ID #669 Comparative Single-Cell RNAseq Analysis Resolves Distinct Immune Microenvironments Between Pediatric Brain Tumor Types

Abstract Background Over the past decade, single-cell RNAseq has charted the cellular heterogeneity in the tumor microenvironment of individual pediatric brain tumor diagnoses. These studies have shown immune subpopulations to be predominantly characterized by multiple myeloid lineage types, in addition to rarer lymphocytic lineage cells. In the present study, we sought to compare these tumor-infiltrating immune populations between different pediatric brain tumor types to identify diagnosis-specific immune microenvironments. To do this we integrated and reanalyzed our existing single-cell RNAseq data from over 100 clinical specimens comprising adamantinomatous craniopharyngioma (ACP), atypical teratoid/rhabdoid tumor (ATRT), ependymoma (EPN), high-grade glioma (HGG), low-grade glioma (LGG) and medulloblastoma (MB). Results Combined single-cell RNAseq data from 115 clinical specimens across the spectrum of pediatric brain tumors totaled 170,929 cells including 72,850 immune cells. Reclustering analysis of immune cells revealed major clusters of myeloid, T-, NK- and B-cells. Consensus non-negative matrix factorization clustering of myeloid cells identified 12 populations that were predominated by non-activated microglia and more differentiated dendritic cells (DCs), neutrophils, M2-polarized, hypoxia-polarized, complement-high and chemokine-high myeloid cells. The proportions of these cell types were significantly different by tumor type. For example, LGG and ACP showed the most disparate myeloid populations, with LGG harboring high proportions of unpolarized microglia and chemokine-high myeloid cells, whereas ACP have scant microglia but abundant DCs and mast cells. Lymphocyte populations were annotated using Azimuth (PBMC reference), revealing proportional enrichments and deficiencies of lymphocyte subtypes in pediatric brain tumors compared to their proportions in PBMC. Differential T-cell phenotypes were also observed between diagnoses, such as a significantly higher ratio of CD8 T central memory to T effector memory in LGG versus other diagnoses. Conclusions Integration of single-cell transcriptomic data from the commonest pediatric brain tumor diagnoses reveals significantly different immune microenvironments that impact tumor biology and immunotherapy strategies.