- Top of page
- Materials and Methods
The presence of high levels of aberrant DNA methylation in gastric mucosae correlates with risk of gastric cancer. Some gastric cancers are known to have methylation of multiple CpG islands (CGI), which is referred to as the CGI methylator phenotype (CIMP). In the present study, we aimed to clarify the possible association between the CIMP in cancers and high methylation levels in their background mucosae by accurate quantitative methylation analysis of 14 carefully selected promoter CGI. Methylation levels were measured in 66 cancers and their background mucosae, along with 19 normal mucosae of healthy volunteers. Methylation in cancers was classified as absent (methylation level = 0%) or positive. The number of methylated CGI in a cancer showed a continuous distribution, and cancers were classified as CIMP high (21 cases), CIMP low (30 cases), or CIMP negative (15 cases). CIMP-high gastric cancer patients had significantly better survival rates than CIMP-negative patients. Of the Epstein–Barr virus-positive gastric cancers studied, eight out of nine presented as CIMP high. Methylation in background mucosae showed a unimodal distribution, and was assessed by their degree. The gastric mucosae of cancer patients showed higher levels than normal gastric mucosae of healthy volunteers. Finally, the CIMP-high, CIMP-low, and CIMP-negative statuses in cancers were not associated with methylation levels of individual genes and their means in the background mucosae. These showed that the CIMP statuses in gastric cancers had no association with methylation levels in the background gastric mucosae. (Cancer Sci 2007; 98: 1853–1861)
Gastric cancer is one of the major causes of cancer death in Asia and some European countries.(1) Regarding the molecular mechanisms of gastric cancers, inactivation of p53, CDH1, CDKN2A, and hMLH1 is well known, and the latter three genes are inactivated more frequently by aberrant DNA methylation of their promoter CpG islands (CGI) than by mutations.(2) As an etiological factor for gastric cancers, Helicobacter pylori infection is known to elevate gastric cancer risk 2.2–21-fold.(3–5) We have recently shown that H. pylori infection potently induces aberrant DNA methylation in gastric mucosae,(6) and that DNA methylation levels in gastric mucosae correlate with risk of gastric cancer in individuals without current H. pylori infection.(6,7) These findings explain why aberrant DNA methylation is frequently associated with gastric cancers.
Methylation of multiple CGI in a cancer was first observed in colorectal cancers.(8) The number of methylated CGI showed a bimodal distribution,(8,9) and the phenotype was designated as the CGI methylator phenotype (CIMP). A recent study using accurate and non-biased quantitative methylation analysis showed that a group of colorectal cancers with CIMP was associated with BRAF mutations.(10) Analysis of methylation levels in the matched non-cancerous background colonic mucosa showed that high methylation levels of specific genes were associated with the presence of CIMP in colorectal cancers.(11) In contrast and surprisingly, although colonic mucosae of patients with ulcerative colitis show accumulation of aberrant methylation,(12,13) CIMP-positive cancers were less frequent in ulcerative colitis-associated colorectal cancers than in sporadic colorectal cancers.(14)
Unlike colorectal cancers, the number of methylated CGI does not show a bimodal distribution in gastric cancers,(15–19) and these cancers have been classified as CIMP high, CIMP low, and CIMP negative for convenience. However, some studies observed that CIMP-high groups were associated with better prognosis,(17,19,20) or with infection with Epstein–Barr virus (EBV).(19–22) These results indicated that CIMP-positive gastric cancers might consist of several different entities. Although gastric cancers arising from gastric mucosae with high methylation levels are likely to have methylation of multiple CGI, the association has not been demonstrated.
Technical limitations may help to explain the ambiguity of CIMP status in gastric cancers. CIMP in gastric cancers has been analyzed only by qualitative methods, such as conventional methylation-specific polymerase chain reaction (PCR), or by combined bisulfite-restriction analysis, which is limited in the number and location of CpG sites that can be analyzed. In addition, a methylation profile of a cancer is dependent on the CGI used for the analysis.(8,16,23) It is known that CGI in different locations relative to a gene show different susceptibility to DNA methylation,(24) therefore CGI with a uniform location relative to a gene should be used. To avoid selection bias of cells with methylation of a CGI, CGI whose methylation does not confer positive or negative selection should be used.
The aim of the present study was to clarify the presence of CIMP in human gastric cancers by an accurate quantitative methylation analysis of selected CGI, and to analyze the effect of methylation in the background non-cancerous gastric mucosae on the CIMP in cancers. We analyzed promoter CGI of one putative tumor-suppressor gene (LOX),(25) and 11 genes that can be methylated not only in gastric cancers but also in non-cancerous gastric mucosae, and are unlikely to cause selection bias.(26) We also analyzed promoter CGI of two tumor-suppressor genes (CDKN2A and hMLH1).
- Top of page
- Materials and Methods
The present study is the first report of concurrent analysis of the methylation levels both in gastric cancers and in their background non-cancerous gastric mucosae. We used quantitative methylation-specific PCR, which is very accurate and can analyze any CpG sites,(6,32) to measure the methylation levels of 14 promoter CGI, consisting of 12 marker genes and two tumor-suppressor genes, in 66 matched cancers and non-cancerous mucosae. As expected from their monoclonal and polyclonal origins, methylation levels in cancers and non-cancerous mucosae showed entirely different patterns. Therefore, methylation in cancers was assessed as positive or negative, and that in non-cancerous mucosae was assessed by its degree using the methylation levels or the deviation values.
The number of methylated genes in a cancer did not show a bimodal distribution, suggesting that CIMP in gastric cancers consists of multiple entities or even does not exist. To classify gastric cancers without a bias, we chose to put them into tertiles. The first group was CIMP negative (15 cancers without methylation), and the remaining 51 cancers were classified as 30 CIMP-low and 21 CIMP-high cancers with a cut-off of five genes methylated. The fraction of CIMP-high gastric cancers (32%) in this classification was similar to those (24–41%) in previous studies.(15–20) Eight of the nine EBV-associated gastric cancers belonged to the CIMP-high cancers,(19–22) and CIMP-high cancer patients tended to have a better prognosis than CIMP-low and CIMP-negative patients, also as reported.(17,19,20) These findings suggested that, although CIMP consists of multiple entities, CIMP itself does exist in gastric cancers. Even when using a different cut-off value (four genes), the above findings did not change (data not shown).
Initially, we expected that gastric cancers arising from a background gastric mucosa with high methylation levels would display a CIMP-high status. However, contrary to our expectation, the CIMP status in cancers did not correlate with the methylation levels in non-cancerous mucosae. This apparent discrepancy can be explained in two ways. First, most of methylation present in gastric cancers could have been induced after a cancer cell was produced. This idea is supported by the fact that some CIMP-high gastric cancers had unique clinicopathological characteristics, and by our previous finding that some gastric cancer cell lines appeared to have an intrinsic abnormality that increased the rate of methylation events.(33,34) Second, methylation levels in the entire gastric mucosae might not reflect methylation levels in the precursor cells for gastric cancers. Although the precise origin of gastric cancer cells has not been clarified, the number of stem and progenitor cells in a gastric gland is known to be relatively small.(35)
We adopted a cut-off value of 6% to score cancer samples with positive methylation. This value was based on the methylation levels of CDKN2A and hMLH1, which were considered to reflect the fraction of cancer cell-derived DNA in the samples. Although some cancers showed methylation levels between 1 and 3%, most methylation-positive cancers had methylation levels higher than 6%. Methylation levels of 1–3% in cancers was considered to be present in a subpopulation of cancer cells, as histological analysis of gastric cancers did not support such a small fraction of cancer cells occurring in a tissue sample. The 6% methylation cut-off value used in the present study is comparable with previous reports using MethyLight technology, in which a cut-off value of 4% was used to best discriminate between normal and malignant tissues.(36,37)
Among the 12 marker genes used in the present study, five genes (LOX, HRASLS, FLNc, HAND1, and THBD) were identified by methylation-sensitive representational difference analysis as methylated in gastric cancers, and were used to analyze gastric cancers for their CIMP statuses in previous studies.(16,19) The remaining seven genes were identified by treating a gastric cancer cell line with a demethylating agent, 5-aza-2′ deoxycytidine, and screening using an oligonucleotide microarray.(26) Because these 12 genes were methylated not only in cancers but also in non-cancerous gastric mucosae, we used these genes as marker genes. Although LOX has tumor-suppressive activity in gastric cancers,(25) the remaining 11 genes were unlikely to have such activity and were considered suitable to analyze the effects of factors that induce aberrant DNA methylation in an unbiased manner.
We recently showed that H. pylori infection, a potent gastric carcinogenic factor, induces methylation of specific genes in non-cancerous gastric mucosae,(6) and methylation levels increase in the order of healthy volunteers, cases with a single gastric cancer, and cases with multiple gastric cancers.(7) The high methylation levels in the non-cancerous mucosae of cancer cases observed here were considered to reflect their current or past exposure to H. pylori, and to be associated with the methylation of tumor-suppressor genes. Because the methylation levels of tumor-suppressor genes are very low in non-cancerous gastric mucosae, as observed for CDKN2A and hMLH1 here, currently their accurate measurement is technically very difficult and the use of marker genes has value.
In conclusion, this is the first study that has revealed no correlation between CIMP status in gastric cancers and methylation levels in their background non-cancerous gastric mucosae.