Background: The mainstay of therapy in metastatic prostate cancer is androgen receptor (AR) signaling inhibition. However, the emergence of early castration resistance or neuroendocrine transformation is associated with poor prognosis. Reliable biomarkers are needed to identify these patients and guide selection of clinical therapy. Methods: mRNA was isolated from EpCAM-positive circulating tumor cells (CTCs) isolated from patients with CSPC, CRPC, or NEPC to measure expression of KLK2, KLK3 (PSA), TMPRSS2, FOLH1 (PSMA), synaptophysin (SYP), and chromogranin (CHGA). Post-hoc retrospective analysis of an institutional review board–approved prospective cohort (N = 98) was performed to identify patterns of gene expression. Samples were considered AR+ if 3 of 4 AR pathway genes (TMPRSS2, KLK2, KLK3, and FOLH1) were positive, and were considered NE+ if either or both SYP or CHGA were positive. Blood samples from two prospective clinical trials of men with mCRPC treated with abiraterone and enzalutamide, respectively, were analyzed to confirm results. Longitudinal samples were collected from 17 patients (6 NEPC and 11 Adenocarcinoma) and cell free DNA was isolated and sequenced using a novel targeted exon panel. Results: AR and/or NE positive patients were found to have a median overall survival (OS) of 8.58 months as compared to a median OS of 29.6 months in the AR and NE negative population (p<0.0001; HR=2.75 [1.60-4.56]). In the subset of castrate resistant prostate cancer (CRPC) patients, AR+ and/or NE+ patients (n=31) were found to have a median OS of 6.74 months vs 18.79 months in the AR- and NE- group (n=39) (p= 0.0009; HR 2.38 [1.36-4.18]). We also tested samples from a phase II ARSI trials with abiraterone and enzalutamide, respectively. None of the baseline samples from these two trials met the above criteria for NEPC (AR- NE+) on their CTCs or histologically. Three of 48 total patients were identified with expression in their baseline blood samples without loss of AR target gene expression (AR+/NE+). Preliminary analysis of this small patient cohort in comparison to patients who were NE- shows that NE+ patients had worse OS (HR= 5.5906 [1.143-27.36), as would be expected by patients with emerging neuroendocrine differentiation. Integrated ctDNA sequencing identified mutations in genes associated with NEPC. Conclusions: The expression of NE genes in liquid biopsies while retaining AR target gene expression is associated with worse OS and may indicate the transition to neuroendocrine differentiation, with clinical characteristics consistent with this phenotype. Early identification of these patients may improve therapeutic decisions and improved patient outcomes. Pairing genomic alteration with changes in gene expression may additionally offer the basis for a new mechanism to assess efficacy of novel therapeutics in future clinical trials.