BIOM-66. GENOME-WIDE DNA METHYLATION PATTERNS IN PERIPHERAL BLOOD OF PATIENTS WITH GLIOBLASTOMA ENROLLED ON THE PHASE II VERTU TRIAL

Abstract BACKGROUND For patients with glioblastoma, the identification of a non-invasive biomarker for longitudinal molecular profiling could improve risk stratification, clinical trial design, and surveillance. In a secondary analysis of the phase II randomized VERTU trial (ACTRN12615000407594) of adult patients with newly diagnosed MGMT-unmethylated glioblastoma, we assessed the feasibility of characterizing peripheral blood genome-wide DNA methylation patterns. METHODS DNA was extracted from whole blood and serum samples at baseline and end-of-treatment (EOT) from patients enrolled on VERTU, which evaluated the PARP inhibitor veliparib combined with radiotherapy and temozolomide. Infinium FFPE QC qPCR and DNA Restoration Kits were used to restore and concentrate fragmented serum DNA. Genomic DNA was plated, bisulfite converted, and profiled for methylation patterns at >935,000 CpG sites using the Infinium MethylationEPIC microarray V2.0. Quality control, pre-processing, leukocyte deconvolution, and calling of MGMT status were completed in R Studio. RESULTS Peripheral blood methylation data were generated and passed quality control for 104/104 baseline and 24/34 EOT samples available for analysis. The median concentration of genomic DNA at baseline and EOT was 104.50 ng/μl (range:1.97-550) and 0.98 ng/μl (range:0.11-19.8). The deconvoluted median leukocyte proportions differed significantly between baseline and EOT for neutrophils (0.64vs0.90,p<0.01), CD4 memory (0.09vs0.00,p<0.01), CD4 naive (0.03vs0.01,p=0.02), CD8 memory (0.04vs0.00,p<0.01), monocyte (0.07vs0.03,p<0.01), natural killer (0.05vs0.00,p<0.01), and T regulatory (0.000vs0.001,p=0.01) cells. There were no differences in baseline leukocyte proportions between patients with (n=6) and without (n=98) biopsy-proven pseudoprogression. 2/104 samples were determined to be MGMT-methylated based on peripheral blood methylation analysis. Unsupervised hierarchical clustering based on the 1,000 most differentially methylated CpG sites generated two patient clusters (n=64, n=40). CONCLUSIONS To our knowledge, this is the first study to demonstrate the feasibility of analyzing peripheral blood methylation patterns from patients with glioblastoma receiving chemoradiation on a randomized clinical trial. Efforts are underway to correlate these blood methylation patterns with previously published tumor tissue methylation patterns and clinical outcomes.