Pax6 Represses Androgen Receptor-Mediated Transactivation by Inhibiting Recruitment of the Coactivator SPBP

The androgen receptor (AR) has a central role in development and maintenance of the male reproductive system and in the etiology of prostate cancer. The transcription factor Pax6 has recently been reported to act as a repressor of AR and to be hypermethylated in prostate cancer cells. SPBP is a transcriptional regulator that previously has been shown to enhance the activity of Pax6. In this study we have identified SPBP to act as a transcriptional coactivator of AR. We also show that Pax6 inhibits SPBP-mediated enhancement of AR activity on the AR target gene probasin promoter, a repression that was partly reversed by increased expression of SPBP. Enhanced expression of Pax6 reduced the amount of SPBP associated with the probasin promoter when assayed by ChIP in HeLa cells. We mapped the interaction between both AR and SPBP, and AR and Pax6 to the DNA-binding domains of the involved proteins. Further binding studies revealed that Pax6 and SPBP compete for binding to AR. These results suggest that Pax6 represses AR activity by displacing and/or inhibiting recruitment of coactivators to AR target promoters. Understanding the mechanism for inhibition of AR coactivators can give rise to molecular targeted drugs for treatment of prostate cancer.


Introduction
The androgen receptor (AR) and androgens have a central role in development and maintenance of the male reproductive system and in the etiology of prostate cancer [1,2], the most commonly diagnosed invasive cancer in men in the USA and other western countries [3]. Since prostate cancers are dependent on androgens for development, growth and survival, inhibition of AR activity is a major therapeutic goal for management of the disease (reviewed in [4]). Unliganded AR is mainly localized to the cytoplasm through interactions between its ligand binding domain (LBD) and heatshock proteins (HSPs). Binding of androgens to the LBD induces conformational changes that cause dissociation of the HSPs [2]. The receptor subsequently dimerizes and becomes phosphorylated, before the receptor-steroid complex is imported into the nucleus. Here it binds with high affinity to specific DNAregulatory elements of target genes [1]. The context of the promoter and the relative levels of coactivators and corepressors determine the resulting transcriptional outcome of the DNA bound receptor [5]. AR has four functional domains; the divergent N-terminal domain (NTD), a central DNA-binding domain (DBD), a hinge-region, and a C-terminal LBD [1,2]. The NTD contains poly-glutamine and -glycine repeats as well as the transcriptional activation function-1 site (AF-1) that is constitutively active independent of androgens [6]. The LBD has a second activator function site (AF-2) that is only active in the presence of androgens [2]. Both AF-1 and AF-2 can activate AR, whereas interactions between them can modulate its activity [7].
Stromelysin-1 PDGF-responsive element Binding Protein (SPBP) is a 220 kDa nuclear protein displaying structural and functional properties of a transcriptional regulator. It contains an N-terminal TAD, three nuclear localization signals (NLS), a DBD containing an AT-hook motif, and a C-terminal extended plant homeodomain (ePHD) [8]. SPBP enhances the activity of the transcription factors c-Jun, Pax6, Sp1 and Ets1 [8,9], and knockdown of SPBP significantly reduces the expression of the Ets1 target gene MMP3 in cells [10]. SPBP has also been identified as an interaction partner for activated Estrogen Receptor a (ERa), and found to work as a corepressor of this particular steroid hormone [11]. Further, SPBP is shown to bind and enhance the activation potential of the transcriptional cofactor RNF4 [12]. RNF4 was originally identified as a coactivator of AR, enhancing both basal and ligand-induced AR-mediated transcriptional activity [13].
Pax6 is an evolutionary conserved transcription factor consisting of two DBDs, the paired domain (PD) and the paired-type homeodomain (HD), separated by a glycine-rich linker sequence [14][15][16]. In addition, a TAD is located in the C-terminal region [17]. The N-terminal PD is structurally and functionally divided into the PAI and RED subdomains [18,19]. These subdomains, as well as the HD, adopt helix-turn-helix motifs that can bind independently to DNA [19]. The PD can also bind to the HD of Pax6, and to other HD containing proteins [20,21]. Pax6 is important for embryogenesis of the eye, brain, spinal cord, pancreas, and olfactory epithelium [14,16,22]. The developmental role of Pax6 and the other Pax family proteins has been established for quite some time. More recently, it has been acknowledged that PAX genes are expressed in cancer cells and tissues where they are not normally found [23]. While most Pax family members are associated with an unfavorable outcome of the cancer, Pax6 has been proposed to have a tumor suppressor function in both glioblastoma [24][25][26][27][28] and prostate cancer [29].
In this study, we have identified SPBP as a transcriptional coactivator of AR. SPBP interacts with the DBD of AR via its own DBD. We also mapped the interaction between AR and the AR repressor and developmental transcription factor Pax6 to involve their respective DBDs. Importantly, Pax6 inhibited the SPBPmediated enhancement of AR activity on the AR targeted probasin promoter. This repression was partly reversed by increased expression of SPBP. Chromatin immunoprecipitations showed that over-expression of Pax6 resulted in less SPBP associated with the probasin promoter. In vitro binding studies revealed that Pax6 and SPBP compete for binding to the AR(DBD). This suggests that Pax6 represses AR activity by displacing and/or inhibiting recruitment of coactivators to AR target promoters.

Plasmid constructs
Plasmids and primers used in this study are listed in Tables S1 and S2, respectively. Details on their construction are available on request. Primers for PCR and DNA sequencing were purchased from Eurogentec and Operon. PCR amplification of cDNA was done using either Pfu Turbo (Stratagene) or ExTaq (TaKaRa) polymerases. Many of the plasmids were made using the Gateway cloning system (Invitrogen). All constructs were verified by DNA sequencing with BigDye v3.1 (Applied Biosystems).

Reporter gene assays
Subconfluent HEK293 or LNCaP cells in 24-well tissue culture dishes (Becton Dickinson) were transiently transfected using Metafectene Pro (Biontex) according to the manufacturer's protocol. The cells were cultured in 5% charcoal treated serum one day before transfection. Medium with 0.5% charcoal treated serum and 10 27 M synthetic androgen R1881 (Perkin-Elmer) was (if indicated) added to the cells 1 hour before transfection. All wells were cotransfected with 75 ng pDestHA-AR, 75 ng of the luciferase reporter p285PB-Luc [31], and 5 ng pCMV-b-galactosidase (Stratagene). In addition, the HEK293cells were cotransfected with 0-25 ng pDestHA-Pax6, and 50 or 150 ng pDestHA-SPBP, as indicated in Figure 1A. The LNCaP cells were in addition cotransfected with 5-200 ng pDestHA-Pax6 as indicated in Figure 1B. pcDNA3HA (Invitrogen) was used to equalize the concentration of DNA in each transfection. Extracts were prepared 20 hours after transfection using a Dual-Light luciferase and b-galactosidase reporter gene assay system (Tropix), and subsequently analyzed in a Labsystems Luminoskan RT dual injection luminometer. The assays were performed in triplicates and repeated at least three times.

Real-Time PCR
Subconfluent LNCaP cells in 6 well dishes were tansfected with 20 nM SPBP siRNA [10] or scrambled siRNA [10] using RNAiMAX (Invitrogen). The cells were stimulated by R1881 (10 27 M) for 48 hours post transfection. RNA was isolated using RNAeasy Plus Minikit (QIAGEN), cDNA made using Transcriptor Universal cDNA Master (Roche), and RT-PCR performed on a STRATAGENE 6300 amplification system using FastStart Universal SYBR Green Master (Roche). Primers were PSA-fw, PSA-rev, Actin-fw and Actin-rev (Table S2). The reactions were run for an initial step at 95uC for 10 min, followed by 40 cycles of amplification at 95uC for 15 s and 60uC for 1 min. All data were collected during the extension step, and a melting curve was obtained at the end of the PCR reaction to verify that only one product was produced.

Confocal microscopy
GFP-AR expression was induced from the GFP-AR 3108 cell line by removing the tetracycline from the medium 3 days before the experiments were performed. G418 and puromycin were also removed from the medium to improve the expression. The cells were cultured in 8-chambered cover slides (Nunc) and transiently transfected with 100 ng pDestCherry-Pax6 or 150 ng pDestCherry-SPBP using TransIT-LT1 (Mirus Bio) according to the manufacturer's protocol. As described previously, the cells were starved in medium with charcoal-treated serum and hormone induced with R1881. Subconfluent HeLa cells in 8-well coverglass slides were transiently co-transfected with 175 ng pDestEGFP-SPBP and 25 ng pDestmCherry-Pax6 using TransIT-LT1. Live cell images were taken the following day after transfection using a Zeiss LSM510 confocal laser scanning microscope. Colocalization was studied using the Zeiss LSM 510 software.

Coimmunoprecipitation
For coimmunoprecipitation of AR and Pax6, expression of 36Flag-Pax6 was induced from HeLa FlpIn 36Flag-Pax6 cells by adding 1 mg/ml tetracycline. Subconfluent cells in 10 cm dishes (Nunc) were transiently transfected with either pDestEGFP-AR or pEGFP-C1 using Metafectene Pro (Biontex) according to the manufacturer's protocol. R1881 (10 27 M) was added to stimulate the AR. For coimmunoprecipitation of AR and SPBP, 2610 5 HeLa Figure 1. Pax6 represses SPBP-mediated enhancement of AR activity. (A) Pax6 represses AR-mediated expression from the 285PB-Luc reporter. LNCaP cells were cultured in medium with charcoal treated serum and cotransfected with 5-200 ng pDestHA-Pax6, 75 ng pDestHA-AR, 75 ng p285PB-Luc, and 5 ng pCMV-bgal using Metafectene Pro (Biontex). pcDNA3HA was used to equalize the amount of DNA in each well. 10 27 M synthetic androgen (R1881) was added as indicated. The mean luciferase/b galactosidase value of AR-mediated expression from 285PB-Luc after stimulation with R1881 was set to 1. The data is based on the mean values of three independent experiments performed in triplicate. (B) SPBP enhances AR-mediated expression, a coactivation that is lost by coexpression of Pax6. HEK293 cells were cultured in medium with charcoal treated serum and cotransfected with 75 ng pDestHA-AR, 50 or 150 ng pDestHA-SPBP, 0-25 ng pDestHA-Pax6, 75 ng p285PB-Luc, and 5 ng pCMV-bgal. The experiment was performed and data obtained as described in A. (C) Chromatin Immunoprecipitation assays show that both SPBP and AR associate with the probasin promoter, and that overexpression of Pax6 decreases the amount of SPBP associated with the promoter. Extracts from HeLa cells cotransfected with 285PB-Luc, pDestHA-SPBP, pSG5-AR, and/or increasing amounts of pDestHA-Pax6 were immunoprecipitated with preimmune serum, polyclonal anti-SPBP antibody or polyclonal anti-AR antibody. PCR analyses on the immuno- cells were seeded in 10 cm dishes in MEM supplied with 5% charcoal treated serum, and incubated for 18-20 hours. The cells were then transiently cotransfected with 3 mg pDestEGFP-AR and 7 mg pDestHA-SPBP, using the Calcium Phosphate transfection procedure [32]. 3-4 hours after transfection, the cells were stimulated using R1881 (10 27 M). The cells from both experiments were harvested 20 hours post transfection and immonoprecipitaed essentially as described [10]. Primary antibodies used in the western blot were mouse anti-Flag M2 (1:5.000; Stratagene), rabbit anti-SPBP (1:500, [10]) or rabbit anti-GFP (1:1.000; Abcam).
For in vitro coimmunoprecipitation assays, expression vectors pDestHA-SPBP and pDest53-AR were in vitro translated in the presence of 35 S-methionine and immunoprecipiated as described [10].

Fluorescence resonance energy transfer (FRET)
FRET with photobleaching of the acceptor was done using a Zeiss LSM510 META confocal microscope as described [20]. HeLa cells were cotransfected with 50 ng of both CFP-and YFPtagged expression constructs using LipofectAmine PLUS (Invitrogen). Three to five cell nuclei expressing a 1:1 ratio of the CFP-and YFP-tagged proteins were examined per protein pair. A CFP-YFP fusion protein with a linker of 52 residues (pDestECFP-re7-EYFP), and a CFP-Pax6-YFP fusion protein (pDestECFP-Pax6-EYFP) were used as positive FRET controls. The following protein pairs were used as negative controls: CFP (pDestECFP-C1) and YFP (pDestEYFP-C1), CFP-AR (pDestECFP-AR) and YFP, and CFP and Pax6-YFP (pDestPax6-EYFP). Percent FRET efficiency between CFP-and YFP-tagged proteins was calculated as the difference between CFP fluorescence before and after bleach using the formula: ((A1-A0)/A1)-((B1-B0)/B1)/BkYFP-Pax66100%. A1 and A0 are CFP emission in the region of interest (ROI) in the bleached cell after and before photobleaching, while B1 and B0 are CFP emission in a non-bleached ROI in a control cell after and before photobleaching, respectively [33]. The bleach constant Bk for YFP-Pax6 was experimentally found to be 0.68.

GST and MBP pulldown assays
GST and GST fusion proteins were expressed in Escherichia coli BL21STAR (lDE3) pLysS (Novagen) and purified from extracts using glutathione-sepharose 4 fast flow beads (Amersham Biosciences). MBP and MBP fusion proteins were expressed in E. coli DH5a and purified on amylose resin (New England Biolabs). In vitro translation and dilution of 35 S-labeled proteins were performed as described for in vitro coimmunoprecipitation. GST or MBP bound to beads were added to the samples for 30 minutes at 4uC on a rotating wheel to remove unspecific binding. The supernatants were subsequently mixed with 1-2 mg of the indicated fusion proteins immobilized on beads and incubated at 4uC for 1 hour. After 5 times washing in NET-N buffer, the beads were resuspended in 26 SDS gel loading buffer and boiled for 5 minutes. If indicated, 10 27 M synthetic androgen R1881 was added to the reactions. The samples were separated on 10% SDS-PAGE gels together with a molecular weight marker (New England Biolabs #P7703S), stained with Coomassie brilliant blue (CBB) and vacuum dried. Signal from 35 S-labeled proteins was detected with a Fujifilm bioimaging analyzer FUJI-BAS5000.

Pax6 represses SPBP-mediated enhancement of AR activity
Recent reports show Pax6 to be a repressor of AR activity [29], and the PAX6 gene to be hypermethylated in prostate cancer cells [34]. p285PB-Luc contains a fragment of the AR targeted rat probasin promoter upstream of the luciferase gene [31]. Using p285PB-Luc as a reporter, we confirm that Pax6 represses ARmediated transactivation in LNCaP prostate cancer cells ( Figure 1A). These results correlate well with other reports suggesting Pax6 to have a tumor suppressor function [24][25][26][27][28].
We have previously found that the nuclear protein SPBP acts as a coactivator of Pax6 [8]. Interestingly, SPBP also interacts with phosphorylated ERa and represses its transcriptional activity [11]. These results encouraged us to investigate whether Pax6 and SPBP together would have any impact on AR activity. To this end, luciferase reporter gene assays were performed in HEK293 cells using p285PB-Luc as a model system. Figure 1B shows that SPBP alone enhanced the transcriptional activity of AR in a dose dependent manner, hence acting as a transcriptional coactivator of AR. Coexpression of SPBP and increasing concentrations of Pax6 resulted in corresponding increased repression of the SPBPmediated enhancement of AR activity. However, increasing ectopic expression of SPBP seems to partly restore the activity suppressed by Pax6, suggesting a competitional recruitment to AR.
Chromatin immunoprecipitation assays were performed to investigate whether SPBP was recruited to the AR target promoter probasin, and if Pax6 could affect this recruitment. In order to obtain expression of the three proteins in the same cells, with varying amounts of Pax6, HeLa cells were cotransfected with the probasin-luciferase promoter construct and vectors expressing AR, SPBP and/or Pax6. The results presented in Figure 1C (upper panel) clearly show that SPBP and AR are associated with the same fragment of the probasin promoter. Furthermore, expression of increasing amounts of Pax6 reduces the amount of SPBP associated with the promoter. An upstream fragment of the Cathepsin D promoter was amplified on the same samples to verify specificity ( Figure 1C, lower panel). Hence, Pax6 seems to inhibit recruitment of SPBP to the AR target probasin promoter.
As shown in Figure 1B, SPBP enhances AR activity only modestly in reporter gene assays using exogenous proteins. To determine whether SPBP has an effect on the expression of an endogenous AR target gene, SPBP siRNAs were transfected into LNCaP cells followed by monitoring the expression level of PSA mRNA by RT-PCR. SPBP expression is highly knocked down by the SPBP siRNAs ( Figure 1D, right panel). Importantly, PSA mRNA is reduced to approximately 40% in the SPBP knock down cells compared to scrambled siRNA ( Figure 1D, left panel). These results show that SPBP enhances expression of AR target genes also in an endogenous system. The modest effect of SPBP in the reporter gene system may be due to low expression levels of exogenous SPBP compared to exogenous Pax6 and AR ( Figure S1).
Together, these results show that SPBP acts as a potent transcriptional coactivator of AR, while Pax6 acts as a repressor inhibiting recruitment of SPBP to the AR target promoter and thereby reduces the SPBP-mediated enhancement of AR activity.

AR, SPBP and Pax6 display a very similar nuclear distribution pattern in androgen stimulated cells
To further evaluate the putative association among Pax6, SPBP and AR, their nuclear distribution patterns were investigated by confocal laser scanning microscopy. The mammary adenocarcinoma cell line GFP-AR 3108 stably expressing GFP-AR [30] was transiently transfected with plasmids expressing Cherry-Pax6 or Cherry-SPBP (Figure 2A and 2B, respectively). Both Cherry-Pax6 and Cherry-SPBP strongly colocalized with GFP-AR in the nucleus of nearly all 3108 cells upon synthetic androgen (R1881) stimulation. The proteins are located throughout the nucleus except from the nucleolus, in a subnuclear localization pattern we have previously denoted as chromatin-rich, nuclear territories [35]. Similar distribution patterns were observed in HeLa cells overexpressing EGFP-AR and Cherry-Pax6 or Cherry-SPBP (data not shown). In line with this, Pax6 and SPBP were found to colocalize almost completely in the nucleus of all transfected HeLa cells ( Figure 2C). Lack of available Pax6 and SPBP antibodies that work well in immunostaining, made it difficult to study the distribution pattern of the endogenous proteins. However, colocalization was assayed on cells that displayed moderate levels of the exogenous proteins. Furthermore, calculation of the Pearson's correlation coefficient of several cells resulted in an average correlation coefficient of 0.7 for Pax6 and AR, SPBP and AR, and Pax6 and SPBP ( Figure 2D). Taken together, these results strongly suggest that AR, Pax6 and SPBP are enriched in the same chromatin-rich territories of R1881 stimulated human cells. To investigate whether Pax6, SPBP and AR are coexpressed in cells, whole cell extracts from nine different cell lines were analyzed by immunoblotting using antibodies against Pax6, SPBP and AR. The blots in Figure 2E show that SPBP is expressed in all of the cell lines tested, while Pax6 is expressed in HeLa, HEK293, B3 and PC3 cells. Expression of Pax6 in the B3 cell line was not unexpected, given that it is a human lens epithelial cell line. We also detected Pax6 at very low levels in Du145 and the human neuroblastoma cell line SKNBe(2), while no expression was detected in the mouse embryonic fibroblasts (MEFs) or the human prostate cancer cell line LNCaP. In contrast, high expression of AR was, as expected, detected in the latter cell line. Human Du145 and PC3 prostate cancer cells are frequently used as AR negative control cells, but we detected some expression of AR in these cell lines as well. AR was also expressed in HeLa, HEK293, Kelly and B3 cells at very low levels, but was not detected in the MEF or SKNBe(2) cell lines. In summary, Pax6, SPBP and AR are coexpressed in the B3 lens epithelial and PC3 prostate cancer cell lines. Interestingly, SPBP is highly expressed in the AR positive prostate cancer cell line LNCaP. SPBP is also expressed in glandular cells of the prostate and in 30% of prostate cancer samples in the Human Protein Atlas (www.proteinatlas.org). This correlates well with the expression pattern of AR and several AR target genes such as PSA, Elk4 (Sap-1a) and the AR-coactivator LSD1 (www.proteinatlas.org, [36]).

FRET analysis reveals a direct interaction between AR and Pax6 in cells
Recently, an association between Pax6 and AR in COS-1 cells has been reported [29]. To evaluate if this interaction is direct in human cells, FRET experiments were performed using the acceptor photobleaching method as previously described [20,33]. HeLa cells were cotransfected with expression constructs for Pax6 and AR tagged with yellow (YFP) and cyan (CFP) fluorescent proteins, respectively. With the CFP and YFP fluorescent proteins, FRET only occurs within a distance of 10 nm [37]. The CFP-AR and Pax6-YFP fusion proteins are schematically illustrated in Figure 3A before and after photobleaching. As shown in Figure 3B, HeLa cells transiently expressing a 1:1 ratio of CFP-AR (pDestECFP-AR) and YFP-Pax6 (pDestPax6-EYFP) were subjected to FRET as described in the methods section. A schematic illustration of CFP-AR and YFP-Pax6 before and after acceptor photobleaching is presented in A. The cell images in B are visualized in pseudo-colors before and after acceptor photobleaching. The YFP acceptor was bleached, and FRET detected as decreased YFP-emission at 532 nm and a corresponding increased CFPemission at 479 nm in the bleached area. The graphical display in B shows the emission spectrum of CFP-AR together with YFP-Pax6 at 479 and 532 nm, respectively, before (blue) and after (red) acceptor photobleaching. (C) Percent FRET between CFP-AR and Pax6-YFP compared with FRET between CFP-YFP, CFP-Pax6-YFP, and CFP-Pax6 -Pax6-YFP. The FRET efficiency was calculated as described in the methods section. Each bar represents the mean of 3-5 experiments. (D) Coimmunoprecipitation of GFP-AR and 36Flag-Pax6. HeLa FlpIn 36Flag-Pax6 cells were induced to express 36Flag-Pax6 and subsequently transfected with pDestEGFP-AR or pEGFP-C1 using Metafectene Pro (Biontex). A GFP-antibody (Abcam) was used to immunoprecipitate GFP-AR-36Flag-Pax6 complexes from the cells. The upper right gel shows the 0.5% input of 36Flag-Pax6 and the lower right gel the 0.5% input of GFP and GFP-AR. The upper left gel shows that 36Flag-Pax6 was immunoprecipitated together with GFP-AR, but not with the GFP control. doi:10.1371/journal.pone.0024659.g003 strong FRET was detected between Pax6 and AR in the nuclei of HeLa cells. Figure 3C shows the calculated percent FRET efficiencies. A FRET efficiency of 2163.7% was detected between the CFP-Pax6 and Pax6-YFP fusion proteins, while the FRET efficiency between Pax6-YFP and CFP-AR was 14.861.5%. A CFP-YFP fusion protein with a linker of 52 non-specific amino acids separating CFP and YFP was used as a positive control, yielding a FRET efficiency of 27.560.4%. Another positive control, CFP-Pax6-YFP, gave a FRET efficiency of 23.362.9%. CFP versus YFP, CFP-AR versus YFP, and CFP versus Pax6-YFP were used as negative controls and displayed little or no FRET (data not shown). The results from the FRET experiments clearly show that Pax6 and AR bind directly to each other in the nucleus of living human cells. This was further confirmed by coimmunoprecipitation of EGFP-AR and 36Flag-Pax6 from HeLa FlpIn cells ( Figure 3D). Expression of Flag-tagged Pax6 was induced by adding tetracycline to the medium, before the cells were transiently transfected with plasmids expressing EGFP-AR. AR was stimulated by adding synthetic androgen to the medium, and a GFP-antibody was used to immunoprecipitate GFP-AR-36Flag-Pax6 complexes from the cells. The upper left panel in Figure 3D shows that 36Flag-Pax6 was immunoprecipitated with GFP-AR, but not with the GFP alone. This further confirms the AR-Pax6 interaction established with the FRET assays.
The PD of Pax6 binds to the DBD of AR Next, pulldown assays were performed in order to map the interacting regions of AR and Pax6. The various domains of both proteins, as indicated in Figure 4A and C, were expressed as GST or MBP fusion proteins in E. coli, while the full-length proteins were in vitro translated in the presence of 35 S-methionine. The GST pulldown assay in Figure 4B shows that AR binds to both the isolated PD and the HD of Pax6, but with highest affinity to the PD. Adding synthetic androgen (R1881) did not affect the interactions significantly. Neither did addition of benzonase ( Figure S2). Hence, the interactions are not dependent on DNA-binding.
To map the region of AR responsible for mediating the reaction, three various AR deletions were constructed; the 500 amino acids long NTD (N-term), the 180 amino acids long central part including the DBD and the hinge region (Central), and the 239 amino acids long C-terminal LBD (C-term) ( Figure 4C). The GST pulldown assay in Figure 4D shows that full-length Pax6 interacts strongly with the central part of AR. Again, adding synthetic androgen did not affect the binding. The interaction between AR(central) and Pax6 is further strengthened by pulldown assays showing that also other Pax proteins interact with this part of AR ( Figure S3).
The central region of AR was further split in the 50 amino acids N-terminal to the DBD (N-DBD), the 90 amino acids long DBD, and the 52 amino acids hinge region ( Figure 4C). MBP fusion proteins of the three constructs were used to pull down in vitro translated Pax6. Compared to the flanking regions, full-length Pax6 interacted strongly with the DBD of AR ( Figure 4E, upper panels). Deleting the PD of Pax6 (Pax6DPD) reduced this interaction significantly ( Figure 4E, middle panels), while deleting the HD (Pax6DHD) did not affect the interaction ( Figure 4E, lower panels). In conclusion, the interaction between AR and Pax6 is mainly between the DBD of AR and the PD of Pax6, and is neither dependent on androgens nor DNA.

SPBP interacts with the DBD of AR
Since SPBP is reported to interact with ERa and here is found to function as a coactivator of AR, we asked whether SPBP was able to associate with AR in cells. HeLa cells were transiently cotransfected with EGFP-AR and HA-SPBP expressing plasmids. An anti-GFP antibody was used to precipitate EGFP-tagged AR from cell lysates, and immunoblotting performed using an anti-SPBP antibody [10]. The upper panel in Figure 5A shows that SPBP coprecipitates with EGFP-AR in HeLa cells stimulated with R1881, but not in unstimulated HeLa cells. The lower panel in Figure 5A confirms precipitation of EGFP-AR in extracts both from stimulated and unstimulated HeLa cells. This strongly suggests that SPBP interacts with AR in the nucleus of human cells.
The interaction between the full length proteins were confirmed by coimmunoprecipitation of in vitro translated proteins ( Figure 5B), followed by mapping of the interacting regions using pulldown assays. First, various regions of AR (schematically illustrated in Figure 4C) were expressed and purified from E. coli as GST or MBP fusion proteins while full-length SPBP were produced in vitro in the presence of 35 S-methionine. The pulldown assays in Figure 5B and 5C show that SPBP, like Pax6, interacts with the central region, and more specifically the DBD, of AR. A weak interaction between SPBP and the C-terminal part of AR was also detected. The interactions are not affected by the presence or absence of synthetic androgen.
SPBP was then split in parts, as indicated in Figure 5D, in vitro translated and used in pulldown assays together with the central part of AR fused to GST. The results of these pulldowns show that amino acids 1333-1666 of SPBP exclusively interacted with AR(Central) ( Figure 5E). Interestingly, the 1333-1666 region of SPBP contains its DBD and also the region (1576-1601) that mediates the interaction with ERa [11]. In addition, this region encompasses three highly conserved motifs (D-box, E-box, and AT-hook). To determine whether DNA binding was important for the SPBP-AR interaction to take place, benzonase was included in some of the GST pulldown assays. However, adding this endonuclease did not have any impact on the binding ( Figure  S2). Hence, here we have mapped a DNA-independent interaction between AR and SPBP, mediated by the DBD of AR and the 1333-1666 region of SPBP containing its DBD.

Pax6 and SPBP compete for binding to AR(DBD)
The findings that coexpression of Pax6 repressed SPBPmediated transactivation of AR activity, inhibited association of SPBP with an AR target promoter, and that both SPBP and Pax6 interacted with the same domain of AR, prompted us to investigate whether there could be a competition between Pax6 and SPBP for AR(DBD) binding. In an attempt to study this, MBP pulldown assays were performed with increasing amounts of SPBP(1333-1960) together with constant amounts of full length Pax6 and MBP-AR(DBD) immobilized on amylose resin. The 1333-1960 region of SPBP binds equally well to AR(DBD) as the full length protein (data not shown), and was therefore used in this assay to reduce the migration differences between Pax6 and SPBP. As shown in Figure 6, Pax6 and SPBP may in fact compete for binding to AR(DBD). Increasing amounts of SPBP reduce the amount of Pax6 coprecipitated with MBP-AR(DBD) while the amount of SPBP binding to AR(DBD) is increased. On the other hand, presence of Pax6 reduces the amount of SPBP bound to AR(DBD) (Figure 6B, right). Taken together, these results suggest that there is a competition between Pax6 and SPBP for binding to AR, and that this competition is a possible mechanism for the repressive effect Pax6 has on SPBP-mediated transactivation of AR activity.

Discussion
AR is critically dependent on recruitment of coactivators and corepressors for transcriptional regulation of its target genes. The differential interaction of AR with various cofactors appears to control the fine balance between proliferation and differentiation important for maintenance of the normal prostate. Changes in cofactor levels may shift the balance between suppressing or facilitating cancer progression (reviewed in [38][39][40]). In this study, we have identified the transcriptional coregulator SPBP as a coactivator of AR, stimulating AR-mediated transcription of the probasin promoter and enhancing the expression of the AR target gene PSA in androgen stimulated LNCaP cells. SPBP is expressed in most cell lines, but highly expressed in the androgen dependent cancer cell line LNCaP. In normal prostate tissue, both SPBP and AR are found in glandular cells (www.proteinatlas.org), and we show that overexpressed AR and SPBP have very similar nuclear distribution patterns in cells stimulated with androgen. Interestingly, 30% of the prostate cancer samples in the Human Protein Atlas (www.proteinatlas.org) display a significant enrichment of nuclear SPBP. Since coregulators of nuclear hormone receptors often have the ability to influence the activity of multiple receptors [38], the AR-SPBP association is strengthened by the previously reported interaction between SPBP and phosphorylated ERa [11].
The developmental transcription factor Pax6 was recently reported to act as a repressor of androgen stimulated AR activity [29]. Here we show that Pax6 in addition inhibits SPBP-mediated stimulation of AR activity and the association of SPBP with an AR target promoter. Both Pax6 and SPBP were found to associate with AR in androgen stimulated cells, and their distribution patterns in the cell nucleus were completely overlapping. Mapping of interacting regions revealed that all three proteins use their DBDs to mediate the interactions. Pax6 contains two DBDs, the PD and the HD. Although both domains are able to interact with AR, the PD seems to bind more strongly. The interaction between Pax6 and AR(DBD) correlates well with the fact that other Pax proteins also interact with the central region of AR. It also correlates with a recent report showing that the HD protein HoxB13 represses AR activity by interacting with the AR(DBD) [41]. The interacting region was in this case mapped using AR(DBD) point mutations in a mammalian two-hybrid assay.
Interestingly, our in vitro competition assays indicated that SPBP and Pax6 compete for AR(DBD) binding. Several AR coactivators are reported to interact with the AR(DBD) (reviewed in [38,39]). Our results suggest that the inhibitory effect of Pax6 on ARmediated transcriptional activity is obtained by masking coactivator binding sites in or displacing coactivators binding to the DBD of AR. This is in line with recent reports on the homeoproteins HOXB13 [41] and HOXC8 [42]. Both proteins act as repressors of AR-mediated transcriptional activity. HOXB13 significantly reduced the recruitment of AR coactivators to specific AR target genes, while HOXC8 was shown to block AR-dependent recruitment of the coactivators SRC-3 and CBP. They also found that overexpression of SRC-3 reverses the HOXC8-mediated blockage. This is similar to our results, demonstrating that increased SPBP expression partly restores the Pax6-mediated inhibition.
Understanding the detailed mechanisms and mapping the specific amino acids important for these interactions and competitions may have a potential in the future development of targeted therapies against prostate cancer. AR plays an important role in both early and advanced stages of prostate cancer etiology. Formation of active AR-directed transcription complexes occurs via a sequential recruitment of coactivators [38][39][40] and alteration in expression and function of an expanding number of cofactors are suggested to underlie pathological conditions. Hence, targeting coregulators for new therapeutics represent a new opportunity (reviewed in [38]). Importantly, cofactors have been suggested to play important roles in the development of androgen insensitive prostate cancer [43], by enhancing the ability of AR to maintain sufficient function in a low androgen environment for maintenance of cell growth and survival. A putative strategy to inhibit or diminish recruitment of coactivators to the receptor is to mimic the AR-Pax6 interaction in cancer cells. One possibility may be to use peptides representing the Pax6 interaction surface. A promising study using a peptide that mimics the function of the short splice variant of the coregulator MTA1 is shown to sequester ERa in the cytoplam [44]. The MTA1s peptide had the ability to inhibit ERamediated transactivation, estrogen-dependent proliferation, anchorage-independent growth and in vivo tumor progression [45]. This indicates that development of peptides abrogating receptorcoactivator interactions may have therapeutical opportunities.
We found that Pax6 and SPBP colocalized with AR in the nucleus of androgen stimulated cells overexpressing the proteins. But since the expression of Pax6 and AR in various tissues is very restricted, few tissues coexpress them both. AR is highly expressed in the epithelium of the tear producing lachrymal glands surrounding the eyes [1], where also Pax6 is found. The expression patterns of AR and PAX6 during development of the pituitary have a transient overlap along the midline border [46,47]. Since the pituitary is the superior hormone-producing gland of the body, a possible common function for Pax6 and AR in this organ is to coparticipate in expression of hormones. Likewise, AR is coexpressed with Pax6 in distinct pyramidal neurons of the hippocampus [48,49], in neurons and astrocytes of the developing and adult forebrain [50,51], as well as in the dopamine-producing substantia nigra of the midbrain [52,53]. Furthermore, our data show that Pax6 and AR are coexpressed in the androgen insensitive prostate cancer cell line PC3, the lens epithelial cell line B3, in addition to the cancer cell lines HEK293 and HeLa. Hence, the Pax6-mediated repression of AR activity may be important and relevant in certain tissues, and also in specific cancer cells aberrantly expressing both proteins.
In conclusion, SPBP and Pax6 modulate AR activity by using their DBDs to interact with the DBD of AR. Elucidating the specific mechanisms involved in the competition between Pax6 and SPBP for binding to AR, may result in important knowledge in the field of targeting AR coregulators for new cancer therapeutics.