptional response to cotransfected Ubc9. The C93S substitution of Ubc9, which prevents SUMO-1 conjugation by abrogating the formation of a thiolester bond between SUMO-1 and Ubc9, does not influence the capability of Ubc9 to stimulate AR-dependent transactivation, implying that Ubc9 is able to act as an AR coregulator in a fashion independent of its ability to catalyze SUMO-1 conjugation.

The androgen receptor (AR) 1 is a ligand-activated transcription factor that belongs to the nuclear receptor superfamily. The N-terminal part of steroid receptors contains a powerful activation function 1, and the C-terminal ligand-binding domain encompasses a second activation function (1,2) that interacts with numerous coactivators (3). The DNA-binding domain (DBD) consists of two zinc fingers and is C-terminally flanked by the hinge region (4). However, binding to specific DNA motifs, the hormone response elements, is not the only function of the DBD. Unlike glucocorticoid receptor (GR) null mice, animals carrying a GR gene mutation encoding a receptor form defective in dimerization and efficient DNA binding are viable (5). AR DBD is important in the transrepression of AP-1 and NF-B-activated genes (6 -8), and it also participates in the interaction with the coactivator CREB-binding protein (6). In view of this, it has not been unexpected that several coactivator proteins interacting with the DBD and hinge regions of steroid receptors have been characterized over the last few years (9 -13).
Activation of steroid receptors is mainly regulated by their cognate ligands. Upon ligand binding, the conformation of the receptors changes; the receptors dissociate from chaperones (14 -20) and are transferred to the nucleus. In nuclei, steroid receptors bind to specific hormone response elements and activate target genes or modulate and interfere with the activity of other transcription factors. The process of nuclear import is energy-dependent and guided by nuclear localization signals (NLS) (21,22). In the case of AR and GR, the cognate ligands facilitate their nuclear transport. Steroid receptors carry a bipartite NLS that encompasses the last residues of the DBD and the N-terminal residues of the hinge region. In addition to NLS, also the activation function 1-containing region and ligand-binding domain modulate the intracellular localization of AR (23)(24)(25).
Ubc9 is a homologue of the E2-type ubiquitin-conjugating enzymes and essential for cell cycle progression in yeast. Its homologues interact with proteins of diverse functions, including the GR. This latter interaction involves the GR DBD and is sensitive to mutations that abolish the ability of GR to repress AP-1 (26). The activity of another transcription factor, Ets-1, is enhanced by coexpressed Ubc9 (27). Also c-Jun (26) and adenovirus E1A (28) have been reported to interact with Ubc9, but similar to GR, functional consequences of these interactions have not been addressed. Likewise, initial attempts to characterize the region of Ubc9 that interacts with GR and c-Jun failed to identify a distinct interface (26). Because of its homology to ubiquitin-conjugating enzymes, Ubc9 has been linked to ubiquitination of target proteins; for example, in vitro ubiquitination of activating transcription factor 2 is facilitated by the addition of purified human Ubc9 (29). More recent results have shown, however, that Ubc9 is not involved in ubiquitin conjugation, but rather in covalent linking of the ubiquitin-like protein SUMO-1 (also known as PIC1, UBL1, GMP1, or Sentrin) to several target proteins (30 -33). In the case of the mammalian Ran GTPase-activating protein RanGAP1, SUMO-1 conjugation acts as a targeting signal to the nuclear pore complex (34 -36). SUMO-1 has targeting functions also within the nucleus; for example, unmodified promyelocytic leukemia gene product resides in nucleoplasm, whereas the SUMO-1-modified form is localized to so-called promyelocytic leukemia gene product nuclear bodies (37). Because the promy-elocytic leukemia gene product nuclear bodies have been reported to contain CREB-binding protein and nascent RNA produced by the action of RNA polymerase II, they might be involved in transcriptional regulation (38). Interestingly, conjugation of SUMO-1 is antagonistic to ubiquitination, in that SUMO-1 is linked to the same lysine of IB as ubiquitin, and by blocking ubiquitination, SUMO-1 modification renders IB resistant to proteasome-mediated degradation (39).
Our recent work has focused on the protein interaction partners of AR DBD and their role in AR-dependent transcriptional regulation (6,10,11,13). Because Ubc9 interacts with GR DBD (26), which is highly homologous to AR DBD, it was pertinent to examine the influence of Ubc9 on the regulation of AR-dependent signaling.

EXPERIMENTAL PROCEDURES
Materials-Protease inhibitors aprotinin, leupeptin, pepstatin, and phenylmethylsulfonyl fluoride were obtained from Sigma. Testosterone was purchased from Makor Chemicals, and 32 P-labeled nucleotides were from Amersham Pharmacia Biotech. Mouse monoclonal M2 anti-Flag antibody, mouse monoclonal anti-Lex antibody, and horseradish peroxidase-conjugated anti-mouse IgG were obtained from Kodak, CLONTECH, and Zymed Laboratories Inc., respectively. The polyclonal anti-AR antibodies were K183 and K333 (40).
Plasmid Constructions-Yeast vectors pVP16, for expressing the transactivation domain of herpes simplex virus VP 16 protein, and pLexN-a, for expressing bacterial LexA protein with the nuclear localization signal of the SV40 large T antigen, were kind gifts from Dr. Stanley M. Hollenberg (Oregon Health Sciences University, Portland, OR). LexN-a AR constructs were generated by inserting polymerase chain reaction-amplified fragments coding for amino acids 554 -644 of human AR (hAR) into the BamHI/SalI site of pLexN-a. Lex-WT1ZF encoding the zinc finger region (residues 312-419) of the Wilms' tumor suppressor gene product WT1 fused to Lex DBD has been described earlier (10). pVP16-Ubc9, a yeast expression vector coding for amino acids 1-156 of mouse Ubc9, was isolated in an unrelated yeast twohybrid screen. Inserts for the expression vectors Flag-Ubc9 and Flag-UbcC93S were generated by polymerase chain reaction and overlap polymerase chain reaction, respectively, and cloned into the HindIII/ EcoRI site of pFlag-CMV-2.
Yeast Two-Hybrid Assay-The LexN-a AR constructs were transformed together with plasmids encoding VP16 AD or VP16 AD fused to Ubc9 into Saccharomyces cerevisiae strain L40 (a gift from S. M. Hollenberg). The transformants were plated on a selective medium devoid of uracil, tryptophan, and leucine, and ␤-galactosidase activities were determined from three separate liquid yeast cultures according to the instructions of the Matchmaker Two-Hybrid System (CLONTECH).
Cell Culture and Transfections-COS-1 cells were obtained from American Type Culture Collection and were maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum and 25 units/ml of streptomycin and penicillin. For transient transfections, 3-3.5 ϫ 10 4 cells were seeded on 12-well plates 24 h prior to transfections, and 4 h before the addition of DNA, the cells received fresh medium with 10% charcoal-stripped fetal bovine serum. Transfections were performed by using FuGene reagent (Roche Molecular Biochemicals) according to the manufacturer's instructions. After 18 h, the medium was changed to Dulbecco's modified Eagle's medium supplied with 2% charcoal-stripped fetal bovine serum and 100 nM testosterone or vehicle. For mammalian two-hybrid assays, 5 ϫ 10 4 COS-1 cells were seeded on 12-well plates. Two hundred ng of VP16, 200 ng of Gal4 fusion protein expression vectors, 200 ng of pG5-LUC, and 50 ng of pCMV␤ were transfected by using FuGene. Luciferase and ␤-galactosidase activities were assayed as described previously (41,42). For preparation of whole cell extracts, COS-1 cells were electroporated with 20 g of DNA/10-cm dish as described (43).
Immunocytochemistry-CV-1 cells, seeded on glass coverslips on 6-well plates, were transfected with FuGene reagent with 150 or 400 ng of hAR expression vector, and the total DNA amount was filled to 1 g with empty pSG5 DNA. Twenty-one hours after transfection, the cells received fresh stripped 2% fetal bovine serum and were cultured for an additional 26 h in the presence or absence of 100 nM testosterone. Cells were fixed in 4% paraformaldehyde in phosphate-buffered saline and permeabilized with 0.4% Triton X-100. AR protein was detected with polyclonal rabbit antiserum K183 (40) raised against full-length rat AR (1:200 dilution). Fluorescein isothiocyanate-conjugated anti-rabbit secondary antibodies (Jackson ImmunoResearch Laboratories, Philadelphia, PA) were used for visualization of the receptor protein by the use of a fluorescence microscope.

Interaction of Ubc9 with AR DBD and Hinge Regions in
Yeast-Human AR DBD and hinge regions (amino acids 554 -644, Fig. 1A) fused to LexN-a as a bait (LexNa-DBDH) was used to evaluate the interaction with Ubc9 cloned in-frame to VP16 activation domain in a yeast two-hybrid system. Strong activation of a ␤-galactosidase reporter was observed when both partners were present (Fig. 1B). Deletion of amino acids 629 -633 in the N-terminal part of the hinge region, containing the RKLKK motif of the bipartite AR NLS (Fig. 1A), weakened the interaction significantly (⌬629 -633). A double substitution in the DBD (R617G and K618A, termed GA), which destroys the N-terminal part of the NLS (Fig. 1A), had a smaller effect on the interaction, but when combined with the ⌬629 -633 deletion (GA⌬629 -633), it resulted in a further decrease in the reporter activity (Fig. 1B). These results are not explainable by differences in the amounts of Lex-AR DBD fusion proteins, as the expression levels of the mutant constructs were similar to each other and somewhat higher than that of wild-type AR DBD in yeast (Fig. 1C). Because the vector encoding the Lex fusion proteins was LexN-a, which contains an NLS from the SV40 large T antigen, the inability of AR DBD/hinge mutants to enter the nucleus should not be responsible for the differences in their interaction with Ubc9.
Interaction of Ubc9 with Full-length AR in Mammalian Cells-To confirm that the AR-Ubc9 interaction also occurs in mammalian cells, mammalian two-hybrid experiments with full-length AR in COS-1 cells were performed (Fig. 2). Cotransfection of a construct encoding rat AR amino acids 3-902 fused to Gal4 DBD (Gal4-AR) with an expression vector for VP16 AD or VP16-CP (VP16 AD-polyoma virus coat protein fusion) did not yield significant reporter gene activation in the presence or absence of testosterone. When the VP16-Ubc9 fusion protein was transfected with Gal4-AR, a weak reporter gene activation was observed. This was substantially augmented by the addition of testosterone, yielding a 5.5-fold increase over that with Gal4-AR and VP16-Ubc9 in the absence of hormone, or a 23fold activation of the reporter in comparison to that with Gal4-AR and VP16-CP in the presence of hormone (Fig. 2).
The association of AR and Ubc9 was further investigated by coimmunoprecipitation (Fig. 3). COS-1 cells were transfected with Flag-tagged Ubc9 and AR expression vectors. Monoclonal anti-Flag antibody was first used to collect the protein complexes containing Flag-Ubc9, and the presence of wild-type or mutant AR proteins in these complexes was subsequently examined by immunoblotting with a polyclonal anti-AR antibody. Wild-type AR but not the GA⌬629 -633 mutant associated with Flag-Ubc9 (Fig. 3, lanes 3 and 5). The presence of ⌬629 -633 and GA mutations alone attenuated but did not abolish completely the interaction between AR and Ubc9 (Fig. 3, lanes 6  and 7), which was in line with the data from functional assays in yeast (cf. Fig. 1B). It was of interest to observe that the deletion of AR amino acids 574 -627, which encompass the major part of AR DBD, did not influence the ability of the receptor to interact with Ubc9.
Intracellular Localization and Transactivation Properties of the NLS Mutants GA, ⌬629 -633, and GA⌬629 -633-To elucidate the role of the bipartite NLS in AR function, the properties of AR mutants GA, ⌬629 -633, and GA⌬629 -633 were investigated (Fig. 4). In transfected CV-1 cells, wild-type AR exhibited predominantly nuclear localization with occasional faint staining in the cytoplasm in the absence of androgen (Fig. 4, A  and B). When both parts of the NLS were destroyed (GA⌬629 -633) (C and D), AR resided in the cytoplasm in the absence of the androgen, but upon hormone exposure, it gained nuclear localization, although cytoplasmic staining with occasional granules remained even in the presence of testosterone. The   FIG. 1. Interaction of Ubc9 and AR in yeast is sensitive to mutations in the AR nuclear localization signal. A, schematic structure of hAR and the mutants used in this study. The numbers refer to the amino acids of hAR (47). The bipartite nuclear localization signal and its mutated forms are shown in bold. TAD, transactivation domain; H, hinge region; and LBD, ligand-binding domain; WT, wild-type. B, the AR constructs were cloned into LexN-a fusion protein expression vectors and transformed into S. cerevisiae strain L40 together with plasmids encoding VP16-AD or VP16-AD fused to Ubc9. The zinc finger region of the WT1 protein, which lacks the ability to interact with Ubc9 (48), was used as a nonspecific control (Lexa-WT1ZF). ␤-Galactosidase activities were assayed from liquid cultures in three separate experiments, each with triplicate samples. The ␤-galactosidase activity of yeast transformed with LexNa-DBDH together with pVP16-Ubc9 is set as 100. C, immunoblot of Lex constructs. The primary antibody was a monoclonal anti-Lex-a antibody (1:1000 dilution). Horseradish peroxidase-conjugated monoclonal anti-mouse IgG (1:3000) was used as secondary antibody, and the detection was carried out employing the ECL system mutants GA (E and F) and ⌬629 -633 (panels G and H) were mostly cytoplasmic in the absence of the ligand. They were able to enter the nucleus in the presence of testosterone but, in contrast to wild-type AR, residual cytoplasmic staining was observed in the presence of androgen. Thus, nuclear transport of the mutants is compromised but not totally abolished.
The transactivation ability of the AR mutants was investigated in transiently transfected COS-1 cells (Fig. 5A). The minimal reporter construct pARE 2 -TATA-LUC was activated poorly, if at all, by the mutants GA and GA⌬629 -633 (GA, 3-fold and GA⌬629 -633, 1.2-fold versus a 35-fold activation by wild-type AR). Surprisingly, the mutant ⌬629 -633 was a 3-10fold more potent activator of the reporter construct than wildtype AR (Fig. 5A). Expression levels of the mutant proteins were determined by immunoblotting, and significant differences were not observed (Fig. 5B). When a reporter driven by the rat probasin promoter was examined under the same experimental conditions, the activity of ⌬629 -633 was 1.4 -2.8fold higher than that of wild-type AR. In contrast, the activities of the mutants GA and GA⌬629 -633 on this promoter were minimal, with the latter exhibiting no activity and the former 1 ⁄5-1 ⁄10 of the wild-type activity. 2 AR-dependent Transactivation Is Stimulated by Ubc9 Independently of the SUMO-1-conjugating Activity-To examine the influence of Ubc9 overexpression on AR-mediated transactivation, COS-1 cells were transiently transfected with expression vectors for hAR and Ubc9 and the reporter construct pARE 2 -TATA-LUC. Coexpression of Ubc9 enhanced AR-dependent transactivation in a dose-dependent fashion, without affecting the reporter activity in the absence of androgen or AR (Fig. 6). The increase in reporter gene activity was not because of increased cellular concentration of AR, because immunoblots (see Fig. 5B) and whole cell ligand-binding assays 2 did not reveal alterations in immunoreactive or biologically active AR protein content. Coexpression of Ubc9 did not increase the DNA binding of AR, because the amount of the receptor interacting with AREs in electrophoretic mobility shift assay remained constant. 2  Whole cell extracts were subjected to immunoprecipitation with mouse monoclonal anti-Flag antibody, and immunoprecipitates were subsequently analyzed by immunoblotting with a rabbit polyclonal anti-AR antibody. A represents an immunoblot of protein complexes precipitated with anti-Flag antibody detected by anti-AR antibody. The arrowhead depicts AR protein forms, and the asterisk depicts the immunoglobulin heavy chain. In B, AR input is shown. 1% cell extract used in immunoprecipitation was immunoblotted with anti-AR antibody. C represents Flag-Ubc9 input. 1% cell extract used in immunoprecipitation was immunoblotted using anti-Flag antibody. wt, wild-type.

FIG. 4. Localization of human AR mutants in CV-1 cells.
CV-1 cells were plated on glass coverslips on 6-well plates, transfected, and maintained with or without 100 nM testosterone. Detection by immunofluorescence was performed with rabbit anti-AR antibody K183 (1: 200 dilution) as primary and fluorescein isothiocyanate-conjugated goat anti-rabbit IgG as secondary antibody. The panels show cellular localization of the AR mutants, with cells cultured in the absence (-T) and presence (ϩT) of testosterone. A and B, wild-type AR; C and D, the GA⌬629 -633 mutant; E and F, the GA mutant; and G and H, the ⌬629 -633 mutant. triguingly, introduction of the C93S substitution into the proteincoding sequence of Ubc9, which renders it incapable of SUMO-1 conjugation (29 -32), attenuated only marginally the ability of Ubc9 to stimulate AR-mediated transactivation (Fig. 6). This suggests that the SUMO-1-ligating activity of Ubc9 is not mandatory for the protein to activate AR function. Indeed, our attempts to detect a SUMO-1-modified AR protein have not yielded results to support the presence of such an AR form. In immunolocalization experiments, coexpression of Ubc9 did not influence AR localization. 3 Deletion of Amino Acids 629 -633 of the Hinge Attenuates Ubc9-mediated Activation of AR Function-The hAR mutant ⌬629 -633 has full transactivation potential compared with wild-type AR in the context of the pARE 2 -TATA-LUC reporter (Figs. 5A and 6). However, the ability of coexpressed Ubc9 to activate the function of this AR form was clearly less than that of the wild-type receptor (Fig. 6). The effect of other NLS mutations on Ubc9-mediated activation of AR function could not be investigated, as the transactivation ability of these mutants is almost negligible.

DISCUSSION
In this study, we show that androgen receptor and Ubc9 interact. The interaction is readily detectable in a yeast twohybrid system employing AR DBD and the first twenty amino acids of the hinge region, the domain of AR harboring the bipartite NLS. Substitutions in the C-terminal DBD and the N-terminal part of the hinge attenuate the interaction between AR and Ubc9. In the case of the Ubc9-hGR interaction, a receptor form carrying two mutations in the second zinc finger (C476W and R479Q) did not recognize Ubc9 (26). The AR-Ubc9 interaction observed in yeast could be verified in a mammalian two-hybrid system with full-length AR and it was enhanced by the presence of ligand, even though a weak interaction also occurred in the absence of androgen. In coimmunoprecipitation experiments, the interaction was again detectable and attenuated by the destruction of the bipartite NLS. Interestingly, a major part of AR DBD could be deleted without a significant effect on the interaction between AR and Ubc9, as the ⌬574 -627 mutant, which lacks the region between the tip of the first zinc finger and beginning of the hinge region as well as the N-terminal part of the NLS, was coimmunoprecipitated with Ubc9 as efficiently as wild-type AR. These results suggest that the AR-Ubc9 interaction is, at least in part, electrostatic, as residues 629 -633 are positively charged and the Ubc9 crystal structure has revealed a strong electrostatic dipole with negative and positive surfaces on the opposite sides of the molecule (44). Because the main function of AR is to regulate gene expression in a ligand-dependent fashion, we investigated the effect of Ubc9 coexpression on transcriptional activation by AR in transiently transfected cells. A dose-dependent increase in pARE 2 -TATA-LUC reporter activity was observed in cells exposed to testosterone, whereas in the absence of ligand, Ubc9 had no effect. Thus, Ubc9 does not enhance transcription in general, and in this sense, it should be viewed as an AR coregulator. In transfections with the ⌬629 -633 mutant, the response to Ubc9 was negligible, as expected on the basis of the interaction data from the yeast two-hybrid system and immunoprecipitations. To address the role of the SUMO-1-conjugating activity of Ubc9 in this event, we used the conjugation-defective mutant C93S that is unable to form thiolester linkage with SUMO-1. As this mutant behaved in a fashion very similar to wild-type Ubc9 in cotransfection experiments, it appears that activation of AR function is not related to SUMO-1 modifications of AR or other proteins. These coactivator-like properties of Ubc9 have been previously reported in the context of transcription factor Ets-1, in that Ubc9 enhanced transactivation by Ets-1, and also in this case, the activity was not abolished by the C93S substitution in Ubc9 (27).
The bipartite NLS of AR seems to be important for the interaction with and transcriptional enhancement by Ubc9, and SUMO-1 modifications catalyzed by Ubc9 have been shown to play multiple roles in cellular targeting of proteins. On the basis of this, it was tempting to assume that Ubc9 is involved in the regulation of AR localization. The transport of AR from cytoplasm to the nucleus is not well understood (24,25,40). In our experiments, mutations of both the N-terminal and Cterminal parts of NLS impair the access of AR to the nuclei of CV-1 cells but do not abolish it entirely, even when combined. Because an AR mutant devoid of the C-terminal part of NLS is incapable of interacting with Ubc9 but still able to enter the nucleus, the primary function of the AR-Ubc9 interaction is probably not related to the targeting of AR to the nucleus. In view of the fact that SUMO-1-conjugating activity of Ubc9 was not mandatory for the activation of AR-dependent transcription, it is unlikely that intranuclear targeting of AR by SUMO-1 modification to transcriptionally more active compartments would explain its ability to regulate AR function. Likewise, SUMO-1 modification of other proteins appears to be excluded as well.
The role of the RKLKK sequence in the previously characterized AR NLS in the interaction and response to Ubc9 is evident. This raises the question of a presently unknown (unrelated to nuclear import) function for this region. The NLS consists of charged amino acids and is potentially available for interactions with nuclear proteins. Should some of these interactions repress AR-dependent transcription, then the binding of Ubc9 could mask this surface and prevent the repressor from influencing AR function. The hinge regions of AR and GR also participate in the targeting of the receptors to nuclear matrix (45,46). As transiently transfected templates are not organized into a native chromatin structure, it is possible that transcription factors not bound to the nuclear matrix do, in fact, stimulate transcription from non-chromatin templates more efficiently than matrix-associated factors. Should Ubc9 indeed prevent putative matrix adaptor/targeting proteins from contacting AR, this would enhance AR-mediated transcription in transiently transfected cells. Likewise, substitutions in the hinge region that influence the potential matrix-targeting residues should generate an AR form that is more potent in transcriptional activation.
Collectively, our results together with those on the modulation of Ets-1-mediated transcription by Ubc9 (27) suggest that Ubc9 has functions other than merely catalyzing SUMO-1 conjugation of different proteins.