EPCO-25. AN IMMUNOMETHYLOMIC PLATFORM INTEGRATING SYSTEMIC IMMUNE PROFILES AND EPIGENETIC AGE IN NEURO-ONCOLOGY. (9th November 2020)

• Record Type: Journal Article
• Authors: Molinaro, Annette; Wiencke, John; Warrier, Gayathri; Lee, JiYoon; Koestler, Devin; Hansen, Helen; Rice, Terri; Bracci, Paige; McCoy, Lucie; Salas, Lucas; Christensen, Brock; Wrensch, Margaret; Kelsey, Karl; Taylor, Jennie; Clarke, Jennifer
• Is Part Of: Neuro-oncology. Volume 22(2020)Supplement 2
• Page Start: ii74
• Page End: ii74
• Publication Date: 2020-11-09
• Subjects: Brain Neoplasms -- Periodicals; Brain -- Tumors -- Periodicals; Brain -- Cancer -- Periodicals; Nervous system -- Cancer -- Periodicals; 616.99481
• DOI: 10.1093/neuonc/noaa215.304
• Languages: English
• ISSNs: 1522-8517

Abstract

Abstract: Lineage-specific DNA methylation marks differentiate leukocyte cell types while individual biological aging mechanisms impact other methylation alterations. Human glioma incidence and survival times have been shown to be associated with aberrant immune profiles and have a strong dependency on age. Here we developed a single epigenetic analysis framework to evaluate both immune cell fractions and epigenetic age in peripheral blood. We examined these measures in archived blood from 197 triple-negative glioma patients (TNG; IDH wildtype, 1p19q intact and TERT wildtype) and 312 frequency-matched controls from the SF Bay Area Adult Glioma Study (AGS). Significant differences were observed with TNG cases having lower CD4 and CD8 T cell, natural killer, and B cell fractions, and higher neutrophil fractions than controls. TNG cases were significantly older than controls in two of three epigenetic age estimates; however, there was no difference in epigenetic age acceleration once immune cell proportions were considered. For the TNG cases, we augmented results from several machine learning methods to delineate risk groups of TNG patients with significantly different overall survival. We compared survival models built by recursive partitioning, random forest, and elastic net methods. The final model was chosen by repeated bootstrap sampling via the Brier score loss function and validated in an independent set of 72 IDH -mutant only or TERT -mutant only glioma patients also from the AGS. The final model indicated important interactions between immune cell fractions (including CD4 and CD8 T cells and neutrophils) and treatment, age, and dexamethasone status when adjusted for the main effects of epigenetic age, glioblastoma status, and the neutrophil-to-lymphocyte ratio. The capacity of immunomethylomics to capture diverse, clinically relevant information and the simplicity of its implementation make this a powerful tool for personalized patient evaluation in the neuro-oncology clinic.