Androgen Receptor-Immune Crosstalk in Metastatic Castration-Resistant Prostate Cancer: From Context-Dependent Remodeling to Biomarker-Guided Translation.

BACKGROUND: Metastatic castration-resistant prostate cancer (mCRPC) rarely responds to immune checkpoint blockade despite clear evidence that androgen receptor (AR) signaling intersects with tumor immunity. The view that androgen deprivation uniformly enhances antitumor immunity does not explain the repeated failure of immunotherapy combinations in biomarker-unselected populations. METHODS: This narrative review synthesizes mechanistic, translational, and clinical evidence and organizes it by cellular compartment, treatment-associated immune state, clinical trial strategy, and biomarker readiness. RESULTS: AR-directed therapy produces context-dependent, compartment-specific immune remodeling. It can alter tumor antigen visibility, thymic output, and T-cell trafficking, while myeloid cytokines and metabolites, cancer-associated fibroblasts, and metastatic niches sustain AR reactivation, T-cell dysfunction, and immune exclusion. These effects are better understood as heterogeneous tumor-immune states than as a fixed treatment window. Randomized phase III trials combining checkpoint blockade with AR pathway inhibitors, PARP inhibitors, or chemotherapy have not improved overall survival in unselected mCRPC, and dual-checkpoint blockade has shown limited activity with substantial toxicity. Deficient mismatch repair or microsatellite instability-high (dMMR/MSI-H) disease is currently the only clinically validated enrichment factor for checkpoint blockade in mCRPC, although this subgroup is uncommon. Composite signatures remain investigational. CDK12, homologous-recombination repair alterations, isolated tumor mutational burden-high disease, PD-L1, and AR-V7 are not independently validated selection markers. T-cell engagers have produced early PSA and radiographic responses by providing a distinct route to tumor-T-cell contact. Their development must balance this activity against cytokine-release syndrome, antigen heterogeneity, variable therapeutic windows, and the lack of comparative evidence. CONCLUSIONS: Progress requires trials that define the relevant tumor-immune state, match treatment to the dominant resistance mechanism, and prospectively test biomarker-treatment interaction. In current practice, dMMR/MSI-H remains the only validated enrichment factor for checkpoint blockade in mCRPC; all other proposed markers and T-cell engager strategies require prospective validation of treatment interaction.