Background: NGS of primary prostate tumors and metastases has illuminated patterns of clonal dissemination, but tissue from multiple sites often is not available in clinical settings. Liquid biopsies are minimally invasive and readily obtained via serial blood draws. We compared genomic alterations detected in primary tissue and metastatic sites with liquid biopsy results from circulating tumor cells (CTCs) and cfDNA. Methods: 7.5ml blood samples from men with metastatic castration resistant PC (mCRPC) were processed using LiquidBiopsy (Cynvenio), an immunomagnetic enrichment platform (EpCAM+CK+DAPI+CD45- ) that yielded a CTC-enriched pellet and matched plasma from each sample. DNA was extracted for targeted NGS (AmpliSeq 2.0, Life Technology) on an Ion Torrent System. Somatic single nucleotide variants (SSNVs) occurring in > 1% of DNA were identified based on presence in CTCs or plasma and absence in a matched WBC pellet and validated by ddPCR (RainDance). Archived FFPE primary and metastatic tumors were analyzed similarly. Results: To date, matched samples have been collected from 5 patients. In one patient with complete data from all tissues, targeted NGS yielded concordant SSNVs in TP53 (p.R175H) and PTEN (p.R130*) in a 2008 prostatectomy tissue (5% and 4% allele frequencies; respectively), a 2014 supraclavicular lymph node biopsy (43% and 36%), and 2014 plasma cfDNA (5% and 4%, confirmed by ddPCR: 7% and 3%). A follow-up plasma cfDNA sample 3 months later had 13% and 5% by ddPCR. NGS of CTC DNA from the same 2014 blood sample identified SSNVs at different loci within the same genes (TP53 c.165-1G > C at 1%; PTEN p.D186N at 1%). Conclusions: Concordant DNA alterations reflecting clonal dissemination were identified by targeted NGS of primary PC and at higher frequency in a metastasis 6 years later. Concurrent liquid biopsy samples detected concordant SSNVs in cfDNA (confirmed by ddPCR), whereas CTC DNA yielded different SSNVs. Expanded prospective cohorts are ongoing to further elucidate which tissue sources, alone or in combination, yield genomic data that best predict clinical outcomes.