SUMO-3 Enhances Androgen Receptor Transcriptional Activity through a Sumoylation-independent Mechanism in Prostate Cancer Cells*

Androgens are important for male sexual development, which depend on the cognate receptor, the androgen receptor. The transcriptional activity of the androgen receptor, like other nuclear receptors, is regulated by accessory proteins that can have either positive or negative effects. Through a yeast functional screen, we have identified SUMO-3 as a regulator of androgen receptor activity in prostate cancer cells. SUMO-3 is one of three eukaryotic proteins that become post-translationally conjugated to their target proteins in a manner analogous to the attachment of ubiquitin. In primary prostate epithelial cells, PrEC, and the prostate cancer cells, PC-3, SUMO-3 has a weak negative effect on androgen receptor transcriptional activity. In contrast, SUMO-3 and it close relative SUMO-2 strongly enhance transactivation by endogenous androgen receptor in LNCaP cells. This positive effect is observed in both androgen-dependent and androgen-independent LNCaP cells. Interestingly, SUMO-1, unlike SUMO-3 and SUMO-2, can inhibit, but not stimulate, androgen receptor activity. Mutational analysis of the androgen receptor and SUMO-3 demonstrates that the SUMO-3-positive activity does not depend on either the sumoylation sites of the androgen receptor or the sumoylation function of SUMO-3. Stable overexpression of SUMO-3 in LNCaP cells significantly enhances the androgen-dependent proliferation of these cells. Additionally, siRNA-mediated repression of SUMO-2 significantly inhibits the growth of both androgen-dependent and -independent LNCaP cells. Collectively, these results suggest (i) a novel mechanism for elevating AR activity through the switch of SUMO-3 from a weak negative regulator in normal prostate cells to a strong positive regulator in prostate cancer cells and (ii) a proliferative role for SUMO-3 and SUMO-2 in the growth of prostate cancer cells that is independent of sumoylation of the androgen receptor.

Androgens are important for male sexual development, which depend on the cognate receptor, the androgen receptor. The transcriptional activity of the androgen receptor, like other nuclear receptors, is regulated by accessory proteins that can have either positive or negative effects. Through a yeast functional screen, we have identified SUMO-3 as a regulator of androgen receptor activity in prostate cancer cells. SUMO-3 is one of three eukaryotic proteins that become post-translationally conjugated to their target proteins in a manner analogous to the attachment of ubiquitin. In primary prostate epithelial cells, PrEC, and the prostate cancer cells, PC-3,

SUMO-3 has a weak negative effect on androgen receptor transcriptional activity. In contrast, SUMO-3 and it close relative SUMO-2 strongly enhance transactivation by endogenous androgen receptor in LNCaP cells. This positive effect is observed in both androgendependent and androgen-independent LNCaP cells. Interestingly, SUMO-1, unlike SUMO-3 and SUMO-2, can inhibit, but not stimulate, androgen receptor activity. Mutational analysis of the androgen receptor and SUMO-3 demonstrates that the SUMO-3-positive activity does not depend on either the sumoylation sites of the androgen receptor or the sumoylation function of SUMO-3. Stable overexpression of SUMO-3 in LNCaP cells significantly enhances the androgen-dependent proliferation of these cells. Additionally, siRNA-mediated repression of SUMO-2 significantly inhibits the growth of both androgen-dependent and -independent LNCaP cells. Collectively, these results suggest (i) a novel mechanism for elevating AR activity through the switch of SUMO-3 from a weak negative regulator in normal prostate cells to a strong positive regulator in prostate cancer cells and (ii) a proliferative role for SUMO-3 and SUMO-2 in the growth of prostate cancer cells that is independent of sumoylation of the androgen receptor.
The physiological functions of the androgens testosterone and 5␣-dihydrotestosterone (DHT) 4 are mediated by the androgen receptor (AR) (reviewed in Ref. 1), a member of the nuclear receptor superfamily (reviewed in Refs. [2][3][4][5][6]. Through the regulation of target genes, androgens and AR play an essential role in male sexual development and the proper development and function of male reproductive organs, such as prostate and epididymis (7). Patients with 5␣-reductase II deficiency, which results in low levels of DHT, have ambiguous external genitalia and a highly underdeveloped and impalpable prostate (8,9). Reduction or loss of AR activity in males results in androgen insensitivity syndrome (10). In addition to normal prostate development, AR is essential for the initiation and progression of prostate cancer. The best demonstration of this is the effectiveness of anti-androgen and androgen ablation therapy in inhibiting the development of prostate cancer in the early stage of the disease (11). However, local recurrences and metastases will eventually develop in most, if not all, patients after therapy, and prostate cancer becomes androgen-independent (12). Since AR is expressed in both androgen-dependent and androgen-independent prostate cancer, this receptor is probably involved in the progression to androgen independence. Moreover, it has been determined that about 10 -20% of prostate tumors harbor mutations in the AR gene, and the frequency of mutation generally is higher in androgen-independent, metastatic tumors compared with untreated lower grade primary tumors (13)(14)(15)(16)(17)(18).
Like other nuclear receptors, the AR is regulated by multiple posttranslational modifications, including phosphorylation (19), acetylation (20,21), and sumoylation (22). Sumoylation represents an important post-translational modification system that regulates the activity of many transcriptional regulators (reviewed in Ref. 23). The continually growing list includes not only AR but also other nuclear receptors and transcriptional activators, coactivators, and corepressors (reviewed in Ref. 24). The biological functions of sumoylation include protein subcellular translocation, subnuclear structure formation, and modulation of transcriptional activity (reviewed in Ref. 25). Sumoylation depends upon the activity of small ubiquitin-related modifier (SUMO), a protein moiety that is conjugated to a specific lysine residue on target proteins (reviewed in Ref. 23). Three SUMO family members exist, SUMO-1/ Smt3C, SUMO-2/Smt3A, and SUMO-3/Smt3B, and all are ubiquitously expressed in mammals (27,28). At the amino acid level, SUMO-2 and SUMO-3 are 87% identical but only ϳ50% identical to SUMO-1 (27). Although they exhibit low homology in amino acid sequence, SUMO-1 and ubiquitin are structurally related and share significant similarity in secondary and tertiary structures (29). Therefore, it is not surprising that the processes of sumoylation and ubiquitination are mechanistically similar (reviewed in Ref. 24). Like ubiquitination, the conjugation of SUMO is mediated by a series of enzymatic reactions catalyzed by E1, E2, and E3 enzymes that are distinct from those enzymes that catalyze ubiquitination (27). The SUMO E1 enzymes SAE1 (SUMO-activating enzyme) and SAE2 activate SUMO and transfer it to the E2 enzyme Ubc9, which then directs the conjugated SUMO to its target substrates (27). In vitro evidence has indicated that Ubc9 is sufficient for binding to the SUMO acceptor site and efficiently transferring SUMO to selected targets (27). However, recent evidence shows that a specific E3 ligase might be required for efficient sumoylation in vivo. Three classes of proteins have been identified to have SUMO E3 ligase activity: the protein inhibitor of activated STAT (PIAS) family proteins (30,31), the polycomb protein Pc2 (32), and RanBP2 (Ranbinding protein 2) (33). The PIAS proteins are reported to act as SUMO-E3 ligases for the SUMO-1 conjugation to AR in vivo and in vitro (34), resulting in inhibition of AR transcriptional activity (22).
Interestingly, several recent studies have shown that the PIAS proteins can have multiple effects on AR activity, depending on the type of PIAS protein, promoter, and cells (reviewed in Ref. 35). The PIAS family is composed of several homologous proteins, including PIAS1, PIAS3, PIASx␣, PIASx␤, and PIASy. Nishida et al. (34) showed that AR-dependent transcription is either repressed by PIAS1 and PIASx␣ in the presence of exogenous SUMO-1 and PIAS RING finger-like domain or enhanced in the absence of sumoylation. PIAS3 inhibits AR transactivation in LNCaP and HeLa cells but enhances AR activity in HepG2 and AR-overexpressing LNCaP cells (36 -38). Moreover, Ubc9, the SUMO E2 enzyme, can stimulate AR transcriptional activity that is independent of its ability to catalyze SUMO-1 conjugation (39). These results demonstrate that the enzymes of the sumoylation pathway can have diverse effects on AR activity. We add to this diversity with the current study, in which we show that SUMO-3 can have a negative or strongly positive effect on AR, depending on the type of prostate cancer cells. Further, the positive effect does not depend on either the sumoylation sites of AR or the sumoylation function of SUMO-3. Finally, SUMO-1 is different from either SUMO-3 or SUMO-2, because it is unable to enhance AR activity.
Cell Transfection and CAT Assays-PC-3 cells were maintained in F12K medium (Sigma) supplemented with 10% fetal bovine serum (FBS) (Hyclone). LNCaP (androgen-dependent and -independent cells) and DU145 cells were maintained in RPMI1640 medium (Sigma) supplemented with 10% FBS. PrEC cells were maintained in the PrEGM medium (Clonetics). C33 and C81 cells (kindly provided by Dr. Lin) (48) were grown in RPMI1640 medium with 5% FBS, 1% glutamine, and 0.5% gentamycin. All cells were cultured with 5% CO 2 at 37°C. These cell lines were grown in RPMI1640 medium with 10% FBS and 0.1 mg/ml neomycin. ␤-Galactosidase assay and CAT assay were described previously (40). For transient transfection of LNCaP and DU145 cells, cells were grown to 80 -90% confluence in RPMI1640 complete medium. Four hours before transfection, medium was changed to RPMI1640 with 5% FBS (dextran-coated charcoal-treated). Transient transfection was performed with FuGene6 reagent (Roche Applied Science). DHT was added 24 h after transfection. LNCaP and DU145 cells were incubated for 24 h in RPMI1640 with 5% FBS (dextran-coated charcoaltreated) with or without DHT, respectively. Reporter analysis (␤-galactosidase assay and CAT assay) were done after the incubation. For PC-3 cells, transient transfection was performed with the calcium phosphate precipitation (CaPO 4 ) method (49). In PrEC cells, transient transfection was performed with FuGene6 in PrEGM medium. Transfection efficiency was standardized by measuring ␤-galactosidase activity, originating from the co-transfected plasmid pCH110 (2 g). Note that for all transfections, empty expression plasmid was used to bring the final plasmid amount to 9 g/transfection.
CAT activity was quantified by scanning with the Bio-Rad Molecular Imager FX of autoradiograms of three independent replicates for each transfection. Thus, each CAT value represents the average of three repetitions plus the S.D.
siRNA Transfections-LNCaP and C81 cells were transfected with a SUMO-2 siRNA (5Ј-GGGAUGAAUCUGUAACUUAtt-3Ј and antisense oligonucleotide) (purchased from Ambion). A luciferase siRNA with 42% GC content was used as control siRNA (GL3 siRNA from Dharmacon). The X-tremeGENE siRNA transfection reagent was used to transfect siRNA into cells following the prescribed protocol (Roche Applied Science).
Generation of Stable Cell Lines-LNCaP cells were grown in 100-mm dishes with 10 ml of RPMI1640 complete medium until 60 -70% confluence, and then cells were transfected with 2 mg of FLAG-SUMO3/ pCI-Neo. The transfection was done with FuGene6 described above. After a 48-h incubation, the LNCaP cells were selected in RPMI1640 complete medium containing 0.9 mg/ml neomycin. The medium was refreshed every 4 days until individual colonies appeared.
The generation of stable LNCaP cell lines C14 and AJ6 and PC-3 lines C-3, A-103, and V-28 has been described (50). The AJ6 LNCaP cells stably express antisense c-Jun. C14 is a control LNCaP cell line stably transfected with an empty pCI-Neo vector. A-103 cells stably express AR, V-28 cells express a fusion protein containing AR and the VP16 activation domain (51), and C-3 cells were transfected with empty PCIneo vector.
Cellular Proliferation Assay-8 ϫ 10 4 LNCaP cells were seeded in 6-well plates with 5 ml of RPMI1640 medium containing 2% dextrancoated charcoal-treated FBS. After a 2-day incubation, either ethanol (vehicle control) or 100 nM DHT was added into the wells. After an additional 2, 4, or 6 days of incubation, the MTT assay was done according to the manufacturer's instructions (Sigma). Cell number was quantified by measuring absorbance at a wavelength of 570 nm. Note that bar graphs represent the averages of thee independent experiments plus S.D.
Western Blot Analysis-For Western blot analysis, COS and LNCaP cells were grown in 100-mm dishes and subjected to transfection. Whole-cell extracts were prepared and subjected to SDS-PAGE and Western blot analysis as described (52). The nitrocellulose blots were probed with the anti-FLAG antibody M2 (Sigma), anti-AR antibody PA1-111A (Affinity Bioreagents), or anti-␤-actin AC-15 (Abcam). The ECL chemiluminescence detection kit (Amersham Biosciences) was used to develop the Western blots.
Northern Blot Analysis-A multiple human tissue blot was obtained from Clontech (catalog no. 7761-1) and probed for SUMO-3 mRNA according to the manufacturer's instructions.
Yeast Transformation for the Functional Screen-The yeast transformation protocol was described previously (53). A 20-ml culture of YPH499/pSH18 -34 (47) was grown and transformed with AR(AB)/ pEG202 in Glu/CMϪUraϪHis liquid dropout medium overnight at 30°C. The culture was diluted into a 300-ml Glu/CMϪUraϪHis liquid dropout medium with 2 ϫ 10 6 cells/ml and incubated at 30°C until A 600 reading reached 0.5, after which it was centrifuged for 5 min at 1000 -1500 ϫ g. The liquid was discarded, and the cells were resuspended in 1.5 ml of TE buffer, 0.1 M lithium acetate. 1 g of P19 library (54) and 50 g of high quality sheared salmon sperm carrier DNA were added to each of 30 sterile 1.5-ml microcentrifuge tubes. 50 l of the resuspended yeast cells were added to each tube, and they were incubated for 30 min at 30°C, and Me 2 SO was then added to 10%. The samples were heatshocked for 10 min at 42°C. For 28 tubes, the complete content of one tube was added per 24 ϫ 24-cm Glu/CMϪUraϪHisϪLeu/X-Gal dropout plate and incubated at 30°C. For the remaining two tubes, 360 l of each tube was spread on 24 ϫ 24-cm Glu/CMϪUraϪHisϪLeu/X-Gal dropout plate. The remaining 40 l from each tube was used to make a series of 1:10 dilution in sterile water. Dilutions were plated on 100-mm Glu/CMϪUraϪHisϪLeu dropout plates. All plates were incubated at 30°C until colonies appeared (2-3 days). Colonies were monitored for color (blue or white). Among 300,000 transformed colonies, nine white colonies appeared on the X-gal medium. Plasmid was isolated from each of the white colonies and used for retransformation. Only four of the isolated plasmids were able to cause a white phenotype upon retrans-formation. DNA sequencing showed that one of the four plasmids matched the open reading frame of mouse SUMO-3, as well as the part of the 5Ј-and 3Ј-untranslated region.
Yeast Liquid ␤-Galactosidase Assay-A single yeast colony was inoculated in 3 ml of YPD (or appropriate selective) medium and incubated overnight at 30°C. 20 -50 l of each overnight culture was inoculated in 4 ml of YPD medium (or appropriate selective medium and/or inducing conditions) and grown to middle or late log phase. This was subjected to a liquid ␤-galactosidase assay using 2-nitrophenyl ␤-D-galactopyranoside as described (55).
DNA Sequencing Analysis-cDNA fragment was isolated from the positive clones using the Yeastmaker TM Yeast Plasmid Isolation Kit (BD Biosciences) and amplified using the upstream oligonucleotide 5Ј-GTTTTTCAAGTTCTTAGATG-3Ј and the downstream oligonucleotide 5Ј-CTGGCAATTCCTTACCTTCC-3Ј (Integrated DNA Technology). DNA sequencing for the nine putative cDNA clones was carried out using the same primers (sequencing done by Ohio State University Plant Microbe Genomics Facility).

SUMO-3 Was Identified Using a Yeast Functional Screen-
In an effort to identify novel repressors of AR, we developed a modified yeast two-hybrid system that we call a "yeast functional screen." Like the yeast two-hybrid system, the yeast functional screen depends on a fusion protein containing the LexA DNA-binding domain fused to a protein of interest and the LexA-responsive LacZ reporter (47). Unlike the yeast two-hybrid system, the protein of interest must have strong transcriptional activity so that all yeast colonies expressing this fusion protein will become blue when grown on X-gal plates. Expression of a protein that blocks the transcriptional activity of the protein of interest will cause yeast colonies to remain white (Fig. 1A). In the present study, the AR(AB) region acted as a bait protein and exhibited strong transcriptional activity in yeast, compared with the activity of LexA alone (Fig.  1B). The AB region harbors most of the AR transcriptional activity (46,56) and is the target of multiple coactivators (44, 46, 56 -63). All yeast transformants expressing LexA-AR(AB) formed blue colonies on X-containing medium. 5 Co-transformation of a cDNA expression library from P19 embryocarcinoma cells (54) resulted in nine white colonies among a total of 300,000 transformants. From these colonies, useful DNA sequence data were obtained for only four, one of which was SUMO-3.  frame of mouse SUMO-3 as well as the part of the 5Ј-and 3Ј-untranslated region.

SUMO-3 Has Different Effects on AR Transactivation in Prostate Cancer
Cell Lines-To verify the negative effect by SUMO-3 on AR transactivation observed in the yeast functional screen, we transiently transfected SUMO-3 and two other members of this gene family into three prostate cancer cell lines (PC-3, LNCaP, and DU145) and PrEC cells (Fig. 2). PC-3 cells were derived from the bone metastasis of a prostate cancer patient and lack endogenous AR (64). Transfected SUMO-3 exhibited a weak but reproducible negative effect (ϳ20%) on AR transactivation in PC-3 cells ( Fig. 2A). SUMO-2 had no effect on AR transactivation, despite the fact that its structure is very similar to that of SUMO-3 (27). SUMO-1 also showed a weak negative effect on AR transactivation, consistent with other reports showing that SUMO-1 can repress several different transcription factors (reviewed in Ref. 65).
To determine whether SUMO-3 can affect the activity of stably expressed AR, we used the PC-3 stable cell lines A103 and V28. A103 cells stably express AR and V28 cells stably express a VP16-AR fusion protein. These two lines exhibited DHT-induced AR transactivation, and VP16-AR has about 10-fold higher transcriptional activity than AR (50). SUMO-3 also had a weak negative effect on AR transactivation in A103 cells (Fig. 2B). In V28 cells, the negative effect was more pronounced (Fig. 2B). These results demonstrate that the weak negative effect of SUMO-3 can be seen with both transiently and stably expressed AR.
To study a potential biological role for SUMO-3 on AR activity, we used LNCaP cells. These cells express endogenous AR and exhibit androgen-dependent gene expression and cellular proliferation (reviewed in Ref. 11). Surprisingly, transfected SUMO-3 significantly enhanced (3.5-fold) AR transactivation in LNCaP cells (Fig. 2C). SUMO-2 also stimulated AR activity, but to a lesser degree than did SUMO-3, whereas SUMO-1 had no effect (Fig. 2C). The endogenous AR gene in LNCaP cells possesses a mutation that replaces residue Thr 877 with Ala, which broadens the AR ligand specificity (15). To study the possibility that this mutation is responsible for the positive effect of SUMO-3 in LNCaP cells, we studied LNCaP cells that were transiently transfected with wild-type AR. Transfected AR has a 3-5-fold higher transcriptional activity than endogenous AR (see In further support of this, we observed a SUMO-3 enhancement of transfected AR activity also in DU-145 cells (Fig. 2D), a prostate cancer cell line derived from a brain metastasis that, like PC-3 cells, lacks endogenous AR (66). These data together show that the direction of the SUMO-3 effect on AR depends on the prostate cancer cell context.
To study the role of SUMO-3 on AR transactivation in normal prostate cells, we used PrEC cells. These cells are normal human prostate epithelial cells that have been widely used as the normal counterpart cell line to prostate cancer cell lines (67). In our experiments, SUMO-3 had a similarly weak negative effect on transfected AR in PrEC cells (Fig. 2E) that was observed in PC-3 cells (see Fig. 2A). Together, these results suggest a novel mechanism of elevating AR activity through the switch of SUMO-3 from a weak negative regulator of AR in normal prostate cells (PrEC) to a strong positive regulator of AR in prostate cancer cells (LNCaP cells). Since enhanced AR activity is probably involved in the development of prostate cancer (reviewed in Ref. 68), SUMO-3 may provide a novel mechanism for up-regulating AR activity in prostate cancer.
SUMO-3 Stimulates AR Transactivation Independent of the Promoter-Our preliminary experiments on the SUMO-3 effect on AR transactivation in LNCaP cells were conducted using the MMTV promoter (see Fig. 2C).
To determine if this effect is unique to the MMTV promoter, we studied the ARE 3 -E1B-CAT reporter, which harbors three AREs and an E1B promoter (5). Since transactivation by endogenous AR of ARE 3 -E1B-CAT is very low in LNCaP cells, 5 exogenous AR was expressed in LNCaP cells together with SUMO-3 (Fig. 3). As is shown in Fig. 3A, transfected AR yielded significantly higher transcriptional activity than did endogenous AR. Importantly, SUMO-3 had a similar positive effect on exogenous AR transactivation of MMTV-CAT (Fig. 3B) and ARE 3 -E1B-CAT (Fig. 3C), demonstrating that SUMO-3 up-regulation of AR transcriptional activity is not promoter-specific.
The SUMO-3-positive Effect on AR Does Not Depend on the Putative AR Sumoylation Sites-SUMO proteins are known to influence the activities of other proteins by direct conjugation to these proteins.  FEBRUARY 17, 2006 • VOLUME 281 • NUMBER 7 Therefore, it is possible that SUMO conjugation to some factor(s) is the mechanism underlying the action of SUMO-3 on AR. This factor may be AR itself or another factor, such as a coactivator. It was recently reported that SUMO-1 can covalently attach to two consensus sumoylation sites found on AR (K385/K519) and inhibit this receptor's transcriptional activity (22). To test if SUMO-3 conjugation to AR is responsible for this protein's positive activity on AR, we obtained a plasmid expressing an AR mutant that has both sumoylation sites disrupted (K385E/K519E). Since the consensus sumoylation site for all three SUMO proteins is the same (27), mutation of Lys 385 and Lys 519 on AR should disrupt any potential SUMO-3 conjugation to AR and possibly block the SUMO-3-positive effect. Surprisingly, in LNCaP cells, SUMO-3 can enhance the transcriptional activity of not only wild-type AR but also the AR mutant K385E/K519E, using either MMTV-CAT or ARE 3 -E1B-CAT as the reporter plasmid (Fig. 4A). We are confident that the transcriptional activity measured above came from transfected AR, since exogenous AR in the experiments yielded significantly higher transcriptional activity than does endogenous AR (see Fig. 3A). To confirm these data from LNCaP cells, we used DU145 cells transfected with either wild-type AR or the AR mutant K385E/K519E. As shown in Fig.  4B, both AR proteins responded positively to co-transfected SUMO-3. Together, the results above indicate that the SUMO-3-positive effect on AR does not depend on the AR putative sumoylation sites, suggesting that the effect is mediated via either other sites on AR or other proteins.

Mutation of AR Sumoylation Sites Alters AR Transactivation-The
consensus SUMO acceptor site, ⌿KXE, overlaps with a synergy control (SC) motif in GR and Sp3, which was demonstrated to repress synergistic transcription from promoters containing multiple binding sites (42). Our results show that mutations in the two SC motifs on the AR N-terminal region (K385E/K519E) led to enhancement of AR transactivation in LNCaP and DU145 cells (Fig. 4, A and B), consistent with previous observations made in COS cells (22). The same result was obtained in PC-3 cells (Fig. 4C). However, in PrEC cells, the AR mutant had lower, not higher, transcriptional activity than did the wild-type AR (Fig. 4D). These data provide the first evidence that mutations of the SC motifs can either enhance or compromise AR transactivation, depending on the type of prostate cancer cell line.
Endogenous c-Jun Is Required for Full Enhancement by SUMO-3 of AR Transactivation-Our published data show that c-Jun can enhance AR transactivation independent of promoter or cell type and targets the AR N terminus (40,46,52,69). In view of these results and the finding that c-Jun is a substrate for sumoylation (70), it is possible that SUMO modification of this protein may be involved in the SUMO-3-positive effect on AR in LNCaP cells. To address this possibility, we used an LNCaP-stable cell line expressing an antisense c-Jun transcript. These cells, called AJ6, exhibit reduced endogenous c-Jun levels, AR transactivation, and androgen-dependent proliferation (50). As expected, AR transactivation in AJ6 cells is significantly lower than in control C14 cells (Fig. 5A), which are transfected with empty expression vector. When analyzed for a SUMO-3 effect, AJ6 cells exhibit a greatly reduced response to transfected SUMO-3 on AR transactivation, suggesting that endogenous c-Jun is required for full enhancement by SUMO-3 of AR transactivation.

TIF-2 Represses SUMO-3-positive Effect on AR Transactivation in LNCaP Cells-
The SRC family proteins represent another group of AR coactivators that act by interacting with the N and C termini of the receptor (57, 58) and recruiting additional cofactors such as CBP/p300 and protein with acetyltransferase or methyltransferase activity (71). Recently, SRC-1 (72) and TIF-2 (73) have been shown to be sumoylated by SUMO-1. Furthermore, mutation of the consensus sumoylation sites found on TIF-2 can impair its coactivation ability on AR (73). Based on these observations, SUMO modification of SRC family proteins may be involved in the SUMO-3-positive effect on AR in LNCaP cells. Transfected SRC-1, SRC-3, and TIF-2 were analyzed for modulatory effects on the SUMO-3 stimulation of AR activity in LNCaP cells. In the absence of transfected SUMO-3, SRC-3 and SRC-1 stimulated AR transactivation, whereas TIF-2, surprisingly, had no measurable effect (Fig. 5B). By contrast, in the presence of transfected SUMO-3, TIF-2 had the strongest effect, markedly impairing the positive effect of SUMO-3 on AR activity (Fig. 7B); SRC-1 and SRC-3 had an appreciably weaker, if any, effect (Fig. 5B). These results show that TIF-2 can down-modulate the SUMO-3-positive effect on AR and therefore may be involved in this effect.
A Conjugation-deficient SUMO-3 Mutant Can Enhance AR Transactivation in LNCaP Cells-The double glycine residues on the C terminus of SUMO proteins have been shown to be critical for the sumoylation pathway (74,75). Deletion of these two glycine residues in SUMO-1 leads to the loss of conjugation to AR (22), a finding confirmed by our results with the SUMO-3(⌬GG) mutant (Fig. 6A). Transfection into COS cells of wild-type SUMO-3 yields multiple bands on a Western blot probed with an anti-SUMO-3 antibody; these multiple bands include unconjugated monomeric SUMO-3 and conjugated SUMO-3 in the upper half of the blot (Fig. 6B). In contrast, only the monomeric form is observed with transfected SUMO-3(⌬GG) (Fig. 5B), indicating that this mutant protein is deficient in conjugation to target substrates. Importantly, however, this SUMO-3 mutant still can enhance AR transactivation in LNCaP cells, although somewhat more weakly than wild-type protein (Fig. 6C), suggesting that SUMO-3 enhances AR transactivation via a conjugation-independent mechanism.

Stable Expression of SUMO-3 in LNCaP Cells Stimulates Cell Proliferation-Our data above show that transient transfection of SUMO-3 can enhance AR transactivation in LNCaP cells.
To determine if this SUMO-3 regulation of AR can modulate a biological function of this receptor, we opted to increase the endogenous levels of SUMO-3 in LNCaP cells and monitor androgen-dependent cellular proliferation. Two LNCaP cell lines were selected that stably express FLAG-SUMO-3 (Fig. 7A). Stable expression of SUMO-3 does not affect the expression of AR protein (Fig.  7A) or mRNA (Fig. 7B). Importantly, these two cell lines also demonstrated enhanced androgen-dependent cellular proliferation, when compared with control C14 cells (Fig. 7C). These data support the hypothesis that SUMO-3 can regulate a biological activity of AR.
siRNA-mediated Repression of SUMO-2 Inhibits the Proliferation of Androgen-dependent LNCaP Cells-To study the role of endogenous SUMO-3 in cellular proliferation, we used siRNA-mediated down-regulation. Unfortunately, no SUMO-3 commercial siRNA oligonucleotides are available, nor is the SUMO-3 nucleotide sequence amenable to custom-designed oligonucleotides. Therefore, we opted to use commercial siRNA oligonucleotides for SUMO-2, a nearly identical protein to SUMO-3 in both amino acid sequence (27) and enhancing activity on AR (see Fig. 2C). Transfection of SUMO-2 siRNA strongly reduces the expression of SUMO-2 mRNA, whereas a control siRNA did not (Fig.  8A). The SUMO-2 siRNA had no effect on SUMO-3 mRNA expression (Fig. 8A). Significantly, siRNA-mediated reduction of SUMO-2 expression resulted in a marked decrease in androgen-induced LNCaP proliferation, whereas control siRNA had no effect (Fig. 8B). These results demonstrate a direct correlation between endogenous SUMO-2 protein levels and LNCaP cellular proliferation, strongly suggesting an essential role for SUMO-2, and probably SUMO-3, protein.
siRNA-mediated Repression of SUMO-2 Inhibits the Proliferation of Androgen-independent LNCaP Cells-Prostate cancer can become androgen-independent during disease progression or androgen deprivation. Numerous molecules have been shown to be involved in regulating cell growth in androgen-independent prostate cancer, including  AR, p53, Bcl2, IGF1, and several other factors (reviewed in Ref. 76). The positive effect of SUMO-3 on AR activity may also contribute to AR transcriptional activity in androgen-independent prostate cancer cells that express endogenous AR. To address this issue, we have compared the effect of transfected SUMO-3 on AR in androgen-dependent and androgen-independent prostate cancer cells. C81 represents LNCaP cells that were cultured for a long time to exhibit androgen-independent growth, whereas C33 cells are the androgen-dependent parental line from which C81 was derived (48). AR expression in these androgenindependent LNCaP cells is similar to expression in androgen-dependent LNCaP cells (48). Transfected SUMO-3 can enhance androgeninduced AR transactivation in C33 cells (Fig. 9A). Interestingly, SUMO-3 also had a positive effect on the endogenous AR in C81 cells, but the magnitude of the effect was attenuated (Fig. 9A). This reduced SUMO-3 effect may reflect the lower androgen responsiveness of AR transactivation in androgen-independent LNCaP cells (Fig. 9A).
To determine if endogenous SUMO proteins are involved in the growth of androgen-independent LNCaP cells, we utilized again siRNA-mediated repression of SUMO-2. As shown in Fig. 9B, SUMO-2 and SUMO-3 mRNAs are both expressed in C81 cells to slightly higher levels than in androgen-dependent LNCaP cells. Similarly, transfection of SUMO-2 siRNA results in a marked decrease in SUMO-2 mRNA expression in C81 cells (Fig. 9C). These siRNA-transfected cells were then studied for growth, which expectedly is largely androgen-independent (Fig. 9D). More importantly, the androgen-independent proliferation of C81 cells is dramatically reduced by SUMO-2 siRNA treatment, in comparison with control siRNA treatment (Fig. 9D). These data illustrate an important role for endogenous SUMO-2 in androgen-independent growth of LNCaP cells.
SUMO-3 mRNA Is Highly Expressed in Human Prostate and Testis and Prostate Cancer Cell Lines-To detect the expression of SUMO-3 in human tissues, Northern blot analysis was carried out using a human

SUMO-3 Activates AR-dependent Transcription
multiple tissue blot (from Clontech). SUMO-3 mRNA was ubiquitously expressed in all human tissues studied, with the highest expression found in prostate, testis, and thymus (Fig. 10A).
RT-PCR was employed to measure the expression of the three SUMO genes in LNCaP, PC-3, and PrEC cells (Fig. 10B). SUMO-3 was detected in all three cell types. SUMO-2 and SUMO-1 are also expressed in these cells, suggesting that the sumoylation pathway is abundant and/or three SUMO proteins act on variety of targets. In the sumoylation pathway, the E1 enzymes SAE1 and SAE2 activate SUMO and transfer it to the E2 enzyme Ubc9 (27). Ubc9 conjugates to SUMO and directs it to target substrates (27). If sumoylation is a biological regulator of AR in LNCaP cells, then the SUMO E1 and E2 enzymes must be involved, and the genes encoding these enzymes must be expressed in these cells. As shown in Fig. 9B, SAE1, and SAE2, and Ubc9 are all expressed in LNCaP cells.

DISCUSSION
Sumoylation is a novel pathway for post-translational modification (reviewed in Ref. 23). In mammalian cells, many transcription factors and cofactors have been identified as targets of sumoylation (reviewed in Refs. 24 and 77). SUMO modification can either enhance or repress the transcriptional activity of these factors. In the case of nuclear receptors, exogenous SUMO-1 significantly increases glucocorticoid receptor (GR) transactivation of a glucocorticoid-responsive reporter (78). Similarly, transactivation by either the progesterone receptor and estrogen receptor ␣ can be up-regulated by SUMO-1 (72). Mutations that prevent SUMO-1 from binding to AR can enhance AR transactivation, indicating a negative role for SUMO-1 on AR (22). To date, few studies have focused on SUMO-3. Recently, Li et al. (79) found that SUMO-3 can either reduce or increase the generation of amyloid ␤ peptide, critical to Alzheimer disease.   FEBRUARY 17, 2006 • VOLUME 281 • NUMBER 7

SUMO-3 Activates AR-dependent Transcription
In our study, overexpression of SUMO-3 or SUMO-2 markedly enhanced AR transactivation in LNCaP and DU145 cells. However, importantly, SUMO-3 has no significant effect in normal primary prostate epithelial cells, suggesting that the SUMO-3 effect on AR may be at least partially responsible for the elevated AR activity that is commonly observed in prostate cancer. Additionally, the SUMO-3-positive effect on AR is not observed in PC-3 cells, a cell response difference that may reflect the cellular heterogeneity of prostate tumors.
What causes the different effects of SUMO-3 on AR transactivation? LNCaP cells contain a mutated AR, which leads to low ligand binding specificity (15). Our data show that wild-type AR also can be up-regulated by SUMO-3 in both LNCaP and DU145 cells, demonstrating that the mutation of AR is not responsible for the positive effect of SUMO-3 on AR transactivation. Another possible explanation comes from evidence showing that the family of PIAS proteins may act as E3 ligases in the SUMO pathway, especially for nuclear receptors. Whereas the PIAS proteins were originally identified as an inhibitor of STAT transcription factors (80), recent studies have shown that PIAS1, PIAS3, and PIASx␣ regulate AR transactivation in a cell-and promoter-specific manner (34,36,81,82). Moreover, PIAS1 and PIASx␣ increase the sumoylation of AR (34). Therefore, differences in expression levels of the various PIAS proteins in prostate cancer cell lines may contribute to the cell-specific effects of SUMO-3 on AR transactivation. Interestingly, the positive effect of SUMO-3 on AR in LNCaP cells was observed on both the natural promoter MMTV and the artificial androgen-inducible promoter ARE 3 -E1B, suggesting that this action, unlike the activities of the PIAS proteins, is not dependent on the promoter context. A third possibility is that an unknown factor in LNCaP cells mediates the SUMO-3-positive effect on AR transactivation, and this factor is not expressed in PC-3 or PrEC cells. Such a factor may be a PIAS protein or one of many known AR coactivators.
Our previous data showed that c-Jun targets the N terminus of AR to enhance androgen-dependent transactivation (46). In this study, we observed that in the anti-c-Jun stable LNCaP cell line AJ6, SUMO-3 still possessed a positive effect on AR transactivation, although the magnitude was lower than in control cells. This suggests that endogenous c-Jun is required for full enhancement by SUMO-3 of AR transactivation. However, since there are no c-Jun expression differences among the various prostate cancer cell lines used in this study, 5 it is unlikely that c-Jun is directly responsible for the cell-specific effects of SUMO-3 on AR.
As a coactivator, TIF-2 can increase AR-mediated transactivation of the prostate-specific antigen promoter in LNCaP cells (83). TIF-2 also is modified by SUMO-1 (37). Mutation of the consensus sumoylation sites on TIF-2 impaired its coactivation function and cooperation with PIAS1, a SUMO E3 ligase, on AR transactivation in COS cells (37). In cultured cells, the mRNA levels of TIF-2 are significantly higher in PC-3 cells than in LNCaP cells (84). Furthermore, using Western blot analysis of human prostate tissues, Gregory et al. (85) reported that TIF-2 was not detected in BPH and androgen-dependent tumors but was overexpressed in recurrent prostate tumors. Our results showed that exogenous TIF-2 can partially repress the SUMO-3-positive effect in LNCaP cells. These data support the possibility that the low expression of TIF-2 in LNCaP cells may contribute to the positive effect of SUMO-3 on AR transactivation. The specific mechanism of how TIF-2 is involved in the SUMO-3 effect is unclear. Since TIF-2 is a known target of SUMO-1 conjugation (37), it is possible that sumoylation of TIF-2 titrates SUMO-3 away from the pathway that leads to the positive effect on AR. Moreover, HDACs, PIAS proteins, and SRC-1 can all be modified by SUMO-1 (reviewed in Ref. 24). Since SUMO-3 is a functionally homologous protein to SUMO-1, it is possible that SUMO-3 acts on AR by modifying these AR coregulators.
The consensus SUMO acceptor site, ⌿KXE, found in GR and Sp3 has been referred to as an SC motif, mutation of which abolishes the sumoylation of GR and Sp3 and significantly increases GR and Sp3 transactivation on promoters containing multiple responsive elements (42). Replacement of lysine with arginine in this motif abolishes SUMO-1 conjugation of AR and markedly increases AR transactivation (22). Our results support this earlier finding, since mutation of the two SC motifs found on the N-terminal region (K385E/K519E) of AR led to enhancement of AR transactivation in LNCaP, DU145, and PC-3 cells. However, surprisingly, in the primary prostate epithelial cells, PrEC, the AR K385E/K519E mutant had lower activity than the wild-type AR. These data provide the first evidence that mutation of the SC motifs can either enhance or compromise AR transactivation, depending on prostate cell type.
More importantly, we observed that mutation of the AR SC motifs did not alter the positive SUMO-3 effect on AR transactivation in either LNCaP or DU145 cells. This finding suggests that the SUMO-3-positive effect on AR is mediated either by other unknown AR sumoylation sites or another factor acting on AR. Our data support the second possibility. Using immunoprecipitation and Western blot studies, we have been unable to obtain evidence for either sumoylated forms of AR or AR physical interaction with SUMO-3 in LNCaP, PC-3, and COS cells. 5 This includes the stable LNCaP cell lines expressing SUMO-3 (FS11 and FS31 cells). In all these cell lines, high molecular weight, SUMO-3containing complexes can be obtained upon transfection of SUMO-3; however, no SUMO-3-modified AR was detected. 5 Support for an indi- A, human multiple tissue Northern blot (Clontech) was probed for SUMO-3, using random prime labeling to synthesize the probe. B, total RNA was isolated from LNCaP, PC-3, and PrEC cells and subjected to RT-PCR using primers for SUMO-1, SUMO-2, SUMO-3, SAE1, SAE2, Ubc9, or glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as a control. Note that RT-minus control templates gave no PCR products. 5 rect effect of SUMO proteins on nuclear receptors comes from several studies. Mutation of the consensus sumoylation site on progesterone receptor does not impair the effect of SUMO-1 on this receptor (72). Moreover, estrogen receptor ␣ transactivation was up-regulated by SUMO-1, although it has no consensus sumoylation sites (42,72). For a nonnuclear receptor protein, Li et al. (79) recently showed that SUMO-3 indirectly regulates the processing of amyloid precursor protein (APP), since the APP sequence does not contain the consensus sumoylation sites, and no sumoylated APP was detected in immunoprecipitation analysis.
If SUMO conjugation to AR is not involved in the SUMO-3 enhancement of AR activity, then it is possible that SUMO-3 conjugation to another factor is involved. Whereas this cannot be excluded, our data demonstrate that the sumoylation activity of SUMO is not necessary for mediating AR activity. Indeed, deletion of the two C-terminal glycine residues, SUMO-3(⌬GG), which renders SUMO-3 deficient in conjugation to target proteins, does not disrupt the SUMO-3 enhancement of AR transactivation. These results strongly argue that SUMO-3 can mediate AR-dependent transcription in a sumoylation-independent manner, making it mechanistically distinct from the SUMO-1-dependent repression of AR activity (22). Moreover, our findings suggest that SUMO-3 (and perhaps other SUMO proteins) harbors conjugationindependent activities. This possibility is supported by a recent observation that SUMO-3 can coactivate of EBNA2 (Epstein-Barr virus nuclear antigen 2) in the absence of direct conjugation to EBNA2 (86).
Prostate cancer often becomes hormone-refractory after androgen ablation therapy (reviewed in Ref. 87). In this stage, cell proliferation becomes androgen-independent and does not respond to currently used androgen antagonists. Despite this, AR transactivation remains intact in most androgen-independent tumors as indicated by expression of prostate-specific antigen and other AR target genes (87). This transition from androgen-dependent to androgen-independent prostate tumors has been reproduced in cultured LNCaP cells (48), resulting in the generation of the C81 cell line. C81 cells are androgen-independent LNCaP cells that have similar AR protein expression to the parental cells, C33 (48). We observed that androgen-induced AR transactivation in C81 cells is weaker than in C33 cells. Indeed, Igawa et al. (48) reported that androgen-induced cell proliferation and expression of prostatespecific antigen is weak in C81 cells. Additionally, our results demonstrate that the SUMO-3-positive effect on AR is intact in C81 cells, although the magnitude is lower than in C33 cells. More significantly, reducing the expression by siRNA of SUMO-2 significantly compromises the growth of both androgen-dependent and androgen-independent LNCaP cells. Because of the high amino acid sequence similarity between SUMO-2 and SUMO-3, these two proteins are believed to be functional homologues (27). Indeed, our data here clearly show that both SUMO-3 and SUMO-2 can enhance AR activity. Thus, our SUMO-2 siRNA results suggest that the enhancing activity of either SUMO-2 or SUMO-3 on AR transactivation is involved in the growth of prostate cancer cells, implicating a potentially significant role for these two proteins in androgen-independent prostate tumors.
Although sumoylation appears to represent a widespread mechanism of regulation of protein activity and subcellular localization (reviewed in Ref. 25), the biological consequence of this post-translational modification is not fully understood. Recently, Steffan et al. (26) reported that SUMO-1 can modify the pathogenic fragment of protein Huntington (Httex1p), which is accumulated in Huntington disease. In a Drosophila model of Huntington disease, unsumoylated Httex1p significantly reduced the cytotoxicity of this protein, suggesting that sumoylation may aggravate neurodegeneration disease (26). In this study, we deter-mined that SUMO-3 is ubiquitously expressed in human tissues, especially high in prostate, testis, and thyroid. Importantly, overexpression of SUMO-3 in LNCaP cells leads to a significant increase in androgeninduced cellular proliferation. This finding supports the contention that SUMO-3 is a biological regulator of AR transactivation and, therefore, suggests that SUMO-3 may have a role in prostate carcinogenesis.