Unsupervised clustering reveals noncanonical myeloid cell subsets in the brain tumor microenvironment.

The tumor immune microenvironment (TiME) of human central nervous system (CNS) tumors remains to be comprehensively deciphered. Here, we employed flow cytometry and RNA sequencing analysis for a deep data-driven dissection of a diverse TiME and to uncover noncanonical immune cell types in human CNS tumors by using seven tumors from five patients. Myeloid subsets comprised classical microglia, monocyte-derived macrophages, neutrophils, and two noncanonical myeloid subsets: CD3 myeloids and CD19 myeloids. T lymphocyte subsets included double-negative (CD4 CD8) T cells (DNTs). Noncanonical myeloids and DNTs were explored on independent datasets, suggesting that our DNT phenotype represents γδ T cells. Noncanonical myeloids were validated using orthogonal methods across 73 patients from three independent datasets. While the proportions of classical myeloids agreed with reported malignancy type-associated TiMEs, unexpectedly high lymphocyte frequencies were detected in gliosarcoma, which also showed a unique expression pattern of immune-related genes. Our findings highlight the potential of data-driven approaches in resolving CNS TiME to reveal the mosaic of immune cell types constituting TiME, warranting the need for future studies on the nonclassical immune cell subsets.