Constitutively Active Androgen Receptor Variants Upregulate Expression of Mesenchymal Markers in Prostate Cancer Cells

Androgen receptor (AR) signaling pathway remains the foremost target of novel therapeutics for castration-resistant prostate cancer (CRPC). However, the expression of constitutively active AR variants lacking the carboxy-terminal region in CRPC may lead to therapy inefficacy. These AR variants are supposed to support PCa cell growth in an androgen-depleted environment, but their mode of action still remains unresolved. Moreover, recent studies indicate that constitutively active AR variants are expressed in primary prostate tumors and may contribute to tumor progression. The aim of this study was to investigate the impact of constitutively active AR variants on the expression of tumor progression markers. N-cadherin expression was analyzed in LNCaP cells overexpressing the wild type AR or a constitutively active AR variant by qRT-PCR, Western blot and immunofluorescence. We showed here for the first time that N-cadherin expression was increased in the presence of constitutively active AR variants. These results were confirmed in C4-2B cells overexpressing these AR variants. Although N-cadherin expression is often associated with a downregulation of E-cadherin, this phenomenon was not observed in our model. Nevertheless, in addition to the increased expression of N-cadherin, an upregulation of other mesenchymal markers expression such as VIMENTIN, SNAIL and ZEB1 was observed in the presence of constitutively active variants. In conclusion, our findings highlight novel consequences of constitutively active AR variants on the regulation of mesenchymal markers in prostate cancer.


Introduction
Prostate cancer (PCa) is the most common cancer in men over 50 years of age and the second cause of male mortality due to cancer in Europe. Androgens signaling plays a key role in PCa cells proliferation or survival [1], and androgen withdrawal remains the main treatment for local recurrence and androgendependent metastatic PCa. However, the benefit of this therapy is transient and all tumors ultimately recur as castration-resistant PCa (CRPC).
Genetic and splicing events affecting the androgen receptor (AR) gene have been linked to CRPC. Constitutively active AR variants, lacking the carboxy-terminal region that encompasses the ligand binding domain and the activation function 2, might contribute to the progression of PCa into castration resistance. These constitutively active AR variants result from premature stop codons due to nonsense mutations as reported for the ARQ640X [2,3,4,5] or from alternative splicing with the retention of a cryptic exonic sequence as described for AR-V7 [4,6,7,8,9,10].
The role of constitutively active AR variants in CRPC has been shown in many studies [7,8,11,12]. The expression of these truncated AR variants is increased by a 20-fold in CRPC compared with localized PCa [9], and is correlated with the capacity of PCa cells to grow in vitro and in vivo in the absence of androgen [7]. However, the exact molecular mechanisms leading to their activation and their mode of action in PCa and CRPC remain unclear.
Recent studies suggest that constitutively active AR variants could play a role in tumor progression. Indeed, although these constitutively active AR variants are already expressed in primary prostate tumors, their expression is all the more expressed in bone metastasis [8]. Furthermore, their expression is associated with an increase of NFAT (Nuclear factor of activated T-cell) and AP-1 (Activator Protein-1) activity, two transcription factors involved in cell proliferation, migration and survival [13].
N-cadherin, which belongs to cadherin superfamily, is located at adherens junctions in nervous, endothelial or mesenchymal cells and is involved in tumor progression [14,15]. Indeed, N-cadherin expression is increased in most cancers and promotes tumor cells migration, invasion and survival [14]. Increased N-cadherin expression is also associated with epithelial-mesenchymal transition (EMT), a phenomenon characterized by a decrease of epithelial markers such as E-cadherin and an increase of mesenchymal markers such as Vimentin or N-cadherin [16,17,18,19]. These molecular and cellular modifications play an important role in tumor cells dissemination at secondary sites [20,21].
More recently, studies have shown that castration-resistant PCa is associated with an upregulation of N-cadherin expression in cellular models as well as PCa xenografts and clinical samples of CRPC [22,23,24]. Moreover, monoclonal antibodies against Ncadherin have been shown to delay the emergence of castration resistance and to reduce the growth of CRPC xenografts [23]. Taken together, these data show that there is a correlation between N-cadherin expression and resistance to castration. Nevertheless, molecular mechanisms whereby N-cadherin expression is increased in CRPC remain unknown.
The aim of this work was to show a possible link between the presence of constitutively active AR variants and the expression of tumor progression markers. More particularly, we focused on the impact of constitutively active AR variants on the expression of Ncadherin and other mesenchymal markers. In the present study, we have shown that N-CADHERIN as well as VIMENTIN, SNAIL and ZEB1 are upregulated in the presence of constitutively active AR variants in PCa.

Plasmids and transfection
For immunofluorescence experiments, the wild type androgen receptor (AR) (AR-WT) and the constitutively active AR Q640X and AR Q670X [25] variants were linked to EGFP as previously described [2,3]. For gene expression analysis and Western-blot, pE-ARWT, pE-ARQ640X and pE-AR-V7 plasmids were constructed by inserting the corresponding AR cDNA between the NheI and BamHI sites in pEGFP-C3.
For transfections, the JetPEI TM transfection reagent (Polyplus Transfection, Ozyme, France) was used according to the manufacturer's protocol. LNCaP cells were seeded in 10 cm dishes at 1610 6 cells/dish or in 6-wells plate at 2610 5 /well. Three days later, the medium was changed and cells were transfected with 10 mg of the indicated plasmid using 20 ml of JetPEI transfection reagent for 10 cm dishes or with 3 mg of plasmid using 6 ml of JetPEI for 6-wells plate. Medium was changed 48 h after and cells were incubated up to 9 days according to the experiments. The medium was changed every two days and for incubations beyond 4 post-transfection days, cells were incubated in the presence of 400 mg/mL geneticin (Invitrogen, France).

Impact of androgens on N-cadherin expression
LNCaP cells were seeded in 6-wells plate in complete medium and transfected as previously described. Twenty four hours later, medium was changed to phenol red free RPMI-1640 supplemented with 5% dextran-coated charcoal-stripped FCS (DCC-FCS) and with the indicated concentration of dihydrotestosterone (DHT) (Sigma-Aldrich, France) or vehicle (ethanol).
For experiment with MDV3100, transfected LNCaP cells were incubated in RPMI-1640 supplemented with 5% DCC-FCS containing the indicated DHT dose and 100 nM MDV3100 (Enzalutamide, Selleck Chemicals, Euromedex, France) or vehicle (dimethyl sulfoxide, DMSO). To confirm the effects of androgens on N-cadherin expression, 22Rv1 cells were grown in RPMI-1640 with 100 nM or 1 mM MDV3100, or DMSO.

Cell Sorting
LNCaP cells were seeded in 10cm dishes at 1610 6 cells/dish and were transfected with pEGFP-ARWT or pEGFP-ARQ640X. Four days after transfection, cells were trypsinized and sorted thanks to the green fluorescence (EGFP) with a BD FACSAria-II cell sorter (BD Biosciences, Le Pont de Claix, France). Total RNA was extracted from EGFP negative (non-transfected) and EGFP positive (transfected) cells and was used to analyze gene expression by qRT-PCR.

Quantitative real-time PCR
Total cellular RNA was extracted from cell lines using NucleoSpinH RNA II assay (Macherey-Nagel, France) according to the manufacturer's procedure. RNA concentrations and purity were quantified measuring the absorbance at 260 nm and 280 nm (GeneQuant pro, GE Healthcare, France). The reverse transcription was performed from 400 ng or 1 mg RNA using RT Omniscript assay (Qiagen, Courtaboeuf, France). RNA were diluted into 13 mL and denatured at 65uC during 5 minutes. A 7 mL reaction mix containing 16RT template, 0.5 mM of each dNTP, 1 mM oligo dT, 10U RNase inhibitor and 4U Omniscript Reverse Transcriptase was added and the reaction was incubated 1 h at 37uC. The reaction was stopped by heating to 93uC for 5 minutes. N-CADHERIN, E-CADHERIN, VIMENTIN, SNAIL, TWIST1, and ZEB1 mRNA levels were quantified using real-time PCR with LightCycler 480 (Roche Applied Science, Meylan, France). For PCR reactions, 5 mL LightCyclerH 480 SYBR Green I Master (Roche, Molecular Diagnostics, Mannheim, Germany) and 1 mL specific primers (Table 1) (Qiagen, QuantiTect Primers, Courtaboeuf, France) were mixed with 4 mL of 1:5 cDNA dilution. Results were normalized using housekeeping gene b-ACTIN or PBGD (Porphobilinogen deaminase) (Qiagen, QuantiTect Primer). Amplification specificity was verified by analyzing melting curve and by electrophoresis migration. All experiments were realized in triplicate and repeated 3 times. Relative quantification was used to determinate fold change in expression level by the DDCt method. Each value is expressed as the mean DDCt 6 SEM. Results were analyzed with Student t test and p-value ,0.05 was considered significant.

Immunofluorescence Staining
Lab-Tek II chamber slides (2 wells) were coated with LNCaP medium for two hours and 1610 5 LNCaP cells/well were seeded. LNCaP cells were transfected with 2 mg of pEGFP-WT, pEGFP-ARQ640X or pEGFP-ARQ670X 3 days later and incubated for 4 days. LNCaP cells were rinsed in PBS and fixed with 2% paraformaldehyde. Cells were blocked and permeabilized by 0.1% Triton/1% Bovine Serum Albumin (BSA)/PBS for 30 min at room temperature. Cultures were incubated with 2.5 mg/mL anti N-cadherin mouse monoclonal antibody (catalog no. 610920, BD Biosciences, France) or isotypic antibody (Sigma-Aldrich, Saint-Quentin Fallavier, France) at 4uC overnight. After washing in PBS, LNCaP cells were incubated with 2 mg/mL Alexa Fluor 568conjugated goat anti mouse (Invitrogen, Fisher Scientific, France) for 1 h and nuclei were stained with 0.1 mg/mL DAPI solution for 20 min at 30uC. Images were captured with the Leica LAS AF6000 fluorescence microscope using LAS AF software (Leica).

Constitutively active androgen receptor variants upregulate N-cadherin expression in prostate cancer cells
Constitutively active AR variants have been associated with CRPC. Moreover, some studies showed that CRPC is also associated with an upregulation of N-cadherin expression [22,23]. We investigated whether constitutively active AR variants upregulate N-cadherin expression in PCa cells. N-CADHERIN mRNA level was determined by qRT-PCR in LNCaP cells overexpressing the constitutively active AR Q640X or AR-V7, or the AR-WT as control (Figure 1). N-CADHERIN expression remained unchanged in LNCaP cells overexpressing AR-WT compared with controls. Interestingly, N-CADHERIN expression was increased by a 8,000-fold in the presence of ARQ640X and AR-V7 ( Figure 1A-B). These data were confirmed in C4-2B cells ( Figure S1) and at the protein level in LNCaP cells ( Figure 1C). In addition, a time course experiment revealed that N-cadherin protein levels were consistently increased from day-3 after LNCaP cells transfection with ARQ640X ( Figure 1D).
To confirm these data from transient transfection, a cell-sorting analysis was performed after LNCaP transfection to demonstrate that N-cadherin expression was restricted to cells expressing a constitutively active AR. N-CADHERIN expression was analyzed in EGFP negative (non-transfected cells) or EGFP positive (transfected cells) fractions by qRT-PCR. Consistent with above results, N-CADHERIN expression was undetectable in both EGFPnegative and positive fractions following LNCaP transfection with pEGFP-ARWT. However, upon transfection with pEGFP-ARQ640X, N-CADHERIN expression was increased in EGFP positive cells overexpressing the constitutively active AR, but not in the EGFP negative fraction (Figures 2A-B). These results were further confirmed by immunofluorescence analysis showing an Ncadherin labeling exclusively in EGFP positive cells expressing a constitutively active AR variant ( Figure 2C).
Taken together, these data strongly suggest that constitutively active AR variants upregulate N-cadherin expression in PCa cells.

Androgens negatively regulate N-cadherin expression induced by constitutively active androgen receptor variants
A recent study reported that constitutively active AR variants might require a full-length AR (AR-FL) to activate endogenous target genes. To explore the effect of the endogenous AR-FL present in LNCaP cells on the ability of constitutively active AR variants to induce N-cadherin expression, LNCaP cells overexpressing AR-WT or a constitutively androgen variant were incubated in the presence of 100 nM DHT or vehicle, and Ncadherin expression was analyzed by qRT-PCR. In accordance with our previous results, no N-cadherin expression was observed in cells overexpressing AR-WT. Interestingly, a 1.4-fold decrease in N-cadherin expression level was observed when cells overexpressing AR Q640X or AR-V7 were cultured in the presence of 100 nM DHT compared to vehicle ( Figure 3A). In addition, this androgen-mediated N-cadherin repression was dose-dependent ( Figure 3B). These results suggest that constitutively active androgen receptor variants do not require AR-FL to up-regulate N-cadherin expression. However, DHT-activated AR-FL seems to antagonize effects of constitutively active androgen receptor variants on N-cadherin expression ( Figure 3C). To verify this hypothesis, the novel anti-androgen MDV3100 was used to inhibit DHT-activated AR-FL in transfected LNCaP cells. As expected, a further significant increase of N-cadherin expression was observed in LNCaP cells overexpressing AR variants in the presence of 100 nM MDV3100 ( Figure 3D). These results were also confirmed in castration-resistant 22Rv1 cells, known to express both AR-FL and constitutively active AR variants. A 2-fold increase in N-cadherin mRNA level was observed when 22Rv1 cells were cultured for 4 days in the presence 100 nM and 1 mM of the anti-androgen MDV3100 ( Figure S2). All together, these data suggest that DHT-activated AR-FL could compete with constitutively active androgen receptor for regulating N-cadherin expression.

Constitutively active androgen receptor variants are associated with the expression of mesenchymal markers
It is widely known that in tumor cells, the expression of mesenchymal markers is associated with a down-regulation of epithelial markers. We hypothesized that the upregulation of Ncadherin expression observed in the presence of constitutively active AR variants is accompanied by a decreased expression of Ecadherin. To test this hypothesis, we analyzed E-cadherin expression in LNCaP transfected with ARQ640X, AR-V7 or the wild type AR expression plasmid, or the empty plasmid as control. E-CADHERIN mRNA levels in LNCaP cells upon transfection with ARQ640X or AR-V7 expression plasmid did not show any significant difference compared with controls ( Figure 4A). These results were further confirmed by Western blot analysis (data not shown), suggesting that the expression of constitutively active AR variants in PCa is associated with a marked increase in N-cadherin expression, but is not correlated with a down-regulation of E-cadherin.
We also investigated whether constitutively active AR expression in PCa cells is associated with other mesenchymal markers. Expression levels of VIMENTIN and transcription factors TWIST1, ZEB1 and SNAIL were determined by qRT-PCR at day-9 after LNCaP cells transfection with ARQ640X or AR-V7 expression plasmid, the wild type AR plasmid or the empty vector as controls ( Figure 4B-E). VIMENTIN expression was increased by a 2.5 and 1.5-fold in LNCaP cells overexpressing ARQ640X and AR-V7 when compared with controls respectively ( Figure 4B).
Although Twist1 is known to induce N-CADHERIN expression in PCa, no significant difference in the mRNA levels of TWIST1 was observed ( Figure 4C). However, constitutively active AR variants led to a statistically significant increase of SNAIL and ZEB1 mRNA levels ( Figure 4D, E). ZEB1 upregulation was also confirmed at the protein level ( Figure 4F).

Discussion
The AR signaling is very important for proliferation and survival of prostate cancer cells. The AR pathway remains activated during the progression of PCa towards a castration-  In this study, we have shown that N-cadherin is upregulated in LNCaP cells expressing constitutively active AR variants, but not in LNCaP cells overexpressing a full-length AR. These data suggest for the first time that constitutively active AR variants can induce N-cadherin expression. This finding should be connected to recent studies reporting a correlation between CRPC and Ncadherin upregulation [22,23,24]. Consistent with these studies, our data suggest that constitutively active AR variants signaling could be a mechanism leading to N-cadherin expression in CRPC.
These findings again highlight the link between AR signaling pathway and N-cadherin expression. Recent studies suggest that AR negatively regulates N-cadherin expression [22,23]. Indeed, N-cadherin upregulation is associated with a decreased expression of AR in castration-resistant PCa xenografts [23]. Furthermore, the upregulation of N-cadherin observed in the castration-resistant LNCaP-19 cells can be reversed in the presence of androgens [22,26,27]. In accordance with these data, we have shown that androgens were associated with a decreased N-cadherin expression in our model overexpressing a constitutively active androgen receptor variant. These results suggest that AR-FL and constitutively active AR variants could act differently ( Figure 3C). For example, DHT-stimulated AR-FL might recruit co-repressors and, in turn, represses N-CADHERIN expression. Besides, constitutively active AR variants lacking of carboxy-terminal region might behave differently and induce N-CADHERIN expression. Furthermore, AR-FL and constitutively active AR variants could compete with each other for regulating N-cadherin expression. Consistent with this hypothesis, N-CADHERIN gene contains a cluster of androgen response elements (ARE) repeats in intron 1 [28].
However, constitutively active AR variants could also indirectly control N-CADHERIN expression. For example, in prostate cancer, N-CADHERIN expression was associated with a nuclear translocation of Twist1 [29]. Although our data showed no significant difference in TWIST1 mRNA levels, constitutively active AR variants might enhance nuclear translocation of Twist1, which could in turn induce N-CADHERIN expression after binding to the E-box within the first intron of N-CADHERIN. These hypotheses deserve to be studied in further studies to understand how constitutively active AR variants regulate N-CADHERIN expression.
In this study, we have also shown that constitutively active AR variants were associated with an increased expression of mesenchymal markers as VIMENTIN, SNAIL and ZEB1. These results are consistent with a recent study, which showed an increase of mesenchymal markers in tumors from patients treated with androgen deprivation therapy [24]. Taken together, our findings suggest that constitutively active AR variants could be associated with EMT process. However, these markers are not consistently associated with EMT. For example, SNAIL confers resistance to apoptosis to tumor cells exposed to ionizing radiations and genotoxic drugs, and enables breast cells to become tumorinitiating cells [30,31,32,33,34].
The expression of mesenchymal markers reported here in the presence of constitutively active AR variants was not associated with a downregulation of E-cadherin in our model. The inverse correlation between N-cadherin upregulation and E-cadherin downregulation is still debated. Indeed, McKeithen and colleagues, and more recently Tiwari and colleagues report a coexpression of both E-and N-cadherins in tumor cells [35,36]. In these studies E-cadherin protein displays a different subcellular localization. Moreover, the reported N-cadherin upregulation after castration in LNCaP-19 cells is not accompanied by a decrease of E-cadherin expression in in vitro cell culture [22]. Nevertheless, the expected E-cadherin downregulation in this model is only observed in orthotopic tumors after castration, but not in subcutaneous LNCaP-19 tumors, suggesting an important role of the surrounding prostatic environment for E-cadherin downregulation [22]. Besides, an inverse correlation between castration-induced N-cadherin expression and E-cadherin downregulation has been documented in LAPC9 and LuCaP35 subcutaneous xenografts models [23,24]. However, this cadherins switch has not been reported in two studies focusing on human clinical prostate tumors, from patients with or without androgen deprivation therapy [22,24].
Further studies are warranted to understand functional consequences of N-cadherin and other mesenchymal markers upregulation in the presence of constitutively active AR variants. Ncadherin expression is widely associated with tumor progression notably owing to its role in migration and invasion. Indeed, Ncadherin favors the migration of cancer cells via cytoskeleton reorganization and lamellipodia formation [14]. It also promotes the migration of cancer cells establishing homophilic interactions with neighboring tissues such as the stromal tissue or endothelium [37,38]. N-cadherin expression is also associated with survival in prostate cancer cells and melanoma cells. Indeed, N-cadherin expression can activate the phosphatidylinositol 3-kinase (PI3K)/ AKT pathway to inactivate pro-apoptotic proteins and to induce an increase of anti-apoptotic proteins as Bcl-2 [39,40]. Finally, a recent study showed that N-cadherin could mediate angiogenesis by inducing monocyte chemoattractant protein-1 (MCP-1) expression via the PI3K/AKT pathway [41].
There is presently great interest in the mode of action of constitutively active AR variants in CRPC. In this study, we have shown for the first time that constitutively active AR variants induce N-cadherin expression and other mesenchymal markers in PCa. These findings support the hypothesis that these constitutively active AR variants could contribute to systemic dissemination of PCa cells, and reinforce the importance to target these AR variants in PCa.  E-CADHERIN, B). VIMENTIN, C). TWIST1, D). SNAIL and E). ZEB1 expression levels were analyzed by qRT-PCR at day-9 after transfection. For each sample, expression levels were normalized to PBGD or b-ACTIN and reported as relative value to LNCaP parental cell line. Values are presented as the mean of DDCt 6 SEM. NS: Not Significant * P,0.05, **P,0.01 and ***P,0.001. F). Western Blot showing evolution of ZEB1 expression in LNCaP cells overexpressing constitutively active AR variants. Immunoblot from 100 mg of total protein extracts. b-actin was used as loading control. doi:10.1371/journal.pone.0063466.g004

Supporting Information
Figure S1 N-cadherin expression was upregulated in C4-2B cells in the presence of constitutively active AR variants. N-cadherin expression was assessed by qRT-PCR in C4-2B cells overexpressing AR Q640X variant or transfected with empty plasmid (C3) 4 days after transfection. Parental C4-2B cells were used as control. N-CADHERIN expression was normalized to b-ACTIN and calculated using the DDCt method. Results are presented as the mean of DDCt 6 SEM from three independent experiments. NS: Not Significant * P,0.05, **P,0.01 and ***P,0.001. (TIF) Figure S2 DHT activated AR-FL repressed N-cadherin expression induced by constitutively active AR variants. 22Rv1 cells were cultured in complete medium supplemented with 100 nM and 1 mM of MDV3100 or DMSO. N-cadherin expression was analyzed by qRT-PCR four days after and was normalized to PBGD. The fold change was expressed as relative values to parental cell line 22Rv1 under normal condition. NS: Not Significant * P,0.05, **P,0.01 and ***P,0.001. (TIF)