“Epigenetics”, as compared to “genetics”, is defined as the study of genomic DNA modifications that are heritable during cell division but do not involve a change in the DNA sequence itself, such as DNA methylation and histone modification[12,13,38]. DNA methylation is the covalent modification of a methyl group on the 5-position of cytosine at CpG dinucleotides[38,39]. CpG islands are genomic regions that contain dense CpG dinucleotides, and they are located in the promoter regions of approximately half of all genes. CpG islands are generally free from DNA methylation, allowing for the expression of downstream genes whose transcription is regulated by histone modification[13].

In normal cellular processes, DNA methylation is used for robust gene silencing, such as genome imprinting[40] and X-chromosome inactivation[41]. Moreover, tissue-specific patterns of methylation or changes in methylation during cellular differentiation have been discovered at CpG-poor promoters[42] and inter- and intragenic CpG islands[43]. In addition, when cells encounter foreign nucleic acid, such as viral DNA, host cells take advantage of DNA methylation as a defensive system to inactivate foreign nucleic acid[44].

Broadly speaking, aberrant DNA methylation in cancer is divided into two categories: “genome-overall hypomethylation” and “regional hypermethylation”. The former, global hypomethylation, was discovered in the 1980s[45] and can be defined as a decrease in 5-methylcytosine content throughout the genome. CpG dinucleotides show heterogeneous distribution, especially in repetitive sequences, which are typically methylated in normal tissue[46,47]. In cancers, these repetitive sequences demonstrate aberrant hypomethylation[48], promoting genomic instability and cancer progression[49-51]. Loss of imprinting is another example of an epigenetic alteration related to aberrant hypomethylation[52], and loss of imprinting in IGF2 was shown to be involved in the early events of carcinogenesis and was associated with increased colorectal cancer risk[53,54]. A subset of male germ line-specific genes, specifically the MAGE gene families, was discovered to be a cancer antigen in malignant melanoma[55]. These genes are repressed by promoter methylation in normal somatic tissues but are activated through promoter hypomethylation in several types of cancers[56,57].

The latter type of DNA methylation, regional hypermethylation, arises in CpG islands[58-60]. Aberrant methylation of promoter CpG islands leads to inappropriate transcriptional silencing, and this phenomenon is regarded as one of the major mechanisms for inactivating tumor-suppressor genes[2,12-14]. Promoter methylation in tumor-suppressor genes has been discovered in various cancers, including RB in sporadic retinoblastoma[61], VHL in renal cell carcinoma[62], CDH1 in hepatocellular carcinoma[63], and p16^INK4A in various cancers[64].

In embryonic stem (ES) cells, the Polycomb repressive complex (PRC) plays a significant role in reversibly repressing gene expression. In ES cells, these PRC-target genes are frequently methylated compared to non-PRC-target genes in various cancers[65]. Our comprehensive methylation analysis of gastric cancer revealed significant enrichment of aberrant methylation in PRC-target genes in a subset of gastric cancers with a high-methylation epigenotype[37]. However, another subset of gastric cancer demonstrated an extensively high methylation epigenotype that displayed extended methylation in both PRC-target genes and non-PRC-target genes. This phenotype was detected in EBV-positive gastric cancer, and it will be discussed in detail later in this review.

Among the factors known to cause aberrant DNA methylation in non-cancerous tissues, aging is known to promote the accumulation of DNA methylation[66,67]. Indeed, age-dependent promoter methylation could explain the association between cancer and aging[68]. Recent whole-genome bisulfite sequencing comparing newborn and centenarian genomes demonstrated that centenarian DNA had a lower DNA methylation content throughout the genome and showed the more hypomethylated CpGs in promoters, exonic, intronic, and intergenic regions, whereas a greater level of DNA methylation was observed in CpG island promoters[69]. Another report showed that replicative senescent human cells exhibited features similar to the cancer epigenome, such as widespread DNA hypomethylation and focal hypermethylation[70]. Epidemiological studies have also revealed that the epigenetic status is influenced by various environmental factors[67] and can be associated with cancer incidence or prognosis[71,72].

Among environmental factors, chronic inflammation is a significant inducer of aberrant DNA methylation, as demonstrated by the analysis of non-cancerous tissues, such as colonic mucosae with ulcerative colitis[73], liver tissue with chronic hepatitis[74], esophageal mucosae with inflammatory reflux esophagitis[75], and gastric mucosae with chronic gastritis[76]. In a mouse colitis model induced by dextran sodium sulfate, aberrant CpG island methylation in colonic epithelial cells was shown to accumulate gradually on a monthly basis[77]. Interestingly, even in severe combined immunodeficiency (SCID) mice lacking functional T and B lymphocytes, DNA methylation was induced at the same level as in the background strain of mice, suggesting that functional T and B lymphocytes are not essential for methylation accumulation.

The direct mechanism by which H. pylori contributes to gastric carcinogenesis is attributed to its pathogenicity. Most Gram-negative bacteria exert pathogenicity through the acquisition of an exogenous gene cluster, termed a pathogenicity island (PAI). H. pylori also contains a cag PAI, which consists of an approximate 40-kbp stretch of DNA encoding approximately 30 genes, including those of the type IV secretion system[84]. The pathogenicity of H. pylori depends on whether it contains cytotoxin-associated gene A (CagA) protein or not. Almost all strains of H. pylori in East Asia contain the CagA protein, whereas this frequency in Western strains is limited to 60%. The pathogenicity of the CagA protein is exerted via injection into gastric epithelial cells through type IV secretion systems (Figure 1). The CagA protein contains a conserved motif in the C-terminus, the EPIYA motif, which dictates the severity of its pathogenicity. East Asian strains of CagA exert more aggressive cytotoxicity compared to Western strains[85].

Open in New Tab   Full Size Figure   Download Figure

Figure 1 Schematic representation about infectious condition and pathogenicity of Helicobacter pylori. Pathogenicity of Helicobacter pylori (H. pylori) is exerted through injection of CagA into gastric epithelial cells using type IV secretion system. Inflammatory cell infiltration due to H. pylori might be more significant factor in induction of aberrant DNA methylation (left). Injection of CagA leads to proliferation of epithelial cells, but it is still unclear whether it plays an important role in methylation induction in epithelial cells (right).

Host cellular responses against injected CagA protein display several patterns. These include (1) enhanced cell motility that induces a growth-factor-like phenotype, termed hummingbird, in host gastric cells[86]; (2) disruption of the epithelial apical-junctional complex[87]; and (3) epithelial proliferative and proinflammatory responses associated with the development of chronic gastritis and gastric cancer[88]. Therefore, CagA plays a key role in gastric carcinogenesis, although the direct involvement of CagA or other components of H. pylori in the induction of DNA methylation has not been clarified.

H. pylori indirectly promotes pathogenicity by inducing chronic inflammation. Chronic inflammation generally involves the accumulation of molecular damage through a variety of mechanisms, such as DNA damage by free radicals[89] or aberrant expression of activation-induced cytidine deaminase (AID)[90]. Moreover, chronic inflammation induces aberrant DNA methylation[74]. Rather than H. pylori infection itself, inflammatory cell infiltration by H. pylori might be a more significant factor for the induction of aberrant DNA methylation[91]. In a Mongolian gerbil model, suppression of aberrant DNA methylation by 5-aza-dC treatment reduced, but did not entirely prevent, the incidence of H. pylori-induced gastric cancers[92]. This result demonstrates that aberrant DNA methylation contributes to H. pylori-related gastric carcinogenesis, although some direct influences of H. pylori, without aberrant DNA methylation, may also be significant.

H. pylori infection induces aberrant promoter methylation in tumor-suppressor genes, including p16^INK4A, LOX, and CDH1[34,93]. Although eradication of H. pylori can reduce the level of promoter methylation, a certain amount of methylation remains[91,94]. This observation suggests that not only fully differentiated gastric epithelial cells but also stem/progenitor cells might acquire aberrant methylation. In human ulcerative colitis and hepatitis, increased expression of IL-1β, IL-8, NOS2, and TNF was observed[95-98], and these genes may represent a common factor associated with the induction of aberrant DNA methylation during chronic inflammation. In particular, IL-1β is thought to be significant, as a specific single-nucleotide polymorphism of IL-1β is associated with increased gastric cancer risk and increased incidence of CDH1 promoter methylation in gastric cancers[99,100]. Furthermore, the role of IL-1β in H. pylori-induced gastric inflammation and DNA methylation was confirmed using IL-1 receptor type 1 knockout mice[101].

EBV is a gamma-herpes virus consisting of a double-strand DNA genome approximately 170 kbp in length. EBV may cause infectious mononucleosis during initial infection, and more than 90% of adult individuals become EBV carriers[102], as this virus can be maintained asymptomatically in a latent form in memory B lymphocytes. However, EBV displays the characteristics of an oncogenic virus; indeed, it was initially discovered in human neoplastic cells, specifically a Burkitt’s lymphoma cell line, in 1964[103]. Subsequently, EBV was associated with several types of malignant tumors, such as nasopharyngeal carcinoma[104], Hodgkin lymphoma[105], and opportunistic lymphoma in immunocompromised individuals[106,107]. Moreover, a subgroup of gastric cancer patients infected with EBV was discovered in 1990[108], and this unique subgroup is distributed throughout the world, without regional or racial deviation, at a rate of 7%-15%[109,110].

EBV-positive (EBV⁺) gastric cancers show distinct clinicopathological features. First, EBV⁺ gastric cancers demonstrate EBV infection in almost all neoplastic cells of the tumor, which has been confirmed by in situ hybridization for a non-coding small RNA, EBER, which is abundantly expressed in the nuclei of infected neoplastic cells. In addition, the clinical features of EBV⁺ gastric cancers differ from EBV-negative (EBV^-) gastric cancers due to their male predominance, proximal location, and relatively favorable prognosis[111]. Histopathologically EBV⁺ gastric cancers demonstrate characteristic features of a poorly differentiated adenocarcinoma with marked infiltration of lymphocytes into the stromal tissue, which has been reported as “gastric cancer with lymphoid stroma”[112].

EBV has a double-stranded DNA genome that exists in a linear form in viral particles. After EBV enters the host cell, the viral DNA circularizes via the fusion of terminal repeats at both ends, and it maintains its circular form in the nuclei of latently infected cells without integration into the host genome[113]. Southern blot analysis for terminal repeats has demonstrated that EBV present in neoplastic cells is mono- or oligo-clonal, even in advanced stages[114-116]. Moreover, all of the cancerous cells are positive for EBER-in situ hybridization in all cases of EBV⁺ gastric cancer. This fact indicates that EBV infection occurs at the initial, or a very early stage, of carcinoma development, and it implies a profound association of EBV with gastric carcinogenesis.

The mechanism underlying EBV infection in the gastric mucosal epithelium remains unclear, while the viral receptor molecule for CD21 in B lymphocytes is not expressed on epithelial cells[117]. Because co-cultivation of virus-producing lymphocytes demonstrates a much greater efficiency of infection (up to 800-fold) compared to cell-free infection, direct cell-to-cell contact between B lymphocytes and gastric epithelial cells is the most likely model to explain how EBV infects epithelial cells in vivo (Figure 2)[118]. This hypothesis supports histopathological data showing that the background mucosa of EBV⁺ gastric cancer presents atrophic gastritis with lymphocyte infiltration due to H. pylori infection[119]. However, it remains unclear whether chronic inflammation with H. pylori is a prerequisite for EBV to infect gastric epithelial cells.

Open in New Tab   Full Size Figure   Download Figure

Figure 2 Schematic representation about infectious condition and pathogenicity of Epstein-Barr virus. Direct cell-to-cell contact between B lymphocyte and gastric epithelial cell may perhaps be the most likely model to infect with Epstein-Barr virus (EBV) into epithelial cells in vivo (left). EBV infection induces extensive promoter methylation, which should contribute to tumorigenesis. H. pylori: Helicobacter pylori.

EBV⁺ gastric cancer forms a distinct subgroup of gastric cancer. Previous reports have indicated that promoter methylation is observed more frequently in EBV⁺ gastric cancers than in EBV gastric cancers, despite analyzing a limited number of cancer-associated genes[120-122]. We performed a comprehensive analysis of promoter methylation in clinical gastric cancers and found that gastric cancers clustered into three distinct subgroups. Interestingly, EBV⁺ gastric cancers displayed an extremely high methylation phenotype, termed the EBV⁺ epigenotype[37]. Moreover, genes specifically methylated in EBV⁺ gastric cancers were shown to expand not only within PRC-target genes in ES cells but also to non-PRC-target genes. This result implies that EBV⁺ gastric cancer is methylated via a unique mechanism(s). Subsequently, to clarify the causal role of EBV infection, we performed in vitro EBV infection experiments in low-methylation MKN7 gastric cancer cells to determine whether these cells would acquire extensive methylation and, as a result, the EBV⁺-specific methylation epigenotype. The induced methylation repressed multiple genes, including multiple tumor-suppressor genes, suggesting a role for EBV in tumorigenesis[37].

The inducer of aberrant DNA methylation remains elusive. EBV exists in three latent forms defined by the expression pattern of latent genes. Lymphoblastoid cell line (LCL) and transformed primary B lymphocytes infected with EBV express all latent genes, LMPs (1, 2A, 2B), EBNAs (1, 2, 3A, 3B, 3C, LP), EBERs (1, 2), and BARTs, and this expression program is referred to as type III latency. In contrast, Burkitt’s lymphoma shows type I latency, with the minimum expression of EBNA1, EBERs, and BARTs only. Type II latency, in which LMP1 and LMP2 are expressed in addition to latency I genes, is observed in EBV-associated Hodgkin lymphoma, peripheral natural killer/T-cell lymphoma, and nasopharyngeal carcinoma. EBV⁺ gastric cancer shows type I (or II) latency and expresses EBNA1, EBERs, BARTs, and LMP2A[105,111,123].

Several studies have elucidated the function of latent genes for promoter methylation. LMP1 was reported to down-regulate CDH1 gene expression and induce cell migration using cellular DNA methylation machinery[124,125]. LMP2A plays an essential role in epigenetic abnormalities by inducing promoter methylation of PTEN[126]. EBER1 and EBER2 are small non-coding RNAs of approximately 170 bases in length that are abundantly expressed in the nuclei of latently infected cells, up to 10⁷ copies per cell. Although some oncogenic properties of EBERs have been reported, such as the contribution of efficient growth transformation of B lymphocytes[127,128] or the induction of insulin-like growth factor 1 (IGF-I) acting as an autocrine growth factor in gastric cancer or nasopharyngeal carcinoma cells[129,130], the distinct influence of epigenetic modification remains unclear. Moreover, while these viral genes may play a role in aberrant methylation, methylation induction at the genome-wide scale, which can result from EBV infection, has not been demonstrated through the forced expression of any viral gene[37].

Rather than viral factors, host cellular mechanisms may play more important roles in the induction of aberrant methylation. In type I latency, while host cells induce dense methylation in the viral genome to silence most viral genes, the host genome itself is also extensively hypermethylated[131,132]. In type III latency, such as LCL, neither the viral genome nor the host cellular genome is significantly hypermethylated[132], and this observation implies that a host-driven mechanism that induces DNA methylation in the viral genome may affect methylation of the host genome. Recent exome sequencing analysis demonstrated that ARID1A was frequently mutated in EBV⁺ gastric cancer[8,133], and other chromatin remodelers were also mutated[9]. While it is not known whether mutation of these genes is causally associated with aberrant methylation in EBV⁺ gastric cancer, these chromatin remodelers may play a role in protecting the epigenomic status of the host genome from the pressure of methylation induction. Further investigation is necessary to clarify the roles of host cellular factors in methylation induction.