While HPV DNA testing has greater sensitivity compared to cytology, specificity is lacking, and triage tests are required to distinguish benign HPV infections from precancers. A promising triage option is HPV DNA methylation (DNAm). Increased HPV DNAm has been associated with precancer in four major carcinogenic types (HPV16, 18, 31, 45). We hypothesize that DNAm is an important step in carcinogenesis common to all HPV types. To test this hypothesis, we conducted a nested case-control study evaluating the association of HPV DNAm with cervical precancer for 12 carcinogenic HPV types.
For 12 HPV types, we selected 30 cases of cervical intraepithelial neoplasia grade 3 (CIN3) and 30 controls without abnormalities from a population of HPV-positive women. HPV DNAm in viral L1 and L2 genes (about 9 CpG sites per type) was measured using next-generation bisulfite sequencing. We calculated odds ratios (OR) using logistic regression for the association of DNAm with precancer of DNAm and assessed the possible discrimination between infection and precancer using areas under the curve (AUC). For each HPV type, we calculated specificity at a fixed sensitivity of 85% and weighted back to all HPV-positive women to estimate the risk in methylation positive and methylation negative subjects. These estimates were compared with established management thresholds (colposcopy referral).
We observed significant associations of higher DNAm with precancer in all but 3 sites (OR range 4-28.0). For each HPV type, the highest AUCs were 0.91 (HPV59), 0.86 (HPV18), 0.85 (HPV39), 0.84 (HPV16), 0.82 (HPV45), 0.81 (HPV35), 0.77 (HPV52), 0.74 (HPV58), 0.75 (HPV31), 0.73 (HPV33) and 0.71 (HPV56 and HPV51). At fixed Se of 85%, the Sp for DNAm across HPV types was similar to that of cytology, and ranged from 26.7% (HPV51) to 90.0% (HPV59). Weighting back to all HPV-positive women, the risk of CIN3+ in methylation–positive women was clearly above the colposcopy referral threshold
We observed a strong association of increased HPV DNAm with precancers across 12 HPV types, suggesting that DNAm is a general phenomenon in the transition from infection to precancer. For most types, clinical performance of DNAm was comparable to or exceeded that of cytology. Next, we will analyze a combined panel of DNAm sites from each HPV type in a large screening population. We plan to develop an assay that provides risk stratifying information based on HPV genotyping and DNAm for the clinical management of HPV-positive women, which can be measured in a variety of specimen types, including self-collected samples, which are not amenable for cytology.