Omics integration identified CRIP2 as a key mediator of olaparib resistance in prostate cancer.

BACKGROUNDThe mechanisms driving resistance to poly(ADP-ribose) polymerase inhibitors (PARPis), such as Olaparib, in castration-resistant prostate cancer (CRPC) that emerge after androgen deprivation therapy (ADT) remain largely undefined. This study aimed to systematically identify tumor epithelial cell subpopulations associated with Olaparib resistance by integrating single-cell transcriptomics, spatial transcriptomics (ST), and bulk RNA-sequencing (RNA-seq) data, and to construct a risk prognostic model.

METHODSSingle-cell RNA-sequencing (scRNA-seq) data from six CRPC patients were subjected to quality control, clustering, annotation, and dimensionality reduction to identify distinct epithelial cell subpopulations. The Scissor algorithm, combined with Olaparib half-maximal inhibitory concentration (IC50) phenotype data from the Genomics of Drug Sensitivity in Cancer (GDSC) database, was employed to select the Scissor+ cell subpopulations significantly associated with drug resistance. The BayesPrism algorithm was then used to estimate the proportion of Scissor+ cells in the bulk RNA-seq samples, to explore their relationship with prognosis. ST data from the 10x Genomics platform were integrated to validate the spatial distribution of the Scissor+ cells in tissues. Additionally, bulk RNA-seq data from multiple cohorts [including The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO)] were integrated to construct an olaparib resistance-associated gene (ORAG) risk model, which was optimized through various machine-learning algorithms. Finally, the core gene CRIP2 was validated through in vitro drug sensitivity assays to investigate its role in olaparib resistance.

RESULTSThe Uniform Manifold Approximation and Projection (UMAP) analysis of the single-cell data revealed significant cellular heterogeneity in the CRPC tissues, and the Scissor algorithm identified the epithelial cell subpopulations (Scissor+) strongly linked to Olaparib resistance. The BayesPrism analysis showed that the increased proportion of Scissor+ cells in the bulk samples was significantly associated with a poor prognosis (P=0.04). ST further validated that this subpopulation was predominantly enriched in tumor regions, displaying a clear spatially specific expression pattern. In seven independent bulk RNA-seq cohorts, the ORAG model, which was built using least absolute shrinkage and selection operator (LASSO) and random survival forest (RSF) algorithms, showed excellent prognostic predictive performance [with the highest concordance index (C-index) of 0.722]. CRIP2 was identified as a key risk gene, and found to be correlated with a worse survival prognosis, higher tumor stage (T stage), Gleason score, and prostate-specific antigen (PSA) levels. In vitro experiments confirmed that CRIP2 knockdown significantly increased the sensitivity of LNCaP and C4-2B cells to olaparib.

CONCLUSIONSThis study was the first to systematically identify olaparib resistance-associated tumor epithelial cell subpopulations in CRPC by integrating single-cell, spatial, and bulk multi-omics data. The constructed ORAG risk-prediction model has great potential for clinical application. CRIP2, which was identified as a key regulatory factor of resistance, offers a novel strategy for personalized treatment and the study of resistance mechanisms in CRPC.