Individualized transcriptomic changes in malignant and benign prostate identify markers of disease recurrence

Abstract Better tools to predict the likelihood of a recurrence of prostate cancer (PC) after a radical prostatectomy are needed and would provide more personalized therapies for the patients early enough to efficiently control the disease burden. In this study, the transcriptome profiles from matched malignant and adjacent benign prostate tissue were compared in 41 patients to facilitate the identification of novel markers for PC aggressiveness. Hierarchical clustering separated benign and malignant tissues and identified expected transcriptional changes associated with carcinogenesis. Reproducibility-optimized statistical testing (ROTS) identified 45 genes whose expression change between malignant and benign prostate tissue from the same patient differed in patients with and without biochemical recurrence (FDR < 0.05). The results indicated two interconnected regulatory pathways underlying the pathogenesis: one centered on the NR4A, FOS , and EGR family transcription factors, and another centered on IL6. Expression of these 45 transcripts was also found to differ between more aggressive Luminal B-type PC and Luminal A- and Basal-type PC. A machine learning method (SIVS) was then used to further define five transcripts ( BPIFB2 , NR4A2 , NR4A3 , C11orf96 , DUSP5 ) whose individualized malignant-to-benign expression difference within a patient was most strongly associated with an increased risk of relapse. Of these, BPIFB2 and NR4A2 were responsive to antiandrogen treatment in VCaP xenografts in castrated mice in a direction concordant with the malignant-to-benign expression ratio associated with reduced risk of BCR, and were also co-expressed in the PC epithelium. In summary, this study identified novel tumor-expressed markers whose expression changes during malignancy are associated with PC aggressiveness.