Large-scale bulk and single-cell RNA sequencing combined with machine learning reveals glioblastoma-associated neutrophil heterogeneity and establishes a VEGFA+ neutrophil prognostic model

Neutrophils play a key role in the tumor microenvironment (TME); however, their functions in glioblastoma (GBM) are overlooked and insufficiently studied. A detailed analysis of GBM-associated neutrophil (GBMAN) subpopulations may offer new insights and opportunities for GBM immunotherapy. We analyzed single-cell RNA sequencing (scRNA-seq) data from 127 isocitrate dehydrogenase (IDH) wild-type GBM samples to characterize the GBMAN subgroups, emphasizing developmental trajectories, cellular communication, and transcriptional networks. We implemented 117 machine learning combinations to develop a novel risk model and compared its performance to existing glioma models. Furthermore, we assessed the biological and molecular features of the GBMAN subgroups in patients. From integrated large-scale scRNA-seq data (498,747 cells), we identified 5,032 neutrophils and classified them into four distinct subtypes. VEGFA+GBMAN exhibited reduced inflammatory response characteristics and a tendency to interact with stromal cells. Furthermore, these subpopulations exhibited significant differences in transcriptional regulation. We also developed a risk model termed the "VEGFA+neutrophil-related signature" (VNRS) using machine learning methods. The VNRS model showed higher accuracy than previously published risk models and was an independent prognostic factor. Additionally, we observed significant differences in immunotherapy responses, TME interactions, and chemotherapy efficacy between high-risk and low-risk VNRS score groups. Our study highlights the critical role of neutrophils in the TME of GBM, allowing for a better understanding of the composition and characteristics of GBMAN. The developed VNRS model serves as an effective tool for evaluating the risk and guiding clinical treatment strategies for GBM. Not applicable.