A genome-scale CRISPR screen reveals PRMT1 as a critical regulator of androgen receptor signaling in prostate cancer

Abstract

Androgen receptor (AR) signaling is the central driver of prostate cancer across disease states. While androgen deprivation therapy (ADT) is effective in the initial treatment of prostate cancer, resistance to ADT or to next-generation androgen pathway inhibitors invariably arises, most commonly through the re-activation of the AR axis. Thus, orthogonal approaches to inhibit AR signaling in advanced prostate cancer are essential. Here, via genome-scale CRISPR-Cas9 screening, we identify protein arginine methyltransferase 1 (PRMT1) as a critical mediator of AR expression and signaling. PRMT1 regulates the recruitment of AR to genomic target sites and the inhibition of PRMT1 impairs AR binding at lineage-specific enhancers, leading to decreased expression of key oncogenes, including AR itself. In addition, AR-driven prostate cancer cells are uniquely susceptible to combined AR and PRMT1 inhibition. Our findings implicate PRMT1 as a key regulator of AR output and provide a preclinical framework for co-targeting of AR and PRMT1 in advanced prostate cancer.

Keywords: CRISPR screen; PRMT1; androgen receptor; prostate cancer; splicing; superenhancer; transcription.

Figure 1.. Genome-scale CRISPR/Cas9 screen identifies regulators of AR/AR-V7 expression.

(A) Schematic of 22Rv1/AR-V7-GFP reporter cell line. CRISPR/Cas9 editing and homology-directed repair were used to insert a GFP-containing cassette directly upstream of the cryptic exon 3 (CE3) stop codon in 22Rv1 cells. (B) Schematic of the CRISPR/Cas9 screening strategy used to identify regulators of AR/AR-V7 expression in 22Rv1/AR-V7-GFP cells. (C) Screen hits, plotted by STARS score on day 5 or day 12 after library transduction, determined by enrichment of sgRNAs in the sorted GFP-negative population as compared with the starting library pool. Top: Scatterplot of STARS scores for screen hits on day 5 versus day 12. For plotting purposes, hits that scored at only one timepoint were assigned a STARS score of 0.5 for the timepoint that they were not enriched. Bottom: Plots of false discovery rate (FDR) versus STARS score of hits from day 5 (left) or day 12 (right) timepoints. Selected high-scoring hits are labeled. (D) Arrayed validation of screen hits by RT-qPCR in parental 22Rv1 cells. Heatmap shows relative AR-FL and AR-V7 expression in 22Rv1 cells at the indicated timepoints after knockout of selected screen hits. mRNA levels are normalized to a control sgRNA (control98). Data are presented as the mean of n = 4 technical replicates. (E) Enrichment analysis showing gene ontology (GO) terms significantly enriched among screen hits scoring on either day 5 or day 12 with FDR < 0.25. See also Figure S1 and Tables S1 and S2.

Figure 2.. PRMT1 regulates AR expression and activity.

(A) Relative AR-FL, AR-V7, and PRMT1 expression, as assessed by RT-qPCR, with or without PRMT1 knockdown by dox-inducible shRNA in the prostate cancer cell lines 22Rv1, VCaP, and LNCaP. Expression is shown relative to no dox. Error bars represent mean ± SD, n = 3 biological replicates. (B) Relative luciferase activity upon PRMT1 knockdown in LNCaP cells transduced with an androgen-responsive MMTV-Luciferase reporter. Luciferase activity is normalized to cell viability for each condition and shown relative to no dox. Error bars represent mean ± SD, n = 6 biological replicates. (C) Relative AR-FL expression in LNCaP cells after treatment with the PRMT1 inhibitor furamidine at the indicated doses. Expression is shown relative to DMSO. Error bars represent mean ± SD, n = 3 biological replicates. (D) Relative MMTV-Luciferase activity in LNCaP cells upon treatment with furamidine at the indicated doses. Luciferase activity is normalized to cell viability at each concentration and shown relative to DMSO. Error bars represent mean ± SD, n = 3 biological replicates. For A-D, statistical significance was determined by t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. See also Figure S2.

Figure 3.. PRMT1 expression is associated with advanced prostate cancer.

(A) PRMT1 expression in normal prostate tissue, tumor-adjacent normal tissue, primary prostate tumors, and metastatic tumors from a published dataset (Chandran et al., 2007). The metastatic tumor group includes multiple samples from each of four patients and points are colored by patient of origin. The y-axis represents signal intensity values from an oligonucleotide microarray probe against PRMT1. Statistical significance was determined by Mann-Whitney U test. (B) Left: PRMT1 staining in three representative prostate tumor samples and one representative normal prostate tissue sample, from prostate TMA. Right: Quantification of percent PRMT1 positive nuclei in tumor versus normal cores on the prostate TMA. Statistical significance was determined by t-test. (C) Kaplan-Meier plot showing disease-free survival after prostatectomy among patients with prostate cancer with low, intermediate, or high PRMT1 expression in a published dataset (Hoadley et al., 2018). The survival distributions of the three groups are significantly different (log-rank test, p = 0.03). Also shown are Benjamini-Hochberg adjusted p-values of log-rank pairwise comparisons. (D) Relative expression of AR and target genes KLK2, KLK3, SLC45A3, and NKX3–1 in published mRNA expression data (Abida et al., 2019) from castration-resistant prostate cancer tumors with low or high PRMT1 expression. Statistical significance was determined by Mann-Whitney U test. For A-D, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Figure 4.. Inhibition of PRMT1 impairs AR binding to genomic target sites.

(A) Volcano plot showing differentially expressed genes in LNCaP cells upon PRMT1 knockdown by dox-inducible shRNA, as assessed by transcriptome sequencing. Genes meeting significance and differential expression thresholds of adjusted p-value < 0.05 and |log2FC| > 0.5 are colored in red. Of these, genes proximal to tumor-specific AR binding sites (t-ARBSs) (Pomerantz et al., 2015), are colored in black and are significantly enriched among downregulated genes (Fisher’s exact test, p = 0.002). AR, PRMT1, and four selected canonical AR target genes are labeled. n = 3 biological replicates were used in each condition. (B) Enrichment analysis of genes downregulated upon knockdown of PRMT1. The top 100 downregulated genes are enriched for targets of the transcription factors listed in the left panel and are highly co-expressed with transcription factors listed in the right panel. (C and D) Venn diagram showing overlap of AR binding sites identified by AR ChIP-seq in LNCaP cells expressing either control shRNA (shLacZ) or shPRMT1 (C) or LNCaP cells treated with either DMSO or furamidine (8 μM) (D). (E) Enrichment analysis of genes located in proximity to AR peaks that were lost by both furamidine treatment and PRMT1 knockdown. Top enriched transcription factor target gene sets are shown. (F) Heatmap of AR binding density over AR peaks shared between shLacZ and shPRMT1 conditions (top) or lost upon PRMT1 knockdown (bottom). Peaks are rank-ordered by AR signal within 3 kb flanking the peak center. Profile plots on the right show average AR ChIP-seq signal in the regions displayed in the heatmaps. Two biological replicates are shown for each condition. (G) Same data representation as in (F) for AR peaks shared between DMSO and furamidine treatment (top) or lost upon furamidine treatment (bottom). See also Figure S4 and Tables S3, S4, and S5.

Figure 5.. Suppression of PRMT1 decreases AR target gene expression through reduced AR occupancy and H3K27 acetylation at lineage-specific enhancers.

(A) Venn diagram showing the overlap of H3K27ac peaks in LNCaP cells with AR peaks in LNCaP cells expressing shLacZ or shPRMT1. (B) Distribution of H3K27ac ChIP-seq signal across 32,515 typical enhancers (TEs) or superenhancers (SEs) in LNCaP cells. The 1,120 SEs, characterized by high H3K27ac signal, are colored in black. Gene labels indicate SEs proximal to AR or AR target gene loci. (C) Donut plots showing the proportion of TEs (top) or SEs (bottom) occupied by AR. (D and E) Heatmaps of AR (D) and H3K27ac (E) ChIP-seq signal over AR peaks in SE or TE regions, shown in the context of either control shRNA or shPRMT1. Peaks are rank-ordered by AR signal within 3 kb of the peak center. H3K27ac signal is shown within 10 kb flanking the peak center. Two biological replicates are shown for each condition. (F and G) Profile plots of average AR (F) or H3K27ac (G) signal in the regions shown in (D) and (E). (H) Boxplots showing average signal per million reads (SPMR) per base over each H3K27ac peak within AR-occupied or non-AR-occupied SEs in cells expressing shLacZ or shPRMT1. Data represent average of two biological replicates. p-values were calculated by Mann-Whitney U test. (I) Volcano plot of differentially expressed genes upon PRMT1 knockdown as determined by transcriptome sequencing. Genes meeting significance and differential expression thresholds of adjusted p-value < 0.05 and |log2FC| > 0.5 are colored in dark gray. Of these, genes proximal to AR-occupied SEs (ARBS-SE) are shown in blue while those proximal to non-AR-occupied SEs (non-ARBS-SE) are shown in red. Fisher’s exact test was used to determine enrichment of ARBS-SE-proximal genes (p = 0.029) or non-ARBS-SE-proximal genes (p = 0.131) among those downregulated by PRMT1 knockdown. Selected canonical AR target genes located near SE in LNCaP cells are labeled. (J) AR and H3K27ac ChIP-seq signals at SE regions regulating KLK2, KLK3 (top) or AR (bottom) expression are shown in the context of DMSO or furamidine (Fur) treatment, or control or PRMT1 knockdown. Superenhancers are indicated by black bars. Signals represent average of two biological replicates. See also Figures S5 and S6 and Table S6.

Figure 6.. AR-expressing prostate cell lines exhibit selective dependency on PRMT1.

(A) Western blot showing relative AR-FL and AR-V7 expression in parental prostate cancer cell lines. (B) Proliferation of AR-expressing (LNCaP, 22Rv1, VCaP) or non-AR-expressing (PC3) cell lines with or without dox-induced PRMT1 knockdown. Confluence readings were taken using an IncuCyte live-cell imager. Error bars represent mean ± SD of the following numbers of biological replicates: n = 4 (LNCaP), n = 8 (22Rv1), n = 4 (VCaP), n = 3 (PC3). (C) Doubling times of prostate cancer cell lines with or without PRMT1 knockdown, estimated by nonlinear regression of confluence readings shown in (A). Data are presented as mean with 95% CI. Statistical significance was determined by t-test. (D) Relative viability (normalized to DMSO) of prostate cancer cell lines after 5 days of treatment with furamidine at the indicated concentrations. Error bars represent mean ± SD, n = 3 biological replicates. **p < 0.01, ****p < 0.0001. See also Figure S7.

Figure 7.. Inhibition of PRMT1 synergizes with enzalutamide to suppress growth of CRPC cells driven by enhanced AR signaling.

(A) Heatmaps showing percent viability of prostate cancer cell lines after 7 days of combination treatment with the indicated doses of furamidine and enzalutamide. LNCaP/AR-Enh cells are derived from the parental LNCaP line and contain knock-in of an additional copy of the AR enhancer (Takeda et al., 2018). Viabilities are shown relative to the DMSO condition. Quantized heatmaps of Bliss synergy index are shown below each cell line. A box is drawn around the dose combination in each cell line that resulted in the maximum Bliss synergy score. Data represent the mean of n = 3 biological replicates. (B) Percent viabilities for single-agent compared to combination treatment at the doses indicated by boxes in (A) for each cell line. Dotted lines indicate predicted additive effect of enzalutamide (E) and furamidine (F), calculated by multiplying the percent viabilities upon single-agent treatment at the respective doses. Error bars represent mean ± SD, n = 3 biological replicates. (C) Western blot showing relative AR protein levels in LNCaP and LNCaP/AR-Enh cells upon furamidine treatment in the context of androgen depletion. Cells were seeded in media supplemented with charcoal stripped serum and treated with DMSO or furamidine (8 μM) for 5 days. (D) Densitometric quantification of western blot in (C). AR protein levels are normalized to actin and shown relative to parental LNCaP treated with DMSO. Error bars represent mean ± SD, n = 3 biological replicates. For A-D, statistical significance was determined by t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (E) Impact of enzalutamide (Enz) or furamidine (Fur) monotherapy, or combination treatment, on 22Rv1 xenograft growth. Error bars represent mean ± SEM for the indicated number of biological replicates. *p < 0.05 for enzalutamide + furamidine vs enzalutamide at the final time point (t-test), **p < 0.01 for enzalutamide + furamidine versus each other condition by 2-way ANOVA and Tukey’s method.