Background Brain immunity is largely myeloid cell dominated rather than lymphoid cells in healthy and diseased state including malignancies of glial origins called as gliomas. Despite this skewed myeloid centric immune contexture, immune checkpoint and T cell based therapeutic modalities are generalizably pursued in gliomas ignoring the following facts i) T cells are sparse in tumor brain ii) glioma patients are lymphopenic iii) gliomas harbor abundant and highly complex myeloid cell repertoire. We recognized these paradoxes pertaining to fundamental understanding of constituent immune cells and their functional states in the tumor immune microenvironment (TIME) of gliomas, which remains elusive beyond a priori cell types and/or states.

Methods To dissect the TIME in gliomas, we performed single-cell RNA-sequencing on ~123,000 tumor-derived sorted CD45+ leukocytes from fifteen genomically classified patients comprising IDH-mutant primary (IMP; n=4), IDH-mutant recurrent (IMR; n=4), IDH-wild type primary (IWP; n=3), or IDH-wild type recurrent (IWR; n=4) gliomas (hereafter referred as glioma subtypes) and two non-glioma brains (NGBs) as controls.

Results Unsupervised clustering analyses delineated predominant 34-myeloid cell clusters (~75%) over 28-lymphoid cell clusters (~25%) reflecting enormous heterogeneity within and across glioma subtypes. The glioma immune diversity spanned functionally imprinted phagocytic, antigen-presenting, hypoxia, angiogenesis and, tumoricidal myeloid to classical cytotoxic lymphoid subpopulations. Specifically, IDH-mutant gliomas were predominantly enriched for brain-resident microglial subpopulations in contrast to enriched bone barrow-derived infiltrates in IDH-wild type especially in a recurrent setting. Microglia attrition in IWP and IWR gliomas were concomitant with invading monocyte-derived cells with semblance to dendritic cell and macrophage like transcriptomic features. Additionally, microglial functional diversification was noted with disease severity and mostly converged to inflammatory states in IWR gliomas. Beyond dendritic cells, multiple antigen-presenting cellular states expanded with glioma severity especially in IWP and IWR gliomas. Furthermore, we noted differential microglia and dendritic cell inherent antigen presentation axis viz, osteopontin, and classical HLAs in IDH subtypes and, glioma-wide non-PD1 checkpoints associations in T cells like Galectin9 and Tim-3. As a general utility, our immune cell deconvolution approach with single-cell-matched bulk RNA sequencing data faithfully resolved 58-cell states which provides glioma specific immune reference for digital cytometry application to genomics datasets.

Conclusions Altogether, we identified prognosticator immune cell-signatures from TCGA cohorts as one of many potential immune responsiveness applications of the curated signatures for basic and translational immune-genomics efforts. Thus, we not only provide an unprecedented insight of glioma TIME but also present an immune data resource that can be exploited for immunotherapy applications.

Ethics Approval The brain tumor/tissue samples were collected as per MD Anderson internal review board (IRB)-approved protocol numbers LAB03-0687 and, LAB04-0001. One non-tumor brain tissue sample was collected from patient undergoing neurosurgery for epilepsy as per Baylor College of Medicine IRB-approved protocol number H-13798. All experiments were compliant with the review board of MD Anderson Cancer Center, USA.

Consent Written informed consent was obtained from the patient for publication of this abstract and any accompanying images. A copy of the written consent is available for review by the Editor of this journal