MED1 mediates androgen receptor splice variant induced gene expression in the absence of ligand.

The appearance of constitutively active androgen receptor splice variants (AR-Vs) has been proposed as one of the causes of castration-resistant prostate cancer (CRPC). However, the underlying mechanism of AR-Vs in CRPC transcriptional regulation has not been defined. A distinct transcriptome enriched with cell cycle genes, e.g. UBE2C, has been associated with AR-Vs, which indicates the possibility of an altered transcriptional mechanism when compared to full-length wild-type AR (ARfl). Importantly, a recent study reported the critical role of p-MED1 in enhancing UBE2C expression through a locus looping pattern, which only occurs in CRPC but not in androgen-dependent prostate cancer (ADPC). To investigate the potential correlation between AR-V and MED1, in the present study we performed protein co-immunoprecipitation, chromatin immunoprecipitation, and cell proliferation assays and found that MED1 is necessary for ARv567es induced UBE2C up-regulation and subsequent prostate cancer cell growth. Furthermore, p-MED1 is bound to ARv567es independent of full-length AR; p-MED1 has higher recruitment to UBE2C promoter and enhancer regions in the presence of ARv567es. Our data indicate that p-MED1 serves as a key mediator in ARv567es induced gene expression and suggests a mechanism by which AR-Vs promote the development and progression of CRPC.


INTRODUCTION
Castration-resistant prostate cancer (CRPC) occurs when androgen ablation therapy fails. Patients with CRPC have an average survival time of 16 to 18 months from identification of recurrence [1][2][3]. A variety of mechanisms have been proposed for progression that bypasses current therapies targeting the androgen receptor (AR), including production of intratumoral androgens, increased conversion of adrenal androgen to testosterone, and increased AR expression after hormone deprivation [4][5][6][7]. In addition, various cytokines and growth factors have been shown to activate AR through direct binding or by cross-talk mechanisms [8,9]. Functionally, all of these mechanisms rely on continued activation of the AR through its ligand-binding domain (LBD).
However, the recent identification of androgen receptor splices variants (AR-Vs) provides an alternative explanation for the development of CRPC. AR-Vs have been identified by several independent groups in human prostate cancer cell lines, xenografts, metastases, and circulating tumor cells [10][11][12][13][14][15]. Most characteristically, these variants are devoid of the ligand binding domain (LBD) but retain the capability to engage transcriptional machinery and promote the regulation of AR-target genes. The potential role of AR-Vs in driving prostate cancer progression is supported by several independent correlative clinical studies describing the significant association of AR-Vs with advanced disease progression and a shorter survival period [15][16][17][18]. Among the constitutively active AR-Vs, AR-V7 (or AR3) and AR v567es are the two most commonly described in advanced disease [17,19]. AR-V7 has been reported in many prostate tissues both benign and malignant, while AR v567es has only been seen in malignant prostate glands [10,14,18,19]. Furthermore, the AR3/ V7 and AR v567es transgenic mouse models demonstrated that expression of AR variant in mouse prostate induced high-grade prostatic intraepithelial neoplasia (PIN) [20] and/or invasive prostatic carcinoma [21].
The mechanism of AR-Vs in CRPC transcriptional regulation still remains unclear. Present evidence suggests, in addition to activation of the classical AR target genes, constitutively active AR splice variants are associated with a distinct transcription program in prostate cancer cells as well as in prostate cancer xenografts displaying treatmentinduced AR-Vs expression [22]. Importantly, this distinct expression signature is enriched with cell cycle genes compared to the canonical AR-ligand dependent gene signature. Very interestingly, another study also described a similar transcription program comprised of upregulated cell-cycle genes in the androgen-independent prostate cancer (AIPC) cell line LNCaP-abl [23]. Although the latter research did not involve the role of AR variant, the ubiquitin-conjugating enzyme E2C (UBE2C) was the most up-regulated cell cycle gene in both studies. UBE2C is an anaphase-promoting complex/cyclosome (APC/C) E2 ubiquitin-conjugating enzyme. It can inactivate the M-phase check point and enhance cell growth. UBE2C has been shown to be a prominent oncogene in solid tumors, and it is found overexpressed in various types of solid tumors including late-stage prostate cancer [24][25][26][27]. Taken together, these studies indicate the presence of a distinct gene expression profile in CRPC that is different from the canonical AR-dependent transcriptome, one that might be associated with different transcriptional machinery of AR-Vs. The underlying mechanism on how AR-Vs induce a distinct transcriptional profile remains to be elucidated.
Modulation of androgen receptor (AR) co-regulators might play a pivotal role in CRPC [28,29]. Previous studies have indicated that epigenetic markers and collaborating transcription factors can be ascribed to androgen-independent prostate cancer, including histone markers, [30] FoxA1 [31,32], MED1(Mediator complex subunit 1) [33] and FOXO1. [34] Chen et al showed p-MED1 could drive CRPC cancer growth through a looping pattern on the UBE2C locus [35]. The chromatin looping functional pattern occurred uniquely in CRPC but not in androgen-dependent prostate cancer (ADPC). Given the critical role identified for MED1 in mediating UBE2C expression in CRPC, we asked whether co-activator MED1 could interact with AR-Vs to mediate the downstream events of transcriptional regulation.

MED1 is necessary for AR v567es induced UBE2C regulation and subsequent prostate cancer cell growth
We first confirmed the regulatory effect of AR v567es variant on UBE2C expression in the AR-dependent cell line LNCaP. After performing transient transfection of an AR v567es expression vector 3Flag-CMV-AR v567es , we assayed UBE2C protein and mRNA levels by Western blot and quantitative RT-PCR analysis. As shown in Figure 1A, the expression of UBE2C increased in the LNCaP-AR v567es cells grown in media without DHT, consistent with previous reports [22]. To investigate whether MED1 is involved in this regulatory activity, MED1 expression was silenced by RNAi (Fig. S1). Subsequently, the up-regulation of UBE2C was significantly impaired at both the mRNA and protein level ( Figure 1A). These data suggest MED1 is involved in AR v567es mediated UBE2C expression.
The expression of AR v567es in metastases portends a rapid progression of the tumor [18], which was also suggested by our in vitro study. The AR v567es expressing stable cell line LNCaP-AR v567es (see methods) grows faster than control LNCaP-Lenti cells in the absence, but not in the presence of DHT (Fig. 1B). To investigate whether there is a functional interaction between MED1 and AR v567es , we tested the effect of MED1 silencing on AR v567es induced cell proliferation. Silencing of MED1 decreased proliferation of LNCaP-AR v567es cells compared with scramble control ( Fig. 1C and S2). This effect was more significant in the androgen-deprived condition. This phenomenon indicates MED1 is more involved with AR v567es transcriptional regulation when the ligand is absent, but it is not actively interacting with full-length AR (AR fl ) when androgen is present. QPCR was further performed to investigate the UBE2C expression in the stable cell lines (Fig. S3). Higher UBE2C expression was seen in the LNCaP-AR v567es cell line in the absence of DHT; both DHT and siMED1 significantly inhibited UBE2C expression. Collectively, these data suggest MED1 plays an essential role in the AR v567es induction of UBE2C and subsequent prostate cancer cell growth in an androgen-independent manner.

MED1 is recruited to AR v567es independent of full-length AR
As reported in a previous study we demonstrated that AR v567es binds to full-length AR (AR fl ) and increases the stability of AR fl protein [19]. Here, we investigated whether there is a physical interaction between MED1 and AR v567es , and whether this interaction is mediated by full-length AR. The co-immunoprecipitation assay was performed with the LNCaP cell line transiently transfected with Flag-tagged AR v567es . As shown in Figure 2A, the anti-Flag antibody could pull down p-MED1 as well as AR fl , indicating a physical association of AR v567es with p-MED1 and AR fl in the context of protein activity. Of note, Flagtagged AR fl also pulled down p-MED1. However, with the same amount of input protein lysates (100 ug), AR v567es showed abundance of p-MED1 co-precipitation especially in the absence of DHT, but AR fl pulled down much less p-MED1 protein. This finding indicates AR v567 has more potency to recruit p-MED1 when androgen is depleted, www.impactjournals.com/oncotarget which is exactly in accordance with the impaired cell growth by siMED1 in androgen-depleted condition shown in Figure 1C.
To rule out the possibility that AR fl might mediate the interaction between AR v567es and p-MED1, the AR-null M12 cell model was also used. As shown in Figure 2B, in the cumate inducible M12-AR v567es stable cell line, the reverse pull-down of AR v567es by p-MED1 antibody further demonstrated the physical interaction between these two molecules. The IgG controls showed the specificity of the co-IP binding (Fig. S4A, B). In total, these data indicate AR v567es , p-MED1, and AR fl might form a ternary Up-regulation of UBE2C by AR v567es could be attenuated by MED1 siRNA at mRNA and protein level. (B) AR v567es stable expressing cell line, LNCaP-AR v567es showed significant higher proliferation rate (**p < 0.01) compared with Lenti virus empty vector control cell line in the absence, but not presence of DHT. (C) AR v567es induced LNCaP cell proliferation could be blocked by MED1 silencing (*p < 0.05) when DHT was not present, but not significant with DHT. www.impactjournals.com/oncotarget complex or bind separately, but AR v567es bound to p-MED1 independent of full-length AR.

The binding of phospho-MED1 to UBE2C promoter and enhancer increased when AR v567es was present
Since p-MED1 was recruited by AR v567es , we further tested whether AR v567es could enhance p-MED1 binding to UBE2C transcriptional elements. As described by Chen et al, UBE2C promoter and enhancer regions can form a chromatin loop while triggering transcriptional initiation in CRPC cells [35]. Using ChIP assay, we tested p-MED1 binding capacity to the UBE2C promoter and all identified enhancers in LNCaP-Lenti and LNCaP-AR v567es cells under different conditions. As shown in Figure 3A, when DHT was absent (T+S media), more p-MED1 binding (5-8 fold) occurred at UBE2C transcriptional regions, but not at control regions, in the LNCaP-AR v567es cells compared to LNCaP-Lenti cells. Similar binding occurred when AR fl was inhibited by enzalutamide (MDV3100), a potent AR LBD inhibitor [36]. However, in the presence of DHT, and thus AR fl activation, the increased binding in LNCaP-AR v567es cells diminished to a level seen in the LNCaP-Lenti cells. When both DHT and MDV3100 were present (MDV+DHT), a combination that partially inhibits AR fl activity, the LNCaP-AR v567es cell showed stronger p-MED1 binding than the LNCaP-Lenti cells, but still lower levels of binding than the LNCaP-AR v567es T+S and MDV3100 treated groups (1-2 fold over IgG). These results strongly indicate that AR v567es endows p-MED1 higher binding ability to UBE2C transcriptional regions in a ligand-independent manner. Consistent with the ChIP assay results, the UBE2C mRNA level (Fig. S5) exactly corresponded to the binding capacity of p-MED1 to UBE2C transcriptional elements. This indicates p-MED1 is the key mediator in AR v567es induced UBE2C transcription.   Figure 3A elucidates higher p-MED1 binding induced by AR v567es than AR fl , and Figure 3B more clearly shows the divergent effects of androgen deprivation (T+S), AR fl activation (DHT) and AR fl inhibition (MDV3100) on p-MED1/UBE2C binding in these two cell lines. While androgen deprivation and AR fl inhibition could abate p-MED1 binding to UBE2C in androgen-dependent LNCaP-Lenti cells, they could inversely and vigorously activate the p-MED1 binding to UBE2C in LNCaP-AR v567es cells. This finding could coincidently address the mechanism on how androgen deprivation treatment (ADT) induces growth inhibition of androgen-dependent cancer, and also on how AR v567es contributes to castrationresistance development. AR fl activation (DHT) also has a contrary effect on these two cell lines. While DHT slightly strengthened the recruitment of p-MED1in LNCaP-Lenti cells, it significantly blocked this action in LNCaP-AR v567es cells ( p < 0.01). The negative regulation of DHT on AR v567es activated p-MED1 binding was very consistent with our previous findings: DHT led to decreased cell proliferation (Fig. 1C) as well as lower amounts of p-MED1/ AR v567es precipitation ( Fig. 2A).
In summary, these data lead to the following conclusions: 1) Transactivation of AR fl by ligand confers low p-MED1 binding capacity to UBE2C transcriptional regions; 2) AR v567es is a potent effector driving p-MED1 to the UBE2C locus independent of androgen; 3) When AR fl and AR v567es coexist and DHT is present, AR fl takes priority, and recruits less p-MED1, thus facilitating low levels of binding to UBE2C transcriptional elements; 4) When androgens are depleted, AR v567es takes the stage and recruits more p-MED1 resulting in much higher levels of binding to UBE2C regulatory regions. These findings are consistent with the chromatin-looping hypothesis [35] and could explain why the UBE2C promoter-enhancer loop is more likely to occur in CRPC cells but not in androgendependent prostate cancer cells.
As a control, PSA (KLK3) promoter and enhancer regions were also investigated with p-MED1 ChIP assays in LNCaP-AR v567es cells (Fig. 3C). Very similar to UBE2C regulatory regions, p-MED1 showed high levels of binding to the PSA enhancer element when AR fl was inhibited by MDV3100, and then next highest when DHT was absent (T+S). However, the p-MED1 binding to PSA promoter is not affected in all the treatment groups compared with IgG control, which is consistent with PSA mRNA outcomes (Fig. S5). These findings indicate that AR v567es /p-MED1 might perform alternative transcriptional regulation of AR canonical genes, but not necessarily affect the final gene expression.

Phosphorylated MED1 mediates AR v567es induced UBE2C locus transcription
We further studied the role of MED1 on AR v567es induced transcriptional activity in the UBE2C locus. The 1.2 kb-long enhancer 1 (E1) fragment was used in a luciferase reporter assay (Fig. 4A). E1 is located 20 kb 5' of the UBE2C transcription start site (TSS) and has the highest activity in a 3C assay [35]. Three ~400bp regions in Enhancer 1 termed E1-1, E1-2 and E1-3 were systematically subcloned into the pGL4.10-E4TATA-Luc vector. The reporter activities were measured in M12-Lenti cells and M12-AR v567es cells. The M12 prostate cancer cell line is an AR negative line, so the effect of AR fl could be excluded. While E1-1 showed comparable transcriptional activities in M12-Lenti and M12-AR v567es cells (Fig. S6A, B), E1-2 and E1-3 both induced significantly higher Luc signals when AR v567es was present ( p < 0.01) (Fig. 4B). After the expression of MED1 in M12-AR v567es cells was silenced by siRNA, the Luc expression was inhibited subsequently (Fig. 4C). Similar results were seen in Chen et al's study, with control groups showing higher Luc activities in the LNCaP cell line, but much lower levels in PC3 cells. The reason might be due to very active E4TATA basal promoter activity in M12 and LNCaP cell lines. However, it does not affect these data, which indicated that AR v567es enhanced UBE2C locus transcription and this enhancement was mediated by MED1.

AR v567es /p-MED1 complex interaction with the PI3K/AKT pathway
The PTEN tumor suppressor gene is mutated in 50% of human prostate cancers. In addition, 70% of late stage prostate cancers show alterations in the PTEN/PI3K/ AKT pathway [37]. We investigated whether the PI3K/ AKT pathway is involved in AR v567es /p-MED1/UBE2C transcriptional activity. The PI3K inhibitor LY294002 was used to examine the effect of PI3K/AKT pathway inhibition on MED1 phosphorylation and UBE2C expression in LNCaP-AR v567es and M12-AR v567es cells. As shown in Figure 5(A, B), LY294002 reduced AKT phosphorylation at S473 and MED1 phosphorylation at T1457 leading to reduced UBE2C protein expression. The involvement of MAPK/ERK pathway was also investigated. Inhibition of the MAPK pathway by PD98059 did not have any effect on MED1 phosphorylation and UBE2C expression in M12-AR v567es cells (Fig. 5C). p-ERK protein was nondetectable in LNCaP-AR v567es cell by western blot, and UBE2C expression was not affected after treatment by PD98059 (data not shown). The involvement of PI3K/ AKT, but not the MAPK pathway, in the proliferation of the LNCaP-AR v567es cells was further observed in MTT assays by inhibited cell growth of LY294002 treated cells, but not of PD98059 (Fig. S7).
Another interesting finding we observed in this experiment was the decreased UBE2C protein expression in the LNCaP-AR v567es cells when DHT was added to T+S media (Fig. 5A). This result not only addresses the direct mechanism of impaired LNCaP-AR v567es cell growth under DHT stimulation (shown in Figure 1C), www.impactjournals.com/oncotarget but also consistent with the negative regulatory effects of DHT on p-MED1 co-immunoprecipitation with AR v567es (Figure 2A) and p-MED1 recruitment to the UBE2C locus (Fig. 3A, B).

AR v567es /p-MED1 transcriptional regulation is associated with FoxA1
As reported previously, phosphorylation of MED1 facilitates FoxA1, Pol II and TATA binding protein (TBP) recruitment and mediates their interactions on chromatin, leading to chromatin looping [35]. Here we performed co-IP with p-MED1 and Flag-tagged AR v567es to examine the recruitment of FoxA1, a key component in the loop complex. In cumate inducible LNCaP-AR v567es and M12-AR v567es cells, p-MED1 antibody and Flag antibody were used as the pull-down antibodies followed by western blotting with the immune precipitated protein. As shown in Figure 6A and 6C, FoxA1 is co-precipitated with p-MED1 and Flag in both variant cell lines, which indicates FoxA1 is associated with the transcriptional complex of p-MED1/ AR v567es , and this process is not dependent on AR fl . RNAi against FoxA1 demonstrated decreased expression of UBE2C in both cell lines (Fig. 6B, D), which further indicates the critical role of FoxA1 in AR v567es /p-MED1 transcriptional regulation. IgG controls were included in Figure S8 to show the specificity of the co-IP binding (Fig. S8A, B).

DISCUSSION
The occurrence of ligand-independent, constitutively active androgen receptor splice variants in castrationresistant prostate cancer provides a conceptually simple explanation for the development of resistance to prostate cancer therapies that target the ligand-binding domain. The AR-Vs induce a distinct transcriptome in which overexpression of UBE2C and other cell cycle genes predominate, suggesting there might be different, or unique, transcriptional machinery used by AR variants compared to full-length AR.
MED1, also termed TRAP220, is a 220-kDa subunit of the human thyroid hormone receptorassociated protein (TRAP)-Mediator complex. TRAP Mediator has been described as a co-activator for a broad range of nuclear hormone receptors as well as other classes of transcriptional activators, including glucocorticoid receptor [38], estrogen receptor [39,40] and androgen receptor [41]. Recently, the role of MED1 in prostate cancer oncogenesis and progression has gained recognition [33]. MED1 expression is upregulated in the epithelium of clinically localized human prostate cancer and in CRPC cell lines. Ectopic MED1 overexpression in prostate cancer xenografts significantly promoted tumor growth in nude mice [42]. Notably, MED1 overexpression in CRPC cells leads to upregulation of distinct AR target genes involving cell cycle progression, including UBE2C [23,35]. Knockdown of MED1 resulted in cell-cycle arrest and decreased proliferation, which is also evident in our LNCaP-AR v567es cell line. It has been reported that phospho-MED1 mediates UBE2C locus looping in castration-resistant prostate cancer cells, but not in the androgen-dependent prostate cancer cells [35]. The chromatin looping model nicely elucidates the components in this complex, however it does not specify which transcription factor initiates looping formation given that MED1 interacts with multiple transcription factors as mentioned above. In the castrate-resistant cancer cells, there is low possibility for AR fl still functioning actively with ligand depletion, but another story for constitutive active AR-Vs. Herein, we have identified that AR v567es associates with p-MED1 as a key mediator in CRPC transcriptional activity. We found that: (i) MED1 is necessary for AR v567es induced UBE2C up-regulation and subsequent prostate cancer cell growth; (ii) p-MED1 is recruited to AR v567es independent of full-length AR; (iii) p-MED1 has higher recruitment to UBE2C promoter and enhancer regions in the presence of AR v567es , (iv) AR v567es enhanced UBE2C transcription could be blocked by silencing MED1; (v) AR v567es / p-MED1 signaling crosstalks with the PI3K/AKT pathway but not the MAPK pathway, and (vi) FoxA1 is involved in AR v567es /p-MED1 induced UBE2C long range chromatin interactions.
Previous data demonstrated that an interaction between MED1 and AR worked in a ligand-dependent manner in androgen responsive prostate cells. However, a recent study showed MED1 could functionally interact with androgen receptor in a non-canonical way via a newly discovered binding motif in the AR N-terminal Tau-1 domain. [43] This study is of high interest to us, as the AR v567es splice variant lacks the ligandbinding domain, its interaction with MED1 might be mediated by AR fl . However, based on this study we know that AR v567es , which contains the N-terminus, has the structural base to directly interact with MED1. Consistent with our data, we could see the binding of MED1 and AR v567es through co-IP in AR positive LNCaP cells, but also in the AR-null M12 cells transfected with AR v567es . However, whether AR v567es functionally interacts with MED1 through this particular Tau-1 domain needs to be further investigated.
The most significant finding in this study is the divergent mechanism of AR fl and AR v567es in regulating p-MED1 recruitment, driving p-MED1 to the UBE2C locus, and then affecting UBE2C expression and cancer cell growth. Here we raise a "p-MED1 switch" hypothesis ( Fig. 7) that could reasonably address this: when androgen/ DHT is available (prior to ADT) in androgen dependent prostate cancer (ADPC), more p-MED1 is recruited to AR fl but no chromatin looping forms and AR fl has a low level of activation on UBE2C transcription (Fig. 7A). After ADT therapy, the DHT is depleted and AR fl signaling is inhibited, AR splice variants are formed as a survival mechanism due to the stress on the cells; the variants have higher affinity to p-MED1, and therefore recruit more p-MED1 to the UBE2C promoter and enhancers with the assistance of FoxA1, resulting in chromatin looping. This strongly enhances UBE2C expression, leading to the CRPC stage, which has enhanced cell survival and increased cell proliferation (Fig. 7B).
p-MED1 recruitment in castrate conditions to the UBE2C promoter/enhancer regions was very high in the presence of AR v567es . In the MDV+DHT treated LNCaP-AR v567es cells, however, MDV3100 was not able to fully attenuate the DHT effect (Fig. 3B). The LNCaP-AR v567es cells also express AR fl . MDV3100 has lower affinity to the AR fl LBD and is not able to fully outcompete the ligand. Thus, the inability of MDV to overcome DHT's suppressive effect on p-MED1 recruitment in this cell line suggests that when DHT is present, it will facilitate the "p-MED1 switch" to AR fl signaling leading to low levels of UBE2C transcription.
The recruitment of p-MED1 to AR canonical genes, such as PSA, differs from non-canonical genes, such as UBE2C. PSA ChIP in LNCaP-AR v567es cells demonstrated that p-MED1 is only recruited to PSA enhancer regions, but not to the promoter. In contrast, in the LNCaP-Lenti cells p-MED1 had the strongest recruitment to the PSA promoter in the presence of DHT. PSA transcript levels overall are significantly lower in the LNCaP-AR v567es cells compared with LNCaP-Lenti cells. Perhaps recruitment of p-MED1 to the enhancer regions of PSA is inhibitory whereas recruitment to the promoter region enhances transcription. These data indicate the "p-MED1 switch" hypothesis may not apply to canonical AR genes.
Multiple signaling pathways have been described leading to MED1 phosphorylation in a variety of conditions, including MAPK/ERK signaling [42] [44] [45,46], AMP-activated protein kinase (AMPK) [47] and PI3K/AKT [35,42]. In our study, we investigated ERK and AKT signaling pathways in LNCaP-AR v567es cells and M12-AR v567es cells, and found only the PI3K pathway was involved in phosphorylating MED1 in the context of AR variant. PTEN deletion with AKT pathway activation has long been recognized as one of the most important mechanisms of CRPC [48]. A previous study reported the activity of another AR variant, AR-V7, is regulated in a PTEN-PI3K-AKTdependent way [49]; here we further confirmed the involvement of PI3K in AR-Vs' function.
The concurrence of the presence of constitutively active AR splice variant, increased MED1 and UBE2C, as well as crosstalk with the PI3K-AKT pathway, signifies that CRPC has multiple factors synergistically contributing to the process. Identification of potent inhibitors for AR-Vs and combining agents that target MED1, UBE2C and phospho-PI3K/AKT would provide an effective therapeutic strategy in future clinical trials. www.impactjournals.com/oncotarget

Cell culture, plasmid construction and transient transfection
LNCaP human prostate cancer cell line was obtained from the American Type Culture Collection. Early-passage cells were used in all experiments. These cells were grown in T-medium (Invitrogen, Grand Island NY, USA) supplemented with 10% FBS, 100 units/ml penicillin, and 100 ug/ml streptomycin at 37°C with 5% CO 2 . M12 cells are AR-negative and were obtained from Dr. J. Ware at the Medical College of Virginia. The generation and characterization of the M12 prostate cell line has been described previously [50][51][52][53]. M12 cells were cultured in RPMI 1640 medium (Invitrogen) containing 5% FBS, 10 ng/ ml EGF, 0.02 mM dexamethasone, 5 ug/ml insulin, 5 ug/ml transfection, 5 ng/ml selenium, fungizone, and gentamicin at 37°C with 5% CO 2 . cDNA of the entire AR v567es variant and AR fl were cloned into p3xFlag-CMV-9 vector as described previously [19,53]. The expression constructs were transfected into the human prostate cancer cell lines using TurboFect reagent according to the manufacturer's protocol (Thermo Scientific, Pittsburgh PA, USA).

Proliferation assay
100 ul of cells were seeded at the concentration of 10,000 per well in 96-well plates in serum free media containing transferrin and selenium (T+S). For cells with dihydrogestosterone (DHT) treatment, 10 -9 M was added. Cells were allowed to proliferate for 24, 48, 72, and 96 hrs respectively. 20 ul/well of CellTiter 96 AQueous One Solution reagent was added and incubated at 37°C for two hours prior to reading (Promega, Madison WI, USA). The absorbance was then recorded using iMark Microplate Reader (Bio-Rad) at a wavelength of 490 nm. Each point represents the mean ± SD of 3 replicates.

Western blotting
Cells were washed with PBS and lysed with cold lysis buffer (50 Mm HEPES, 150 mM NaCl, 1.5 mM EGTA, 1% Triton X-100) containing protease inhibitors (Complete Mini Tablets) (Roche, USA) and Phosphatase Inhibitors Cocktail II (Sigma-Aldrich, USA). Protein concentration of cell extracts was determined by the BCA Protein assay (Thermo Scientific). Twenty micrograms of protein was electrophoresed through 4-15% gradient SDS-PAGE and subsequently transferred onto nitrocellulose membranes using Invitrogen iBlot Gel transfer system, and probed with respective antibodies at 4°C overnight. Antibodies used in this study and the working conditions are listed in Table 3. The blots were washed and incubated with a horseradish peroxidase-conjugated secondary antibodies (Pharmacia Biotech, Piscataway, NJ) for 1 hour at room temperature. Immunoreactive proteins were detected by ECL (Pharmacia Biotech). The membranes were stripped for 30 minutes in stripping buffer (Pierce, Rockford, IL) and reprobed with anti-GAPDH antibody as a loading control.

Co-Immunoprecipitation
Cells were lysed in cold lysis buffer using the aforementioned lysis buffer with complete protease Precleared cell lysate was incubated with anti-Flag or anti-p-MED1 antibodies overnight and then with ultralink immobilized protein A/G plus beads (Thermo Scientific) for 24 hours. Beads were washed four times with cold lysis buffer, and then samples were degenerated by boiling. Lastly, immune complexes were applied in Western blotting as previously described.

ChIP assay
ChIP assays were performed using MAGnigy TM Chromatin Immunoprecipitation System (Invitrogen) according to the manufacturer's instructions. Briefly, chromatin was crosslinked for 10 min at room temperature with 1% formaldehyde. After sonication with Diagenode Bioruptor (Denville, NJ), chromatin was sheared into fragments of ~500bp and immunoprecipitated with Dynabeads coupled with anti-p-MED1 antibody or isotope control. After washing, the reversed ChIP DNA was purified and then analyzed by real-time PCR.The primers used are listed in Table 4.

Statistical analyses
All the experiments were performed at least three times. The data are displayed as mean ± SEM. When two groups were compared, 2-tailed Student's t test was used (GraphPad Prism, version5.0d; GraphPad Software, La Jolla, CA). A p value of 0.05 or less was considered significant.