Training and validation of a novel 14-gene expression signature associated with prostate cancer metastasis and postoperative recurrence.

Novel biomarkers are needed to improve prostate cancer risk stratification at diagnosis, as current clinical practice is suboptimal. Here, using a training cohort of 107 newly diagnosed prostate cancer patients, primary staged by 68Gallium prostate-specific membrane antigen (PSMA) Positron Emission Tomography-Computed Tomography (PET/CT) imaging (41 localized and 66 metastatic; Cohort 1), we aimed to train a gene expression signature associated with metastasis, a key hallmark of aggressiveness. By mRNA sequencing of primary tumor tissue samples from 107 patients in Cohort 1 and use of Least Absolute Shrinkage and Selection Operator regression, we trained a novel 14-gene model (CaP14) that distinguished localized and metastatic prostate cancer (MPC). Associations between CaP14 and disease outcomes were tested in three independent cohorts, including 17 MPC and 127 radical prostatectomy (RP) patients (Cohort 2), 473 RP patients (Cohort 3), and 174 RP patients (Cohort 4), respectively. Diagnostic/prognostic performance of CaP14 was assessed by receiver operating characteristics, Kaplan-Meier, uni/multivariate Cox regression, and biochemical recurrence (BCR)-free and prostate cancer-specific survival analyses. A high CaP14 score was associated with metastasis and poor prostate cancer-specific survival in Cohorts 1 and 2, and with early BCR in RP Cohorts 2, 3, and 4 independently of routine clinical variables. This is the first study to develop a gene signature capable of predicting metastasis at initial diagnosis, trained using a cohort of patients whose metastatic spread was evaluated using 68Ga-PSMA PET/CT scans. CaP14 has the potential to guide better and more personalized treatment decisions for newly diagnosed prostate cancer in the future.