Demethylation of MAGE promoters during gastric cancer progression

Melanoma antigen (MAGE)-encoding genes are expressed in various tumour types via demethylation of their promoter CpG islands, which are silent in all non-neoplastic tissues except for the testis and placenta. The clinicopathological significance of demethylation of MAGE genes in gastric carcinoma is not known. We investigated the promoter methylation status of MAGE-A1 and -A3 in 10 gastric cancer cell lines and in surgical specimens from 84 gastric cancer patients by methylation-specific PCR (MSP). Expression of MAGE-A1 and -A3 in the 10 gastric cancer cell lines was also investigated by RT–PCR. Any correlation between the methylation status of the MAGE promoters and clinicopathological characteristics of the gastric cancer patients was then assessed. Eight of the 10 gastric cancer cell lines showed demethylation of both MAGE-A1 and -A3, and the remaining two cell lines did either of MAGE-A1 or -A3. Expression of MAGE-A1 and -A3 was confirmed in seven and nine of the 10 gastric cancer cell lines, respectively. The MAGE-A1 and -A3 promoters were demethylated in 29% (25 out of 84) and 66% (56 out of 84) of the gastric tumour specimens, respectively. Demethylation of both MAGE-A1 and -A3 promoters (n=22) was found more frequently in gastric cancer patients in advanced clinical stages (P=0.0035), and these patients also exhibited a higher incidence of lymph node metastasis (P=0.0007) compared to those patients without demethylation (n=25). Furthermore, demethylation patients tended to have a worse prognosis, although this difference was not statistically significant (P=0.183). Demethylation of MAGE-A1 and -A3 occurs during progressive stages of gastric cancer, and may be associated with aggressive biological behaviour of gastric cancer.

Epigenetic alterations, including hypermethylation of promoter CpG islands and histone deacetylation of tumour suppressor and tumour-related genes (Choi et al, 2001;Jones and Takai, 2001;Kim et al, 2001;Goodman and Watson, 2002;Jones and Baylin, 2002) as well as global DNA hypomethylation Gaudet et al, 2003;Lengauer, 2003), have been recognised as important contributors to carcinogenesis in humans. Global DNA hypomethylation has been observed in carcinomas of the breast, liver, and colon, and is considered to occur in the early stages of tumour development (Goelz et al, 1985;Cravo et al, 1996;Narayan et al, 1998;Lin et al, 2001;Bariol et al, 2003). However, little is known about promoter hypomethylation of specific genes such as oncogenes and growth-related genes, with the exception of the association between demethylation and increased expression of c-abl, c-myc, Ha-ras, and raf (Cheah et al, 1984;Weitzman et al, 1989;Sharrad et al, 1992;Counts and Goodman, 1994).
Human melanoma cells express antigens that are recognised by cytolytic T lymphocytes derived from the blood of tumour-bearing patients or from tumour-infiltrating lymphocytes . A number of such antigens are encoded by genes of the MAGE family ( Van der Bruggen et al, 1991;Lucas et al, 2000). A total of 19 MAGE genes are located on chromosome X (Lucas et al, 2000), and are expressed in other tumours, including gastric cancer, in addition to melanoma (Van der Bruggen et al, 1991;Brasseur et al, 1992;Chambost et al, 1993;Inoue et al, 1995a, b;Gotoh et al, 1998;Takahashi et al, 1998). Although the functions of the various MAGE proteins remain to be elucidated, MAGE gene expression is known to be activated by promoter demethylation in a similar manner as the oncogenes and growthrelated genes described above (De Smet et al, 1996). These genes are silent in normal tissues except for the testis and placenta , and may be targets for future cancer immunotherapies (Marchand et al, 1999;Jang et al, 2001).
In the present study, we investigated the promoter methylation status of MAGE-A1 and -A3, which were the most frequent targets for immunotherapy, in gastric cancers, and analysed the correlation between the MAGE-A1 and -A3 methylation status and clinicopathological parameters of gastric cancer patients, including event-free survival.

Gastric cancer cell lines
We have investigated 10 gastric cancer cell lines with variable histologies that were cultured under appropriate conditions in our laboratory: MKN1, an adenosquamous cell carcinoma; MKN7, a well-differentiated adenocarcinoma; MKN28 and MKN74, moderately differentiated adenocarcinomas; MKN45 and KWS-I, poorly differentiated adenocarcinoma; KATO-III, a signet-ring cell carcinoma; TSG11, a hepatoid carcinoma; and ECC10 and ECC12, endocrine cell carcinomas.

Primary gastric cancers
In total, 84 pairs of cancerous and noncancerous gastric tissues (51 differentiated and 31 undifferentiated carcinomas; 25 early-stage carcinomas that demonstrated a depth of invasion limited to the submucosa and 57 advanced stage carcinomas) were surgically obtained from 84 gastric cancer patients. These tissues were immediately frozen and stored at À801C until analysis. All patients received a median of 36.7 months of follow-up care (range, 1 -77 months). Signed informed consent was obtained from every patient to allow the use of biological materials for biological studies.

DNA extraction
DNA was extracted from 10 gastric carcinoma cell lines and 84 primary gastric cancers and their corresponding noncancerous gastric tissues with SepaGene (Sanko-Junyaku, Tokyo, Japan).

RNA extraction
Total RNA was isolated from 10 gastric carcinoma cell lines with the TRIZOL reagent (Gibco BRL, Life Technologies, Gaithersburg, MD, USA).

Bisulphite modification and methylation-specific polymerase chain reaction (MSP)
Treatment of DNA samples with sodium bisulphite converts all unmethylated cytosines to uracils and does not affect methylated cytosines. Briefly, 2 mg of genomic DNA were denatured with sodium hydroxide and modified by sodium bisulphite. The samples were then purified using Wizard DNA purification resin (Promega, Madison, WI, USA), treated with NaOH, recovered in ethanol, and resuspended in 30 ml of distilled water. Amplification was achieved in a 20 ml reaction volume containing 2 ml of GeneAmp PCR Gold Buffer (PE Applied Biosystems, Foster City, CA, USA), 1.0 mM MgCl 2 , 1 ml each primer, 0.2 mM dNTPs, and 1 U Taq polymerase (AmpliTaq Gold DNA Polymerase, PE Applied Biosystems). After heating at 941C for 10 min, polymerase chain reaction (PCR) was performed in a thermal cycler (GeneAmp 2400, PE Applied Biosystems) for 35 cycles, each of which consisted of denaturation at 941C for 30 s, annealing at 541C for 60 s, and extension at 721C for 60 s, followed by a final 7-min extension at 721C. A positive control (Sss-I methylase-treated DNA) and negative control (distilled water without DNA) were included for each amplification. The PCR products were separated on a 6% nondenaturing polyacrylamide gel. The following primer sets were used: MG1 M forward (5 0 -ATT TAG GTA GGA TTC GGT TTT C-3 0 ) and MG1 M reverse (5 0 -AAA CTA AAA CGT CTT CCC GCG-3 0 ) for the methylated MAGE-A1 sequence; MG1U forward (5 0 -ATT TAG GTA GGA TTT GGT TTT T-3 0 ) and MG1U reverse (5 0 -AAA CTA AAA CAT CTT CCC ACA-3 0 ) for the MAGE-A1 unmethylated sequence; MG3 M forward (5 0 -CGT TTT GAG TAA CGA GCG AC-3 0 ) and MG3 M reverse (5 0 -ACT AAA ACG ACG AAA ATC GAC G-3 0 ) for the MAGE-A3 methylated sequence; MG3U forward (5 0 -TGT TTT GAG TAA TGA GTG AT-3 0 ) and MG3U reverse (5 0 -ACT AAA ACA ACA AAA ATC AAC A-3 0 ) for the MAGE-A3 unmethylated sequence. Methylated and unmethylated PCR products of MAGE-A1 and -A3 in gastric cancer cell lines were sequenced. The PCR products were purified using QIA Quick PCR Purification Kit (QIAGEN, Tokyo). The purified PCR products were sequenced with the BigDye Terminator Cycle Sequencing Ready Reaction Kit (PE Applied Bioststems). Gel electrophotesis, data collection, and analysis were carried out with a Genetic Analyser (model 310, PE Applied Biosystems).

Preparation of positive control
Sss-I methylase (New England BioLabs, Inc.,. Beverly, MA, USA) was used to methylate 100 mg of peripheral blood DNA, which was modified by sodium bisulphite as described above.

Statistical analysis
Statistical comparisons were performed using Fisher's exact test. A Po0.05 was considered significant. Survival analysis was performed using a Kaplan -Meier curve with log rank test.

Correlation between demethylation of MAGE promoters and clinicopathological parameters
Gastric cancer patients who exhibited demethylation of both the MAGE-A1 and -A3 promoters (n ¼ 22) were at a more advanced clinical stage (P ¼ 0.0035). and had a higher incidence of lymph node metastasis (P ¼ 0.0007) compared with those who did not have demethylated MAGE-A1 and -A3 promoters (n ¼ 25) (Table 1). Furthermore, patients with demethylated MAGE-A1 and -A3 promoters tended to have a worse prognosis, although this difference was not statistically significant by the log rank test (P ¼ 0.183) (Figure 4). Patients with demethylation of only one of the two promoters exhibited biological features intermediate to those of the other two groups.    C, E, and G). In gastric cancer specimens, demethyled MAGE-A1 and -A3 are present in lanes 1, 6, and 7 of (B) and in lanes 5-8 of (F), respectively, whereas none of the non-neoplastic gastric tissues exhibit demethylation of MAGE-A1 or -A3 (D and H). Lanes: 1, M244; 2, M245; 3, M246; 4, M248; 5, M251; 6, M254; 7, M256; 8, M257; and 9, M262.

DISCUSSION
MAGE-A1 and -A3 encode tumour-specific antigens that are recognised on melanoma cells by autologous cytolytic T lymphocytes ( Van der Bruggen et al, 1991;Boon et al, 1994). These genes are expressed in a significant proportion of tumours of various histological types, but not in normal tissues, except for male germ line cells and placenta Takahashi et al, 1995). Demethylation of promoter CpG islands in MAGE genes triggers their expression in tumour cells, whereas they are not expressed in cells in which they remain methylated (De Smet et al, 1996). The function of the MAGE peptides are not known, although their tumour-specific expression is clearly of great importance for immunotherapy (Marchand et al, 1999;Nishiyama et al, 2001;Sadanaga et al, 2001). The MAGE-A1 and -A3 peptides are expressed in 67 -73% of gastric cancer cell lines (Inoue et al, 1995a;Li et al, 1996). In agreement with these data, we have demonstrated that MAGE-A1 and -A3 mRNA are expressed in 70 and 90% of gastric cancer cell lines, respectively. MAGE-A1 and -A3 mRNA has also been reported to be expressed in approximately 40% of primary gastric cancers (Inoue et al, 1995a, b;Li et al, 1996). However, these previous studies did not verify the methylation status of the MAGE-A1 and -A3 promoter CpG islands. In the present study, we showed that the methylation status of the MAGE-A1 and -A3 promoters almost directly correlated with their expression status in gastric cancer cell lines.
Global DNA hypomethylation is thought to occur during the early stages of tumour development in gastric and other tissues (Goelz et al, 1985;Narayan et al, 1998;Lin et al, 2001;Bariol et al, 2003). Additionally, in pulmonary carcinogenesis, demethylation of the promoter CpG islands of MAGE genes has been observed not only in tumours but also in the adjacent non-neoplastic lung tissues and bronchial epithelia from smokers . Therefore, MAGE genes may be activated prior to malignant transformation in the lung, possibly by global DNA hypomethylation (Tamura, 2002). However, we have demonstrated that MAGE gene promoters are demethylated more frequently in gastric cancers at advanced clinical stages. Furthermore, demethylation of  MAGE-A1 and -A3 is quite rare in non-neoplastic gastric tissues of gastric cancer patients. In a separate study, we have confirmed that demethylation of MAGE-A1 and -A3 was also very rare in various organs obtained at autopsies, from various age groups (data not shown). Therefore, we hypothesise that demethylation of MAGE genes occurs during progressive stages of gastric carcinogenesis, probably after global DNA hypomethylation. Promoter CpG islands of several tumour suppressor and tumour-related genes are frequently methylated in both neoplastic and nonneoplastic gastric epithelia (Tamura, 2002;Waki et al, 2002). Hypermethylation of different genes increases with age in different organs (Waki et al, 2003). These results suggest that hypermethylation of promoter CpG islands occurs very early in gastric carcinogenesis, in contrast to demethylation of MAGE gene promoters. Several studies have analysed MAGE gene expression Inoue et al, 1995a, b;Cravo et al, 1996;Li et al, 1996;Sadanaga et al, 2001), but none have evaluated the demethylation status of their promoters in gastric cancer. Inoue et al (1995a, b) detected MAGE expression in about 40% of primary gastric cancers, but failed to find any significant correlation between MAGE expression and clinicopathological parameters (Inoue et al, 1995a, b). In the present study, demethylation of either MAGE-A1 or -A3 was not significantly correlated with clinicopathological parameters, but demethylation of both genes significantly correlated with advanced clinical stage and lymph node metastasis. Furthermore, we have noticed that patients with tumours showing demethylation of both MAGE-A1 and -A3 tend to have a worse prognosis, although this difference was not statistically significant. In contrast, hypermethylation of the hMLH1 gene promoter is a marker of a better prognosis (Yamamoto et al, 1999). No correlation has been observed between demethylation of the MAGE genes and hypermethylation of hMLH1 or p16 (data not shown).
In summary, demethylation of the MAGE-A1 and -A3 promoters frequently occurs during progressive stages of gastric carcinogenesis and may be associated with aggressive biological behaviour of gastric cancer.