Abstract Prostate cancer (PCa) is a heterogeneous disease. Responses to therapy and prognosis vary significantly from patient to patient, and biomarkers that can predict therapy response and prognosis are urgently needed. In an attempt to identify metabolic mechanisms of hormonal therapy for PCa, we discovered that an important function of androgen receptor (AR) in advanced PCa is to upregulate the expression of glutaminase 1 (GLS1) which is critical for glutamine utilization by cancer cells to compensate for their inability to produce sufficient ATP and metabolic intermediates from glucose due to the Warburg effect. Hormonal therapy targeting AR inhibits GLS1 expression and glutamine utilization, starving cells to death. GLS1 has two isoforms, kidney-type glutaminase (KGA) and glutaminase C (GAC). Our work demonstrates that early stage, hormone-sensitive PCa mostly expresses kidney-type glutaminase (KGA), the weaker, AR-dependent GLS1 isoform, while late stage, therapy-resistant PCa mostly expresses the more potent and AR-independent glutaminase C (GAC). We have shown that this switch in the expression of GLS1 isoforms is a molecular basis for the important clinical phenomenon including tumors’ initial sensitivity to hormonal therapy and the eventual therapy resistance. Importantly, we also observed significant heterogeneity in the GLS1 isoform expression from case to case. For example, while the majority of hormone sensitive cancers expresses KGA, a minority expresses GAC. Similarly, while most cases of late stage PCa express GAC, a minority expresses KGA. The heterogeneous expression of GLS1 isoforms of different enzymatic activities by PCa of various disease stages was intriguing and prompted us to determine their possible correlations with the heterogeneous clinical outcomes observed in PCa patients. We will study the value of GLS1 isoforms as potential biomarkers with the hypothesis that tumors that predominantly express KGA respond better to hormonal therapy and have better prognosis while those that predominantly express GAC respond poorly to hormonal therapy and have worse prognosis. This hypothesis will be tested in large, highly valuable patient tumor cohorts of hormone sensitive prostate cancer and CRPC. Additionally we will test if differential expression of the GLA isoforms is associated with the two histologic types of PCa, adenocarcinoma that is AR dependent and usually has a protracted disease course vs small cell neuroendocrine carcinoma which is AR-independent and rapidly lethal.