Abstract 104: Profiling peripheral blood to predict response to targeted androgen receptor axis therapy in mice

Abstract Background: Targeting the androgen receptor (AR) signaling axis (ARAT) continues to be the mainstay treatment for advanced-stage prostate cancer. Using second-generation antiandrogens like abiraterone acetate (Abi) and apalutamide (Apa) has contributed to the survival of prostate cancer patients. However, resistance to these therapies remains a problem and most of these patients progress to castration-resistant prostate cancer (CRPC). Myeloid-derived suppressor cells (MDSCs) have been shown to drive CRPC. We previously showed that unfavorable responses to androgen deprivation therapy (ADT) or ADT plus Apa were associated with polymorphonuclear MDSCs (PMN-MDSCs) cells enriched in prostate tumors from conditional Pten/Trp53-double knockout (DKO) mice. Here, we explore the association between pre-treatment immune cell subsets blood and antitumor to ARAT therapy DKO mice. Methods: Twenty-eight-week-old DKO mice were surgically castrated and after four weeks were randomized to treatments with vehicle, Apa or Abi for an additional four weeks (n=20). Individual treatment responses were assessed and classified according to tumor burden as follows: high tumor burden (HTB), ≥20% of median; moderate tumor burden (MTB), -20% to 20% of median; or low tumor burden (LTB), ≤-20% of median. Peripheral blood was collected both before and after treatment and analyzed by multi-color flow cytometry panels for lymphoid and myeloid subsets. Blood samples from healthy wildtype and untreated DKO mice with castration-naïve prostate tumors were collected as reference samples. Subsets of PMN-MDSCs and monocytic-MDCSs (m-MDSCs) were represented as % of total abundance and 5 of parental population. revealed a signature that was correlated with treatment responses. Profiles were constructed and the data was used to train supervised learning models to predict treatment responses. Shapely additive explanations (ShAP) were used to determine feature importance. The model was tested and validated on an independent cohort of 20 DKO mice treated with ADT and ADT plus Abi. Results: Moderate prediction from baseline blood could be identified for tumor burden following ARAT using a Naïve Bayes classifier. Model evaluation, using an independent cohort, yielded favorable performance with area under the curve (AUC) mean, 0.707; HTB, 0.627; MTB, 0.750; and LTB, 0.745. The most important features for prediction were identified. Conclusion: Our study revealed that MDSC subsets present in blood were associated with treatment responses to ARAT. Citation Format: Marco A. De Velasco, Yurie Kura, Kazuko Sakai, Yoshitaka Saito, Takafumi Minami, Kazuhiro Yoshimura, Kazutoshi Fujita, Kazuto Nishio, Hirotsugu Uemura. Profiling peripheral blood to predict response to targeted androgen receptor axis therapy in mice [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 104.