Methylation Signatures In Circulating Cell-free DNA For Early Detection Of Esophageal Cancer

Background:The lack of effective minimally invasive screening test for esophageal cancer(ESCA)partially contributes to the insufficient early diagnosis and the high mortality rate.Recently,emerging technology of liquid biopsy is expected to bring breakthroughs.Among various biomarkers,circulating free DNA(cfDNA)methylation signature has shown good detection efficiency in the diagnostic trials of many cancers.This study aimed to investigate the feasibility and performance of cfDNA methylation signatures in the early detection of esophageal cancer.Methods:A total of 24 esophageal tumor and paired adjacent tissues from Guangdong Provincial People’s Hospital were sequenced for methylation profiles.ESCA-specific methylation markers were identified and optimized by comparing with the differentially methylated regions(DMRs)of ESCA from the Cancer Genome Atlas(TCGA).Patients with treatment-naive ESCA(n=85),benign esophageal diseases(n=10),and age-matched healthy controls(n=125)were subsequently recruited and randomized into the training and testing groups,with their peripheral blood sequenced for the aforementioned ESCA-specific DMRs using Next-generation Sequencing(NGS).A classifier was then developed in training group by support vector machine(SVM)-based machine learning to differentiate ESCA from healthy controls and benign esophageal diseases.This diagnostic model was internally validated in the testing group and the model with the best diagnostic performance was locked for the subsequent analysis.External independent validation was conducted in the cohort of 83 ESCA patients and 98 healthy controls recruited from the First Affiliated Hospital of Zhengzhou University and Chongqing Cancer Hospital,respectively.Results:A total of 921 DMRs between tumor and adjacent tissues were identified.A diagnostic model based on these DMRs was developed in the training group and then was validated in the testing group.The sensitivity and specificity to discriminate ESCA patients from benign diseases and healthy controls was 76.2%[95%confidence interval(Cl),60.5%-87.9%]and 94.1%(95%Cl,85.7%-98.4%),respectively.Subgroup analyses indicated that the performance(defined by area under the curve,AUC)of the diagnostic model was consistently stable irrespective of age(<55 years vs.≥55 years,0.953 vs.0.941,p=0.68),gender(male vs.female,0.968 vs.0.902,p=0.08)and pathological subtype(squamous cell carcinoma vs.adenocarcinoma,0.951 vs.0.903,p=0.33).In the independent external validation cohort,similar diagnostic efficiency was observed with a sensitivity of 74.7%(95%CI,64.0%-83.6%)and a specificity of 95.9%(95%CI,89.9%-98.9%).In particular,sensitivity for detecting early-stage ESCA(stage 0-Ⅱ)was 58.8%(95%CI,44.2%72.4%),while the sensitivity for detecting advanced ESCA(stage Ⅲ-Ⅳ)reached 100%(95%Cl,89.1%-100.0%).Conclusions:We demonstrated the feasibility of using cfDNA methylation signatures in the differentiation between patients with ESCAs and healthy individuals or benign esophageal diseases.Although further validation in prospective studies is warranted,this diagnostic model showed a good sensitivity and specificity,and provided preliminary evidence for the future prospective screening of ESCA using liquid biopsy in high-risk communities.