5-Hydroxymethylcytosines of Circulating Cell-Free DNA and Prognosis in Diffuse Large B-Cell Lymphoma

Background: Elevated levels of circulating cell-free DNA (cfDNA) have been associated with poor prognosis and relapse in patients with diffuse large B-cell lymphoma (DLBCL). However, the tumor-specific molecular targets in cfDNA that have prognostic significance remain unclear. We investigated the association between 5-hydroxymethylcytosine (5hmC), a mark of active demethylation and gene activation, in cfDNA from plasma and prognosis in DLBCL.

Methods: We used the 5hmC-Seal, a highly sensitive chemical labeling technique integrated with next-generation sequencing (NGS), to profile 5hmC in plasma cfDNA samples from Caucasian patients at the University of Chicago who were newly diagnosed with DLBCL (n=43) or follicular lymphoma (FL, n=28), the two most common histological subtypes of non-Hodgkin lymphoma (NHL), in 2011-13. Baseline clinical, laboratory, and vital status data were abstracted from medical records. Patients were followed through December 31, 2017 and those who relapsed after completion of treatment, lost to follow-up, or died were censored. We profiled 5hmC with 1-2 ng of cfDNA extracted from ~2 ml of plasma for library construction and the NGS. We obtained ~25 million reads per sample, providing a depth of coverage ~600X in terms of gene bodies. We normalized read counts and identified differential 5hmC markers using DESeq2. Cox proportional hazards model were used to estimate the association between 5hmC markers and overall survival.

Results: We found that in cfDNA from DLBCL patients, 5hmC markers were enriched within gene bodies and depleted in CpG islands. The cfDNA-based 5hmC profiles at diagnosis differed between DLBCL and FL, and in DLBCL, the 5hmC profiles differed between Ann Arbor stage (stage 1/2 vs stage 3/4), lactate dehydrogenase (LDH) level (normal vs elevated), and cell-of-origin (germinal center B-like and activated B-cell-like DLBCL). In addition, genome-wide 5hmC distribution patterns in cfDNA samples are highly correlated with those found in cfDNA-paired tumor tissues, supporting the tumor relevance of cfDNA in a patient. Next, we evaluated the prognostic significance of cfDNA-based 5hmC in DLBCL using a two-step approach. In the discovery phase (7 DLBCL patients with relapse and 12 age- and sex-matched patients without relapse within two years following treatment), a substantial number of 5hmC markers were associated with relapse (449 gene bodies at 5% false discovery rate [FDR]). These relapse-associated 5hmC signatures showed high sensitivity, specificity, and overall accuracy (area under curve [AUC]=0.91) in predicting relapse in the independent validation set (relapse=5, no relapse=13). Finally, we identified a panel of 128 5hmC markers (fold change >20% and p-value <0.05) that were associated with 4-year overall survival.

Conclusion: These findings suggest that 5hmC signatures in cfDNA at the time of DLBCL diagnosis correlate with standard clinical prognostic indices and hold promise as non-invasive markers for prognosis and survival.