A gene expression signature associated with structural differentiation of prostate acini predicts postoperative recurrence of prostate cancer

Abstract

Abstract Glandular malignancies such as prostate carcinoma display a considerable heterogeneity in the degree of structural organization while the prognostic role of the associated molecular changes awaits further exploration. Using a three-dimensional (3-D) tissue organization culture model that permits the structural differentiation of prostate glandular epithelium we profiled the transcriptomes of multicellular organoids formed by untransformed prostate epithelial cells (PECs) or prostate carcinoma cells as well as those cultured as cell monolayers. Interrogation into the multiple (n = 6) sets of differentially expressed genes (DEGs) identified that those altered significantly during the structural differentiation process of polarized PEC acini strongly associated with the probability of recurrence in a cohort of 50 prostate cancer patients receiving radical prostatectomy (RP). By contract, the DEGs related to prostate tumor spheroids did not correlate significantly with prognosis. To further explore this association, we prioritized the genes and applied supervised classification algorithms whereby identified a 39-gene structure-associated prognosis signature (SAPS) that shows a remarkable predictivity for the likelihood of post-RP recurrence (overall accuracy 100%, P < 0.001). The SAPS outperformed standard clinicopathological variables, including Gleason scores, as well as existing multigene prognostic predictors of prostate cancer. Our results suggest a distinct contribution by the glandular structural differentiation program to prostate cancer pathology and the usefulness of a robust and biologically relevant genomic predictor of the prognosis of prostate cancer (Supported by NHRI CA-099-PP-19 and Department of Health of Taiwan H99-TD-C-111-004 to KKC).