Viral Genome Sequencing for Ultrasensitive Detection of Circulating Tumor DNA

Abstract

Accurate detection of circulating tumor DNA (ctDNA) has the potential to guide personalized cancer therapy. DNA viruses such as Epstein-Barr virus (EBV) and human papillomavirus (HPV) are major oncologic drivers of nasopharyngeal cancer (NPC) and cervical cancer (CC), respectively, and provide a convenient etiological marker for measurement of ctDNA. Polymerase chain reaction (PCR)-based methods have been widely studied, but more accurate tests are needed. We hypothesized that a next generation sequencing (NGS) approach could outperform PCR-based methods for viral ctDNA detection. We developed a novel NGS method for viral ctDNA detection. Viral genome sequencing for EBV (EBV-seq) or HPV (HPV-seq) was achieved by hybrid capture NGS. Unique molecular identifiers were incorporated for accurate copy number quantification. We implemented a custom bioinformatics workflow to (1) quantify viral sequences, (2) report viral subtype, and (3) measure ctDNA fragment length. The lower limit of detection (LLOD) was compared with published PCR-based methods in cell line dilution series. To test performance on clinical samples, we applied EBV-seq to a cohort of 19 newly diagnosed EBV-associated NPC patients for whom the clinical EBV quantitative PCR test had provided a false negative result and to a control cohort of 20 healthy volunteers. We applied HPV-seq to 40 plasma samples from a cohort of 18 CC patients treated with chemoradiotherapy and to a control cohort of 5 healthy volunteers. The LLOD of EBV-Seq was ≤0.01 EBV genome-equivalents (GE), 100X lower than quantitative PCR. The LLOD of HPV-seq was ≤0.2 HPV GE, 10X lower than digital PCR. For EBV-seq, 19/19 NPC patients had detectable ctDNA (sensitivity=100%, range=0.08-141 EBV GE); 10/20 controls had detectable ctDNA (specificity=50%, range=0.02-0.71 EBV GE). 17/19 (89%) patients, including 6/6 with stage I-II disease, had EBV-seq levels that were >20X above the highest level seen in the control cohort (i.e., 89% sensitivity with 100% specificity). For HPV-seq, 6/6 baseline CC samples, 10/17 (59%) end-of-treatment samples, 1/14 (7%) 3-month post-treatment samples, and 3/3 samples taken upon recurrence had positive HPV-seq results (sensitivity=100%, range=0.02-559 HPV GE) that were highly correlated with digital PCR (Pearson R=0.97) but with 60X higher signal than digital PCR; 0/5 controls had detectable ctDNA (specificity=100%). EBV-seq and HPV-seq were able to report viral subtype in 49/49 positive samples. Viral ctDNA was consistently shorter than non-cancer-derived plasma DNA (median difference=15bp). For DNA virus-associated cancers, viral genome NGS provides dramatically greater sensitivity than PCR-based methods and enables viral subtyping and ctDNA fragment length analysis. Further evaluation is warranted to test the clinical utility of ultrasensitive ctDNA detection using this approach.