Epigenetic changes such as DNA methylation are recognized as an important mechanism in cancer initiation and progression [
13]. Inactivation of TSGs occurs as a consequence of promoter hypermethylation with gene silencing in many cancer types [
14]. In NPC, a vast number of TSGs have been found to be inactivated by promoter hypermethylation [
15]. Interestingly, EBV infection induces increased genome-wide gene methylation, resulting in the formation of a unique epigenotype with high CpG methylation in tumor cells [
16]. Given its important functions in cancer initiation and progression, DNA methylation is being explored as a biomarker for cancer, including NPC.
Many methods are available to examine DNA methylation at single-base resolution. These are broadly classified into two categories, depending on whether they are based on microarrays or next-generation sequencing. Microarray-based technologies use a fixed number of probes with the limitation of low genome coverage and the advantage of low cost [
17]. Whole-genome bisulfite sequencing can overcome this limitation but elevates the costs tremendously [
17]. It is only practical to conduct whole-genome bisulfite sequencing on a limited number of samples, and coverage is usually in the range of 5–15 reads per CpG, limiting the statistical significance of results [
18]. Methyl-Cap sequencing is an attractive intermediate solution to increase the methylome coverage in large sample sets [
17]. We utilized Methyl-Cap sequencing and cDNA microarray analysis to explore TSG candidates with highly methylated promoter CpG islands and gene down-regulation, resulting in 150 possibilities. Of these 150 candidate genes, several had already been reported to be epigenetic silencing of tumor suppressor genes in NPC, such as
ZFP82 [
19],
ADAMTS8 [
20],
INPP4B [
21], and
ATOH8 [
22]. Several conventional NPC tumor suppressor genes [
23], such as
RASSF1 and
CDKN2A (
p16), were methylated at promoter regions in NPC patients from our Methyl-Cap sequencing data, but their expression levels were not significantly down-regulated in NPC by cDNA microarray analysis. Gene
ZMYND10 (
BLU) was significantly down-regulated in NPC, but there was no significant difference in Methyl-Cap sequencing between NNE and NPC (data not shown). Since we combined Methyl-Cap sequencing (more than 3-fold) and cDNA microarray (less than 0.5-fold) data, these TSGs were not included in our candidate gene list (Additional file
2: Table S1). Our literature review resulted in the selection of seven target genes. These genes were previously reported to exhibit DNA methylation (
CR2 [
24],
ITGA4 [
25],
RERG [
26],
RRAD [
4],
SHISA3 [
27],
ZNF549, and
ZNF671 [
28]). Schwab and Illges found that premature B lymphocytes contained a methylated CpG island and did not express
CR2 (
CD21) [
24]. Interestingly, viral capsid protein mediated EBV binding on CR2 [
29]. Gerecke et al. showed that methylation markers in the promoters of
ITGA4,
TFPI2, and
VIMENTIN seemed to be suitable risk markers for inflammation-associated colon cancer [
25], and Chang et al. demonstrated
ITGA4,
SFRP2, and
p16 promoter methylation in stool samples from patients with colorectal adenomas and carcinoma [
30].
RERG was reported to be a tumor suppressor gene in colorectal cancer [
31] and breast cancer [
26]. Our previous study demonstrated that
RRAD was frequently methylated in EBV-associated NPC, and it functioned as a tumor suppressor by inhibiting cell proliferation, colony formation, and migration in
RRAD-overexpressing NPC cells [
4].
SHISA3 was a novel tumor suppressor identified in lung cancer [
32], and was found to be epigenetically inactivated in a substantial fraction of patients with colorectal cancer [
27]. Yeh et al. demonstrated that
ZNF671, an epigenetically silenced novel tumor suppressor, was a potential non-invasive biomarker for predicting urothelial carcinoma relapse [
33]. Lleras et al. reported the epigenetic silencing of Kruppel-type zinc finger protein genes, including
ZNF549 and
ZNF671, on chromosome 19q13 in oropharyngeal cancer [
28]. Our results demonstrated the utility of BAS in validating findings from genome-wide methylation analysis, by showing that all seven candidates had significantly higher average CpG methylation rates in NPC than NNE.
BAS is an efficient, cost-effective, and robust high-throughput technique for assessing DNA methylation at targeted loci of interest [
18]. In our experiment, BAS coverage attained an average of over 1000 reads per CpG. BGS is another method of targeted bisulfite sequencing that includes subcloning and clone selection steps, which limits the total numbers of sequenced clones and sample sets [
34]. Therefore, BGS coverage is usually in the range of around 10 clones per CpG. Due to the significantly increased throughput of next-generation sequencing, a large number of differentially methylated genes can now be identified in a single experiment, and the traditional methods of experimental validation, such as methylation-specific PCR and BGS, are no longer sufficient to keep up with increasing demand. As our results show, BAS can quantitatively and accurately measure CpG methylation levels in genomic regions of interest in a high-throughput manner, and this approach may replace traditional validation methods in the future.