P04.13.B SPATIAL HETEROGENEITY OF CAR-T ANTIGEN EXPRESSION IN GLIOBLASTOMA

Abstract BACKGROUND Chimeric antigen receptor T-cell (CAR-T) therapy has emerged as a breakthrough immunotherapy for multiple cancers, yet its efficacy remains limited in glioblastoma due to tumor heterogeneity, immune-evasion of the microenvironment, and rapid T-cell dysfunction. The aim of our study is to investigate the spatially differential expression of CAR-T targets in glioblastoma. We want to focus on differential expression of CAR-T targets in tumor cell subtypes, differences between tumor core and infiltration zone as well as reactive immune zones. MATERIAL AND METHODS We analyzed spatial transcriptomic data from 62 glioblastoma specimens using SPATA2 software, augmented by advanced computational tools such as scVI (data integration), Cell2Location (single-cell deconvolution), graph neural networks, gene set enrichment analysis (GSEA), and spatial annotation screening (SAS). CAR-T antigens were chosen from existing published clinical and pre-clinical CAR-T cell trials. RESULTS Spatial mapping revealed a pronounced gradient of NPC-like cells with a higher abundance of NPC-like cells in the infiltration zone and resection margins—regions critical for residual disease persistence after resection. Clinical CAR-T targets (e.g., ERBB2, PTPRZ1) were highly expressed in AC-like and OPC-like dominated niches while nearly absent in NPC-rich areas, suggesting a blind spot for current CAR-T therapies. Single-cell deconvolution further correlated NPC-like dominance to increased T-cell infiltration, while GSEA and clonal dynamics analysis of the T-cell compartment suggested downregulated functioning subclones in NPC-rich spatial neighborhoods, underscoring potential immune evasion mechanisms in these regions. CONCLUSION Our study exposes a fundamental limitation of current CAR-T designs: their inefficacy against NPC-like cells that dominate the infiltration zone and therefore the post-surgical glioblastoma ecosystem. By pinpointing spatial and phenotypic resistance hubs, we provide a roadmap for refining CAR-T targets to address this gap, potentially enhancing adjuvant immunotherapy for glioblastoma. Protein-level validation of key findings is currently ongoing to further substantiate the results.