Genome-wide dna methylation patterns in vertu: a randomized phase ii trial of veliparib, radiotherapy and temozolomide in patients with mgmt-unmethylated glioblastoma

Abstract BACKGROUND VERTU was a randomized phase II trial evaluating veliparib, a brain-penetrant PARP inhibitor, combined with radiotherapy and temozolomide, for patients with newly diagnosed MGMT-unmethylated glioblastoma. As part of planned correlative work after study completion, we assessed genome-wide DNA methylation patterns to predict methylation class, glioblastoma subtype and MGMT status. METHODS Patients were randomized 2:1 to experimental (60Gy/30 fractions with veliparib 200mg bid, then temozolomide 150-200mg/m2 D1-5 + veliparib 40mg bid D1-7 Q28D for 6 cycles) versus standard arm (60Gy/30 fractions with temozolomide 75mg/m2 daily, then temozolomide 150-200mg/m2 D1-5 Q28D for 6 cycles). The primary objective to improve 6-month progression-free survival (PFS-6m) was not met (doi: 10.1093/neuonc/noab111). Methylation data were generated using the Illumina Infinium Methylation EPIC bead chip array. Tumor tissues were categorized using the Heidelberg methylation-based classifier. RESULTS Methylation data were successfully generated for 98/125 patients (poor quality DNA [n = 12], no consent [n = 11], insufficient tissue [n = 4]). Those with classifier scores below 0.5 (n = 25), tumor microenvironment only (n = 6) and rediagnosis as pleomorphic xanthoastrocytoma (n = 1) were excluded, leaving n = 66. Methylation classes were GBM RTK II (n = 23, PFS-6m 43% [95%CI 23-62]), RTK I (n = 20, PFS-6m 50% [95%CI 27-69]), MES (n = 20, PFS-6m 40% [95%CI 19-60]), MID (n = 2) and G34 (n = 1). Glioblastoma subtypes were mesenchymal (n = 28, PFS-6m 50% [95%CI 30-66]), proneural (n = 24, PFS-6m 50% [95%CI 29-68]) and classical (n = 14, PFS-6m 36% [95%CI 13-59]). MGMT status were unmethylated (n = 58, PFS-6m 48% [95%CI 35-60]) and methylated (n = 8, PFS-6m 38% [95%CI 9-67]). There was no evidence of interaction between treatment arm and methylation class (excluding GBM MID and G34, P = 0.45), glioblastoma subtype (P = 0.68) or MGMT status (P = 0.52). CONCLUSIONS Genome-wide DNA methylation patterns in VERTU identified a spectrum of methylation-defined subgroups, reflecting tumoral heterogeneity. This may have utility for future clinical trials and practice. The effect of veliparib in VERTU appeared to be consistent across subgroups. ACTRN12615000407594.