Abstract
Metastatic prostate cancer is initially sensitive to androgen receptor inhibition, but eventually becomes castration-resistant prostate cancer (mCRPC). Early use of more intensive therapies targeting androgen receptor and other oncogenic drivers in treatment-naïve primary prostate cancer (PC) may be more effective than that in advanced mCRPC. However, analysis of primary tumors may not reveal targetable metastatic drivers that are subclonal in the primary tumor or acquired at metastatic sites. METHODS: PC samples spanning one patient's clinical course: diagnostic biopsies, pre- or post-enzalutamide metastatic biopsies, and rapid autopsy samples including a patient-derived xenograft (PDX) were analyzed by targeted exome sequencing followed by phylogenetic analysis. RESULTS: Left- and right-lobe primary PC tumors appeared to diverge, with the right acquiring additional shared mutations and striking differences in copy number alterations that later appeared in metastatic samples during the treatment course and at autopsy, whereas the left base tumor maintained a quiet copy number alteration landscape and partitioned into a dead-end node. RB1 loss, a common finding in advanced castration-resistant disease, was identified throughout mCRPC samples, but not in the primary tumor. Significantly, a truncal EGFR-activating mutation (R108K) was identified in the primary tumor and was also found to be maintained in the mCRPC samples and in a PDX model. Furthermore, the PDX model remained sensitive to the EGFR inhibitor erlotinib, despite the presence of both RB1 and BRCA2 losses. CONCLUSION: These findings indicate that truncal alterations identified in primary PC can drive advanced mCRPC, even in the presence of additional strong oncogenic drivers (ie, RB1 and BRCA2 loss), and suggest that earlier detection and targeting of these truncal alterations may be effective at halting disease progression.