P02.05.A COMPREHENSIVE ANALYSIS OF TUMOR-ASSOCIATED MACROPHAGES IN GLIOBLASTOMA

Abstract BACKGROUND Glioblastoma (GBM) is the most common and extremely aggressive primary tumor of the central nervous system. Tumor-associated macrophages (TAM) can take up around 50% of cells in the tumor microenvironment (TME) of GBM and are thought to play a role in therapy resistance by creating an immunosuppressive TME. Researchers have used the M1/M2 dichotomy to describe the pro- and anti-inflammatory states of TAM, respectively. However, an expanding number of studies have found that this classification strategy oversimplifies the nature of TAM as seen in patients. Therefore, a more comprehensive strategy to represent the continuous spectrum of cellular states of TAM is essential. In this study, we aim to identify the cellular states of TAM, and further explore their impact on clinical outcomes and the TME. MATERIAL AND METHODS A bioinformatic pipeline based on principal component analysis was developed to identify the cellular states of TAM, which addressed the high inter-tumoral heterogeneity of GBM. A public Image Mass Cytometry dataset is used to verify the existence of the TAM states. The correlation analysis and survival analysis were conducted to identify the clinical relevance of the TAM states. Three spatial transcriptomic datasets were used to explore the role of these TAM states in the TME. A vessel/neuron impact index was developed to describe the impact of vessel/neuron on the TME. An 8-color multiplex immunohistochemistry is currently being developed to explore the role of VIM-TAM in the TME with spatial analysis in our cohort with 74 GBM patients. RESULTS Four public GBM scRNA-seq datasets with 94 samples were input into our pipeline. Four TAM states together with a proliferating state were identified: HLA-TAM (HLA-DR, SPP1, APOE, TREM2, C3, CD74), VIM-TAM (VIM, S100A4/6/10, CD44), IL1B-TAM (IL1B, CCL3, CD83), and HSP-TAM (HSP90AA1, BAG3). By mapping TAMs into a 3D plot based on these state scores, we revealed a continuous spectrum of TAM. Prognosis analysis found that the mean IL1B-TAM score showed a protective effect for OS (P = 0.065), while the mean VIM-TAM score had a risk impact on PFS (P = 0.045). Spatial transcriptomic analysis showed that the VIM-TAM is negatively correlated with neural-progenitor-like cancer cells and the neuron impact index. CONCLUSION A continuous spectrum of TAM could be described by these 4 identified TAM states. The IL1B-TAM and the VIM-TAM are predictors for OS and PFS, respectively. Also, the VIM-TAM is related to NPC-like cancer cells and neurons in the TME.