TMIC-60. Spatial profiling of IDH-wildtype glioblastoma using CosMx Spatial Molecular Imaging technology to establish a large-scale FFPE tissue cohort

Abstract BACKGROUND Glioblastoma, particularly IDH-wildtype, represents a highly heterogeneous r with limited spatial transcriptomic studies conducted on large-scale cohorts. Challenges in cell typing in formalin-fixed, paraffin-embedded (FFPE) tissues, compounded by tumor heterogeneity and reference matrix constraints, have impeded progress. Addressing these gaps, we present one of the largest FFPE-based glioblastoma tissue microarray (TMA) cohorts to date, comprising 21 WHO Grade 4 IDH-wildtype cases, profiled using CosMx Spatial Molecular Imaging NanoString technology. METHOD FFPE tumor specimens collected between 2018 and 2022 were subjected to RNA quality control before TMA construction. Spatial profiling was performed using the CosMx platform with the 6K Human Discovery Panel. Raw data were processed semi-supervised using AtoMx Spatial Informatics Platform (v2.0). Cell typing was optimized using three reference matrices: NanoString GBM Cell Profile Library, Core GBMap, and an in-house single-nucleus RNA sequencing (snRNA-seq)-derived profile. RESULTS A total of 449 fields-of-view across 21 cases were captured, profiling 6,175 genes and segmenting 337,622 cells using DNA, membrane, Pan-CK, and CD45 markers. The mean transcript count per cell was 312 (10th percentile: 25; 90th percentile: 727). No significant batch effects were observed by PCA and UMAP. Cell typing with the NanoString library offered limited annotation granularity, broadly categorizing malignant cells as “Neoplastic 1–3” and leaving many cells unannotated. GBMap and in-house snRNA-seq profiles enabled finer resolution of malignant subtypes (AC-like, MES-like, OPC-like, NPC-like) and immune lineages (CD4+/CD8+ T cells, B cells, TAM-BDM, TAM-MG, monocytes), reflecting the tumor heterogeneity. CONCLUSION This large-scale spatial transcriptomic dataset using FFPE tissues enables high-resolution profiling of malignant subtypes and immune lineages. Its scalability and compatibility with archived specimens provide a robust platform for advanced tumor microenvironment studies, including cell–cell interactions and ligand-receptor modeling, augmenting glioblastoma research.