In this study, genome-wide surveillance of hypermethylation alterations in NotI sites by MSFLP-array revealed that somatic hypomethylation was lower in SSAs compared with MSI or MSS carcinomas. These benign tumors also occurred in younger individuals compared with MSI carcinomas. This is consistent with the proposed hypothesis of demethylation as a gradual accumulation of methylation replication errors during aging  assuming SSAs being the precursors of the MSI carcinomas. In contrast, there were no major differences in global hypermethylation between these groups of benign and malignant tumors regardless of their location in intergenic, intragenic, promoter, or 3′end regions. Unsupervised clustering analysis revealed no clear differences in the patterns of hypermethylatuon between or within the four different tumor groups. Only after applying an ANOVA approach was possible to discern that MSS cancers and TAs shared similar epigenetic features, so did MSI and SSA, as reported previously [12–17]. The study also disclosed distinct profiles of genes relevant for colorectal cancer such as homeobox genes, transcription factors, growth factors and genes in the Wnt signaling pathway, including AXIN2.
Several papers estimated the frequency of Wnt signaling activation in SSAs but they are controversial [40–44]. Possible explanations to account for the discrepancies may include that some SSAs were misdiagnosed and wrongly categorized due to the complication in the definition of serrated polyps . Therefore, a standardized diagnosis of SSA formulated recently [30, 32] was applied in this study.
Recent genome-scale exome sequencing analysis of 276 colorectal tumors, DNA copy number, promoter methylation and messenger RNA and microRNA expression conducted by the Cancer Genome Atlas project,  indicated that 92% of MSI cancer and 97% of MSS cancers exhibited at least one alteration of genes involved in the Wnt pathway including LRP5, FZD10, FAM123B, AXIN2, APC, CTNNB1 (β-catenin), TCF7L2, FBXW7 and SOX7. Thus, Wnt signaling pathway seems to play a critical role in colorectal carcinogenesis in general, although the spectrum of alterations may vary depending on the distinct oncogenic pathways.
AXIN was identified as a component of the complex in Wnt signaling pathway to regulate the levels of β-catenin along with the wild type of adenomatous polyposis coli (APC) gene . AXIN1 plays as a scaffold protein on which the complex for phosphorylation of β-catenin by glycogen synthase kinase-3β (GSK-3β) is assembled . AXIN2 / Conductin was identified as an AXIN homolog, which also played a scaffold protein, and was found mutated in a subset of colorectal cancers [47, 49]. AXIN1 appears to be a constitutive component of β-catenin degradation complex for maintenance of basal life activity while AXIN2 is considered to be an inducible component that is upregulated in response to increases in β-catenin levels and thus serves to limit the duration and intensity of the Wnt signal [50, 51]. AXIN2 has been only found expressed in colon tissues (Additional file 1: Figure S3).
Epigenetic silencing of AXIN2 in MSI colon cancer was reported in 2006 . However, aberrant methylation of AXIN2 in SSA has not been previously reported. In addition, we identified an apparent increase in methylation alterations of AXIN2 from SSAs to MSI carcinomas, suggesting that its expression deregulation by methylation associates with the serrated adenoma-MSI cancer pathway.
The hierarchical tree identified three clusters according to methylation profiles, MSI, SSAs epigenetically close to MSI and SSAs far from MSI. The expression levels of AXIN2 in these three groups associated with the levels of methylation of AXIN2 in each group, respectively (Figure 3). Our results suggest that expanding of methylation in the promoter region of AXIN2 in SSAs lead to the suppression of the AXIN2 gene expression gradually, which contributes to a stepwise acquisition of the epigenetic features seen in MSI colon cancer. Koinuma et al.
 reported that overexpression of AXIN2, either by treatment with 5′-azacytidine or by transfection with AXIN2 cDNA, resulted in rapid cell death in a MSI CRC cell line, which supports the functional significance of AXIN2 changes in methylation and expression in our study. Dong et al.
 reported progressive methylation of several genes during the serrated pathway. In contrast with the epigenetic silencing of AXIN2 in MSI colon cancer, up-regulation of AXIN2 mRNA was reported in MSS cancers. Indeed, in our study, AXIN2 was frequently hypomethylated in MSS cancers, suggesting that the epigenetic change of AXIN2 specifically associates with the MSI pathway for colon cancer. The fact that down-regulation is not always accompanied by methylation (Figure 3) shows that additional mechanisms may be at play to inactivate the suppressor function of the AXIN2 protein. For instance, frameshift mutations of AXIN2 in MSI colon cancers may be one such additional mechanism [49, 53, 54].
The epigenetic influence on MSI manifestation is shown by the hypermethylation and silencing of MLH1
. High level of hypermethylation has been also associated with MSI cancers [19, 56], and also in SSAs [9, 13, 15, 16, 24–26, 57, 58]. However, MLH1 methylation was not detectable in SSAs in contrast with the common presence observed in MSI cancers. This shows that the epigenetic silencing of MLH1 is not involved in SSA development where it must occur sometime during the adenoma expansion. But silencing of MLH1 then appears to drive the adenoma cells towards the carcinoma state by the generation of many subsequent mutations. The difference in age between the patients with SSAs and MSI carcinomas also supports this suggestion, implying a necessary additional step after SSA development for the accumulation of oncogenic mutations responsible for the carcinoma transition.
AXIN2 aberrant methylation appears to occur during adenoma growth like MLH1 methylation. The assumption here is that no methylation of MLH1 is found at the SSA stage because once it occurs it may lead to the carcinoma transition in the absence of further clonal expansion, since mutator genes do not alter the growth properties of the cells. The association observed between aberrant methylation and down-regulation with small size SSAs without crypt branching could be interpreted assuming that the occurrence of MLH1 methylation may speed the transition to carcinoma in the absence of a need for further expansion of the adenoma.