A 12-gene panel in estimating hormone-treatment responses of castration-resistant prostate cancer patients generated using a combined analysis of bulk and single-cell sequencing data

BACKGROUND: Castration-resistant prostate cancer (CRPC) represents one type of advanced prostate cancer (PCa) with a median survival time of 1–2 years. Currently, there is a lack of reliable gene panels in predicting hormone treatment (HT) responses due to limited knowledge of CRPC-specific tumor-microenvironment (TME) characteristics. METHODS: In this study, we first screened for up-regulated genes in CRPC samples using bulk-sequencing data retrieved from TCGA online database, and further investigated the expression status of these genes in four sets of downloaded single-cell RNA sequencing (scRNAseq) data: GSE117403 containing 16 normal human prostate samples; GSE141445 containing 13 PCa samples; GSE176031 containing 11 PCa samples and GSE137829 containing 6 CRPC samples. RESULTS: We identified a series of CRPC-specific TME characteristics including an enriched number of PEG10(+) neuroendocrine cells, elevated expression of PPIB/CCDC74A/GAPDH/AR genes in tumor cells, increased expression of FAP/TGFB1 in cancer-associated fibroblasts (CAFs), suppressed immune environment featured by enhanced M2 macrophage polarization, T cell exhaustion and increased number of regulatory B cells. We further established a 12-gene panel using these characteristics and showed that this panel could separate CRPC samples from PCa samples (AUC of 0.78), and CRPC patients with higher panel scores tended to have treatment failure or progression (R = −0.47, p = 0.019). CONCLUSIONS: Based on these unique TME characteristics of CRPC, we established a prediction tool for estimating the duration of HT responses in PCa treatment. Our results suggest mechanisms by which prostate cancer becomes castrate resistant. Further study of PEG10 (and/or others) to evaluate therapeutic efficacy should be considered.