Unsupervised clustering of brain tumor leukocytes to reveal atypical lymphoid cell subsets.

25 Background: The brain tumor immune microenvironment (TiME) is characterised by suppression of tumor-infiltrating leukocytes. The lymphocyte niche activity is particularly downregulated within this TiME milieu. Methods: Here, using data-driven approaches in flow cytometry and RNA sequencing analysis, we aimed at dissecting leukocyte cell types infiltrating human brain tumor TiME. Results: Unsupervised clustering provided T lymphocyte subsets including double-negative (CD3 + CD4 - CD8 - ) T cells (DNT). Our DNT phenotype was validated with an independent dataset, suggesting DNT to be compatible with γδ T cell phenotype. In our cohort, myeloid immune cell frequencies were associated with malignancy type as previously reported, while a gliosarcoma notably presented unexpectedly high lymphocyte frequencies. Consistently, double-positive (CD4 + CD8 + ) T cells (DPT) had higher frequency in the gliosarcoma than other tumors. The flow cytometry results were supported by transcriptome patterns and the activity of immune-related genes in these tumors. Conclusions: Our findings feature the benefit of data-driven approaches for brain TiME lymphocyte analysis, and show that primary brain tumors can entail atypical lymphocyte subsets like DNTs and DPT cells. Unbiased lymphocyte characterisation can potentially outline novel targets for immune checkpoint blockade therapy.