Abstract
Glioblastoma (GBM) is the deadliest primary brain tumor that frequently infiltrates surrounding brain tissue, causing therapy resistance and recurrence. The molecular characteristics of infiltrating GBM cells and their interactions with brain cells remain poorly understood, partly due to limited spatial tools for distinguishing tumor cells from those in the tumor microenvironment (TME). Here, we introduce Spatially-resolved Chimeric AnalyzeR (SCAR), a computational tool for dissecting tumor-TME gene expression in spatial transcriptomics of human-mouse chimeric cancer models. SCAR reveals spatially distinct characteristics of GBM and their interactions with TME, identifying that infiltrative GBM up-regulates creatine kinase brain type (CKB) at the astrocyte-enriched leading edge compared to the tumor core. Mechanistically, GBM-secreted CKB catalyzes the extracellular conversion of astrocyte-supplied adenosine triphosphate (ATP) and creatine into phosphocreatine, providing metabolic support for infiltrative tumor growth, which can be blocked by cyclocreatine. These results provide proof-of-concept validation of SCAR and demonstrate spatial context-dependent metabolic rewiring of GBM cells, with implications for therapies targeting infiltrative GBM.