识别胶质母细胞瘤浸润边缘细胞的分子特征作为肿瘤侵袭和复发的驱动因素

Glioblastoma (GBM) is the most common malignant brain tumor in adults. Despite extensive research, there haven't been remarkable gains in resolving the seeds of GBM recurrence, and the outcomes for many patients suffering from this devastating disease remain poor. Our knowledge on GBM heterogeneity is mostly restricted to the surgically resectable tumor core, while functional characterization of tumor cells at the infiltrating edge remains largely elusive due to the presence of normal functional brain tissue in the peritumoral lesion. Edge-derived cells exhibit larger capacity for infiltrative expansion and are the main drivers of treatment failure and tumor recurrence, making them action targets for novel treatment approaches.In this study, we present a first-of-its-kind integrative spatial investigation of GBM, combining two complementary spatial omics modalities high-definition spatial transcriptomics (ST - Visium HD) and spatial proteomics (SP - COMET) to achieve a comprehensive morphological, transcriptomic, and proteomic characterization of invasive tumor edge in situ. This multimodal spatial framework enabled to resolve the complex molecular landscape of the GBM periphery and to identify druggable biomarkers specific to edge-derived malignant cell populations.By integrating pathologically annotated H&E images with high-resolution spatial gene expression, we delineated patterns of tissue architecture and captured continuous gradients of gene expression across tumor-brain interface. Using NicheCompass, we dissected tissue hierarchies and spatially localized cellular processes within the infiltrative compartment of GBM. Through unsupervised phenotyping, we identified top spatially variable genes and active gene programs, revealing modules indicative of tumor cell hijacking of neuronal pathways. These findings align with previously described mechanisms of glioma-neuron synaptic coupling and formation of neurite-like tumor microtubes, consistent with enrichment of OPC- and NPC-like cellular states at tumor margin.Complementing the transcriptomic layer, SP was used to guide single-cell segmentation for ST analysis, infer cell types based on canonical phenotypic markers, and characterize cell state and function through protein-level profiling. This dual-modality approach allowed for precise spatial mapping of highly invasive edge cell populations and their functional states.By integrating multiple spatial omic layers, our study provides an unprecedented multimodal view of GBM invasion and identifies novel, spatially defined biomarkers that distinguish malignant edge-derived cells. These findings hold potential translational value as diagnostic and prognostic tools, enabling early assessment of treatment response and facilitating personalized therapeutic strategies aimed at mitigating GBM progression and recurrence.