Consistency of the S5 DNA methylation classifier in formalin-fixed biopsies versus corresponding exfoliated cells for the detection of pre-cancerous cervical lesions.

Abstract

Methylation biomarkers are promising tools for diagnosis and disease prevention. The S5 classifier is aimed at the prevention of cervical cancer by the early detection of cervical intraepithelial neoplasia (CIN). S5 is based on pyrosequencing a promoter region of EPB41L3 and five late regions of HPV types 16, 18, 31, and 33 following bisulfite conversion of DNA. Good biomarkers should perform well in a variety of sample types such as exfoliated cells, fresh frozen or formalin‐fixed paraffin‐embedded (FFPE) materials. Here, we tested the performance of S5 on 315 FFPE biopsies with paired exfoliated cervical samples using four different conversion kits (Epitect Bisulfite, Epitect Fast Bisulfite, EZ DNA Methylation, and EZ DNA Methylation‐Lightning). The S5 values from FFPE biopsies for all kits were significantly correlated with those obtained from their paired exfoliated cells. For the EZ DNA Methylation kit, we observed an average increased methylation of 4.4% in FFPE. This was due to incomplete conversion of DNA (73% for FFPE vs. 95% for cells). The other kits had a DNA conversion rate in FFPE similar to the cells (95%–97%). S5 performed well at discriminating <CIN2 lesions from CIN2+ lesions on the FFPE with all kits given optimized adjustments to the cut‐off. The area under the curve (AUC) for S5 on FFPE was not significantly different from the paired cells (0.74–0.79 vs. 0.81). The best sensitivity and specificity were obtained for EZ DNA Methylation after the adjustment of the cut‐off to reflect its lower conversion rate. Consistent methylation results can be obtained from FFPE material regardless of the conversion kit used. The S5 classifier performed as well on FFPE material as on exfoliated cells with adjusted cut‐off allowing easier clinical implementation.