DNA methylation is the main epigenetic modification after replication in humans.1 DNA (cytosine-5)-methyltransferase (DNMT) catalyzes the transfer of a methyl group from S-adenosyl-L-methionine (SAM) to the C5 of cytosine within CpG dinucleotide sequences in genomic DNA of higher eukaryotes.2–4 The DNMT1 recognizes hemimethylated DNA and adds a methyl group to the newly synthesized daughter strand, thus maintaining the methylation pattern of the parental DNA.5 The recently identified DNMT3a and DNMT3b are essential for de novo methylation.6 In 1998, two study groups found that a methyl-binding protein (MeCP2) forms a complex with histone deacetylase (HDAC) and a transcriptional repressor (Sin-3).7,8
Usually, the CpG-rich region of DNA is termed CpG island, and the promoter region and the first exon of various genes are very rich in CpG island.9,10 In normal tissues, isolated CpG dinucleotides in bulk chromatin are often methylated, whereas cytosines in CpG islands are unmethylated. In neoplasms, this pattern of methylation is commonly reversed. The mechanisms underlying the generation and maintenance of de novo methylation have, however, remained a subject of considerable debate. In particular, hypermethylation of normal unmethylated CpG islands in many tumor suppressor genes correlates with the loss of expression;1 also, alteration in the pattern of DNA methylation plays a key role in the process of carcinogenesis. In contrast, methylation of the CpG island in the promoter is known to cause transcriptional silencing of many tumor suppressor genes, and demethylation of the hMLH1 promoter has been shown to restore the hMLH1 protein expression and mismatch repair (MMR) capacity in some cancer cell lines.11–14
The mechanisms of DNA methylation that may play important roles in carcinogenesis are as follows: (i) DNA cytosine methylation facilitates gene mutation, as 5-methylcytosine is deaminated to thymine;15,16 (ii) DNA methylation occurs frequently in clusters of CpG dinucleotides near regulatory regions of genes17 and affects the transcription of specific genes;18–20 and (iii) aberrant DNA methylation may be associated with allelic loss.21–23
Detection of DNA methylation
DNA methylation can be assessed by using many different methods, but generally these methods can be classified into two categories: the methods to detect total genomic DNA methylation level, and other methods to determine the methylation pattern in a specific gene region.
Assay of total genomic DNA methylation
This is a modified methyl accepting assay.24 DNA is incubated with 3H-S-adenosylmethionine incorporation (3H-SAM) and methylase HpaII or M.Sss I, which catalyzes the transfer of methyl groups from SAM to all cytosine residues associated with guanine in the doublet CpG. The reaction is carried out under the condition of enzyme excess. Radioactivity in the disks is measured by scintillation counting using aqueous scintillation fluorescence. A higher degree of incorporation of 3H-methyl groups into DNA reflects a lower intrinsic methylation status.
High pressure liquid chromatography
Twenty years ago, Beranek et al. described this assay method for methylation study.25 The DNA methylation level is valued by using the percentage of 5-methylcytosine (5-Mcyt), which is calculated by the equation: 5-Mcyt (%) = 5-Mcyt/(5-Mcyt + cytosine) × 100%.26 This method provides an assessment of the overall methylation status of CpG, including the quantitative assay.
Analysis of methylation in a special gene
Digestion by Hpa II/Msp I and the use of Southern blotting
Until 5 years ago, Southern blotting is the most frequently used method for the DNA methylation assay. DNA is digested with the restriction enzymes, Hpa II/Msp I. Both Hpa II and Msp I are cut at the sequence CCGG, but Hpa II is not cut when the internal cytosine is methylated. The digested DNA samples are electrophoresed, then transferred onto nitrocellulose membranes, and hybridized with a DNA probe to study the special gene methylation. Therefore, the different size and frequency of fragments in Hpa II digested DNA are indicative of a different methylation status.27
Bisulfite modification and sequencing (Bisulfite mapping)
Bisulfite mapping is used for determining the methylation pattern in a special region of most genes. Clark et al. developed a genomic sequencing technique that is capable of detecting every methylated cytosine.28 First, sodium bisulfite is used to convert cytosine residues to uracil residues in single-stranded DNA, under the conditions where 5-methylcytosine remains non-reactive. The converted DNA is amplified with specific primers and sequenced. All the cytosine residues remaining in the sequence represent previously methylated cytosines in the genome.
Methylation specific polymerase chain reaction
Herman et al. first used this method to analyze CpG island methylation of a specific gene region.29 This assay method consists of the initial modification of DNA by sodium bisulfite, and subsequent PCR with primers specific for methylated versus unmethylated DNA. The methylated specific polymerase chain reaction (MSP) requires only small quantities of DNA, and its false-positive results inherent to previous PCR-based approaches relied on differential restriction enzyme cleavage to distinguish methylated from unmethylated DNA.
Methylation-sensitive single nucleotide primer extension
The methylation-sensitive single nucleotide primer extension method (Ms-SNuPE) is a rapid quantitative method for assessing methylation differences at specific CpG sites based on bisulfite treatment of DNA, and is followed by a single nucleotide primer extension.30
Combined bisulfite restriction analysis
A quantitative technique for the methylation study of gene restriction enzyme digestion is used to reveal methylation-dependent sequence differences in PCR products of sodium bisulfite-treated DNA.31
Differential methylation hybridization
Huang et al. developed a DNA array-based method to detect DNA methylation, which was applied to determine the methylation status of > 276 CpG island loci.32
Aberrant DNA methylation in gastrointestinal carcinogenesis
Changes in total genomic DNA methylation
Some observations from several laboratories have shown that the total 5-methylcytosine content of DNA is decreased during hepatocarcinogenesis and colonic carcinogenesis.33,34 Sawady et al. also showed that a reduced level of DNA methylation in colonic cancer was a biochemical characteristic that could appear in adenomas before malignant transformation.35
The data from our previous experiment showed that the level of DNA methylation in cancer tissue was significantly less than that of non-cancerous gastric mucosa and paracancerous tissue.36 In addition, we also found that there is significant hypomethylation of total genomic DNA in human hepatocellular cancer (HCC). DNA hypomethylation also had a strong correlation with gross pathology: the levels of total genomic DNA methylation in multifocal and nodular type cancers were significantly decreased. Pathological analysis has indicated that the nodular type of HCC has more complications of cirrhosis than the massive type, and that the former is more or less infiltration orientated. Although a multicentric origin of the multifocal type of HCC is a possibility, multifocal intrahepatic spread from a unicentral origin may be more reasonable. It can be demonstrated that the decrease of the total genomic DNA methylation level is more significant in the case when both infiltration and intrahepatic metastasis are present. Some pathological factors, such as the size of tumor, the involvement of the portal vein, and the degree of cellular differentiation have been proven to be prognostic predicators.37
Hypomethylation of oncogenes
In general, there is hypomethylation of oncogenes in carcinogenesis. Some reports have demonstrated that several specific oncogenes such as c-myc, c-Ha-ras and c-Ki-ras were hypomethylated in rat liver cancer, coinciding with the decreased level of total genomic DNA methylation.26,38 In the progression of human hepatocarcinogenesis, alterations in DNA methylation have so far been observed only in c-myc,39c-fos40 oncogenes and in α-fetoprotein (AFP).41
In order to study the status of DNA methylation in specific oncogenes and the relationship between them and the pathological changes, the methylated status of c-myc and c-Ha-ras oncogenes in human gastric cancer with the use of Southern blotting was analyzed. The results suggested that the c-myc and c-Ha-ras oncogenes are hypomethylated in gastric cancerous and paracancerous tissues than in non-cancerous tissue, but there were no significant correlations between the methylation level and pathological changes, including macroscopic and histopathological changes, the depth of infiltration, and the status of perigastric lymph node metastasis.42
The c-myc and c-N-ras oncogenes are strongly related with HCC in China, and c-N-ras, as a trend forming gene, possibly could become the target genes in HCC. We examined the methylation pattern of c-myc and c-N-ras oncogenes and measured the level of total genomic DNA methylation in cancerous and paracancerous tissues. The results showed that the hypomethylation rate of c-myc and c-N-ras oncogenes were 30 and 61%, respectively, in human HCC, which coincide with a decreased level of total genomic DNA methylation. The DNA hypomethylation was highly significant in cases with a tendency of tumor infiltration or metastasis.43
Many studies have implicated methyl group deficiency as a factor in carcinogenesis in rodents, and it has been shown that amounts of methylcytosine are rapidly decreased in DNA isolated from animals on a methyl-deficient diet, and there is also an alteration in gene expression.44 Instances of deficiency is then made up by the de novo synthesis of methyl groups via the folate pool of the coenzyme. We also indicated that the folic acid concentration on plasma in patients who showed hypomethylation was lower than it was in patients showing normal methylation. These data suggested that the decrease of folic acid in maintaining DNA methylation is involved in human gastric carcinogenesis.45
Hypermethylation of tumor suppressor genes
The p16INK4A was first identified in a yeast two-hybrid screen for proteins that interact with human cyclin-dependent kinase 4;46 it inhibits the catalytic activity of the CDK4/cyclin D complexes and blocks G1-to-S transition in the cells.46,47 The overexpression of p16INK4A in cultured cells inhibits cell cycle progression and suppresses cell growth.48 The p15INK4B binds to CDK4 and CDK6, and inhibits their activities.49 In addition, it is clear that hypermethylation of the 5′CpG island in the p16INK4A gene is associated with the loss of its expression in esophageal squamous cell cancer.50,51 A report showed that the p15INK4B gene was unmethylated at this Eag I site in all normal tissues,52 but the methylation of p15INK4B CpG island in primary colonic carcinoma was infrequent.18
Recently, Lee et al. found that deletions or mutations of the p16INK4A and p15INK4B genes were uncommon in primary gastric carcinomas, but aberrant methylation may result in a downregulation of transcription of these genes.53 They suggested that DNA methylation might be one of the causes of inactivation of the p16INK4A gene that leads to the development of gastric cancer. The undetectable p16INK4A mRNA expression results from abnormal DNA methylation of the 5′region and exon 1 of the p16INK4A gene in gastric cancer cell lines, and transcriptional blockage could be reversed by treatment with 5-aza-2′-deoxycytidine (5-Aza-dC). Toyota et al. revealed that the p16INK4A in 28% (11 of 40) of colonic cancer cases is hypermethylated.54 In the review by Rocco and Didransky, the frequency of p16INK4A mutation and hypermethylation in human gastric cancer is 0–2% and 32–42%, respectively.55 In comparison with p16INK4A, the promoter of p21WAF1, another cyclin-dependent kinase inhibitor gene, was not promoter hypermethylated in gastric cancer cells56 and colorectal carcinogenesis,57 although DNMT1 expression is abnormal in the latter.
Wong et al. extracted DNA from the tumor tissues and peripheral blood plasma or serum of 22 HCC patients.58 The p16INK4A methylation was found in 73% of HCC tissues with the use of methylated specific PCR (MSP). Among the 16 cases with aberrant methylation in the tumor tissue, similar changes were also detected in the plasma/serum samples of 81% of cases.58 Saito et al. showed that DNA hypermethylation on CpG islands of the p16INK4A gene in human liver cancer is 66%, but it is 8% in non-cancerous liver tissues.59
A recent experiment by Kaneto et al. demonstrated that methylation of the p16INK4A promoter was detected in patients with HCC (72.7%, 16 of 22), cirrhosis (29.4%, 5 of 17) and chronic hepatitis (23.5%, 4 of 17).60 All methylation positive HCC, cirrhosis, and chronic hepatitis samples in the present study showed a loss of p16INK4A expression, and a significant correlation was found between methylation and loss of expression.
Hypermethylation-in-cancer (HIC-1) is identified at the D17S5 locus, and it is a candidate tumor suppressor gene. The incidence of DNA hypermethylation at this locus was significantly higher in patients with HCC (90%) than in those who had non-cancerous liver tissues. A loss of heterozygosity (LOH) at the D17S5 locus, which was preceded by DNA hypermethylation at the same locus, was detected in 54% of HCC. Moreover, wild-type p53 did not overcome DNA hypermethylation at the D17S5 locus to activate HIC-1 in HCC. It was suggested that aberrant DNA methylation at this locus and reduced HIC-1 mRNA expression participated in hepatocarcinogenesis during both early developmental stage and malignant progression of HCC.61 In addition, Kanai has also found frequent DNA hypermethylation at the D17S5 locus in precancerous conditions; it may play a role in gastric carcinogenesis.62
E-cadherin is a tumor invasion suppressor gene, which is located on the chromosome 16q22.1;63 CpG methylation around its promoter region correlated significantly with reduced E-cadherin expression in HCC.19 Aberrant DNA methylation may be involved in hepatocarcinogenesis through predisposing some chromosomes to LOH and/or the silencing of some specific genes, including the E-cadherin gene, even during early developmental stages of HCC. A study showed the hypermethylation of the promoter region in the E-cadherin gene in 46% of colorectal cancers studied, and there was also a trend for methylation to be associated with a more advanced stage of cancer, although this did not reach statistical significance.64
The fragile histidine triad (FHIT) gene is also a tumor suppressor gene.65,66 Regarding the alteration of DNA methylation in esophageal cancer, Tanaka et al. examined the aberrations of the FHIT gene in 23 esophageal squamous cell carcinoma cell lines and 35 primary tumors, and found aberrant expression in several cell lines (30%), including a shorter transcript in two cell lines and loss of apparent transcript in five cell lines. All methylated cell lines exhibited re-expression of the FHIT gene and demethylation in the CpG island after treatment with a demethylating agent, 5-Aza-dC. Hypermethylation was also found in five out of 35 primary tumors, whereas corresponding normal tissue showed no methylation. These findings suggested that methylation of the 5′CpG island of the FHIT gene was closely associated with transcriptional inactivation and might be involved in esophageal cancer development.67 Tanaka et al. also thought that the distribution of methylated cytosines within each cell line and tumor was heterogeneous.
Tissue inhibitor of metalloproteinase-3 (TIMP-3) antagonizes matrix metalloproteinase activity and can suppress tumor growth, angiogenesis, invasion, and metastasis. A loss of TIMP-3 has been related to the acquisition of tumorigenesis. Bachman et al. showed that the loss of TIMP-3 expression is associated with dense methylation of the 5′CpG island in cell lines from many common human cancers and can be restored in colonic cancer cell lines after 5-Aza-dC induced demethylation.68 Genomic bisulfite sequencing data indicated that the pattern of methylation for TIMP-3 is complex, and that the density of CpG island methylation, particularly the 5′ start of transcription, correlates best with TIMP-3 silencing. Furthermore, in studies of primary cancers, Bachman et al. suggested that the methylation associated with the silencing of TIMP-3 is tumor specific and is associated with the lack of TIMP-3 protein.68
The human adenomatous polyposis coli (APC) gene located at the chromosome 5q21, decodes a large 300 kDa protein that interacts with β-catenin. The APC mutations are present in more than 85% of sporadic colorectal cancer patients.69 Sakamoto et al. revealed that methylation of the CpG loci in the 5′-untranslated region of the APC mRNA repressed steady-state expression of the gene in colorectal cancer cell lines.70
The Fas antigen is a widely expressed cell surface receptor. Expression of Fas in the colon is progressively reduced during the transformation of normal epithelium to benign neoplasms, adenocarcinoma, and ultimately to metastases.71 Butler et al. have examined the Fas promoter region CpG island for evidence of hypermethylation in colorectal tumors.72
Cyclo-oxygenases (COX) have two isoforms, COX1 and COX2. Cyclo-oxygenase 2 is overexpressed in patients with various forms of carcinogenesis.73,74 However, a low expression of COX2 has been reported in a subset of colorectal and gastric cancers.75,76
Aberrant methylation of COX2 was detected by Toyota et al.,77 and they found that among the 33 cell lines examined, dense methylation (> 70%) of COX2 was detected in five of them, and partial methylation was detected in 10. Moreover, the loss of expression of COX2 transcription was closely correlated with methylation of exon 1. It is possible that tumors with COX2 methylation may be less sensitive to treatment of specific COX2 inhibitors.
Hypermethylation of the hMLH1 gene
Mismatch repair is an important mechanism by which cells correct errors in DNA replication during proliferation to maintain the fidelity of the genome. Both hMLH1 and hMSH2 have been cloned and demonstrated to participate in DNA-MMR.
The loss of hMLH1 expression in colonic cancer cell lines and tissues was shown to correlate with its cytosine methylation.78,79 Some findings have established that the epigenetic silencing of the hMLH1 gene promoter is an important event in many sporadic human colonic cancers. This epigenetic silencing mechanism is closely associated with methylation of the hMLH1 gene promoter. The causal role of the MMR gene inactivation in initiating the pathway of microsatellite instability (MSI) in colonic carcinogenesis suggests its presence in the tumors.
Kane et al. has analyzed the expression of hMLH1 in 66 sporadic colorectal tumor patients, and identified four patients that did not express hMLH1. Cytosine methylation of the hMLH1 promoter was found in these four patients who lost hMLH1 expression.78 It suggests that methylation of the hMLH1 promoter plays an important role in the pathogenesis of colonic cancer.
Deng et al. examined the methylation status of all CpG sites in the hMLH1 promoter in 24 colorectal cancer cell lines by using bisulfite mapping.80 The methylation status invariably correlated with the lack of hMLH1 expression, and the former could be induced to re-express hMLH1 by a methyl transferase inhibitor (5-Aza-dC).
The results obtained by Herman et al. showed methylation of the hMLH1 promoter occurred commonly in both cell lines and primary cancers included in colonic cancer with MMR deficiency; such methylation was correlated with a decreased expression of the hMLH1 gene at both the RNA and protein levels, and most importantly, demethylation of the hMLH1 promoter resulted in the re-expression of hMLH1 in each of the three cell lines tested.13
Cells with MMR defects show mutation rates up to 1000-fold. Mismatch repair can be measured by using MSI analysis, and has been detected in tumors from patients with hereditary non-polyposis colorectal cancer14,81 and sporadic colorectal cancer.82,83 Promoter hypermethylation of hMLH1 was present in the majority of sporadic colonic carcinomas with MSI.13 The loss of hMLH1 expression was further shown to correlate with cytosine methylation of CpG sites in its promoter in colonic cancer cell lines and tissues, and methylation-sensitive enzyme (e.g. Hpa II) digestion and MSP have been successfully applied to determine the methylation status of the hMLH1 promoter.78,79,84
Ahuja et al. have determined the methylation patterns of selected genes in colorectal cancers with or without MSI, which results from defects in one of several base MMR genes.85 A total of 47 colorectal cancers were analyzed of which 15 were MSI+ (32%). Hypermethylation of the p16INK4A gene was found in 60% of MSI+ cancers, compared with only 22% in MSI− cancers. Similarly, hypermethylation of the thrombospondin-1 (TSP-1) gene, an angiogenesis inhibitor was increased in MSI+ cancers (27 vs 0%). Extensive methylation of insulin-like growth factor II (IGF 2) gene was observed in 60% of MSI+ cancers, as contrasted with 6% of MSI− cancers.85
Macrosatellite instability has been found in 15–39% of sporadic gastric carcinomas worldwide.86–88 Leung et al. investigated a series of 35 sporadic gastric cancers stratified into high-frequency MSI (MSI-H), low frequency MSI (MSI-L) and microsatellite stable (MSS) groups and found that hypermethylation of the CpG island in the hMLH1 promoter region was present in 100% of MSI-H sporadic gastric cancer. In 90% of cases, there was an associated complete loss of hMLH1 mRNA level. This loss of hMLH1 protein occurred in the MSI-H invasive tumor but not in the adjacent cancer in situ or dysplastic components that were MSS. The MSI-L and MSS forms of gastric cancer all showed predominantly unmethylated hMLH1 promoter, positive hMLH1 protein and high hMLH1 mRNA level.89
Microsatellite instability has been reported to be responsible for this phenotype. In order to determine how family gastric cancer (FGC) acquires MSI, Yangisawa et al. examined the MSI status, hMLH1-protein expression and methylation status of the hMLH1-promoter region in FGC cases, and found that methylation in the hMLH1 promoter region was hypermethylation, and the mechanism of inactivation of hMLH1 is epigenetic, and that there are other genes responsible for FGC.90
Hypermethylator phenotype, termed CpG island methylator phenotype (CIMP), includes methylation of several tumor suppressor genes such as p16INK4A and hMLH1, and it may be present in human gastric cancer.91
Intervention for methylation
The nucleoside analog 5-azacytidine (5-Aza-C) was first synthesized in Czechoslovakia in 1963. It and 5-Aza-dC induce the expression of repression of repressed genes in eukaryotic cells, and act as an inhibitor of DNA methylation.92
The organoselenium compounds benzyl selenocyanate (BSC) and 1,4-phenylene-bis methylene selenocyanate (p-XSC), as well as sodium selenite, are effective in the treatment of chemically induced tumors. Cox and Goorha, and Jones made the pioneering finding that sodium selenite is a strong inhibitor of DNMT; the relevance of this finding to the cancer chemopreventive activity of selenium has not been completely clear.92–94 Fiala et al. found that selenite, BSC and p-XSC inhibited the activation of DNMT and decreased the level of total genomic DNA methylation in samples extracted from human colonic cancer.95 They suggested that inhibition of DNMT may be a major mechanism of chemoprevention by selenium compounds at the postinitiation stage of carcinogenesis.
Treatment with folic acid or folate
A growing body of evidence has suggested that DNA methylation can be altered by dietary manipulation of methyl group donors. Folic acid and folate are the main methyl group. Folate deficiency produces a marked decrease in the ratio of SAM to S-adenosylhomocysteine.96 Experimental evidence suggests that folate depletion plays a role in carcinogenesis.44,97–99
Cravo et al. have evaluated the effect of folate supplementation on total genomic DNA methylation status in the rectal mucosa of 20 patients with resected colonic adenomas in a prospective, controlled, cross-over study.96 They found that folate supplementation may decrease the degree of DNA hypomethylation, but only in patients with one single polyp. In those with multiple lesions, a case-control study examining folate intake was conducted.97 Giovannucci et al. found that folate and methionine could influence methyl group availability, and a methyl-deficient diet may be linked to early stages of colorectal neoplasia. A dietary pattern that increases methyl availability could reduce the incidence of colorectal cancer.98 The material from review by Choi and Mason support efforts to increase dietary folate in parts the particular population having diets with low intakes of this nutrient.99
Gene therapy of DNMT1
There have been several reports on the studies for the isolation, cloning, characterization and expression of the gene coding for a DNMT.100–102 Mikovits et al. have found that the introduction of an antisense DNMT construct into lymphoid cells resulted in a markedly decreased DNMT expression, hypomethylation throughout the IFN-γ gene and increased INF-γ production, demonstrating a direct link between DNMT and IFN-γ gene expression.103 The ability of increased DNMT activity to downregulate the expression of genes like the IFN-γ gene may be one of the mechanisms for dysfunction of T cells in HIV-1 infected individuals. A report demonstrated that reduced DNMT1 protein also showed an upregulation of p21WAF1 gene in human breast cancer cell line MDA-231 cells transfected by the antisense DNMT1 plasmid pCMV-TMH.104 However, until recently, we have not found any reports on the therapy or treatment of DNMT for aberrant DNA methylation during GI carcinogenesis.