Viral cfDNA Profiling Reveals Distinct EBV Subtypes and Stratifies Risk in Hodgkin Lymphomas

Introduction: Approximately 10-40% of classic Hodgkin Lymphoma (cHL) tumors are Epstein-Barr virus (EBV) positive, and immunodeficiency can predispose to cHL, suggesting that viral infections are important in cHL pathogenesis. We hypothesized that circulating viral genomic and fragmentomic features would inform pathogenicity and evolutionary selection of EBV in cHL and potentially allow for risk stratification of cHL subtypes. We developed Viral CAncer Personalized Profiling by deep Sequencing (VirCAPP-Seq), a novel capture-based method for sequencing viral cell-free DNA (cfDNA). Using VirCAPP-Seq, we surveyed the composition of the circulating virome in cHL patients, and identified EBV mutations under positive selection in cHL.

Methods: 352 peripheral blood plasma specimens were collected from 315 cHL patients (mean age = 32, SD = 16) and 44 controls, including cord blood donors and healthy adults (mean age = 43, SD = 16). cfDNA was extracted from plasma and processed as previously described (Newman et al, 2016). Libraries were subjected to hybrid capture with a custom VirCAPP-Seq viral capture panel spanning 180 viral species, including EBV types 1 and 2, and other common oncotropic viruses. Human sequencing reads were subtracted before alignment to a composite reference genome of targeted viral species. Read coverage and plasma cfDNA concentration were used to calculate absolute viral abundance, and expressed as genome copies per mL plasma. Additional EBV cfDNA sequencing data were analyzed from 55 nasopharyngeal carcinoma patients and 24 EBV+ controls (Lo et al, PNAS 2018). EBV variants were called with GATK HaplotypeCaller (v4.1.2).

Results: 48% of cHL plasma samples profiled by VirCAPP-Seq had detectable EBV cfDNA above the background level in healthy controls. Plasma EBV load (PEL) was significantly higher in cHL patients with Epstein-Barr-encoded small RNA (EBER) positive tumors (N = 24, mean = 3810, SD = 8083 copies / mL plasma) compared to healthy controls (N = 29, mean = 0.1, SD = 0.25) (p < 0.001, Wilcoxon T-test). PEL strongly predicted tumor EBER status with an accuracy of 94% at a threshold of 32 copies EBV / mL plasma.

Unexpectedly, patients with EBER-neg tumors also demonstrated elevated detectable PEL (N = 137, mean = 11, SD = 81 copies / mL plasma). PEL was correlated with International Prognostic Score (IPS) (r = 0.42, p < 0.001), and patients with “high” pre-treatment PEL (≥ 10 copies / mL plasma) had a significantly lower 5-year PFS (73%) compared to patients with “low” pre-treatment PEL (90%) (p = 0.03, Figure A). In contrast, tumor EBER status showed no association with PFS.

Hierarchical clustering of circulating viral load for 180 targeted viral species identified subclusters of cHL patients bearing distinct metaviromic signatures, including 18% of cHL patients with elevated plasma Anellovirus burden, an observation associated with compromised immune function (De Vlaminck et al, Cell 2013). Mutation analysis performed on 92 EBV isolates from cHL patients revealed non-synonymous variants in BARF0, BHLF1, and EBNA-1 under strong positive selection by dN / dS analysis.

Fragment length analysis showed that EBV cfDNA molecules were, on average, longer in EBER-pos cHL patients (median = 157 bp, IQR = 154 - 158 bp) compared to healthy controls (median = 147 bp, IQR = 143 -149 bp) (Figure B). Strikingly, EBV cfDNA fragment length profiles were similar between EBER-pos and EBER-neg cHL patients (Median = 155 bp, IQR = 151 – 161 bp), suggesting that a significant proportion of EBV cfDNA may yet be tumor-derived in patients with EBER-neg tumors, reflecting either the limited sensitivity of clinical EBER ISH or, perhaps, “hit-and-run” viral oncogenesis.

Conclusions: VirCAPP-Seq enables ultrasensitive detection, quantification, and genotyping of EBV cfDNA in cHL patients, and identifies viral variants under strong evolutionary selection. Our findings provide a proof-of-concept for viral cfDNA sequencing as a powerful tool for discovering prognostic biomarkers.