Assessment of EBV DNA methlyation to guide antiviral use in EBV-associated lymphoma

Abstract

Abstract Intro/Objective: Epstein-Barr Virus (EBV), an oncogenic herpesvirus, can drive human lymphoma development. EBV uses DNA methylation to silence its genome, switching from a lytic (active) to a latent (dormant) state, and it's widely understood that EBV remains latent in lymphoma. Current diagnostics provide no information on EBV activation state, though it could be useful for clinicians to pick appropriate antiviral (lytic) or lymphoma-directed (latent) therapy. The antiviral ganciclovir (GCV) is effective specifically against lytic EBV due to viral BGLF4 expression. Our group discovered BGLF4 expression in latent EBV+CNS lymphomas, which was associated with favorable outcomes to a GCV-containing treatment regimen (Dugan et al, Clin Ca Res, 2018). BGLF4 expression encoding potential for GCV response in other types of lymphomas remains unknown. We sought to first evaluate the landscape of BGLF4 methylation across EBV+lymphoma subtypes. We hypothesized that methylation loss at BGLF4 is associated with GCV response in EBV+lymphoma types, potentially serving as a high throughput and inexpensive cell-free DNA (cfDNA) biomarker to identify GCV candidates. Methods: We established a high-throughput PCR mass spectrometry assay quantifying methylation of n=12 CpG sites at the BGLF4 promoter (iPLEX). The assay was validated against the EpiTYPER methylation assay (r^2=0.766), methylation standards, and the limit of detection was determined. We assessed n=162 EBV+patient plasma samples from a cross-sectional study cohort including EBV-reactivations, post-transplant lymphoproliferation (PTLD), B-, T- and NK-cell lymphomas. Transcriptional activity, 5'RACE, and BGLF4 DNA methylation data were overlaid to determine methylation and expression association. Retrospective chart review was performed to determine GCV use and response association with BGLF4 methylation. Inclusion criteria for GCV response assessment was a negative 0.5log(10) change in viral load, lack of recent or current chemotherapy, rituximab, reduction in immunosuppression (RIS), and un-interrupted GCV use. Results: Our cohort consisted of n=23 NK/T cell lymphoma, n=15 Hodgkin lymphoma, n=39 PTLD, n=25 diffuse large B cell lymphoma (DLBCL), n=21 EBV viremia, and n=39 “other” samples, which were excluded from analysis. The average age was 55 years old (yo), median age was 59 yo (range: 20-86 yo). The iPLEX assay can be run on cfDNA from plasma in 10-12 hours. We identified DNA methylation loss at single CpG nucleotides correlated with BGLF4 expression (r2=0.73), demonstrating methylation can be used as a surrogate for gene expression. Unsupervised clustering of average methylation separated samples into 2 methylation patterns: group 1, of intermediate methylation, that contains a high methylation subgroup (subgroup 1a), and group 2, of low methylation. Cluster assignment was significantly associated with disease entity (p=0.0278), and post-hoc analysis showed that PTLDs were underrepresented in subgroup 1a (z=-2.89), and overrepresented in group 2 (z=2.49). NK/T cell lymphomas were underrepresented in group 2 (z=-2.20). We identified n=4 CpG sites within BLGF4 that were heterogeneously methylated among samples, deemed the BGLF4 core promoter. Strikingly, 42% of samples had ≤50% average core promoter methylation, demonstrating hypomethylation across disease groups. After applying inclusion criteria to assess GCV response, preliminary analysis showed that patients who responded to GCV (n=6) had a significantly lower BGLF4 methylation (p=0.0012) than patients who did not respond to GCV (n=7). N=5 GCV responders belong to group 2 designated by unsupervised clustering, and all GCV non-responders belong to group 1, presenting a methodology for stratifying out potential responders. These results suggest: (1) the novel finding that EBV exhibits site specific hypomethylation across multiple lymphoma types that correlates with BGLF4 gene expression; and (2) BGLF4 demethylation is clinically superior to viral load because it adds context to the epigenetic and histologic state, predicting antiviral sensitivity. Conclusions: We identified site specific BGLF4 methylation loss to be widespread in EBV+lymphoma, challenging the current dogma of EBV latency in lymphoma. Our BGLF4 demethylation assay is high throughput, inexpensive, and proves clinically superior to viral load alone, serving as a potential biomarker to successfully guide GCV administration in EBV-driven lymphomas.