Koos Boeve

152 Chapter 8 levels for each sample were calculated as ratios using absolute measurements: the average DNA quantity of the gene of interest divided by the average DNA quantity of ACTB and then multiplied by 10,000. The PubMed electronical database was searched for DNA methylation biomarkers that were reported to be hypermethylated in saliva of head and neck SCC patients compared to saliva of healthy controls. This search revealed four genes, EDNRB [19] , HOXA9 [24] , NID2 [24] and TIMP3 [20] which were used as a reference. QMSP primers and probes were selected from literature for EDNRB, HOXA9, NID2 and TIMP3 [19,20,24] (Table 2). Statistical analysis The Mann-Whitney U test was used for comparing MethylCap-Seq read counts of OSCC and leukocytes and was also used for comparing methylation levels between saliva of patients and controls. Optimal cut-offs were determined by ROC-curves. The diagnostic potential of the biomarkers in detecting OSCC in saliva was determined by the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV). Differences in methylation levels between the tumors and saliva within the OSCC patients were compared using the related Wilcoxon signed-rank test. All tests were performed two-tailed. Results were considered significant when p < 0.05 or FDR < 0.05. Statistical analysis was performed with IBM SPSS Statistics 23 (Statistical Package for the Social Sciences, Inc., Chicago, IL, USA). RESULTS Selection of OSCC specific methylation markers With the MethylCap-Seq analysis, a total of 11.6 to 22.3 x 10 6 reads were sequenced per sample. Approximately 6.91 to 14.6 x 10 6 unique reads could bemapped back to the genome per sample [22,23]. Statistical analysis of reads around the transcription start site resulted in a ranking list of the 5000 most significant equally methylated regions among the 12 OSCCs. In total 335 methylation cores (MCs) representing 319 genes were significantly differentially methylated between the 12 OSCC samples and the two leukocyte pools (Supplementary data 1). Of these 335 MCs, 53 MCs were not hypermethylated in the leukocytes (≤2 reads). Seven of these MCs were hypermethylated (≥3 reads) in all OSCC and had a 100% positive and negative predictive value for hypermethylation in OSCC and leukocytes. These seven MCs were associated with six genes: C11orf85 , CMTM2 , FERMT3 , KCNA5 , SIPA1 and TBX4 . Semi- quantitative comparison with the methylation data in the Map of the Human Methylome showed no methylation in a panel of 80 reference samples that were considered as samples not associated with OSCC (thus considered as negative controls). For TBX4 , CMTM2 and FERMT3 no suitable QMSP primers/probes could be designed or (Q)MSP did not pass the