Androgen receptor nuclear localization correlates with AR-V7 mRNA expression in circulating tumor cells (CTCs) from metastatic castration resistance prostate cancer patients.

Androgen receptor (AR) signaling drives prostate cancer (PC) progression and remains active upon transition to castration resistant prostate cancer (CRPC). Active AR signaling is achieved through the nuclear accumulation of AR following ligand binding and through expression of ligand-independent, constitutively active AR splice variants, such as AR-V7, which is the most commonly expressed variant in metastatic CRPC (mCRPC) patients. Most currently approved PC therapies aim to abrogate AR signaling and activity by inhibiting this ligand-mediated nuclear translocation. In a prospective multi-institutional clinical study, we recently showed that taxane based chemotherapy is also capable of impairing AR nuclear localization (ARNL) in circulating tumor cells (CTCs) from CRPC patients, whereas taxane induced decreases in ARNL were associated with response. Thus, quantitative assessment of ARNL in CTCs can be used to monitor therapeutic response in patients and help guide clinical decisions. Here, we describe the development and implementation of quantitative high throughput (QHT) image analysis algorithms to aid in CTC identification and quantitative assessment of percent ARNL (%ARNL). We applied this algorithm to fifteen CRPC patients at the start of taxane chemotherapy, quantified %ARNL in CTCs, and correlated with expression of AR-V7 mRNA (from CTCs enriched via negative, CD45+ depletion of peripheral blood) and with biochemical (prostate specific antigen; PSA) response to taxane chemotherapy. We found that CTCs from AR-V7 positive patients had higher baseline %ARNL compared to CTCs from AR-V7 negative patients, consistent with the constitutive nuclear localization of AR-V7. In addition, lower %ARNL in CTCs at baseline was associated with biochemical response to taxane chemotherapy. High inter- and intra-patient heterogeneity was also observed. As ARNL is required for active AR signaling, the QHT algorithms described herein can provide prognostic and/or predictive value in future clinical studies.