Can variant allelic frequencies of driver mutations predict false-negative MGMT pyrosequencing results in adult-type diffuse gliomas?

Abstract MGMT promoter methylation predicts favorable temozolomide (TMZ) response in gliomas. Accurate test results depend on adequate tumor cellularity in analyzed samples, which is usually estimated (with sub-optimal accuracy) via light microscopy. We evaluated driver mutation variant allelic frequency (VAF) as a quantitative metric for tumor cellularity in MGMT promoter methylation testing. In a cohort of 691 adult-type diffuse gliomas, MGMT promoter methylation was tested by either pyrosequencing (N = 445) or DNA methylation array (N = 246). VAF of TERT and IDH driver mutations was quantified by next generation sequencing (NGS). We compared frequency of methylated (positive) and unmethylated (negative) results (Fisher’s exact test) and promoter methylation levels (unpaired t-test) according to VAF. We correlated MGMT promoter methylation, VAF, visually estimated tumor cellularity, and cellularity inferred from VAF via linear regression. We assessed survival (Log-Rank test) for 201 patients with IDH-wildtype GBM treated with TMZ. In samples with VAF <0.325 (identified by CutoffFinder), pyrosequencing yielded fewer positive results (37% versus 56%, p < 0.0001) and lower methylation levels (13% versus 19%, p < 0.0001) than samples with VAF≥0.325. Methylation array showed less frequent positive results at low VAF for IDH-mutant astrocytoma, but not for IDH-wildtype GBM. Microscopic examination tended to overestimate tumor cellularity, especially when VAF was low (R2 = 0.558). GBMs with unmethylated MGMT by pyrosequencing and TERT VAF <0.145 showed better responses to TMZ versus cases with VAF≥0.145 (median survival = 16.8 versus 13.3 months). Twelve pyrosequencing samples were re-tested by DNA methylation array and droplet digital PCR (ddPCR). RESULTS: changed for 2/6 samples with TERT VAF≤0.1, and 0/6 samples with VAF >0.1. Driver mutation VAF is useful for quality assurance in MGMT promoter methylation testing. Tumor samples with low VAF are at greater risk of false-negative results by pyrosequencing, which can be corrected by DNA methylation array and ddPCR.