Resistance to androgen deprivation therapy (ADT) is the main challenge for advanced fatal prostate cancer (PCa), which can gradually develop into metastatic castration-resistant prostate cancer (mCRPC). However, the pathologic mechanisms of mCRPC are still far from clear. Given the high incidence and mortality related to mCRPC, understanding the causes and pathogenesis of this condition as well as identifying potential biomarkers are of great importance. In the research reported here, we integrated several gene expression profiles from hormone sensitive prostate cancer (HSPC) and mCRPC datasets to identify differentially expressed genes (DEGs), key biological pathways, and cellular components. We found that extracellular matrix (ECM) genes were significantly enriched, and further filtered them using Pearson correlation analysis and stepwise regression to find ECM signatures to differentiate between the HSPC and mCRPC phenotypes. Six ECM signatures were input into K-nearest neighbor, logistic regression, naive Bayes, and random forest classifiers models. Random forest algorithm with the six-gene prognostic signatures showed best performance, which had high sensitivity and specificity for HSPC and mCRPC classification and further the six ECM signatures were validated in organoid models. Among the six ECM genes, SPP1 was identified as the key hub signature for PCa metastasis and drug resistance development; we found that both protein and mRNA expression levels of SPP1 were remarkably up-regulated in mCRPC compared with HSPC in organoid models and could regulate the androgen receptor signaling pathway. Therefore, SPP1 is a potential novel biomarker and therapeutic target for mCRPC. Further understanding of the role of SPP1 in mCRPC development may help to explore effectively therapeutic approaches for the prevention and intervention of drug resistance and metastasis.
Keywords: SPP1; castration-resistant prostate cancer; extracellular matrix; metastasis; organoid model.
Copyright © 2019 Pang, Xie, Zhang, Liu, Wu and Cui.