单细胞层面阐明 IDH 突变型胶质瘤中分子上不同的状态及治疗易感性

IDH-mutant gliomas, including oligodendrogliomas (IDH-O) and astrocytomas (IDH-A), are a molecularly defined class of fatal primary brain tumors characterized by mutations in the IDH1 and IDH2 genes. Despite advances in standard-of-care approaches, prognosis remains poor with median survival rates of 5-15 years. Substantial inter- and intra-tumor heterogeneity limits the efficacy of current monotherapies and necessitates ad hoc combination strategies targeting distinct tumor subpopulations. To comprehensively characterize the cellular landscape of IDH-mutant gliomas, we generated >250k high-quality single-nucleus transcriptomic profiles from 20 IDH-mutant glioma tumors (grade II and grade III; 10 IDH-O; 10 IDH-A) obtained from the Molecular Pathology Shared Resource (MPSR) Tumor Bank at Columbia University. Gene expression analysis revealed glioma cells, microglia, neurons and mature oligodendrocytes as the predominant cell types in both IDH-O and IDH-A, with a striking depletion of microglia in IDH-O. Network-based VIPER analysis of single-nucleus profiles identified Master Regulator (MR) proteins representing molecular dependencies of three previously described glioma states: astrocyte-like (AC), oligodendrocyte-like (OC) and neural-progenitor-like (NPC). Notably, the AC state was transcriptionally orthogonal to OC, whereas OC and NPC largely shared their regulatory architecture. To predict rational combination therapy candidates, we employed NYS CLIA-certified OncoTarget and OncoTreat algorithms. OncoTarget identifies small molecule inhibitors targeting individual state-specific MRs, while OncoTreat predicts candidate drugs by assessing their ability to invert the activity of glioma-specific MRs, leveraging large-scale drug perturbation assays generated by PLATE-seq. Specifically, we generated a library of genome-wide RNA-seq profiles from patient-matched in vitro models, one adherent cell line (SF10417) and one neurosphere (SUA03), 24 hours after treatment with 374 compounds from a library of FDA approved and investigational compounds. OncoTarget uncovered distinct pathway dependencies across glioma states, including STAT3/PI3K/AKT in AC cells, PDGFRA in OC, and RTK/EGFR/MET signaling in NPC cells. OncoTreat identified CNS-permeable agents predicted to invert glioma-specific MR programs. Future work will include validation of candidate compounds in patient-derived acute slice cultures to assess single-cell responses within an intact microenvironment. Complementary Xenium spatial profiling will map the microenvironmental context and cell-cell communication of molecularly distinct subpopulations, informing rational therapeutic strategies. Overall, our study establishes a generalizable framework for precision oncology in transcriptionally complex tumors and provides actionable targets for clinical translation.