A Genome-Wide CRISPR Activation Screen Identifies PRRX2 as a Regulator of Enzalutamide Resistance in Prostate Cancer

Abstract

Androgen receptor (AR) pathway inhibitors are the mainstay treatment for advanced prostate cancer, but resistance to therapy is common. Here, we used a CRISPR activation screen in metastatic castration-sensitive prostate cancer cells to identify genes that promote resistance to AR inhibitors. Activation of the TGFβ target gene paired-related homeobox2 (PRRX2) promoted enzalutamide resistance. PRRX2 expression was the highest in double-negative prostate cancer (DNPC), which lack AR signaling and neuroendocrine differentiation, and a PRRX2-related gene signature identified a subset of patients with DNPC with reduced overall survival. PRRX2-expressing cells showed alterations in the CDK4/6/Rb/E2F and BCL2 pathways. Accordingly, treatment with CDK4/6 and BCL2 inhibitors sensitized PRRX2-expressing, castration-resistant tumors to enzalutamide. Overall, PRRX2 was identified as a driver of enzalutamide resistance. The PRRX2 signature merits investigation as a biomarker of enzalutamide resistance in prostate cancer that could be reversed with CDK4/6 and BCL2 inhibitors.

Significance: PRRX2 mediates enzalutamide resistance via activation of the E2F and BCL2 pathways, which can be targeted with CDK4/6 and BCL2 inhibitors to reverse resistance.

Trial registration: ClinicalTrials.gov NCT02452008 NCT03751436 NCT03900884.

Conflict of interest statement

Conflict of Interest: The authors declare no potential conflicts of interest

©2022 American Association for Cancer Research.

Figures

Fig. 1.. CRISPRa genome-wide screen identifies PRRX2 as a driver of enzalutamide resistance.

A) CRISPRa Screen design. Cells were transduced with the Calabrese library at a low MOI (0.3–0.5) and 4 days after transfection, cells were treated with vehicle (Veh) or 25 μM enzalutamide (ENZ) for 7.5 or 9 weeks before sequencing. ENZ treated cells at 7.5 weeks were sorted into GFP (+) and GFP (–) cells using FACS. We sequenced the sgRNAs from week 0 (T0), Vehicle (Veh), ENZ-GFP (+), ENZ-GFP(−) and ENZ unsorted cells at week 9 (ENZ-All). B) Violin plot of log2 sgRNA counts representation within the different populations. Lines represent the median and the interquartile range. C,D) Ranked-ordered plot of normalized counts for the fold change (FC) of the ENZ-All population vs T0 (C) and the ENZ-All vs Veh (D). E) Fold change of the sgRNA counts comparison between GFP (+) vs GFP (−) for genes shown in graph. F) Significantly over-represented pathways (q<0.15) for genes with >3.5 FC in the ENZ-All vs T0.

Fig. 2.. PRRX2 mediates enzalutamide resistance in vitro and in vivo.

A) Cell counts of LNCaP-sgNC and LNCaP-sgPRRX2 cells after treatment with ENZ (5μM) at different time-points. Insert: Western blot comparing PRRX2 levels in LNCaP-sgNC vs LNCaP-sgPRRX2 cells. B) Cell counts of LNCaP-sgNC and LNCaP-sgPRRX2 cells in charcoal stripped (CSS) media at different time-points. C) Colony formation assay after 14 days of treatment with ENZ (5μM) or DMSO. D) Growth in low attachment plates (GILA) assay for 3D spheroids of LNCaP-sgNC vs LNCaP-sgPRRX2 cells grown in CSS and different concentrations of ENZ for 7 days. E) Cell proliferation assay of 22RV1 cells after PRRX2 knockdown using shRNA. F) Colony formation assay of 22RV1-shNC and 22RV1-shPRRX2–1 cells after treatment with ENZ (40μM) for 14 days. G) In vivo experimental design. H) Relative tumor growth of LNCaP-sgNC and LNCaP-sgPRRX2 xenografts in castrated mice treated with ENZ (10mg/kg) IP 3 times/week. I) Tumor incidence (volume ≥ 100mm3) after 1 month. All viability experiments were performed in at least 3 independent biological replicates. Statistical tests: p-value determined by two-sided t-test (A-H). Chi-squared test for I. Error bars represent S.E.M. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Fig. 3.. PRRX2 modulates the expression of enzalutamide-regulated genes.

A) Experimental design of RNA-seq experiment. LNCaP-sgNC and LNCaP-sgPRRX2 cells were treated with 5μM ENZ (E) or DMSO (D) for 7 days. B) Heatmap of RNA-seq data in LNCaP-sgNC-DMSO (NC-D), LNCap-sgNC-ENZ (NC-E), LNCaP-sgPRRX2-DMSO (PR-D) and LNCaP-sgPRRX2-Enz (PR-E). Clusters were generated using K-means clustering function in Morpheus software. C, D, E) Top significant pathways enriched in Cluster 1, 2 and 4 respectively. F) GSEA plot of Androgen Response geneset (Hallmark dataset) enriched in LNCaP-sgPRRX2 cells in DMSO conditions. G) GSEA analysis of different signatures in PR-E vs NC-E cells.

Fig. 4.. The AR and TGF-β pathways converge to cross-regulate PRRX2.

A) Spearman r correlation between AR score and PRRX2 log2 expression levels. AR score and expression data were obtained from cBioPortal TCGA provisional database (n = 499) and SU2C (n = 209). B) PC patients from Stelloo et al (n = 83) were clustered using K-means clustering according to their PRRX2 and AR expression (z-values) and number of AR ChIP peaks (z-values). C) PRRX2 expression in paired patient samples (Rajan et al) before and after 22 weeks of ADT treatment. * *p<0.05 by paired two tail t-test. D) ChIP-seq data of AR and SMAD3 binding peaks near the PRRX2 promoter in different PC patients and cell lines. IGV was used for visualization E) Spearman r correlation between PRRX2, TGFB1 and TGFBR2 expression. Data obtained from cBioPortal (SU2C, 2019) n=208. F) Scatter plot of AR and PRRX2 expression. Circle color represents TGFB1 expression. Spearman r coefficient shown for AR and PRRX2 expression. Data obtained from Stelloo et al. (n = 82) G, H) PRRX2 expression measured by RT-qPCR (G) or Western blot (H) in LNRII (LNCaP cells expressing TGFBR2) after treatment with ENZ (5μM), TGFB1 (5ng/ml) or the combination during 48h. β-Actin served as loading control. Experiments performed in 3 independent biological replicates. Statistical analysis in G done with unpaired two-tail t-test. Error bars represent S.E.M. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Fig. 5.. PRRX2 expression and a PRRX2 signature are elevated in DNPC.

A) Scatter plot of AR and NEPC scores within the SU2C patient dataset. ARPC defined as AR Score < 0.5 and NEPC Score <0.5 (n=85). NEPC defined as NEPC Score > 0.5, AR Score < 0.5 (n=22). DNPC defined as AR score < 0.5 and NEPC score < 0.5 (n=71). ARPC/NEPC defined as both AR Score and NEPC score >0.5 (n= 5). B) PRRX2 expression levels in ARPC, NEPC, NEPC/ARPC and DNPC subgroups. C) Heatmap of the top 21 differentially expressed genes between LNCaP-sgNC-DMSO (NC-D) and LNCaP-sgPRRX2-DMSO (PR-D) used for the PRRX2 signature. D) Similarity matrix of patients in the SU2C cohort using AR, NEPC and PRRX2 signature genes. E) Hierarchical clustering of the SU2C, 2019 patients according to the expression of AR, NEPC and PRRX2 signature genes. F) Overall survival plot of DNPC patients classified into PRRX2 high (25th percentile of PRRX2 score) and PRRX2 low (rest of the patients). G) Colony formation assay of DU145 after PRRX2 knockdown using two different shRNAs. Experiment performed in 3 independent biological replicates H) LAPC9-AI organoid size after PRRX2 knockdown with two different shRNAs. Experiment performed in 3 independent biological replicates I) Tumor growth of LAPC9-AI-shNC and LAPC9-AI-shPRRX2–1 inoculated into castrated mice. Clustering: D, E Clustergrammer was used to build the matrix using 1-cosine distance and average linkage. Statistical analysis: Log-rank analysis used for survival analysis. Unpaired two tail t-test for G, H, I. Error bars represent S.E.M. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001

Fig. 6.. PRRX2 regulates the Rb1/E2F and BCL2 pathways.

A) Volcano plot of ATAC-seq genes. Colored genes represent p-value < 0.05 B) Venn diagram showing the overlap between RNAseq (p<0.05) and ATAC-seq (p<0.05) genes. C) Enrichment analysis of different pathways using the list of genes that overlap (n=333) between RNAseq and ATAC-seq experiments in B. D) Enrichment analysis using the genes up-regulated in PR-D vs NC-D RNAseq (p<0.05) and the open chromatin regions in the PR-D vs NC-D (p<0.05) (n=100). E) Western blot analysis of LNCaP-sgNC and LNCaP-sgPRRX2 cells treated with ENZ (5μM) for 10 days. F) Western blot analysis of LNCaP-sgNC and LNCaP-sgPRRX2 cells grown in CSS for 7 days. G) Proposed link of PRRX2 to the Rb/E2F pathway. H) Palbociclib (Palbo) IC50 of LNCaP-sgNC and LNCaP-sgPRRX2 cells after treatment for 4 days. I) Area quantification of GILA assay spheroids. Cells were grown in CSS media and treated with ENZ (5μM) for 7 days. Representative images shown. Scale bar = 50μm. J) Compusyn was used to calculate the combination index (CI) for cells were treated with Palbo and ENZ for 4 days. K) Western blot analysis of LNCaP-sgNC and LNCaP-sgPRRX2 cells treated with ENZ (5μM) or grown in CSS media for 7 days. L) GILA assay for LNCaP-sgPRRX2 cells grown in androgen free CSS media and treated with ENZ (5μM), 12μM venetoclax (Veneto) or the combination for 4 days. M) Sphere size quantification of experiment in C. Scale bar = 200μm. All experiments were performed at least in 3 independent biological replicates. Statistical tests: Unpaired two tail t-test (H, I, L, M). Error bars represent S.E.M. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Fig. 7.. Pharmacological inhibition of CDK4/6 and BCL2 sensitizes PRRX2 over-expressing CRPC to enzalutamide.

A) Established LNCaP-sgPRRX2 xenograft bearing NSG mice were castrated. When tumor re-grew, they were treated with vehicle (Veh), 10mg/kg enzalutamide thrice weekly (ENZ), 100 mg/kg/d venetoclax (Veneto), 100 mg/kg palbociclib, 5 days on, 2 days off (Palbo) alone or in combinations as indicated. Veh (n=4), ENZ (n=6), ENZ + Veneto (n=6), ENZ + Palbo (n=6), ENZ + Veneto + Palbo (n=6). Statistics: ANOVA comparison between (Veh, ENZ, ENZ+Veneto) and between (ENZ + Veneto, ENZ + Palbo, ENZ + Veneto + Palbo) B) Immunostaining (IHC) images of Ki67 and cleaved caspase 3 (CC3) (quantified as number of cells per optical field POF). Scale bar = 50 μm. C) Quantification of data from B. Four tumors were stained per condition and 3 images/slide were analyzed. Statistical tests: Unpaired two tail t-test. Error bars represent S.E.M. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.