Background: We recently identified mechanisms underlying the clonal evolution of castration-resistant prostate adenocarcinoma (CRPC-Adeno) to a neuroendocrine resistance phenotype (Beltran et al, Nat Med 2016). We aimed to develop a non-invasive approach to identify patients that are developing NEPC. Methods: We performed whole exome sequencing of matched ctDNA, germline DNA, and metastatic biopsies from patients with CRPC-Adeno and NEPC. After applying ad hoc partial duplication filtering, we used FACETS and extended CLONET to calculate the fraction of tumor DNA and clonality of genomic lesions. Results: 64 CRPC patients were prospectively enrolled. The spectrum of alterations captured by WES of ctDNA was consistent with those commonly observed in CRPC validating the feasibility of the approach. The similarity of copy number alterations between tumor tissue and ctDNA was higher in NEPC compared to CRPC-Adeno (p = 0.0001) suggesting less heterogeneity in NEPC. There was enrichment of RB1 and TP53 loss in NEPC ctDNA and ARgains in CRPC-Adeno. The overall fraction of mutations shared by ctDNA and tumoral tissue was ~80%. We compared three different tumor biopsy time-points of patient PM161—CRPC-Adeno (lymph node), CRPC-Adeno (bone), NEPC (liver). Unexpectedly the baseline ctDNA profile (at time of CRPC-Adeno) displayed genomic features most similar to the NEPC liver biopsy. These data suggest that NEPC alterations are detectable in the circulation potentially prior to the development of NEPC clinical features. We compared the ctDNA of another patient PM0 with 6 sites of NEPC metastases obtained 6 days later at autopsy; the relative contribution of tumor alterations in ctDNA was highest for the liver metastasis (similarity 0.59) versus other sites suggesting differential contribution of metastatic sites in the circulation, with implications for the interpretation of single site clinical biopsies. Conclusions: This is the first study to show that WES of ctDNA is feasible in CRPC and can help elucidate intra-patient heterogeneity and identify the spectrum and frequency of NEPC genomic changes. ctDNA may improve the detection of patients transforming towards NEPC.