ABSTRACT: We have recently demonstrated that the undifferentiated PSA-/lo prostate cancer (PCa) cell population harbors self-renewing long-term tumor-propagating cells that are refractory to castration, thus representing a therapeutic target. Our goals here are, by using the same lineage-tracing reporter system, to track the dynamic changes of PSA-/lo and PSA+ cells upon castration in vitro, investigate the molecular changes accompanying persistent castration, and develop large numbers of PSA-/lo PCa cells for drug screening. To these ends, we treated LNCaP cells infected with the PSAP-GFP reporter with three regimens of castration, i.e., CDSS, CDSS plus bicalutamide, and MDV3100 continuously for up to ~21 months. We observed that in the first ~7 months, castration led to time-dependent increases in PSA-/lo cells, loss of AR and PSA expression, increased expression of cancer stem cell markers, and many other molecular changes. Meanwhile, castrated LNCaP cells became resistant to high concentrations of MDV3100, chemotherapeutic drugs, and other agents. However, targeted and medium-throughput library screening identified several kinase (e.g., IGF-1R, AKT, PI3K/mTOR, Syk, GSK3) inhibitors as well as the BCL2 inhibitor that could effectively sensitize the LNCaP-CRPC cells to killing. Of interest, LNCaP cells castrated for >7 months showed evidence of cyclic changes in AR and the mTOR/AKT signaling pathways potentially involving epigenetic mechanisms. These observations indicate that castration elicits numerous molecular changes and leads to enrichment of PSA-/lo PCa cells. The ability to generate large numbers of PSA-/lo PCa cells should allow future high-throughput screening to identify novel therapeutics that specifically target this population.