Androgen Receptor Splice Variants Contribute to the Upregulation of DNA Repair in Prostate Cancer

Abstract

Background: Canonical androgen receptor (AR) signaling regulates a network of DNA repair genes in prostate cancer (PCA). Experimental and clinical evidence indicates that androgen deprivation not only suppresses DNA repair activity but is often synthetically lethal in combination with PARP inhibition. The present study aimed to elucidate the impact of AR splice variants (AR-Vs), occurring in advanced or late-stage PCA, on DNA repair machinery.

Methods: Two hundred and seventy-three tissue samples were analyzed, including primary hormone-naïve PCA, primary metastases, hormone-sensitive PCA on androgen deprivation therapy (ADT) and castration refractory PCA (CRPC group). The transcript levels of the target genes were profiled using the nCounter platform. Experimental support for the findings was gained in AR/AR-V7-expressing LNCaP cells subjected to ionizing radiation.

Results: AR-Vs were present in half of hormone-sensitive PCAs on androgen deprivation therapy (ADT) and two-thirds of CRPC samples. The presence of AR-Vs is highly correlated with increased activity in the AR pathway and DNA repair gene expression. In AR-V-expressing CRPC, the DNA repair score increased by 2.5-fold as compared to AR-V-negative samples. Enhanced DNA repair and the deregulation of DNA repair genes by AR-V7 supported the clinical data in a cell line model.

Conclusions: The expression of AR splice variants such as AR-V7 in PCA patients following ADT might be a reason for reduced or absent therapy effects in patients on additional PARP inhibition due to the modulation of DNA repair gene expression. Consequently, AR-Vs should be further studied as predictive biomarkers for therapy response in this setting.

Keywords: AR-V7; BRCA1; BRCA2; DNA repair; androgen deprivation therapy; androgen receptor; castration-refractory prostate cancer; prostate cancer; splice variant.

Figure 1

Flow chart of the main steps in this study: material processing, data acquisition, biostatistical analyses and experimental validation.

Figure 2

(A) Structure of the samples in study cohort and group labels used in further Figures. Among primary prostate cancer samples, 33 ductal adenocarcinoma (DA) samples were analyzed. Abbreviations: AA—acinar adenocarcinoma, ADT—androgen deprivation therapy (primary tumor and metastases), BPH—benign prostatic hyperplasia, CRPC—castration-refractory prostate cancer (primary tumor and metastases), CTRL—normal prostate tissue, GG—grade group, NEC—neuroendocrine carcinoma, PCA—prostate cancer, WHO—World Health Organization. (B) Composition of the gene panel for mRNA expression analysis using nCounter technology. AR—androgen receptor, FL—full-length. Additionally, four housekeeping genes (HPRT1, ALAS1, ARF1, PGK1) were included in the panel.

Figure 3

(A) Analysis of samples expressing any of the four androgen receptor splice variants in the four main study groups. BPH and CTRL groups were negative for AR-Vs. Numbers represent percentages of samples containing AR-Vs. (B–D) Stratification of the samples expressing any of the AR-V splice variants (V1, V3, V7, V9) in study groups. Boxes represent single samples. Stratification according to histological subtype (acinar or ductal (DA) adenocarcinoma) in primary hormone-naïve tumor samples. As other AR-Vs are co-expressed with AR-V7 in CRPCs, the latter appears a reasonable surrogate marker for the presence of AR-Vs. Correlation (co-existence) measure is presented for AR-V1, -V3 and -V9 compared to AR-V7 splice variant. “n” represents the number of samples positive for single AR splice variants. (E) Expression of full-length AR (AR-FL) and two other splice variants (AR45 and ARv567es) in tumor and benign study groups. Statistical significance (p < 0.05): * vs. PRIM group, ^ CRPC vs. ADT group, # vs. CTRL group (Mann–Whitney U-test).

Figure 4

(A) Distribution of AR scores in study groups (cumulative score of AR pathway activation, calculated based on mRNA expression of 10 transcriptional targets of AR). Light blue points express any of the AR-V splice variants (V1, V3, V7, V9); dark blue points do not express AR-V splice variants. “Positive” area represents AR score in “activated” range compared to reference group (CTRL: benign non-hyperplastic prostate tissue). “Negative” area represents depression in AR signaling. (B) mRNA expression of proliferation marker MKI67. (C) Analysis of AR score distribution in four “tumor” study groups in relation to dependence on AR-V splice variant expression. (D) Analysis of MKI67 mRNA expression in ADT and CRPC groups in relation to dependence on AR-V splice variant expression. (E) Analysis of UBE2C mRNA expression in ADT and CRPC groups in relation to dependence on AR-V splice variant expression. p-levels calculated using Mann–Whitney U-test. Statistical significance (p < 0.05): * vs. PRIM group, # vs. CTRL group, § vs. PRIM MTS group.

Figure 5

Heatmap plot of mRNA expression levels of DNA repair genes and AR-V splice variants as a quantitative parameter (left side). Samples represent pooled tumor samples from ADT and CRPC groups. Note similarities in the expression levels of the upper cluster of DNA repair genes (RAD51 to BRCA2) and AR-V splice variants, with the exception of 15 samples on the left side (mostly ADT samples). Clusters are separated by yellow lines.

Figure 6

(A) Distribution of DNA repair scores in study groups (cumulative score based on mRNA expression of 20 DNA repair genes). Light blue points express any of the AR-V splice variants (V1, V3, V7, V9); dark blue points do not express AR-V splice variants. (B) Dependence of DNA repair score on the presence of AR-V splice variants (V1, V3, V7, V9) in ADT and CRPC groups. p-levels calculated using Mann–Whitney U-test. Statistical significance (p < 0.05): * vs. PRIM group, # vs. CTRL group. (C) Correlation analysis shows dependence of the AR score and DNA repair score in PRIM, ADT and CRPC groups.

Figure 7

DNA repair genes differentially expressed in samples from different ADT, CRPC and PRIM groups (p < 0.05; in italic—p = 0.05–0.1) in relation to AR-V expression. Detailed expression analysis of individual DNA repair genes is provided in Supplementary Figures S4–S7. Upregulated genes are shown with a red background and downregulated genes with a blue background. (A) Cumulative analysis of all samples in study groups independent of AR-V status. (B) Analysis of genes affected in samples positive for any of AR-V splice variants (V1, V3, V7, V9) compared to those without AR-V splice variant expression.

Figure 8

(A) The presence of AR-V7 enhances DNA repair in vitro. Increased DNA repair is visualized by a diminished number of residual γH2Ax foci 24 h after irradiation. Nuclear counterstain with DAPI from left to right: LNCaP/AR (0 Gy), LNCaP/V7 (0 Gy), LNCaP/AR (2 Gy), LNCaP/V7 (2 Gy). (B) Quantification of residual γH2A.X foci 24 h after irradiation (2 Gy, blue). Non-irradiated cells were used for comparison (0 Gy, black). In total, >200 nuclei were counted per cell line, irradiative condition and experiment, with a mean of three independent experiments. ** = p < 0.01.

Figure 9

Validation of upregulated DNA repair genes of the AR splice variant expressing CRPC. LNCaP overexpressing AR-V7 served as a surrogate for CRPCs expressing AR-Vs. DNA repair genes upregulated in CRPC+AR-Vs (Figure 7B) were validated by qRT-PCR under irradiated (6 Gy, IR, blue bars) and non-irradiated (0 Gy, Ø, black bars) androgen-deprived conditions. The latter condition, with generally lower AR-V7-specific expression, served as the threshold (light gray box) to identify genes (CHEK1, EXO1, RAD54L, XRCC2) strongly upregulated by irradiation.

Figure 10

(A) Impact of AR splice variants in clinical CRPC via the validation of deregulated DNA repair genes in vitro. LNCaP cells overexpressing AR-V7 served as a surrogate for CRPCs expressing AR splice variants. DNA repair genes deregulated in CRPC vs. PRIM (Figure 7A) were validated by qRT-PCR under irradiated (6 Gy, IR, blue bars) and non-irradiated (0 Gy, Ø, black bars), androgen-deprived conditions. The latter condition, with generally lower AR-V7-specific expression, served as the threshold (light grey box) to identify genes visibly upregulated by irradiation such as RAD54L, EXO1, RMI2. (B) Under the same experimental conditions, genes such as ATM, NBN, MPCH1 were confirmed as downregulated.