Distinct Transcriptional Programs Mediated by the Ligand-Dependent Full-Length Androgen Receptor and Its Splice Variants in Castration-Resistant Prostate Cancer

Androgen deprivation therapy (ADT) for advanced prostate cancer is designed to disrupt the androgen receptor (AR) pathway (1). The intended therapeutic target is the full-length androgen receptor (AR-FL), complete with an intact ligand-binding domain (LBD). Prostate tumors that progress despite first-line ADT (e.g., LHRH analogues), generally termed castration-resistant prostate cancer (CRPC), frequently show continued androgen receptor signaling driven by intratumoral androgens as well as elevated levels of AR-FL (2, 3). In support of the importance of ligand-driven AR-FL signaling in CRPCs, a number of clinically effective endocrine therapies targeting AR-LBD were recently developed to treat patients with CRPCs (e.g., abiraterone, MDV3100; refs. 4, 5). Nevertheless, the majority of patients progress shortly after treatment, again with reactivated androgen receptor signaling (4, 5).

Androgen receptor splice variants that lack the functional LBD (AR-Vs) were recently decoded and characterized (6–11), with some (e.g., AR-V7, ARV567ES; refs. 7, 9, 10) but not all (8) facilitating ligand-independent androgen receptor signaling in CRPCs. AR-Vs originate from androgen receptor transcripts with insertions of cryptic exons downstream of the coding sequences for androgen receptor DNA-binding domain (DBD; refs. 6, 9–11), or with deletions of exons coding for AR-LBD (7, 8). These alterations in androgen receptor transcripts disrupt the androgen receptor open reading frame, leading to truncated androgen receptor proteins with the intact N-terminal domain (NTD), AR-DBD, and a short variant-specific peptide replacing the functional AR-LBD. AR-Vs are capable of activating canonical AR-FL–regulated genes (e.g., KLK3, TMPRSS2, NKX3.1) in the absence of AR-FL signaling (6–11), raising the possibility that AR-Vs and AR-FL may direct a similar transcriptional program and that elevated AR-Vs in CRPC may compensate for AR-FL signaling. However, previous studies also suggest that AR-Vs are not as potent as AR-FL in inducing the expression of AR-FL genes (9). Because AR-FL and AR-Vs are both overexpressed in CRPCs and AR-Vs are less abundant than AR-FL (6, 9), the role of ligand-independent androgen receptor signaling in the context of suppressed, yet still active, AR-FL signaling is not clear. In addition, regulation of expression of endogenous AR-Vs and their functional relationships with AR-FL signaling have not been discerned in the context of ADTs. In this study, we used a set of detection and targeting tools that differentiate the AR-FL and AR-Vs to investigate the functional interplay and distinctions between AR-FL and AR-V. The cumulative in vitro and in vivo evidence supports the importance of AR-V–mediated signaling in mediating responses to CRPC therapies targeting AR-LBD.

LNCaP, CWR22Rv1, and VCaP human prostate cancer cell lines were obtained from the American Type Culture Collection. Early-passage cells were used in all experiments. LNCaP95 is an androgen-independent cell line derived from the parental LNCaP cells (provided by Dr. Alan K. Meeker, Johns Hopkins University, Baltimore, MD). LNCaP95 cells and the parental LNCaP cells show qualitatively similar androgen receptor transcriptional responses (Supplementary Fig. S1). Previous studies have evaluated and established the expression status for both AR-FL and AR-V in these 4 cell lines (9). Stable clones of AR-FL–positive and -negative LNCaP cells were developed by E. Bluemn (Supplementary Methods). In all comparisons involving the treatment of cells with or without R1881 (NEN), cells were cultured for indicated time periods in culture media without phenol red (Invitrogen) supplemented with 10% charcoal-stripped FBS (CSS; Invitrogen) with or without 1 nmol/L R1881 (NEN). For siRNA treatment, cells reaching approximately 70% confluence were treated with siRNA for 24 hours and then cultured for 48 hours with or without 1 nmol/L R1881. Target sequences for specific knockdown of different androgen receptor molecules were described previously (9). MDV3100 was obtained from Medivation. VCaP and LNCaP95 cells were treated with MDV3100 dissolved in dimethyl sulfoxide (DMSO) at the indicated concentrations for 24 hours with or without 1 nmol/L R1881 (NEN). Transient transfection with AR-Vs was conducted as described previously (7–9).

Antibodies used in this study include anti-AR N20 (Santa Cruz Biotechnology); anti-AR-V7 and anti-PSA (Santa Cruz Biotechnology); anti-ERG (C-17, Santa Cruz Biotechnology); anti-UBE2C (Boston Biochem); and anti-β-actin (Sigma-Aldrich). The anti-AR-V7 antibody is a mouse monoclonal antibody developed using peptide sequences specific to AR-V7 (CKHLKMTRP; Supplementary Fig. S2).

Cells were grown on chamber slides to approximately 80% confluence in culture media supplemented with 10% FBS. At 24 hours after the indicated treatments, cells were fixed for 10 minutes using freshly made 4% paraformaldehyde and then with 0.2% Triton X-100 in ×1 PBS for 5 minutes at room temperature. Fixed cells were incubated with the primary anti-AR-V7 antibody for 2 hours at room temperature. Secondary antibody was an Alexa Fluor 488–conjugated goat anti-mouse antibody (1:200 dilution; Invitrogen).

Expression data were generated using the Agilent Whole Genome Expression Arrays (Agilent Technologies) at Johns Hopkins University (by J. Luo) and the University of Washington (Seattle, WA; by P.S. Nelson), analyzed using GeneSpring GX 11.5 (Agilent Technologies), and subjected to Gene Set Enrichment Analysis (12). Other details are described in Supplementary Methods. Raw expression microarray data have been submitted to GEO (GSE36549).

Abiraterone-treated xenograft tumor samples derived from the LuCaP35CR xenograft line was obtained from E.A. Mostaghel. Detailed methods for treatment, tumor collection, and mRNA analysis were fully described previously (13).

CRPC tissues on the tissue microarray (TMA) were procured as previously described (14). Immunohistochemical staining for AR-V7, AR-FL, and UBE2C was optimized and conducted by J. Edwards. Deparaffinized TMA slides were placed in 10 mmol/L Tris, 1 mmol/L EDTA (pH 9.0) and steamed for 40 minutes for antigen retrieval. The primary antibodies were incubated with each slide for 1 hour at room temperature. The EnVision reagent (DAKO Corp.) was used for color development. Protein expression levels were scored by a semi-automated method as previously described (15). Scoring data were analyzed using SPSS (SPSS Inc.) and GraphPad (GraphPad Software, Inc.).

AR-FL and AR-V are both overexpressed in clinical CRPC specimens (9) and induced in castrate conditions in CRPC xenografts (6), suggesting that increased AR-V levels may be coupled with enhanced transcription of the androgen receptor gene and that AR-V function may require the presence of AR-FL (6). Because individual AR-V levels are typically lower than AR-FL (6, 8, 9), we used 2 cell lines, LNCaP95 and VCaP, that recapitulate the relative expression levels of AR-FL and AR-Vs in clinical CRPC specimens, that is, detectable but lower levels of AR-Vs than AR-FL (9). AR-FL signaling in the presence of R1881 was suppressed (Fig. 1A) by siRNA targeting the AR-LBD (AR-LBD siRNA), by ligand depletion (R1881–), or by MDV3100, a potent anti-androgen that targets the AR-LBD (4). We show that suppression of ligand-mediated AR-FL signaling by all 3 of these AR-LBD targeting strategies leads to an increase of the aggregate AR-V signal (AR-Vs; Fig. 1A). Increased AR-V protein expression following suppression of AR-LBD was confirmed by an increase in AR-V7, as shown by Western blot analysis (Fig. 1A) and immunofluorescent staining (Fig. 1B) using a variant-specific monoclonal antibody (Supplementary Fig. S2). We conclude that AR-V protein expression may not parallel expression of AR-FL following suppression of AR-FL signaling (Fig. 1A). Instead, AR-V protein levels are negatively regulated by ligand-mediated AR-FL signaling irrespective of the AR-FL protein levels, suggesting an adaptive shift toward AR-V–mediated signaling after therapy targeting AR-LBD.

To dissect the transcriptional programs induced by AR-V–mediated signaling, we took 2 approaches: first, we examined transcriptional changes driven by forced expression of AR-Vs in the presence or absence of AR-FL signaling by Gene Set Enrichment Analysis (12). Transient expression of exogenous AR-V7 in parental LNCaP cells (Supplementary Fig. S3) induced expression of cell-cycle genes, under both androgen-depleted and -stimulated conditions (Supplementary Fig. S4). On the other hand, top gene sets increased by ligand-dependent AR-FL are dominated by those related to biosynthesis, metabolism, and secretion (Supplementary Fig. S5). For illustration, we generated 2 gene sets, AR-V7 UP and AR-FL UP, each containing 25 probes that contributed to the core enrichment of top ranked gene sets driven by AR-V7 and AR-FL. The 2 representative gene sets show clearly independent expression patterns (Fig. 2A). Other canonical AR-FL–regulated genes (KLK3, TMPRSS2, NKX3.1) follow the same pattern as those in the AR-FL UP gene set (not shown). To further determine whether expression of the AR-V genes requires the presence of endogenous AR-FL, we generated stable LNCaP clones with or without endogenous AR-FL protein (Supplementary Methods), and transiently transfected these clones with AR-V7 or ARV567ES in androgen-deprived conditions (Supplementary Figs. S6 and S7). In this independent series of expression profiles (Fig. 2B), the AR-V7 UP gene set remains as the top-ranked gene set induced by ARV567ES or AR-V7 (not shown), and the absence of endogenous AR-FL did not attenuate induction of the AR-V7 UP signature (Fig. 2B; Supplementary Fig. S8). Thus, the presence of AR-FL is not required for induction of cell-cycle genes by constitutively active AR-V7 and AR-V567ES.

As a second approach to further corroborate the functional distinctions between AR-FL and AR-Vs, we analyzed gene expression correlates of endogenous AR-Vs following suppression of canonical AR-FL signaling only or suppression of both AR-FL and AR-Vs (Fig. 2C; Supplementary Figs. S9–S12). In LNCaP95 cells, the AR-V7 UP signature is again the top gene set enriched for upregulation following an increase of endogenous AR-V7 induced by AR-FL suppression (Supplementary Fig. S11). Knockdown of both AR-FL and AR-Vs abrogated expression of the AR-V7 UP gene set (Fig. 2C), confirming the essential role of AR-Vs. In contrast, the canonical AR-FL target genes showed a change of direction opposite to those driven by AR-V, with the AR-FL UP gene set as the top-ranked gene set enriched for downregulation following suppression of AR-FL (Supplementary Fig. S11). Increased expression of the AR-V7 UP gene set was further confirmed in VCaP cells following suppression of AR-FL signaling by either AR-LBD siRNA or MDV3100 (Supplementary Fig. S12).

The AR-V7 UP gene set included UBE2C (Fig. 2). Previous studies established a direct regulation of UBE2C expression by androgen receptor in LNCaP-abl cells under androgen-deprived conditions (16). Our findings (Fig. 2) suggest that UBE2C expression in our CRPC models may be driven by AR-Vs but not by AR-FL. To validate the relationship between UBE2C and AR-V expression following suppression of AR-FL signaling, we first show that in LNCaP95 cells treated with MDV3100, the expression of UBE2C parallels that of AR-Vs but not AR-FL (Fig. 3A). Both AR-FL and constitutively active AR-Vs are elevated in the LuCaP35CR xenografts following treatment with abiraterone (13). We further show that in those abiraterone-treated xenografts, AR-Vs, but not AR-FL, are significantly correlated with UBE2C and other cell-cycle genes at mRNA levels (Fig. 3B and 3C; Supplementary Table SI). From these findings (Figs. 1–3), we reason that an adaptive shift toward AR-V–mediated signaling may contribute to resistance to MDV3100 and abiraterone.

Suppression of AR-FL does not induce AR-V expression in the parental LNCaP and CWR22RV1 cells (Supplementary Fig. S13), in which high levels of AR-V7 may be caused by genomic alterations not frequently seen in clinical specimens (17). Consistent with the role of AR-V, we did not observe increased cell-cycle gene expression after suppression of AR-FL signaling in these 2 cell lines (not shown). Thus, cell-type–specific regulation of endogenous AR-V expression may explain why suppression of AR-FL inhibits tumor growth in some models of CRPC but not the others (6). Our results nevertheless suggest that AR-Vs, rather than AR-FL, may play a key role in supporting castration-resistant growth in at least a subset of CRPC. We investigated the relative importance of AR-FL and AR-V in vivo by immunohistochemistry analyzing the correlation between AR-FL and UBE2C, and between AR-V7 and UBE2C, in transurethral resection of the prostate (TURP) specimens derived from patients who developed obstructive urinary symptoms after hormone therapy (Fig. 4A; Supplementary Fig. S14 and Supplementary Table SII). AR-V7, but not AR-FL, is significantly correlated with UBE2C (Fig. 4B and C), further supporting that AR-Vs, rather than AR-FL, mediate cell-cycle gene expression in at least a subset of CRPC specimens. This protein expression data are consistent with findings from an mRNA-based study showing elevated cell-cycle gene expression that are detected in bone metastasis express high levels of AR-Vs (18).

In summary, this study reveals the functional interplay between AR-FL and AR-Vs when AR-LBD is rendered inactive by ADTs. Importantly, the combined in vitro and in vivo data predict an adaptive shift toward AR-V–mediated signaling with effective CRPC therapies targeting AR-LBD. These studies indicate that therapeutic efficacy of agents targeting AR-LBD may be compromised as a consequence of this adaptive shift and that early detection of this shift may be used to guide treatment decisions. In addition, novel agents for CRPCs (19) may be designed to suppress the activation of transcriptional programs directed by the AR-Vs. Overall, given the recently expanded therapeutic options for metastatic CRPCs (20), the present findings may stimulate efforts to target adaptive AR-V signaling for treatment selection and to overcome resistance to CRPC therapy.

No potential conflicts of interest were disclosed.

Conception and design: R. Hu, C. Lu, P.S. Nelson, S.R. Plymate, J. Luo

Development of methodology: R. Hu, C. Lu, S.R. Plymate, J. Luo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Hu, C. Lu, E.A. Mostaghel, J. Edwards, P.S. Nelson, E. Bluemn, S.R. Plymate, J. Luo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Lu, E.A. Mostaghel, S. Yegnasubramanian, M. Gurel, J. Edwards, W.B. Isaacs, P.S. Nelson, S.R. Plymate, J. Luo

Writing, review, and/or revision of the manuscript: R. Hu, C. Lu, E.A. Mostaghel, E. Bluemn, S.R. Plymate, J. Luo

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Tannahill, S.R. Plymate, J. Luo

Study supervision: J. Luo

The authors thank John Kim, Shuanzeng Wei, Ilsa Coleman, and Shihua Sun for computational and technical support.

The study was supported by NIH/NCI Specialized Program in Research Excellence (SPORE) in Prostate Cancer grant P50CA58286 (Johns Hopkins University), the Patrick C. Walsh and David H. Koch Prostate Cancer Research Foundation (J. Luo), the Prostate Cancer Foundation (J. Luo), the Pacific NW Prostate Cancer SPORE P50 CA97186 (P.S. Nelson), PO1 CA085859 (S.R. Plymate, P.S. Nelson), and the Veterans Affairs Research Program (S.R. Plymate).